Agonist-dependent Repression Mediated by Mutant Estrogen Receptor (cid:1) That Lacks the Activation Function 2 Core Domain*

Nuclear receptor corepressor (N-CoR) and silencing mediator of retinoid and thyroid hormone receptors (SMRT) form heterogeneous complexes with various histone deacetylases (HDACs). In this report, we found that ER (cid:1) - (cid:2) AF2, a mutant estrogen receptor (cid:1) (ER (cid:1) ) deleted for the C-terminal activation function 2 (AF2) core domain, directs estradiol (E 2 )-dependent repression and impairs E 2 -induced transactivation by wild type ER (cid:1) . This repression required coexpressed BRG1 in SW-13 cells that lack BRG1, the ATPase constituent of the chro-matin-remodeling SWI (cid:1) SNF complex, and was abolished by HDAC inhibitor trichostatin A. We further demonstrated that ER (cid:1) - (cid:2) AF2 constitutively associates with SMRT but binds DNA in an E 2 -dependent manner in vivo . These results suggest that ER (cid:1) - (cid:2) AF2 and similar mutant receptors recently found associated with certain tumors may actively perturb the normal E 2 signaling via SWI/SNF, N-CoR/SMRT, and HDAC. The nuclear receptor superfamily is a group of ligand-de-pendent transcriptional regulatory proteins that function by binding to specific DNA sequences named hormone response elements in the promoters of target genes (for a review, see Ref. 1). The superfamily includes receptors for a variety

The nuclear receptor superfamily is a group of ligand-dependent transcriptional regulatory proteins that function by binding to specific DNA sequences named hormone response elements in the promoters of target genes (for a review, see Ref. 1). The superfamily includes receptors for a variety of small hydrophobic ligands such as steroids, T3, and retinoids as well as a large number of related proteins that do not have known ligands, referred to as orphan nuclear receptors. The C terminus of the ligand binding domain of these proteins harbors an essential ligand-dependent transactivation function, activation function 2 (AF2) 1 (1), whereas the N terminus of some nuclear receptors includes activation function 1. Genetic studies have implicated that transcription coregulators (or cofactors) with no specific DNA binding activity are essential components of transcriptional regulation that ultimately led to the identification of a series of nuclear receptor-interacting coregulatory proteins (for reviews, see Refs. [2][3][4]. These proteins appear to function by either remodeling chromatin structures and/or acting as adapter molecules between nuclear receptors and the components of the basal transcriptional apparatus. Transcriptional coactivators of nuclear receptors include the steroid receptor coactivator-1 (SRC-1) family, cAMP-response element-binding protein (CREB)-binding protein/p300, p/CAF, thyroid hormone receptor (TR)-associated protein/vitamin D3 receptor-interacting protein, activating signal cointegrator-1, activating signal cointegrator-2, and many others (2)(3)(4). Interestingly unliganded retinoic acid receptor (RAR) and TR bind to their target genes and repress transcription. Silencing mediator of RAR and TR (SMRT) and nuclear receptor corepressor (N-CoR) are known to mediate this repression (2)(3)(4). SMRT and N-CoR harbor transferable repression domains that associate with histone deacetylases (HDACs), consistent with the concept that histone hypoacetylation correlates with gene repression. SMRT and N-CoR were originally thought to exclusively act as adaptor molecules between target nuclear receptors and the mSin3⅐HDAC1 complex (5,6) and were subsequently shown to interact directly with class II HDAC4 and HDAC5 (7). In addition, multiple steady-state N-CoR/ SMRT complexes have recently been identified (8 -11). Enzymatically active HDAC3 complexes that contained both SMRT and N-CoR were isolated from HeLa nuclei (8,9). Two multiprotein N-CoR complexes, designated N-CoR-1 and N-CoR-2, were also isolated from HeLa nuclei (10). N-CoR-1 contained HDAC3, the SWI/SNF-related proteins BRG1, BAF 170, BAF 155, and BAF 47/INI1, and the corepressor KAP-1, whereas N-CoR-2 contained predominantly HDAC1 and HDAC2 as well as several other subunits that are found in the Sin3A complex (5,6). Similarly, Jones et al. (11) reported the presence of at least three distinct N-CoR complexes from Xenopus egg extract: one complex contained Sin3, Rpd3, and RbAp48, the second complex contained a Sin3-independent HDAC, and the third complex lacked HDAC activity. Steroid hormone receptors do not appear to interact with SMRT/N-CoR in the presence or absence of agonists, whereas both the estrogen receptor (ER) and the progesterone receptor can interact with these corepressors in the presence of their respective antagonists (12)(13)(14)(15). Similarly, binding of antagonists or the deletion of the AF2 domain is known to enhance the binding of N-CoR/SMRT to TR and RAR (16,17).
