Glucocorticoid Ligands Specify Different Interactions with NF-κB by Allosteric Effects on the Glucocorticoid Receptor DNA Binding Domain*

Glucocorticoids inhibit inflammation by acting through the glucocorticoid receptor (GR) and powerfully repressing NF-κB function. Ligand binding to the C-terminal of GR promotes the nuclear translocation of the receptor and binding to NF-κB through the GR DNA binding domain. We sought how ligand recognition influences the interaction between NF-κB and GR. Both dexamethasone (agonist) and RU486 (antagonist) promote efficient nuclear translocation, and we show occupancy of the same intranuclear compartment as NF-κB with both ligands. However, unlike dexamethasone, RU486 had negligible activity to inhibit NF-κB transactivation. This failure may stem from altered co-factor recruitment or altered interaction with NF-κB. Using both glutathione S-transferase pull-down and bioluminescence resonance energy transfer approaches, we identified a major glucocorticoid ligand effect on interaction between the GR and the p65 component of NF-κB, with RU486 inhibiting recruitment compared with dexamethasone. Using the bioluminescence resonance energy transfer assay, we found that RU486 efficiently recruited NCoR to the GR, unlike dexamethasone, which recruited SRC1. Therefore, RU486 promotes differential protein recruitment to both the C-terminal and DNA binding domain of the receptor. Importantly, using chromatin immunoprecipitation, we show that impaired interaction between GR and p65 with RU486 leads to reduced recruitment of the GR to the NF-κB-responsive region of the interleukin-8 promoter, again in contrast to dexamethasone that significantly increased GR binding. We demonstrate that ligand-induced conformation of the GR C-terminal has profound effects on the functional surface generated by the DNA binding domain of the GR. This has implications for understanding ligand-dependent interdomain communication.

gene expression. The anti-inflammatory activities of GCs are caused, in part, by transrepression of the proinflammatory transcription factor NF-B.
The GR has distinct functional domains, an N-terminal transactivation domain (AF-1), a central DNA binding domain (DBD), and a C-terminal domain that includes ligand binding (LBD) and transactivation activities (AF-2) (1-3). However, although the domains can function separately, there is evidence for functional interdependence within the intact GR, which may be necessary for authentic activity (4). For example, the conformation of the DBD, induced by either DNA sequence or protein binding, can determine whether transcription is induced or repressed (5).
The chemical structure of glucocorticoid ligands alter the function of the receptor as they induce conformational changes to the LBD that specify whether co-activators or co-repressors are recruited (6,7). For example, both the GR agonist dexamethasone and the antagonist RU486 promote nuclear translocation and DNA-binding of the GR (8 -11); however, when RU486 is bound, the position of helix 12, as shown by the recent crystal structure, is altered and partially blocks the co-activator pocket (7). This leads to the recruitment of co-repressors, such as NCoR, and not co-activators, such as SRC1, which are recruited to the agonist bound GR (12,13). However, the effects of these ligand-induced conformational changes on the function of the DNA binding domain have not been fully explored.
The mechanism of NF-B repression by GR has yet to be fully defined, but it is known that GR and the p65 subunit of NF-B physically interact (14 -17) with the GR DBD binding to the Rel homology domain (RHD) of p65 (18). This interaction occurs on the NF-B response element resulting in "tethering" of the GR to DNA. GR repression of p65 requires the LBD, in addition to the DBD (19). A repressor protein or complex, recruited to the tethered GR, has been proposed, with evidence implicating GRIP-1, a member of the p160 family (18,20,21). However, it is not clear if functional differences between agonist and antagonist are due to defective interaction with NF-B or subsequent defective recruitment of the co-repressor protein.
The function of the GR DBD has been thought to be independent of the LBD, but RU486 causes subtle alteration of DNA binding kinetics for the GR compared with dexamethasone (22).
In this paper, we explore the influence of ligand on functional interactions between the ligand binding domain and DNA binding domain of the GR. Improved understanding of how GR interaction with p65 occurs and how this can be modulated by ligand has major implications for effective drug design and screening. Here we describe a previously unsuspected role for the GR LBD in regulating recruitment of NF-B to the GR DNA binding domain.
For the construct encoding the p65:GFP 2 fusion protein, the enzymes used were HindIII and BamHI, and the recipient vector was pGFP 2 1-C1(FG)Zeo (Biosignal Packard). Similarly, the construct encoding the GFP 2 :p65 fusion protein was obtained by using XhoI and HindIII, and the recipient vector was pGFP 2 1-N2(FG)Zeo (Biosignal Packard).
