The Coactivator p/CIP/SRC-3 Facilitates Retinoic Acid Receptor Signaling via Recruitment of GCN5*

p/CIP/SRC-3 is a member of a family of steroid receptor coactivators/nuclear receptor coactivators (SRC/NCoA) proteins that mediate the transcriptional effects of nuclear hormone receptors (NRs). Using deletion analysis we have mapped the location of two distinct activation domains in p/CIP (AD1 and AD2) capable of activating transcription in mammalian cells when fused to the Gal4-DNA binding domain. In addition to AD1 being coincident with the interaction domain for CBP, we demonstrate a novel in vivo interaction between the AD1 and GCN5. Overexpression of a Gal4-AD1 fusion protein in yeast leads to growth arrest that is relieved by mutation of genes encoding components of the SAGA complex including GCN5, ADA3, and SPT7. In addition, the AD1 of p/CIP and the ADA3 gene are shown to be essential for retinoic acid receptor α-dependent transcription in yeast. Transient transfection assays in mammalian cells indicate that GCN5 cooperates with p/CIP as a coactivator of RARα-dependent transcription. Down-regulation of GCN5 using small interfering RNA in mammalian cells indicates that the AD1 domain and the RARβ promoter activity are dependent, in part, on GCN5. Mutational analysis of AD1 has identified two helical motifs that are required for interactions with GCN5 and CBP. Taken together, these results support a model by which p/CIP functions as a ligand-dependent adapter, through specific protein-protein interactions with AD1, to recruit members from at least two distinct families of acetyltransferase proteins to NRs.

Derivatives of vitamin A such as all-trans-retinoic acid (RA) 1 play essential roles in a variety of biological processes such as embryonic development and cell differentiation. RA exerts its effects primarily by binding to and activating the retinoic acid receptor (RAR) family of proteins that form heterodimers with the retinoid X receptors (RXRs) to regulate the expression of target genes (1). A critical step in the activation of the RAR is ligand binding, which triggers a conformational change in the receptor that results in the recruitment of coactivator proteins. The mouse p300/CBP interacting protein (p/CIP) (2), which has been identified in humans as ACTR/AlB1/RAC3/SRC-3/ TRAM-1 (3)(4)(5)(6)(7), is a member of the p160/steroid receptor coactivator (SRC) family of proteins that also includes SRC1 (8) and GRIP1/TIF2 (9,10). Multiple lines of evidence have shown that the p160 proteins play an integral role in NR signaling events. It has been demonstrated that all the p160 proteins bind to the RAR, as well as other NRs, in a ligand-and AF2-dependent manner, both in solution and on DNA (3)(4)(5). Transient overexpression assays have shown that individual p160 proteins can enhance the transcriptional activities of NRs in response to their respective ligands (8). Single cell microinjection of antibodies against specific p160 proteins block ligand-dependent stimulation of reporter genes containing RAR response elements suggesting that they are required for some NR signaling events (2). In vitro transcription assays using chromatinized DNA templates containing NR response elements have shown that p160 proteins can enhance the transcriptional activity of ligand-bound RARs (11), the estrogen receptor and progesterone receptor (12,13). More recently, chromatin immunoprecipitation assays have shown that p/CIP/SRC-3 can be rapidly recruited to RAR-regulated target genes in response to RA (14).
The mechanism of transcriptional regulation by the p160 proteins has been the subject of intense investigation. The amino terminus of p/CIP contains a conserved basic helix-loophelix PAS (bHLH-PAS) domain also found in a family of transcription factors that includes the dioxin receptor, hypoxia inducible factor 1␣, and several circadian oscillatory proteins (15). The PAS domain is a functionally diverse domain that confers dimerization specificity to many bHLH-PAS proteins (16,17). The function of the bHLH-PAS domain within the context of the SRC proteins is not known although it has been reported that the bHLH-PAS domain of GRIP-1 mediates in vitro interactions with myogenin (18). The nuclear receptor interaction domains (NRID) of all SRC proteins contain three highly conserved motifs, or NR boxes, with a consensus amino acid sequence LXXLL (2, 19 -21). It has been established that the LXXLL motifs mediate direct interaction with the AF-2 domain of RARs, as well as other NRs, upon hormone binding. Furthermore, amino acids directly adjacent to each motif appear to specify the affinity for particular NR homo-or heterodimers (20).
Transcriptional activation is mediated by 2 major activation domains (ADs) found within the COOH-terminal half of all the p160 proteins. The AD1 is the major transactivation domain and serves as an interaction surface for CBP/p300 (2, 3). The CBP and p300 family of coactivators are highly conserved pro-teins that possess protein acetyltransferase activity and can acetylate histone and non-histone proteins. CBP/p300 interact directly with p160 proteins through the AD1 domain, in vitro and in vivo (22)(23)(24)(25)(26). Overexpression of SRC-1 and CBP enhance transcription synergistically (27), and in single-cell microinjection experiments the inhibition of NR signaling by anti-p/CIP antibodies is reversed by the simultaneous overexpression of both p/CIP and CBP (2). Furthermore, mutations that abolish interaction with CBP/p300 impair the ability of p160 proteins to enhance transcription by NRs in vitro (26). Activation domain 2 (AD2) is located in the extreme carboxyl terminus of the p160 proteins. It has been reported that the AD2 domain possesses protein acetyltransferase activity (3,28) and can also interact directly with the coactivator-associated arginine methyltransferase 1 (CARM-1) protein, a member of the argininespecific protein methyltransferase family (29). Transient transfection and in vitro transcription assays using chromatinized templates have shown that CARM-1 can cooperate with CBP/ p300 to enhance the transcriptional effects of RARs as well as other NRs (11,30).
Although the p/CIP CBP/p300 complex may mediate many of the transcriptional effects of NRs, it is conceivable that other factors are also targeted by p/CIP through direct interactions with the AD1 domain. This hypothesis is substantiated by recent evidence that has shown that the protein acetyltransferase p/CAF can function as a NR coactivator by interacting with the AD1 domain of p/CIP (31,32). Furthermore, liganddependent transcriptional activation by the TR in yeast is dependent on both the presence of GRIP1 and the yGCN5 complex (33).
