TFIIH Interacts with the Retinoic Acid Receptor γ and Phosphorylates Its AF-1-activating Domain through cdk7*

Retinoic acid receptor γ (RARγ) is phosphorylated in COS-1 cells at two conserved serine residues located in the N-terminal region (serines 77 and 79 in RARγ1 and serines 66 and 68 in RARγ2) that contains the activation function AF-1. These serines are phosphorylated in vitro by cdk7, a cyclin-dependent kinase associated to cyclin H and MAT1 in the CAK complex (cdk7·cyclin H·MAT1), that is found either free or as a component of the transcription/DNA repair factor TFIIH. RARγ is more efficiently phosphorylated by TFIIH than by CAK and interacts not only with cdk7 but also with several additional subunits of TFIIH. RARγ phosphorylation and interaction with TFIIH occur in a ligand-independent manner. Our data demonstrate also that phosphorylation of the AF-1 function modulates RARγ transcriptional activity in a response gene-dependent manner.

Retinoic acid receptor ␥ (RAR␥) is phosphorylated in COS-1 cells at two conserved serine residues located in the N-terminal region (serines 77 and 79 in RAR␥1 and serines 66 and 68 in RAR␥2) that contains the activation function AF-1. These serines are phosphorylated in vitro by cdk7, a cyclin-dependent kinase associated to cyclin H and MAT1 in the CAK complex (cdk7⅐cyclin H⅐MAT1), that is found either free or as a component of the transcription/DNA repair factor TFIIH. RAR␥ is more efficiently phosphorylated by TFIIH than by CAK and interacts not only with cdk7 but also with several additional subunits of TFIIH. RAR␥ phosphorylation and interaction with TFIIH occur in a ligand-independent manner. Our data demonstrate also that phosphorylation of the AF-1 function modulates RAR␥ transcriptional activity in a response gene-dependent manner.
The pleiotropic effects of retinoids are transduced by two nuclear receptor families, the retinoic acid receptors (RARs) 1 and the retinoid X receptors (RXRs), that are ligand-dependent transregulators belonging to the nuclear receptor superfamily (1)(2)(3)(4). RARs are activated by all-trans and 9-cis retinoic acid, whereas RXRs are activated by 9-cis retinoic acid only. There are three RAR (␣, ␤, and ␥) and three RXR (␣, ␤, and ␥) isotypes, and for each isotype there are at least two main isoforms that differ in their N-terminal region (1,5,6).
As do other members of the nuclear receptor superfamily, RARs and RXRs exhibit a conserved modular structure with six variably conserved regions (A to F) ( Fig. 1) (1,5). The Nterminal A/B region of RARs contains a ligand-independent transcriptional activation function, AF-1 (7,8). Although the B regions of the three RAR isotypes are moderately conserved, their A regions are unrelated and differ for each isoform of a given RAR isotype (5). The highly conserved C region encompasses the central DNA binding domain. The function of region F, if any, is unknown. Region E is more complex, as it contains the ligand binding domain, a dimerization interface, and the ligand-dependent transcriptional activation/repression domain AF-2 (1,9). The activity of AF-2 is entirely dependent on the integrity of a conserved sequence referred to as the AF-2 AD core, located in ␣-helix 12 at the C-terminal end of the ligand binding domain. Ligand binding induces a major conformational change that includes helix 12 and creates a new surface for coactivator binding while corepressors are released, thus resulting in a transcriptional-competent nuclear receptor relayed to the transcriptional machinery and the chromatin template (1, 10 -12). The AF-2 and AF-1 activities synergize with each other in a response element-and promoter context-dependent manner (1,8,13).
