Recruitment of cAMP-response Element-binding Protein and Histone Deacetylase Has Opposite Effects on Glucocorticoid Receptor Gene Transcription

Glucocorticoids control the synthesis of the glucocorticoid receptor (GR) in various tissues through a negative feedback regulation of the mRNA. In this study, we have identified feedback regulatory domains in the human GR gene promoter and examined the roles of GR, the cAMP-response element-binding protein (CREB), and HDAC-6 in association with promoter elements of the human GR gene. Using breast cancer T47D and HeLa-GR cells, we identify specific negative glucocorticoid-response elements in the GR gene. The feedback regulatory domains were also involved in interactions with CREB. GR-bound negative glucocorticoid-response elements recruited HDAC-6, and this was dependent on treatment with dexamethasone. Both CREB and HDAC-6 formed complexes with GR-dexamethasone. The HDAC-6 LXXLL motif between amino acids 313 and 418 made direct contact with the GR AF-1 domain. Interestingly enough, although the level of GR decreased in CREB knockdown cells, it was elevated in HDAC-6 knockdown cells. Our results suggest that CREB-P is dephosphorylated and that HDAC-6 is recruited by the GR, and they play opposite roles in the negative feedback regulation of the GR gene.

The hypothalamic-pituitary-adrenal axis is regulated by corticotrophin-releasing hormone (CRH), 2 arginine vasopressin (AVP), and glucocorticoids. In response to stress, the neuroendocrine system releases CRH and arginine vasopressin to regulate adrenocorticotropin (ACTH) hormone secretion from the anterior pituitary and glucocorticoids from the adrenal cortex (1,2). Increasing levels of circulating glucocorticoids downregulate the synthesis of the glucocorticoid receptor (GR) in various tissues (3) in an endocrine feedback loop and produce inhibitory effects on the secretion of CRH and ACTH (4,5) to bring back the system to its resting state. When this balance is perturbed, as in glucocorticoid resistance, the decrease in the negative feedback control of cortisol in the hypothalamic-pituitary-adrenal axis leads to increased secretion of ACTH and cortisol and resistance to adrenal suppression by DEX (6,7). The diverse pathway through which glucocorticoids regulate gene transcription is illustrated in reports where glucocorticoids directly affect expression of arginine vasopressin through a glucocorticoid-response element (GRE) (8) and negative glucocorticoid-response elements (9,10) and is a receptor-mediated interaction with nonreceptor factors (11,12). Glucocorticoids inhibit CRH secretion and in doing so inhibit pro-opiomelanocortin synthesis (13). Furthermore, the proopiomelanocortin nGRE suppresses pro-opiomelanocortin synthesis in a hormone-dependent manner (14). Like the AVP gene, response to glucocorticoids by the phosphoenolpyruvate carboxykinase gene involves a functional interaction between the glucocorticoid-response element and cAMP-response element (CRE), which binds the cAMP-response element-binding protein, CREB (15). Despite this wealth of information, the precise mechanism of feedback regulation remains elusive.
In our previous studies, we demonstrated the negative feedback regulation of the GR gene in Att-20 cells (16). In target tissues, biological actions of glucocorticoid hormones are transmitted by the GR. At the cellular level, GR⅐hormone complexes mediate cell-specific response by binding to specific DNA sequences to modulate the expression of target genes. The human GR gene promoter lacks the classical TATA element, and the authentic GR transcription start site was identified by S1 nuclease mapping (17)(18)(19)(20). Using extensive biochemical, recombinant, and affinity chromatographic techniques, we have characterized several factors binding to the promoter region of the human GR gene (21). We have previously reported the presence of the feedback regulatory (FBR) sequence motif within the Ϫ2846 bp (29,335 bp) upstream of the GR cap site (32,181 bp of the GR gene) using S1 nuclease mapping and primer extension (20). There are several regulatory sequences within this region as follows: at least two nGRE motifs at Ϫ1796 and Ϫ1485 and potential cAMP motifs at Ϫ1460 and Ϫ1073.
The GR is down-regulated by its ligand in a mechanism of negative feedback regulation of its mRNA. Upon translocation to the nucleus, the GR⅐hormone complex binds to specific DNA sequences (22). In the nucleus, however, eukaryotic chromosomes are compacted together, and most DNA sequences are inaccessible to transcription factors. Alteration in chromatin structure modulates the accessibility of regulatory sites to DNA-binding proteins. The core of the nucleosome, the basic subunit of chromatin, consists of an octamer of two copies each of four core histones as follows: an H3-H4 tetramer and dimers of H2A-H2B around which 147 bp of DNA are wound (23). This core is hydrophobic and conceals the C-terminal regions of the histones, whereas the N-terminal tails with highly conserved lysines protrude. Acetylation of these lysines is one of the first steps of actively transcribed chromatin (24) followed by the removal of acetyl groups on the N-terminal tails by histone deacetylases (HDACs) returning chromatin to its condensed structure where accessibility to DNA is restricted (25).
In this study, we have identified FBR sequences in the human GR gene promoter and examined the roles of GR, CREB, and HDAC-6 in the negative feedback regulation associated with promoter elements of the human GR gene. Because the overexpression of factors involved in regulatory pathways could lead to misinterpretation, we used HeLa-GR stable transfectants and T47D breast cancer cells expressing endogenous GR as the source for nucleosomal DNA in our core experiments. We demonstrate that the enhancer and suppression activities contained within the promoter elements of the GR gene occur through the interaction of GR and the active recruitment of HDAC-6 and phosphorylated CREB (CREB-P) by the ligandbound GR. We demonstrate here a direct involvement between GR, CREB, and HDAC-6 in the feedback regulation.
Cell Culture and Transfection-Cells were grown in minimal essential medium or Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum (FBS) in a 5% CO 2 incubator. Cells (HeLa, HeLa-GR, and T47D) were plated at a density of 10 6 cells per well in 6-well plates containing 1 ml of regular medium supplemented with dextran charcoal-treated (DCC)-FBS 12 h before transfection with 0.5 g of CsCl-purified reporters and 0.5 g of pcDNA 3.1 ␤-galactosidase (Invitrogen) using GenJuice from Novagen. After 12 h, cells were washed; 5% DCC-FBS containing 500 l of medium supplemented with ligands (10 Ϫ11 to 10 Ϫ7 M) were added, and the incubation was continued for 12 h in a CO 2 incubator at 37°C. Cells were collected by mild trypsinization and resuspended in 50 l of luciferase extraction buffer (Promega) containing protease inhibitors and 1 mM phenylmethylsulfonyl fluoride (PMSF). The cells were lysed by three cycles of freeze-thaw and clarified by centrifugation at 4°C. An aliquot of 10 l was used to determine ␤-galactosidase activity using the ␤-galactosidase assay kit (Stratagene). Aliquots containing 10 units of ␤-galactosidase were used to determine luciferase activity. Experiments were repeated at least three times, and the data were plotted using SigmaPlot. For the ChIP assays, cells were grown in 250-ml culture flasks as indicated and counted before sonication.
Construction of pGL 4.14 -TK-Luciferase and Other Reporter Vectors-The herpes simplex virus TK promoter was amplified using forward primer 5Ј-CAAGATCTGAATTCGAACACG-CAGATGCAGTCGGGGCGG-3Ј and reverse primer 5Ј-TGCCAAGCTTCTGCAGGGTCGCTCGGTGTTCGAGGC-CACA-3Ј from pBL-TK-CAT vector, restriction-digested with BglII ϩ HindIII, and ligated into pGL 4.14 -luciferase (Neo and puromycin) vectors (Promega). The minimum synthetic TK promoter does not contain the cAMP-binding site present in the wild type herpes simplex virus TK promoter. The 3ϫGRE synthetic palindrome was inserted into the puromycin-resistant TK-luciferase vector to generate the pGRE 3ϫ -TK-luciferase reporter. The ChIP-generated PCR fragments of the human GR gene flanking 5Ј-BamHI-EcoRI-3Ј were cloned into pcDNA4.1-HIS/MAX-Topo cloning vector (Invitrogen), and the plasmids were digested to purify the inserts and cloned into pGL 4.14 -TK-luciferase (Neo) vector. The sequence and orientation of the cloned inserts were confirmed by DNA sequencing.