The pcDNA3-GR plasmid containing full-length wild type GR has been previously described (12). The human GR DNA binding domain/ligand binding domain (DBD/LBD) between codons 289 and 777 was amplified from pcDNA3-GR using the HF2 PCR kit and inserted into the pACT plasmid (Promega Corp. Southampton, UK) to give VP16-DBD/LBD. The specific primers had BamHI and NotI restriction enzyme sites added for ligation into the vector (forward primer for VP16-DBD/LBD, 5Ј-GGATCCCTGGGGTAATTAAGCAAGAGAAACTGG-3Ј; reverse primer, 5Ј-GCGGCCGCCTTTTGATGAAACAGAAGTTTTTTG-3Ј). The CMV-Renilla plasmid (pRL-CMV) was obtained from Promega and has been previously described (12).
All plasmid constructs were sequenced to confirm the presence of the predicted changes and to exclude introduction of errors.
Interaction of p65-RHD and the Glucocorticoid Receptor in Vitro-Glutathione S-transferase (GST) and GST-RHD fusion proteins were expressed in Escherichia coli strain DH5␣ (Invitrogen) and were purified as described (29). Briefly, a 50-ml culture of the expression vector was stimulated with 0.5 mM isopropyl-1-thio-␤-D-galactopyranoside (Sigma) and grown for 3 h. The bacteria were then pelleted and resuspended in NETN buffer (20 mM Tris, pH 8.0, 100 mM NaCl, 1 mM EDTA, 0.01% Nonidet P-40) supplemented with protease inhibitors ("Complete"; Roche Applied Science). The bacteria were treated with lysozyme (100 g/ml) for 15 min at 4°C and then disrupted using a 50-watt, 20-kHz sonicator (Jencons, Leighton Buzzard, UK) on full power for 5 ϫ 30 s. The samples were then centrifuged to remove the debris, and glutathione-Sepharose beads (Amersham Biosciences) were added to the supernatant. The fusion proteins were allowed to bind to the beads for 1 h at 4°C. The beads were then washed three times in NETN plus protease inhibitors and stored at 4°C. 35 S-Labeled GR was synthesized using the TNT kit (Promega) following the manufacturer instructions with the full-length, human GR cDNA in pcDNA3 or VP16-DBD/LBD as template, with no ligand, 10 M Dex, or 10 M RU486 and [ 35 S]methionine (Amersham Biosciences). Labeled GR was incubated with the GST proteins bound to glutathione-Sepharose beads for 1 h at 4°C. The beads were then washed three times in NETN, and boiling in 2ϫ SDS-PAGE sample buffer eluted the bound proteins. Proteins were analyzed by SDS-PAGE, and the gel was stained in Sypro Red to visualize the GST fusion proteins. The 35 Slabeled GR was then visualized by exposure of the gel to a phosphor imaging plate (BasIII; Fuji).
The Sypro Red-stained acrylamide gels were analyzed under UV light (AlphaImager 2000; Flowgen, Lichfield, UK). The amount of radiolabeled protein present on the gel was quantitated using a phosphor imager (BAS1800; FujiFilm) and Aida 2.0 analysis software (Raytest, GMBH).
Transrepression Assay-COS7 cells were seeded at 5 ϫ 10 4 cells into 24-well plates and were transfected with 2.4 g of NRE-luc and 1.2 g of GR construct (pcDNA3-GR or VP16-DBD/LBD) and 7.5 ng of GAL4-p65. Cell lysates were subjected to both firefly and Renilla luciferase assays; firefly luciferase was then corrected for Renilla luciferase expression as previously described (12).
On the day of transfection, the medium was replaced by growth medium supplemented with 10% charcoal/dextran-treated fetal bovine serum (Hyclone), and cells were transfected using LipofectAMINE 2000 TM (Invitrogen) according to the manufacturer's instructions. Cells were transfected with GR fused to Renilla luciferase at the C or N terminus and p65 fused to GFP 2 at the C or N terminus using RenLuc-GFP 2 at different ratios (4:1, 4:4, 1:4, and 0.5:4). As a negative control for interaction, GR-RenLuc was also cotransfected with GFP 2 -KaiB plasmid.
48 h post-transfection, cells were harvested using Versene (PBS and 0.5 mM EDTA); centrifuged for 5 min at 1200 rpm, resuspended in BRET buffer, Dulbecco's PBS (Invitrogen) supplemented with 2 g/ml aprotinin (Sigma, Poole, UK); and counted. Cells were recentrifuged and resuspended in BRET buffer to a cell density of 1.5 ϫ 10 6 cells/ml. 5 ϫ 10 4 cells were added to each well of a 96-well shallow Optiplate (PerkinElmer BioSignal Inc.) followed by stimuli diluted in BRET buffer so that the final concentrations were 1 M dexamethasone and 10 ng/ml TNF-␣. The treatment was performed at room temperature for 1 h. Triplicate determinations for each condition were prepared.