RARs and RXRs are phosphoproteins (14 -16), and their phosphorylation involves several kinases. RAR␣ can be phosphorylated in its AF-1-containing B region by the cyclin-dependent kinase cdk7 (14), which together with MAT1 and cyclin H forms the CAK complex that is found either free or as a component of the general transcription/DNA repair factor TFIIH (17)(18)(19)(20). This phosphorylation, which results from an interaction with cdk7, is crucial for RAR␣ transcriptional activity and modulates its ligand-induced degradation by the ubiquitin-proteasome pathway. 2 RAR␣ can also be phosphorylated by protein kinase A at a residue located in the ligand binding domain (15), and this phosphorylation is required for differentiation of mouse embryonal carcinoma F9 cells into parietal endoderm-like cells upon RA and cAMP treatment (13). Similarly, RXR␣ was found to be phosphorylated in its N-terminal A/B region and shown to be hyperphosphorylated in the same region by c-Jun N-terminal kinases upon UV activation (16).
Mutations of putative phosphorylation sites located in the AF-1 domain of mRAR␥2 were found to prevent the RA-induced differentiation of F9 cells (13), thus indicating that RAR␥2 phosphorylation in this domain could be required for this differentiation. Moreover, phosphorylation in the same domain was recently shown to be crucial for the ligand-induced degradation of RAR␥ by the ubiquitin-proteasome pathway. 2 Because of our previous demonstration that RAR␣ can be phosphorylated by cdk7 present within TFIIH, we assumed that RAR␥ could also be phosphorylated by TFIIH. In the present study, we demonstrate that the B region of the two major human or mouse RAR␥ isoforms, RAR␥1 and RAR␥2, are phosphorylated in a ligand-independent manner, by the cyclin Hand MAT1-dependent protein kinase cdk7. We also show that phosphorylation of RAR␥ is more efficient when cdk7 is present within TFIIH. In addition, we reveal the existence of multiple RAR␥-TFIIH interactions that involve not only cdk7 but also additional subunits of TFIIH. Finally, phosphorylation of the A/B region was found to modulate RAR␥-induced transcription in a response gene-dependent manner.
Cells, Transfections, and CAT Assays-Mouse embryocarcinoma F9 cells (WT and RAR␥ Ϫ/Ϫ ) were cultured as monolayers (34,35). COS-1 cells were grown and transiently transfected by using the Ca 2ϩ phosphate precipitation technique (36). In addition to expression vectors and reporters described in figures, all transfections included Bluescript DNA as a carrier and the ␤-galactosidase expression vector pCH110 (1 g) to correct for variations in transfection efficiency. After a 20-h incubation with calcium phosphate-precipitated DNA, cells were washed and incubated for an additional 20 h in the appropriate medium in the absence or presence of RA (10 Ϫ7 M) or the RAR␥ agonist (BMS961) (10 Ϫ7 M). Cells were harvested 48 h after transfection, and CAT assays were performed using the enzyme-linked immunosorbent assay method (Roche Molecular Biochemicals). All assays were normalized to equal ␤-galactosidase activity, and the results were expressed as pg of CAT/unit of ␤-galactosidase.
Immunoprecipitations-Sf9 cells were infected with combinations of baculoviruses, and whole cell extracts were prepared (17). Nuclear extracts were prepared from F9 cells as described (30). Extracts were incubated with protein A-Sepharose beads cross-linked with the indicated monoclonal antibodies in lysis buffer (20 mM Tris-HCl, pH 7.8, 200 -250 mM NaCl, 10% glycerol, 0.1 mM EDTA, 1 mM dithiothreitol, and 0.1% Nonidet P-40) (14,19). The beads were washed several times with the same buffer, resuspended in Laemmli buffer, and boiled, and the immunoprecipitated proteins were resolved by SDS-PAGE and revealed by immunoblotting and chemiluminescence.
For phosphorylation in transfected cells, COS-1 cells were transfected with wild type or mutated RAR␥ expression vectors using the standard calcium phosphate procedure and labeled with [ 32 P]orthophosphate as described (15). Whole cell extracts were immunoprecipitated and resolved by SDS-PAGE, and after electrotransfer, the proteins were revealed by autoradiography and immunoreaction. Two-dimensional phosphoamino acid and tryptic phosphopeptide separations were carried out on thin layer cellulose plates using the Hunter thin-layer electrophoresis (HTLE) system as described (15).

