Mechanistic differences in the activation of estrogen receptor-α (ERα)- and ERβ-dependent gene expression by cAMP signaling pathway(s)

Although increases in intracellular cAMP can stimulate estrogen receptor-α (ERα) activity in the absence of exogenous hormone, no studies have addressed whether ERβ can be similarly regulated. In transient transfections, forskolin plus 3-isobutyl-1-methylxanthine (IBMX), which increases intracellular cAMP, stimulated the transcriptional activities of both ERα and ERβ. This effect was blocked by the protein kinase A inhibitor H89 (N-(2-(p-bromocinnamylamino)-ethyl)-5-isoquinolinesulfonamide) and was dependent on an estrogen response element. A 12-O-tetradecanoylphorbol-13-acetate response element (TRE) located 5′ to the estrogen response element was necessary for cAMP-dependent activation of gene expression by ERβ but not ERα, indicating that the former subtype requires a functional interaction with TRE-interacting factor(s) to stimulate transcription. Both p160 and CREB-binding protein coactivators stimulated cAMP-induced ERα and ERβ transcriptional activity. However, mutation of the two cAMP-inducible SRC-1 phosphorylation sites important for cAMP activation of chicken progesterone receptor or all seven known SRC-1 phosphorylation sites did not specifically impair cAMP activation of ERα. The E/F domains of ERα are sufficient for activation by forskolin/IBMX, and this is accompanied by an increase in receptor phosphorylation. In contrast, cAMP signaling reduces the phosphorylation of the corresponding region of ERβ, and this correlates with the lack of forskolin/IBMX stimulated transcriptional activity. Our data suggest that cAMP activation of ERα transcriptional activity is associated with receptor instead of SRC-1 phosphorylation. Moreover, differences in the cofactor requirements, domains of ERα and ERβ sufficient for forskolin/IBMX activation, and the effect of cAMP on receptor phosphorylation indicate that this signaling pathway utilizes distinct mechanisms to stimulate ERα and ERβ transcriptional activity.

The biological effects of estrogens are mediated by two estrogen receptors (ERs), 1 ER␣ and ER␤, which belong to a large superfamily of nuclear hormone receptors. These ligand-regulatable transcription factors possess six structural domains labeled A through F (1). The A/B domain encompasses activation function-1 (AF-1); the C and D domains correspond to the DNA binding domain and the hinge region, respectively; the E region encompasses a second activation function (AF-2) and an overlapping ligand binding domain; and the F domain, located at the extreme carboxyl terminus, is thought to play a modulatory role in ER activity. Both ERs possess similar binding affinities for estradiol and their cognate DNA binding site (estrogen response element; ERE), which is likely caused by the high degree of sequence homology they share in their ligand and DNA binding domains (2)(3)(4)(5)(6). The AF-2 domain of each receptor is regulated by ligand-induced changes in receptor conformation, but the activities of the poorly conserved AF-1 domains are ligand-independent and can be modulated by phosphorylation (7)(8)(9)(10). Notable for ER␣, in most cases the AF-1 and the AF-2 domains interact functionally to enhance transcription in a cooperative manner (7,11).
In the best studied mechanism of ER␣ and ER␤ activation, hormone diffuses into the cell, binds to the receptor, and induces a conformational change in the receptor ligand binding domain (1). Receptors, bound to their EREs either as ER␣ or ER␤ homodimers or ER␣:ER␤ heterodimers, can then recruit coactivators to the promoter region of estrogen target genes via their interaction with the receptor activation domains (2,3,12,13). There, these molecules can stimulate transcription by bridging ERs to the general transcriptional machinery and promoting the formation of a stable preinitiation complex (12,13). Notably, various coactivators also possess ubiquitin ligase, arginine methyltransferase, or histone acetyltransferase enzymatic activities that may facilitate chromatin remodeling and gene activation (14 -17).
In addition to this relatively well characterized mode of activation, ER␣ can be activated via the cAMP signaling path-way in the apparent absence of estrogens (for review, see Ref. 18). In MCF-7 cells, endogenous ER␣ target genes, including the progesterone receptor (PR; Ref. 19), pS2 (20), Liv-1 (20), and cathepsin-D (21), can be stimulated either with the cAMP analog 8-bromo-cAMP or a combination of the phosphodiesterase inhibitor 3-isobutyl-1-methylxanthine (IBMX) and cholera toxin, which increases cAMP production via a G protein-mediated signal transduction pathway. Importantly, in these experiments as well as in later studies the cAMP-induced responses are inhibited by treatment with the pure antiestrogen ICI 164,384, signifying their receptor dependence (19 -21). Demonstrating the need to transfect ER␣ into cells lacking endogenous receptor further supported this receptor dependence (22).
In addition to these ER␣ studies a number of reports indicate that the transcriptional activities of several other nuclear receptors can be modulated by the cAMP signaling pathway. For example, the chicken PR (23), androgen receptor (24), retinoic acid receptor (25), retinoid X receptor (26), and peroxisome proliferator-activated receptor-␦ (27) can be activated by cAMP signaling, demonstrating that this mode of ligand-independent activation is not exclusive to ER␣. However, human PR (28) cannot be activated ligand-independently via this mechanism, nor can the unliganded glucocorticoid receptor, although cAMP stimulation increases the hormone-dependent responses of these receptors (29,30). Collectively, these reports demonstrate the specific nature by which the cAMP signaling pathway can cross-talk with different nuclear receptors.
Some progress has been made in the effort to understand the mechanisms involved in cAMP activation of nuclear receptordependent transcription. Although an increase in receptor phosphorylation accompanies the cAMP-mediated activation of ER␣ (31,32), there is no increase in chicken PR phosphorylation associated with its activation in cells treated with 8-bromo-cAMP (33). Rather, cAMP/protein kinase A (PKA) signaling enhances chicken PR-dependent transcription, in part by increasing phosphorylation of a receptor-interacting coactivator, steroid receptor coactivator-1 (SRC-1), and thereby promotes a more stable interaction between this coactivator and p300/ CREB-binding protein (CBP)-associated factor and facilitates functional synergism between SRC-1 and CBP (34). Although it is still unknown how the unliganded ER␣ is activated by cAMP/ PKA, there is evidence that the transcription factor, CREB, can interact functionally with ER␣ and thereby mediate synergism between the 17␤-estradiol (E 2 )-dependent and cAMP-dependent signaling pathways (35).