There is a large and increasing body of experimental and clinical data supporting the existence of variant ER proteins in both normal and neoplastic estrogen target tissues including human breast (for review, see Ref. 18). The functions of these variant ER proteins, either physiological or pathological, remain largely unclear. However, possible tissue-specific expression suggests that ER variants may have a role in tissuespecific estrogen action. In particular, ER variants lacking internal exons and representing dominant positive and negative activity may be involved in the initiation and/or progression of endocrine-dependent tumors. Interestingly a series of alterations and/or truncations in exon 8, which contains the AF2 core region, were identified with uterine tumor tissues (19) as well as a very aggressive and poorly differentiated form of breast cancer tissues recently isolated from African-American women (20).
In this report, we demonstrate that a mutant ER␣ deleted for the C-terminal AF2 core domain (ER␣-⌬AF2) directs basal repression and impairs transactivation by wild type ER␣, both in an estradiol (E 2 )-dependent manner, via SWI/SNF, N-CoR/ SMRT, and HDAC. Our results suggest that ER␣-⌬AF2 and similar receptors associated with certain tumors (19,20) can perturb the normal E 2 signaling, which may play an active role in cancerogenesis.

EXPERIMENTAL PROCEDURES
Plasmids and Ligands-Polymerase chain reaction fragments encoding ER␣-⌬AF2 were inserted into EcoRI and XhoI restriction sites of the LexA fusion vector pEG202PL, the B42 fusion vector pJG4-5, and the mammalian expression/in vitro translation vector pcDNA3. B42 and LexA fusions to the ER␣, mammalian expression vectors for SMRT, N-CoR, BRG1, and ER␣, the reporter constructs estrogen response element (ERE)-Luc and LexA-␤-gal, and the transfection indicator construct pRSV-␤-gal were as described previously (10,21,22). E 2 and tamoxifen (TAM) were purchased from Sigma, and trichostatin A was from Wako Pure Chemical Industries, Ltd. (Osaka, Japan).
Cell Culture and Transfection-CV-1 or SW-13 (23) cells were grown in 24-well plates with medium supplemented with 10% charcoalstripped serum. After a 24-h incubation, cells were transfected with 100 ng of ␤-galactosidase expression vector pRSV-␤-gal and 100 ng of ERE-Luc reporter gene along with expression vectors for ER␣, ER␣-⌬AF2, N-CoR, and BRG1. Total amounts of expression vectors were kept constant by adding decreasing amounts of pcDNA3 to transfections. Twelve hours later, cells were washed and refed with Dulbecco's modified Eagle's medium containing 10% charcoal-stripped fetal bovine serum. After 12 h, cells were left unstimulated or were stimulated with 0.1 M of the indicated ligand. Cells were harvested 24 h later, and luciferase activity was assayed as described previously (24), and the results were normalized to the ␤-galactosidase expression. Consistent results were obtained in more than two similar experiments.
Yeast Two-hybrid Tests-The cotransformation and ␤-galactosidase assays in yeast were performed as described previously (24). For each experiment, at least three independently derived colonies were tested.
Gel Mobility Shift Assays-Gel mobility shift assays were performed as described previously (24). Radiolabeled ERE oligonucleotides were incubated with in vitro translated ER␣ and ER␣-⌬AF2, and the reaction products were analyzed by native polyacrylamide gel electrophoresis and autoradiography as described previously (24).
