Glucocorticoid Receptor Domain Requirements for Chromatin Remodeling and Transcriptional Activation of the Mouse Mammary Tumor Virus Promoter in Different Nucleoprotein Contexts*

The glucocorticoid receptor (GR) contains several activation domains, τ1 (AF-1), τ2, and AF-2, which were initially defined using transiently transfected reporter constructs. Using domain mutations in the context of full-length GR, this study defines those domains required for activation of the mouse mammary tumor virus (MMTV) promoter in two distinct nucleoprotein configurations. A transiently transfected MMTV template with a disorganized, accessible chromatin structure was largely dependent on the AF-2 domain for activation. In contrast, activation of an MMTV template in organized, replicated chromatin requires both domains but has a relatively larger dependence on the τ1 domain. Domain requirements for GR-induced chromatin remodeling of the latter template were also investigated. Mutation of the AF-2 helix 12 domain partially inhibits the induction of nuclease hypersensitivity, but the inhibition was relieved in the absence of τ1, suggesting the occurrence of an important interaction between the two domains. Further mutational analysis indicates that GR-induced chromatin remodeling requires the ligand-binding domain in the region of helix 3. Our study shows that the GR activation surfaces required for transcriptional modulation of a target promoter were determined in part by its chromatin structure. Within a particular cellular environment the GR appears to possess a significant degree of versatility in the mechanism by which it activates a target promoter.

Nuclear receptors have provided fruitful models for studying the mechanisms by which transcription factors work (reviewed in Ref. 1). The steroid receptors are a ligand-inducible class of this family and the glucocorticoid receptor (GR) 1 cDNA was the first of these to be cloned and sequenced (2). The GR has a ubiquitous expression pattern in mammals and is involved in regulation of a number of biological processes through regula-tion of various target genes. In the mammary gland, it is essential for the differentiation process that leads to lactation.
The mechanism of GR transactivation has been the focus of a large number of studies, and domains of the GR involved in transcriptional activation have been defined. One of these domains, 1 or AF-1, is located in the amino-terminal region of the receptor (3)(4)(5). Its core region is unstructured in solution but can form an ␣-helical structure in a hydrophobic environment (6,7). Helix-breaking proline substitutions (7) or mutations in hydrophobic amino acid residues decrease its transactivation potential (8,9). The 1 domain is also important in mediating transcriptional repression by the GR (10). It has been shown to interact with a number of factors and complexes including TBP (11), TFIID (11,12), CBP (9), members of the DRIP/vitamin D receptor-interacting proteins complex (13), the yeast histone acetyltransferase complex SAGA (14), and the ATP-dependent chromatin remodeling complex, SWI/SNF (15).
Two other GR transactivation domains are located in the ligand-binding region. The first, 2, is a small region at the amino-terminal end of the ligand-binding domain (LBD) (5). It contains sequences that are conserved among the steroid receptors and has been shown to mediate transactivation when fused to a DNA-binding domain (DBD) (5,16). It also harbors a nuclear matrix targeting signal and interacts with Hic-5, a protein that associates with the nuclear matrix and also potentiates GR action (17,18). The other known transactivation domain of the GR, termed AF-2, is a larger region within the COOH terminus of the LBD. In other nuclear receptors it is thought to comprise a surface formed by ␣-helices 3, 5, 6, and 12 (19,20). Helix 12, termed the AF-2 AD core, undergoes a conformational change upon binding of ligand that generates a surface for interaction with either corepressors or coactivator proteins, dependent on whether the ligand is an agonist or antagonist (reviewed in Ref. 1). Coactivators include SRC1, TIF2/GRIP1, CBP/p300, and P/CAF, some of which are histone acetyltransferases (reviewed in Ref. 21).
Consistent with the fact that it interacts with histone acetyltransferases and ATP-dependent remodeling complexes, the GR is known to induce modification of chromatin structure. A number of target genes have glucocorticoid-inducible nuclease hypersensitive sites (22)(23)(24). Perhaps the best studied of these is the mouse mammary tumor virus (MMTV) promoter. The GR has been shown to activate this promoter by two distinct mechanisms, depending on its chromatin structure (25). When the promoter is organized into replicating chromatin, either as an episome or integrated into the genome, the GR induces a chromatin remodeling event in the nucleosome regions containing its binding sites (24,26). This is mechanistically necessary for transcriptional activation because it allows access of previously * The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18  excluded transcription factors NF1 and Oct1 (25,27,28). The GR also facilitates the association of the basal transcription machinery with the promoter, which presumably activates transcription (28). Therefore, this form of the MMTV promoter is derepressed and then activated through GR action.