The identification of ER␤ has increased our awareness of the diversity of potential mechanisms by which ER-dependent and estrogenic responses may be achieved (6). Notably, the relative magnitude of ER␣-and ER␤-mediated estrogen activation of ERE-containing reporters typically varies depending on the cell type and promoter context (36). In the absence of ligand, both ER␣-and ER␤-dependent transcription can be modulated by a mitogen-activated protein kinase (MAP kinase) signaling pathway (8 -10). It is unknown, however, whether ER␤ can be activated by the cAMP/PKA signaling pathway, and we therefore examined this using transient transfection assays and synthetic agents that increase intracellular cAMP levels. In so doing, we have defined mechanistic differences between cAMP activation of ER␣ and ER␤, particularly with respect to phosphorylation and the influence of cross-talk with AP-1 transcription factors. Moreover, we report that all of the p160 coactivators, as well as CBP, can coactivate ER␣ and ER␤ transcriptional activity stimulated by cAMP pathways, and we demonstrate that the importance of SRC-1 phosphorylation to cAMP activation of gene expression is receptor-dependent.
Cell Culture and Transfections-HeLa (human cervical carcinoma) cells were routinely maintained in Dulbecco's modified Eagle's medium (DMEM) supplemented with 10% fetal bovine serum (FBS). 24 h prior to transfections for transactivation assays, cells were plated in six-well culture dishes at a density of 3 ϫ 10 5 cells/well in phenol red-free DMEM with 5% charcoal-stripped fetal bovine serum (sFBS). For transfections, medium was replaced with serum-free medium, and DNA was introduced into cells in the indicated amounts using LipofectAMINE (Invitrogen) according to the manufacturer's guidelines. 5 h later, serum-free medium was replaced with phenol red-free DMEM supplemented with 5% sFBS. 12 h thereafter, cells were treated with the indicated amounts of various hormones. After 24 h of hormone treatment (12 h for inhibitor experiments), cells were harvested, and extracts were assayed for CAT activity, as described previously (45,46) using butyryl-coenzyme A (Amersham Biosciences) and [ 3 H]chloramphenicol (PerkinElmer Life Sciences). The quantity of resulting radiolabeled product was determined by scintillation counting using biodegradable counting scintillant (Amersham Biosciences) and a Beckman LS 6500 scintillation counter, and then normalized to total cellular protein measured by Bio-Rad protein assay. Experiments were done in duplicate, and values represent the average Ϯ S.E. of at least three individual experiments.
Western Blot Analysis-To determine ER expression levels, cells were transfected as above and harvested. Cell pellets were resuspended in a 50 mM Tris (pH 8.0) buffer containing 5 mM EDTA, 1% Nonidet P-40, 0.2% Sarkosyl, 0.4 M NaCl, 100 M sodium vanadate, 10 mM sodium molybdate, and 20 mM NaF and incubated on ice for 1 h. The lysates were subsequently centrifuged at 21,000 ϫ g for 10 min at 4°C. The resulting supernatants were mixed with SDS-PAGE loading buffer, resolved on a 7.5% SDS-polyacrylamide gel, and subsequently transferred to nitrocellulose membrane. The membranes were blocked using 1% nonfat dried milk in 50 mM Tris (pH 7.5), 150 mM NaCl, and 0.05% Tween 20, and sequentially incubated with an anti-FLAG M2 antibody (Sigma) and a horseradish peroxidase-conjugated, anti-mouse antibody. Blots were visualized using enhanced chemiluminescence (ECL) reagents as recommended by the manufacturer (Amersham Biosciences).
Gel Mobility Shift Assays-HeLa cells were harvested in phosphatebuffered saline using a cell scraper and subsequently resuspended in ice-cold lysis buffer (10 mM Hepes (pH 7.9), 1 mM EDTA, 60 mM KCl, 1 mM dithiothreitol, Complete Mini-Tablet protease inhibitor (Roche Diagnostics), and 0.5% Nonidet P-40), and incubated on ice for 5-15 min to lyse the cell membranes. The resulting lysates were centrifuged for 5 min at ϳ 1,400 ϫ g to pellet the nuclei. Nuclei were washed in 2.5 ml of lysis buffer lacking Nonidet P-40 and repelleted as before. Thereafter, the nuclei were resuspended in a minimum volume of 250 mM Tris (pH 7.8) buffer containing 60 mM KCl and protease inhibitors and subjected to three cycles of rapid freeze/thaw. The nuclear lysates were centrifuged for 15 min at 21,000 ϫ g and 4°C, and the protein concentrations of the resulting supernatants were determined by Bradford assay.
The radiolabeled oligonucleotides were combined with nuclear extract in binding reaction buffer (50 mM Tris (pH 7.5) containing 2.5 mM EDTA, 5 mM MgCl 2 , 2.5 mM dithiothreitol, 250 mM NaCl, 20% glycerol, and 0.25 mg/ml poly(dI-dC)) and incubated for 15 min at room temperature. For oligonucleotide binding competition and antibody supershift experiments, unlabeled oligonucleotides or antibody (anti-c-Jun (Upstate Biotechnology, Lake Placid, NY) or anti-c-Fos (Santa Cruz Biotechnology, Santa Cruz, CA)), respectively, were incubated with the nuclear extract for 30 min on ice before the addition of 32 P-labeled probe. Free and complexed DNAs were resolved on a nondenaturing acrylamide gel. To avoid any possible influence of the loading dye on complex mobility, the gel-loading buffer containing 250 mM Tris (pH 7.5), 0.2% bromphenol blue, and 40% glycerol was added to the minusextract control sample only. After electrophoresis at 350 V, the gel was transferred to 3MM Whatman paper, vacuum dried, and exposed to X-Omat film for autoradiography.
