The Transcriptional Corepressor, PELP1, Recruits HDAC2 and Masks Histones Using Two Separate Domains*

PELP1 (proline-, glutamic acid-, and leucine-rich protein 1) has been recognized as a coactivator of estrogen receptor (ER)-recruiting p300/CREB-binding protein histone acetyltransferase to the target chromosome. The present study shows that PELP1 does indeed coactivate ER-mediated transcription but also serves as a corepressor of other nuclear hormone receptors (NR)- and non-NR sequence-specific transcription factors tested, including GR, Nur77, AP1, NF-κB, and TCF/SRF. PELP1 expression also retarded the proliferation of mouse fibroblast cell lines in which there was no detectable ER. This was due, at least in part, to the suppressed activation of serum-response genes such as c-fos that in turn resulted from the blocked histone hyperacetylation of nucleosomes containing the c-fos promoter region. The N-terminal leucine-abundant region of PELP1 was observed to interact with HDAC2 and exhibited repressive activity when tethered to the chromatin. In addition, the C-terminal glutamic acid-abundant region bound to the hypoacetylated histones H3 and H4 and prevented them from becoming substrates of histone acetyltransferase. Thus PELP1 promotes and maintains the hypoacetylated state of histones at the target genomic site, and ER binding reverses its role to hyperacetylate histones through an as yet unidentified mechanism.

Eukaryotic chromatin is a highly compact DNA-protein complex composed of an array of nucleosomes. The histone octamer, containing two copies each of histones H2A, H2B, H3, and H4, provides the protein core of a nucleosome into which the DNA is packaged (1). The modulation of chromatin compactness, i.e. chromatin remodeling, is important for rendering transcription factors and the basic transcription machinery accessible to their target gene promoters. Chromatin remodeling can be achieved by at least two enzyme complexes, the histone-modifying enzyme complexes and the ATP-dependent remodeling complexes (2)(3)(4)(5)(6). Hyperacetylation of the lysine residues contained in histones is believed to be a mechanism for the destabilization of nucleosomes and the facilitation of transcription factor binding (7,8).
The steady state acetylation level of the core histones is maintained by a balance between the opposing activities of HATs 1 and HDACs (8). Many transcriptional coactivators, in-cluding GCN5/PCAF, CBP/p300, and NcoA, exhibit intrinsic HAT activity (5), whereas most transcriptional corepressors studied thus far recruit specific HDACs to the chromatin of target genes (5,9). For example, Sin3, Mi-2/NuRD, and retinoblastoma protein (pRb) associate with HDAC1 and -2, CtBP with class II HDACs, and NcoR and SMRT with HDAC3 (10 -16). The presence of these divergent complexes of corepressors and HDACs has been surmised to support the specific transcriptional regulation of target genes.
Some transcriptional coregulators for NR-mediated transcription, such as ZAC1, NSD1, and RIP14, have been demonstrated to possess dual functions, acting as both coactivators and corepressors (17)(18)(19)(20). Recently, the nuclear protein PELP1 (proline-, glutamic acid-, and leucine-rich protein 1) has been proposed to be another dual-function coregulator for NR-mediated transcription; i.e. a coactivator for ER and a corepressor for glucocorticoid receptor (GR) (21). PELP1 is ubiquitously expressed in most tissues but more abundantly in neuronal and endocrinal organs such as the brain and the mammary gland. The discovery of up-regulated expression of PELP1 messages in breast cancer tissues underlined the significance of the ER-dependent coactivator function of PELP1 in the tumorigenesis of mammary gland cells.
PELP1 possesses 10 putative NR-interacting LXXLL motifs in its N-terminal half, followed by a proline-abundant region and an extremely acidic region containing an unusually long stretch of about 70 glutamic acids at the other side of the protein (21). PELP1 does not appear to have intrinsic HAT activity, but it does appear to exert a coactivational activity for ER via binding to both ER␣ and p300 through its N-terminal region, which contains 10 LXXLL motifs. PELP1 also plays a permissive role in 17␤-estradiol (E2)-mediated cell cycle progression via interaction with retinoblastoma protein through its N-terminal 330 amino acids (22). In addition to the transcriptional coregulator functions for PELP1, it has also been suggested that it plays a nongenomic role, bridging ER and Src tyrosine kinase through its LXXLL and PXXP motifs, respectively (23).
In the present study, we report that PELP1 definitely functions as a corepressor of sequence-specific transcription factors, including some NR and non-NR transcription factors such as AP1, TCF/SRF, and NF-B. Successful transcriptional repression requires both the N-terminal leucine-abundant region (LAR) and the C-terminal glutamic acid-abundant regions (GAR), which are associated with HDAC2 and hypoacetylated core histones, respectively. PELP1 can be considered to belong to a novel class of corepressor, in that it suppresses histone acetylation via active deacetylation, using associated HDAC and masking core histones from HAT-mediated acetylation.

EXPERIMENTAL PROCEDURES
cDNA and Plasmids-The full-length cDNA of PELP1 was obtained by rapid amplification of cDNA ends from the HeLa cell cDNA library; its sequence was confirmed by automatic DNA sequencing, and it was cloned into the EcoRI and XhoI restriction sites of pcDNA3.1 A/His.Myc vector (Invitrogen). Deletion mutants of PELP1 were generated by PCR and subcloned into pcDNA3.1. To express the GAR alone in the nucleus, the portion of PELP1 cDNA encoding for the GAR was subcloned into the XhoI site of pEF/nuc vector (Invitrogen) and modified in order to add six copies of histidine. The INHAT domain of TAF-1␤ (amino acids 195-277) was amplified by PCR and inserted into the XhoI site at the end of PAR in PELP1⌬GAR. All GAL4 fusion constructs were subcloned into the SalI site of pGH250 that contained GAL4DB. To generate GST-GAR fusion proteins, the cDNA encoding the GAR was subcloned into the EcoRI and XhoI sites of pGEX 4T-1 (Amersham Biosciences). The cDNAs encoding constitutively active calmodulin-dependent kinase IV (CaMK IVc), Nur77, p300, SRC-1, and ER␣ were obtained from Dr. J. W. Lee (Baylor University) and SRF was from Dr. K. C. Chung (Yonsei University, Korea). Reporter plasmids were obtained from Stratagene with the followings exceptions: pERE-Luc and pNurRE-Luc were gifts from Dr. J. W. Lee; pMMTV-Luc was from Dr. H. Choi (Chonnam University, Korea); and pGH250 and pUAS(G5)-TK-Luc were from Dr. J. H. Ahn (Sungkyunkwan University, Korea). pTK-R-Luc constitutively expressing Renilla luciferase under the thymidine kinase promoter was obtained from Promega Corp.