When the MMTV promoter is introduced into cells as a transiently transfected reporter construct, it adopts a disorganized and accessible chromatin structure to which NF1 and Oct1 bind constitutively rather than in a hormone-dependent fashion (25,28). This template does not undergo remodeling but is activated by GR, probably through increased association of the basal machinery. Once the MMTV template in organized chromatin is remodeled and derepressed, it is unclear whether GR activates it by the same mechanism employed at the transiently transfected MMTV template. One way to approach this question would be to assess the effect of various GR activation domain mutants on the ability of the two templates to be activated. These domains were defined using transiently transfected reporter constructs. It is not known whether these same domains would be necessary for activation of a promoter in repressed chromatin.
There is conflicting information on the domains required for chromatin remodeling induced by the GR. Several studies indicate that the SWI/SNF family of ATP-dependent remodeling complexes are involved in transactivation by GR in mammalian cells (15,29,30). Both in vivo (31) and in vitro (32,33) studies suggest a role for the SWI/SNF complex in the induction of nuclease hypersensitivity by GR at the MMTV promoter. GR has also been shown to interact physically with the SWI/SNF complex (15,30,31,34). Recently, Wallberg et al. (15) showed that activation by GR in yeast was dependent on the presence of Snf6, a member of the ATP-dependent nucleosome remodeling complex, SWI/SNF. This dependence was mediated through the 1 domain, which was shown to interact directly with the yeast SWI/SNF complex in vitro and, when fused to a heterologous DNA-binding domain, activate transcription on in vitro assembled chromatin in a SWI/SNF-dependent manner. However, Muchardt and Yaniv (29) showed that expression of the SWI/SNF ATPase, brahma, greatly potentiated GR-mediated promoter activation in a manner dependent on the GR DBD. Further complexity emerged from the studies of DiRenzo et al. (35) who showed that the estrogen receptor interacts functionally and physically with this complex in an AF-2-dependent fashion. Thus, multiple receptor domains have been implicated in interactions with the SWI/SNF complex but none of these studies assayed chromatin remodeling directly.
In this study we examine the role of each of the activation domains in GR function in vivo. We have taken advantage of a GR mutant, C656G, which binds its ligand with higher affinity than wild type GR (36). We showed previously that this receptor is capable of remodeling chromatin and activating transcription from an MMTV promoter in ordered chromatin (37). In the context of the full-length C656G receptor, mutations were introduced into the previously defined activation domains. These mutations were assayed for their role in transcriptional activation of the transiently transfected and stably replicating MMTV templates as well as in GR-dependent chromatin remodeling at the latter. The results show that the GR uses its domains differently in the transactivation of the MMTV promoter in distinct nucleoprotein environments. In addition, through direct measurement of chromatin remodeling we have determined that the ligand-binding domain plays a critical role in mediating this process. Our findings strongly support the idea that the mechanism by which the GR modulates transcription is not only dependent on the nature of the target promoter but also on its chromatin configuration.

EXPERIMENTAL PROCEDURES
Cloning of GR Mutants-The C656G expression vector, pCInH6HA-C656G, was made, as previously described (37), by insertion of C656G cDNA (kindly provided by Stoney Simons) into a pCI-nH6HA vector, such that the full-length receptor was fused with a histidine tag and a hemagglutinin A (HA) epitope at its NH 2 terminus. The ⌬1 mutant was created as follows. A SalI site in the multiple cloning sequence was deleted from the C656G expression vector by digestion with XbaI (in the multiple cloning sequence) and EagI (just downstream of the multiple cloning sequence) and subsequent ligation with linker sequences. The resulting vector (C656G⌬SalI) was digested with SalI and Bsp120I, which cleave the receptor cDNA at sites within or at the COOHterminal end of the 1 domain, respectively. The cleaved ends were then ligated with the following complementary oligonucleotides: 5Ј-TCGAC-CAGCGTT-3Ј and 5Ј-GGCCAACGCTGG-3Ј. The resulting product has amino acids 157-317 in the 1 domain deleted (⌬1). The AF2dm12 mutant (M770A/E773A) and ⌬1AF2dm12 mutants were created using a Chameleon TM double-stranded, site-directed mutagenesis kit (Stratagene). The mutagenic primer used to change methionine 770 and glutamic acid 773 of the helix 12 domain of C656G and ⌬1 to alanines was 5Ј-AATTCCCAGAGGCGTTAGCTGCAATCATCACTAATCAG-3Ј.