Assessment of Forskolin/IBMX-induced Phosphorylation-For metabolic labeling studies, 2 ϫ 10 6 HeLa cells were plated onto 150-mm Petri dishes, and 24 h thereafter they were transfected with pBind-ER␣ EF or pBind-ER␤ EF using a nonrecombinant adenovirus DNA transfer procedure (47). Briefly, plasmid DNA (200 ng/plate) was mixed with polylysine-coupled adenovirus and incubated for 30 min at room temperature. Before the addition of the adenovirus-plasmid particles to the cells, the medium was aspirated and replaced with DMEM. The adenovirus-plasmid particles were added to the cells at a 400:1 adenovirus: cell ratio. After incubation with the cells for 2 h at 37°C, an equal volume of DMEM with 5% sFBS was added to each dish resulting in a final concentration of 2.5% sFBS. 24 h after transfection, the medium was aspirated from the dishes and replaced with phosphate-free DMEM, and the cells were incubated for 1 h at 37°C. The medium was removed and replaced with phosphate-free DMEM containing 1% dialyzed FBS before the addition of 0.14 mCi/ml [ 32 P]H 3 PO 4 . After a 14 -16-h incubation, the plates were incubated for 90 min with either 0.1% dimethyl sulfoxide (vehicle) or 10 M forskolin ϩ 100 M IBMX.
After treatments, cells were harvested and incubated with lysis buffer (10 mM Tris (pH 8.0), 0.4 M NaCl, 1 mM EDTA, 1 mM EGTA, 10 mM ␤-mercaptoethanol, 50 mM potassium phosphate, 50 mM sodium fluoride, 0.1% Triton X-100, 0.1 mM phenylmethylsulfonyl fluoride, and a protease inhibitor mixture of 1 g/ml each leupeptin, antipain, aprotinin, benzamidine HCl, chymostatin, and pepstatin), vortexed, and centrifuged at 20,000 ϫ g for 10 min at 4°C. Protein A-Sepharose beads (Amersham Biosciences) were incubated with 2 g of antibody to the Gal4 DNA binding domain (sc-577; Santa Cruz Biotechnology) for 1 h at room temperature and then incubated with the cell lysate supernatant for 2 h while rotating at 4°C. The beads were washed with 100 volumes of phosphate-buffered saline. After the addition of Laemmli sample buffer and boiling for 5 min, the samples were electrophoresed on a 10% SDS-polyacrylamide gel and transferred to a nitrocellulose membrane for autoradiography. The same membrane was subsequently used for Western blotting to detect pBind-ER␣ EF and pBind-ER␤ EF using a horseradish peroxidase-conjugated antibody to the Gal4 DNA binding domain (sc-510HRP; Santa Cruz Biotechnology) and ECL. Autoradiography signals were quantitated by densitometric scanning of films or quantitation of the ECL signal using a Kodak Image Station 440CF.

Activation of ER␣ and ER␤ by a cAMP Signaling Pathway-
Several studies have demonstrated that agents that stimulate increases in cAMP can promote ER␣-dependent gene expression in the apparent absence of hormone (19 -22, 32, 48). To determine whether ER␤ could be activated in a similar manner, HeLa cells were transiently transfected with expression vectors for either human ER␣ or ER␤ along with an ERE-E1b-CAT synthetic target construct that possesses an ERE from the Xenopus vitellogenin A2 promoter linked to the adenoviral E1b TATA box and CAT reporter gene. Cells were subsequently stimulated with either E 2 or a combination of forskolin and IBMX (forskolin/IBMX), which increases cAMP production by activating adenylate cyclase and inhibiting phosphodiesterases, respectively, and CAT activity was measured. As expected, E 2 stimulated both ER␣-and ER␤-dependent transcription of this reporter (Fig. 1). Consistent with previous reports (22,32), stimulation with forskolin/IBMX also resulted in a robust stimulation of ER␣ transcriptional activity, although a longer (24 h) hormone treatment greatly enhanced the E 2stimulated relative to forskolin/IBMX-stimulated response (compare Figs. 1 and 2). Importantly, under the same conditions ER␤ was also activated upon stimulation of cells with forskolin/IBMX (Fig. 1). Pretreatment with a PKA-selective inhibitor, H89 (49), blocked forskolin/IBMX-induced gene expression by both receptor subtypes, thus demonstrating that forskolin/IBMX induction of ER␣ and ER␤ activity is mediated by a cAMP/PKA signaling pathway in these cells. Moreover, this inhibition was specific for forskolin/IBMX induction because H89 treatment did not significantly alter basal or E 2stimulated responses for either ER␣ or ER␤.
An ERE Is Required for ER␣ and ER␤ Activation of ERE-E1b-CAT Expression by the cAMP/PKA Signaling Pathway-To test whether the cAMP signaling response was mediated through an ER genomic mechanism in which the ER binds directly to the promoter, the effects of forskolin/IBMX on ER␣ and ER␤ was assessed on the E1b-CAT reporter, which lacks an ERE. The expected ER␣-and ER␤-dependent responses were observed for the ERE-E1b-CAT reporter. As anticipated, basal transcription of the E1b-CAT promoter was minimal, and E 2 did not further increase CAT gene expression in cells transfected with either ER␣ or ER␤ (Fig. 2). Moreover, forskolin/ IBMX did not stimulate E1b-CAT reporter gene activity in the presence of transfected ER␣, and only a very weak forskolin/ IBMX-dependent E1b-CAT expression was observed in the presence of ER␤. These data indicate that ER␣-and ER␤-dependent transcription in response to forskolin/IBMX requires the presence of an ERE.