Limited Proteolyses-ER␣ and ER␣-⌬AF2 were in vitro translated/ labeled with [ 35 S]Met, and the digestions were done for 10 min at room temperature with 50 g/ml trypsin in the presence of the increasing amount of the indicated ligands. The reaction products were analyzed by sodium dodecylsulfate-polyacrylamide gel electrophoresis and autoradiography as described previously (25).
Immunoprecipitation and Chromatin Immunoprecipitation-293T cells were transfected with ER␣, ER␣-⌬AF2, or SMRT expression vector and treated with E 2 or TAM for 1 h. Cells were harvested and immunoprecipitated with a monoclonal antibody against SMRT (a kind gift of Dr. Dean Edwards at University of Colorado Health Sciences Center) and blotted with monoclonal antibody against ER␣ (Santa Cruz Biotechnology). For chromatin immunoprecipitation, soluble chromatin from these cells was prepared and immunoprecipitated with ER␣ antibody as recently described (26). The final DNA extractions were amplified using pairs of primers that cover the ERE promoter region. The experiments were repeated at least three times and were highly reproducible.  (19,20) in tumorigenesis, we constructed a mutant ER␣ that is deleted for the C-terminal 58 amino acids containing the core AF2 region (i.e. ER␣-⌬AF2). Interestingly ER␣-⌬AF2 repressed transactivation mediated by the EREdriven luciferase reporter construct in an E 2 -dependent manner, whereas the wild type ER␣ directed E 2 -dependent transactivation as expected (Fig. 1A). In addition, ER␣-⌬AF2 impaired E 2 -mediated transactivation by the wild type ER␣ in a dose-dependent and dominant negative fashion (Fig. 1B). These results suggest that ER␣-⌬AF2 and similar mutant ERs associated with tumors (19,20) may actively perturb the normal E 2 signaling within the cell, which may play a role in the possible tumorigenesis by these receptors.

RESULTS AND DISCUSSION
Constitutive Recruitment of N-CoR/SMRT and E 2 -dependent DNA Binding by ER␣-⌬AF2-To understand how ER␣-⌬AF2 achieves this repression, we examined various molecular properties of this mutant receptor. First, LexA fusions to ER␣-⌬AF2 and ER␣ enhanced transactivation mediated by B42 fusions to ER␣, ER␣-⌬AF2, and ER␤ in a ligand-dependent manner ( Fig. 2A). It was interesting to note that LexA-ER␣ stimulated transactivation mediated by B42-ER␤ even in the absence of ligand. These results clearly demonstrate that ER␣-⌬AF2 binds ligands and still retains the ability to dimerize with ERs. Second, ER␣-⌬AF2 was found to bind ERE either as a homodimer or heterodimer with ER␣ as demonstrated by gel mobility shift assays (Fig. 2B). Next we examined the tertiary structure of ER␣-⌬AF2 by using a limited proteolysis experiment (25). Two previously described fragments of 31 and 28 kDa (25) as well as a smaller fragment of ϳ10 kDa were generated in an E 2 -dependent manner from the full-length, [ 35 S]Met-labeled ER␣ when subjected to 50 g/ml trypsin (Fig.  3A). In contrast, only the 28-kDa fragment was visible when the ER antagonist TAM was used. Interestingly ER␣-⌬AF2 produced a pattern similar to ER␣/TAM in the presence of either E 2 or TAM, producing a fragment of ϳ28 kDa. Thus, ER␣-⌬AF2, in the presence of either E 2 or TAM, may adopt a conformation that resembles that of TAM-bound ER␣. From these results along with the results in which N-CoR/SMRT bindings were shown to be enhanced either by antagonists or the deletion of the AF2 domain with TR and RAR (16,17) and with steroid hormone receptors, observed only in the presence of antagonists (12-15), we hypothesized that ER␣-⌬AF2 may recruit corepressor SMRT/N-CoR in an E 2 -dependent manner. To test this idea, 293 cells cotransfected with expression vectors for the full-length ER␣ and SMRT either in the absence or presence of E 2 or TAM were immunoprecipitated with SMRT antibody and probed with an antibody against ER␣ in Western analysis (Fig. 3B). ER␣ exhibited TAM-dependent interactions as reported previously (12)(13)(14)(15). Surprisingly, however, ER␣-⌬AF2 constitutively interacted with SMRT. Similar results were also