Both Human and Mouse RAR␥ Overexpressed in COS-1 Cells Are Phosphorylated in Their N-terminal A/B Region, whereas
Mouse RAR␥ Is Additionally Phosphorylated in Its F Region-To determine whether wild type human RAR␥1 (hRAR␥1WT) is a phosphoprotein, COS-1 cells were transfected with the corresponding expression vector and labeled with [ 32 P]orthophosphate in the absence or presence of RA (10 Ϫ7 M). Whole cell extracts were immunoprecipitated with a RAR␥-specific monoclonal antibody and resolved by SDS-PAGE, and the phosphorylated proteins were analyzed either by autoradiography or by immunoblotting. hRAR␥1 was phosphorylated irrespective of the addition of RA to the culture medium ( Fig. 2A, lanes 1 and 2). Phosphoamino acid analysis indicated that this phosphorylation was restricted to serine residues (Fig. 2B).
Tryptic phosphopeptide mapping yielded 3 phosphopeptides a, b, and c ( Fig. 2C, panel 1). To characterize these phosphopeptides, COS-1 cells were transfected with hRAR␥ 1 deleted for the A/B region (hRAR␥⌬AB) and labeled with [ 32 P]orthophosphate. hRAR␥⌬AB was not phosphorylated ( Fig. 2A, lane 3), suggesting that phosphopeptides a, b, and c are located in the A/B region. Similarly, transfection in COS-1 cells of a chimeric construct expressing the A/B region of hRAR␥1 fused to the DNA binding domain (C) of the human estrogen receptor (ER) showed that the corresponding chimeric protein, hRAR␥1(AB)-ER(C), was phosphorylated ( Fig. 2A, lane 8) and yielded the three phosphopeptides a, b, and c ( Fig. 2C, panel 2).
Altogether, our results indicate that hRAR␥1WT is phosphorylated at serines 77 and 79. They also indicate that phosphopeptide a contains serine 77, whereas phosphopeptides b and c contain serine 79. These two b and c peptides may be partial digestion products due to the presence of low efficiency sites for trypsin cleavage in human RAR␥1 (37). In addition, the fact that mutation of serine 79 eliminates all phosphopeptides suggests that phosphorylation of serine 77 depends on that of serine 79 (37). Finally, as both serines 77 and 79 are located in a same tryptic peptide, it is not excluded that peptides a, b, and c are phosphoisomers, spot "a" representing the diphosphorylated peptide (at serines 77 and 79) and spots "b" and "c," the monophosphorylated one (at serine 79) (37).
The phosphorylation sites that are located in the B region of RAR␥ are conserved between human and mouse and identical in the RAR␥1 and RAR␥2 isoforms (21, 23) (Fig. 1). In contrast, human and mouse regions F are different ( Fig. 1) (21). Therefore, the two mouse RAR␥ isoforms (mRAR␥1 and mRAR␥2) were overexpressed in COS-1 cells and labeled with [ 32 P]orthophosphate. Both receptors were phosphorylated in a RA-independent manner (Fig. 3A, lanes 1-4). Tryptic phosphopeptide mapping of mRAR␥1 yielded, in addition to phosphopeptides a, b and c described above for hRAR␥1, two other phosphopeptides, d and e (compare Fig. 3B, panel 1 to Fig. 2C, panel 1). The same d and e peptides were present in the tryptic digest of mRAR␥2, whereas phosphopeptides aЈ and bЈ differed from phosphopeptides a and b of RAR␥1, and phosphopeptide c was not seen (Fig. 3B, compare panels 1 and 2; see also panel 6), most probably because regions A of these two RAR␥ isoforms are unrelated in their sequence (23). As expected, mRAR␥⌬AB was less phosphorylated than mRAR␥2WT (Fig. 3A, compare lanes 5 and 6) and lacked phosphopeptides aЈ and bЈ (Fig. 3B,   panel 4). Moreover, the tryptic digest of mRAR␥2S66A/S68A lacked phosphopeptides aЈ and bЈ (Fig. 3B, panel 5), indicating that mRAR␥2 is phosphorylated in its A/B region at serines 66/68.