In the present study we show that, in addition to CBP, GCN5 is recruited to p/CIP via direct interaction with the AD1 domain. Using yeast as a model system we demonstrate that the minimal AD1 domain of p/CIP is a potent inhibitor of yeast growth when expressed as a fusion protein with the Gal4p DNA binding domain (DBD). Importantly, the growth inhibition and ligand-dependent transcriptional activation by RAR␣ in yeast requires the AD1 domain of p/CIP, yGCN5 as well as other components of the SAGA complex. In mammalian cells, loss of function experiments using siRNA against GCN5 indicates that both the AD1 transcriptional activity and RA-mediated transcriptional activation of the RAR␤ promoter are, in part, dependent on GCN5. Mutagenesis analysis of amino acids found within the minimal AD1 domain indicates that the interaction surface between p/CIP and CBP or GCN5 is highly overlapping. Importantly, we have identified mutations that severely compromise interactions with GCN5 while leaving the interaction with CBP intact. Collectively these results provide support for a model in which p/CIP functions as an essential bridging protein, which may be capable of assembling complexes containing distinct proteins at nuclear receptor target genes.
Plasmids and Mutagenesis-PCR reactions were performed using PCR Supermix High Fidelity (Invitrogen) according to the manufacturer's recommendations. Mutations in p/CIP were introduced using a QuikChange TM site-directed mutagenesis kit according to the manufacturer's directions (Stratagene, La Jolla, CA). PCR amplified products from both wild-type and mutant pCMX-p/CIP constructs were subcloned into the PJG4 -5 yeast expression vector in-frame with the B42 activation domain, or the pMGal vector, in-frame with the Gal4 DBD. The fragments of p/CIP corresponding to AD1 (aa 985-1085) and AD2 (aa 1300 -1398) were also subcloned into the yeast expression vector pAS1U in-frame with the Gal4 DNA binding domain. The human GCN5-short form (aa 292-390) and mouse CBP (aa 2058 -2170) were PCR amplified and the products were subcloned into the PEG 202 yeast expression vector in-frame with the LexA-DBD.
Yeast Interaction Assays-Yeast two-hybrid interaction assays were performed as described (35). The EGY48 yeast strain containing the LexA ␤-galactosidase reporter plasmid (pSH18 -34) was grown overnight at 30°C in selection media containing 2% glucose. Cells were cotransformed the following day with PEG202 and PJG4 -5 plasmids containing the various in-frame coding sequences of interest as indicated in the figure legends. Cells were plated onto 2% glucose selection media and allowed to grow at 30°C for 2-3 days. Isolated colonies were grown in the same liquid selection media containing 2% galactose, and grown at 30°C to an A 600 nm of ϳ0.5-0.8 to induce expression from the PJG4 -5 plasmids. Liquid ␤-galactosidase assays were normalized to cell density. For ligand-dependent interactions, two separate cultures were induced and treated with either vehicle or 10 Ϫ6 M RA. Values reported are the average -fold of activation over PJG4 -5 alone, which was arbitrarily set to 1.
To examine the effect of p/CIP on yeast growth, parental and knockout yeast strains were transformed with the plasmid pAS1U containing the ADH1 promoter directing the expression of Gal4-DBD alone, the Gal4 DBD fused in-frame with AD1 (aa 985-1085) or AD2 (aa 1300 -1398). Cells were plated onto synthetic selection media minus uracil containing 2% glucose and allowed to grow for 3 days at 30°C before growth was recorded photographically with an ImageMaster VDS (Amersham Biosciences).
Transactivation Assays in Mammalian Cells-HeLa S3 cells and C127 cells (transformed mouse mammary gland) cells were routinely maintained in a humidified 37°C incubator with 5% CO 2 in T-75 flasks in 10 ml of Dulbecco's modified Eagle's medium containing 10% fetal bovine serum, 0.1 unit/ml penicillin G, and 0.1 g/ml streptomycin sulfate. To assess transcriptional activity of Gal4-DBDp/CIP fusion proteins, transient cotransfections were performed according to a modified version of the calcium phosphate procedure. HeLa S3 cells were seeded at ϳ500,000 per 60-mm plate and cotransfected with 1 g of luciferase reporter plasmid containing 6 binding sites for Gal4 upstream of a minimal promoter, 1.5 g of a CMV driven ␤-galactosidase reporter plasmid as an internal control to monitor transfection efficiency, and 1 g of the pMGal vector (CMV) directing the expression of Gal4-DBD or p/CIP deletion mutants fused in-frame with Gal4 DBD. Total DNA was adjusted to 10 g per 60-mm plate with empty plasmid and transfected cells were incubated in 5% CO 2 at 37°C for 24 h. Cell extracts were prepared in 300 l of 1ϫ reporter lysis buffer (Promega). 50-l aliquots of the extract were used for subsequent determination of luciferase and ␤-galactosidase activity according to the manufacturer's instructions (Promega). Luciferase values were normalized to the output of the internal ␤-galactosidase control plasmid and the folds of activation are relative to that of pMGal alone, which was arbitrarily set to 1.
For the NR signaling assays, HeLa S3 cells were seeded at 840,000 cells per well of a 6-well plate and transiently cotransfected according to the LipofectAMINE Plus protocol with 150 ng of a luciferase reporter plasmid containing two consensus (DR5) RAR response elements upstream of a minimal promoter, 150 ng of a CMV-driven ␤-galactosidase reporter plasmid, 50 ng each of CMV vectors directing the expression of hRAR␣ and hRXR␣, and increasing amounts of a CMV vector driving the expression of mGCN5 and/or p/CIP. Total DNA was adjusted to 1 g/well of a 6-well plate with empty CMV vector. The following day, cells were treated with either vehicle alone or 10 Ϫ6 M RA and incubated in 5% CO 2 at 37°C for 24 h. Cells were harvested and extracts were prepared in 200 l of lysis buffer. 50-l aliquots of the extract were used for subsequent determination of luciferase and ␤-galactosidase activity according to the manufacturer's instructions (Promega). For the lossof-function experiments, C127 cells were transfected using Lipo-fectAMINE 2000 as described. Synthesized small interfering RNA (siRNA), obtained from Dharmacon using the SMART protocol, were cotransfected into the cells. Luciferase values are normalized to the output of the internal ␤-galactosidase control plasmid and all values are relative to that of the basal condition (RAR/RXR alone) in the absence of ligand that was arbitrarily set to 1. The relative expression of endogenous GCN5 was monitored by Western blotting of cell extracts isolated from control or siRNA-treated cells, using a polyclonal antibody against GCN5 (Santa Cruz).