Human GR Gene 5Ј and 3Ј Deletion Mutant Reporter Vectors-The construction of HGR2.7CAT was described previously (20). 30-bp oligonucleotides containing the 5Ј end GR gene sequences Ϫ2786, Ϫ2486, Ϫ2196, Ϫ1523, Ϫ1443, Ϫ971, Ϫ761, Ϫ631, Ϫ531, and Ϫ278 flanking 5Ј XhoI and 3Ј oligonucleotide at Ϫ40 flanking BglII were used to amplify the 5Ј deletion mutants using GR gene genomic DNA isolated from human EMBL genomic library (supplemental Table 1). The amplifica-tion conditions were 94 to 56 to 68°C for 35 cycles. Amplified products were cloned into pcDNA4.1-HIS/MAX-Topo vector (Invitrogen), and plasmids were purified and sequenced before isolating the XhoI-BglII fragments and ligated into pGL 4.14 -TK-luciferase vector (Neo) lacking the CRE present in the herpes simplex virus TK promoter. The 3Ј deletion mutants were similarly amplified using synthetic primers containing ϩ8-, Ϫ278-, Ϫ531-, Ϫ631-, Ϫ731-, Ϫ971-, Ϫ1443-, Ϫ1523-, Ϫ2196-, and Ϫ2486-flanking BglII and a 5Ј primer flanking the XhoI site at Ϫ2786 and cloned following Topo cloning and sequencing into pGL 4.14 -TK-luciferase (Neo) vector. The amplified product from ϩ8 to Ϫ2786 and ϩ8 to Ϫ1523 was cloned into promoter-less pGL 4.14 -luciferase (Neo). For ChIP assays, additional GR gene reporters were generated using 5Ј end primers and 3Ј primer from the region ϩ8 containing endogenous GR gene promoter and Cap site. The reporter plasmids were purified by two CsCl gradient centrifugations, and 0.5 g/well were used with 0.25 g of pcDNA3.1-␤-galactosidase for transfecting T47D cells using GenJuice (Novagen). ␤-Galactosidase activities were determined using the ␤-galactosidase kit (Stratagene) as described before. Total light emission during the initial 20 s of reaction was measured with a luminometer (Berthold Luminat LB 9501). Stable transfectants were generated in HeLa and T47D cells by selection with G418.
Point Mutants of the FBR Domain-Single base mutagenesis was performed using QuikChange XL site-directed mutagenesis kit (catalog no. 200516) from Stratagene and pGL 4.14 Ϫ1523/ϩ8-luciferase as template. Oligonucleotides containing mutants (supplemental Table 2) and complementary strands were synthesized and purified on 5% acrylamide gels according to the manufacturer's suggestions. The sequences of the mutants were confirmed by DNA sequencing.
GST Pulldown-GST-GR WT, GST-GR S113A, GST-GR S141A, GST-GR S203A, GST-GR S211A, or GST-GR S226A was expressed in Escherichia coli BL21 cells and prepared by standard methods (26,27). Full-length substitution mutants of [ 35 S]methionine-labeled factors were synthesized by in vitro transcription and translation (TNT Promega Corp., Madison, WI). Equivalent amounts of GST or GST fusion proteins were used in vitro binding assays as described previously.
GR Antibody Characterization-Cells were grown in medium supplemented with 10% FBS to 80% confluence and harvested with the aid of a rubber policeman, washed twice with ice-cold phosphate-buffered saline, and resuspended in 500 l of a buffer containing 50 mM Hepes, pH 7.4, 100 mM NaCl, 0.25 mM EDTA, 1 mM dithiothreitol, 10% glycerol, 0.5% Nonidet P-40, 1 mM PMSF, and complete protease inhibitor mixture (extraction buffer). Following sonication, the cell extracts were clarified by centrifugation at 15,000 rpm for 15 min at 4°C. Aliquots of 10 -25 g of total proteins were diluted with equal volumes of 2ϫ SDS sample buffer and size-fractionated on 12% SDS-polyacrylamide mini-gels by electrophoresis with prestained molecular weight markers (New England Biolabs). The gels were electroblotted onto Immobilon-P membranes (Millipore), and respective regions were cut out with the help of a scalpel and incubated with 2% dry milk in blocking buffer (Tris-HCl, pH 7.6, 120 mM KCl, 0.25 mM EDTA, 1 mM dithiothreitol, 10%, 0.05% Tween 20, 1 mM PMSF, and complete protease inhibitor mixture). The blots were incubated with polyclonal antibodies against GR, CREB, HDAC-1-6, acetyl histone H3, or acetyl histone H4 following the manufacturer's suggestions, washed, and incubated with 1:1000 diluted secondary horseradish peroxidase-conjugated anti-rabbit IgGs or anti-goat IgGs coupled to peroxidase and visualized using the Visualizer EC detection kit (Upstate). For immunoprecipitation assays, the cells growing in log phase in 100-mm Petri dishes were transferred into medium without methionine for 24 h before adding [ 35 S]methionine containing Dulbecco's modified Eagle's medium and grown in the labeling medium for 36 h before collection. The labeled cells were washed several times with ice-cold phosphate-buffered saline and resuspended in 250 l of extraction buffer, sonicated, and clarified by centrifugation. Aliquots of 10 -25 g of total cell extracts were used for immunoprecipitation using protein G magnetic beads from Active Motif. The immunoprecipitates were boiled in SDS sample buffer and analyzed by SDS-PAGE followed by fluorography using Enhancer. The dried gels were autoradiographed using Kodak BioMax films and intensifying screens.
Transfection and Selection of Cells with G418-HeLa cells were plated in 6-well plates and transfected using 1 g of pTAP-HIS-GR using GenJuice. The cells were supplemented with selection medium containing G418 (Invitrogen) 12 h posttransfection, and selection was continued until the appearance of colonies (4 -6 weeks). Six colonies were isolated with the help of a pipette tip held under a microscope, resuspended in selection medium, and grown in 25-ml culture flasks to confluence. For immunoprecipitation and transactivation analyses, cells were further amplified in selection medium in 250-ml culture flasks. For the ChIP and transactivation experiments, one of the selected HeLa-GR cultures was used.
Time Course of DEX Treatments-HeLa-GR and T47D cells grown to 70 -80% were plated in 250-ml culture flasks overnight in DCC-FBS-containing medium. Cells were treated with 10 Ϫ7 M DEX for 5, 10, 30, 60, and 120 min and overnight before collection for ChIP assays. Identical experiments were performed at least three times to evaluate the kinetics of promoter occupancy.
ChIP Assays and Amplification of GR-binding Sites Using Primers Designed for the GR-binding Site-ChIP assays were performed using the ChIP-IT Express kits from Active Motif (supplemental Table 3). Sonication was optimized to yield 600 -1000 bp of sheared DNA in chromatin samples from 5 ϫ 10 7 cells/ml of ChIP buffer, and 50 l containing (15 g of DNA) was used for immunoprecipitation with specific preimmune IgG (PI) antibodies. Briefly, after overnight incubation with IgGs ϩ protein G magnetic beads, samples were washed once with 1 ml of low salt ChIP buffer, once with 0.5 M NaCl containing high salt ChIP buffer, followed by 1 ml of LiCl/detergent buffer (28,29). Finally, the beads were washed with 500 l of TE and resuspended in 8 l of 5 M NaCl to reverse crosslinking, and samples were incubated overnight at 65°C in a water bath. The next morning, the samples were briefly centrifuged and separated from the magnetic beads, and 1 l of RNase A was added and incubated for 30 min at 37°C. A mixture of 4 l of 0.5 M EDTA, 8 l of 1 M Tris-HCl, pH 7.8, and 1 l of proteinase K was added to each sample and incubated at 42°C for 2 h. The volume was adjusted to 200 l with TE, and the DNA was purified by phenol/chloroform extraction and precipitation with ethanol in the presence of 20 g of glycogen.
After centrifugation for 15 min at 4°C at high speed, the solvents were removed, and the pellet was dissolved in 100 l of TE. The DNA was reprecipitated using 20 l of 2 M potassium acetate and 250 l of ethanol. The samples were dissolved in 25 l of TE, and 5 l were used for amplification (94 to 56 to 68°C, 35 cycles) using 25 l of Advantage PCR Supermix (Invitrogen) and 1 l each of primers Ϫ7820/Ϫ7531, Ϫ2846/Ϫ2486, Ϫ1806/Ϫ1443, Ϫ1523/Ϫ971, and Ϫ971/Ϫ530 of human GR gene 5Ј-flanking sequences and control glyceraldehyde-3phosphate dehydrogenase primers.
