Suppression of β-Amyloid Precursor Protein Signaling into the Nucleus by Estrogens Mediated through Complex Formation between the Estrogen Receptor and Fe65

Abstract

The C-terminal fragment of the β-amyloid precursor protein produced after cleavage by γ-secretase, namely, APPct or AICD, has been shown to form a multimeric complex with the adaptor protein Fe65 and to regulate transcription through the recruitment of the histone acetyltransferase Tip60. The present study shows that 17β-estradiol inhibits the transcriptional and apoptotic activities of the APPct complex by a process involving the interaction of estrogen receptor alpha (ERα) with Fe65. ERα-Fe65 complexes were detected both in vitro and in the mouse brain, and recruitment of ERα to the promoter of an APPct target gene (KAI1) was demonstrated. Our studies reveal a novel mechanism of estrogen action, which may explain the well-known neuroprotective functions of estrogens as well as the complex role of this female hormone in the pathogenesis of neuronal degeneration diseases.

Estrogens are pleiotropic hormones that regulate the growth and differentiation of many diverse tissues. The actions of estrogens are mediated through estrogen receptors alpha and beta (ERα and ERβ, respectively), which belong to the steroid/thyroid nuclear receptor superfamily, a group of ligand-regulated, zinc finger-containing transcription factors (11, 45). The two ERs share a similar structural organization composed of an amino-terminal A/B region containing activation function 1 (AF-1), a central DNA binding domain (DBD), and a carboxyl-terminal ligand-binding domain (LBD) containing activation function 2 (AF-2). While AF-1 of the ER is constitutively active, the activity of AF-2 is strictly ligand dependent. In response to ligand activation, the ER forms a homodimer that binds estrogen response elements (EREs) and recruits multiple transcriptional coactivator complexes (23), leading to transcriptional stimulation.

Besides their well-established role in the female reproductive system, estrogens also play an important role in regulating the neuronal activities of the pituitary gland, the hypothalamus, and other specific brain regions. It is generally believed that estrogens are neuroprotective and that the decrease in their abundance after menopause contributes to the development of neurodegenerative diseases such as Alzheimer's disease (AD). Multiple animal experiments and molecular analyses overwhelmingly support a neuroprotective role for estrogens (14, 31, 46). For example, estrogens have been shown to protect neurons from ischemic stroke, an activity that depends on ERα (9). 17β-Estradiol was shown to exert neurotrophic effects in tissue explants derived from the developing mouse hypothalamus and preoptic area (44). Synaptogenic effects of 17β-estradiol were demonstrated to occur in adult brain regions, such as the hippocampus, that are important for cognitive function (17, 51). For various models of neuronal damage, such as that caused by ischemic stroke, estrogens were reported to enhance neuronal survival (31, 46, 50).

The β-amyloid precursor protein (APP) plays important roles in AD pathogenesis. It is a membrane-spanning protein with classical features of an orphan G protein-coupled receptor that is physiologically processed by cleavages with α- or β- and γ-secretase. After cleavage by β-secretase, APP is further cleaved in the transmembrane region by γ-secretase. The resulting extracellular amyloid β-peptide (Aβ) is deposited in the senile plaques in the brains of AD patients, which are considered the hallmark of AD. Besides its role in neurodegeneration, APP has multiple cellular functions, which include cell adhesion and motility (37), synaptic transmission and plasticity (38), and memory (7). It is also important that APP is ubiquitously expressed and not restricted to the brain. Increased expression of APP protein has been described for oral squamous cell carcinoma (24), suggesting additional biological functions for APP in nonneuronal and cancer cells.

Perhaps one of the most interesting findings about APP in recent times is the demonstration of its potential role in transcriptional regulation (4). Besides Aβ, γ-cleavage of APP releases its intracellular C-terminal fragment, or APPct, which moves to the nucleus with Fe65 (20, 22). Fe65 binds Tip60 (34), a histone acetyltransferase that is involved in regulating higher-order chromosome structure and consequential gene expression (21). In the present study, we demonstrate that estrogens inhibit the transcriptional activity of and apoptosis induced by the APPct transcriptional complex. Mechanistic investigations reveal that the transcriptional repression is due to the “tethering” of ERα to the APPct complex through Fe65. The potential functional significance of this inhibition and its relevance to the well-documented neuroprotective activities of estrogens and to their suspected role in preventing AD are discussed.

MATERIALS AND METHODS

Materials.

