Characterization of the Physical Interaction between Estrogen Receptor (cid:1) and JUN Proteins*

Activated estrogen receptor (cid:1) (ER (cid:1) ) modulates transcription triggered by the transcription factor activator protein-1 (AP-1), which consists of Jun-Jun homodimers and Jun-Fos heterodimers. Previous studies have demonstrated that the interference occurs without binding of ER (cid:1) to DNA but probably results from protein (cid:1) protein interactions. However, involvement of a direct interaction between ER (cid:1) and AP-1 is still debated. Using glutathione S -transferase pull-down assays, we demonstrated that ER (cid:1) bound directly to c-Jun and JunB but not to FOS family members , in a ligand-independent manner. The interaction could occur when c-Jun was bound onto DNA, as shown in a protein-protein-DNA assay. It implicated the C-terminal part of c-Jun and amino acids 259– 302 present in the ER (cid:1) hinge domain. ER (cid:1) but not an ER (cid:1) mutant deleted of amino acids 250–303 (ER241G), also associated with c-Jun in intact cells, in the presence of estradiol, as shown by two-hybrid and coimmunoprecipitation assays. We also show that ER (cid:1) , c-Jun, and the p160 coactivator GRIP1 can form a multiprotein complex in vitro and in intact cells and that the ER (cid:1)

Activated estrogen receptor ␣ (ER␣) modulates transcription triggered by the transcription factor activator protein-1 (AP-1), which consists of Jun-Jun homodimers and Jun-Fos heterodimers. Previous studies have demonstrated that the interference occurs without binding of ER␣ to DNA but probably results from protein⅐protein interactions. However, involvement of a direct interaction between ER␣ and AP-1 is still debated. Using glutathione S-transferase pull-down assays, we demonstrated that ER␣ bound directly to c-Jun and JunB but not to FOS family members, in a ligand-independent manner. The interaction could occur when c-Jun was bound onto DNA, as shown in a protein-protein-DNA assay. It implicated the C-terminal part of c-Jun and amino acids 259 -302 present in the ER␣ hinge domain. ER␣ but not an ER␣ mutant deleted of amino acids 250 -303 (ER241G), also associated with c-Jun in intact cells, in the presence of estradiol, as shown by two-hybrid and coimmunoprecipitation assays. We also show that ER␣, c-Jun, and the p160 coactivator GRIP1 can form a multiprotein complex in vitro and in intact cells and that the ER␣⅐c-Jun interaction could be crucial for the stability of this complex. VP16-ER␣ and c-Jun, which both interact with GRIP1, had synergistic effect on GAL4-GRIP1-induced transcription in the presence of estradiol, and this synergistic effect was not observed with the ER␣ mutant VP16-ER241G or when c-Fos, which bound GRIP1 but not ER␣, was used instead of c-Jun. Finally, ER241G was inefficient for regulation of AP-1 activity, and an ER␣ truncation mutant encompassing the hinge domain had a dominant negative effect on ER␣ action. These results altogether demonstrate that ER␣ can bind to c-Jun in vitro and in intact cells and that this interaction, by stabilizing a multiprotein complex containing p160 coactivator, is likely to be involved in estradiol regulation of AP-1 responses.
Estrogens play a pivotal role in the control of growth and differentiation of estrogen target tissues. Their action is mediated through estrogen receptors (ER), 1 which belong to a su-perfamily of nuclear receptors that act as ligand-activated transcription factors and can be subdivided to six regions (A-F) exhibiting different degrees of evolutionary conservation (1). Domain C encompasses the highly conserved DNA binding domain (DBD). The moderately conserved region E contains the ligand binding domain (LBD) and a ligand-dependent transcription activation function (AF-2). The quite divergent A/B domains contain, in some nuclear receptor members such as ER, a transcription activation function (AF-1), which can activate transcription constitutively in the absence of ligand. Upon binding to their cognate ligands, nuclear receptors activate transcription by interacting with specific DNA sequences present in target gene promoters (reviewed in Refs. 2,3). Coactivators, among them the cAMP-response element-binding protein CBP/p300 and a group of highly related molecules called p160 proteins, comprising SRC-1, TIF2/GRIP1, and RAC3/p/CIP/ AIB1/ACTR, associate with receptors in a ligand-and AF-2-dependent manner to enhance their transactivation potential. They function as bridging proteins to the components of the basal transcriptional machinery, and some of them, such as CBP/p300, SRC-1, and ACTR, possess an intrinsic histone acetyltransferase activity that could influence the accessibility of transcription factors to the chromatin template (reviewed in Refs. 4,5).