Cells, Transfection, and Sorting-Cell line 1471.1 was maintained in Dulbecco's modified Eagle's medium (Invitrogen) supplemented with 10% fetal bovine serum (Hyclone). This cell line was derived from C127 mouse mammary adenocarcinoma cells and contains multiple stably replicating copies of MMTV-chloramphenicol acetyltransferase fusions in the context of bovine papilloma virus sequences, as previously described (37). The 1471.1 cells express moderate levels of GR.
Transfections were carried out by electroporation using a BTX Electro SquarePorator T820 (Genetronics) as described previously (37). Cells were transfected with 5 g of pCMVIL2R (interleukin 2 receptor Tac subunit expression vector) for cell sorting and varying amounts of receptor expression vector DNA (3-12 g). Cells were also transfected with either 10 g of pLTRluc (full-length MMTV long terminal repeat driving luciferase) or 10 g of pMTVbgln (full-length MMTV long terminal repeat driving expression of the rabbit ␤-globin gene) for transient template assays. After electroporation, cells were plated in phenol red-free Dulbecco's modified Eagle's medium (Invitrogen) supplemented with 10% charcoal/dextran-treated fetal bovine serum (Hyclone). The following day cells were treated with dexamethasone as indicated in the figure legends and sorted with goat anti-mouse IgGcoated magnetic beads (Dynal) coupled with interleukin 2 receptor monoclonal antibody (Upstate Biotechnologies, Inc.), as previously described (37). Sorted cells were either used immediately or frozen for later analysis.
RNA and Luciferase Analysis-RNA was isolated as described previously (38). RNA induction was measured by S1 nuclease protection assay, as previously described (37). Briefly, isolated RNA (10 g) was hybridized with probes specific for MMTV and ␤-actin sequences overnight and then digested with S1 nuclease (25 units/10 g of RNA). Digestion products were separated on 8% denaturing gels that were dried and exposed to phosphorimaging screens. Quantitation was carried out using ImageQuant software (Amersham Biosciences).
Cell extracts for luciferase assays were made from transfected, nonsorted cells as previously described (39). Luciferase activities were measured using a Microlumat LB 96P luminometer (EG&G Berthold). For each sample luciferase activity was normalized to the amount of cellular protein used in the assay. Dose-response curves and EC 50 values were generated using GraphPad Prism software.
Western Blotting-Whole cell extracts were generated as follows. Sorted cells were resuspended in HEGDM (10 mM HEPES, pH 8.0, 1 mM EDTA, 10% glycerol, 2 mM dithiothreitol, and 10 mM sodium molybdate) containing 0.1% Nonidet P-40 and a protease inhibitor mixture (Calbiochem), after which NaCl was added to a final concentration of 250 mM. After incubation on ice for 20 min, cell lysates were centrifuged at 30,000 ϫ g. The supernatants were stored at Ϫ80°C. Generation of cytosolic extracts was similar except that NaCl addition was omitted and incubation on ice was reduced to 5 min before centrifugation at 12,000 rpm in a microcentrifuge.
Nuclei Digestion and DNA Analysis-Nuclei were isolated from magnetically sorted cells as previously described (37). Aliquots of nuclei containing 75-100 g of DNA were resupended in 100 l of nuclei digestion buffer (2.5% glycerol, 1 mM MgCl 2 , 50 mM NaCl, 50 mM Tris, pH 8.0, and 1 mM ␤-mercaptoethanol). Nuclei were digested with SacI (10 units/g of DNA) for 15 min at 30°C or with -exonuclease (100 units/ml) and HaeIII (1000 units/ml) at 37°C for 15 min. Digestion was terminated with 5 volumes of nuclei stop buffer (10 mM Tris, pH 7.5, 10 mM EDTA, 0.5% SDS) containing 100 g/ml proteinase K, and samples were incubated at 37°C overnight. DNA was purified by phenol/chloroform/isoamyl alcohol extraction, precipitated, and resuspended in 10 mM Tris, pH 8.0, 1 mM EDTA. DNA from nuclei cleaved with SacI was digested to completion with DpnII. Digestion products were detected by linear amplification using Taq polymerase and a radiolabeled oligonucleotide primer (MMTV sequence from ϩ27 to ϩ1 bp) followed by separation on 8% denaturing gels. Dried gels were exposed to Phospho-rImager screens. Quantitation of the radiolabeled digestion products was carried out using ImageQuant software (Amersham Biosciences).