After incubation, the luminescence reaction was started by adding 10 l of the substrate DeepBlueC TM coelenterazine (PerkinElmer Bio-Signal), to give a final concentration of 5 M; plates were read after 20 min on a Fusion Universal Microplate Analyzer (PerkinElmer BioSignal) using the following filter settings: 410 nm for luciferase emission and 515 nm for enhanced yellow fluorescent protein (EYFP) emission.
After subtraction of the average counts obtained with untransfected cells from the readings of individual wells, the BRET 2 signal was determined as the ratio between GFP 2 emission (515 nm) and Renilla luciferase emission (410 nm).
BRET 1 Assay-The day before transfection, HEK 293T cells were seeded at 2 ϫ 10 6 cells per 10-cm tissue culture dish containing DMEM, 2 mM glutamine, and 10% fetal bovine serum. Cells were transfected using Fugene 6 transfection reagent following the manufacturer's instructions (Roche Applied Science). The medium on the cells was replaced with 10 ml of Opti-MEM I reduced serum medium (Invitrogen). The transfection solution containing a 1:5 DNA/Fugene ratio, using 150 l of Opti-MEM I reduced serum medium, 30 l of Fugene, and 6 g of total DNA, was then added on the cells after 20 min at room temperature. Cells were cotransfected with C-terminal tagged GR-Renilla and the other protein encoding cDNA fused to EYFP (p65, SRC1, or NCoR) at a ratio of 1:4, 1:4, and 1:8, respectively. After 4 h, medium was replaced by DMEM, 2 mM glutamine, and 10% charcoal/dextran-treated fetal bovine serum (Hyclone). 48 h post-transfection, cells were harvested, resuspended in BRET buffer, and seeded in 96-well plates as for the BRET 2 assay. Cells were then treated with GR ligands diluted in BRET buffer so that the final concentration range was between 0.01 and 1000 nM. The incubation was performed at 37°C in the presence of 5% CO 2 for 1 h. Triplicate determinations for each condition were prepared. After incubation, the luminescence reaction was started by adding 10 l of the substrate coelenterazine, to give a final concentration of 5 M, and plates were read after 15 min on a Fusion Universal Microplate Analyzer (Packard BioScience, Pangbourne, UK) using the following filter settings: 485 nm for luciferase emission and 535 nm for EYFP emission.
The BRET ratio was then corrected for the BRET signal obtained from cells transfected with the Renilla-fused protein alone (Ren) determined as follows: BRET ratio ϭ (emission 535 nm-emission 485 nm ϫ cf)/emission 485 nm, where cf ϭ emission 535 nm Renilla alone/emission 485 nm Renilla alone. Curves were then constructed, and EC 50 values were calculated using a four-parameter curve fit.
IL-8 ELISA-HeLa cells were cultured in DMEM (Invitrogen) supplemented with 10% charcoal/dextran-treated fetal calf serum (Hyclone) for 48 h. Cell-conditioned medium was then collected from HeLa cells treated with 0.5 ng/ml TNF-␣, and a dose response of either Dex or RU486 (1-1000 M) for 16 h. IL-8 concentration was measured using an IL-8 ELISA kit (R & D Systems), following the manufacturer's instructions.
Immunofluorescence of Endogenous Protein-HeLa cells were seeded into each well of 8-chamber slides at 30,000 cells/chamber using DMEM supplemented with Glutamax-1 (Invitrogen Life Technologies) and 3% charcoal/dextran-stripped serum. 24 h later, the cells were treated with TNF-␣ (10 ng/ml) and either 100 nM Dex or 100 nM RU486 for 1 h. Cells were then washed with PBS and fixed for 15 min at room temperature with 3.7% paraformaldehyde. After fixing, cells were washed three times with TD buffer (10 mM Tris-HCl, pH 8, 150 mM NaCl) and treated with blocking buffer (TD plus 1% bovine serum albumin, 0.2% Triton X-100) for 1 h at room temperature. A 1:200 dilution of primary antibody, P20 sc-1002 for the GR, sc-109-G for p65 (both from Santa Cruz Biotechnology, Inc., Santa Cruz, CA) in washing buffer (TD plus 1% bovine serum albumin, 0.05% Triton X-100), was added for 2 h at room temperature. Cells were then washed three times with wash buffer, and a 1:200 dilution of the appropriate secondary antibody (Alexa Fluor 488 goat anti-rabbit IgG or Alexa Fluor 568 rabbit anti-goat IgG) (Molecular Probes, Inc., Eugene, OR) in washing buffer was added for 1 h at room temperature. The cells were then washed three times and mounted on slides using Citifluor (glycerol/PBS) (Citifluor Ltd.). Coverslips were sealed and stored at 4°C. Images were taken using a Leica TCS-4D confocal microscope (Leica Microsystems, Heidelberg, Germany) using a ϫ63 water immersion objective. To visualize Alexa 488, cells were excited at 488 nm using an argon laser, and emission was collected using a band pass filter of 525 Ϯ 25 nm and a dichroic beam splitter of 500. Alexa 568 cells were viewed using an excitation filter of 568 nm, and emission was collected using a 590-nm long pass filter.