Cell Culture and Transfection-NIH3T3 and C3H10T1/2 cells (American Type Cell Collection, ATCC) were maintained in Dulbecco's modified Eagle's medium supplemented with 10% calf serum and antibiotics. COS-7 cells were maintained in Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum. To establish NIH3T3 and C3H10T1/2 cell lines stably expressing PELP1 and PELP1 lacking the GAR (PELP1⌬GAR), cells were transfected with pcDNA3.1-PELP1 and PELP1 ⌬GAR by using LipofectAMINE 2000 reagent (Invitrogen), diluted, and grown in complete media in the presence of 500 g/ml neomycin.
Electrophoresis Mobility Shift Assay-Confluent NIH3T3 cells transfected with expression plasmids encoding PELP1 and SRF were starved for 24 h before stimulation with 10% serum for 1 h, and nuclear extracts of the cells were prepared as described (24). Desalted nuclear extracts (8 g of total protein) were incubated in the presence of 1 ng of radiolabeled SRE oligonucleotide probe (Santa Cruz Biotechnology) in binding buffer (10 mM Tris-HCl, pH 7.5, 1 mM dithiothreitol, 1 mM EDTA, 25% glycerol, and 1 g of poly(dI-dC)). For the supershift assay, the nuclear extracts were incubated with 200 ng of appropriate antibodies for 30 min at room temperature and then incubated for 15 min with the probe. Reaction mixtures were electrophoresed on 4% acrylamide gel, dried, and visualized by autoradiography.
Reporter Assays-NIH3T3 cells were plated in triplicate at a density of 2 ϫ 10 4 per well (1.8 cm 2 ) and transfected with either 100 or 200 ng of reporter plasmids, along with the indicated amount of expression plasmids, using FuGENE 6 reagent (Roche Applied Bioscience). The total amount of plasmids was kept constant by adjustments with empty vector, and 1 ng of pTK-R-Luc was added to the DNA mixtures to monitor the efficiency of the transfections. After stimulation with proper ligands, cells were extracted with passive lysis buffer (Promega Corp.), and firefly and Renilla luciferase activities were assessed using a Dual-Luciferase® reporter assay kit (Promega Corp.). For the pERE-Luc and pMMTV-Luc assays, cells were cultured in phenol red-free RPMI 1640 medium containing 10% fetal bovine serum, stripped with dextran-charcoal for at least 24 h, and then transfected. In experiments using HDAC inhibitors, cells were pretreated with the inhibitors for 5 h before stimulation.
Immunoprecipitation and Immunoblotting-Cells were lysed by two cycles of 15-s sonication in immunoprecipitation buffer (50 mM Tris-HCl, pH 8.0, 150 mM NaCl, 0.5% Nonidet P-40, and 1 mM EDTA) containing protease inhibitor mixture and phosphatase inhibitors. The cell homogenates were collected by centrifugation at 10,000 ϫ g for 10 min and incubated with specific antibodies for 16 h at 4°C. The immune complexes were extensively washed four times with immunoprecipitation buffer, separated by SDS-PAGE, and transferred to nitrocellulose membranes. The membranes were immunoblotted with appropriate antibodies and visualized by enhanced chemiluminescence (Amersham Biosciences).
In Vitro Binding Assays-Purified GST or GST-GAR proteins (1 g) were mixed with 2 g each of histone H2A/H2B complex, H3/H4 complex, or core histone complex in a binding buffer containing 50 mM Tris-HCl, pH 7.9, 150 mM NaCl, 1 mM MgCl 2 , 0.5 mM phenylmethylsulfonyl fluoride, and 0.1% Nonidet P-40, and were incubated for 2 h at 4°C. GST fusion proteins and bound histones were pulled down using glutathione beads, washed extensively with binding buffer containing 150or 300 mM NaCl, and eluted by boiling in the SDS sample buffer. Proteins were separated by SDS-PAGE (13.5% acrylamide) and visualized by silver staining. For peptide competition or peptide-PELP1 interaction experiments, H3 peptides (amino acids 1-20; ARTKQTARK-STGFKAPRKQL) unacetylated or acetylated on the K9 and/or K14 were purchased (Upstate Biotechnology, Inc.) and used. In another set of experiments, PELP1 and PELP1⌬GAR were translated in vitro in the presence of [ 35 S]methionine using TNT TM -T7 coupled reticulocyte lysate (Promega Corp.) and incubated with 5 g of either H3 or bovine serum albumin for 30 min on ice. Reaction mixtures were then diluted with phosphate-buffered saline and immunoprecipitated using anti-Pan histone antibody. Precipitated proteins were separated by SDS-PAGE and visualized by autoradiography.
RNA Isolation and Semi-quantitative PCR-Total RNA was isolated from NIH3T3 cells using the RNeasy® kit (Qiagen). Semi-quantitative RT-PCR was carried out with 10 ng of RNA using a one-step kit (Qiagen) for an appropriate number of cycles. The following primers were used: a pair of 5Ј-cgcagagcatcggcagaagg-3Ј and 5Ј-tgagaaggggcagggtgaagg-3Ј for c-fos; a pair of 5Ј-gcctgcacagtgagcctccg-3Ј and 5Јagttggcacccactgttaac-3Ј for c-jun; and a pair of 5Ј-gtgaacggatttggccgtattg-3Ј and 5Ј-agtcttctgggtggcagtgat-3Ј for gapdh. Equal volumes of PCRs were resolved on 2% agarose gel and visualized by ethidium bromide staining.