Effects of Mutations in the 1 and AF-2 Helix 12 Domains on
Transactivation-Defining domains of the GR required for activation or repression of endogenous mammalian target promoters has been hampered by the fact that most cultured cell lines that express a known endogenous target gene also express endogenous GR. This makes it difficult to separate effects of transfected mutant receptors from those of the wild type resident receptor. We have circumvented this problem by using the rat GR mutant, C656G, which has higher affinity for ligand than the wild type GR. The receptor has been shown previously to fully activate the MMTV promoter at concentrations of ligand that are too low to induce the endogenous GR in our murine mammary adenocarcinoma-derived cell lines (37). Thus, the activity of C656G and any activation domain mutants derived from it can be determined in the presence of endogenous GR. Although the C656G contains a point mutation in the LBD, we have shown that it does not affect its ability to induce remodeling or activation of the MMTV promoter in organized chromatin (37).
As shown in Fig. 1A, we introduced mutations into the C656G receptor to disable either the 1 or AF-2 helix 12 domains, or both. The 1 domain was originally characterized as a rather large region of the NH 2 -terminal area of the GR (5). Later, a smaller core region was defined that was sufficient, when fused to a DBD, to activate transcription (40). However, we deleted most of the 1 domain to avoid missing any relevant sequences that might be necessary for activation in chromatin. Mutation of the AF-2 domain was more problematic because of the possibility of deleterious effects on ligand binding. We introduced two mutations, M770A and E773A, into the helix 12 region, which had been reported to have varying effects on ligand binding (41)(42)(43).
When introducing mutations into the LBD of steroid receptors it was important to assess the effects of the mutation on ligand binding to separate ligand binding defects from actual transcriptional defects. To ensure that our receptors with LBD mutations respond appropriately to ligand, we carried out doseresponse experiments to measure their ability to activate a transiently transfected MMTV-luciferase reporter. The EC 50 values are shown in Table I. The dose-response curve for the AF2dm12 mutant, shown in Fig. 1B, is right-shifted relative to that of C656G. Comparison of EC 50 values for C656G and the AF2dm12 mutant indicates that the mutant was about half as sensitive to ligand as C656G (Table I). Whereas C656G has reached maximal activation at 1 nM dexamethasone, the AF2dm12 mutant reaches that point at 3 nM dexamethasone. However, the endogenous receptor was partially active at 3 nM so we were limited to using the lower dose of dexamethasone. At 1 nM dexamethasone the AF2dm12 mutant has reached 80 -90% of its maximal activity. Thus, small (10 -20%) transcriptional defects observed with receptors containing this mutation, as well as others with similar EC 50 values, will be interpreted as a ligand binding effect in the experiments to follow. Deletion of the 1 region alone had no effect on the dose response relative to C656G (data not shown). However, removal of the 1 domain in the context of the AF2dm12 muta- FIG. 1. GR mutants. A, mutations were introduced into the C656G receptor as shown. B, dose response assays were carried out using a transfected MMTV-luciferase reporter construct. Cells were also transfected with expression vectors for C656G or AF2dm12 or no expression vector (endogenous GR). Treatments with dexamethasone (Dex) were 4 h in duration. Curve fitting and calculation of EC 50 values were carried out using GraphPad Prism software.
To compare the activity of the various mutant receptors on the transient and stable MMTV templates in the same cells, we co-transfected cell line 1471.1 with expression vectors for the GR mutants, the Tac subunit of the interleukin 2 receptor, as well as an MMTV-␤-globin reporter, which served as the transient template. The interleukin 2 receptor subunit serves as a tag for magnetic affinity cell sorting (44). This was carried out after transfection and dexamethasone treatment to isolate a population of cells highly enriched for the presence of both MMTV templates and the expression of the GR mutants. RNA was collected from the transfected cell populations and subjected to S1 nuclease analysis with an MMTV probe designed to detect transcripts from both the transient and stable MMTV templates as well as a ␤-actin probe. A representative S1 analysis is shown in Fig. 2A. Whole cell extracts were prepared from the transfected cell populations and subjected to Western analysis to ensure that roughly equivalent amounts of the various receptor mutants were expressed (Fig. 2B).