Immunofluorescence of Overexpressed YFP Constructs-HeLa cells were pelleted, resuspended in medium supplemented with 10% charcoal/dextran-stripped serum (Hyclone), and seeded onto 22-mm glass coverslips at a density of 3 ϫ 10 5 cells/slide. Cells were transfected with 1 g of either p65-EYFP, SRC1-RID-EYFP, or EYFP-NCoR-RID using Fugene 6 transfection reagent (Roche Applied Science). 24 h posttransfection, cells were visualized on a Leica TCS-4D confocal microscope, using a ϫ 63 water immersion objective. Cells were excited with an argon laser at 488 nm, and emission was collected using a band pass filter of 525 Ϯ 25 nm.
Chromatin Immunoprecipitation-HeLa cells were cultured for 3 days in DMEM supplemented with 10% charcoal/dextran-stripped fetal bovine serum (Hyclone) to 95% confluence in T75 flasks (ϳ 5 ϫ 10 6 cells). Cells were treated with vehicle, TNF-␣ (10 ng/ml), or TNF-␣ and GR ligand (1 M) for 2 h and then washed twice with PBS. 1% formaldehyde was then added, and cells were incubated at room temperature for 10 min. Glycine (final concentration 0.125 M) was then added for 5 min to quench the formaldehyde. The protocol described by Shang et al. (24) was then followed. Briefly, cells were then washed twice with ice-cold PBS, collected into 100 mM Tris-HCl (pH 9.4), 10 mM dithiothreitol, incubated for 15 min at 30°C, and centrifuged for 5 min at 2000 ϫ g. Cells were then sequentially washed with 1 ml of ice-cold PBS, buffer I (0.25% Triton X-100, 10 mM EDTA, 0.5 mM EGTA, 10 mM HEPES, pH 6.5), and buffer II (200 mM NaCl, 1 mM EDTA, 0.5 mM EGTA, 10 mM HEPES, pH 6.5). Cells were resuspended in 0.5 ml of lysis buffer containing 1% SDS, 10 mM EDTA, 50 mM Tris-HCl, pH 8.1, 1ϫ protease inhibitor mixture (Roche Applied Science), and sonicated three times for 30 s each at the half-power setting (Jencons) followed by centrifugation for 10 min. Supernatants were collected and made up to 1 ml in lysis buffer, and 100 l was set aside to represent input. The input sample was heated at 65°C overnight to reverse the cross-links, and then DNA was extracted using the DNeasy tissue kit (Qiagen). The remaining supernatant was immunocleared with 2 g of sheared salmon sperm DNA, 20 l of preimmune serum, and 25 l of protein G-Sepharose (Santa Cruz Biotechnology) for 2 h at 4°C. Immunoprecipitation was performed overnight at 4°C, using either 2.5 g of GR M20 or 2 g of p65 sc-109 antibody (Santa Cruz Biotechnology). After immunoprecipitation, 25 l of protein A-Sepharose and 2 g of salmon sperm DNA were added, and the incubation was continued for another 2 h. Precipitates were then washed sequentially for 10 min each in TSE I (0.1% SDS, 1% Triton X-100, 2 mM EDTA, 20 mM Tris-HCl, pH 8.1, 150 mM NaCl), TSE II (0.1% SDS, 1% Triton X-100, 2 mM EDTA, 20 mM Tris-HCl, pH 8.1, 500 mM NaCl), and buffer III (0.25 M LiCl, 1% Nonidet P-40, 1% deoxycholate, 1 mM EDTA, 10 mM Tris-HCl, pH 8.1). Precipitates were then washed three times with TE buffer and extracted twice with 1% SDS, 0.1 M NaHCO 3 . Eluates were pooled and heated at 65°C HeLa cells were treated with TNF-␣ (10 ng/ml) and GR ligand (100 nM) for 1 h before fixing in 4% paraformaldehyde. Proteins were identified first using anti-GR antibody (P20) followed by anti-rabbit Alexa 488 secondary antibody and then using anti-p65 antibody (sc109-G) followed by anti-goat Alexa 568 secondary antibody. Proteins were visualized by sequential scanning on a Leica TCS-4D confocal microscope. a and b, representative cells treated with Dex or RU486, respectively. overnight to reverse the formaldehyde cross-linking. DNA fragments were purified using the DNeasy tissue kit (Qiagen). Following DNA extraction, real time quantitative PCR was performed (Stratagene M3000), using the SYBR Green master mix (Sigma), 0.4 M primers for promoter regions of the IL-8 and U6SnRNA genes, which have been previously described (18) (IL-8 promoter region Ϫ121/ϩ61, forward (5Ј-GGGCCATCAGTTGCAAATC-3Ј) and reverse (5Ј-TTCCTTCCGGTGG-TTTCTTC-3Ј); U6SnRNA promoter region Ϫ248/ϩ85, forward (5Ј-GG-CCTATTTCCCATGATTCC-3Ј) and reverse (5Ј-ATTTGCGTGTCATCC-TTGC-3Ј) and 5 l of extracted DNA product. Amounts of DNA precipitated in each experiment were calculated by comparison with standard curve values obtained from amplification reactions carried out with serial dilutions of genomic DNA. These values were then corrected for the amount of DNA precipitated from the IgG control and the amount of DNA present in the input sample. To control the specificity of the amplification and to ensure that only one product was generated, products were examined with both a dissociation curve program and by gel electrophoresis.