ChIP and Immunodetection-After a final wash of the immunoprecipitated chromatin⅐IgG complexes using 1 mg of total chromatin proteins, protein G magnetic beads were washed three times as described, and the beads were suspended in 50 l of 2ϫ SDS sample buffer and boiled for 30 min in a water bath to reverse the formaldehyde cross-linking (30). Aliquots of 25 l were resolved on 12% polyacrylamide gels by electrophoresis with pre-stained protein markers. The gels were blotted onto Immobilon-P, and the respective regions were cut out for immunodetection using specific antibodies against GR, CREB-P, HDAC, acetyl-H3K9, and acetyl-H4K12 rabbit or goat polyclonal antibodies and respective second peroxidase antibody-coupled antibodies. Western blotting detection was performed using Visualizer EC Western detection kit from Upstate (catalog no. 64-301). Primer used for ChIP assay was hGR gene ID 2808, GenBank TM accession number NC_000005.8, transcription start site 32181 (supplemental Table 2).
Ligand Treatment and CREB and GR Interaction in T47D Cells-CREB occupancy of GR gene promoter was analyzed by ChIP assays using chromatin prepared from ligand-treated T47D cells and antibodies against GR and CREB using primers Ϫ2846/Ϫ2486, Ϫ1806/Ϫ1443, Ϫ1523/Ϫ971, and Ϫ971/Ϫ530 of human GR gene-flanking sequences. T47D cells were treated with DEX, 8-Br-cAMP, DEX ϩ 8-Br-cAMP, Bt 2 -cAMP, OKA, DEX ϩ Bt 2 -cAMP, TSA, and DEX ϩ TSA to determine the effect of cAMP-mediated response to CREB-P binding and to study the possible recruitment HDACs in GR gene silencing.
Dose Response to GR Transactivation and GR Gene Feedback Regulation-Dose response to DEX was analyzed by treating the pGRE 3ϫ -TK-luciferase-T47D transfectants with 0.1-100 nM DEX as a control. The GR gene reporters Ϫ1523/Ϫ971-TKluciferase, Ϫ1806/Ϫ1443-TK-luciferase, Ϫ1806/Ϫ971-luciferase, and Ϫ2846/ϩ8-luciferase were transfected into T47D cells and treated with 0.01 to 100 nM DEX, and the dose response was measured.
Coimmunoprecipitation of GR-HDAC and GR-CREB-P-HeLa cells were transfected with control vector, GR, CREB, HDAC-6 alone or in combinations, and the transfectants were labeled with [ 35 S]methionine. Whole cell extracts were prepared and 10 g used for immunoprecipitation with CREB-P, GR, and His 6 rabbit polyclonal antibodies and Dynabeads protein G (Invitrogen). The samples were boiled in SDS sample buffer, separated on SDS-polyacrylamide gels, and visualized by fluorography.
Knockdown of HDACs and CREB in T47D Cells-Silencing oligonucleotides (supplemental Table 4) for cloning into pSilencer 4.1-CMV Hygro were derived from the insertion of the complete cDNA sequences of HDAC-1, HDAC-2, HDAC-3, HDAC-4, HDAC-5, HDAC-6, and CREB into the Ambion design software according to manufacturer's instructions. Plasmids were purified and sequenced to confirm the presence of appropriate inserted sequences. The effect of silencing on the GR-mediated transactivation was determined by dose response using pGRE 3ϫ -TK-luciferase reporter in silenced and hygromycin-selected stable T47D cells. HDAC-2, HDAC-6, and CREB silenced T47D cells were tested with controls for response to DEX, cAMP, and TSA using pGL 4.14 Ϫ1523/ϩ8-luciferase reporter. Two different knockdown vectors derived from HDAC-1-6 (a and b) were used. Knockdown was further analyzed by rescue of CREB and HDAC-6 in knockdown cells and Western blotting and RT-PCR using actin as control (supplemental Table  5). Western blots were performed with extracts from knockdown cells for the expression of GR using GR IgGs.
Rescue and Expression of GR in CREB and HDAC-6 Knockdown Cells-The CREB and HDAC-6 rescue was performed in knockdown cells by transient transfection of pCMV-CREB or pcDNA1-HDAC-6 or naked expression vectors. The transfectants were labeled with [ 35 S]methionine, and whole cell extracts were prepared. 20 g of the total proteins were incubated with 2.5 l of CREB IgGs or control actin IgGs and immunoabsorbed using 50 l of Dynabeads and analyzed by SDS-polyacrylamide gels and visualized by fluorography. Total RNAs were prepared as indicated, and RT-PCR (kit) was performed using CREB and actin-specific primers (supplemental Table 5). GR expression was analyzed using 20 g of the whole cell extracts from nonlabeled cells by Western blotting using GR IgGs or actin IgGs as controls. Similar rescue and expression analysis was performed using HDAC-6 knockdown and control T47D cells using HDAC-6 antibodies in immunoprecipitation and HDAC-6specific primers (supplemental Table 5) for RT-PCR. The GR expression was analyzed by Western blotting using actin as controls.
Expression of GR WT and Mutants and CREB in Bacteria-Human GR mutants S113A, S141A, S203A, S211A, and S226A were generated by oligonucleotide (supplemental Table 6) mutagenesis using hGR single-stranded DNA template as described (28). Other GR-GST fusion proteins were generated by cloning the fragments into pGEX vectors. Expression using pRSET vectors (Invitrogen) was used for the production of CREB, CREB-133, hGR, and HDAC-6.
In Vivo Protein-Protein Interaction-HDAC-6 contains four LXXLL motifs. A set of HDAC-6, CREB, and CREB-133 deletion mutants was generated by PCR using synthetic primers flanking suitable restriction sites for cloning into pVP16 vector. Human GR WT and AF-1 single serine to alanine mutant cDNAs were cloned into pM vector, and plasmids were purified by CsCl gradient centrifugation. The two-hybrid assay was performed using pG 5 luciferase reporter.
CREB Phosphorylation-Purified bacterially expressed CREB and CREB-133 were phosphorylated after minor modifications as described by Wadzinki et al. (31). The incubation buffer contained 0.5 mM PMSF and complete protease inhibitor mixtures (Roche Applied Science). CREB-32 P was purified as described, and GST-GR interaction with CREB-32 P was performed using CREB-32 P.
Immunoprecipitation of GR and CREB-P in Vivo-T47D cells were grown in 5% DCC-FBS-supplemented medium without phenol red and methionine to 75% confluency and treated with 100 nM DEX, 1 M Br-cAMP, DEX ϩ Br-cAMP, 1 M OKA, DEX ϩ OKA, OKA ϩ Br-cAMP, 100 nM TSA, DEX ϩ TSA for 4 h with the addition of [ 35 S]methionine as described (27). Cells were washed thoroughly with ice-cold phosphate-buffered saline and collected by scraping with a rubber policeman. Whole cell extracts were prepared in 25 mM Hepes, pH 7.6, 50 mM KCl, 50 mM NaCl, 1 mM dithiothreitol, 1 mM sodium orthovanadate, 0.5 mM PMSF, 0.05% Tween 20 and complete protease inhibitor mixture (Roche Applied Science). Protein concentration was determined using an aliquot with Bio-Rad protein assay reagent. Duplicate aliquots of 20 g of total proteins were incubated in duplicate with GR-IgG-Sepharose or CREB-P IgGs or antiactin IgGs in a final volume of 200 l. After 2 h, 50 l of protein-G magnetic beads (1:1 suspension) were added to the CREB-P and control actin incubates. Beads were washed five times, and the bound materials were eluted and subjected to SDS-PAGE (12% gel).

DNA-Protein Interaction by Mobility Shift Assays and Supershift
Assays-5Ј end-labeled double-stranded oligonucleotides used for generating point mutants of the FBR motif and wild type were end-labeled using [␥-32 P]ATP and kinase labeling kit from GE Healthcare. The labeled products were purified using Qiagen purification column, and 1500 cpm of the labeled oligonucleotides were incubated with extracts containing factors prepared from transfected HeLa cells. Competitions were performed with 100-fold molar excess of nonlabeled competitor double-stranded oligonucleotides. For supershift assays, 1 l of the purified IgGs was used during incubations. Fig. 1 shows the 5Ј-regulatory domains of the GR gene promoter. This map is derived from GR gene sequences compiled in the GenBank TM . Potential regulatory sites are marked, and their positions are indicated. In this study, we have focused on the two potential negative GR-responsive motifs, nGRE 5Ј-CGTCCA-3Ј (Ϫ1796, 30,385) and 5Ј-CGTCCA-3Ј (Ϫ1485, 30,696) between Ϫ2846 and the cap site of GR and two potential cAMP regulatory motifs 5Ј-AGAGGTCA-3Ј (Ϫ1460, 30,720) and 5Ј-TGCCGGCA-3Ј (Ϫ1073, 31,108). The second numbers are reference to the GR gene NC_000005.8.

