Distinct Roles of the Steroid Receptor Coactivator 1 and of MED1 in Retinoid-induced Transcription and Cellular Differentiation*

Retinoic acid receptors (RARs) are the molecular relays of retinoid action on transcription, cellular differentiation and apoptosis. Transcriptional activation of retinoid-regulated promoters requires the dismissal of corepressors and the recruitment of coactivators to promoter-bound RAR. RARs recruit in vitro a plethora of coactivators whose actual contribution to retinoid-induced transcription is poorly characterized in vivo. Embryonal carcinoma P19 cells, which are highly sensitive to retinoids, were depleted from archetypical coactivators by RNAi. SRC1-deficient P19 cells showed severely compromised retinoid-induced responses, in agreement with the supposed role of SRC1 as a RAR coactivator. Unexpectedly, Med1/TRAP220/DRIP205-depleted cells exhibited an exacerbated response to retinoids, both in terms transcriptional responses and of cellular differentiation. Med1 depletion affected TFIIH and cdk9 detection at the prototypical retinoid-regulated RARβ2 promoter, and favored a higher RNA polymerase II detection in transcribed regions of the RARβ2 gene. Furthermore, the nature of the ligand impacted strongly on the ability of RARs to interact with a given coactivator and to activate transcription in intact cells. Thus RAR accomplishes transcriptional activation as a function of the ligand structure, by recruiting regulatory complexes which control distinct molecular events at retinoid-regulated promoters.

of transcription factors, other coactivators or components of the basal transcription machinery. They possess enzymatic activities, catalyzing post-translational modifications of histones and of other transcriptional regulators. Their functional integrity is therefore required to recruit the basal transcriptional machinery to activate gene transcription. Transcriptional activation by liganded nuclear receptors is a paradigm to study promoter activation in response to small hydrophobic molecules, and this process is achieved by the sequential dismissal of corepressors and recruitment of distinct classes of coactivators, each serving one or several specific functions. Many of these functions are targeted to chromatin which, under its compacted form, precludes gene expression.
All-trans retinoic acid receptors (RARs) belong to the nuclear hormone receptors (NRs) superfamily and act as ligand-inducible transcription factors. The acquisition of a transcriptional activity by RARs results from structural transitions occurring in the ligand binding domain or LBD, leading to the formation of an hydrophobic coactivator binding pocket, with which an LXXLL motif from a coactivator molecule will interact (1,2). A charge clamp stabilizes this interaction, allowing the docking of multiprotein coactivator complexes. It is thought that transcriptional activation by RARs requires the sequential recruitment of (reviewed in Ref. 3): (i) ATP-dependent chromatin remodeling complexes which affect the mobility of nucleosomes to alleviate, in most cases, chromatin-based repression. More specifically, tight binding of RXR/RAR heterodimers to DNA requires an ATP-dependent ISWI-mediated chromatin remodeling activity (4). (ii) Acetyl transferases such as the p160-related coactivator family (SRC1, 2, and 3), the integrator complex CBP/p300 and pCAF (5,6). The recruitment of these coactivators favors histone acetylation at least for some retinoid-regulated promoters (7). (iii) The mediator complex (DRIP/TRAP/SMCC, Ref. 8), which allows the phosphorylation of RNA polymerase II (RNApol2) by TFIIH and its conversion into an elongation-competent form (8).
Whereas being relatively well established for specific model systems (7,9), this mechanism is however not universal. We have established that the retinoid-controlled RAR␤2 promoter is, in P19 embryonal carcinoma cells, highly responsive to retinoid stimulation and that histones associated to this promoter are constitutively acetylated (10,11). Transcriptional activation of this promoter is correlated with histone H3 phosphorylation, a post-translational modification reminiscent of those occurring in immediate-early gene promoters (11). A much less explored area is the actual contribution of each class of coactivator to RAR-mediated transactivation.