Nuclear receptors also modulate transcription without receptor⅐DNA interaction by functional interference with other transcription factors such as activating protein-1 (AP-1) (reviewed in Refs. 6 -9). AP-1, which is implicated in diverse cellular processes, including differentiation, cell proliferation, and transformation (reviewed in Ref. 10), predominantly consists of various combinations of JUN (c-Jun, JunB, JunD) and FOS (c-Fos, Fra-1, Fra-2, FosB) proteins. JUN proteins can form homodimers or more stable heterodimers with proteins of the FOS family that do not homodimerize. Jun⅐Jun and Jun⅐Fos dimers regulate gene transcription through interactions with a specific DNA sequence, the TPA-responsive element (TRE) (11)(12)(13). We and others have previously shown that estradiol could modulate AP-1-dependent transcription (14 -18). Estrogen regulation of AP-1 activity is generally positive (14 -17), although it can also be negative in breast cancer cells expressing high Fra-1 level (18). ER␣ whose expression is necessary for the estradiol effect, does not bind to TRE (14). In addition, ER␣ bearing a point mutation in the first zinc finger (18) or complete deletion of the DBD (14,17) was shown to be efficient in regulating AP-1 responses, thus demonstrating that modulation of AP-1-dependent transcription is directly induced by the activated receptor. These data together suggest that the mechanism by which estrogens regulate AP-1 activity is triggered by protein⅐protein interactions. A physical interaction between c-Jun and nuclear receptors has been proposed to be responsible for negative or positive cross-talk between nuclear receptor and AP-1 (17, 19 -25). Little is known, however, about the domains of nuclear receptors involved in the physical interaction with c-Jun or about the actual role of this direct interaction in the regulation of AP-1-dependent transcription. C-Jun and c-Fos also directly interact with coactivators CBP/ p300 (26) and SRC-1 (27), and these transcriptional integrators regulate AP-1 activation of transcription, suggesting their involvement in nuclear receptor⅐AP-1 cross-talk. It has been proposed that the mutual inhibition observed between some nuclear receptors and AP-1 depends on the competition for limited amounts of CBP/p300 (28). In addition, some coactivators likely participate in the positive interference between ER␣ and AP-1, because deletion of ER␣ helix12 (29) or mutations in AF-2 that prevent binding of p160 coactivators drastically inhibit estradiol regulation of AP-1 activity (29). 2 In the present study, we provide evidence that ER␣ binds to c-Jun in vitro and in vivo and further characterize the physical interaction between ER␣ and AP-1 factors. We also show that this interaction, which participates in a multiprotein complex containing the p160 coactivator GRIP1, is likely to be involved in estradiol regulation of AP-1 responses.
Transient Transfection, CAT, and Luciferase Assays-Twenty-four hours after plating, the medium was changed and cells were transfected for 16 h using the calcium phosphate DNA coprecipitation method as previously described (15). When cells were transfected by an expression vector, the same amount of empty vector was transfected in control cells. One microgram of the ␤-galactosidase expression plasmid pCMV ␤ (CLONTECH Laboratories, Palo Alto, CA) for MCF7 cells, and 2 g of the ␤-galactosidase expression plasmid PCH110 (Amersham Pharmacia Biotech) for COS cells, were used for internal control of transfection efficiency. PSPT19 DNA was added up to 5 g of total DNA per well. Cells were washed twice with phenol red-free medium and treated, as indicated, for 24 h in phenol red-free medium containing 1% DCC for MCF7 cells and 3% DCC for COS cells. CAT enzyme assays were performed in whole cell extracts after normalization for ␤-galactosidase activity (15). Acetylated and nonacetylated forms of [ 14 C]chloramphenicol were separated by TLC. Quantification was performed with a Fuji BAS1000 Bioimaging Analyzer (Raytest, Paris, France). For luciferase assays, cells were lysed for 15 min in the cell culture lysis reagent from Promega. Luciferase activity was measured using an LKB luminometer (LKB Instruments, Rockville, MD) and normalized for ␤-galactosidase activity as described by Roux et al. (44).
Expression, Purification of GST Fusion Proteins, and GST Pull-Down-Overnight cultures of E. coli transformed with parental or recombinant pGEX plasmids were diluted 1:10 in L-broth with 50 g/ml ampicillin and incubated at 37°C with shaking to an A 600 of 0.5. Isopropyl-␤-D-thiogalactopyranoside was then added to a final concentration of 0.1 mM. After a further 3-5 h of growth, cells were pelleted at 5000 ϫ g for 10 min at 4°C and resuspended in a 1:5 (v/v) solution for plasmids recombinants and in 1:10 (v/v) for parental plasmid of the original culture volume of NETN (0.5% Nonidet P-40, 1 mM EDTA, 20 mM Tris, pH 8, 100 mM NaCl) containing proteases inhibitors (Complete, Roche Molecular Biochemicals). Cells were then sonicated and centrifuged at 10,000 ϫ g for 5 min at 4°C. GST fusion protein suspension beads (50 l) were incubated overnight at 4°C with 35 S-labeled proteins generated by the TnT in vitro transcription-coupled translation system from Promega. After three washes with NETN, samples were boiled in 2ϫ SDS sample buffer and analyzed by SDS-PAGE. Signals were amplified by fluorography (Amplify, Amersham Pharmacia Biotech) and gels exposed at Ϫ80°C. Quantification of 35 S proteins was performed with a Fuji BAS1000 Bioimaging Analyzer (Raytest, Paris, France).