obtained with the glutathione S-transferase pull-down assays (results not shown). Overall these results were somewhat similar to the case with mutant TR and RAR deleted for the AF2 domain (16,17). It is important to note that ER␣-⌬AF2 was not capable of binding SRC-1 in the glutathione S-transferase pull-down and yeast two-hybrid assays as expected (results not shown). To further examine whether ER␣-⌬AF2 binds ERE in an E 2 -dependent manner in vivo, as it was recently shown with ER␣ (26), we used chromatin immunoprecipitation assays. As shown in Fig. 3C, both ER␣ and ER␣-⌬AF2 were recruited to ERE in an E 2 -or TAM-dependent manner. Taking all of these results together, we suspect that ER␣-⌬AF2 constitutively associates with SMRT/N-CoR in solution, and this complex is recruited to ERE in a strictly ligand-dependent manner in vivo, leading to the observed E 2 -dependent repression (Fig. 1). As expected from this prediction, ER␣-⌬AF2 also repressed ERE-dependent transactivation in a TAM-dependent manner (results not shown).
Requirement of BRG1 and HDAC in ER␣-⌬AF2-mediated Repression-Multiple steady-state complexes containing N-CoR/SMRT and distinct HDAC proteins have recently been isolated (8 -11). In particular, N-CoR-1 contained HDAC3, the SWI/SNF-related proteins BRG1, BAF 170, BAF 155, and BAF 47/INI1, and the corepressor KAP-1 that is involved in silencing heterochromatin (10). When we examined the transcriptional properties of ER␣ and ER␣-⌬AF2 in SW-13 cells (23) that lack BRG1 and the similar protein hBRM, the -fold repression observed with both receptors was negligible (Fig. 4A). However, the previously described TAM-dependent repression by ER␣ was slightly increased in the presence of cotransfected BRG1 expression vector and synergistically increased when BRG1 was coexpressed with N-CoR (Fig. 4A, left). Similarly the E 2 -dependent repression by ER␣-⌬AF2 was also enhanced by cotransfected BRG1 (Fig. 4A, right). As expected, trichostatin A, a HDAC inhibitor, abolished the E 2 -dependent repression by ER␣-⌬AF2 (Fig. 4B). Overall these results suggest that the E 2 -dependent repression by ER␣-⌬AF2 may function through recruitment of a distinct N-CoR complex containing both HDAC and BRG1, such as N-CoR-1 (10), although other similar complexes yet to be identified may also play a role in this repression. Alternatively multiple N-CoR/SMRT complexes (i.e. BRG1 and non-BRG1 complexes) may function in sequence or combination. The SWI/SNF family of nucleosome-remodeling complexes has been shown to play important roles in gene expression throughout eukaryotes via establishing a local chromatin structure that is permissive for increased access of transcription factors for their binding sites. Notably SWI/SNF is required for both transcriptional activation and repression of some genes. Our results, along with the recent results in which BRG1 was shown to be essential for E 2 -dependent transactivation of ERs (27), suggest that SWI/SNF is also involved with both activation and repression by ER and mutant ERs such as ER␣-⌬AF2.
Importantly our results unraveled a novel type of functional transformationinwhichaspecificmutationconvertedanagonistdependent transcriptional activator (i.e. ER␣) into a potent, FIG. 2. Dimerization and DNA binding properties of ER␣-⌬AF2. A, the indicated B42 and LexA plasmids were transformed into a yeast strain containing an appropriate LacZ reporter gene as described previously (24). Closed, open, and shaded boxes indicate the presence of no ligand, 0.1 M E 2 , and 0.1 M TAM, respectively. The data are representative of at least two similar experiments. B, gel mobility shift assays were executed as described previously (24). Shown is the specific retarded band, which is competed by unlabeled ERE oligonucleotides (lanes 5 and 6) but not with unrelated p53 oligonucleotides (results not shown). When ER␣ and ER␣-⌬AF2 are mixed, the predominant species appears to be a heterodimeric complex (lane 3) based on the mobility of the retarded band, which runs between ER␣ and ER␣-⌬AF2 homodimers (lanes 2 and 4).