A summary and statistical analysis of the S1 data are shown for the transient and stable templates in Fig. 2C. Relative to C656G, activation of the transient MMTV template (left panel) by AF2dm12 was about 60% lower, whereas activation by ⌬1 was only about 20% lower. The double mutant, ⌬1/AF2dm12, was a less efficient activator than AF2dm12. In contrast, both domains assayed separately had significant effects on dexamethasone activation of the stable MMTV template (right panel), ⌬1 resulting in a 60% loss of activation and AF2dm12 resulting in a 40% loss. The double mutant was only 25% as effective as C656G and had less activity than either of the two mutations alone. The same pattern of effects for both templates was also observed in a different cell line having integrated copies of an MMTV-Ras transcription unit (data not shown). The results show that activation of the transient template was driven largely by the AF-2 helix 12 domain with a small contribution from the 1 domain. However, in the same cells activation of the stable MMTV template requires both domains with the larger contribution coming from 1. Mutation of both domains does not have a synergistic effect on loss of activation at either template but does not eliminate activation altogether.
Hydrophobic residues in the 1 core region (amino acids 208 -264 in rat GR) have been shown to be important for its activation function. In particular, a combination of three mutations was found to severely impair 1-dependent activation of transiently transfected glucocorticoid-regulated reporters in F9 embryonal carcinoma cells (10). In addition, the same set of mutations prevents the interaction of GR with members of the DRIP complex (13). We constructed a receptor containing these mutations (Fig. 3A) and tested its ability to transactivate the two MMTV templates. A representative Western blot and S1 RNA analysis are shown in Fig. 3, panels B and C, respectively. At the transient template (Fig. 3D, left panel) the 1EFW mutant was only 65-70% as effective in activation as C656G. Its efficiency was somewhat lower than that of ⌬1, which carries a large deletion, indicating that these amino acids were probably responsible for the small contribution of the 1 domain to activation at the transient template. However, this combination of amino acid substitutions may cause 1 to be somewhat inhibitory to receptor function in the context of the transient template relative to removal of the entire domain, as in ⌬1. At the stable MMTV template the 1EFW mutant was 80% as effective as C656G (Fig. 3D, right panel). This provides a sharp contrast to the ⌬1 mutant, which was only 40% as effective.
Because the ⌬1/AF2dm12 receptor has a significant amount of residual activity at the two MMTV templates, we considered the possibility that other regions of the AF-2 domain partially compensate for the two mutated residues in the AD core. Met 770 and Glu 773 lie within the helix 12 region. However, the  Fig. 1A were expressed in 1471.1 cells by transient transfection. Dexamethasone (Dex) treatments were 3.5-4 h in length. A, RNA was isolated from transfected cells after magnetic affinity cell sorting, hybridized with MMTV and actin probes, and subjected to S1 nuclease digestion. A representative gel is shown. B, whole cell extracts were isolated from transfected cells in the absence of dexamethasone and subjected to magnetic affinity cell sorting. Transfected receptors were detected by Western blotting using an antibody against the HA epitope. C, data from 5 to 14 independent experiments were subjected to statistical analysis. In each experiment MMTV RNA levels were normalized to those of actin in the same sample. The normalized MMTV RNA level induced by activation of C656G with 1 nM dexamethasone was set to 100 and other levels were expressed as a percentage. Error bars represent S.E. AF-2 surface to which various coactivators bind in other nuclear receptors also includes residues in helices 3, 5, and 6 (19,20). Mutation of a lysine residue at the COOH-terminal edge of helix 3 has been shown to impair transactivation in the estrogen receptor (45), thyroid receptor (19), and vitamin D receptor (46) without a severe effect on ligand binding. This mutation was introduced into the C656G, AF2dm12, or ⌬1/AF2dm12 receptors (see Fig. 1A). Dose-response experiments were carried out (Fig. 4A) and EC 50 values are shown in Table I. The K597A receptor has near maximal activity (90%) at 1 nM dexamethasone, and its EC 50 value was close to that of C656G. The EC 50 values of the two receptors carrying both the K597A and AF2dm12 mutants vary depending on the presence of the 1 domain, as we observed for the AF2dm12 and ⌬1/AF2dm12 receptors. The ⌬1/AF2dm12/K597A also has near maximal activity (80%) at 1 nM dexamethasone and an EC 50 slightly lower than that of AF2dm12. However, the activity of the AF2dm12/K597A mutant was less than 50% of maximum at 1 nM (Table I), making it unsuitable for this analysis.