Organization of the Human GR Gene Promoter-
An FBR Element Is Located between Ϫ1523 and Ϫ1443-We used a set of 5Ј ( Fig. 2A, left) and 3Ј deletion mutants (Fig. 2B, left) of the FBR region to generate luciferase reporter vectors (supplemental Table 1) and then transfected T47D cells. Treatment with DEX alone resulted in a 3-fold decrease in luciferase expression from the upstream promoter deletions Ϫ2786/Ϫ40, Ϫ2486/Ϫ40, Ϫ2196/Ϫ40, and Ϫ1523/Ϫ40 when compared with untreated cells ( Fig. 2A, top right control). No reporter activity was observed with downstream Ϫ1443/ Ϫ40, Ϫ971/Ϫ40, Ϫ761/Ϫ40, Ϫ631/Ϫ40, Ϫ531/Ϫ40, and Ϫ278/ Ϫ40 5Јdeletion mutants ( Fig. 2A, lower right). However, when 1 M cAMP was added, promoter activity increased 3-4-fold with the upstream promoter deletions. When both cAMP and 100 nM DEX were added, cAMP-induced activity of the upstream reporters decreased 5-fold. This was even lower than the level with untreated cells. Although reporter expression with deletions downstream of Ϫ1523/Ϫ40 in the presence of cAMP was observed, the level was significantly lower. The addition of cAMP and 100 nM DEX did not have any significant effect. These results suggest that the regulatory region of GR gene and the nGRE motif is located between Ϫ1523 and Ϫ1443.
Similar analyses were performed to map the pattern of cAMP, DEX, and TSA response by generating a battery of 3Ј deletion mutants (supplemental Table 1). The results of the experiments are compiled in Fig. 2B. Deletion of upstream promoter sequences as in Ϫ2786/Ϫ1523, Ϫ2786/Ϫ2196, and Ϫ2486/Ϫ2486 resulted in complete loss of responsiveness with DEX, cAMP, and TSA alone or in combination. Reporter expression from downstream promoter sequences Ϫ2786/ϩ8, Ϫ2786/Ϫ278, Ϫ2786/Ϫ531, Ϫ2786/Ϫ731, Ϫ2786/Ϫ971, and FIGURE 1. Human glucocorticoid receptor gene 5-regulatory domains. Schematic presentation of human GR gene compiled from GenBank TM information. Potential recognition sites were analyzed using the Gene Construction kit. The analysis of the 5Ј region extending to Ϫ7531 was considered in this study. FEBRUARY 12, 2010 • VOLUME 285 • NUMBER 7

JOURNAL OF BIOLOGICAL CHEMISTRY 4493
Ϫ2786/Ϫ1443 in the presence of DEX, cAMP, and TSA alone or in combination was similar to that of the 5Ј upstream deletion mutants. The possibility that this domain contained DNA elements for the interaction of GR, CREB, and HDAC became clear from these studies. Furthermore, the proximity of the GRbinding site at nGRE Ϫ1485 and the CRE at Ϫ1460 suggested that there could be functional interaction between GR and CREB.
GR recruits deacetylases to inhibit expression of genes, and TSA releases this restraint. As TSA is a general deacetylase inhibitor, we did observe a basal level of induction as a result of its effect on chromatin remodeling. However, our results also indicate that the lack of reporter activity in treatments with TSA alone can also be attributed to the failure of the defective nGRE to recruit GR⅐DEX complexes. These data confirm the presence of a composite nGRE-CRE element between Ϫ1523 and Ϫ1443.

T47D and HeLa-GR Cells Express High Levels of GR-
The analysis of how GR induces feedback control required not only cells expressing sufficient levels of endogenous GR (T47D) but also cells with intact GR signaling pathways (HeLa). As a first step, we determined GR expression and ligand response in HeLa, T47D, and HeLa-GR cells (Fig. 3A). Luciferase activity in HeLa-GR cells (Fig. 3A, right) was about seven times higher than with control HeLa cells (left) where response, although weak, still indicated an intact GR signaling pathway. Similarly, the level of induction in T47D (Fig. 3A, center) cells was 2-fold when compared with HeLa cells. Using antibodies against GR, we compared the level of GR expression in HeLa, T47D, and HeLa-GR cells (Fig. 3B). No GR band was observed in the PIprecipitated sample (Fig. 3B, lane 1). Although the GR protein was visible in HeLa cells (Fig. 3B, lanes 3 and 4), quantities were higher in T47D cells (lane 5). The increase in dose response of HeLa-GR cells was indeed due to high expression of GR indi-FIGURE 2. Human GR gene 5 and 3 deletion mutation and analysis of feedback regulation. A, left, 5Ј end of the deletion mutants were used for the construction of luciferase reporter vectors. Right, upper and lower panels, T47D cells were transfected with 5Ј deletion mutants as described under "Experimental Procedures," and reporter activity was determined following treatment with ligands as indicated. B, left, 3Ј end deletions of GR gene used for reporter constructs. Right, upper and lower panels, similar transfection protocol as in A but with 3Ј deletion mutants, and reporter activity was measured following treatments with various ligands as indicated. Values are the average of triplicates. cating that HeLa-GR cells (Fig. 3B, lanes 2 and 6) could very well be used as a source of GR in subsequent analyses. We used anti-actin as a control measure to show that the same amount of total proteins was used in immunoprecipitation assays and to record differences.
GR Binds to Specific Elements in the GR Promoter in T47D Cells and Kinetics of GR Recruitment-To define the role of the GR gene promoter motifs, a series of overlapping synthetic primers were designed. The supplemental Table 3 shows the sequences of the oligonucleotides used for the ChIP assays. Using antibodies against GR, the resulting precipitated DNA was analyzed by PCR with primers covering the regulatory domain of the GR gene. We then determined the time course of GR recruitment. In T47D cells, GR recruitment peaked at 30 min post-DEX treatment. Of the four domains examined, three covering the motifs between Ϫ1523 and Ϫ1443 showed similar patterns of GR recruitment (Fig. 4A). There was no signal with primers between Ϫ971/Ϫ530 either with PI IgGs or GR IgGs and thus was not considered further. The housekeeping gene glyceraldehyde-3-phosphate dehydrogenase served as another control, and its absence in the GR IgG-immunoprecipitated nucleosomal DNA indicates the specificity of GR IgGs. A similar analysis with HeLa-GR demonstrated the high specificity of GR interaction with the GR gene promoter domains (Fig.  4B). The reporter pGL 4.14 Ϫ7820/ Ϫ7531-TK-luciferase did not induce a response (data not shown) and hence was not included in subsequent investigations.
CREB Is Recruited by the GR Gene Promoter-The primer set between Ϫ1806 and Ϫ971 contains two putative CRE sequences suggesting that FBR elements could play a role in CREB recruitment. To address this, chromatin prepared from ligand-treated T47D cells was analyzed by ChIP assays using CREB-IgG (Fig. 5A). Although analogs of cAMP increased recruitment (Fig.  5A, lanes 3, 5, and 7), DEX treatment of the cells was inhibitory (lanes 2, 4, 6, 8, and 10) when compared with untreated cells (lane 1). However, CREB levels were higher with the promoter 1523/Ϫ971, which contains both potential CREB sites. These results suggest that binding of CREB to sites within the promoter regions of the GR gene is affected by the concomitant binding of GR to the nGRE motifs.
It has been well documented that the phosphorylated form of CREB (CREB-P) binds transcription factors (11,12). Western blots (Fig. 5B) were probed with antibodies against HDAC-1-5 (results not shown), GR IgGs, and a mixture of HDAC-6, CREB-P, acetyl-H3K9, and acetyl-H4K12 IgGs. Although treatment with DEX (Fig. 5B, lane 2) alone inhibited recruitment of CREB-P, cAMP or its derivatives alone showed a substantial enrichment of CREB-P (lanes 3, 5, and 7). Simultaneous incubation with DEX had a negative effect on the recruitment of CREB-P (Fig. 5B, lanes 4, 6, and 8) when compared with nontreated cells (lane 1). Intriguingly, enhanced levels of GR and HDAC-6 were observed in the presence of DEX (Fig. 5B, lanes 2, 4, 6, 8, and 10), suggesting that there was an interaction between GR and HDAC-6. Our results seem to indicate that CREB was not recruited in the presence of TSA; however, a slightly higher level of CREB was present during TSA ϩ DEX treatment (Fig. 5B, lanes 9 and  10). These experiments suggest that DEX-bound GR prevents CREB binding to the promoter. A possible theory could be that this occurs either through dephosphorylation of CREB-P or through steric hindrance of CREB-P binding by the recruitment of GR⅐DEX complexes. Preliminary experiments using antibodies against histones 1-5 were negative and hence were also not included in subsequent studies. Control actin was used to monitor that equivalent quantities of the chromatin were used for the ChIP assays.