CBP and p160/SRC Coactivators Enhance cAMP-stimulated ER␣ and ER␤ Responses-ER␣-and ER␤-dependent transcriptional differences are partly attributable to the selectivity these receptors possess for the different coactivators in the presence of various ligands (50,51). To extend this concept, we overexpressed the p160 coactivators (SRC-1, TIF2, and RAC3) as well as the general coactivator/cointegrator, CBP, in cells to determine whether any of them might distinguish between ER␣ and ER␤ in their ability to potentiate activation by cAMP-dependent signaling. Each of these coactivators strongly enhanced ER␣-and ER␤-dependent transcription compared with re-porter activity in the absence of exogenous coactivator (Fig. 3). However, none of the coactivators selectively enhanced the activity of estrogen-activated receptor over cAMP-activated receptor, nor of one receptor subtype over the other. These results suggest that each of the p160 coactivators and CBP can contribute significantly to the forskolin/IBMX-induced activation of both ER␣ and ER␤.
Activation of ER␣ and ER␤ by cAMP Depends on Promoter Context-In general, ER-mediated transcription depends on the promoter context in which an ERE is found (11,52,53). We therefore tested the extent to which forskolin/IBMX could stimulate ER-dependent gene expression in the context of various synthetic and natural ERE-containing promoters. The pS2 construct contains the Ϫ1100 to ϩ10 region of the E 2 -responsive human pS2 promoter subcloned into a CAT reporter plasmid (41). The pATC0, pATC1, and pATC2 are synthetic reporters that possess 0, 1, or 2 EREs, as indicated in their nomenclature. The ERE-E1b-CAT reporter construct was included in these experiments as a control. There was no E 2 -induced response for pATC0 because it had no ERE, a weak response for pATC1, and a synergistic E 2 -dependent response for pATC2, for both ER subtypes (Fig. 4, A and B). Notably, there was no forskolin/IBMX-induced response on either of these promoters, demonstrating that the number of EREs, in and of itself, had no effect on the forskolin/IBMX-induced activation. Moreover, although the expected E 2 -dependent increases were present for both ER␣ and ER␤ on the pS2 promoter, there was no significant increase in reporter activity in response to forskolin/ IBMX. Thus, the cAMP/PKA signaling pathway mediates ER␣and ER␤-dependent gene expression in a promoter-dependent fashion.
The Upstream TRE Enhancer in the ERE-E1b-CAT Reporter Is Required but Not Sufficient for Activation of ER␤-dependent Gene Activation by Forskolin/IBMX-Promoter differences in forskolin/IBMX-induced responses suggested that cis-acting factor(s) in addition to EREs contributed to cAMP-stimulated ER␣ and ER␤ transcriptional activities. It has been reported that ER␣ and ER␤ can mediate ligand-dependent responses at TREs, independent of EREs, when they are tethered to the promoter via AP-1 transcription factor complexes (54, 55). Previous sequence analysis (56) revealed that a putative TRE (TGACACA) that differs from the consensus TRE sequence (TGAGTCA) by two nucleotides resides in the backbone of many plasmid vectors. In ERE-E1b-CAT such a putative TRE is located ϳ255 bp upstream from the ERE. To determine whether this TRE played any role in the ability of forskolin/ IBMX to stimulate the activity of either ER, we examined the ability of forskolin/IBMX to increase CAT gene expression from a reporter plasmid (ERE-E1b-CAT(mTRE)) in which the TRE was mutated to GGACTCA, a mutation previously demon-strated to abolish AP-1 binding (39,57). As shown in Fig. 5A, the TRE mutation decreased overall reporter gene activity in the presence of both ER␣ and ER␤ whether cells were treated with vehicle, E 2 , or forskolin/IBMX. However, E 2 was still able to increase ERE-E1b-CAT(mTRE) activity above basal for both ERs, and forskolin/IBMX stimulated this reporter activity in the presence of ER␣. In contrast, forskolin/IBMX was unable to stimulate ER␤-dependent transcription of ERE-E1b-CAT- Values for ERE-E1b-CAT and pS2-CAT are normalized to their respective vehicletreated samples, which were arbitrarily set as 100. Values for pATC0, pATC1, and pATC2 are normalized to vehicle treatment for pATC2, which was set as 100. The results are the averages Ϯ S.E. of three experiments.
(mTRE), suggesting that this putative TRE was necessary for cAMP-induced ER␤-mediated increases in ERE-E1b-CAT reporter gene expression. Similar results were obtained when this TRE was removed by an NdeI to Eco0109I deletion of a 195-bp region surrounding the site (39) as opposed to mutating it (data not shown). Western blot analysis indicated that cAMP activation of ER␣ on the ERE-E1b-CAT reporter (with or without the TRE) was not caused by an increase ER␣ protein levels (Fig. 5B). In contrast to ER␣, ER␤ protein expression is elevated modestly by forskolin/IBMX. However, this does not permit cAMP-stimulated ERE-E1b-CAT(mTRE) activity by ER␤. Thus, cAMP activation of ER␤ is dependent on a putative TRE site, but loss of this cis element only partially attenuates ER␣ activity and indicates that activation of ER␣ and ER␤ by cAMP signaling is distinct. Taken together with the above data that demonstrated a requirement for an ERE, these data suggest that forskolin/IBMX promotes synergism between either ER␣ or ER␤ and a factor(s) that is bound to a neighboring DNA response element (i.e. the putative TRE).
To determine whether c-Jun and/or c-Fos expressed in HeLa cells could bind to the putative TRE site, an electrophoretic mobility shift assay was performed. As shown in Fig. 6, a 27-bp oligonucleotide corresponding to the TRE-containing region of the target gene bound to factors present in a HeLa cell nuclear extract, and this binding could be competed with an excess of cold oligonucleotide. The addition of antibodies against c-Jun or c-Fos to the binding reaction resulted in the supershift of the oligonucleotide-protein complex, indicating that both of these proteins associate with this DNA fragment. Notably, equivalent levels of oligonucleotide-bound material were observed regardless of whether the nuclear extract was prepared from vehicle-or forskolin/IBMX-treated cells, indicating that the forskolin/IBMX-induced activation of ER-dependent gene ex-pression was not a result of increased Jun/Fos occupancy of the promoter region. This is consistent with the comparable levels of c-Jun and the small increase in c-Fos expression detected by Western blot analyses of extracts prepared from vehicle-and forskolin/IBMX-treated cells (data not shown).