Deletion of the F region in mRAR␥2⌬F decreased the overall phosphorylation level (Fig. 3A, lane 8), and phosphopeptides d and e disappeared (Fig. 3B, panel 6). Moreover, a chimeric receptor construct expressing the DEF regions of mRAR␥ fused to the DNA binding domain of the yeast transactivator Gal4 [Gal4-mRAR␥(DEF)] was also phosphorylated when overexpressed in COS-1 cells (Fig. 3A, lane 11) and yielded phosphopeptides d and e (Fig. 3B, panel 3). When the region F serine 440, which belongs to a potential phosphorylation site for proline-dependent kinases (Fig. 1), was mutated to alanine, the corresponding mutant construct, mRAR␥2S440A, did not yield the two phosphopeptides d and e (data not shown), which, as mentioned above, may correspond to partial trypsin digestion products (37). Thus, mouse RAR␥ contains a phosphorylation site in its F region, that is absent in human RAR␥.
Mouse RAR␥2 is the main RAR␥ isoform present in F9 embryonal carcinoma cells (21,23). This endogenous RAR␥ isoform was also phosphorylated (Fig. 3C) in its B and F regions, as it yielded an array of tryptic phosphopeptides similar to that observed with recombinant mRAR␥2 overexpressed in COS-1 cells (compare Fig. 3D with Fig. 3B, panel 2).
Because serines 77 and 79 belong to consensus motifs for proline-directed kinases (38 -40), we also investigated whether RAR␥ could be a substrate for mitogen-activated protein kinases. Although hRAR␥1 was phosphorylated by p44 MAPK (Fig.  4B, lane 4), tryptic phosphopeptide mapping yielded only phosphopeptide a (Fig. 4C, panel 3). Only the S77A mutation and not the S79A one abrogated this RAR␥ phosphorylation (Fig.  4B, compare lanes 4 to 7). Thus, although RAR␥ can be a substrate for MAPK in vitro, phosphorylation by this kinase is different from that achieved with cdk7.
RAR␥ and RAR␣ Bind Both to cdk7 and to Several Subunits of the Core of TFIIH-As previously reported in the case of RAR␣ (14), RAR␥ directly interacted with cdk7. Purified bacterially expressed hRAR␥1 was indeed bound by recombinant cdk7 immunoabsorbed onto protein A-Sepharose beads crosslinked with cdk7 antibodies, (Fig. 5, lane 2). Similarly to RAR␣⌬AB (14), RAR␥⌬AB was also bound by cdk7 (Fig. 5, lane  3), indicating that cdk7 does not bind to the A/B region.
The observation that RAR␥ was more efficiently phosphorylated by cdk7 within TFIIH than within CAK prompted us to investigate whether this resulted from a tighter binding of RAR␥ to TFIIH than to free CAK. TFIIH is a multisubunit complex composed of CAK and of six additional subunits (p34, p44, p52, p62, XPB, and XPD) referred to as the core of TFIIH (17). Sf9 cells were therefore coinfected with combinations of baculoviruses expressing RAR␥⌬AB and either the whole TFIIH (nine subunits, IIH9) or some TFIIH subcomplexes, such as the "core TFIIH/XPD" (six subunits, IIH6), the core TFIIH (IIH5), or CAK (17). Infected cell extracts were then immunoprecipitated with monoclonal antibodies directed against either the p44 subunit of TFIIH or RAR␥. The immunoprecipitated fractions (mentioned as "Bound (B)" in Fig. 6) were resolved by SDS/PAGE and analyzed by immunoblotting using antibodies against either XPB and p62 subunits of TFIIH or cdk7, cyclin H, and MAT1 subunits of CAK. mAb-p44 retained the corresponding IIH complexes (Fig. 6A, lanes 1-8) and also RAR␥ (Fig. 6A, lanes 4, 6, and 8). Conversely, mAb-RAR␥ retained not only RAR␥ but also the TFIIH subcomplexes, whether the core of TFIIH was produced or not in association with CAK (Fig. 6B, lanes 2 and 6), indicating that RAR␥ can bind not only CAK but also the core of TFIIH. RAR␥ also coimmunoprecipitated with cdk7, cyclin H, and MAT1 when coinfected with CAK only (Fig. 6B, lane 10).