We investigated this latter question by assessing the respective role of p160 coactivators and of the mediator complex in retinoid-induced transcription and cellular differentiation. P19 cells are multipotent and differentiate into endoderm, mesoderm, or ectoderm depending on the chemical inducer and culture conditions (12). Retinoids promote P19 cells differentiation into neurons and glial cells, and the ␣-isotype of RAR is critical for this process to take place (13). P19 cells are thus an appropriate developmental system to study the role of retinoids, RARs and their coactivators in neuronal differentiation. Moreover, the expression of RAR␤2 is critical for all-trans retinoic acid (atRA)-induced differentiation of P19 cells (14), thus providing a model in which transcriptional and differentiation processes can be studied simultaneously. We have manipulated p160-related factors and DRIP205/TRAP220/Med1 (hereafter termed Med1 according to the unified nomenclature, Ref. 15) expression levels in P19 cells using RNAi, and the consequences of mRNA knock-down on transcriptional and differentiation events have been characterized. In initial experiments, we discovered that a decreased level of expression of SRC2 or of SRC3 induced cell death, suggesting a contribution of these proteins to yet unknown critical biological processes. We therefore focused our study on SRC1 and Med1, which are two potentially important players in retinoid-induced transcription. SRC1 has been shown to play a critical role, although partially filled by SRC2, in steroid-induced tissue development (16,17) and transcriptional regulation by steroid and thyroid hormones (18,19). Its contribution to retinoid-controlled transcription is much less documented, but SRC1 binds physically to RAR (6,20) and its overexpression in P19 cells increases RXR/RAR heterodimers transcriptional activity (6,10). The acetyl transferase activity of SRC1 (21) is dispensable for atRA-induced transcription (5).
Hypomorphic Med1 mice highlighted the role of this coactivator in hepatic and cardiovascular development (22), and mouse embryonic fibroblasts (MEFs) isolated from Med1 Ϫ/Ϫ mice show strongly impaired thyroid hormone receptordependent transcription (23). One-hybrid assays revealed a moderate contribution of Med1 to the AF2 function of RAR␣, which was not detected at the level of p21 expression, an endogenous gene regulated by RAR␣ (23). Thus, although Med1 interacts physically with RAR␣ through LXXLL motifs (24), it remains to be established whether it actually serves as a coactivator for RARs.

MATERIALS AND METHODS
Plasmids-Oligonucleotides encoding for small hairpin RNAs were synthesized with appropriate loop and cohesive ends sequences according to the plasmid provider instructions. Oligonucleotides were cloned into the pSHAG plasmid (obtained from G. Hannon) or into pSilencer 2.1-U6 (Ambion Inc., Austin, TX). Oligonucleotide sequences were selected according to siRNA design guidelines (25). Two or three siR-NAs were designed for each target mRNAs and tested for their efficiencies to selectively decrease target gene expression. Selected siRNAs were: si-Luciferase: cttacgctgagtacttcga; si-SRC1: tgaccgcaccatccatcct; si-Med1: cgtacccacagccagtgtc. All constructions were checked by sequencing. Detailed sequences are available upon request.
Cell Culture and Transfections-HeLa cells were grown in Dulbecco's modified Eagle's medium (DMEM) supplemented with 10% fetal calf serum (Biowhittaker), 1,000 units of penicillin and 10 g of streptomycin per ml. One day prior to transfection, cells were plated in 6-well plates in DMEM supplemented with 10% fetal calf serum, 1,000 units/ml penicillin, and 10 g/ml streptomycin. Each well was transfected using Lipo-fectamine2000 according to the manufacturer's instructions (Invitrogen), with a DNA mixture including 1 g of a RA-responsive firefly luciferase reporter gene containing three repeats of a composite GRE-RARE (GRARE 3 -tk Luc), 0.25 g of the expression vectors pSG5hRAR␣ and pSG5RXGR␣, the latter containing mutations in the P box to confer specific binding to a glucocorticoid response element (26), 50 ng of a control plasmid tk-luciferase (tk-Renilla, Promega), and 2.5 g of pSilencer Control, pSilencerMed1 or pShagSRC1. The total amount of transfected DNA was adjusted to 4 g per well. After a 5-h incubation with the DNA mix, cells were washed and cultured in fresh medium. 16 h later, cells were challenged with 1 M atRA overnight. Cells were harvested and luciferase activities were assayed with the Dual-Glo Luciferase Assay System (Promega). Firefly luciferase values were normalized to those of Renilla luciferase values.
Generation of P19 Cell Subclones-P19 cells with a significantly decreased coactivator expression were obtained as follows: the pSHAG-SRC1 or the pSilencerMed1 plasmid was cotransfected with a vector encoding the green fluorescent protein (pEGFP, Clontech) in a 1:10 ratio. Selection of resistant clones was carried out for 10 -15 days in DMEM containing 300 g/ml G418 (SRC1) or 350 g/ml hygromycine (Med1). After selection based on antibiotics resistance and EGFP expression, resistant colonies were isolated, expanded, and characterized for Med1 or SRC1 expression by RT-PCR and Western blot analysis.
Cellular Extracts Preparation and Western Blot Analysis-Cellular extract preparation and Western blot analysis were carried out as previously described (10,27). Immunodetections were performed using a polyclonal anti-SRC1 antibody (M-341, Santa Cruz Biotechnology), a polyclonal anti-Med1 antibody (C-19, Santa Cruz Biotechnology) and an anti-actin monoclonal antibody (ac-15, Sigma).