17β-Estradiol was purchased from Sigma (St. Louis, MO). Alexa Fluor 594-goat anti-mouse immunoglobulin G (IgG [heavy plus light chains]), Alexa Fluor 488-goat anti-rabbit IgG, and Alexa Fluor 594-chicken anti-goat IgG (heavy plus light chains) were obtained from Molecular Probes (Eugene, OR). Anti-hemagglutinin (anti-HA; 12CA5) antibody was obtained from Roche. Gal4 DBD antibody was purchased from BD Biosciences Clontech (Palo Alto, CA). Anti-FE65 (clone 3H6) was purchased from Upstate (Charlottesville, VA). The following antibodies were obtained from Santa Cruz Biotechnology (Santa Cruz, CA): anti-ERα (F-10 and H-184), -Gal4 (DBD) (RK5C1), -c-Myc (9E-10 and A-14), -KAI1 (C-16), -β-amyloid (D-17), and -amyloid A4 (N-18). Anti-APP C-terminal antibody (CT-15) has been described previously (25). Lipofectamine was purchased from Gibco-BRL Life Technologies (Rockville, MD). The ECL reagents for immunoblotting were obtained from Amersham Pharmacia Biotech Inc. (Piscataway, NJ). Luciferase assay systems were obtained from Promega Corporation (Madison, WI), and 4′,6′-diamidino-2-phenylindole (DAPI) mounting medium was obtained from Vector Laboratories (Burlingame, CA). All other reagents were reagent grade.

Expression vectors for APP, Tip60, and Fe65 were gifts from Thomas C. Südhof (4). pRST7-ERα, ER282G, ER179C, and ER-TAF1 were gifts from Donnald P. McDonnell (47). pLEN-hERα, pLEN-hERαY537A (26, 49), pLENβgal, EREe1bLuc (26), and Galluc (53) were described previously.

Tissue preparation and sectioning.

For immunofluorescence and chromatin immunoprecipitation (CHIP) studies, 5-month-old wild-type and B6C3 APPswe/PSEN1(A246E) transgenic mice were deeply anesthetized with pentobarbital (50 mg/kg of body weight) and perfused transcardially with 0.1 M phosphate-buffered saline (PBS; pH 7.4) followed by ice-cold phosphate-buffered 4% paraformaldehyde. Brains were taken out of the skulls, fixed by submersion in 4% paraformaldehyde in 0.1 M phosphate buffer for 24 h at 4°C, and stored at 4°C in phosphate-buffered sucrose (30%). Tissues were sectioned serially at 30 μm on a freezing microtome and then mounted onto aminopropyltriethoxysilane-coated glass slides.

Transfections and reporter assays.

HeLa and ER-positive Ishikawa cells were plated on Dulbecco's modified Eagle's medium containing 10% fetal bovine serum at 1.5 × 10⁵ cells/well in six-well plates. One day after plating, transfections were performed with Lipofectamine. After transfection, cells were placed in Dulbecco's modified Eagle's medium containing 1% charcoal-stripped fetal bovine serum and treated with vehicle and estrogens. At 48 h posttransfection, cellular extracts were prepared by directly adding lysis buffer containing 25 mM Tris-phosphate (pH 7.8), 2 mM dithiothreitol, 2 mM 1,2-diaminocyclohexane-N′,N′,N′,N′-tetra-acetic acid, 10% glycerol, and 0.2% Triton X-100 to the cells on ice. Luciferase activity was determined with luciferase assay systems following the company's protocol. β-Galactosidase activity was determined as previously described (27, 28).

Immunoprecipitation and immunoblotting analyses.

To prepare cells for assay of the interaction of ectopically expressed proteins, cells were plated in Dulbecco's modified Eagle's medium containing 10% fetal bovine serum at 1.0 × 10⁶ cells per 100-mm dish. One day after plating, transfections were performed with Lipofectamine. Transfected cells were placed in Dulbecco's modified Eagle's medium containing 1% charcoal-stripped fetal bovine serum and treated with different reagents. Seventy-two hours later, cellular extracts were prepared in a buffer containing 20 mM Tris-HCl (pH 7.5), 0.5% NP-40, 250 mM NaCl, 3 mM EDTA, 3 mM EGTA, 10 μg of aprotinin per ml, 10 μg of leupeptin per ml, 10 mM benzamidine, and 1 mM phenylmethylsulfonyl fluoride.

For assay of the interaction of endogenous proteins, mouse brain tissue was removed and homogenized in a lysis buffer containing 50 mM Tris (pH 7.4), 1 mM dithiothreitol, and 0.1 mM EDTA for 10 strokes on ice with a Polytron homogenizer. The homogenate was then sonicated for 15 seconds on ice.