Statistics-Where appropriate, data were analyzed using analysis of variance followed by the Bonferoni post hoc test.

RESULTS
Dexamethasone but Not RU486 Repressed IL-8 Production-To investigate the ability of GR ligands to repress p65mediated transactivation, HeLa cells were treated for 16 h with TNF-␣ (0.5 ng/ml) in the presence or absence of steroid. IL-8 was measured in cell-conditioned medium by ELISA. TNF-␣ treatment increased IL-8 concentration 6.5-fold; this was significantly suppressed by Dex. In contrast, RU486 did not significantly repress IL-8 secretion (Fig. 1).
Co-localization of GR and NF-B p65 in Glucocorticoidtreated Cells-Both Dex and RU486 allow GR translocation, and both promote DNA binding (22), but RU486 failed to effectively repress NF-B-dependent transactivation; therefore, we wanted to analyze the effect of these two ligands on the nuclear colocalization of GR and p65. Endogenous GR was found in clusters throughout the nucleus when liganded with either Dex or RU486, with no apparent difference seen (Fig. 2, a and b). The intranuclear distribution of p65 was more diffuse compared with GR, but there was clearly significant overlap and co-localization in discrete foci (Fig. 2, a and b). Therefore, the failure of RU486 to repress NF-B was despite the GR sharing a similar intranuclear distribution with both ligands.
Analysis of Ligand-dependent Interaction between p65 and GR-The GR DBD and the Rel homology domain of p65 interact directly. To measure the effects of ligand on this interaction, a GST pull-down approach was used. The Rel homology domain of p65 bound full-length GR in the absence of ligand (Fig. 3a). This interaction was not significantly altered by Dex, but, surprisingly, it was significantly inhibited by RU486. Furthermore, the same pattern of ligand-dependent recruitment was seen when the experiment was repeated with the isolated DBD/LBD of the GR (Fig. 3, a and b), the minimal GR construct active in transrepression (18). To ensure that our VP16-DBD/ LBD construct retained the ability to repress NF-B, we compared its activity to wild type GR in an NRE-luc reporter gene assay. In response to dexamethasone, the ability of VP16-DBD/ LBD to repress p65 transactivation did not significantly differ from wild type GR at any ligand concentration (Fig. 3c).
Analysis of Protein-Protein Interaction in Living Cells-Using the in vitro GST pull-down approach, it appeared that GR ligand influenced interaction with p65. However, in living cells, other constraints may operate. Therefore, confirmation of the role of ligand was sought in living cells using BRET. Since BRET relies on energy transfer from Renilla luciferase to a fluorescent protein with appropriate spectra overlap, the relative orientation of the fluorescence tags to each other is crucial. We analyzed fusion proteins with both tags on either the N or the C termini using a BRET 2 assay. This approach revealed that Renilla luciferase fused to the C-terminal of the GR produced the most robust BRET signal with a variant of green fluorescent protein (GFP 2 ) fused to the C-terminal of p65 after treatment with Dex and TNF␣ (Fig. 4). This suggests molecular proximity of the C termini of GR and p65 in their final conformation. The maximal response was reached when cells were transfected with an excess of p65-GFP 2 relative to GR-Renilla; ratios between GR and p65 of 1:4 and 0.5:4 produced a -fold induction of 2.2 and 2.4, respectively. TNF-␣ alone did not increase the BRET ratio (data not shown). A robust BRET signal was also produced in the presence of GR-Renilla with either C-terminal tagged p65-EYFP or C-terminal tagged SRC1-RID-EYFP or N-terminal tagged EYFP-NCoR-RID (see Fig. 5a for schematic of BRET 1 constructs).