FIG. 3. Constitutive association with SMRT and E 2 -dependent DNA bindings of ER␣-⌬AF2.
A, limited proteolyses were executed with radiolabeled ER␣ and ER␣-⌬AF2 as described previously (25). The digestions were done for 10 min at room temperature with 50 g/ml trypsin in the presence of the increasing amounts of either E 2 or TAM (10, 100, and 1000 nM, respectively) (25). Approximately 40% of the total reaction was loaded as input. B, 293T cells transfected with expression vectors for SMRT and either ER␣ or ER␣-⌬AF2 were immunoprecipitated with SMRT monoclonal antibody and blotted against ER␣ antibody. 100 nM E 2 and TAM were used as indicated. I.P., immunoprecipitation; W, Western blot. C, chromatin immunoprecipitation (ChIP). 293T cells were transfected with ER␣ or ER␣-⌬AF2 expression vector and treated with 100 nM of the indicated ligand. Soluble chromatin from these cells was prepared and immunoprecipitated with monoclonal antibody against ER␣ as described previously (26). The final DNA extractions were amplified using pairs of primers that cover the ERE promoter region. The experiments were repeated at least three times and were highly reproducible. agonist-dependent repressor (i.e. ER␣-⌬AF2). With regard to this switch of function by mutations, it is interesting to note that resistance to thyroid hormone can result from an aberrant interaction between mutant TR␤1 and N-CoR/SMRT (28 -30). For instance, a natural TR␤1 mutant (G345R) with poor T3 binding affinity, originally isolated from resistance to thyroid hormone, formed a TR␤1⅐N-CoR complex, both in the absence and presence of T3, but could not form a TR␤1⅐SRC-1 complex (29). An artificial TR␤1 mutant, which bound T3 with normal affinity and containing a mutation in the AF2 core region (E457D), had normal interactions with N-CoR but could not bind SRC-1 (29). Similarly another mutant, TR␤1, derived from resistance to thyroid hormone (L454S), bound T3 and weakly interacted with SRC-1 in the presence of T3 (30). In contrast, in the absence of T3, the L454S mutant interacted much more strongly with N-CoR than did the wild type receptor, and the T3-dependent release of N-CoR was markedly impaired (30). All of these mutants had strong dominant negative activity on wild type TR transactivation (28 -30). Interestingly treatment with retinoids is known to induce differentiation of leukemic blast cells and disease remission in patients with promyelocytic leukemia RAR␣ fusion but not from patients with promyelocytic leukemia zinc finger-RAR␣ due to the ability of the latter to bind N-CoR much more tightly via a second, retinoid-resistant N-CoR-binding site in the promyelocytic leukemia zinc finger N-terminal region (31). As exemplified from these results, it is noted that altered bindings to transcriptional coactivators and corepressors could result in perturbation of the normal cellular signalings and serious diseases. Accordingly both specific mutations and fusions/translocations to other proteins with switched functions are expected to exist with other members of the nuclear receptor superfamily. Finally it will be interesting to test whether ER␣-⌬AF2 as well as those mutant ERs associated with tumors (19,20) can directly cause cancers when ectopically overexpressed in mice.
In summary, we found that ER␣-⌬AF2 directs E 2 -dependent repression and impairs E 2 -mediated transactivation by wild type ER␣ via N-CoR/SMRT, BRG1, and HDAC. Perturbation of the normal E 2 signaling by ER␣-⌬AF2 and similar mutant receptors associated with tumors (19,20) may represent a novel mechanism for the tumorigenesis processes in vivo.