Histones H3 and H4 Are Acetylated in Transcriptionally Active GR Gene Nucleosomes-We employed CREB association assay to detect proteins bound to chromatin in vivo in treated cells. ChIP samples were further processed by reversing formaldehyde cross-linking and by  Table 2. On the left, primer sets used are shown, and the antibodies are indicated on the right. Specificity was determined using preimmune IgGs, and amplification of GR-IgGs precipitated chromatin DNA with glyceraldehyde-3-phosphate dehydrogenase (GADPH) primers. B, protein G magnetic beads containing the chromatin immunoprecipitates were washed three times as described, and the beads were suspended in 50 l of 2ϫ SDS sample buffer, and boiled for 30 min in a water bath to reverse the formaldehyde cross-linking (30). Aliquots of 25 l were resolved on 12% polyacrylamide gels by electrophoresis with pre-stained protein markers. The gels were blotted onto Immobilon-P and probed with a mixture of HDAC-6, GR, CREB-P, acetyl-H3K9, and acetyl-H4K12 IgGs; lower panel shows the actin IPs used as control. C, proteins were induced and prepared as described under "Experimental Procedures." In vitro DEX-dependent dephosphorylation of CREB-P was analyzed using bacterially expressed CREB (lanes 1 and 2) and GR (lane 3); center, CREB-32 P input (lane 1), 32  Western blotting. We also analyzed the consequence of CREB-P recruitment on the expression of histones H3 and H4 in the transcriptional activation of the GR gene (Fig. 5B). The modified histones H3 and H4 were barely visible or absent in DEX or in combination with cAMP derivative-treated cells in which the presence of both GR and HDAC-6 bands were predominant (Fig. 5B, lanes 2, 4, 6, and 8). This would suggest that GR⅐DEX complex recruits HDAC-6, which deacetylates H3 and H4. The absence of HDAC-6 bands in TSA-treated cells (Fig.  5B, lane 9), the enhanced recruitment of CREB-P in cAMP and derivative-treated cells, and the presence of acetylated H3K9 and H4K12 strongly suggest that GR, CREB-P, and HDAC-6 are directly involved in the regulation of the GR gene.
Ligand-bound GR-captured Phosphatase Activity Dephosphorylates CREB-P in Vitro-To study whether GR mediated dephosphorylation of CREB-P, GR wild type and CREB were expressed in BL-21 cells and purified using Talon affinity adsorbent (Fig. 5C). CREB was labeled in vitro with ␥-ATP and protein kinase A (Fig. 5C, center, lanes 1 and 2). Next, we ascertained the status of CREB-P dephosphorylation (Fig. 5C, right). Bacterially expressed control CREB-P (Fig. 5C, lane 1) and control GR in the absence of CREB-P (lane 2) are shown. Talon matrix bound GR treated without DEX (Fig. 5C, lane 3) and with DEX (lanes 4 and 6) before incubation with whole cell extract from HeLa cells, washed, and then incubated with CREB-32 P. The decreased intensity of CREB-P band was due to an activity captured by GR in the presence of DEX from HeLa cell extracts. Extracts were treated with DEX and OKA (Fig. 5C, lane 5) to examine whether the dephosphorylation observed (lanes 4 and 6) was inhibited by OKA, a phosphatase inhibitor. OKA treatment increased the intensity of CREB-P (Fig. 5C, lane  5), and DEX treatment alone decreased it (Fig. 5C, lanes 4 and 6) when compared with untreated cells (lane 3). These results demonstrate that there was enhanced dephosphorylation of CREB-P by the matrix-bound GR⅐DEX complexes, which recruited specific phosphatase from the HeLa cell extract in vitro.
Specific Elements in GR Gene Promoter Suppress Transcription-We asked if specific nGRE motifs in the FBR elements would be involved in suppressing transcription. To do so, dose response to DEX was performed with T47D transfectants (Fig.  6A). Reporter vectors were generated using PCR fragments Ϫ1523/Ϫ971, Ϫ1806/Ϫ1443, and Ϫ1806/Ϫ971 and pHGR 2.7 -Luc spanning the promoter sequences between Ϫ2846 and ϩ8 from ChIP samples and cloned into heterologous synthetic TK promoter (without the CREB-binding site) containing the luciferase reporter. As control, pGL 4.14 -GRE 3ϫ -TK-luciferase was used to measure ligand response. All three promoter elements containing potential nGREs showed a similar down-regulation of transcription that was dose-dependent.
Inhibition of CREB Dephosphorylation by OKA Increases Transcriptional Activity-DEX-mediated suppression was persistent in all four reporters as follows: Ϫ1523/Ϫ971-, Ϫ1806/ Ϫ1443-, and Ϫ1806/Ϫ971-TK-luciferase and pGL 4.14 Ϫ2846/ ϩ8-luciferase when compared with the vehicle-treated controls indicating the negative role of the nGRE elements in this region (Fig. 6B). Bt 2 cAMP-stimulated transcription from GR promoter was enhanced with all promoters under study.
But when cells were treated with DEX and Bt 2 cAMP, luciferase activity declined 2-fold. In cells treated with OKA, Bt 2 cAMPstimulated transcription from the GR gene promoter was enhanced 2-2.5 times when compared with Bt 2 cAMP alone. These results lead us to conclude that the phosphorylation and dephosphorylation of CREB defines the recruitment, and this is one of the major events regulating transcription and GR synthesis. Even though the addition of DEX reduced the OKAinduced levels, simultaneous treatment of the transfectants with cAMP resulted in the highest levels of luciferase activity. This again confirmed that CRE-bound CREB was maintained in its active form, CREB-P, by OKA inhibition of a specific phosphatase recruited by the GR⅐DEX complex. This results in a remodeling of chromatin, which permits the regulatory factors to access DNA elements and to stimulate transcription. Furthermore, the down-regulation role of nGRE with DEX was relieved in the presence of TSA, and luciferase activity increased severalfold. This indicates that HDAC-6 alone is not recruited to these promoter sites and that GR is necessary for recruiting HDAC-6 and for TSA inhibition to work.

HDAC-6 and CREB-P Form Complexes with GR in Vivo-
To investigate whether ligand-bound GR recruited HDAC-6 and CREB-P, a co-immunoprecipitation assay was performed using HeLa cells transfected with GR, CREB, and HDAC-6 expression vectors either alone or in combinations and labeled with [ 35 S]methionine (Fig. 7). Control actin was used to standardize protein levels. GR polyclonal IgGs coimmunoprecipitated with HDAC-6 ( Fig. 7, left) and CREB-P (right) complexes only in the presence of DEX. These experiments support the hypothesis that both HDAC-6 and CREB-P form complexes with GR in vivo and that the complex formation is ligand-dependent.
nGRE at Ϫ1485 Determines Feedback Regulation-To elucidate the mechanism of feedback regulation defined by the 5Ј and 3Ј deletion mutagenesis, the FBR motif between the GRE-1485 and CRE-1461 was considered for detailed mutational analysis ( Fig. 8A and supplemental Table 2). Two nGRE mutants (M1 and M2) and three CRE mutants (M3-M5) were generated, and reporter vectors were constructed containing authentic GR transcription initiation sites (supplemental Table  3). Transfected cells were treated with various ligands (Fig. 8B), and with mutations of the nGRE site in either M1 or M2, reporter activity was unchanged in the presence of DEX when compared with control, confirming that the inhibitory effect of DEX was abolished in these nGRE mutants. cAMP stimulated similar levels of reporter activity with Ϫ1528/ϩ8 and both M1 and M2, but inhibition by simultaneous treatment with DEX was eliminated. We attribute this to the lack of GR recruitment with DEX by these mutants. When deacetylases were inhibited with TSA, reporter activity of wild type, M1, and M2 decreased, consistent with earlier results. TSA ϩ DEX treatment increased transcriptional activity in the wild type reporter, but had no effect on M1 and M2 due to absence of GR recruitment. Any change in the CRE sites as in mutants M3, M4, and M5 abolished not only stimulation with cAMP but also DEX-mediated effects. However, the TSA ϩ DEX generated a 6-fold increase in luciferase activity suggesting that ligand-bound GR determines the state of chromatin remodeling. These results indicate that GR⅐DEX complexes play a role in CREB-P dephosphorylation and recruitment of deacetylases.