Protein-Protein-DNA Assay-Protein-Protein-DNA assay was performed as described by Thénot et al. (45). The double-stranded oligonucleotide, corresponding to the collagenase TRE (18), was labeled by Klenow enzyme in the presence of [ 32 P]dCTP. C-Jun-primed reticulocyte lysate (15 l) was preincubated with the TRE (2 nM) in TKE buffer (10 mM Tris, 75 mM KCl, 0.5 mM EDTA) plus 0.5 mM dithiothreitol, 0.1 g/l poly(dIdC) and protease inhibitors. GST fusion proteins preloaded on glutathione-Sepharose and resuspended in TKE were then added, and binding reactions were performed overnight at 4°C. After two washes in TKE, bound molecules were analyzed on a 12% polyacrylamide denaturing gel and visualized by autoradiography.
Immunoprecipitation and Immunoblotting-For immunoprecipitation, transfected COS cells were harvested in lysis buffer containing 20 mM HEPES (pH 7.5), 0.4 M KCl, 0.1 mM EDTA, 0.1% Nonidet P-40, and a mixture of protease inhibitors (Complete, Boehringer). Cell lysates were clarified by centrifugation before incubation overnight at 4°C with monoclonal antibodies (clone 1D5, Dako, Glostrup, Denmark) reacting with the A/B domain of ER␣ (dilution 1:40). Pre-washed protein G-Sepharose (Amersham Pharmacia Biotech) was then added and the incubation continued for 2 h at 4°C. Immunoprecipitates were recovered by centrifugation, washed four times in lysis buffer, and resolved by SDS-PAGE. Proteins were analyzed by Western blotting using polyclonal anti-c-Jun rabbit antibodies (N-G, Santa Cruz Biotechnology, Santa Cruz, SA, dilution 1:300) followed by horseradish peroxidaseconjugated goat anti-rabbit immunoglobulin G (Sigma-Aldrich, Saint Quentin Fallavier, France, dilution 1:4000). Signals were visualized by chemiluminescence (Renaissance, PerkinElmer Life Sciences, Le Blanc Mesnil, France).
Purification of Hexahistidine Fusion Proteins-Overnight cultures of E. coli transformed with pDS56-c-Jun and pDS56-c-Fos (46) were diluted 1:20 in L-broth and incubated at 37°C with shaking to an A 600 of 0.6. Isopropyl-␤-D-thiogalactopyranoside was then added to a final concentration of 0.5 mM. After a further 4 h of growth, cells were pelleted at 5000 ϫ g for 10 min at 4°C and resuspended in a 1:20 (v/v) solution of NETN. Cells were then sonicated and centrifuged at 10,000 ϫ g for 5 min at 4°C. The pellet was then dissolved in the same volume of inclusion body solubilization reagent (Pierce, Rockford, IL). After 20min incubation at 4°C, the solution was centrifuged for 30 min at 15,000 ϫ g, and the supernatant was incubated with nickel-nitrilotriacetic acid silica (Qiagen, Courtaboeuf, France) for 1 h at room temperature. After three washes with DWB buffer (6 M urea, 20 mM NaH 2 PO 4 , 500 mM NaCl, pH 8), recombinants proteins were eluted in the same buffer in the presence of 0.3 M imidazole. The purified proteins were then dialyzed against 6 M urea for 6 h, and 25 mM Tris, pH 7.5, was added, eight times every 2 h, until an urea concentration of 2 M was reached. Finally, proteins were dialyzed for another 6 h against 25 mM Tris, 150 mM NaCl. The purity of histidine-tagged Fos and Jun was ϳ95% as determined by SDS-PAGE. Protein renaturation was verified by gel retardation assay using a consensus collagenase TRE, as previously described (18).

ER␣ Directly Interacts with c-Jun in Vitro-
To test whether c-Jun associates directly with ER␣ in vitro, we first performed glutathione S-transferase (GST) pull-down experiments in which GST and GST-c-Jun fusion proteins, preloaded on glutathione-coupled beads, were incubated with in vitro translated ER␣. As shown in Fig. 1B, 35 S-labeled ER␣, which was not retained by GST, associated with the bead-bound GST-c-Jun fusion protein. The ability of 35 S-labeled in vitro translated c-Jun to interact with a GST-ER␣ fusion protein was also tested in a reciprocal experiment. Contrary to GST-c-Jun, only a small fraction of the hybrid ER␣ protein was expressed as a full-length protein. Although efficacy of interaction between c-Jun and ER␣ was lower than in the reciprocal experiment, the assay confirmed the direct in vitro binding between the two proteins (data not shown).