Expression and Purification of Recombinant Proteins-For the GST pull-down assays p/CIP (985-1085) was subcloned into the pGEX bacterial expression vector in-frame with the GST moiety, and the resulting plasmid was transformed into the Escherichia coli strain BL-21. Single colonies were grown overnight in 3 ml of LB and the following day, cultures were diluted 1:100 in 300 ml of LB and grown at 37°C with shaking to an A 600 nm of ϳ0.5, at which time protein expression was induced with 0.5 mM isopropyl-1-thio-␤-D-galactopyranoside (final) for 5 h at room temperature. Cells were pelleted by centrifugation at 4°C for 10 min at 10,000 ϫ g and resuspended in 10 ml of lysis buffer A (50 mM Tris, pH 8.0, 300 mM KCl, 0.1% Nonidet P-40, 10% glycerol, 0.5 mM EDTA, 0.5 mM EGTA, 0.5 mM dithiothreitol, 0.2 mM phenylmethylsulfonyl fluoride, and 5 g/ml each of leupeptin, pepstatin, and aprotinin). The cells were then sonicated (4ϫ 15 s bursts at 50%) prior to centrifugation of the samples at 30,000 ϫ g for 1 h at 4°C. 500 l of GST-Sepharose beads prewashed in resuspension buffer were added to the clarified supernatant and the samples were incubated at 4°C with rocking for 1 h. The beads were pelleted by centrifugation at 4°C for 2 min at 1000 rpm and subsequently washed 5 times with 12 ml of resuspension buffer. The beads were resuspended in 1 ml of resuspension buffer containing 10 mM reduced glutathione and incubated for 5 min with rocking. Proteins were dialyzed against a buffer consisting of 20 mM Tris, pH 7.9, 100 mM KCl, 10% glycerol, 0.5 mM EDTA, and 0.5 mM dithiothreitol and stored at Ϫ80°C.
FLAG-tagged CBP, FLAG-tagged GCN5, FLAG-tagged p/CIP, and HA-tagged p/CIP (574 -1398) cDNAs were subcloned into the pFastbac vector (Invitrogen) and recombinant proteins were expressed using the Bac-to-Bac expression system according to the manufacturer's instructions (Invitrogen). Briefly, Sf9 cells were maintained by suspension in serum-free media prior to infection with the recombinant baculoviruses (multiplicity of infection ϳ 5). After a 48-h incubation the infected cells were harvested and resuspended in lysis buffer A and Dounce homogenized. Extracts were centrifuged 37,000 ϫ g for 30 min and proteins were then purified by immunoaffinity chromatography using the appropriate immunoaffinity resin. Proteins were eluted with 20 mM Tris buffer, pH 7.9, 100 mM KCl, 10% glycerol, 0.5 mM EDTA, and 0.2 mg/ml of the appropriate peptide competitor. Proteins were analyzed by SDS-PAGE and stained with Coomassie Brilliant Blue R-250 or analyzed by Western blotting.
GST Interaction Assays-1 g of either GST alone or GST-AD1 were incubated with GST-Sepharose and recombinant CBP or GCN5 at 4°C for 1 h in 200 l of buffer A containing 1 mg/ml bovine serum albumin. The protein complexes were then centrifuged, washed four times with lysis buffer A, and then analyzed by SDS-PAGE and Western blotting using a monoclonal ␣FLAG antibody (Sigma). For the coimmunoprecipitation assays, Sf9 cells were coinfected with the appropriate viruses as indicated in the figure legends. After 48 h cells were lysed in buffer A, centrifuged, and the resulting supernatant was incubated with 100 l of ␣HA-Sepharose immunoaffinity resin for 2 h at 4°C with rocking. The complexes were washed four times with lysis buffer A and proteins were eluted with buffer containing 0.2 mg/ml HA peptide.

Deletion Mapping Reveals Two Independent Transcriptional
Activation Domains in p/CIP-Previous studies have identified two distinct transcriptional activation domains common to all p160 family members (3,10,20,21,37,38). To confirm the location of these domains in p/CIP, cells were transiently transfected with a luciferase reporter plasmid containing Gal4 DNA binding sites upstream of a minimal promoter, and an expression vector directing the synthesis of various deletion mutants of p/CIP fused to the Gal4 DNA binding domain (Fig. 1A). Two independent domains capable of activating the reporter plasmid were identified (Fig. 1B). The fragment of p/CIP corresponding to amino acids 985-1085 (P6) exhibited the strongest activity (ϳ1389-fold above Gal4 alone) and is designated AD1. Interestingly, when this domain was extended to include the glutamine-rich region between amino acids 947 and 985 (P5) the activation of the reporter was decreased ϳ7-fold compared with the shorter P6 fragment. The mechanism by which inclusion of the poly(Q) tract represses the activation in this assay is not known although differential expression levels or improper protein-folding cannot be ruled out. It is interesting to note that shorter CAG repeat lengths in the androgen receptor are also associated with increased transcriptional activation (39). The location of the poly(Q) tract immediately N-terminal to AD1 is unique to p/CIP as the human homologues appear to be differentially spliced such that this region is found C-terminal to the AD1. The second domain, designated AD2, corresponds to Cterminal amino acids 1300 -1398, enhanced the output of the reporter ϳ26-fold above Gal4 alone. The relative strength of activation between AD1 and AD2 of p/CIP is consistent with previous reports that have identified homologous domains in related p160 family members (3,21,37,38,40). An additional activation domain corresponding to the N-terminal 93 amino acids of SRC-1 has also been reported (37). However, this region of p/CIP did not appear to activate the reporter in our assay (Fig. 1B). The ability of AD1 and AD2 to independently activate transcription when fused to a heterologous DNA-binding domain indicates that they are functionally separable and suggests that they exert effects on transcription through distinct mechanisms.