Knockdown of HDAC-6 Results in Augmented GR Expression, and CREB Knockdown Shuts Down the GR Gene Response
Completely-Dose response to DEX was measured in stable HDAC-1-6 T47D cells containing endogenous GR (Fig. 9A).
The oligonucleotides used for the construction of silencer vectors are shown in supplemental Table 4. Knocking down of HDAC-1, -3, -4, and -5 did not affect dose response when compared with control. Identical data were obtained with the second set of knockdown sequences (data not shown). However, knocking down of HDAC-2 and HDAC-6 abolished the GR-

. Dose response to feedback regulation in transfected T47D cells and effect of TSA on GR gene feedback regulation by GR⅐DEX complexes.
A, T47D cells were co-transfected with pGL 4.14 -GRE 3ϫ -TK-luciferase, pGL 4.14 Ϫ1523/Ϫ971-TK-luciferase, pGL 4.14 Ϫ1806/Ϫ971-TK-luciferase, and pGL 4.14 Ϫ2786/ϩ8-luciferase in the presence of pcDNA3.1-␤-galactosidase. The transfectants were treated with 0.01-100 nM DEX as indicated for 12 h, and whole cell extracts were prepared. Luciferase activity was determined using extracts containing 10 units of ␤-galactosidase activity. The results are from three independent dose-response experiments. B, T47D cells were cotransfected with pGL 4.14 -GRE 3ϫ -TK-luciferase, pGL 4.14 Ϫ1523/Ϫ971-TK-luciferase, pGL 4.14 Ϫ1806/Ϫ971-TK-luciferase, and pGL 4.14 Ϫ2786/ϩ8-luciferase in the presence of pcDNA3.1-␤-galactosidase. The transfectants were treated with DEX, Bt 2 cAMP, Bt 2 cAMP ϩ DEX, OKA, OKA ϩ DEX, OKA ϩ Bt 2 cAMP, TSA, and TSA ϩ DEX as indicated. mediated transcriptional activation. As previous experiments showed the presence of HDAC-6 in the precipitated chromatin with CREB-P and GR (Fig. 5B), the next analysis was performed using knockdown HDAC-2 as a control to compare the effects of HDAC-6 knockdown (Fig. 9B) and pGL 4.14 Ϫ1523/ϩ8-luciferase reporter. All three, namely control T47D cells, control vector selected T47D cells, and HDAC-2 knockdown T47D cells, showed similar response to treatments with various ligands. Only HDAC-6 knockdown showed variations. Although induction with cAMP was similar to controls, the addition of DEX failed to reduce response. Treatment with OKA, a phosphatase inhibitor, did not increase the level of induction compared with controls; however, OKA in the presence of cAMP enhanced transcription only to the same levels as cAMP alone indicating the absence of activity due to OKA inhibition. TSA treatment did not show any response as well as TSA in the presence of DEX, indicating that the GR⅐DEX complex recruits HDAC-6 and not HDAC-2. We also obtained similar results with HDAC-6 knockdown b T47D cells (data not shown). Similar experiments were performed to analyze the knockdown effects of CREB (Fig. 9C). Reporter activity was completely abolished in both stable CREB knockdown A and B. These data suggest that CREB expression is crucial for the regulation of the GR gene, and that HDAC-6-mediated chromatin remodeling is uncoupled from the CREB response.
Rescue of GR Expression in Knockdown Cells by Western Blotting-Actin served as an internal control and as a measure to ensure that equimolar quantities of extracts and RNA were used. To investigate whether the effects measured in reporter assays were indeed due to the silencing effects of both CREB and HDAC-6, we performed Western blotting (Fig. 10A, left,  top panel) and RT-PCR (Fig. 10A, left, lower panel). Extracts were prepared from T47D-CREB and T47D-HDAC-6 knockdown cells. The CREB knockdown A and B as well as HDAC knockdown a and b were performed with two different silencing sequences (supplemental Table 4). Nontransfected control T47D and silencing vector without any inserts served as controls (Fig. 10A, left, lanes 1 and 2). CREB expression was completely abolished in CREB knockdown A and CREB knockdown B (Fig. 10A, left, top panel, lanes 3 and 4). Transfection of pCMV-CREB rescued the expression of CREB at higher levels than controls in both CREB knockdowns (Fig. 10A, left, lanes 5 and 6). This is clearly visible in the RT-PCR analysis (supplemental Table 5 and Fig. 10A left, lower panel). Knockdown was 75-80% compared with controls, and in reinstated cells CREB levels were elevated (Fig. 10A,  left, lanes 3-6). It was also evident in the same RT-PCR that GR mRNA levels decreased with CREB mRNA levels (Fig. 10A, left, lanes 3 and 4). Western blot analysis was performed (Fig. 10A, right) using 20 g of whole cell extracts from T47D, and GR expression was analyzed by probing the Western blot with GR IgGs. The results show that GR expression was knocked out in CREB knockdown A and CREB knockdown B (Fig. 10A, right,  lanes 3 and 4). Rescue of CREB expression by transfection with pCMV-CREB (Fig. 10A, right, lanes 5 and 6) resulted in GR protein levels similar to controls (lanes 1 and 2). Neither the transfection with CREB-133 (Fig. 10A, right, lane 7), a mutant unable to initiate transcription (32), nor with KCREB (lane 8), a second mutant that forms inactive dimers with endogenous CREB and blocks its ability to bind to CRE (33), showed such a response indicating the importance of intact CREB machinery in GR gene regulation. Similar experiments were performed using the T47D-HDAC-6 knockdown cells (Fig. 10B, left, top). The knockdown effects clearly demonstrate that GR-mRNA (Fig. 10B left, lower panel, lanes 3 and 4) expression in these cells was enhanced, and transfection of pCMV-HDAC-6 inhibited GR mRNA expression (lanes 5 and 6). We analyzed by Western blotting if these results correlated to the GR protein levels (Fig. 10B, right). Control T47D (Fig. 10B, right, lane 1) and control vector selected T47D (lane 2) showed similar levels of GR expression. The level of GR expression was elevated in the HDAC-6 knockdown cells (Fig. 10B, right, lanes 3 and 4). Rescue of HDAC-6 by transfection of pCMV-HDAC-6 decreased GR expression in the knockdown T47D cells (Fig. 10B, right,  lanes 5 and 6). HDAC-2, used as control (Fig. 10B, right, lane 7), did not affect GR mRNA level at all. This HDAC-6 knockdown experiment provides a direct link between chromatin remodeling through histone deacetylation and GR⅐DEX complexes in the regulation of the GR gene.
HDAC-6 and CREB Bind Directly to GR⅐DEX Complexes in Vivo-HDAC-6 has four LXXLL signature motifs that are important in the assembly of nuclear receptor⅐co-activator complexes. Serine residues in the GR AF-1 domain play an important role in transcription regulation (13). We generated HDAC-6 deletion mutants to assess the importance of the four LXXLL motifs in HDAC-6 (Fig. 11A). As we have shown in co-immunoprecipitation analyses that GR and HDAC-6 interact with one another (Fig. 7), we used GR serine substitution mutants (supplemental Table 6) in mammalian two-hybrid assays to map domains of the interaction between GR and HDAC-6. Fig. 11B shows that not only is an intact GR AF-1 domain essential for HDAC-6 binding to GR⅐DEX complexes HeLa cells were transiently transfected with pcDNA 1 GR alone or in the presence of pcDNA3.1-His 6 HDAC-6 or pCMV-CREB. Cell extracts were prepared and processed as described under "Experimental Procedures." Control actin IgG was used for standardization. Left panel, GR IgG and HDAC-6 precipitation in the absence and presence of DEX. HDAC-6 was identified with HDAC-6 His 6 IgG. Right panel, precipitation of GR-CREB-P by GR IgG. CREB-P was identified with CREB-P IgG in independent assays.  Table 3) are shown in boldface letters in the boxes. B, functional analysis of two FBR mutants, Mut 1 and Mut 2, and CRE mutants Mut 2, Mut 4, and Mut 5 and feedback regulation were analyzed by transfection into HeLa-GR cells followed by ligand treatment as indicated.
but that serine residues Ser-141, Ser-203, Ser-211, and Ser-226 are important in establishing contact with HDAC-6. Next we determined the importance of the four LXXLL HDAC-6 motifs. Reporter activity was determined with various constructs containing the LXXLL motifs in the presence and absence of ligand. Our results demonstrate that the motif between amino acids 313 and 418 in HDAC-6 made contact with the GR⅐DEX complexes.