The subcellular localization of the EYFP-tagged proteins was also examined. HeLa cells were transfected with either p65-EYFP, SRC1-RID-EYFP, or EYFP-NCoR-RID constructs. Cells that expressed low levels of the p65-EYFP protein retained the expected cytoplasmic localization, whereas high expressing cells displayed a predominantly nuclear localization (Fig. 5b), consistent with previous observations (28). Both the SRC1-RID-EYFP and EYFP-NCoR-RID constructs were expressed diffusely throughout the cell, probably due to the lack of their nuclear localization signals (Fig. 5b). Similar results were also obtained in HEK293 cells, and the expression of the three EYFP constructs was confirmed to be similar by Western blotting for the shared fluorophore tag (data not shown).
A basal BRET signal of about 0.05 was usually observed (data not shown), suggesting a possible basal interaction between GR and p65 as demonstrated in the GST pull-down assay. The addition of either Dex or RU486 produced a further dose-dependent increase in BRET signal with an EC 50 of 8.7 nM for Dex and 6.88 nM for RU486 (Fig. 6, a and b); this was demonstrated using either BRET 2 (data not shown) or BRET 1 assays. Although both ligands produced similar EC 50 values, the increase in BRET signal induced by RU486 was only 49% of the Dex response. To examine the impact of activation of NF-B p65 by the TNF pathway, we incubated cells with recombinant TNF␣. The addition of 10 ng/ml TNF-␣ to the transfected cells had no significant effect on the maximal BRET ratio.

FIG. 4. Increased BRET signal is detected upon activation of C-terminal tagged GR and C-terminal tagged p65. GR and p65 fused
to Renilla or GFP 2 , respectively, were cotransfected into HEK293T cells using different combinations of C-terminal (e.g. GR:Rluc) and N-terminal (e.g. Rluc:GR) tagged constructs at different ratios as described in b. Cotransfection of GR-Renilla and GFP 2 -KaiB was performed as control for interaction, and the negative control was untransfected cells. Cells were split into 96-well plates and treated with 1 M Dex and 10 ng/ml TNF-␣ or mock-treated for 1 h at room temperature . After incubation, DeepBlueC coelenterazine was added, and the cells were assayed for light emitted at 410 and 515 nm. The 515/410 ratio was corrected using the untransfected cells. The column numbers correspond to the conditions labeled in b (in brackets). Results are shown as mean Ϯ S.E. of one representative experiment performed in triplicate (a).
When GR binds agonists, it is able to recruit members of the p160 co-activator family, such as SRC1, to its ligand binding domain, whereas the binding of the antagonist RU486 prevents such interaction (12). We found that Dex increased the GR: SRC1-RID BRET signal in a dose-dependent manner with an EC 50 of 0.5 nM (Fig. 6c); in contrast, RU486 failed to produce a BRET signal (Fig. 6d). The BRET signal induced by 10 nM Dex was competed by increasing concentrations of RU486 (data not shown). RU486 fails to recruit coactivators to the GR, and recent work has shown that corepressor proteins such as NCoR are recruited instead (12,13). We found that RU486 increased the GR:NCoR-RID BRET signal in a concentration-dependent manner with an EC 50 of 1.15 nM (Fig. 6f), whereas Dex failed to enhance the BRET signal (Fig. 6e).

RU486 Causes Reduced Tethering of GR to the IL-8 Promoter-
The BRET studies showed that ligand recognition influences the functional surface on the GR for recruitment of NF-B. However, on target genes, there may be an additional constraint imposed by the DNA sequence. Therefore, we sought interaction between the GR and NF-B on a well characterized, tethering GRE on the IL-8 promoter using chromatin immunoprecipitation. As a control, we used a region of the U6SnRNA promoter, which is not regulated by either NF-B or GR (18).
Using an anti-p65 antibody, we observed a significant induction (6.58 Ϯ 0.98) of NF-B p65 binding to the IL-8 element in response to treatment of the cells with TNF-␣, but no further changes were observed upon treatment with TNF-␣/Dex or TNF-␣/RU486 (Fig. 7a).