For immunoprecipitations, cellular and mouse brain extracts were then incubated with 4 μg antibodies overnight at 4°C and subsequently with protein G-agarose beads for an additional hour. The agarose beads were then washed three times, and immunoprecipitates were heated in sodium dodecyl sulfate (SDS)-polyacrylamide gel electrophoresis sample buffer.

For immunoblotting, cellular extracts or immunoprecipitates were separated in SDS-polyacrylamide gels, transferred to a nitrocellulose membrane, and probed with different antibodies. Proteins were detected with ECL reagents.

GST pull-down assays.

ERα and Fe65 were translated in vitro from PRST-ERα and HA-Fe65, respectively, using T7 polymerase and a coupled transcription/translation kit (Promega, Madison, WI). GST-APPct protein was produced in and purified from Escherichia coli strain BL21(DE3). Glutathione S-transferase (GST) pull-down assays were performed with glutathione beads as described previously (29). The proteins in the precipitates were detected by immunoblotting analysis.

Confocal and deconvolution immunofluorescence imaging.

Cells grown on coverslips and frozen sections of mouse brain mounted on glass slides were washed with PBS three times and fixed in 2% paraformaldehyde for 15 min at room temperature. After additional washing with PBS, fixed cells were made permeable with 1% Triton X-100 and 1% bovine serum albumin for cells or with goat serum for tissue. The slides were then incubated with primary antibodies for 2 h at room temperature or overnight at 4°C, followed by incubation with secondary antibodies, i.e., IgG conjugated to Alexa Fluor 594 for red or to fluorescein isothiocyanate for green, for another 1 to 2 h at room temperature. The slides were then washed three times in PBS and stained with DAPI in antifade mounting medium. Fluorescent images were obtained with a confocal laser scanning microscope or a Leitz Orthoplan 2 microscope.

CHIP assays.

For CHIP assays, mouse brain tissues were prepared as described above. Cells were plated in Dulbecco's modified Eagle's medium containing 10% fetal bovine serum at 1.0 × 10⁶ cells per 100-mm dish. One day after plating, transfections were performed with Lipofectamine. Transfected cells were placed in phenol red-free Dulbecco's modified Eagle's medium containing 1% charcoal-stripped fetal bovine serum and treated with ethanol (EOH) or estradiol (E2). Seventy-two hours later, cells were washed with PBS and cross-linked with 1% formaldehyde. After being washed twice with cold PBS, cells were lysed in buffer containing 5 mM PIPES [piperazine-N,N′-bis(2-ethanesulfonic acid) (pH 8.0)], 85 mM KCl, 0.5% Nonidet P-40, and protease inhibitors. Cell nuclei were pelleted and resuspended in a buffer containing 50 mM Tris-Cl (pH 8.1), 10 mM EDTA, 1% SDS, and protease inhibitors.

Soluble chromatin was prepared by sonication of transfected cells or brain tissue, with an average size of sheared fragments of about 400 bp, and diluted in buffer containing 16.7 mM Tris-Cl (pH 8.1), 0.01% SDS, 1.1% Triton X-100, 1.2 mM EDTA, 167 mM NaCl, and protease inhibitors. Immunoprecipitates were prepared as described above, and DNA fragments were detected by PCR, using 1 μl of a 50-μl DNA extract as the template for 25 cycles. The sequences of the PCR primers for the human KAI1 gene promoter were 5′-GGATGGGGTGGGCTCGAAG-3′ (upstream) and 5′-CGCCCCCAGAAGACACGC-3′ (downstream). For the mouse KAI1 gene promoter, the sequences of the primers were 5′-ACCGTTAGGCAGCGCCGTGAG-3′ (upstream) and 5′-CTTGGGAAGGCGGTGCGCTC-3′. The amplified DNA fragment corresponds to positions −413 to −69 of the KAI1 gene promoter, containing an NF-κB site (2, 8, 39).

Cell survival and apoptosis assays.

Transfected cells were washed with phosphate-buffered saline and fixed in 3.7% formaldehyde. The viability of transfected cells in each well was determined by counting the total number of green cells in each well under a fluorescence microscope. For the detection of apoptotic cells, fixed cells were stained with DAPI, and cells positive for fluorescence were observed with a Leitz Orthoplan 2 microscope. Representative micrographs were captured by a charge-coupled device camera with the Smart Capture program (Vysis, Downers Grove, IL). The apoptotic indices of transfected cells were determined by scoring 300 green fluorescent protein (GFP)-positive cells for chromatin condensation and apoptotic body formation.