FIG. 6. hRAR␥1 interacts with cdk7 associated to CAK and with the core of TFIIH. A, Sf9 cells were coinfected with baculoviruses encoding for hRAR␥⌬AB and for each TFIIH subunit that form either the IIH9, IIH6, or IIH5 complexes. Equal amounts of extracts were immunoprecipitated (IP) with mAbp44 and immunoblotted with RP␥(F) (lower panel) or with antibodies raised against the representative TFIIH subunits (upper panels, p62, XPB, cdk7, cyclin H, and MAT1). Controls were performed in the absence of RAR␥ (lanes 1 and  2). Indicated are the loaded (L) and bound (B) materials. B, extracts from Sf9 cells coinfected with baculoviruses encoding for hRAR␥⌬AB and for each TFIIH subunit that form either the IIH9, IIH6, or CAK complexes were immunoprecipitated with mAb4␥(F) and immunoblotted as in A. Controls were performed in the absence of RAR␥ (lanes 3, 4, 7, 8, 11, and 12). L and B are as in A. C, Sf9 cells were coinfected with baculoviruses encoding for FLAG-tagged full-length RAR␣1 and for each TFIIH subunit that form either the IIH9, IIH6, or IIH5 complexes. Extracts were immunoprecipitated with mAbp44 and immunoblotted with RP␣(F) (lower panel) or with antibodies against the TFIIH subunits (upper panel) as in A. Controls were performed in the absence of RAR␣ (lanes 1, 2, 11, and 12). L and B are as in A. D, extracts from Sf9 cells coinfected with baculoviruses encoding for FLAG-tagged fulllength RAR␣1 and for each TFIIH subunit that form either the IIH9, IIH5, or CAK complexes were immunoprecipitated with monoclonal anti-FLAG antibodies and immunoblotted with RP␣(F) (lower panel) or with antibodies against the TFIIH subunits (upper panel) as in A. Controls were performed in the absence of RAR␣ (lanes 3, 4, 11, and 12) or in the absence of TFIIH (lanes 1 and 2). L and B are as in A. E, extracts from Sf9 cells coinfected with baculoviruses encoding for FLAG-tagged RAR␣1 and for either XPB, XPD, p62, or p52 were immunoprecipitated with monoclonal anti-FLAG antibodies and immunoblotted with RP␣(F) (lower panel) or with antibodies against the TFIIH subunit (upper panel). Controls were performed in the absence of RAR␣ (lanes 3, 4, 7, 8, 11, 12, 15, and 16). L and B are as in A.
cipitated RAR␣ whether it was coexpressed with the entire TFIIH or only with the TFIIH core (Fig. 6C, lanes 4, 6, and 8) (14). Reciprocally, monoclonal anti-FLAG antibodies coimmunoprecipitated RAR␣ and subunits from the TFIIH core, whether the latter was produced or not in association with CAK (Fig. 6D, lanes 6 and 8). RAR␣ also coimmunoprecipitated with the three subunits of CAK when coinfected with CAK only (Fig.  6D, lane 10).