However, using GR antibody, we could detect no enrichment in IL-8 promoter DNA being precipitated when cells were treated with TNF-␣ alone (Fig. 7, b and d). When cells were treated with both TNF-␣ and dexamethasone, there was a 5.5-fold increase in the amount of IL-8 promoter DNA (p Ͻ 0.0001) immunoprecipitated. When cells were treated with TNF-␣ and RU486, the increase in IL-8 DNA precipitated compared with untreated or TNF-␣-treated cells was significantly less than the increase seen with dexamethasone (p Ͻ 0.0001) (Fig. 7, b and d). Furthermore, amplification of a region on the U6SnRNA promoter, which is not regulated by either GR or p65, revealed minimal amplification post-IP but had similar C t values to the IL-8 amplification from the input samples (Fig. 7c). Similarly replacing the primary antibody with control IgG also resulted in no amplification using either primer pair (Fig. 7b). DISCUSSION In this study, we demonstrate, both in vitro and in living cells, that glucocorticoid receptor ligands regulate recruitment of NF-B p65 to the GR DNA binding domain. It has long been established that ligands for the nuclear hormone receptors act to generate different functional surfaces at the C-terminal and so affect recruitment of transcriptional co-factors. However, we now show that ligand-induced conformational changes extend to the DNA binding domain of the GR, previously thought to be autonomous in function.
Since previous studies had reported divergent action of RU486 on cytokine gene repression, we initially analyzed glucocorticoid repression by both Dex and RU486 in HeLa cells. We found that there was negligible repression of IL-8 secretion, even when high concentrations of RU486 were used. In contrast, maximal repression of IL-8 was seen by 1 nM Dex, suggesting an IC 50 for this effect of less than 1 nM.
Studies of GR intracellular trafficking induced by RU486 or Dex did not reveal marked differences in nuclear translocation or DNA binding (8 -11). Although Htun et al. (9) reported clear differences in the intranuclear distribution of overexpressed GR induced by either RU486 or Dex, this was not apparent either with endogenous protein or with overexpressed EYFP-GR in our cell system ( Fig. 2 and data not shown).
There are several differences between the experimental systems used in this study and previously by Htun et al. (9). First, in the Htun et al. study, a rat GR containing the C656G mutation, an N-terminal His tag, and a heamagglutinin epitope was used. The GFP cDNA was then inserted between the two N-terminal tags and the GR cDNA. Second, the cell line model was a murine adenocarcinoma cell line with multiple copies of murine mammary tumor virus long terminal repeat stably integrated. The influence of the N-terminal modifications and the cell line employed may be important. For these reasons, we examined the intranuclear distribution of the endogenous GR. We did not observe any consistent difference in the intranuclear distribution of the GR in the presence of dexamethasone compared with RU486. It is possible that overexpression of the modified GR unmasks differential distribution induced by the two ligands, but in our studies using immunofluorescence and chromatin immunoprecipitation, we used endogenous, wild type GR, and we propose that this approach is more physiological. Therefore, the different effects of Dex and RU486 on NF-B function are unlikely to be explained by differences in trafficking or intranuclear sequestration of GR away from p65 (Fig. 2).
Many theories have been presented to explain the mechanism of GC transrepression, including up-regulation of IB␣, competition for co-factors, and sequestration of PKAc (25) (16, 26 -28). It has more recently been shown that activated GR does not displace NF-B from its binding site on the IL-8 promoter, but rather recruits repressor proteins in a tethering mechanism (18). Recent work by Rogatsky et al. (20) indicated that the transcriptional co-activator GRIP-1 is able to mediate GC repression of NF-B activity on the collagenase gene pro- moter and found no role for histone deacetylases in this mechanism. The failure of the RU486-bound GR to recruit GRIP-1 as a co-repressor may explain its failure to repress this template. However, as RU486 causes recruitment of the potent repressor protein NCoR, it might be expected to inhibit NF-B transactivation, possibly in a similar manner to that of the thyroid hormone receptor, which represses NF-B by recruiting HDAC activity to the promoter (20,21).