Cell Cycle Analysis-10 6 cells were trypsinized, washed once with culture medium, twice with PBS 1ϫ, and fixed with 90% ethanol/phosphate-buffered saline overnight at Ϫ20°C. Cells were rehydrated, washed twice in PBS 1ϫ, and stained for 30 min in a 50 g/ml propidium iodide solution containing 0.25 g/ml RNase and 0.1% Triton X-100. The cell cycle repartition of fixed cells was analyzed by flow cytometry in a EPICS XL-MCL TM cytometer (Beckman-Coulter), and quantified with the WinCycle software (Phoenix Flow Systems). When indicated, cells were treated with 40 M of the caspase inhibitor z-VAD.fmk (Bachem) for 24 h.
Neuronal Differentiation-Neural differentiation was induced as follows: 10 7 P19 cells were cultured for 4 days in 60-mm bacterial grade Petri dishes in DMEM-10% fetal calf serum. Cell aggre- JULY 21, 2006 • VOLUME 281 • NUMBER 29 gates were dissociated by trypsin treatment and grown in tissue culture dishes in DMEM-10% fetal calf serum for 1 day. Adherent cells were then grown in the presence of atRA for 48 h at the indicated concentrations, followed by 48 h in the absence of atRA. Neuronal differentiation was assessed by Western blot analysis of whole cell extracts using a mouse anti-␤III tubulin antibody (clone 2G10, Upstate Biotech, Inc.).
Quantitative PCR-mRAR␤2 transcripts were detected as described in Ref. 28. CRABPII transcripts were assayed using an "Assay on Demand" kit from Applied Biosystems. The 18 S primers and Vic probe were purchased from Applied Biosystems. Reactions (40 cycles) and data analysis were carried out on an ABI Prism 7700 using the SDS software (PerkinElmer-Applied Biosystems). Expression levels of RAR␤2 and CRABPII transcripts were normalized to 18 S RNA levels, and relative levels of expression of each transcript were calculated using the 2 Ϫ⌬⌬Ct method.
, anti-cdk9 (H-169), and anti-Med17 (G-17) were from Santa Cruz Biotechnology. Anti-phospho-Ser 5 RNApolII (H14) was from Covance. ChIP analysis was performed at least in triplicate using distinct DNA preparations. When indicated (Fig. 3), DNA was quantified by Q-PCR using the ABI PRISM 7700 sequence detection system. The RAR␤2 promoter sequence was in this case amplified from Ϫ78 to ϩ38 with the following primers: Forward: TTGAAGGTTAGCAGCCCGG, Reverse: CTTCTGTCACACGGAATGAAAGAT, probe FAM/TAMRA: AAGGTTCACCGAAAGTTCACTCGCA. DNA was quantified and results expressed relative to input DNA, after subtracting nonspecifically bound DNA as assayed using nonspecific IgG in the ChIP assay.

The Transcriptional Activation Process in EC P19 Cells-P19
cells express all isotypes of RAR and RXR, and many of the established or putative corepressors (CoRs) and CoAs for RARs, including SRC1 and Med1 (see supplementary Fig. S1). The detection of transcriptional regulators, as well as of histone modifications occurring at the RAR␤2 promoter ( Fig. 1A) upon atRA challenge were carried out in this cellular background by ChIP assays (Fig. 1B). Histones H3 and H4 were constitutively acetylated, whereas ligand-dependent histone H3 phosphorylation was observed. No phosphorylated H1 could be detected on this promoter. SRC1 and RAR␣ binding was constitutive and ligand-insensitive, whereas the basal level of Med1 detection increased in the presence of atRA. Mediator complex recruitment is facilitated, in a chromatinized environment, by histone acetylation (31), providing a molecular basis for the constitutive high detection of the mediator complex to the RAR␤2 promoter, where both H3 and H4 are acetylated. Also consistent with this hypothesis is the finding that SRC1 is constitutively associated to the promoter, thus promoting the permanent recruitment of CBP/p300, and tethering of HAT activity to the RAR␤2 promoter. Thus SRC1 and Med1 are likely to play a role in the transcriptional control of the RAR␤2 promoter, which exists in a state poised for transcription.