GCN5 and CBP Can Interact with the AD1 of p/CIP-The functional importance of the AD1 region of p160 proteins is highlighted by several reports that have linked the activation function of AD1 to a specific interaction with other classes of acetyltransferase proteins such as CBP/p300 and p/CAF (2,20,21,38). Based on the high degree of homology between p/CAF and GCN5, it was of interest to determine whether AD1 of p/CIP could also interact specifically with the ADA2 binding region of GCN5. The amino acids encompassing the ADA2 binding region of hGCN5 (aa 292-390) were fused in-frame with the LexA DBD and tested for interaction with AD1 of p/CIP in yeast. As indicated in Fig. 2A, GCN5 exhibits ϳ30-fold interaction with AD1 of p/CIP over vector alone. Importantly, these results demonstrate a novel interaction between GCN5 and AD1 of p/CIP and delineate the minimal interaction domain of GCN5 to lie between amino acids 292 and 390 corresponding to the ADA2 binding region. As previously demonstrated, the C terminus of CBP (aa 2058 -2170) also interacted with the region of p/CIP designated here as AD1, exhibiting approximately a 10-fold interaction over vector alone.
To determine whether the interaction with p/CIP is direct, recombinant FLAG-tagged GCN5 (fGCN5) or FLAG-tagged CBP were incubated with GST alone or GST-p/CIP-(985-1085) and the complexes were then subjected to pull-down assays using GST-Sepharose (Fig. 2B). Western blot using an anti-FLAG antibody identified a specific band corresponding to the molecular size of either CBP or GCN5 (Fig. 2C). In contrast, no such band was present when GST alone was used in the pulldowns suggesting that the interaction between the AD1 and GCN5 and CBP is most likely direct. Co-immunoprecipitations were also performed using extracts isolated from Sf9 cells infected with baculoviruses for HA-tagged p/CIP-(573-1398) (Fig. 3A) in combination with either FLAG-tagged GCN5 or FLAG-tagged CBP (Fig. 3B). FLAG-tagged CBP and GCN5 are specifically immunoprecipitated using anti-HA-Sepharose only when they are coexpressed with HA-tagged p/CIP (Fig. 3B). Furthermore, HAT activity associated with the immunopurified complexes could only be detected when either FLAGtagged CBP or GCN5 were coexpressed with p/CIP (Fig. 3C). Based on these results, it was of interest to examine the HAT activity of p/CIP more directly. Acetylation assays were performed using recombinant full-length p/CIP alone, or in the presence of limiting quantities of either CBP or GCN5 (Fig.  3D). These experiments indicate that p/CIP alone possesses little or no intrinsic acetyltransferase activity under the conditions of our assay. A slight increase in the acetylation of H4 was observed after prolonged exposure. In addition, increasing amounts of p/CIP had no effect on the intrinsic HAT activity associated with CBP or GCN5.
The RAR Requires Both p/CIP and GCN5 for Transcriptional Activation-The direct interaction between GCN5 and p/CIP supports a model by which p/CIP specifically recruits a GCN5 complex to facilitate gene activation by NRs. To test this hypothesis, several lines of investigation were initiated. First, transcription-based assays were established using yeast as a model system. Yeast represent a complimentary approach in which to evaluate mechanistic aspects of NR signaling as they lack any known NRs, as well as specific coactivators, found in mammalian cells such as the p160 proteins and the CBP/p300 family (41). In contrast, several GCN5 complexes have been identified that are well conserved between yeast and mammalian cells thus allowing us to exploit a genetic approach to assess the requirement for GCN5 in NR signaling events.
A common feature of many transcriptional activation domains is their ability to suppress growth when fused to a heterologous DNA binding domain and overexpressed in yeast (42,43). For example, the herpes simplex virus protein 16 (VP16) activation domain is a potent suppressor of growth in yeast. One possible explanation for growth suppression is the interaction of the activation domain with essential components of the yeast transcriptional machinery and/or adapter molecules, and sequestration of these components to Gal4-regulated promoters preventing the transcription of genes necessary for growth (44,45). To determine whether the activation domains of p/CIP could suppress growth, several yeast strains were transformed with a plasmid containing the constitutive ADH promoter directing the expression of either Gal4-DBD alone, or Gal4-DBD fused in-frame with AD1 or AD2 of p/CIP. Overexpression of Gal4-AD1 suppressed the growth of all yeast strains tested and was comparable with the growth suppressing properties of Gal4-VP16 when compared with Gal4 alone (Fig. 4 and data not shown). In contrast, the COOH-terminal activation FIG. 1. p/CIP contains two autonomous transcriptional activation domains. A, schematic representation of p/CIP functional domains. Depicted are the activation domains AD1 and AD2, the bHLH, the Per-Arnt-Sim A and B domains (PAS-A/PAS-B), the NRID, the glutamine-rich region (Q), and the relative locations of the LXXLL and DRALGI ␣-helical amino acid motifs (*). P1 to P9 represent fragments of p/CIP fused inframe with the Gal4-DBD. B, two regions of p/CIP activate transcription when tethered to DNA. HeLa S3 cells were cotransfected with 1 g of a luciferase reporter plasmid containing a minimal promoter and 5 upstream binding sites for Gal4, 1.5 g of a CMV driven ␤-galactosidase reporter plasmid, and 1 g of the pMGal vector (CMV) directing the expression of either the Gal4-DBD alone or the indicated p/CIP deletion mutants fused inframe with Gal4-DBD. Luciferase values were normalized to the output of the internal ␤-galactosidase control plasmid and the -folds of activation were determined relative to pMGal alone, which was arbitrarily set to 1. Values are the average of three independent experiments performed in duplicate Ϯ S.E. domain of p/CIP (Gal4-AD2) did not suppress the growth of any of the yeast strains tested. To determine whether the growth suppressing properties of GAL4-AD1 requires yGCN5, Gal4-AD1 was overexpressed in a GCN5 knockout yeast strain (⌬gcn5). Deletion of the GCN5 gene eliminated the growth suppressing properties of Gal4-AD1 providing genetic evidence that the interaction of p/CIP with yGCN5 is required for suppression of yeast growth by p/CIP. yGCN5 is the catalytic subunit of at least two endogenous yeast transcriptional activation complexes including a large multiprotein complex referred to as SAGA (SPT/ADA/GNC5 acetyltransferase) and a smaller ADA complex (46 -48). Deletion of one or more genes encoding various subunits of the SAGA results in similar yeast phenotypes such as slow growth, temperature sensitivity, and reduced activation by VP16 and yGCN4 activation domains (49 -51), suggesting that inactivation of any one of the subunits is sufficient to disrupt the activity of the entire complex. To determine whether the growth suppressing properties of Gal4-AD1 was because of the specific inactivation of the SAGA complex, Gal4-AD1 of p/CIP was also overexpressed in strains deleted for either ADA3/ NGG1 (⌬ada3) or SPT7 (⌬spt7), in parallel with the parental strains. Fig. 4 demonstrates that deletion of either the ADA3/ NGG1 or SPT7 gene reverses the ability of Gal4-AD1 of p/CIP to suppress yeast growth providing convincing evidence that the suppression of yeast growth in the wild-type strains is because of the inactivation of the SAGA complex. Importantly, the growth suppression because of overexpression of Gal4-AD1 in the wild-type yeast strains is much more severe than the slow growth that results from simply deleting the GCN5, ADA3/NGG1, or SPT7 genes (compare expression of Gal4 alone in the knockout strains with expression of Gal4-AD1 in the wild-type strains, Fig. 4). This strongly suggests that in addition to the sequestration of yGCN5, the overexpression of Gal4-AD1 may also result in the sequestration of the entire SAGA complex or possibly other components necessary for transcription of essential genes (52).
We also tested the requirement for yGCN5 directly on RAR activity in yeast. It has been demonstrated that NR exhibit very little activity in yeast cells (53,54) suggesting that yeast may lack a ligand-dependent adapter that bridges the gap between the LBD of the NR and the yeast transcription machinery. Therefore, it was of interest to determine whether p/CIP could function as such an adapter and if this ability was dependent on both the AD1 region of p/CIP and a functional yGCN5 pathway. Parental and ADA3 knockout yeast strains were cotransformed with LexA-RAR LBD and either p/CIP-(aa 591-940), which contains the NRID but lacks the region designated as AD1, or p/CIP (aa 591-1085), which contains both the NRID and an intact AD1 region. These were then tested for their ability to activate a ␤-galactosidase reporter plasmid containing LexA binding sites. As shown in Fig. 5, LexA-RAR LBD

FIG. 2. The AD1 domain of p/CIP interacts directly with GCN5 and CBP.
A, yeast two-hybrid interaction of p/CI-P(AD1) with GCN5 and CBP. The yeast strain EGY48 containing the LexA ␤-galactosidase reporter plasmid (pSH18 -34) was cotransformed with PEG202 plasmids containing an ADH promoter directing the expression of LexA DNA binding domain fusions of either hGCN5 (aa 292-390) or CBP (aa 2058 -2170) with the PJG4 -5 plasmid directing expression from the GAL1 promoter of either the B42 activation domain alone or fused in-frame with p/CIP(AD1). ␤-Galactosidase activity was determined as described under "Experimental Procedures." Values are the average -folds of activation over PJG4 -5 alone (arbitrarily set to 1) from three independent experiments performed in triplicate Ϯ S.E. B, Coomassiestained SDS-PAGE gels of the recombinant proteins utilized for interaction analysis. C, GST pull-down assays demonstrating a direct interaction of the GST AD1 domain with CBP or GCN5. Pulldowns were performed using 5 g of either purified GCN5 or CBP as described under "Experimental Procedures." The presence of either CBP or GCN5 was detected by Western blotting using an anti-FLAG antibody. Input lane represents 10% of the total input. fGCN5, FLAGtagged GCN5; fCBP, FLAG-tagged CBP.
alone, or when coexpressed with p/CIP (aa 591-940), weakly activated the reporter in the presence of RA. However, coexpression of p/CIP containing AD1 (aa 591-1085) with LexA-RAR-LBD resulted in a dramatic enhancement of reporter activity (ϳ28-fold over LexA-RAR alone) indicating that the region corresponding to AD1 of p/CIP was critical for activation of the RAR in yeast. Importantly, ligand-dependent activation of the reporter mediated through AD1 of p/CIP is significantly reduced in the ADA3 knockout strain suggesting that the activation is in part mediated through the yGCN5 pathway.
Experiments were also performed in mammalian cells using reporter-based assays. First, HeLa cells were transfected with The corresponding complexes were then purified and tested for HAT activity as described under "Experimental Procedures." D, in vitro histone acetylation activity of purified coactivator proteins. Recombinant baculovirus proteins for CBP and GCN5 were purified from Sf9 cells and tested for HAT activity in the absence or presence of full-length p/CIP as indicated.

FIG. 4. Suppression of wild-type yeast growth by overexpression of Gal4-AD1 of p/CIP is relieved by mutation of genes in the yGCN5 pathway.