To investigate the interaction of GR with p65, we initially adopted a GST pull-down approach. The native GR was able to interact with the RHD of p65 in vitro; this was unaltered by Dex, but, surprisingly, interaction was opposed by RU486. Constitutive and agonist-mediated interaction between GR and p65 in the GST assay has been previously reported (18). In principle, since both the GR and p65 are held in chaperone complexes in the cytoplasm while inactivated, it seems unlikely that such an interaction would be seen under physiological conditions in the absence of ligand. However, recent work by Widén et al. (29) suggests that there may be a cytoplasmic interaction between the GR and the p65, p50, and IB␣ complex. The impairment of interaction between GR and p65 seen with RU486 is surprising and suggests allosteric effects from the GR-LBD on the GR-DBD. RU486 similarly inhibited interaction between p65 and a truncated GR protein, VP16-DBD/ LBD, that retains transrepression activity when it is bound to GR agonist. Identification of a previously unsuspected effect of ligand on GR protein recruitment in vitro prompted development of a complementary in vivo assay. BRET allows the monitoring of protein-protein interactions in living cells by tagging one protein of interest with Renilla-luciferase and the other with either EYFP for BRET 1 or a variant of green fluorescent protein (GFP 2 ) for BRET 2 . The strongest BRET signal was observed with a C-terminal GFP 2 -tagged p65 and a C-terminal Renilla luciferase-tagged GR. This suggests that the C termini of both proteins are in close proximity to each other. All subsequent analyses were performed using the BRET 1 assay, because its substrate was much more stable, and therefore the assay is more robust for high throughput screening of compounds. The DNA binding domain of GR and the Rel homology domain of p65 have been identified as the interacting domains. Nevertheless, as the C termini of both proteins are also in close proximity, it is feasible that other stabilizing interactions may occur between the proteins or that close opposition of the C termini is required for functional interaction. The increase in BRET signal upon the addition of both Dex and RU486 suggests an increase in binding between GR and p65 or a change in conformation of the protein complex. The addition of TNF-␣ to the assay failed to significantly alter the GC-induced increase in the GR:p65 BRET ratio. This may be due to the fact that only cells expressing low levels on p65-YFP displayed a cytoplasmic localization. Therefore, TNF-induced translocation of protein from these low expressing cells would not significantly increase the GR:p65 BRET signal. Both Dex and RU486 produced a similar EC 50 in the assay comparable with the similar, high affinities the two ligands share for binding to the GR. In contrast, the total increase in BRET ratio was halfmaximal for the antagonist RU486 compared with the agonist Dex. There are two possible interpretations of this observation: 1) the conformational change in the GR LBD induced by RU486 may move the Renilla tag on GR away from the EYFP tag on p65, thereby reducing the efficiency of energy transfer, or 2) RU486 may result in significantly less interaction with p65 associated with a conformational change in the GR LBD that moves the Renilla tag on GR closer to the EYFP tag of p65.
This result is compatible with the reduced interaction observed in the GST pull-down assay but still implies that a fraction of RU486-liganded GR is bound to NF-B p65. Differences between in vitro and in vivo assays could be due to the presence of cofactors crucial for GR-p65 interaction in living cells.
Recent work in our laboratory and by others has revealed that Dex promotes the recruitment of SRC1 but not NCoR to the GR LBD, whereas RU486 promotes the recruitment of NCoR but not SRC1 (12,13). To further explore ligand-dependent protein recruitment to the GR, living cells were examined using the BRET assay. Since both SRC1 and NCoR are large proteins (160 and 270 kDa, respectively), truncated proteins were used to maximize molecular proximity. The truncated NCoR construct included the three receptor-interacting domains (RIDs) outlined by Schulz et al. (13), and we also used a well characterized SRC1 fragment containing three LXXLL motifs, as previously described (12,30). Dex promoted BRET and, by inference, interaction between GR and SRC1; in contrast, RU486 promoted BRET between GR and NCoR, as predicted from earlier biochemical approaches (12,13). Importantly, this provides independent, in vivo confirmation of the earlier observations as well as validating the BRET approach for investigating GR interactions in vivo.
However, demonstration of altered interaction between the GR and NF-B p65 either in vitro or in vivo does not take into account the possible effects of DNA binding on NF-B structure. Therefore, we set out to analyze GR interaction with the DNA-bound NF-B on the well characterized IL-8 promoter in HeLa cells.
TNF-␣ treatment of the HeLa cells causes induction of NF-B p65 binding to the IL-8 NF-B element, to a similar level as previously reported in A549 cells (18). Furthermore, this study also found that GR ligand did not affect the amount of p65 on the promoter, in accordance with our findings. We found no change in GR occupancy of the promoter in response to TNF-␣, but a clear induction was seen with combined TNF-␣ and Dex. There was also a small increase in GR occupancy seen in response to RU486, which possibly explains the marginal repression of IL-8 previously described. Since RU486 impairs both interaction between the GR and NF-B and also recruitment of GR to DNA-bound NF-B, this mechanism contributes to the failure of RU486 to inhibit inflammation.
These data show that ligand-induced folding of the GR does indeed regulate interaction between GR and NF-B, in addition to influencing recruitment of transcriptional co-factors to the GR C terminus. This allosteric effect was greater than expected and reveals previously unrecognized cross-talk between the C terminus and DNA binding domain of the GR. These observations identify new ways in which the functional domains of the GR interact to regulate target gene transcription.