Characterization of P19 Cells Deficient in SRC1 or Med1-P19 subclones overexpressing small hairpin RNA (shRNA) targeting either the luciferase gene (control), SRC1, or Med1 were selected. Efficiencies of SRC1 and Med1 mRNAs knock-downs were characterized in several subclones and compared with wild type (wt) cells. Wild-type cells behaved similarly to cells expressing an shRNA targeting the luciferase gene, which left unchanged SRC-1 and Med1 protein levels and did not affect RA-induced gene transcription (data not shown). Results obtained with representative subclones are shown (Fig. 1C) and were similar for several subclones: mRNA and, more importantly, protein levels were significantly decreased in both cases by more than 80% (Fig. 1D). In the case of Med1, we noted that the protein content decreased more strongly than mRNA levels, in agreement with observations showing that RNAi may also act on RNA translation. We monitored Med1 mRNA levels in SRC1RNAi cells and vice versa, as well as that of SRC2, which have been shown to be up-regulated in SRC1 Ϫ/Ϫ mice (16). No nonspecific alteration of the expression of coactivators was observed (Fig. 1E), showing that selected shRNAs affected specifically the stability of targeted mRNAs. OAS1 expression was The steady-state level of each mRNA species was assayed as described under "Materials and Methods" and normalized to 18 S RNA level. Results are expressed relative to the basal level observed in wtP19, which was arbitrarily set to one. B, Q-PCR analysis of CRABPII transcripts. C, transcriptional activation of a retinoic acid-inducible reporter gene in HeLa cells. HeLa cells were co-transfected with expression vectors coding for RXR␣, RAR␣, and the DR5tk-Luc reporter gene. Vectors encoding for a scrambled, antiSRC1, or antiMed1 shRNA were transfected as indicated, and luciferase activity were assayed and normalized as indicated under "Materials and Methods." The luciferase activity detected in control cells upon atRA treatment was arbitrarily set to 100%.

Med1-independent Transcriptional Activation by Retinoids
not up-regulated in P19 subclones (data not shown), indicating that siRNAs did not induce an interferon-like response (32). Thus by these criteria, shRNA overexpression triggered a specific degradation of targeted mRNAs.
SRC1 and Med1 Have Distinct Contributions to the Transcriptional Regulation of Retinoid-regulated Genes-Functional consequences of SRC1 or Med1 knock-down were assessed by monitoring the activity of two retinoid-responsive genes (Fig.  2). Retinoid responsiveness is conferred to the RAR␤2 gene promoter by a retinoic acid response element (RARE) organized as a direct repeat of a hexanucleotide separated by a 5-bp spacer (DR5) (Refs. 33 and 34, see Fig. 1A). Quantitative PCR analysis ( Fig. 2A) showed that RAR␤2 transcripts accumulated in wtP19 cells according to a first order kinetics, reaching a plateau after an 8-h treatment with 1 M atRA. Accumulation of RAR␤2 mRNAs in SRC1RNAi cells followed similar kinetics, although the maximal level was decreased when compared with wtP19 (45-fold induction versus 20-fold). Most notably, Med1 knock-down promoted a much faster mRNA synthesis at earlier time points, without affecting significantly the maximal level which was reached, as in wtP19 cells, after an 8-h treatment. Of note, the basal expression of the RAR␤2 gene was strongly increased in Med1RNAi cells (4 -5-fold), hinting at a repressive role for this protein in this cellular context. Overexpression of the appropriate coactivator was able to rescue transcriptional activity in depleted cells (data not shown).
To investigate whether both coregulators exerted a similar control on other retinoid-regulated genes, we monitored cytoplasmic retinoic acidbinding protein II (CRABPII) gene expression by Q-PCR, whose expression is regulated through a DR1 and a DR2 RAREs (35). In wtP19, CRABPII transcripts accumulated very slowly, on an almost linear, zero order kinetics (Fig. 2B). Knocking down SRC1 expression dramatically decreased atRA-induced CRABPII mRNA synthesis. Med1 depletion increased the basal level of expression of CRABPII but, in contrast to the RAR␤2 gene, allowed a more efficient ligand-dependent accumulation of CRABPII transcripts. Similar results were obtained with different P19 subclones. Med1 and SRC1 depletion impacted similarly on the response of a retinoid-inducible synthetic reporter gene in transiently transfected HeLa cells (Fig. 2C).
CRABPII participates in retinoidmediated transcription by interacting directly with RXR/RAR heterodimers and increasing their transcriptional activity. This activity requires the binding of retinoids to CRABPII (36). In light of these data, the decreased expression of the RAR␤2 gene in SRC1RNAi cells could be interpreted as a result of a decreased expression of CRABPII. However, we obtained similar results using a panel of synthetic retinoids binding or not to CRABPII (data not shown), excluding a possible contribution of a decreased CRABPII expression to the observed phenotype.
Taken together, our data suggest that SRC1 acts as a coactivator in distinct cellular backgrounds, whereas Med1 exerts a repressive activity on various retinoid-regulated promoters, irrespective of the cell type.