Parental and yeast knockout strains (as indicated at the top of the figure) were transformed with pAS1U plasmids directing expression from the ADH promoter of either Gal4-DBD alone, or the Gal4-DBD fused in-frame with AD1 (aa 985-1085) or AD2 (aa 1300 -1398), of p/CIP as described under "Experimental Procedures." Transformed cells were allowed to grow on glucose selection media for 3 days before growth was recorded photographically. a reporter plasmid containing two consensus DR-5 hormone response elements upstream of the minimal promoter and luciferase gene along with expression vectors directing the expression of hRAR␣ and hRXR␣ as well as increasing amounts of mGCN5. The output of the reporter was determined in the presence of RA or vehicle. Cotransfection of mGCN5 resulted in a dose-dependent enhancement of RA signaling with 600 ng of plasmid exhibiting ϳ3-5-fold enhancement of ligand-dependent activation of the reporter over vector alone (Fig. 6A). In addition, we observed a further enhancement of retinoic acid signaling when p/CIP was coexpressed with mGCN5 compared with GCN5 alone, providing evidence that mGCN5 and p/CIP may function in the same pathway to facilitate retinoic acid signaling.
To further substantiate the contribution of GCN5 in RARdependent transcriptional activation we utilized an RNA interference approach. The results in Fig. 6B demonstrate that siRNA directed against GCN5 resulted in a specific downregulation in the levels of endogenous GCN5. To analyze the effects of down-regulation of GCN5 on p/CIP activity, the transcriptional activity of the Gal4-AD1 domain was cotransfected with siRNA directed against GCN5. Down-regulation of GCN5 resulted in a 53% reduction in transcriptional activation of the Gal4-dependent reporter consistent with the hypothesis that the transcriptional effects of p/CIP are, in part, mediated through recruitment of GCN5. siRNA interference experiments were also used to test the activity of the RAR␤ promoter, a RA-dependent promoter that has been shown to be regulated by p/CIP recruitment (14). Similarly, we observed a 50% reduction in luciferase reporter activity in the RA-dependent activation of the RAR␤ reporter compared with control siRNA, indicating that the regulation of this promoter is also dependent, in part, on the presence of adequate levels of GCN5. Collectively, these results support the hypothesis that GCN5, through interactions with p/CIP, functions as a coactivator for the RAR.

Mutational Analysis of AD1 Defines Molecular Determinants
Critical for the Binding of GCN5 and CBP-The observation that distinct acetyltransferase proteins interact within the same functional region of p/CIP led us to investigate the molecular determinants of these interactions to define the mode of coactivator recruitment mediated by AD1 of p/CIP. The AD1 region of p/CIP reveals three conserved motifs that have the potential to form ␣-helices (Fig. 7A). The first of these amino acid motifs (H1) contains the amino acid consensus sequence LXXLL and closely resembles the leucine charged domains, also known as NR boxes, shown to specifically mediate interaction with the NRs (2,19,55). The second conserved sequence has a slightly different amino acid consensus (LLXXL) and is found within H2. The third conserved amino sequence DRALGI (H3) is 100% conserved between the p160 family members. To examine the individual importance of these motifs in mediating specific protein-protein interactions, they were systematically mutated within the context of AD1 of p/CIP and tested in parallel with wild-type AD1, for their ability to bind GCN5 or CBP by yeast two-hybrid assay. As shown in Fig. 7B, mutation of H1 alone (AD1.1) had no significant effect on the binding of GCN5. However, mutation of the H2 motif (AD1.2) resulted in a 50% reduction in interaction with GCN5 compared with wild-type (AD1). Deletion of the H3 motif (AD1.3) resulted in a 75% decrease in interaction with GCN5. Similar results were obtained when the homologous region in p/CAF was used (data not shown). These results demonstrate that full interaction of GCN5 and p/CAF to AD1 of p/CIP is dependent predominantly on the H2 and H3 motifs. The nearly identical binding preferences for GCN5 and p/CAF are expected for highly homologous proteins and demonstrate that they are likely recruited to p/CIP in a mutually exclusive manner. Surprisingly, the binding pattern of CBP is different from that of GCN5 and p/CAF ( Fig. 7B and data not shown). Deletion of H2 resulted in a 50% decrease in binding, whereas mutation of either H1 or the H3 motif alone had no significant effect on the ability of CBP to interact with p/CIP AD1. In fact, although not statistically significant, mutation of either the H1 (AD1.1) or H3 (AD1.3) motifs appeared to increase the binding of CBP in the yeast assay. These results indicate that members from two distinct families of acetyltransferase proteins, CBP and GCN5, have overlapping but distinct binding preferences within the context of AD1 of p/CIP. The identification of specific amino acid motifs important for the binding of GCN5 and CBP by the AD1 in p/CIP suggests that the transcriptional activity of AD1 in mammalian cells is mediated through the specific recruitment of these proteins. AD1 and mutant versions of AD1 were fused to the Gal4 DNA binding domain and transiently transfected into cells along with a reporter plasmid containing Gal4 DNA binding sites upstream of a minimal promoter and a luciferase reporter gene. As seen in Fig. 8B, mutation of the H1 (AD1.1), H2 (AD1.2), and DRALGI (AD1.3) motifs significantly reduced the transcriptional activation of AD1. The observation that both the leucine charged domains and the DRALGI motifs are required for the full transcriptional activity of AD1 suggests that the activity of AD1 in mammalian cells is due, in part, to the recruitment of GCN5, p/CAF, and/or CBP. Furthermore, deletion of the DRALGI motif alone, which has no effect on binding of CBP to the AD1, significantly diminished transcriptional activation suggesting that the loss of activity of this mutant is because of a loss of binding of either GCN5 or p/CAF. Interestingly, mutation of the H1 motif (AD1.1), which had no significant effect on the binding of GCN5 or CBP as assessed by yeast two-hybrid assay (Fig. 7B), demonstrated a significantly diminished transcriptional activity suggesting that other factors may