In vitro translated ER␣ mutant proteins (Fig. 1A) were then analyzed for binding to GST-c-Jun to localize the ER␣ domain(s) required for interaction with c-Jun (Fig. 1B). Deletion of the C-terminal part of ER␣ (mutant HE15) totally abolished binding with the fusion protein. On the contrary, the ER␣ mutant protein HE19, deleted of the N-terminal part of ER␣, was still able to interact with GST-c-Jun. HE11, deleted of the entire DBD, also bound to GST-c-Jun although less efficiently than HEO or HE19. To assess the role of ER␣ domains in ER␣-mediated regulation of AP-1 activity, increasing concentrations of the same ER␣ deletion constructs (1) were transfected in MCF7 cells together with the (AP-1)4-TK-CAT reporter plasmid (Fig. 1C). Mutant HE19, lacking AF-1, was as efficient as HEO in increasing the hormonal effect. In contrast, mutant HE15, deleted of the LBD and AF-2, and, in agreement with previous results (18), mutant HE11, lacking the DBD, had no effect on AP-1 activity in MCF7 cells. Both LBD and DBD therefore appeared to be important in ER␣-mediated regulation of AP-1 activity in these cells.
The ER␣ Hinge Domain Is Implicated in the Protein⅐Protein Interaction-To more accurately define the borders of c-Jun binding sites on ER␣, a series of GST-ER␣ deletion mutants were tested in pull-down experiments (Fig. 2). The GST-ER-(2-184) fusion protein, which only contains the A/B ER␣ domain, did not interact with radiolabeled c-Jun, in agreement with results obtained in the reciprocal experiment (Fig. 1B). In fact, deletion of 250 amino acids from the ER␣ N terminus (hybrid protein GST-ER-(251-595)) did not impair the interaction. By contrast, deletion of amino acids 251-312 (compare results obtained with GST-ER-(251-595) and GST-ER-(313-599)) totally abolished c-Jun binding, indicating that an important motif is localized in the ER␣ hinge region (domain D). Conversely, binding of c-Jun to GST-ER-(179 -312) and GST-ER-(251-312) demonstrated that the C terminus of ER␣ was also dispensable. The fact that the binding efficiency of c-Jun to both fusion proteins was equivalent also showed that DBD did not participate in the protein⅐protein interaction. Finally, c-Jun binding was retained by protein GST-ER-(259 -302) but not by GST-ER-(283-330).
ER␣ Interacts with the C-terminal Domain of c-Jun but Not with Fos Proteins-To specify the c-Jun domain(s) involved in the interaction with ER␣, several c-Jun deletion mutants were translated in vitro in the presence of [ 35 S]methionine and tested in GST pull-down assays for their ability to bind GST-ER-(251-595) (Fig. 3). C-Jun mutants deleted of amino acids 6 -194 or 146 -221 still bound to the GST-ER␣ fusion protein.
In agreement with these results, the c-Jun N terminus (mutant ⌬224 -334) alone did not interact with GST-ER-(251-595). By contrast, deletion of the 238 residues from the N terminus only moderately affected this interaction. We therefore conclude These bands most likely correspond to Ser-63, Ser-73, or both phosphorylated forms of c-Jun, as previously described by Bannister et al. (26). In fact, they were only detected for c-Jun mutants containing the N terminus part of the protein (Fig. 3). All phosphorylated forms of c-Jun bound to the same extent to GST-ER-(251-595) (Fig. 3) and to all GST-ER␣ fusion proteins containing residues 259 -302 (Fig. 2), indicating that, at least in our in vitro GST pull-down assay, phosphorylation of Ser-63 or Ser-73 did not modify the interaction with ER␣. The bZIP region is highly conserved between members of the Jun family (38). Because our results showed that the c-Jun C terminus was implicated in the in vitro interaction with ER␣, we analyzed the ability of JunD and JunB to bind ER␣. As shown in Fig. 4, JunB was as efficiently retained by GST-ER-(251-595) as c-Jun (for both proteins, 15-25% of the total input was specifically bound to the hybrid protein in at least four experiments). On the contrary, JunD only weakly hybridized with the fusion protein (1-4% of the total input in four independent experiments). The potential of Fos proteins to physically interact with ER␣ was also examined. Neither c-Fos nor FosB, Fra-2 or Fra-1 significantly interacted with the GST-ER-(251-595) fusion protein. We also used pull-down experiments with GST-ER-(2-184) or GST-ER-(179 -312) to investigate the possibility that Fos proteins bind an ER␣ domain other than that bound by Jun proteins. We did not detect any specific interaction of Fos proteins with both fusion proteins (data not shown).