FIG. 2. Effects of 1 and AF-2 mutations on activation of two MMTV templates. Receptors shown in
Expression levels of the K597A and ⌬1/AF2dm12/K597A receptors were measured by Western blotting after transfection and found to be similar (data not shown). Their activity at the two MMTV templates was assessed by S1 nuclease assay (Fig.  4B). As seen in Fig. 4, C and D, mutation of Lys 597 alone reduced activation at both MMTV templates to approximately the same extent as shown for the AF2dm12 mutant (dashed line and Fig. 2C). Thus helices 3 or 12 were important for activation of the MMTV promoter in either structural context. The Lys 597 mutation in the context of ⌬1/AF2dm12 (⌬1/ AF2dm12/K597A) resulted in a further decrease in activation efficiency relative to the K597A and the ⌬1/AF2dm12 receptors at both templates, presumably reflecting the contributions the 1 and helix 12 domains make to the overall activity of the GR. Notably, ⌬1/AF2dm12/K597A has residual activity at both templates.
The remaining activity of GR with disabled 1 and AF-2 domains may reflect a contribution from the third known activation domain in the GR, 2. Because 2 lies within the ligandbinding domain, it was difficult to mutate without disrupting ligand binding. However, Milhon et al. (16) had characterized a mutation in the mouse GR 2 region that bound ligand with normal affinity but impaired transactivation. This mutation was introduced into the C656G, AF2dm12, or ⌬1/AF2dm12 receptors (see Fig. 1A). As seen in Table I, dose-response analysis showed that the S573A receptor had an EC 50 value close to that of C656G. However, combination of the S573A and AF2dm12 mutations had a deleterious effect on the EC 50 independent of the presence of the 1 domain (see EC 50 values for the AF2dm12/S573A and ⌬1/AF2dm12/S573A receptors in Table I). Therefore, the S573A receptor was transfected into cells and tested for expression levels (Fig. 5A) as well as transactivation potential (Fig. 5B). The summary of results, as shown in FIG. 3. Effects of hydrophobic residues in the 1 core on MMTV activation. A, mutations were introduced into the 1 core region as shown. B, cytosols were isolated from cells transfected with either C656G or 1EFW and sorted by magnetic affinity cell sorting. Receptors were detected by Western blotting and an anti-HA antibody. C, results from a representative S1 nuclease analysis of RNA from transfected cells. Dexamethasone (Dex) treatments were 3.5-4 h in length. D, data from three independent experiments were analyzed as described in the legend to Fig. 2.

FIG. 4. Effects of an LBD helix 3 mutation on MMTV activation.
A, expression vectors for the K597A and ⌬1/AF2dm12/K597A receptors were transfected into cells with the MMTV reporter construct, pLTRluc. Dexamethasone dose-response assays were carried out as described under "Experimental Procedures." Receptor activity was expressed as -fold induction (relative to activity at 1 pM dexamethasone). Results from at least three independent experiments are shown. B, a representative S1 nuclease experiment. Dexamethasone (Dex) treatments were 3.5 h in length. C, data from at least three independent experiments were analyzed as described in the legend to  Effects of Activation Domain Mutations on Chromatin Remodeling at the Stable MMTV Template-To determine which GR domains were essential for the structural transition in chromatin at the stable MMTV template, we tested the ability of various mutants to induce nuclease hypersensitivity and binding of NF1 in the proximal promoter region. Fig. 6 shows the results of restriction enzyme access experiments using SacI, which cleaves the promoter within the induced region of hypersensitivity (26). A representative set of data is shown in Fig. 6A. In each experiment, the dexamethasone-induced change in fractional cleavage was calculated for each receptor. The average changes in fractional cleavage, expressed as a percentage, are shown in Fig. 6B. As shown previously, C656G induces hypersensitivity at 1 nM dexamethasone to the same extent as the endogenous receptor (indicated as mock) at 100 nM dexamethasone (37). Deletion of the 1 domain has only a slight effect on the induction of SacI cleavage, indicating that it functions downstream of chromatin remodeling and derepression in the mechanism of activation. The receptor mutated in helix 12 (AF2dm12) was significantly impaired for induction of hypersensitivity, being less than half as effective as C656G. This degree of impairment correlates well with its relative ability to activate transcription from this template (see Fig.  2D), which underscores the fundamental role depression plays in the mechanism of GR action in the chromatin context. Surprisingly, the double mutant, ⌬1/AF2dm12, has normal remodeling activity. These results indicate that the helix 12 region of AF-2 contributes to the chromatin remodeling activity of GR only in the presence of an intact 1 domain.