SRC1 and Med1 Knock-downs Affect Transcriptional Events at the RAR␤2 Promoter-Both RAR␤2 and CRABPII gene transcription are affected by SRC1 or Med1 knock-downs. To focus solely on early events leading to transcriptional activation, we selected to study those occurring at the RAR␤2 gene promoter after a 2-h stimulation by atRA. Indeed, at this early time point, coactivator knock-down does not alter significantly CRABPII transcript accumulation, whereas RAR␤2 gene transcription is clearly affected. Although knock-down effects are more pronounced at later time points for the CRABPII gene, it is likely that events unrelated to primary transcriptional regulation will be involved in this process. We thus investigated whether SRC1 and Med1 knock-downs altered H3 and H4 post-translational modifications and transcription factors detection at the RAR␤2 promoter (Fig. 3). . Histone post-translational modifications and transcription factor loading at the RAR␤2 promoter. P19 cells were treated or not by 1 M atRA for 1 h and the association of phosphorylated histone H3, acetylated histone H3 or H4, RNApol2, RAR␣, SRC1, and of subunits of the mediator complex to the RAR␤2 promoter was analyzed by ChIP assays using antibodies as indicated in Fig. 1. The association of phosphorylated RNApol2, cdk8, cdk9, and TFIIH/ERCC3 to the RAR␤2 promoter was similarly monitored by ChIP assays. Immunoprecipitated DNA was quantified by Q-PCR and expressed as a percentage of input DNA corrected from the background signal.
To monitor Med1 loading, ChIPs experiments were initially carried out using a mix of two antibodies directed against Med1 and Med17, two components of the core complex (also named PC2). As expected, knocking down Med1 expression prevented its association to the RAR␤2 promoter, and therefore that of Med17 as well. Of note, SRC1 knock-down also affected Med1 detection, and most especially the basal level of association with the RAR␤2 promoter.
Because the mediator complex is organized into submodules which can be isolated functionally and physically, we concluded that either the mediator complex is absent from the promoter in Med1RNAi cells, or that it undergoes significant conformational/structural changes preventing epitope recognition or that the mediator complex may exist under a Med1-free form. We thus used another antibody targeted at the tail of the core mediator complex (37). Several proteins are part of the tail of the core complex, including Med24/TRAP100. Med24 has a broad contribution to transcriptional activation (38), and despite the presence of 6 LXXLL motifs, does not interact in vitro with RAR␣ or RXR␣ (39,40). In wtP19, a clear liganddependent association of Med24 was detected in the presence of atRA, and a similar pattern was observed in SRC1RNAi and Med1RNAi cells, suggesting that Med1 depletion does not affect the association of other components of the mediator complex to the RAR␤2 promoter. cdk8, a component of the repressive kinase complex of the mediator complex (41), is associated to the promoter under basal conditions and cannot be detected after agonist treatment (Fig. 3). Med1 knock-down did not alter cdk8 interaction with the promoter region, which was not detected on the promoter in all three cellular backgrounds. However, cdk8 detection at the RAR␤2 promoter was sharply reduced in SRC1-depleted cells, suggesting that SRC1 affects, directly or indirectly, cdk8 association to the promoter, a phenomenon which could be related to the lower density of promoter-bound Med1&Med17 in SRC1 RNAi cells. The possibility that cdk8 is still present but not detectable in this configuration cannot however be excluded at this point.
The constitutive, ligand-insensitive SRC1 binding was severely compromised in the SRC1-deficient background but not affected in Med1 RNAi cells. As expected from its role as a primary transcriptional activator, RAR␣ binding was not significantly altered in SRC1-or in Med1-deficient cells and remained constitutively high in each cellular background.
Histone post-translational modifications were then monitored. In wtP19 cells, a constitutive acetylation of H3 and H4 was detected, as previously shown (Fig. 1). In both SRC1RNAi and Med1RNAi cells, H3 acetylation was not significantly altered. In contrast, SRC1 knock-down caused a slight but consistent decrease in the basal acetylation level of H4, which became ligand-sensitive, increasing 2-fold upon atRA treatment. Med1 knock-down impacted moderately on H4 acetylation levels, with no detectable increase upon atRA challenge. H3 phosphorylation increased in wtP19 upon atRA treatment, as well as in Med1-deficient cells. This post-translational modification was abrogated in SRC1RNAi cells, confirming the relationship between transcriptional activation of the RAR␤2 promoter and H3 phosphorylation.