The Carboxyl Terminus of ER␣ and ER␤ Mediates Forskolin/ IBMX-induced Transcription, Which Can Be Enhanced Further by the Amino Terminus of ER␣ but Not ER␤-Two previous reports have characterized the physical interactions between
ER␣ and the AP-1 transcription factor family member, c-Jun (55,58). Although one of these studies demonstrated that ER␣ interaction with c-Jun is predominantly mediated by the centrally located hinge region (domain D) of the receptor (58), both indicated that an interaction with the amino terminus of ER␣ is also possible. Therefore, to assess the potential contribution of the amino-terminal A/B domain in mediating forskolin/ IBMX-induced responses, ER␣ and ER␤ deletion mutants lacking their A/B domains (ER␣-179C and ER␤-143C) were constructed, and the ability of the resulting receptors to stimulate ERE-E1b-CAT reporter activity in response to forskolin/IBMX stimulation was examined. In the absence of transfected ER (Fig. 7A, Empty), there is very weak CAT gene expression in response to forskolin/IBMX treatment. This minimal promoter activity is substantially less compared with activity in the presence of transfected ERs. As shown by comparison of ER␣-179C (ER␣ amino acids 179 -595) with wild type ER␣, deletion of the amino terminus of ER␣ reduced forskolin/IBMX as well as basal and E 2 -stimulated activities, which is consistent with the constitutively active amino-terminal AF-1 domain of this receptor contributing to E 2 -dependent ER␣ responses (7,11). In contrast, removing the A/B region of ER␤ to generate ER␤-143C (ER␤ amino acids 143-530) did not reduce forskolin/ IBMX induction of ER␤-dependent target gene expression. No- tably, the E 2 -stimulated activity of ER␤-143C is much higher than that of wild type ER␤, which is in agreement with a previously reported inhibitory function for the amino terminus of ER␤ (59).
As shown in Fig. 7B, ER␣ but not ER␤ retained the ability to stimulate CAT expression from the TRE-minus reporter (ERE-E1b-CAT(mTRE)) in response to forskolin/IBMX (see also Fig. 5). Interestingly, much of the E 2 -induced and all of the remaining forskolin/IBMX-induced ER␣ activity are lost when the A/B domain is removed, as shown for ER␣-179C, thus supporting the above data that this domain can mediate cAMP-dependent activation of ER␣. Taken together, these results demonstrate that domains C through F of ER␣ and ER␤ are sufficient for cAMP/PKA signaling pathway activation of either receptor provided that an AP-1 DNA binding site is present on the promoter; these results also indicate that this functional interaction is enhanced by the A/B domain of ER␣ but not ER␤.
Several studies have focused on the ability of the MAP kinase signaling pathway to stimulate the AF-1 activity of ER␣ (8,10). Because it is known that the cAMP/PKA signaling pathway can cross-talk with the MAP kinase signaling pathway (34,60), we examined whether the MAP kinase-directed phosphorylation site in the amino terminus of ER␣ (Ser 118 ) might be important for the forskolin/IBMX-induced activity of the ER␣ AF-1 domain. However, mutating this serine to an alanine (ER␣-S118A) did not alter the ability of ER␣ to stimulate transcription of the ERE-E1b-CAT(mTRE) reporter in response to forskolin/IBMX (data not shown). Similarly, alanine mutation of the other known amino-terminal phosphorylated residues in ER␣ (Ser 104/106/118 or Ser 167 ) did not inhibit forskolin/IBMX induction of reporter gene expression (data not shown), indicating that these phosphoserine residues did not account for the forskolin/IBMX-dependent activity of the ER␣ AF-1 domain. Thus, ER␣ AF-1 activity in response to forskolin/ IBMX is likely to be mostly the result of cAMP/PKA signaling to a factor(s) that can interact with the A/B domain rather than altering the phosphorylation of the ER␣ amino terminus itself.

Functional Interactions with the TRE-bound Factor(s) Can Be Mediated by the EF Region of ER␣ but Not ER␤-It has been
demonstrated that amino acids 259 -302 of ER␣ constitute a major interaction site with c-Jun (58), and because the ability of ER␣-179C to mediate forskolin-induced activation of CAT expression was dependent on a TRE site within the reporter gene that has been shown to support c-Jun physical and functional interactions (39,56), we investigated the possibility that cAMP activation of ER␣-179C was caused by a direct interaction between the hinge region of the receptor and c-Jun. To test this hypothesis, the EF domain of ER␣ (amino acids 302-595) and the corresponding ER␤ fragment (amino acids 254 -530) were fused to the Gal4 DNA binding domain (Gal-ER␣ EF and Gal-ER␤ EF) and examined for their abilities to stimulate expression of 17mer-E1b-CAT, which contains four Gal4 binding sites upstream from the TATA box and CAT gene. This reporter also possesses the TRE site upstream of the Gal4 binding sites. As expected, both Gal-ER␣ EF and Gal-ER␤ EF were stimulated by E 2 (Fig. 8). Importantly, forskolin/IBMX stimulated Gal-ER␣ EF-dependent expression of the 17mer-E1b-CAT but not the TRE-minus reporter, 17mer-E1b-CAT(⌬TRE). This demonstrates that an ER␣ construct lacking all of the c-Jun binding sites can still be activated; this indicates that the interaction between the TRE-binding factor and ER␣ could be indirect and may be mediated via a trans-acting factor that can bind to both the EF domain of ER␣ and a TRE-binding factor such as c-Jun. In contrast, the EF region of ER␤ is insufficient to activate reporter gene expression regardless of the presence or absence of a TRE site, indicating that regions within the DNA binding domain and/or hinge of ER␤ are important for activation of target gene expression in response to cAMP signaling. Taken together, these results indicate that the EF domains of ER␣ and ER␤ differ in their abilities to mediate ER cooperation with a TRE-bound factor in response to cAMP.