The binding of NF1 to the stable MMTV template was tightly correlated to the induction of nuclease hypersensitivity by GR. We tested the ability of the various GR mutants to induce NF1 binding as measured by exonuclease block assay. As shown in Fig. 7, the C656G receptor was able to induce NF1 binding when activated by 1 nM dexamethasone to approximately the same level as that induced by the endogenous GR at 100 nM. The ⌬1 mutant induces NF1 binding efficiently, but the AF2dm12 mutant does not. The double mutant, ⌬1/AF2dm12, was also able to induce NF1 binding to the MMTV promoter. These results correlate well with the SacI access data.
To better define the region of the GR important for induction of chromatin remodeling, we assayed the K597A mutation alone or in the context of the ⌬1/AF2dm12 receptor, which induces wild type levels of nuclease access. As shown in Fig. 8, both the K597A and ⌬1/AF2dm12/K597A receptors were significantly impaired in their ability to remodel MMTV chromatin. The ⌬1/AF2dm12/S573A receptor was also assayed and found to be ϳ50% as effective as C656G in inducing SacI access. This was likely a reflection of its impaired dose response rather than a contribution to remodeling because its EC 50 falls at 1 nM dexamethasone (Table I) 6. Effects of GR domain mutations on induction of nuclease accessibility. Nuclei were isolated from sorted cells expressing various GR mutants and subjected to digestion with SacI. DNA was purified and digested to completion with DpnII. Digestion products were detected as described under "Experimental Procedures." A, results were from a representative experiment. Dexamethasone (Dex) treatments were 1 h in length. B, data from at least three independent experiments were analyzed as follows. In each experiment fractional cleavage by SacI was calculated by dividing the intensity of the SacI digestion product by the added intensities of the SacI and DpnII products for each sample. The change in cleavage induced by dexamethasone for each receptor was calculated by subtracting fractional cleavage in the absence of dexamethasone from that in the presence of dexamethasone. This value was then converted to a percentage. Error bars represent S.E. that the helix 3 region of the GR LBD was also involved in chromatin remodeling. DISCUSSION Our study into the mechanism by which the GR activates the same promoter in different nucleoprotein contexts shows that the GR was a highly versatile protein that utilizes distinct domains and activation mechanisms to regulate transcription. A previous study reported that distinct domains are required for GR activation in different cellular contexts (47); whereas 1 was highly active in CV1 cells, it had much less transactivation capacity in HeLa cells where the LBD was much more active. They speculated that these differing domain dependences reflected varying cellular concentrations of cofactors. However, our study shows that this was true even within a particular cellular environment and was dependent on nucleoprotein context of a specific target promoter. In addition, we have defined a region of the GR important for productive interaction with machinery that catalyzes chromatin remodeling in mammalian cells and have provided evidence for a functionally important cross-talk between the AF-2 and 1 domains. The results of our study have implications for understanding the tissue-specific regulation of target genes by steroid receptors, which may be determined by the availability of various cofactors as well as the chromatin structure of the target promoter.
The GR domains required for various steps in activation of the two structurally distinct MMTV templates are summarized in Fig. 9. Activation of a transiently transfected MMTV promoter construct by the GR was largely dependent on the GR LBD in our cell lines. Mutation of amino acids in helices 3 or 12 significantly reduced activation. However, a mutation in the 2 region had no effect, which differs from the results of Milhon et al. (16). The disparity may lie in the species of receptor used (rat versus mouse) or the cell type assayed (mammary versus kidney origin). In contrast to the LBD mutations, deletion or mutation of 1 impaired activation of the transient template by only 20 -30%. The importance of the 1 region for activation of transiently transfected promoters has been found to vary depending on the cell line used (5,10,47,48).
Activation of the stably replicating MMTV promoter required both the 1 domain and amino acid residues in helices 3 and 12 of the AF-2 domain. Interestingly, the hydrophobic residues in 1 required for full activation of the transient template made only a small contribution to activation of the stable MMTV template. Thus the region of 1 necessary for activation in the context of ordered chromatin does not completely coincide with that required for activation of transient templates and interaction with the DRIP complex (13). This result strongly suggests that 1 interacts with different sets or domains of transcriptional cofactors at the two MMTV templates.