RNApol2 detection in the promoter region suggested that RNApol2 binding could decrease in the Med1-deficient background, suggesting a faster promoter clearance. Conversion of RNApol2 to an elongation-competent form depends on the orchestrated activity of cyclin-dependent kinases. We therefore monitored the recruitment of two cyclin-dependent kinases involved in the regulation of RNApol2 activity, cdk7 and cdk8. cdk8 binding was monitored using an antibody directed against cdk8 itself, whereas cdk7 loading was assayed by immunoprecipitating the ERCC3 subunit of TFIIH, to which cdk7 is associated. In addition, phosphorylation of Ser5 of RNApol2 CTD was followed as an index of TFIIH activity. Upon atRA treatment, phosphorylation of Ser 5 increased concomitantly to TFIIH detection in wtP19. Very strikingly, this correlation was lost in Med1RNAi cells, in which TFIIH was constantly detected to the promoter, and where no Ser5P RNApol2 could be detected. In addition, cdk9 ChIP assays evidenced a stronger association of this kinase to the RAR␤2 promoter in atRA-treated. These data suggested that the high density of TFIIH at the RAR␤2 promoter in Med1depleted cells could promote a faster dissociation of RNApol2 from promoter sequences, although we cannot not rule out at this stage that this reflects merely a RNApol2 epitope masking in these conditions.
To test further this hypothesis, we carried out ChIP assays to detect RNApol2 on exon 3 of the RAR␤2 gene (Fig. 4). RNApol2 detection increased in a ligand-dependent manner in transcriptionally active cells (wt and Med1 RNAi cells), thus establishing a strict correlation between RNApol2 detection at exon 3 and

Med1-independent Transcriptional Activation by Retinoids
transcriptional activity. Quite strikingly, RNApol2 detection was ligand-insensitive in Med1-deficient cells, providing a molecular basis for the increased basal transcriptional activity of the RAR␤2 gene in this background. A model summarizing these observations and interpreting ChIP assays as a change in association of the monitored factors, is presented in Fig. 5.
SRC1 and Med1 Regulate atRA-induced Cellular Differentiation-atRA-induced P19 differentiation is accompanied by cell growth arrest (42), apoptosis, and accumulation of neuronal markers such as ␤III tubulin. wtP19 accumulated ␤III tubulin as a function of atRA concentration (Fig. 6A). SRC1RNAi cells exhibited a lower ␤III tubulin expression, underlining a lower propensity to differentiate upon atRA treatment. In contrast, Med1 knock-down increased the ability of P19 cells to express spontaneously ␤III tubulin, and this accumulation remained ligand-dependent.
Differentiation is intimately linked to apoptosis in P19 cells (43). Apoptosis was quantified in wt, SRC1RNAi and Med1RNAi cells by flow cytometry (Fig. 6B). atRA-induced cell death was dose-dependent in wtP19 cells, to reach 7% of the cellular population. SRC1RNAi cells were highly resistant to apoptosis, which increased only marginally above control. In contrast, Med1RNAi cells underwent massive apoptosis, affecting 20% of the cells after a 48-h treatment with 1 M atRA (44,45). atRA-stimulated Med1RNAi cells were thus treated with the caspase pan-inhibitor zVAD-fmk. After a 24-h treatment, this inhibitor prevented atRA-induced apoptosis (Fig. 6C), but was barely active after a 48-h atRA treatment (data not shown), indicating that apoptosis becomes, at this stage, irreversible.
Thus Med1 knock-down favors differentiation-induced apoptosis of P19 cells, showing that altering an upstream event, i.e. transcriptional activation by retinoids, impacts on long term processes such as cellular differentiation and apoptosis.
SRC1 Knock-down Impairs Selectively the Transcriptional Response to Synthetic Retinoids-We demonstrated that the structure of retinoids affect the ability of RXR␣/RAR␣ heterodimers to interact with coactivators (20). A prediction from these results is that the loss of expression of a given coactivator could selectively impair the transcriptional response to a specific retinoid. Using GST pull-downs experiments designed to monitor the association of coactivators to a RXR␣/RAR␣ heterodimer bound to a DR5 RARE (20), we assessed the ability of several RAR␣ ligands to promote heterodimers interaction with SRC1 or Med1 in vitro (Fig. 7A), Med1 interacted with RAR␣ in a ligand-dependent manner, irrespective of the nature of the retinoid. In contrast, the interaction with SRC1 was conditioned by the nature of the ligand, displaying the strongest interaction in the presence of TTNPB, a RAR synthetic panagonist (46). The ability of these ligands to activate the RAR␤2 promoter in the wt, SRC1-, and Med1-deficient cellular background was quantified (Fig. 7B). In wtP19 cells, atRA, TTNPB, and Ch55, another synthetic RAR panagonist (47), were equally able to stimulate mRNA transcription from the RAR␤2 promoter. In SRC1-deficient cells, the efficacy of these retinoids to promote transcription was strongly affected, as it could be predicted (see Fig. 2). However, the activity of TTNPB was, in this assay, much more dramatically affected than those of atRA and Ch55. Med1 deficiency affected positively RAR␤2 basal expression, as shown in Fig. 2. Thus TTNPB-mediated activation of the RAR␤2 promoter is highly sensitive to SRC1 levels, showing that synthetic retinoids may select among several coactivators to regulate transcription. Similar assays were carried out for the CRABPII gene. Whereas Med1-deficient cells exhibited a stronger response irrespective of the ligand used, as observed for the RAR␤2 gene, the differential effect of SRC1 depletion could not be precisely quantified, due to the very low residual activity of this gene in a SRC1-deficient background (see Fig. 2 and data not shown).