Caspase-3 activity assay.

Caspase-3 activity was determined by use of a caspase-3 fluorometric assay kit (R&D Systems), following a protocol from the vender. Absorption at 405 nm was measured in an MRX microplate reader (Dynex Technologies, Chantilly, VA). Caspase-3 activity is expressed in optical density units (at 405 nm) per μg of protein.

RESULTS

17β-Estradiol inhibits the Fe65-dependent transcriptional activity of the APPct protein fused to Gal4 DNA binding domains.

To test the effect of estrogens on the activity of the APPct complex, ER-negative HeLa cells were transfected with Fe65 and Gal4-APPct together, with or without ERα. The activity of the cotransfected Gal4 reporter was determined in the presence or absence of 17β-estradiol. As shown in Fig. 1A, cells transfected with a mutant APPct that does not interact with Fe65, APPct* (4), contained no detectable reporter activity, whereas those with wild-type APPct contained large amounts of activity, confirming the transcriptional activity associated with APPct and its dependency on Fe65 (4). In cells expressing ERα, 17β-estradiol reduced the activity of the APPct complex (Fig. 1A). Inhibition by 17β-estradiol was not observed in cells lacking ERα expression, and the degree of inhibition depended on the amount of ERα (Fig. 1A and B). 17β-Estradiol did not affect Gal4-APPct or Fe65 expression (Fig. 1B) and thus apparently inhibits the specific activity of Gal4-APPct per molecule. As expected, 17β-estradiol induced the activity of an ERE-based reporter, and the degree of induction depended on the amount of ERα (Fig. 1C). Ectopically expressed ERα is known to have constitutive transcriptional activity, probably explaining why the expression of ERα also decreased the activity of the APPct complex to a small extent in the absence of 17β-estradiol (Fig. 1A).

FIG. 1.

Inhibition of transcriptional activity of Gal4-APPct by estrogens. (A) HeLa cells were transfected with 0.5 μg of 3×17merLuc, 0.1 μg of pLENβGal, 0.3 μg of HA-Fe65, 0.3 μg of Gal4-APPct* (APPct*) or Gal4-APPct (APPct), and the indicated amounts of pLENhERα (ERα) and treated with 10⁻⁸ M 17β-estradiol (E2) or ethanol (EOH) as a vehicle control for 48 h. Luciferase activity was determined and normalized with cognate β-galactosidase activity. (B) Cells were transfected and treated as described for panel A. Immunoblotting analyses were performed with the indicated antibodies. (C) HeLa cells were transfected with 0.5 μg EREe1bLuc, 0.5 μg pLENβGal, and the indicated amounts of pLENhERα. Cells were treated and luciferase activity determined as described for panel A. (D) ER-positive Ishikawa cells were transfected as described for panel A, but without pLENhERα. Cells were treated, and luciferase activity was determined.

To rule out the possibility that the inhibition of APPct activity by estrogens is an artifact of ectopic ERα expression in ER-negative cells, the transcriptional activity of the APPct complex was measured in ER-positive Ishikawa cells in the presence or absence of 17β-estradiol. As shown in Fig. 1D, 17β-estradiol significantly reduced the activity of the APPct complex. The degree of reduction increased as the time of treatment was prolonged. The data suggest that endogenous ERα is sufficient to mediate the inhibition of the APPct complex by estrogens.

ERα forms a complex with the Fe65-APPct complex.

ERα interacts with the transcriptional coactivator Tip60 (3), as does the APPct complex (4). Thus, ERα may inhibit the activity of the complex by sequestering Tip60. To test this possibility, we asked whether ectopic expression of Tip60 would negate the inhibitory effects of 17β-estradiol on APPct activity in cells expressing APPct and ERα. As shown in Fig. 2A, 17β-estradiol inhibited the activity of the APPct complex in cells cotransfected with Tip60. The degree of inhibition was comparable to that in cells lacking ectopic Tip60 expression (Fig. 1A). This finding suggests that activated ERα does not inhibit APPct activity by competing with the APPct complex for the limited pool of Tip60. Ectopic Tip60 decreased APPct activity, likely due to the “squelching” of other cofactors away from the APPct transcriptional complex.

FIG. 2.