The 1 domain was not required for chromatin remodeling, which indicates that its main contribution to transactivation at the stable template was downstream of template derepression and NF1 binding, and may be mediated through interactions with the basal transcription machinery. Consistent with this idea is the fact that the 1 domain has been shown to interact with TBP and the TFIID complex (11,12) as well as CBP (9). The AF-2 helix 12 domain also functions downstream of template derepression. This was evidenced by the fact that the ⌬1/AF2dm12 receptor is a less efficient activator than the ⌬1 receptor even though it can fully remodel chromatin at the stable MMTV template. The helix 12 domain may thus participate in a common activation step at both MMTV templates.
Our study clearly shows that the GR LBD plays an important role in the induction of chromatin remodeling at the stable MMTV template. This is consistent with our previous work showing that a different LBD mutant of GR failed to induce remodeling (49). Mutation of helix 12 residues causes a significant impairment of the ability of the GR to induce chromatin remodeling and subsequent NF1 binding, but only in the presence of the 1 domain. This observation suggests that helix 12 is not directly necessary for productive interaction with remodeling machinery but may facilitate this process through crosstalk with 1. A mutation in LBD helix 3 either alone or in the context of the ⌬1/AF2dm12 receptor significantly impaired chromatin remodeling, suggesting that this region of the LBD plays a role distinct from that of helix 12. However, remodeling was never completely abolished. The helix 3 region may provide part of an interaction surface for the remodeling complex and/or facilitate its interaction with another domain. Previous studies indicated that the DBD of the GR may be important for interaction and function of SWI/SNF complexes (29,34). In addition, a recent report showed that the SWI/SNF complex makes physical and functional interactions with transcription factors, which like the GR, contain zinc finger DBDs (50). The proximity of helix 3 to the GR DBD leaves open the possibility that both of these domains may serve to recruit or interact with the remodeling machinery.
A recent report showed that 1 interacted physically and functionally with purified yeast SWI/SNF complex (15). Our results showing that 1 was dispensable for chromatin remodeling at the MMTV promoter in organized chromatin may differ based on subunit differences between yeast and mammalian SWI/SNF complexes and their interactions with other basic transcription complexes (51). The promoter context may also influence the function of various GR domains. In addition, it is possible that a complex other than SWI/SNF can remodel MMTV chromatin in mammalian cells (52). Our results are more in line with those of DiRenzo et al. (35), who showed a physical and functional interaction of human SWI/SNF complex with the LBD of estrogen receptor. However, their transactivation assays were carried out on transiently transfected reporters that may not undergo the type of remodeling observed at the MMTV promoter in organized chromatin. Our study is the only report to date that directly assays effects of various receptor domains on the induction of nuclease hypersensitivity at a target promoter in mammalian cells.
Our results have also provided evidence for a functionally important cross-talk between the 1 and AF-2 domains. First, the presence of the 1 domain has a deleterious effect on dose response of the GR containing mutations in helices 3 and/or 12 (AF2dm12 and AF2dm12/K597A) because when it is removed (⌬1/AF2dm12 and ⌬1/AF2dm12/K597A) the EC 50 improves to a value close to that of the C656G receptor (see Table I). This same effect was observed in chromatin remodeling assays as described above (AF2dm12 versus ⌬1/AF2dm12, Fig. 6). Although there is no evidence for a direct interaction between 1 and the GR LBD, some steroid receptor coactivators are capable of interacting with both domains (13,53,54). It is possible that coincident interaction of the two activation domains with each other and/or bridging proteins may cause a conformational change in the receptor that facilitates stable binding of ligand and allows another region of the GR to make contact with the remodeling machinery.
Why are different GR domains required for activation of the same promoter in distinct nucleoprotein contexts? It has been established that the nucleoprotein structure of the MMTV promoter influences its mechanism of activation by the GR (25). Incorporation of the promoter into organized chromatin necessitates remodeling and derepression prior to activation of transcription. A distinct set of factors is likely to be recruited to the MMTV promoter in organized chromatin to participate in remodeling and the subsequent transcriptional activation and elongation in the context of positioned nucleosomes and linker histones (55). This greater complexity may be reflected by the fact that both transcriptional domains of the GR (1 and the LBD) are required for activation of the stable template, whereas activation of the transient template is largely dependent only on the LBD. Another interesting possibility is that the interaction of GR with nucleosomes may serve as an allosteric effector, much like its interaction with various GREs (56,57). At a nucleosome the GR may present various domains necessary for activation in that context.