DISCUSSION
Transcriptional activation by NRs is dependent on the sequential recruitment and dismissal of corepressors and coactivators at the promoter of regulated genes. A plethora of proteins with RAR corepressive or coactivating activities have been identified but their relative role in retinoid-regulated transcription has not yet been clearly established (reviewed in Ref. 48). In a seminal study, Kawasaki et al. (49) demonstrated that p300 and CBP had distinct roles in atRA-induced differentiation and transcription in F9 cells. Using hammerhead ribozymes, they showed that p300 is required, but not CBP, for atRA-induced differentiation of F9 cells. p300 and CBP had opposite roles in the transcriptional regulation of inhibitors of cyclin-dependent kinases, p21 Cip1 and p27 Kip1 . Korzus et al. (5) showed that pCAF is required for retinoid-induced transcription in Rat-1 cells. Finally, RAR-driven transcription is only mildly affected in Med1 Ϫ/Ϫ MEFs (23), despite the broad implication of the Mediator complex and Med1 ligand-dependent interaction in vitro with multiple NRs, including RARs (24).
We aimed at defining the respective role of p160 coactivators and of Med1, the mediator complex subunit which mediates its physical interaction with RAR␣ in vitro, in P19 cells, a relevant model of retinoid action at the transcriptional and cellular differentiation levels. Selective inhibition of the expression of either SRC1 or Med1 in these cells led to distinct transcriptional effects: SRC1 knock-down impaired the transcriptional activation of two retinoid-regulated genes, CRAB-PII and RAR␤2, in agreement with its proposed role as a coactivator. More unexpectedly, Med1 knockdown greatly increased retinoid sensitivity of both genes. Notably, the basal level of transcription of these genes was enhanced by 3-5fold, suggesting that Med1 exerts a repressive activity on both promoters. The mediator complex associates to the RAR␤2 promoter in the absence of ligand (our results and Ref. 50). Cdk8, which is part of the repressive kinase module of the mediator complex, phosphorylates the cyclin H subunit of TFIIH, inhibiting TFIIH kinase activity. This prevents RNApol2 CTD phosphorylation, thereby exerting a repressive effect on transcription (51). The behavior of the cdk8 module was not altered in any case, excluding the possibility that an altered cdk8 structure or association could lead to transcriptional derepression. However, it should be noted that Med1 depletion induces gene-specific repression or induction in yeast, and that functional links between yMed1 and Srb10, the ortholog of mammalian cdk8, have been identified (52).
Retinoid-induced transcription occurs efficiently either with severely diminished (our data) or abolished Med1 expression (23). The constitutive detection of RNApol2 in wtP19 is also observed in the Med1-deficient background, showing that the conversion to the elongation-competent state may be Med1-independent. ChIP experiments with antiMed1 and antiMed17 antibodies failed to detect the association of these two subunits to the RAR␤2 promoter. However, Med24 was detected on the promoter, showing that the mediator complex, which has probably a different molecular composition in the absence of Med1 (52), is present. Thus a compromised recruitment of Med1 promotes a more efficient transcriptional activation by retinoids in P19 cells. This finding is consistent with the mild positive effect of Med1 knock-out on RAR-mediated transcription in MEFs (23), and with the hypothesis that the mediator complex is recruited indirectly to the promoter through interactions with CBP/p300 (31). In the same report, study of mediator complex recruitment to p300 in a chromatinized environment provided data suggesting that Med1 recruitment to p300 is facilitated by histone acetylation. This would explain why we and others (see Figs. 1, 3, and Ref. 50) could detect constitutively the mediator complex, since both H3 and H4 are constitutively acetylated at the RAR␤2 promoter. Also consistent with this hypothesis is the finding that SRC1 is always detected at the RAR␤2 promoter, which in turn would recruit CBP/p300 and thus tether HAT activity to the RARb2 promoter. Indeed, acetyl-H4 levels are decreased in SRC1-depleted cells, and we noted that Med1/Med17 loading was affected in this setting (Fig. 3).