Collectively, these results show that RAR␥ and RAR␣ interact with TFIIH through both CAK and the TFIIH core. Moreover, a direct interaction between RAR␣ (Fig. 6E) and RAR␥ (data not shown) and either one of the core TFIIH subunits, XPB, XPD, p52 (Fig. 6E, lanes 2, 6, and 14), p44, or p34 (data not shown), could be revealed, indicating that these subunits may interact with RARs within TFIIH. Note that no interaction was observed between p62 and RARs, whether the coimmunoprecipitations were carried out with mAb-p62 or mAb-RAR (Fig. 6E, lane 10, and data not shown).
To investigate whether RAR␥ was also intracellularly associated with TFIIH, we used the murine F9 cell line in which mRAR␥2 is phosphorylated (see above). Nuclear extracts from F9 cells were immunoprecipitated with monoclonal antibodies against p44 and analyzed by SDS-PAGE/immunoblotting using antibodies against either p62, cdk7, or RAR␥. mAb-p44 coimmunoprecipitated TFIIH and RAR␥ (Fig. 7A, lane 3). Reciprocally, mAb-RAR␥ coimmunoiprecipitated RAR␥ and cdk7 (Fig.  7B, lane 5). These interactions were specific, as they were not revealed in control immunoprecipitations performed with mAb against the unrelated glutathione S-transferase protein (Fig.  7B, lanes 3 and 4) or with nuclear extracts from RAR␥ null F9 cells (RAR␥ Ϫ/Ϫ F9 cells) (Fig. 7B, lanes 2 and 6 and Fig. 7A,  lane 4). Similar results were observed with RAR␣ in the same cells (data not shown). Altogether these data indicate that a fraction of RAR␥2 and RAR␣ is associated with TFIIH in F9 cells.
Phosphorylation by cdk7 Modulates the Transcriptional Activity of RAR␥-To investigate whether phosphorylation of the A/B region of RAR␥ could play a role in the ligand-independent AF-1 transactivation function, we first tested the activity of hRAR␥1(AB)-ER(C) (a fusion of the A/B region of hRAR␥1 with the ER DNA binding domain, see above) in transient transfection assays using the CAT reporter construct, mCRBPII(17m-ERE)-CAT (8). hRAR␥1(AB)-ER(C) activated 3-fold the expression of the reporter, whereas mutation of serines 77 and 79 into alanine abrogated this stimulation, indicating the importance of these serines for AF-1 activity (Fig. 8).
The role of phosphorylation of the AF-1 domain on the transactivation properties of RAR␥ was further investigated using the full-length receptor and a reporter construct containing the CAT gene under the control of a RA-inducible promoter, the natural mRAR␤2 promoter (7). Transcription from this promoter was stimulated by hRAR␥1WT in the presence of a selective ligand (the RAR␥ agonist, BMS961) at 10 Ϫ7 M (Fig.  9A). Deletion of the A/B region abrogated transcriptional activation, in agreement with previous reports (7). Mutation into alanine of Ser-77 and Ser-79 located in the A/B region reduced the transcriptional activity of hRAR␥1 (Fig. 9A), confirming that these phosphorylation sites are required for optimal transcription. That phosphorylation of serines 77 and 79 by cdk7 could be responsible for efficient transcription was further supported by the observation that overexpressed cdk7 significantly enhanced transcription by hRAR␥1WT but not by hRAR␥1S77A/S79A (Fig. 9B). This increase did not occur with cdk7 devoid of kinase activity through mutation within its ATP binding site (cdk7m) (Fig. 9C). Importantly, overexpression of other proline-dependent kinases such as cdk1 or p44 MAPK did not enhance the transcriptional activity of hRAR␥1 (data not shown).