Similar two-hybrid assays were performed with CREB to delineate residues involved in contact (Fig.  11C). We used CREB-133, defective in phosphorylation, to determine whether the phosphorylated form of CREB associated with GR. As shown in Fig. 11C, CREB contact with GR AF-1 domain amino acids Ser-211 and Ser-226 followed the same pattern as HDAC-6. CREB substitution mutant CREB-133 did not make contact with GR wild type in the presence of DEX. This further indicated that it was indeed phosphorylated CREB-P contacting GR wild type.
HDAC-6 and CREB-P Bind GR-DEX in Vitro-To delineate GR domains that interacted with HDAC-6 and CREB-P, a set of GR wild type and GR AF-1 (S113A, S141A, S203A, S211A, and S226A) were generated as GST fusion proteins (supplemental Table 5), and GST pulldown experiments were performed. Fig. 12A shows the purified GST fusion proteins. Interaction of both CREB (Fig. 12B) and HDAC-6 ( Fig. 12C) with GR was ligand-dependent. Mutating serine at position 113 did not affect CREB association with GR (Fig. 12C, lane  5). However, S203A and S211A were unable to interact with CREB (Fig. 12C, lanes 6 and 7). GR wild type made contact with HDAC-6 only in the presence of DEX (Fig.  12C, lane 5). Binding of the AF-1 mutants GST-GR S113A (Fig. 12C,  lane 6) and GST-GR S141A (lane 7) to HDAC-6 was weaker than with GST-GR wild type in the presence of hormone (lane 4). HDAC-6 associated weaker with GR S203A (Fig. 12C, lane 8), GR S211A (lane 9), and GR S226A (lane 10). These studies confirm that intact AF-1 domain, DNA binding, and hormone binding domains of the GR are essential for this FBR mechanism.
Analysis of CREB-P Regulation by DNA-Protein Interaction-To demonstrate that CREB-P was involved in the regulation of the GR gene through contact with GR⅐DEX complex, 32 P-labeled FBR or mutants (M1-5) were incubated with cell extracts containing GR wild type, GR AF-1 mutants, CREB-P, CREB-133, or KCREB (Fig.  13). When labeled wild type FBR was used in the presence of DEX (Fig. 13A), specific bands were observed with GR wild type (lane 4) and with GR AF-1 S113A and S141A (lanes 6 and 7). The disappearance of the band with wild type GR in the presence of 100 nM DEX indicates that 100-fold molar excess of nonlabeled FBR (Fig. 13A, lane 5) competed for GR binding. The ligand specificity is shown in Fig.  13A, lanes 3 and 11. The wild type FBR unit bound to GR AF-1 mutant S113A (Fig. 13A, lane 6) and GR S141A (lane 7), but mutants GR S203A (Fig. 13A, lane 8), GR S211A (lane 9), and GR S226A (lane 10) failed to recognize this motif in the presence of DEX. GR binding to the FBR motif used in transcriptional activation studies (Fig. 8, A and B) was analyzed using FBR point mutants M1 and M2 (Fig.  13A, lanes 14 and 15) and CRE mutants M3, M4, and M5 (lanes 16 -18). The 32 P-labeled M1 and M2 failed to bind GR in the presence of DEX, and the CRE M3 (Fig.  13A, lane 16), M4 (lane 17), and M5 (lane 18) bound the GR⅐DEX complexes, where M4 intensity was higher than with either M3 or M5.
Supershift analyses with antibodies against GR and CREB-P determined the identity of proteins in the DNA complexes FIGURE 9. Knockdown of HDAC-1-6 in T47D cells and dose response to GR transactivation. A, stable HDAC-1-6 knockdown T47D cells were established, whole cell extracts prepared, and luciferase activity determined as described under "Experimental Procedures." B, stable knockdown of T47D cells using vector without any silencing sequences, HDAC-2A, and HDAC-6a silencer sequences were established by G418 selection. Control T47D and the selected cells were transiently cotransfected with pGL 4.14 Ϫ1523/ϩ8-luciferase and pcDNA3.1-␤-galactosidase. Transfectants were treated with ligands as indicated. C, CREB knockdown A and B in T47D cells and feedback regulation.  (Fig. 13C). The GR-specific interaction is shown in Fig. 13C,  lanes 7 and 8, where the GR⅐IgG retarded complex disappeared by incubation with 100-fold molar excess of nonlabeled FBR element (lane 8). CREB⅐FBR complexes were supershifted by CREB-P IgG (Fig. 13C, lanes 9 and 13) and disappeared after treatment with competitor FBR (lane 14). No supershifted FIGURE 11. In vivo interaction of HDAC-6 with GR. A, structure of HDAC-6 is shown for reference. The location and position of nuclear receptor recognition motifs (LXXLL) are indicated. The deletion mutants used for the construction of pVP16 fusion constructs are shown. B, mammalian two-hybrid assay of HDAC-6 interaction with pM-GR as described under "Experimental Procedures." C, mammalian two-hybrid assay of CREB interaction with pM-GR. The hGR AF-1 serine mutants S113A, S141A, S203A, S211A, and S226A were cloned into pM, and CREB derivatives were cloned into pVP16 vectors. The reporter vector pG5-TKluciferase was co-transfected with pcDNA3.1-␤-galactosidase, and 10 units of ␤-galactosidase activity were used to determine luciferase activity. FIGURE 12. Production and purification of GST-GR. A, GST-GR WT, GST-GR S113A, GST-GR S141A, GST-GR S203A, GST-GR S211A, and GST-GR S223A fusion proteins were purified from isopropyl 1-thio-␤-D-galactopyranoside-induced cultures using glutathione magnetic beads, and GST pulldown assays were performed as described under "Experimental Procedures." GST was used as negative control. B, in vitro interaction of CREB with GST-GR WT and AF-1 mutants. C, in vitro interaction of HDAC-6 with GST-GR WT and AF-1 mutants. GST was used as negative control.
CREB Phosphorylation and HDAC-6 Inhibition Increase GR mRNA Synthesis-To examine if the above results had any consequences on the synthesis and regulation of GR mRNA transcription, T47D cells were treated with ligands as indicated. and total protein extracts and total RNA were prepared. Using actin as control, equal amounts of protein samples were resolved by electrophoresis and blotted onto Immobilon-P membranes. The blots were probed using antibodies against GR, CREB-P, and actin. Synthesis of GR was enhanced by Br-cAMP treatment when compared with DEX (Fig. 14, upper panel, lanes 2  and 3). Simultaneous treatment with Br-cAMP and DEX decreased GR levels considerably to levels observed with nontreated cells (Fig. 14, upper panel, lane 1). Inhibition of CREB-P dephosphorylation by OKA increased GR protein and mRNA levels (Fig. 14, upper panel, lane 5) significantly suggesting that recruitment of GR-DEX blocked CREB-P binding to the FBR element through a direct dephosphorylation mechanism. The highest level of GR protein was observed when cells were treated with OKA and Br-cAMP (Fig. 14, upper panel, lane 7), correlating well with the enhanced levels of GR mRNA observed by RT-PCR (Fig. 14, lower panel, lane  7). TSA alone had little effect, but TSA with DEX enhanced the GR protein and mRNA level indicating that GR⅐DEX-recruited HDAC-6 complexes were inhib-ited, enabling histone acetylation and thus enhancing GR synthesis. Fig. 15 illustrates this mechanism.

DISCUSSION
A typical promoter contains regulatory elements to which a number of factors bind. These factors not only regulate transcription and coordinate with the transcriptional machinery but also maintain positive and negative control of gene transcription (8, 34 -37). GREs targeted by the ligandbound GR can act either as enhancers or suppressors, nGREs, of gene regulation. Numerous genes contain such elements recognized by receptor as well as nonreceptor factors. Glucocorticoids increase transcription of the phosphoenolpyruvate carboxykinase gene (38,39) and require both GRE and CRE (15,40) to do so. Although acting as suppressor sequences, nGREs are involved in feedback mechanisms in several stress-response systems. GREs can be located as far upstream as 2.5 kb from the transcription start site (41) or as close as 395 bp and can span anywhere from 25 to 110 bp (42). The CRE in the CRH promoter is 18 bp long, and positive regulation and glucocorticoid repression are located between Ϫ232 and Ϫ215 bp (43). In this study, we examined a 3.2-kb region between Ϫ1806 and Ϫ971 from the Cap site containing two potential nGRE (Ϫ1795, Ϫ1485) and two potential CRE (Ϫ1460, Ϫ1073) sites and their role in the regulation of GR gene expression in the presence of glucocorticoids.