ChIP Assay-The ChIP assay was carried out according to a protocol optimized for mammalian cells (25). Briefly, NIH3T3 cells (1 ϫ 10 6 cells per 10-cm dish), either serum-starved or stimulated, were fixed with 270 l of 37% formaldehyde for 10 min and extensively washed with ice-cold phosphate-buffered saline containing 1 mM phenylmethylsulfonyl fluoride. Cells were extracted in SDS lysis buffer (1% SDS, 10 mM EDTA, and 50 mM Tris-HCl, pH 8.1) and sonicated three times by 9-s pulses under the 30% amplitude setting of a Vibra Cell TM (Sonics and Materials Inc.). The extracts were diluted 10-fold in ChIP dilution buffer (0.01% SDS, 1% Triton X-100, 1.2 mM EDTA, 16.7 mM Tris-HCl, pH 8.1, and 167 mM NaCl) and immunoprecipitated with anti-PELP1, anti-SRF, anti-HDAC2, anti-acetyl H3, or H4 antibodies, extensively washed, eluted using 1% SDS and 0.1 M sodium bicarbonate, and incubated at 65°C for 4 h to reverse the cross-links. DNA fragments in the immune complex were recovered by phenol extraction and ethanol precipitation and amplified by quantitative PCR, using two sets of primers specific for the c-fos gene as follows: a pair of 5Ј-cgtcagcaggtttccacggc-3Ј and 5Ј-gcgcctttatagaagcgctg-3Ј for amplification of the immediate downstream of SRE, and another pair of 5Ј-gtgacacctgagagctggta-3Ј and 5Ј-gaccacctcgacaatgcatg-3Ј for amplification of the 3Ј-end of c-fos gene as a negative control. For the re-ChIP assay, immune complex, which had been eluted in buffer containing 1% SDS and 0.1 M sodium bicarbonate, was diluted in ChIP dilution buffer and then proceeded according to the same ChIP assay protocol as above using the proper antibody. In an experiment for the evaluation of in vivo binding, sonicated extracts of HeLa, MCF-7, or COS-7 cells transfected with plasmid encoding either wild-type or mutant PELP1 were immunoprecipitated with rabbit anti-PELP1 antibody according to the ChIP protocol. The immune complexes were separated by SDS-PAGE, immunoblotted with anti-histone H3 antibody, and visualized by Lumigen TM TMA-6 reagents (Amersham Biosciences).
In Vitro HAT Assay-CBP or p300 immunoprecipitates from 400 g of HeLa nuclear extracts were incubated with 1 l (50 Ci/ml) of [ 14 C]acetyl coenzyme A (Amersham Biosciences) and indicated amounts of histone H3 and H4 in a 50-l reaction buffer (50 mM Tris-HCl, pH 8.0, 10% glycerol, 1 mM dithiothreitol, and 10 mM sodium butyrate). Purified GST or GST-PELP1GAR was incubated with histones for 30 min before the commencement of the in vitro HAT reaction. The reaction was stopped by the addition of SDS sample buffer, and proteins were separated on SDS-PAGE (13% acrylamide) and visualized by fluorography.

PELP1 Is a Coactivator of ER but Functions as a Corepressor
of Other Sequence-specific Transcription Factors-PELP1 has been characterized recently as a coactivator of ER␣ (21). The coregulatory role of PELP1 on transcriptions mediated by other sequence-specific transcription factors including GR, Nur77, NF-B, AP1, and TCF/SRF was analyzed in NIH3T3 cells (Fig.  1). NIH3T3 cells express endogenous PELP1 at a low level and exhibit undetectable ER levels (see below). Expression of both ER and increasing amounts of PELP1 in NIH3T3 cells enhanced the E2-induced transactivation of ER-responsive element (ERE) in a dose-dependent manner, as expected (Fig. 1A). On the other hand, expression of PELP1 efficiently repressed dexamethasone-induced transactivation of the mouse mammary tumor virus (MMTV) promoter (Fig. 1B). In addition, the CaMK IVc-induced activation of Nur77-mediated transcription was also repressed by the expression of PELP1 (Fig. 1C). Furthermore, PELP1 successfully repressed the transactivation of reporter genes for non-nuclear hormone receptor transcription factors, including NF-B, AP1, and TCF/SRF, in a dose-dependent manner (Fig. 1, D-F). Most interestingly, the basal transcription of the MMTV-Luc, NurRE-Luc, NF-B-Luc, AP1-Luc, and SRE-Luc reporter genes were also repressed by PELP1, in the absence of proper stimuli (Fig. 1F). However, constitutive expression of the Renilla luciferase gene under the control of the thymidine kinase promoter was affected in no way whatsoever in both the presence and the absence of serum stimuli (Fig. 1G). The dose-dependent expression of PELP1 in the reporter assays was confirmed by immunoblot analysis of all extracts with anti-PELP1 antibody ( Fig. 1H and data not shown). These results suggest that PELP1 behaves as a dualfunction transcriptional coregulator, i.e. as a corepressor of several sequence-specific transcription factors, but as a coactivator of ER.
ER␣ Inhibits the PELP1-mediated Transcriptional Corepression of SRE-The ER signaling pathway has been implicated in the progression of breast cancer tumorigenesis. As PELP1 functions as an ER coactivator and is expressed at a 3-5 times higher rate in breast tumors than in normal tissues, the role of PELP1 overexpression in hormone-dependent tumor progression also warrants careful consideration (21). For this to be the case, breast cancer cells highly expressing PELP1 should be able to elude or obviate the corepressive activities of PELP1, in order to induce the transcription of SRE that is important in signaling for proliferation. Most interestingly, in the mammary gland cancer cell line, MCF-7, which expressed both endogenous ER and PELP1, serum-induced SRE transcription remained unaffected by the overexpression of PELP1 ( Fig. 2A). On the other hand, in a similar fashion to NIH3T3 cells (Fig.  1F), overexpression of PELP1 successfully repressed SRE transcription in an ER-negative cervical epithelial cell line, HeLa, although this repression required a somewhat higher PELP1 level than in the NIH3T3 cells (Fig. 2B). These results indicate that ER may interfere with the corepressive role of PELP1 with regard to SRE transcription. Indeed, the PELP1-mediated transcriptional repression of SRE in the NIH3T3 cells was completely inhibited by ER␣ coexpression (Fig. 2C). Taken together, it would be that the basic role of PELP1 is to repress the transcriptional activation of several sequence-specific ciselements, including SRE, and that the binding of ER to PELP1 may modify PELP1 to obtain a coactivator function, albeit by an as yet uncharacterized mechanism. In our study, the biochemical mechanism of PELP1 underlying its signal-dependent transcriptional corepression was further studied, using SRE as the target gene.