ER␣ Interacts with c-Jun Bound onto DNA-To determine whether ER␣ could interact with c-Jun when AP-1 complexes were bound onto DNA, a protein-protein-DNA binding assay was then performed. 35 S-Labeled in vitro translated c-Jun was preincubated with a 32 P-labeled double-stranded oligonucleotide containing the collagenase TRE before it was tested for its ability to bind GST-ER-(251-595). As shown in Fig. 5, in the presence of c-Jun, the labeled TRE was retained onto GST-ER-(251-595) but not onto GST. Retention of TRE by the fusion protein was c-Jun mediated, because no specific binding of TRE to GST-ER-(251-595) was observed using unprogrammed reticulocyte lysate. Moreover, addition of 50 nM cold TRE did not significantly modify the in vitro interaction between ER␣ and c-Jun (not shown) demonstrating that, at least in vitro, binding of c-Jun on DNA did not influence the ER␣⅐c-Jun physical interaction.
In Vitro Effect of ER␣ Ligands-Because ER␣ regulation of AP-1 activity was dependent on the presence of ligand, we then examined the ability of estradiol and antiestrogens to modulate the in vitro physical interaction between ER␣ and c-Jun. Interaction of ER␣ with the coactivator SRC-1 (47), which was reported to be hormone-dependent, was tested in parallel as a control. In vitrotranslated c-Jun or SRC-1 were incubated with GST-ER-(251-595) in the presence of 1 M 17␤ estradiol, 4-hydroxytamoxifen, or ICI164,384, or in the absence of ligands. As shown in Fig. 6, only modest binding of SRC-1 to ER␣ was observed when the receptor  35 S-labeled c-Jun deletion mutants were incubated overnight with GST (lanes 2, 5,8,11,14) or the GST hybrid protein GST-ER-(251-595) (lanes 3, 6,9,12,15) in a GST pull-down assay as described under "Experimental Procedures" and Fig. 1. Inputs (lanes 1, 4, 7, 10, and 13) represent 10% of the different radiolabeled c-Jun mutants used in the assays. was either free or occupied with either antiestrogen as compared with the strong binding detected in the presence of estradiol. In contrast, significant amounts of c-Jun bound to ER␣ irrespective of whether the receptor was unoccupied or occupied with agonist or antagonists.

FIG. 3. Mapping of the c-Jun interaction domain. A, schematic representation of c-Jun mutants used in B. B, in vitro
Interaction between ER␣ and c-Jun in Mammalian Cells-The interaction between ER␣ and c-Jun in intact cells was evaluated using a mammalian cell two-hybrid system. Fulllength human ER␣ was fused to the transcriptional activator VP16 (VP16-ER␣) and c-Jun to the DNA binding domain of GAL4 (GAL4-c-Jun). Expression vectors for the hybrid proteins were cotransfected with a GAL4-responsive luciferase reporter (pG5-luc) in COS cells (Fig. 7A). As expected, c-Jun increased luciferase activity when tethered to DNA by the GAL4 DBD, due to the intrinsic c-Jun transactivating activity. VP16-ER␣ did not have any significant effect either in the absence or presence of estradiol or antagonists 4-hydroxytamoxifen and ICI164,384. However, when VP16-ER␣ and GAL4-c-Jun were coexpressed, GAL4-c-Jun transcriptional activity was enhanced after estradiol addition but not after antihormone treatment or in control cells. Moreover, both antiestrogens inhibited estradiol-induced luciferase activity. The two-hybrid system does not distinguish whether the c-Jun⅐ER␣ interaction is direct or mediated by another unknown factor in assembling a multiprotein complex with c-Jun and ER␣. To evaluate the importance of the direct interaction between ER␣ and c-Jun in the observed enhancement of luciferase activity, an ER␣ mutant deleted of amino acids 250 -303 (ER241G (32)) and unable to bind c-Jun in vitro (not shown) was fused to the transcriptional activator VP16 and used in the same experiment. This ER␣ mutant, which was mostly nuclear in the presence of estradiol (not shown), was totally inefficient in increasing GAL4-c-Jun transcriptional activity. The association between ER␣ and c-Jun was further investigated by coimmunoprecipitation. COS cells were cotransfected with c-Jun and ER␣ expression vectors. Proteins associated with ER␣ were first precipitated with monoclonal antibodies directed against the A/B domain of the receptor and subsequently analyzed by immunoblotting with c-Jun-specific