DISCUSSION
This study has mapped the minimal regions of p/CIP, between amino acids 985-1085 and 1300 -1398, as the location of two independent transcriptional activation domains (AD1 and AD2, respectively). The relative location of these domains is consistent with homologous regions found in both SRC-1 and TIF2/GRIP1 (3,10,20,21,37,38). Previous studies have demonstrated that deletion or mutation of the AD1 region significantly reduces the ability of p160 proteins to enhance transcriptional activation by NRs in vitro and in vivo (2,3,21,37). In addition, several studies have suggested that the function of AD1 is primarily mediated by its ability to bind CBP/p300 (2, 3, 20 -23, 38, 40, 56). A particularly novel and significant finding in this study is the demonstration that the transcriptional effects of AD1 can be mediated by GCN5 as well as CBP. This hypothesis is substantiated by several interrelated findings. First, GCN5 can interact with the AD1 of p/CIP in vitro and in vivo and this interaction coincides with the recruitment of HAT activity by p/CIP. Second, the potent suppression of yeast growth by the Gal4DBD-AD1 fusion protein is dependent on yGCN5 as well as other components of the SAGA complex. Third, activation of the RAR in yeast requires the AD1 domain of p/CIP in addition to a functional SAGA complex. Fourth, transient overexpression assays in mammalian cells suggest that mGCN5 can cooperate with p/CIP for full transcriptional activity of the RAR. Finally, using siRNA to down-regulate endogenous GCN5, we have shown that the transcriptional activity of the AD1 domain and RA-dependent stimulation of the RAR␤ promoter are impaired. Based on the relatively weak activation capacity of the RAR in yeast in the absence of p/CIP it is likely that the enhancement of RAR activity by GCN5 is indirect via its interaction with the AD1 domain of p/CIP. GCN5 and PCAF contain a highly conserved HAT domain and this activity has been shown to be of critical importance in transcriptional activation by these proteins (32,(57)(58)(59)(60)(61)(62). There is ample evidence suggesting that human GCN5 and p/CAF FIG. 6. GCN5 is required for RAR signaling in mammalian cells. A, HeLa S3 cells were transiently cotransfected with a luciferase reporter plasmid containing DR-5 retinoic acid response elements upstream of a minimal promoter as described under "Experimental Procedures" and increasing amounts of a CMV vector driving the expression of mGCN5 and/or p/CIP. Cells were treated for 24 h with either vehicle or 10 Ϫ6 M all-trans-RA and luciferase and ␤-galactosidase activity was determined. All values are relative to that of the basal condition (RAR/RXR alone) in the absence of ligand, which was arbitrarily set to 1. B, transfection of siRNA directed against GCN5 inhibits endogenous GCN5 expression. siRNA was transfected as described under "Experimental Procedures." After 48 h, total proteins were extracted and Western blotting was performed to monitor GCN5 expression. The right panel is the corresponding densitometric scan of the Western blot of cells treated with the indicated concentrations of siRNA. C, AD1 activity is compromised in siGCN5-treated cells. C127 cells were cotransfected with 0.5 g of a luciferase reporter plasmid containing a minimal promoter and 5 upstream binding sites for Gal4, a plasmid directing the expression of either the Gal4-DBD alone or gal4-p/CIP-(985-1085), and siRNA (0.5 nmol) directed against GCN5 or green fluorescent protein as a control. Luciferase values were normalized to the output of the internal ␤-galactosidase control plasmid and the -folds of activation were determined relative to pMGal alone, which was arbitrarily set to 1. Values are the average of three independent experiments performed in duplicate (ϮS.E.). No effect of siRNA on GAL4 DBD activity alone was detected (data not shown). D, RAR activity is compromised in siGCN5-treated cells. C127 cells were transiently cotransfected with a luciferase reporter plasmid downstream of the RAR␤ core promoter, along with siGCN5 RNA or siGFP RNA as a control as described under "Experimental Procedures." After 24 h, cells were treated with either vehicle or 10 Ϫ6 M all-trans-RA and luciferase and ␤-galactosidase activity were determined. All values are relative to that of the basal condition (RAR/RXR alone) in the absence of ligand, which was arbitrarily set to 1. Values are the average of three independent experiments performed in duplicate Ϯ S.E. complexes purified from mammalian cells are structural and functional counterparts of the SAGA complex of yeast (63)(64)(65)(66). In addition to ADA2, the human p/CAF and GCN5 complexes also contain human homologues to ADA3, SPT3, and various TAF proteins (65). Recent studies have also identified TFTC (TATA-binding protein-free TAFII-containing complex) and FIG. 7. Identification of binding site preferences of GCN5 and CBP to AD1 of p/CIP. A, amino acid sequence of the core AD1 and AD1 mutants (AD1. 1-1.4) and the location of the ␣-helices designated H1, H2, and H3 within the core AD1 domain of p/CIP (aa 985-1085). B, yeast two-hybrid interaction assay of p/CIP-AD1 with GCN5 and CBP. The yeast strain EGY48 containing the LexA ␤-galactosidase reporter plasmid (pSH18-34) was cotransformed with a PEG202 plasmid containing an ADH promoter directing expression of LexA-DBD fusions of either hGCN5 (aa 292-390) or CBP (aa 2058-2170), along with the PJG4-5 plasmid directing expression from the GAL1 promoter of either the B42 activation domain alone or fused inframe with wild-type or mutant fragments (AD1.1-AD1.4) of AD1 of p/CIP. Liquid cultures were induced and ␤-galactosidase activity normalized to cell density was determined as described under "Experimental Procedures." The relative -folds of interaction over PJG4-5 alone were determined for each interaction with the -fold of the wildtype interaction arbitrarily set to 100%. Values are the average of three independent experiments performed in triplicate Ϯ S.E. *, p Ͼ 0.05.