Both the LAR and the GAR Are Important for the Corepressor Function of PELP1 in a Full-length Context-To demonstrate that PELP1 is in the SRF complex on SRE, we performed gel mobility shift assays, using a high affinity SRE probe, nuclear extract of NIH3T3 cells ectopically expressing PELP1 and SRF, and antibodies against SRF and PELP1. The anti-PELP1 antibody used in the experiment recognized only PELP1 in the MCF-7 cell extract (Fig. 3A, lane 1), which was confirmed by blocking recognition with the epitope-specific PELP1 peptide (amino acids 1052-1138) (Fig. 3A, lane 2). Although PELP1 does not bind to DNA by itself (21), supershift of the DNAprotein complex band by anti-PELP1 or anti-SRF antibody suggests that PELP1 is able to form a complex with SRF on the SRE probe (Fig. 3B, lanes 4 and 2, respectively). In addition, the recruitment of PELP1 to the endogenous SRE on the c-fos gene in PELP1-expressing NIH3T3 cells was further verified by a ChIP assay using anti-PELP1 antibody (Fig. 3C). Two sets of primers were used to amplify the immediate downstream region of an SRE in the 5Ј-flanking region (Ϫ290 to Ϫ21 nt) of the c-fos gene, and also a region in its 3Ј-flanking region (ϩ2830 to ϩ3169 nt) as a control (Fig. 3C). The region encompassing Ϫ290 to Ϫ21 nt, but not ϩ2830 to ϩ3169 nt, was prominently amplified from both anti-PELP1 and anti-SRF immunoprecipitates, supporting the conclusion that PELP1 specifically targeted to the promoter region of the c-fos gene (Fig. 3C). In order to confirm further the presence of PELP1 in the SRF complex, PELP1 or SRF immunoprecipitates were subjected to re-ChIP assays, using anti-SRF or anti-PELP1 antibodies, respectively. As shown in Fig. 3D, SRF was detected in PELP1 immunoprecipitates, and PELP1 was detected in SRF immunoprecipitates, indicating that PELP1 does indeed form a complex with SRF on the SRE. Thus, it is expected that the corepressive activity of PELP1 with regard to TCF/SRF-mediated transcription can be achieved in a manner specific to the bound transcription factor.
The primary structure of PELP1 can be divided into three structurally unique regions as follows: the N-terminal LAR, the internal PAR, and the C-terminal GAR. To access the inherent structure-function relationships, various His 6 epitope-tagged PELP1 mutants, containing a deletion of one or two of these structurally unique regions, were prepared, and their nuclear expression characteristics were confirmed (Fig. 4A). For nuclear expression of the GAR alone, the nuclear localizing signal of SV40 was tagged to its C terminus. The results of luciferase reporter assays showed that expression of full-length PELP1, but not the expression of any of the deletion mutants, effectively repressed transactivation of SRE (Fig. 4B). Identical results were observed for all other cis-elements examined, including the reporter genes for GR, Nur77, NF-B, and AP1 (data not shown). Expression of the GAR alone activated instead both the basal and signal-dependent transactivations of SRE, probably because of its extremely acidic nature. Nevertheless, these results indicate that the proper corepressor function of PELP1 can be achieved only in a full-length context. However, we found that overexpression of the LAR and the GAR almost completely inhibited PELP1-mediated transcriptional repression, in a dose-dependent manner, whereas the effect of the PAR was found to be minimal at best (Fig. 4C). This indicates that overexpression of the LAR or the GAR exerts a dominant-negative effect on PELP1-mediated repression and suggests that the LAR and the GAR contain major functional motifs of the corepressive activity of PELP1, although these separate domains appear to be unable to function as efficient corepressors, in and of themselves.
The LAR Has an Intrinsic Repressive Activity When Tethered to DNA-Various deletion mutants of PELP1 were fused to GAL4 DNA binding (DB) domain and transfected into NIH3T3 cells, along with a reporter plasmid containing five copies of the GAL4 binding element and the luciferase gene, under control of the thymidine kinase promoter (G5-tk-luc), and a luciferase assay was carried out (Fig. 4D). The results demonstrated that GAL4-PELP1 repressed the transcription of G5-tk-luc by 70%, as expected. Expression of GAL4-LAR and GAL4-⌬GAR repressed the transcription more efficiently, by up to 90%. On the other hand, the fusion of GAR, PAR, or both to GAL4DB repressed the transcription less efficiently, in the range of 20 -40%. The common feature, shared by the three effective constructs, GAL4-LAR, GAL4-⌬GAR, and GAL4-PELP1, is the presence of the LAR. Considering the inefficient corepression by the LAR alone (Fig. 4B), it seems likely that the PAR and/or the GAR contain domains important for delivering the LAR to the target chromatin site, as GAL4DB does in this experiment. In addition, although the LAR seems to play a critical role in this process, the GAR and/or the PAR may also contribute to the corepressive activity of PELP1.
The GAR Binds to Histones-The GAR contains an unusually long stretch of about 70 glutamic acids (21). If the GAR has a function to tether PELP1 to the chromatin, then the GAR, an extremely acidic domain, would prefer to interact with basic proteins such as histones rather than with DNA. To test the possibility, purified histone proteins and either GST-GAR or GST alone were mixed and pulled down using glutathione beads (Fig. 5, A and B). GST-GAR interacted strongly with histone 3 (H3) in any context, whether it manifested as H3 monomer, H3-H4 complex, or as a part of the core complex. On the other hand, it interacted only weakly with H1 and H4 monomers and H2A-H2B complex, after washing with the buffer containing 300 mM NaCl (Fig. 5B, upper panel), whereas GST alone did not pull down any histone or histones in any forms (Fig. 5A). Under less stringent washing conditions (150 mM NaCl), increased precipitation of H2A/B and H4 with GST-GAR was observed when H2A/B and H4 were incubated with H3 (Fig. 5B, lower panel). Furthermore, in vitro translated full-length PELP1, but not PELP1⌬GAR, was coimmunoprecipitated with H3 (Fig. 5C). These results suggest that PELP1 interacts directly with H3 through the GAR. To confirm this interaction of PELP1 with H3 in vivo, COS-7 cells were transfected with plasmids encoding PELP1 or PELP1⌬GAR, lysed, and immunoprecipitated using anti-PELP1 antibody. Subsequent immunoblot analysis revealed that H3 was coimmunoprecipitated with PELP1 but not with PELP1⌬GAR (Fig. 5D). Furthermore, anti-PELP1 antibody but not normal IgG coimmunoprecipitated H3 with endogenous PELP1 from HeLa cell extracts (Fig. 5E), confirming the physiological interaction between PELP1 and H3.