antibodies. As shown in Fig. 7B, c-Jun protein was detected in immunoprecipitates from cells transfected with the wild-type ER␣ expression vector (HEGO) and cultivated in the presence but not in the absence of estradiol. The same experiment was also performed using ER241G FIG. 4. In vitro interactions between ER␣ and Jun and Fos family members. c-Jun, JunB, JunD, c-Fos, FosB, Fra-2, and Fra-1 were labeled with [ 35 S]methionine by in vitro translation and incubated with GST (lanes 2, 5,8,11,14,17,20) or GST-ER-(251-595) (lanes 3, 6,9,12,15,18,21) immobilized on glutathione beads. The input lanes (1,4,7,10,13,16,19) contain 10% of the radiolabeled proteins used in the binding experiments. Tripartite Complex between ER␣, GRIP1, and c-Jun-ER␣ mutants unable to bind coactivators drastically decrease estradiol regulation of AP-1-mediated transcription and overexpression of the coactivator GRIP1 (43) enhanced the estradiol effect on AP-1 activity (29 and not shown). Moreover, the closely related p160 protein SRC-1 was reported to interact with both c-Jun and c-Fos in vitro (27). We therefore analyzed whether GRIP1 could participate in a multiprotein complex containing ER␣ and c-Jun. Binding of GRIP1 on pre-formed ER␣⅐c-Jun complexes was first tested in vitro, in GST pull-down assays. GST-ER-(251-595) preloaded on glutathione-coupled beads was preincubated with an excess of purified unlabeled c-Jun protein and unlabeled c-Fos, which does not bind to ER␣, was used as a control. After extensive washes, beads were incubated with in vitro 35 S-labeled c-Jun or GRIP1 in the absence or the presence of ER␣ ligands (Fig. 8). Preincubation with c-Jun drastically decreased the consecutive interaction of ER␣ with labeled c-Jun demonstrating that most GST-ER-(251-595) molecules were bound to unlabeled c-Jun in these experimental conditions (Fig. 8A). In contrast, 35 S-labeled GRIP1 efficiently interacted with the bead-bound ER␣⅐c-Jun complexes in a liganddependent manner (Fig. 8B). GRIP1 binding, which was increased by the presence of c-Jun in control conditions, in agreement with a direct interaction of GRIP1 with c-Jun, was further enhanced by estradiol addition whereas antiestrogens had no effect.
The direct interaction between GRIP1 and c-Jun was confirmed in intact cells. c-Jun or c-Fos overexpression increased (␣-c-Jun) as described under "Experimental Procedures." As a control, 5% cell extracts used in immunoprecipitations were analyzed by Western blotting to monitor the amounts of ER␣ and c-Jun expressed in transfected cells. luciferase activity driven by GRIP1 fused to the GAL4 DBD (GAL4-GRIP1) in MCF7 cells cotransfected by a GAL4-responsive luciferase gene reporter (Fig. 9A). The same experiment was then performed in the absence or presence of the hybrid protein VP16-ER␣. As shown in Fig. 9A, and as expected, an enhancement of GAL4-driven luciferase activity was measured when GAL4-GRIP1 and VP16-ER␣ alone were coexpressed in estradiol-stimulated cells. Note that estradiol had no effect in the absence of VP16-ER␣ indicating that endogenous ER␣ concentration was likely negligible compared with that of overexpressed proteins. The addition of VP16-ER␣ together with GAL4-GRIP1 and c-Jun or c-Fos, did not significantly modify reporter gene transcription, in the absence of hormone. However, it had a synergistic effect in estradiol-treated cells in the presence of c-Jun. In contrast, in the same experimental conditions, an additive rather than a synergistic effect was observed when c-Fos was used instead of c-Jun. As we had shown that ER␣ interacted with c-Jun but not with c-Fos (Fig. 4), these results suggested that binding of ER␣ to c-Jun was important for the synergy. To try to confirm this hypothesis, VP16-ER241G mutant, deleted of the ER␣ part interacting with c-Jun, was therefore used in the same experiment. As shown in Fig. 9B, in the presence of GAL4-GRIP1 alone, VP16-ER241G increased reporter gene transcription as efficiently as the VP16 fusion protein containing wild-type ER␣. However, contrary to the results obtained with VP16-ER␣, no synergistic effect was detected on luciferase activity induced by GAL4-GRIP1 and c-Jun with VP16-ER241G, thus demonstrating the role of the ER␣⅐c-Jun interaction in the observed phenomenon.