Because the phosphorylation sites located in the B region are the same in RAR␥1 and RAR␥2 isoforms and are conserved between human and mouse (see above and Fig. 1), similar results were expected with mRAR␥2. In fact, transactivation by mRAR␥2 was also reduced by mutation of serines 66 and 68 into alanine (Fig. 9D) and overexpressed cdk7-enhanced tran-  scription by mRAR␥2WT but not by mRAR␥2S66A/S68A (Fig.  9E). Interestingly, mRAR␥2⌬F and mRAR␥2S440A were as active as mRAR␥2WT (Fig. 9D), indicating that, at least under these conditions, the phosphorylation site located in region F is dispensable for the transcriptional activity of mouse RAR␥.
Phosphorylation Modulates the Transcriptional Activity of RAR␥ in a Responsive Gene-dependent Manner-Because the transactivation activity of the A/B region of RARs has been shown to be responsive gene-dependent (7,8), the role of the phosphorylation of this region was also studied with another RA-inducible promoter, the synthetic (TRE3)3tk promoter (29). Both hRAR␥1 and mRAR␥2 stimulated transcription in the presence of the ligand (Fig. 10, A and B). However, with this promoter, hRAR␥1S77A/S79A and mRAR␥2S66A/S68A exhibited higher activities (Fig. 10, A and B). Similar results were obtained with hRAR␥⌬AB and mRAR␥⌬AB (data not shown). Moreover, overexpressed cdk7 was without effect (Fig. 10A and data not shown). Note, however, that as described above with the mRAR␤2 promoter, mRAR␥2⌬F and mRAR␥2S440A were as active as mRAR␥2WT (Fig. 10B). Thus, phosphorylation of the A/B region of RAR␥ modulates transcription in a responsive gene-dependent manner, as it clearly affects differentially transcription from the (TRE3)3tk and mRAR␤2 promoters. DISCUSSION In this study, we have demonstrated that RAR␥ is phosphorylated in the N-terminal region that contains the activation function AF-1 (8) and plays an essential role in RA-induced primitive endodermal differentiation of F9 cells (13). This phosphorylation, which involves two serine residues, is ligand-independent and appears to be most efficiently performed by the cyclin H-dependent kinase cdk7, a component of the general transcription/DNA repair factor TFIIH (17,41,42). Interestingly, this phosphorylation modulates the activity of AF-1 in a responsive gene-dependent manner.
RAR␥ Is Phosphorylated by cdk7 in Its B and F Regions- RAR␥ is phosphorylated at two phosphorylation sites located in the B region. These sites are present in both the ␥1 and ␥2 isoforms and are conserved between human and mouse (21,23). They have been identified to serines 66 and 68 in RAR␥2 and to serines 77 and 79 in RAR␥1 (see Fig. 1). In this latter case, our data show that phosphorylation of serine 77 depends on that of serine 79. As previously reported for RAR␣ (14), we demonstrate that these sites are phosphorylated by cdk7, a cyclin H-and MAT1-dependent kinase. Indeed, overexpression of wild type cdk7, but not of cdk7 mutated at its ATP binding site or of any other cdk, results in a higher level of phosphorylation of the various RARs so far tested (Ref. 14). 3 Moreover, similarly to RAR␣, the pattern of phosphorylation of RAR␥ appears to be independent of the phases of the cell cycle (Ref. 14). 3 As previously reported for RAR␣, RAR␥ is also phosphorylated in its F region. However, this phosphorylation concerns only mouse RAR␥ (either the ␥1 or the ␥2 isoforms) and not its human counterpart, due to the lack of conservation of this region between human and mouse (see Fig. 1). In contrast to phosphorylation of region B, no role has yet been found for this F region phosphorylation, either in RAR␣ or in RAR␥.