To assess the effect of forskolin/IBMX treatment on Gal-ER␣ EF and Gal-ER␤ EF phosphorylation, expression vectors for Gal-ER␣ EF or Gal-ER␤ EF were transfected into HeLa cells using an adenovirus-mediated DNA transfer technique to increase protein expression levels (47) followed by in vivo labeling with [ 32 P]H 3 PO 4 and incubation with forskolin/IBMX or vehicle for 90 min. The results of a representative experiment are shown in Fig. 9, A and B, and demonstrate that when 32 P signals were normalized to protein levels that there was an increase in the overall level of Gal-ER␣ EF phosphorylation after incubation with forskolin/IBMX, whereas the same treatment induced a decrease in Gal-ER␤ EF phosphorylation. This experiment was repeated three times, and the averaged results reveal that forskolin/IBMX increased Gal-ER␣ EF phosphorylation by 44 Ϯ 6% but decreased Gal-ER␤ EF phosphorylation by 45 Ϯ 17% (Fig. 9C).
Forskolin/IBMX-stimulated ER␣ Activity Is Not Caused by cAMP/PKA-dependent SRC-1 Phosphorylation-It had been demonstrated previously that the chicken PR is not phosphorylated in response to cell treatment with 8-bromo-cAMP, but rather cAMP activation of transcription seems to be mediated by an increase in SRC-1 phosphorylation (34). Moreover, SRC-1 binds to both the EF domain of ER␣ as well as c-Jun (61, 62) and is therefore a good candidate for mediating indirect interactions between these two transcription factors, as described

FIG. 6. c-Jun and c-Fos binding to the putative TRE is similar in vehicle-and forskolin/IBMX-treated HeLa cells.
Nuclear extracts prepared from vehicle (Ϫ) or 10 M forskolin/100 M IBMXtreated cells (ϩ) were subjected to electrophoretic mobility shift assays using a 32 P-labeled TRE as probe. Competitions with a 100-fold excess of unlabeled probe (TRE) or a DNA fragment containing a mutation in the TRE site (mTRE) are shown in lanes 3 and 6, and 4 and 7, respectively. Supershift assays are shown with c-Jun (lanes 9 and 11) or c-Fos (lanes 8 and 10) antibodies. Lane 1 contains no nuclear extract. A representative experiment is shown.
above. Therefore, we examined whether the ability of cAMP/ PKA to modulate SRC-1 phosphorylation might also contribute to forskolin/IBMX-dependent activation of ER␣. SRC-1 expres-sion vectors containing substitutions for the two cAMP-induced phosphorylatable residues (T1179/S1185A) or substitutions for all of the seven previously mapped phosphorylation sites (at positions 372, 395, 517, 569, 1033, 1179, and 1185; Ref. 63) were introduced into cells, and the abilities of these mutant coactivators to enhance ER␣-dependent reporter activity was assessed. Compared with wild type SRC-1 there is an ϳ20 and ϳ35% decrease in the ability of the SRC-1 T1179/S1185A and the seven-alanine mutant (SRC-1 7Ala ), respectively, to potentiate forskolin/IBMX-induced ER␣ activity (Fig. 10, A and B). Nonetheless, decreases in ER␣ coactivation by both the SRC-1 T1179/S1185A and the SRC-1 7Ala mutants were equal for basal as well as for E 2 -stimulated and forskolin/IBMX-stimulated responses, suggesting that SRC-1 phosphorylation does not specifically modulate cAMP-dependent activation of ER␣. DISCUSSION In this report, we demonstrated that forskolin/IBMX, through increased intracellular cAMP and activation of PKA, can stimulate ER␤-dependent transcription, as was shown previously to be the case for ER␣. However, there are significant differences in the ability of ER␣ and ER␤ to be ligand-independently activated by this mechanism. In particular, a TRE upstream from the ERE was necessary for ER␤-dependent transcription in response to stimulation with forskolin/IBMX, whereas this sequence contributed to, but was not required for, activation of full-length ER␣. Furthermore, functional interactions with the TRE-bound factor(s) could be mediated by the EF region of ER␣ but not ER␤, and forskolin/IBMX treatment increased the phosphorylation of Gal-ER␣ EF but decreased the phosphorylation of the corresponding ER␤ construct. Overexpression of the p160 and CBP coactivators enhanced forsko- lin/IBMX-induced ER␣ and ER␤ activity, indicating that these coactivators can form functional complexes with the unliganded receptor. However, in contrast to the previously reported importance of SRC-1 phosphorylation for cAMP-mediated activation of chicken PR (34), the contribution of these phosphorylation sites to SRC-1 coactivation of cAMP-induced ER␣ was minor and not specific to ER activation by forskolin/IBMX. Taken together, these data demonstrate that the cAMP signaling pathway can stimulate ER-dependent transcription by promoting functional interactions among TRE-bound transcription factor(s), coactivators, and either ER␣ or ER␤ bound to an ERE.