The glutamic and aspartic acid-rich INHAT domain of SET/ TAF-1␤ (amino acid 225-277) has been shown to interact with histones and to contribute to the corepressive activity of the INHAT complex by suppressing histone acetylation (26). To confirm further the importance of the GAR-histone interaction in the corepressive activity of PELP1, the GAR of PELP1 was replaced by the INHAT domain of SET/TAF-1␤ (PELP1⌬GAR-INHAT). This chimera successfully repressed the serum-induced transactivation of SRE, in fact, even more efficiently than did PELP1 (Fig. 5F). Thus, the GAR of PELP1 probably has a function similar to the INHAT domain, in regard to interaction with histones, which may be important in tethering PELP1 to the chromatin.
Expression of PELP1 Repressed the Transcription of c-fos and c-jun and Retarded Fibroblast Proliferation-As PELP1 exhibits corepressive activities for AP1 and SRE (Fig. 1), it is expected that PELP1 might play a role in cell proliferation. Two mouse fibroblast cell lines, NIH3T3 and C3H10T1/2, were stably transfected with cDNAs encoding PELP1, PELP1⌬GAR, and neo, and their proliferations were measured and compared (Fig. 6A). The stable expression of full-length PELP1 severely retarded the proliferation of all cloned cells of both fibroblast lines. During the proliferation study, no significant death of PELP1-expressing cells was observed (data not shown). On the other hand, the growth rate of cells stably expressing  4) antibodies, and then incubated with the SRE oligonucleotide radiolabeled with [␥-32 P]ATP. Radiolabeled bands containing SRE, SRF, and PELP1 (arrowhead), and bands supershifted by antibody against SRF or PELP (arrow) are marked. C, schematic diagram of mouse c-fos gene and ChIP assay. Two regions of DNA subjected to amplification in the ChIP assay were marked with barbell lines at Ϫ290 to Ϫ21 nt and ϩ2830 to ϩ 3169 nt. Filled square indicates SRE. NIH3T3 cells were transfected with pcDNA3-PELP1 and subjected to the ChIP assay using normal rabbit IgG (nIgG) and anti-PELP1 antibody. The data from the ChIP assay represents three independent experiments. D, Re-ChIP assay was performed according to the protocol described under "Experimental Procedures" using NIH3T3 cells ectopically expressing PELP1 and SRF and antibodies against anti-PELP1 and SRF antibodies. The region corresponding to Ϫ290 to Ϫ21 nt of the c-fos gene was amplified.
PELP1⌬GAR was largely identical with that of the control neo-transfected cells, although the expression level of PELP1⌬GAR was similar to or higher than that of PELP1 (Fig. 6B).
PELP1-induced growth retardation (Fig. 6A) and the repressed transcription of SRE and AP1 (Fig. 1, E and F) indicated that immediate-early serum-response genes such as c-fos and c-jun might be target genes of the corepressive activity of PELP1. To examine this in the cells expressing relatively high levels of PELP1, the serum-induced activation of both c-fos and c-jun genes was accessed by semi-quantitative RT-PCR, using extracts of NIH3T3 cells transfected with proper cDNAs and enriched by G418 selection for 10 days (Fig. 6, C-E). Upon serum stimulation, the expression of both c-fos and c-jun in the PELP1⌬GAR-expressing cells was apparently up-regulated to levels similar to those in the control neo-expressing cells (Fig. 6,  C and D, respectively). On the other hand, overexpression of PELP1 repressed the serum-induced activation of c-fos by ϳ50% and repressed the c-jun activation almost completely. Thus, the retarded proliferation of NIH3T3 and C3H10T1/2 cells as a result of PELP1 expression is likely due, at least partly, to the repressed transcription of genes required for cell proliferation, a category that includes c-fos and c-jun.
PELP1 Inhibits Target Gene Expression by Reducing Histone Acetylation-Upon mitogen stimulation, histone acetylation is induced at the very initial step of the transcriptional activation of immediate-early genes, such as c-fos and c-jun (27)(28)(29). The possibility that PELP1 affects the covalent modification of nucleosomes near the SRE of the c-fos gene was investigated by ChIP assay, using antibodies specific for acetylated H3 and H4 (anti-AcH3 and anti-AcH4, respectively). The two sets of prim-ers described in Fig. 3C were used in this assay. The region encompassing Ϫ290 nt to Ϫ21 nt, but not the region encompassing ϩ2830 nt to ϩ3169 nt, was prominently amplified from both anti-AcH3 and anti-AcH4 immunoprecipitates, supporting the hypothesis that H3 and H4 in the promoter region of c-fos gene were specifically acetylated (Fig. 7, A and B). Serum stimulation enhanced the acetylation of both H3 and H4 at the c-fos promoter region in the NIH-PELP1 ⌬GAR and NIH-Neo cell lines (Fig. 7, A and B, lanes 14 and 16), compared with their basal levels (lanes 13 and 15). Acetylation of H3 in NIH-PELP1 was increased upon serum treatment, but in less dramatic fashion than in the NIH-Neo cell line (Fig. 7A, lanes 17 and 18  compared with lanes 13 and 14, and Fig. 7C, left panel). On the other hand, acetylation of H4 in the NIH-PELP1 line was never increased, even after serum stimulation (Fig. 7B, lanes 17 and  18 compared with lanes 13 and 14, and Fig. 7C, right panel). Thus, the repressed transcription of target genes, including c-fos, by PELP1 is likely due, at least in part, to restricted chromatin remodeling caused by reduced histone acetylation.