The ER␣ Hinge Domain Contributes to the Regulation of AP-1 Activity-We further questioned whether the physical interaction between ER ␣ and c-Jun actually took part in estradiol regulation of AP-1-dependent transcription. On a first approach, the contribution of the ER␣ hinge domain on AP-1directed transcription was tested in MCF7 cells transfected with the (AP-1)4-TK-CAT reporter plasmid and increasing concentrations of the ER␣ mutant expression vector ER241G. As shown in Fig. 10A, overexpression of the hinge deleted mutant had no significant effect on AP-1 activity compared with the transfection of HEO and HE19 in a same experiment (Fig. 1). We then constructed an ER␣ mutant encompassing the interaction domain with c-Jun as determined by in vitro proteinprotein assays. If the protein⅐protein interaction was important in vivo, this truncated ER␣, by competing with the endogenous receptor for binding to c-Jun, should act as a dominant negative mutant on AP-1 activity. Increasing concentrations of pCI-ER␣-(249 -306) were therefore transfected in MCF7 cells with the (AP-1)4-TK-CAT reporter plasmid (Fig. 10B). In the absence of estradiol, ER␣-(249 -306) overexpression had no significant effect on basal AP-1 activity. However, ER␣-(249 -306) inhibited the estradiol effect on AP-1-mediated transcription. Estradiol induction of CAT activity decreased by more than 2-fold with the highest amount of pCI-ER␣-(249 -306). In all experiments and irrespective of the amount of pCI-ER␣-(249 -306) used, total inhibition of the estradiol effect was, however, never achieved. To determine whether ER␣-(249 -306) overexpression specifically inhibited estradiol-induced AP-1 activity, the effect of increasing ER␣-(249 -306) expression was tested in parallel in cells cotransfected by an ERE-containing reporter plasmid. Neither basal transcription nor estradiol induction of the ERE-␤-globin-luciferase construct was significantly altered by ER␣-(249 -306) overexpression. These results altogether suggested that physical interaction between activated ER␣ and c-Jun participated in estradiol regulation of AP-1 responses.

DISCUSSION
Previous transfection experiments using ER␣ mutants demonstrated that ER␣ could modulate AP-1 responses without binding to DNA, therefore indicating that cross-talk between the two transcription factors resulted from protein⅐protein interactions (14,17,18). However, involvement of a direct interaction between ER␣ and AP-1 complexes in this regulation is still debated (48).
We evaluated whether AP-1 family members could interact in vitro with ER␣ and showed that some of them do indeed bind to ER␣. ER␣ efficiently bound to c-Jun and JunB but only weakly interacted with JunD. ER␣ did not directly bind to any Fos family members. ER␣ thus behaved like other nuclear receptors for which an interaction with c-Jun has been described (19 -23). In most studies, no interaction between GR or retinoic acid receptor and c-Fos was detected in the absence of c-Jun (19,(21)(22)(23). Only the group of Tourray et al. (20) reported an interaction of GR with c-Fos, which was, however, less stable than with c-Jun. The C-terminal part of c-Jun containing FIG. 9. Multiprotein complex between ER␣, c-Jun, and GRIP1 in intact cells. Steroid-stripped MCF7 cells were transfected with 1 g of GAL4luc and 1 g of GAL4-GRIP1 together with 1 g of VP16-ER␣ (A) or 1 g of VP16-ER241G (B). They were cotransfected when indicated with 1 g of pCI-c-Jun or 1 g of pCI-c-Fos. Cells were then incubated with vehicle (C) or 10 nM estradiol (E2), and luciferase activity was evaluated in whole cell extracts. The results shown represent the mean (ϮS.D.) of luciferase activities calculated from triplicate wells from one experiment representative of three separate assays.
both the DBD and the leucine zipper was implicated in the association with ER␣. However c-Jun⅐c-Jun homodimers bound on TRE were still retained by ER␣ in vitro, demonstrating that the interaction did not prevent either dimerization or binding onto DNA (Fig. 5). We dissected ER␣ to determine the region of interaction with c-Jun. In contrast with the findings of Webb et al. (17), which showed that a GST fusion protein harboring the N-terminal part of ER␣ (residues 1-185) bound to c-Jun, no or only a very weak interaction could be detected with this ER␣ domain ( Figs. 1 and 2). In fact, our data demonstrated that ER␣ amino acids 259 -302 located in the hinge D domain were sufficient for binding to c-Jun. The fact that neither ER␣ residues 1-282 (Fig. 1C) nor residues 283-330 (Fig. 2B) hybridized with c-Jun in GST pull-down assays also demonstrated that an important motif for the in vitro interaction was localized around amino acid 282. This region belongs to one of the less conserved domain of nuclear receptors, which might suggest that different regions are implicated in interactions between other receptors and c-Jun.
In agreement with the in vitro studies, ER␣ truncation mutant HE19 (amino acids 179 -595) functionally interacted with AP-1 whereas HE15 (amino acids 1-282) did not. Although deletion mutant HE11 harbors the 259 -302 ER␣ region, it repeatedly bound to c-Jun with a lower efficiency than wildtype ER␣ or mutant HE19. This may suggest that residues present in the DBD directly participate in the protein⅐protein interaction, as already suggested for other nuclear receptors. However, this is not consistent with results obtained with a series of truncated ER␣ GST fusion proteins (Fig. 2B). Conversely, deletion of the DBD could induce conformational changes in the hinge region, leading to a reduced affinity for c-Jun. It is worth mentioning that HE11 has been reported to increase (14,17) or to have no effect (14,16,17) on AP-1 activity in different cellular or promoter contexts in which ER␣ was a potent activator. In the case of the ovalbumin promoter (14), mutant HE11 coactivated when cotransfected with c-Fos but not c-Jun, which may suggest that the Jun partner in AP-1 complexes could modulate the strength of the interaction with ER␣. Further experiments are, however, needed to definitively answer this question.