Our goal was to define the minimal DNA sequence required for glucocorticoid inhibition of the human GR gene. We report here that the functional response of GR gene transcription regulation is located between Ϫ1523 and Ϫ971, involving both nGREs and CREs in a manner reminiscent of the AVP gene where multiple CREs (9) and glucocorticoids (12) regulate expression in the presence of CREB. As reported in an earlier study, we showed the presence of the FBR element within the Ϫ2846 upstream of the GR Cap site of the GR gene promoter (18). Our results show that this FBR motif is crucial to the negative regulation of the GR gene by glucocorticoids, and thus the FIGURE 13. Interaction of GR and AF-1 serine mutants with feedback regulatory element. A, labeled FBR element was incubated with whole cell extracts containing GR and GR AF-1 mutants expressed in HeLa cells as described under "Experimental Procedures." Control extracts (lanes 1 and 2), GR WT with FBR in the absence (lanes 3 and 11) and presence of DEX (lanes 4 and 12), with competitor ( lanes 5 and 13), GR AF-1 serine mutants S113A (lane   Upper panel, in the cortisol-depleted state, phosphorylated CREB binds to the CRE in the composite feedback regulatory element and CBP⅐p/CAF complexes. This interaction results in the acetylation of core histones and chromatin decondensation, which turns on the GR gene. Lower panel, in the presence of active glucocorticoids, nGRE recruits GR⅐ligand complexes that tether HDAC-6 and phosphatase complexes. The GR-DEX-bound HDAC-6 deacetylates core histones, and the GR-DEX-tethered phosphatase dephosphorylates CREB-P. Deacetylation leads to chromatin condensation and dephosphorylation of CREB-P, which results in its removal from the CRE which turns off the GR gene. separation of two independent responsive elements (44) for two classes of transcription factors allows interactions to take place without mutual hindrance.
As gene transcriptional activation and inhibition implicates accessibility of DNA sequences and as eukaryotic chromosomes are compacted together in the form of nucleosomes, alteration of chromatin structure is an important step in gene activation. Acetylation of lysines in target histones, one of the signs of actively transcribed chromatin, and CREB-P recruitment of CBP histone acetyltransferase activity, which relaxes chromatin, indicate a link between GR, CREB, and HDAC-6. We used stable HeLa-GR cells and T47D cells that express endogenous GR to establish which of the GR gene promoter elements was targeted by the ligandbound GR. Although CREB IgG precipitated CREB from all samples, DEX reduced CREB intensity. Both GR and CREB are DNA-binding proteins, and Western blot analyses with anti-CREB-P demonstrated that GR-DEX bound to the nGRE decreased level of CREB and inhibited cAMP-stimulated CREB-induced transcriptional activation. This is in contrast to the thyroid receptor where association with CREB is increased in the presence of triiodothyronine (45,46). The GR is not only a phosphoprotein (47,48) but, when activated, affects phosphorylation status of transcription co activators such as CREB. This would indicate that CREB dephosphorylation is catalyzed by a nuclear protein phosphatase, a process that is dependent on DEX (49 -52). Several candidate phosphatases, in particular PP2A, are targets of OKA, which leave the CREB-P bound to the CREs (31,51). Higher levels of cAMP as used in the experiments here abrogate the function of PP2A, allowing an indeterminate quantity of CREB-P bound to the FBR element and GR transcription intact even in the presence of recruited GR⅐DEX complexes. Among steroid hormone receptors, transcription is regulated through a multitude of pathways. Phosphorylation of the estrogen receptor itself affects its interaction with the CREB-binding protein, CBP, even in the absence of ligand (53) and activates transcription through its histone acetyltransferase activity (54, 55). The phosphorylated form of CREB interacts with CBP and is a co-activator of transcription of cAMP-responsive genes (56,57) such as AVP, somatostatin, and c-fos among others (58 -60). The level of CREB-P and GR decreased with DEX and increased with either Br-cAMP or OKA alone or in combination. This explains the highest level of transactivation observed with FBR motifs. Data presented here show that the phosphorylation status of CREB defines its recruitment by the GR⅐DEX complex and that this is one of the major events regulating transcription and GR synthesis. In our study, interaction of CREB with GR was dependent on ligand, whereas binding of CREB to thyroid receptor was not (45). The GR AF-1 region is involved in contacts with various factors that regulate transcription. Specifically, our data show that GR Ser-203, GR Ser-211, and GR Ser-226 are crucial for CREB and HDAC-6 interaction.
Nuclear factor-1 inhibits cAMP-induced transcription from the phosphoenolpyruvate carboxykinase gene by binding to block CREB access to CBP (61). The PCR-amplified GR gene promoter elements between Ϫ1523 and Ϫ971 we used in this study include the same nGRE Ϫ1485 -cAMP Ϫ1461 regulatory motifs separated by 25 bp but exclude the putative nuclear factor-1 Ϫ1530 recognition motif. Furthermore, we show ligand dependence in transfection assays with this region of the GR gene. We propose that the stimulation of GR gene transcription is due to the inhibition of HDAC-6 catalytic activity recruited by GR-DEX. As the nuclear factor-1 site at Ϫ1530 is not part of the region we investigated, it also rules out the direct involvement of nuclear factor-1 in the down-regulation of the GR gene.
The mechanism of deacetylases in the transcriptional repression of target genes has only just moved into the spotlight, although it has been known for a while that glucocorticoids repress inflammatory genes such as cytokines (62) and are effective therapy for asthma (63). Sequence homology and subcellular localization differentiates three classes of HDACs. Numerous reports document class I and class II HDACs and most focus on class I HDAC. Class I HDAC-1-3 and HDAC-8 are ubiquitously expressed and are strictly nuclear (64). Class II HDAC-4 -7, HDAC-9, and HDAC-10 are tissue-specific, cytoplasmic, and nuclear (65). GR inhibits granulocytemacrophage colony-stimulating factor expression by recruiting HDAC-2, which inhibits acetylation of H4K12 by CBP-associated histone acetyltransferase activity (66). In interleukin-1␤-induced gene expression, the activated GR complex inhibits histone acetylation and recruits HDAC-2 (66). Thyroid receptor recruits CBP to trigger histone acetylation (46). HDAC-3 forms a complex with SMRT⅐NCoR (67,68). Not much is known about the functional role of HDAC-6. What is clear though is that HDAC-6 forms complexes with other proteins and catalyzes deacetylation of substrates that are involved in neurodegenerative disorders, cell migration associated with actin, and tumor suppression.
We confirmed a direct interaction between HDAC-6 and GR with ligand-bound GR by co-immunoprecipitation assays. The inhibitory role of FBR Ϫ1523/ϩ8-bound GR-DEX was relieved by TSA confirming that activated GR recruits HDAC-6. We examined the role of HDAC-1-6 in GR-mediated negative regulation and found that although knocking down HDAC-1, HDAC-3, HDAC-4, and HDAC-5 had no effect on transcriptional activation from wild type GR gene promoter, knocking down HDAC-2 or HDAC-6 abolished GR-DEX induction. By further expanding our investigation to include DEX, cAMP, and TSA, we found significant differences that confirmed that it was indeed HDAC-6 and not HDAC-2 that interacted with the GR⅐DEX complex. Of the HDACs examined, only HDAC-6 has the nuclear receptor signature motif LXXLL. Our deletion mutation analyses provide further confirmation that the LXXLL 313-418 motif in HDAC-6 made contact with the GR⅐DEX complex. We also provide clear evidence for its role in the regulation of the GR gene by glucocorticoids.
As CREB protein decreased and HDAC-6 increased with glucocorticoid treatment, silencing HDAC-6 would disrupt feedback regulation of GR by glucocorticoids. This is further substantiated by the fact that acetylated H3 and H4 proteins are present in immunoblots in TSA-treated T47D cells but are absent in DEX-treated cells. Increased HDAC expression results in the removal of acetyl groups from histones causing compacting of DNA and thereby inhibiting expression of genes that include tumor suppressor genes. Therefore, knowledge of the HDAC-6 mechanism would provide a deeper understanding behind the molecular mechanisms of the evolution of cancer. Fig. 14 summarizes the cellular context of the negative regulation of the GR gene by GR⅐DEX complexes, where DEX decreases GR mRNA, protein, and CREB phosphorylation suggesting a direct mechanism dependent on the recruitment of a specific cellular phosphatase by the glucocorticoid receptor ligand complexes.