RNApol2 detection is however not indicative of promoter activation: in the SRC1-deficient background, RNApol2 association is ligand-sensitive, but transcriptional activation is severely blunted. In this background, H4 acetylation also becomes ligand-sensitive and is reproducibly decreased compared with wt and Med1RNAi levels. HAT activity has been associated to SRC1 (21), but is not selective for either H3 or H4. Similarly, CBP and p300 displays HAT activities targeted at H3 and H4, and pCAF acetylates preferentially H3 (53). It is therefore unclear whether the observed decrease in H4 acetylation is due solely to the loss of SRC1 association to the promoter, or to the loss of an associated HAT such as CBP/p300. In this regard, it is worth noting that binding of transcriptional activators to HAT dramatically alters HAT activity (54), thus making a formal identification of the H4 acetylase associated to the RAR␤2 promoter still elusive. Histone H3 phosphorylation is detected in the wt and Med1-deficient background, in which efficient transcription is observed, and H3 phosphorylation is lost in SRC1RNAi cells, in which a weak transcriptional activity occurs. Interestingly, phosphorylation of H3 increases H3 K14 acetylation by yeast GCN5 (55), and we observed that H3 acetylation is decreased, but not abolished, in the SRC1-deficient background. Whether a similar process occurs at the RAR␤2 promoter will require kinetics studies and the identification of the histone H3 kinase, which is distinct from Rsk2 and Msk1 kinases (11).
RNApol2 detection was decreased at the RAR␤2 promoter in the Med1-deficient background, suggesting a faster promoter clearance. Conversion of RNApol2 to an elongation-competent form depends on the orchestrated activity of cyclin-dependent kinases. Cdk8 exerts a repressive activity by phosphorylating RNApol2 CTD on Serine 2 and 5, prior to preinitiation complex assembly, whereas cdk7, a component of the TFIIH complex, promotes the conversion toward an elongation mode by phosphorylating Ser 5 . cdk9, a component of the positive transcription elongation factor b (P-TEFb), phosphorylates Ser 2 of the CTD and favors transcript elongation, but associates to a number of RNApol2-regulated promoters and phosphorylates RNApol2 CTD when RNA pol2 is in the promoter clearance mode (56). Upon atRA treatment, phosphorylation of Ser 5 increased concomitantly to TFIIH detection in wtP19. This correlation was lost in Med1RNAi cells, in which TFIIH was always detected on the promoter and where no Ser5P RNApol2 could be detected. In addition, association of cdk9, which decreased upon atRA treatment in wtP19, bound more strongly to the RAR␤2 promoter in Med1RNAi cells challenged with atRA. Taken together, these data suggest that Med1 depletion may favor both the conversion of RNApol2 into an elongationcompetent form, and relieves the elongation blockade shortly after the initiation of transcription, thus facilitating its access to transcribed regions.
P19 cells differentiate into neurons and glials cells upon exposure to atRA. Most neurons exhibit a GABAergic or a cholinergic phenotype (57,58), and express neuronal markers including ␤III tubulin. atRA also reduces the proliferation rate due to an increased duration of the S phase (42). P19 differentiation is linked to atRA-induced apoptosis and is an RAR-and RXR-dependent process (43). Flow cytometry studies revealed that Med1RNAi cells, which expressed a differentiation marker in an atRA-independent fashion, were prone to caspase-regulated apoptosis. This suggests that altering an upstream event, i.e. transcriptional activation by retinoids, impacts on a longterm process, i.e. cellular differentiation. This relationship is further demonstrated by the fact that SRC1-deficient cells, which display a weak transcriptional response to atRA, exhibit a strongly diminished expression of neuronal markers. Ligand docking into the ligand binding pocket induces structural changes which are a function of retinoid structure. Indeed, atRA or 9-cis RA-bound RAR␣ exhibits distinct abilities to release SMRT (59), and synthetic agonists confer to RAR␣ a varying selectivity for coactivators in vitro, which is also conditioned by DNA binding (20). Our data suggest that liganded RAR␣ interacts with SRC1 with different avidities in vivo: SRC1 knock-down affected more severely TTNPB-mediated gene activation, in agreement with the higher affinity of TTNPB⅐RAR␣ complex for SRC1 in vitro. Finally, it suggests that alteration of the expression level of a given coactivator, for example as a consequence of a pathological condition, will lead to a dramatic change in ligand sensitivity of target cells.