There has been considerable interest in understanding the relative contributions of the ER␣ and ER␤ AF-1 domains in mediating ER-dependent transcription. The A/B domain of ER␣, which encompasses the AF-1, is generally more active than that of ER␤ (59). Interestingly, although a MAP kinase signaling pathway can induce ER␣ and ER␤ AF-1 activity in the absence of ligand (8 -10), our results demonstrate that cAMP signaling can stimulate the AF-1 activity of ER␣, but not ER␤. Although it has been reported that the cAMP signaling pathway can stimulate MAP kinase activity, mutation of the previously identified amino-terminal MAP kinase phosphorylation site in ER␣ (8, 10) does not inhibit its activation by forskolin/IBMX. Consistent with this result, our work and that of LeGoff et al. (31) demonstrates cAMP-induced phosphorylation of the ER␣ carboxyl-terminal domain; however, the location of these phosphorylation sites in vivo remains to be identified. It is therefore likely that forskolin/IBMX activation of ER␣ via the AF-1 domain is caused by cAMP/PKA signaling effects on the recruitment and/or activity of a coactivator(s) that interacts selectively with the AF-1 domain of ER␣. Interestingly, there are several coactivators, including the p72/ p68 RNA-binding DEAD box proteins (64,65) and the RNA coactivator, SRA, 2 which have been found to interact selectively with the AF-1 domain of ER␣ but not that of ER␤. Moreover, in addition to p68 being a phosphorylated protein (66), all three of these coactivators are found in complexes with other coactivator molecules, such as CBP, SRC-1, TIF2, and AIB1 (64,65), which are known to be phosphoproteins (34,(67)(68)(69).
The carboxyl-terminal ligand binding domains of ER␣ and ER␤ possess considerably higher sequence homology than do the A/B domains. Interestingly, however, there are also differences in the abilities of the ER␣ and ER␤ EF domains to be 2 K. M. Coleman and C. L. Smith, unpublished data. Gal-ER␣ EF and Gal-ER␤ EF were immunopurified and then electrophoresed on a 10% SDS-polyacrylamide gel. The proteins were transferred to a nitrocellulose membrane, and the membrane was exposed to X-AR film for autoradiography (left panels). After autoradiography, the membrane was subjected to Western blotting with anti-Gal-horseradish peroxidase antibody and detection using enhanced chemiluminescence (right panels). C, autoradiography signals were quantitated by densitometric scanning of films, and Western blot signals were quantitated by either densitometric scanning of films or quantitation of the ECL signal using a Kodak Image Station 440CF. Normalized phosphorylation values for Gal-ER␣ EF and Gal-ER␤ EF were calculated by dividing the value for the 32 P signal by the value for the protein signal. Data are plotted as change in phosphorylation relative to vehicle treatment (set at a value of 1). The experiment was repeated three times with the mean Ϯ S.E. reported. activated by cAMP signaling, as indicated by the ability of Gal-ER␣ EF but not Gal-ER␤ EF to stimulate 17mer-E1b-CAT activity in response to forskolin/IBMX. Interestingly, forskolin/ IBMX induced opposite changes in overall phosphorylation for these two constructs. Consistent with a previous demonstration of cholera toxin and IBMX-induced phosphorylation of an ER␣ mutant lacking its A/B domains (31), forskolin/IBMX increased the phosphorylation of the Gal-ER␣ EF chimera. There are four protein kinase A consensus sites (XRRXSX or SKKIX-SIX) in the carboxyl terminus of ER␣: Ser 236 , Ser 305 , Ser 338 , and Ser 518 . However, mutation of the Ser 236 , Ser 305 , and Ser 518 sites to alanines does not block cAMP-induced ER␣ transcriptional activity (35), and the Ser 236 site is not present in the Gal-ER␣ EF construct. Although a S338A mutation does inhibit cholera toxin/IBMX-induced ER␣ activity, substitution of a glutamic acid at this position does not mimic the ligandindependent response, suggesting that mutation of this residue may affect cAMP signaling to the receptor by inducing structural perturbations (35). The only other phosphorylation sites identified in the ER␣ carboxyl terminus, Thr 311 (70) and Tyr 537 (71), are also unlikely to be the targets of the forskolin/IBMXinduced phosphorylation site because cAMP signaling has been reported to induce only serine phosphorylation of ER␣ (31). The location of the phosphorylation site(s) induced by forskolin/ IBMX in vivo, as well as their importance for cAMP activation of ER␣ transcriptional activity therefore remain to be characterized; this is however, beyond the scope of this report. It should be noted that the forskolin/IBMX-induced site need not lie within a protein kinase A consensus sequence because the cAMP signaling has been shown to cross-talk with other kinase pathways (34,60). Apart from an in vitro demonstration of p38-induced phosphorylation of the AF2 domain of ER␤ (72), little is known about phosphorylation of the ER␤ EF domain. Our results therefore provide the first demonstration that the ER␤ carboxyl-terminal region is phosphorylated in vivo, as well as novel information indicating that the basal level of ER␤ EF domain phosphorylation is reduced by forskolin/ IBMX treatment in HeLa cells. Although this may seem paradoxical, cAMP has been shown to inhibit MAP kinase pathways through cross-talk with the G protein Rap1 and Raf-1 (60).