The LAR of PELP1 Recruits HDAC2 to the Chromatin-Such reduced histone acetylation by PELP1 can be achieved via the recruitment of HDACs to the chromatin or by the suppression of HAT-mediated events. We examined whether the recruitment of HDACs is involved in PELP1-mediated transcriptional repression of the SRE. Expression of PELP1 effectively repressed the serum-induced transactivation of SRE. Treatment with the HDAC inhibitor, trichostatin A (TSA), recovered the serum-induced transcription of SRE in the PELP1-expressing cells (Fig. 8A). Treatment of cells with TSA or another HDAC inhibitor, sodium butyrate, also efficiently rescued the transcription of the G5-tk-luc reporter gene from GAL4DB-LARmediated repression (Fig. 8B). However, nicotinamide, an inhibitor of class III HDAC, appeared to have no effect on the corepressive activities of either GAL4DB-LAR or PELP1 (data not shown). These results suggest that PELP1 may be associating with class I or II HDACs, through the LAR. Indeed, PELP1 and PELP1-LAR, but not PELP1-⌬LAR, coimmunoprecipitated HDAC2 from the lysates of the transfected COS-7 cells (Fig. 8, C and E). Furthermore, in the HeLa cells in which endogenous PELP1 was modestly expressed, PELP1 was coimmunoprecipitated with endogenous HDAC2 (Fig. 8D), confirming the existence of a physiological interaction between PELP1 and HDAC2. Although the involvement of other HDACs cannot FIG. 6. Retarded proliferation of the cells overexpressing PELP1 due to suppressed expression of c-fos and c-jun. A, growth of NIH3T3 and C3H10T1/2 cloned cells stably expressing His-tagged PELP1, but not PELP1 ⌬GAR, was retarded. The stable cell lines (5 ϫ 10 4 ) were seeded on 60-mm dishes in triplicate and cultured at subconfluent states for 10 or 12 days. The number of cells was counted with a hemocytometer. B, expression of exogenously expressed His-tagged PELP1 and PELP1⌬GAR in several NIH3T3 and C3H10T1/2 stable cloned cells. Total cell extracts (50 g) were immunoblotted using anti-His antibody. C and D, expression levels of c-fos and c-jun in the absence and presence of serum stimulation were analyzed by semi-quantitative RT-PCR. The intensity (INT) of bands was quantified by densitometry. The equal amounts of loaded mRNA were verified by the amplification of gapdh mRNA (E). be excluded, HDAC2 showed the most prominent binding, both in vivo and in vitro, among the HDAC1, -2, and -3 used in our experiments (data not shown). Furthermore, a set of ChIP assays, using anti-HDAC2 antibody and an extract of the NIH3T3 cells expressing either PELP1, HDAC2, or both, revealed that HDAC2 was strongly recruited to the c-fos promoter region but only in the presence of PELP1 (Fig. 8F, lane  12). Therefore, the LAR can be considered to contribute to the corepressive activity of PELP1 by recruiting HDAC2 to the chromatin of the target genes.
The GAR of PELP1 Binds to Hypoacetylated Histones and Masks Acetylation Sites-PELP1 was reported to interact with p300/CBP upon ER binding (21). As the GAR is important in H3/H4 acetylation (Fig. 7), although it is dispensable in binding to HDAC2 (Fig. 8), the possibility that the GAR either inhibits HAT catalytic activity or masks the acetylation sites of histones was examined. The region encompassing the HAT domain of p300 was fused to GAL4DB and coexpressed with subdomains of PELP1 and the G5-tk-luc reporter gene (Fig.  9A). Most interestingly, only full-length PELP1 and PELP1GAR resulted in the repression of the HAT-mediated transcription of the reporter gene, up to 50% (Fig. 9A). Thus, it may be that the GAR contributes to the corepressive activity of PELP1 via the suppression of p300/CBP-mediated histone acetylation.
If the reduced histone acetylation is indeed because of the masking effect of GAR by its direct binding to histones, excess histones over GAR proteins in vitro should provide free substrate histones to the available HAT enzymes. On the other hand, if the GAR has an ability to inhibit the catalytic activity of HATs, even the GAR-free histones in the reaction mixture would not be acetylated. Immunoprecipitated p300 and CBP from HeLa cell nuclear extracts efficiently acetylated purified H3 and H4 (Fig. 9B, lanes 1 and 5). Addition of GST into the reaction mixture did not suppress histone acetylation (Fig. 9B,  lane 2). On the other hand, when equal amounts of GST-GAR and histones were incubated with the immunoprecipitated p300 or CBP, histone acetylation underwent a dramatic reduction (Fig. 9B, lane 6). However, addition of histones in excess of GST-GAR resulted in the recovery of H3 and H4 acetylation (Fig. 9B, lanes 7 and 8). Furthermore, overexpression of p300 did not resulted in the recovery of the transcriptional activity of the SRE reporter gene from the PELP1-induced repression (Fig. 9C). In the experiment, the HAT activity of overexpressed p300 was confirmed in two ways. (i) Immunoprecipitated p300 from the cells overexpressing p300 exhibited increased HAT activity (Fig. 9C, inset). (ii) The ectopic expression of p300 efficiently and synergistically coactivated SRE-mediated transactivation in response to serum if another HAT coactivator, SRC-1, was coexpressed, as has been reported (30, 31) (Fig. 9C,  lane 7). These results strongly suggest that reduced histone acetylation by the GAR is most likely due to the interference of substrate histones for acetylation, rather than the inhibition of the catalytic activities of HATs.
Most interestingly, the binding of PELP1 to H3 was completely inhibited by 30 M of unacetylated H3 N-terminal peptide but not by H3 peptide diacetylated at lysine residues 9 and 14 (AcK9 and AcK14) even at a concentration of 90 M (Fig.  9D), suggesting that PELP1 binds preferentially to hypoacetylated histones. Furthermore, unacetylated H3 peptide-immobilized beads efficiently pulled down translated PELP1 in vitro (Fig. 9E, lane 3), whereas monoacetylated H3 peptide (either AcK9 or AcK14) beads pulled down PELP1, but notably less efficiently (Fig. 9E, lanes 4 or 5). Most interestingly, PELP1-H3 interactions were more profoundly affected by Lys-9 acetylation than by Lys-14 acetylation. On the other hand, diacetylated H3 peptide beads were unable to pull down any PELP1 at all (Fig. 9E, lane 6). These results suggest that the position of acetylation, as well as its extent, is important in determining the affinity between PELP1 and histone H3. Taken together, PELP1 functions as a corepressor of NR and non-NR sequencespecific transcription factors, via the suppression of histone acetylation in the target chromatin, using two separate domains, the LAR and the GAR, which recruit HDACs and mask histone acetylation sites, respectively.