FIG. 8. Intact protein-protein interaction motifs are required for full activation by p/CIP AD1. A, amino acid sequence of the core AD1 and AD1 mutants (AD1.1-1.4) and the location of the ␣-helices designated H1, H2, and H3 within the core AD1 domain of p/CIP (aa 985-1085). B, transcriptional activation of Gal4-p/CIP AD1 and the corresponding mutants. HeLa S3 cells were transiently transfected with 1 g of a luciferase reporter plasmid containing 5 upstream binding sites for Gal4 upstream of a minimal reporter, 1.5 g of a ␤-galactosidase reporter plasmid, and 1 g of the pMGal vector directing the expression of either the Gal4-DBD alone or various deletion mutants of p/CIP fused in-frame with the Gal4 DBD. The output of the reporters was determined as described under "Experimental Procedures." Luciferase values were normalized to the output of the internal ␤-galactosidase control plasmid and the relative -folds of activation over Gal alone were determined for each condition with the -fold of the wild-type activation arbitrarily set to 100%. Values are the average of three independent experiments performed in duplicate Ϯ S.E. *, p Ͼ 0.05. STAGA (SPT3-TAFII31-GCN5-acetyltransferase) (66 -68) as GCN5-containing complexes purified from human cell lines that have been shown to compliment NR function. Interestingly, each of these complexes also contain several subunits found in the TFIID complex including histone-like TAFs (47,65,69). It is interesting to note that the ADA2 binding region of GCN5 is required for interaction with p/CIP. The ADA2 subunit of the SAGA complex (46) is a target for various transcriptional activators. Mutations within ADA2 have been shown to affect transcriptional activation by VP16 and GCN4, but not by HAP4 (50,52). This suggests a mechanism whereby distinct activation domains can interact with one or more components of the GCN5 complex. A distinguishing feature between two of the GCN5-containing complexes isolated from humans is the presence and absence of the ADA2 subunit (66,70). Thus, it is conceivable that the ADA2 subunit confers differential responsiveness of these two complexes to ADA2-dependent and ADA2-independent transcriptional activators. For example, the absence of the ADA2 subunit would presumably leave the complex unresponsive to activation by VP16 or GCN4, but may allow other activators such p/CIP to bind to the ADA2 binding site in GCN5.
This study has also identified the molecular determinants of AD1 that mediate specific interactions with GCN5 and CBP. Our data indicates that H2, and not H1, of p/CIP is a critical determinant for interaction with both GCN5 and CBP. This is consistent with previous reports demonstrating that the H2 motif of TIF2 and SRC-1 are required for interaction with CBP (20,21,71). Furthermore, the recent NMR structural determination of the ACTR/CBP interface indicates that H1 is not a component of the hydrophobic core that can participate in direct interactions with CBP (72). Our data also suggests that both H2 and the H3 are required for the full binding to GCN5. This is consistent with a recent study demonstrating that mutations within H2 compromise the ability of TIF2 to activate NR signaling in yeast (21). Because yeast do not contain a CBP or p300 homologue, other proteins with a similar recognition surface to CBP must exist in yeast. Based on the evidence presented in the present study, this is most likely yGCN5. Surprisingly, deletion of the H3 motif, which is conserved among all members of the p160 family, did not significantly affect interactions with CBP in our yeast assay. However, mutations within both H2 and H3 decreased the interaction with CBP to a greater degree than mutations within H2 alone suggesting that H3 may provide a level of stability for interaction with CBP/p300. In contrast, the interaction observed between the AD1 and GCN5 is clearly dependent on a functional H3 motif. Collectively, these findings suggest that the H3 motif is a differential binding determinant for GCN5/pCAF and CBP/ p300 and consequently may dictate the specificity and order of assembly of components of the p160 coactivator-containing complexes. Based on the relatively high degree of overlap that exists for the interactions with GCN5 and CBP we conclude that the recruitment of GCN5, p/CAF, or CBP by p/CIP is most likely mutually exclusive.
To establish a functional correlation between the interaction of the AD1 with GCN5 and CBP, we also examined the ability of the AD1 mutants to activate transcription in mammalian cells. We observed that mutations in either H2 or H3 significantly impaired transcriptional activity when tethered to Gal4. Interestingly, mutation of H1, which has negligible effects on the binding of CBP, p/CAF (not shown), or GCN5, significantly diminished transcriptional activation suggesting that other factors may also be involved in potentiating the transcriptional activity of AD1. Similar mutations of the H2 and H3 motif of SRC-1 and TIF-2 have also been found to significantly abolish activation by the TR and RAR (20).
GCN5 and PCAF exhibit remarkable similarities in structure as well as in vitro function. Unlike CBP/p300, they do not appear to exhibit dosage effects in heterozygous null mice. Mice heterozygous for either GCN5 or P/CAF appear to be free of defects (73). Furthermore, PCAF null mice appear to be normal, whereas homozygous deletions of the GCN5 gene in mice is lethal and mice die between 10 and 11 days post coitum. This suggests that the specific patterns of histone modifications created by particular HATs may be critical for generating patterns of tissue-specific gene expression for normal growth and development. In this regard, the p/CAF and GCN5 mRNAs have a different tissue distribution in mice suggesting that they may not have overlapping functions (73,74). Targeting of distinct HATs has been shown to be important for specific transcriptional events and differential HAT recruitment may be important for the regulation of different sets of genes. For example, microinjection studies have demonstrated that the HAT activity of p/CAF is required for activation by RAR, whereas the HAT activity of CBP was required for CREB function (32). In addition, p300 and p/CAF have been shown to have different roles in muscle differentiation (75,76). Nonhistone proteins in the vicinity of a promoter can also serve as substrates for acetyltransferase enzymes. For example, CBP/ p300 has been shown to acetylate p53 (77), ACTR (14), TFIIE and TFIIF (78).
In summary, we have demonstrated that members from two distinct families of HATs can specifically interact with AD1 of p/CIP. Furthermore, the observation that the AD2 of p160 proteins interacts with the methyltransferase CARM1 (11,30) indicates that p/CIP functions primarily as a ligand-dependent adaptor protein to recruit proteins that possess distinct types of enzymatic activity to facilitate transcription.
Recent studies have indicated that overexpression and amplification of the human of p/CIP/SRC-3 is found in a significant percentage of breast and ovarian cancers (4) and that the relative levels of p/CIP may be rate-limiting for hormone dependent growth (79). Consequently, the sequestration of critical coregulatory proteins, such as GCN5 and CBP, to NR-dependent genes may provide a selective advantage for tumor growth.