RAR␥ Is More Efficiently Phosphorylated by TFIIH than by Free CAK-Cdk7 is associated with cyclin H and MAT1 in the CAK complex, and in the cell CAK is found either free or complexed with TFIIH, a general transcription factor also involved in DNA repair (17)(18)(19)(20). Interestingly, we have shown that RAR␥ is more efficiently phosphorylated in vitro by cdk7 when included in TFIIH rather than in CAK, as previously reported for RAR␣ (14) and for the CTD of RNA polymerase II (19,41,43,44). That RARs are more efficiently phosphorylated by TFIIH than by CAK may result from RAR interactions not only with cdk7, as previously reported (Ref. 14 and Fig. 5), but also with core subunits of TFIIH. Indeed, in coimmunoprecipitations experiments performed with insect cells coinfected with baculoviruses expressing different subunits of TFIIH and either RAR␥ or RAR␣, we found that both RARs interact not only with cdk7 in CAK and TFIIH but also with several subunits of the TFIIH core. Thus, these multiple interactions may account for more efficient phosphorylation by cdk7 within TFIIH than within free CAK. Note that other transcription factors such as p53 and E2F-1, which have been shown to be phosphorylated by cdk7, also interact with the core subunits of TFIIH (45)(46)(47).
How cdk7 and the different TFIIH subunits interact with RARs remains to be investigated, but it is already clear from this and previous (14) studies that the N-terminal A/B region is not mandatory for these interactions. Moreover, the interaction of TFIIH with RARs is not sensitive to deletion of the AF-2AD core/helix 12, 3 which is involved in the coactivator binding surface of the ligand binding domain (1). This is in accordance with our observation that the interaction of RARs with cdk7 and TFIIH is ligand-independent. Moreover, it suggests that the interaction between RARs and TFIIH involves another surface and is therefore mechanistically distinct from that described between RARs and coactivators (see the Introduction).
Regulation of Transcription by RAR␥ through Phosphorylation of the AF-1 Domain by TFIIH-associated cdk7-To investigate whether phosphorylation of the B region of RARs could modulate the ligand-induced activation of transcription, two reporter genes under the control of different responsive elements and promoters were tested: the natural mRAR␤2 promoter, which contains a RARE with directly repeated motifs separated by 5 nucleotides (DR5), and the synthetic (TRE3)3tk promoter, which contains inverted (palindromic) repeated motifs. In both cases, RAR␥ activated transcription upon ligand binding. However, mutation of the phosphorylation sites located in the A/B region reduced transcription from the mRAR␤2 promoter-based reporter gene, whereas it enhanced that from the (TRE3)3tk promoter-based reporter gene. The three-dimensional conformation of bound RXR/RAR heterodimers is most likely different on the two types of response elements. This may result in distinct steric conformations of the AF-1-activating domain and, therefore, in different interactions with putative AF-1 coactivators, which could be differentially modulated by phosphorylation. In this respect we note that interactions between coactivators and the AF-1-activating domain of either the estrogen receptor ER␤ or the nuclear receptor SF-1 have been recently shown to be modulated by AF-1 phosphorylation (48,49). In any event, such a possibility is in accordance with our previous report showing that phosphorylation of RAR␥ AF-1 is differentially required for RAinduced expression of target genes in F9 cells (13). Additionally, RAR phosphorylation may modulate the activity of TFIIHassociated cdk7 and/or regulate the enzymatic activity of some TFIIH subunits such as XPB and XPD that possess ATPase and helicase activity (18, 50 -54) and are involved in distinct transcriptional steps (17).
In conclusion, we have demonstrated that, as other transcriptional regulators such as p53 and E2-F (45,46,55), RARs are targets for phosphorylation by cdk7 and interact with TFIIH. Thus, phosphorylation by TFIIH may be a general way of modulating the activity of transcriptional regulators. However, our present data also show that another proline-dependent kinase, p44 MAPK , can also phosphorylate RAR␥ in vitro. Even though phosphorylation by this kinase is different from that achieved with cdk7 and overexpression of MAPK in COS cells does not affect the phosphorylation level and the transactivation properties of RAR␥, our present data do not rule out the possibility that p44 MAPK could modulate RAR␥ activity in other cell types upon activation of the growth factor/MAPK cascade, as previously reported for ER␣ (56).