DISCUSSION
The present study demonstrates that PELP1 is a histonebinding corepressor of various non-NR sequence-specific transcription factors, including NF-B, AP1, and TCF/SRF in addition to the NRs, with the exception of ER. Several bifunctional coregulators of NR have been reported to exist, including ZAC1, NSD1, and RIP140, which tend to behave either as coactivators or as corepressors under distinct circumstances (17)(18)(19)(20). Although the switching mechanisms responsible for the bifunctionality of these coregulators have yet to be elucidated, mouse variant ZAC1 (mZac1b), which binds to many other coregulators including GRIP1, CBP, and p300, has FIG. 7. Reduced acetylation of histone H3 and H4 by PELP1 expression. A and B, serum-induced acetylation of histones on nucleosomes encompassing SRE was specifically suppressed in the NIH3T3.PELP1 (clone 30), but not in NIH3T3.PELP1⌬GAR (clone 35) or NIH3T3.Neo (clone 38). Sonicated cell extracts were immunoprecipitated (IP) using anti-AcH3 (A) or AcH4 (B), and the two DNA regions described in Fig. 3 were amplified using the immunoprecipitates as templates and specific primer sets. Normal rabbit IgG was used as a control (lanes [7][8][9][10][11][12]. The data are representative of three independent experiments. C, the intensity (INT) of bands amplified at the c-fos promoter region (Ϫ290 to Ϫ21 nt) was quantified by densitometry. Wt, wild type. been proposed to decide its coactivational or corepressive activity according to the relative concentrations or activities of the cellular components (19). On the other hand, NSD1 appears to use its intrinsic activation and repression domains for its bifunctional activity, but it is not clear when and how one domain functions dominantly over the other (18). Most interestingly, PELP1 coactivates ER-mediated transcription but shows corepressive activity for several of the other transcription factors thus far tested. This functional specificity was also maintained in a few other cell lines, including HeLa, COS-7, and SaOS2 ( Fig. 2B and data not shown), suggesting that distinct cellular contexts may not affect the decision of whether PELP1 functions as a coactivator or as a corepressor. As PELP1 recruits HDAC2 through the LAR (Fig. 8) and binds to and masks histones through the GAR (Fig. 9), in order to express the coactivational activity of PELP1, ER binding must somehow dismantle both corepressive tools of histone deacetylation and acquire the means to induce hyperacetylated chromatin structures. CBP/p300 recruitment of PELP1 upon ER binding (21) would explain the latter function. However, it is not clear how ER-PELP1 interactions can induce the inactivation of PELP1-bound HDAC activity or how such interactions could ameliorate the histone-masking effect resulting from PELP1histone interaction. Nevertheless, further studies are required to understand the mechanisms underlying this functional switching.
A few transcriptional corepressors carry two arms to induce and maintain the hypoacetylated state of histone-recruiting HDACs for active deacetylation, and binding to histones for inactivation of HAT or for masking histones from HAT-mediated acetylation. SMRT would be a good example having two adjacent SANT1 and SANT2 motifs, each of which is utilized in interactions with HDAC3 and hypoacetylated histones, respectively (32). This led to a "feed-forward" model of progressive repression, in that the SMRT-HDAC3 complex initiates local histone deacetylation, which, in turn, increases the affinity of the SMRT for histones and further promotes the deacetylation of adjacent histones. In addition, as proposed in a study using another two armed corepressor, Groucho/Tub1, the locally deacetylated heterochromatin may further recruit more corepressor molecules, resulting in the formation of a large transcriptionally silent domain (33). PELP1 seems to operate in a similar mode of action in the repression of target gene transcription.
Deletion of either the LAR or GAR resulted in marked reductions in the corepressive activity of PELP1 (Fig. 4B), whereas expression of isolated LAR or GAR inhibited the corepressive activity of PELP1 in a dominant-negative manner (Fig. 4C). Thus, although individual functions facilitated by the LAR and the GAR are essential in PELP1 activity, simple summation of recruited HDAC2 and the histone masking effect may not represent overall corepressive activity of PELP1. Rather, the presence of these two domains together in cis, as normally found in PELP1, may synergistically repress the target gene transcription.