The inability of the AF-1-deletion mutants (ER␣-179C and ER␤-143C) and the Gal-ER␣ EF chimera to stimulate the activities of reporters lacking TREs (ERE-E1b-CAT(mTRE) and 17mer-E1b-CAT(⌬TRE), respectively) in response to forskolin/ IBMX also suggests that activation of domains EF of ER␣ and FIG. 10. Alanine mutation of SRC-1 phosphorylation sites decreases its coactivation of ER␣ but is not specific to cAMP/PKA-dependent signaling. A, a representative experiment in which HeLa cells were transiently transfected with 10 ng of expression plasmid for ER␣ (pCMV 5 -ER␣) along with 1 g of expression plasmid for either wild type or mutant SRC-1a (SRC1 T1179/S1185A or SRC-1 7Ala ) or the empty vector (pCR3.1) and 1 g ERE-E1b-CAT reporter. Cells were subsequently treated with vehicle, 1 nM E 2 , or 10 M forskolin ϩ 100 M IBMX (F/I). B, combined results from seven experiments. Relative coactivation was determined by dividing reporter activity in the presence of wild type SRC-1 by values obtained in the absence of coactivator (pCR3.1) for each treatment group (vehicle, E 2 , and F/I) and defining this value as 100. Values for mutant SRC-1 coactivation are given relative to wild type SRC-1 coactivation. Results are the averages Ϯ S.E. of seven experiments. C through F of ER␤ require a functional interaction between these receptor domains and a TRE-bound transcription factor. Based on sequence information, this TRE-binding factor most likely belongs to the AP-1 transcription factor family. This includes Jun (c-Jun, JunB, JunD), Fos (c-Fos, FosB, Fra1, Fra2), and ATF (ATFa, ATF2, ATF3) proteins, which can form homo/heterodimers among themselves and promote transcription via binding to palindromic sequences (TGA(C/G)TCA) that are found in a number of promoters (for review, see Ref. 73). Although the sequence (TGACACA) present in our ERE-E1b-CAT reporter diverges from the consensus TRE sequence, it has been shown to bind to in vitro translated Jun and Fos (56), and our gel mobility shift assays reveal that the levels of both c-Jun and c-Fos expressed in HeLa cells are sufficient to bind to this DNA. Moreover, our laboratory demonstrated previously through overexpression studies that c-Jun can enhance ER␣mediated transcription of ERE-E1b-CAT but not of the corresponding TRE mutant reporter, ERE-E1b-CAT(mTRE) (39). Because we demonstrate a requirement for the ERE and TRE DNA sites, our data suggest that cooperation between ER and AP-1 transcription factors bound to their respective target promoter sequences results in robust forskolin/IBMX activation of target gene expression. Based on the ability of just the EF domains of ER␣ to be activated by cAMP even though this portion of ER␣ does not bind c-Jun or c-Fos (58), it is likely that the interactions between ER and AP-1 transcription factors need not be direct, although we cannot rule out the possibility of other TRE-binding factors directly contacting ER␣ and/or ER␤.
One potential mechanism through which TRE and ERE binding factors could interact indirectly with one another is through coactivators. We have demonstrated that CBP as well as all three p160 coactivators can enhance forskolin/IBMXinduced transcription of ERE-E1b-CAT by ER␣ and ER␤, suggesting that these coactivators can form functional complexes with ERs in the absence of hormone. Moreover, overexpression of these coactivators does not compensate for the inability of forskolin/IBMX to stimulate ER␤-dependent transcription of the ERE-E1b-CAT(mTRE) target construct (data not shown), indicating that their ability to stimulate cAMP-induced ER function is derived from ER and TRE-binding factor interactions. There are a number of potential candidates that possess the ability to bind to c-Jun as well as ERs. These include the coactivators SRC-1, JAB1, and CAPER as well as the integrator protein, CBP (61, 74 -77). In all cases but CAPER, the coactivator protein utilizes distinct sites to bind to c-Jun and nuclear receptors, suggesting that these coactivators are well suited to act as physical bridges between these two classes of transcription factors. Although all of the p160 coactivators and CBP contributed to forskolin/IBMX-induced activation of ER target gene expression, we were unable to observe a similar activity by the c-Jun activation-binding protein, JAB1 (data not shown).
Cyclic AMP signaling leads to phosphorylation of SRC-1 at Thr 1179 and Ser 1185 residues contributing to stabilizing interactions between CBP-P and p300/CBP-associated factor and functional synergy between CBP and SRC-1 (34). Moreover, mutation of these two amino acids to alanines reduced both progesterone-stimulated and, in an even more marked fashion, cAMP-stimulated chicken PR activity in COS cells. However, these mutations did not completely block the ability of SRC-1 to enhance cAMP-dependent activation of chicken PR, suggesting that phosphorylation of another cofactor(s) may contribute to activation of this receptor by cAMP. The same mutations only slightly impaired the ability of SRC-1 to enhance ER␣ activity stimulated by E 2 and cAMP. Moreover, mutation of all seven SRC-1 phosphorylation sites identified by Rowan et al. (63) also reduced the overall efficacy of this coactivator but, again, regardless of receptor stimulus. These data suggest that SRC-1 is not specifically involved in the activation of ER␣ by cAMP and that this ligand-independent activity can be mediated by another ER␣-interacting cofactor(s). It should be noted that growth factor and protein kinase C signal transduction pathways have been shown to alter the phosphorylation and/or coactivation potential of GRIP1/TIF2, AIB1/RAC3, and p300/ CBP coactivators (67)(68)(69), and it is possible that cAMP crosstalk with one or more of these factors is critical for activation of ER transcriptional activity. An examination of this possibility awaits identification of cAMP-induced phosphorylation sites in these coactivators. Taken together, our data indicate that cAMP activation of cPR and ER␣ differ in the extent to which SRC-1 phosphorylation is required for this process as well as whether the respective receptors are themselves phosphorylated. In addition, ER␣ and ER␤ also differ in their phosphorylation, their dependence on promoter TRE sites, and the minimal region of receptor required to respond to cAMP signaling. Overall, this argues that multiple mechanisms contribute to cAMP activation of nuclear receptor transcriptional activity.
The promoters of endogenous genes typically consist of binding sites for many distinct transcription factors. Importantly, the human pS2 promoter contains binding sites for ERs as well as AP-1 transcription factors (78), indicating that expression of this gene, which had previously been demonstrated to be activated by cAMP in an ICI 164,384-inhibited manner, might involve cross-talk between ER and AP-1. Unexpectedly, pS2-CAT was not activated by ER␣ or ER␤ in response to forskolin/ IBMX treatment. This could be the result of cell type differences and/or loss of a promoter region critical for cAMP activation of ER during construction of the pS2-CAT reporter. DNA sequence analyses have enabled us to identify several other target gene promoters containing both TRE and ERE sites. Thus, the ability of the cAMP signaling pathway to stimulate ER-dependent transcription via ER-AP-1 interactions might be applicable to many other ER target genes. Based on our studies, it is clear that ER-dependent responses cannot be predicted on the presence of an ERE alone, but consideration must be given to the complexity of such promoters and how these receptors interact with various nonreceptor transcription factors, either directly or through coactivator/cointegrator molecules. Undoubtedly, the ability of coactivators to integrate responses through various classes of transcription factors adds another level of control and specificity to regulation of gene expression.