In addition to the convergent in vitro evidences, we demonstrated using a mammalian two-hybrid system or performing coimmunoprecipitation assays, that c-Jun and ER␣ form a protein complex in intact cells. Direct interaction between ER␣ and c-Jun appeared crucial in the complex formation, because it was not observed when the ER241G mutant (Fig. 7), which is unable to bind c-Jun and inefficient in regulating AP-1-mediated transcription (Fig. 10A), was used instead of ER␣. Moreover, the dominant negative effect of ER␣-(249 -306) (an ER␣ truncation mutant encompassing the c-Jun⅐ER␣ interaction region) on estradiol regulation of AP-1-dependent transcription is further evidence in favor of a direct interaction between the two proteins within cells, strongly suggesting that this physical interaction actually participated in estradiol-induced AP-1 activity. However, total inhibition of ER␣-mediated regulation of AP-1 activity was never obtained. This observation and the low amplitude of the effect in the two-hybrid system (Fig. 7A) suggested that one or more additional factors could also take part in this regulation and stabilize c-Jun⅐ER␣ complexes.
We show that an additional partner, i.e. the nuclear receptor coactivator GRIP1, which increased estradiol-regulated AP-1 activity (29 and this study) could bind preformed ER␣⅐c-Jun complexes (Fig. 8). Moreover, in a modified two-hybrid system, c-Jun and ER␣ had a synergistic effect on GAL4-GRIP1-driven transcription (Fig. 9). Synergy was not observed when c-Fos was present instead of c-Jun or when an ER␣ mutant unable to bind c-Jun was used, thus enlightening the crucial role of the ER␣⅐c-Jun interaction in the tripartite complex formation. Therefore, our results altogether indicate that ER␣ does not only link the pre-existing Jun⅐coactivator complexes via contacts with p160s (48) but could stabilize the c-Jun⅐GRIP1 interaction through binding to the coactivator and c-Jun. Interestingly, similar stabilization of a protein⅐protein complex by a third factor has recently been described (49) concerning the progesterone receptor⅐SRC-1 complex and JAB1, a c-Jun coactivator. JAB1 potentiates the transactivation properties of most receptors, among them ER␣ (49), reflecting the high complexity of the cross-talk between ER␣ and c-Jun. Moreover, it has been suggested that stabilization by CBP/p300 could mediate the observed cooperation between Myb and the b-Zip protein NF-M, which both bind directly to the same target DNA sequence (50), and also the positive cross-talk between thyroid hormone and retinoic acid receptors and the bZIP protein p45/ NF-E2 (51). CBP/p300, which associates with c-Fos (52), c-Jun (26), and ER␣ and ERAP160/SRC-1 (28,53) and cooperatively enhances AP-1-mediated transcription (27), may also participate in the multiprotein complex recruited by c-Jun.
Neither estradiol nor the estrogen antagonists 4-hydroxytamoxifen and ICI164,384 influenced the in vitro binding of ER␣ to Jun (Fig. 6). This was not the case in vivo: Mammalian two-hybrid experiments revealed the interaction in the presence of estradiol but not in steroid-stripped cells or after treatment with antiestrogens (Fig. 7A). Similar differences in hormone dependence in vivo and in vitro have been reported for interactions between nuclear receptors and some corepressors (54,55) or coactivators (43,49). It has been suggested that in vitro translated nuclear receptors could be in an active conformation, even in the absence of ligand (43). It is, however, tempting to speculate that the interaction between ER␣ and c-Jun is labile or weak in vivo in the absence of hormone, but enhanced by estradiol, which promotes the recruitment of nuclear receptor coactivators and further stabilizes the multiprotein complex. In addition, the fact that a coactivator is required for a stable interaction may explain the different efficiencies of ER␣ and ER␤ in regulating AP-1 activity (56), 3 whereas both proteins bound to c-Jun in vitro (data not shown). SRC-3, which belongs to the same coactivator family as GRIP1, was reported to differentially interact with the two receptors and enhance ER␣-but not ER␤-stimulated gene transcription (57). Moreover, some LXXLL peptides were shown to selectively interact with both ERs (58).
In conclusion, our present study demonstrates that the ER␣ hinge domain binds to c-Jun in vitro. This interaction also occurs in intact cells and is likely to be involved in the regulation of AP-1-induced responses. Whereas direct ER␣⅐c-Jun binding may not be sufficient by itself to trigger estradiol regulation of AP-1 activity, it could be crucial for the stability of a multiprotein complex containing c-Jun, ER␣, and a nuclear receptor activator such as GRIP1.