However, the GAR does not seem to possess HAT inhibitory activity (Fig. 9) and is dispensable with respect to LAR-mediated transcriptional repression (Fig. 4D). Instead, this synergy HDAC2. Sonicated extracts of NIH3T3 cells transfected with plasmids encoding HDAC2 and GAL4DB, GAL4DB-LAR, or GAL4DB-⌬LAR were immunoprecipitated with anti-GAL4DB, and the precipitated proteins were analyzed by immunoblotting using anti-HDAC2 antibody. 5% of extracts used for immunoprecipitation were immunoblotted with anti-HDAC2 and anti-GAL4DB antibodies (input). F, overexpressed HDAC2 recruits to the c-fos promoter region by PELP1. NIH3T3 cells transfected with PELP1 and/or HDAC2 were subjected to ChIP assay, using anti-HDAC2 and normal rabbit IgG. may be explained, at least in part, by another role of the GAR, i.e. tethering of PELP1 to the target chromatin. This is supported by the fact that LAR by itself, when tethered to the chromatin by GAL4DB, effectively repressed transcription of the reporter gene (Fig. 4D). However, this notion raises an important question as to how PELP1 decides its target gene specificity. Most transcriptional corepressors thus far described are guided to the target chromatin by bound transcription factors, including NRs. As PELP1 contains 10 LXXLL putative NR-binding sites in the LAR, which indeed has been shown to bind ER (29), and as PELP1 formed both in vitro and in vivo complexes with SRF on the SRE (Fig. 3, B and C), the target genes of the coregulatory role of PELP1 can be specified by bound transcription factors. However, if PELP1 can also be tethered to the chromatin by the GAR, through binding to hypoacetylated histones, PELP1 should also be able to function as a general silencing mediator of the heterochromatin. Efficient repression of the basal transcription of several reporter genes in a dose-dependent manner (Fig. 1) may then be explained in a way that PELP1 preoccupies hypoacetylated his- FIG. 9. The GAR exhibits repressive activity through the binding to and masking of hypoacetylated histones. A, the GAR repressed HAT-mediated transcription. NIH3T3 cells were cotransfected with G5-tk-Luc, 400 ng of plasmid encoding GAL4DB-HAT domain (amino acids 1284 -1669 of p300), and 400 ng of plasmids containing PELP1, GAR, ⌬GAR, LAR, or PAR, and then analyzed. Data are representative of three independent experiments. B, GAR inhibited the acetylation of histones H3 and H4 by masking their acetylation sites. Purified GST (2 g, lanes 2-4) or GST-GAR (2 g, lanes 6 -8) was mixed with histone H3 and H4 (2.5 g in lanes 1, 2, 5, and 6; 5 g in lanes 3 and 7; and 10 g in lanes  4 and 8). The mixtures were then subjected to in vitro histone acetylation assays, using immunoprecipitated p300 (upper panel) or CBP (middle panel) isolated from 400 g of HeLa nuclear extracts. 10% of input was visualized by silver staining (lower panel). C, exogenous expression of CBP/p300 was unable to ameliorate the PELP1-mediated repressed transcription of SRE. NIH3T3 cells were cotransfected with 100 ng of SRE-Luc, 300 ng of pcDNA-PELP1, along with 50 ng of pcDNA-p300 or SRC-1, and reporter activity was analyzed both before and after serum treatment. HAT activity of p300 in the transfected cells is shown in the inset. D, PELP1 preferentially binds to hypoacetylated histone tails. Increasing amounts of unacetylated H3 peptide or di-acetylated H3 peptide (see "Experimental Procedures") were preincubated with 500 ng of purified GST or GST-GAR, and then 500 ng of histone H3 was added, and the sample was further incubated for 10 min. The mixtures were pulled down using glutathione beads, and bound histones were separated on 13.5% SDS-PAGE and visualized by silver staining. E, PELP1 preferentially binds to hypoacetylated H3 tail peptides. In vitro translated and labeled PELP1 was pulled down by using biotinylated H3 N-terminal peptides and immobilized on streptavidin beads, and then bound PELP1 was visualized by autoradiography. tones through the GAR, even in the absence of bound transcription factor, and promotes deacetylation via the tethered LAR-HDAC complexes, resulting in repressed basal transcription.
Most interestingly, PELP1 interacts with histones using a long acidic amino acid stretch for binding to histones (Fig. 5), instead of using sequence-specific motifs, such as the SANT motif in SMRT and other coregulators. A few histone chaperones, including Nucleophosmin B23, NASP, NAP-1, and TAF-1, have also been demonstrated to interact with histones in a similar manner, using their acidic amino acid stretches (34 -37). Considering the highly basic nature of histones, electrostatic force would play a major role in such interactions. The acidic domain-histone interaction was also described in the process of transcriptional repression. TAF-I␣ and -␤ and pp32 of the INHAT complex bind to and mask histone proteins through their long stretches of acidic amino acids, prevent HAT-mediated acetylation of the histones, and thereby inhibit transcription (26,38). The GAR likely contributes to the corepressive activity of PELP1 in a fashion similar to that of the INHAT complex (Fig. 5F).
Most interestingly, the GAR of PELP1 binds to H3 with a higher affinity in vitro (Fig. 5B), but H4 acetylation was far more severely inhibited than was H3 acetylation, at least in nucleosomes near the SRE of the c-fos gene, when PELP1 was expressed in the cell (Fig. 7). Similarly, the preferential acetylation of H4 over H3 by p300 was also reported to occur during glucose-induced insulin gene transcription, although broad specificity of p300 for the acetylation of H3 and H4 has been well documented (39 -41). As described in the case of the IN-HAT complex, it is possible that PELP1 may form complexes with other component(s) that change histone binding specificity, resulting in the preferential inhibition of H4 acetylation in vivo. Alternatively, the HDAC2 recruited through the LAR (Fig. 8) may be localized at a more accessible position and further deacetylate, H4 whereas the GAR maintains its preferential binding to H3.
Although generally conserved motifs for interaction with HDACs are not currently available, the SANT motif for HDAC3 binding is shared by SMRT and NcoR, and the LXXLL motif in the C-terminal half-of HIRA has been described recently to be essential in interactions with HDAC2 (42). PELP1 does not possess a SANT-like domain but recruits HDAC2 through the LAR, which contains 10 LXXLL motifs (Fig. 8). At present, it is not clear whether any LXXLL motif in the LAR is directly or indirectly involved in interaction with HDAC2, whether the LXXLL motif is specialized with regard to HDAC2 binding, or can be competitively occupied by HDAC2 and NR. It would be of interest if ER and HDAC2 compete for binding to a specific LXXLL motif on the LAR of PELP1. In this case, ER binding could switch the function of PELP1 to that of a coactivator by acetylating histones using recruited p300/CBP HATs, while minimizing the HDAC2-mediated histone deacetylation.
Alternatively, because HDAC2 is a pRb-binding protein and is known to mediate pRb-mediated repression (43), and because a portion of the LAR (amino acids 1-303) interacts with pRb (22), the PELP1-mediated transcriptional repression of SRE could be due to the indirect recruitment of HDAC2 via pRb. However, PELP1 still efficiently repressed the transcriptional activation of SRE in the mouse embryonic fibroblasts lacking the Rb gene (data not shown). Thus, PELP1 does not appear to use pRb in its corepressive activity for the transcriptional activation of SRE.
Until now, the functions of PELP1 have been studied from the perspective of ER pathways, either as a coactivator of ER-mediated transcription, or as a modulator of the nongenomic activity of ER (21,23). In the present study, we report that PELP1 is able to repress the transcription induced by various sequence-specific transcription factors, including AP1, TCF/SRF, NF-B, as well as some NRs with the exception of ER. This corepressive activity was achieved mainly via the recruitment of HDAC activity, and also by histone binding, employing two separate domains. However, as the present study narrowly focused on the biochemical mechanisms underlying the corepressive activity of PELP1, more studies are required to understand better the biological role of PELP1. It would be of particular and immediate interest to determine how PELP1 overcomes its corepressive activity and manifests its ER coactivator function.