Complex formation between ectopic ERα and Fe65 in HeLa cells. (A) Cells were transfected as described in the legend to Fig. 1A, with or without 0.3 μg of HA-Tip60, and treated for 48 h. Luciferase activity was determined and normalized with β-galactosidase activity. (B) Cells were transfected with 2 μg of pLENhERα, with or without 4 μg of HA-Fe65, and treated with EOH or 10⁻⁸ M E2 for 72 h. Immunoprecipitation (IP) analyses were performed, followed by immunoblotting (IB) with the indicated antibodies.

To define the mechanism underlying the estrogen inhibition of APPct activity, coimmunoprecipitation analyses were performed to test whether ERα forms a complex with Fe65, the adaptor essential for APPct to regulate transcription (4). In HeLa cells coexpressing HA-Fe65 and ERα, both proteins were present in anti-HA immunoprecipitates, regardless of whether cells received vehicle or 17β-estradiol (Fig. 2B, left panels). Because 17β-estradiol decreased the level of total ERα protein in the cells, the data in Fig. 2B suggest that the hormone enhances the specific interaction between ERα and Fe65, even though the ERα signals are comparable between anti-HA immunoprecipitates prepared from cells treated with vehicle and 17β-estradiol. Anti-HA antibody did not coprecipitate ERα in the absence of HA-Fe65 and thus does not cross-react with ERα. The protein detected in anti-HA precipitates of cotransfected cells was indeed ERα, as it was only detected when ERα was expressed. Reciprocal coimmunoprecipitation analyses showed that anti-ERα also precipitated Fe65 together with ERα, confirming the complex formation (Fig. 2B, right panels).

Since Fe65 forms a complex with both APPct and ERα, it is very likely that ERα and APPct coexist in the same complex, using Fe65 as an intermediate. Consistent with this idea, our immunological analyses detected little interaction (Fig. 3A) of ectopically expressed ERα and Gal4-APPct in HeLa cells without Fe65. In cells cotransfected with HA-Fe65, however, ERα and Gal4-APPct were able to interact, as indicated by the presence of Gal4-APPct in anti-ERα precipitates (Fig. 3A). This interaction occurred in the presence of either vehicle or 17β-estradiol.

FIG. 3.

Complex formation between ERα and APPct mediated through Fe65. (A) HeLa cells were transfected with 4 μg of Gal4-APPct and pLENhERα, with or without 4 μg of HA-Fe65, and treated for 72 h. Coimmunoprecipitations were performed with the indicated antibodies. (B) ERα protein, either alone (lanes 3 and 4) or together with Fe65 protein (lanes 1 and 2), was produced by in vitro transcription-coupled translation in the absence or presence of E2. GST pull-down assays were performed, and the precipitated proteins were detected by immunoblotting. An image of Coomassie blue (C.B.) staining was included to show that similar amounts of GST and GST-APPct were used for the pull-down assays.

To clearly show the Fe65-dependent association of ERα and APPct, pull-down assays were performed using GST-APPct and recombinant Fe65 and ERα produced by in vitro transcription-coupled translation reactions. When all three proteins were included in the assay, GST-APPct precipitated both ERα and Fe65 (Fig. 3B). GST alone had no effect, and GST-APPct did not precipitate ERα in the absence of Fe65. Coomassie blue staining showed that equal amounts of GST-APPct protein were used for the pull-down reactions.

Using GST pull-down assays, we further mapped the domain of ERα that interacts with Fe65 and APPct (Fig. 4A). In reaction mixtures containing Fe65 in vitro, similar amounts of full-length ERα protein and truncated ERα protein lacking the amino-terminal A/B region (ER179C) were specifically precipitated by GST-APPct. Consistently, the ERα protein lacking the carboxyl-terminal LBD (ERN282G) did not interact with Fe65 and APPct in vitro. More importantly, GST-APPct failed to precipitate a full-length ERα protein that was mutated at three amino acids essential for AF-2 activity (ER-TAF1) (47). Consistent with the binding analyses, reporter assays showed that 17β-estradiol inhibited APPct activity in cells expressing N-terminally truncated ERα but did not inhibit APPct activity in cells expressing C-terminally truncated ERα or the AF-2 mutant (Fig. 4B). Overall, the analyses show that the AF-2 region is essential, whereas the amino-terminal A/B region, including AF-1, is expendable for the interaction of ERα with Fe65 and APPct. They also show that the region required for interaction is also required for inhibition of the activity of the APPct complex by estrogens.

FIG. 4.

Defining the domain of ERα required for interaction with and inhibition of the APPct complex. (A) Full-length ERα (lanes 1), ERα with a truncated N-terminal A/B region (ER179C) (lanes 2), ERα with a truncated C-terminal LBD (ER282G) (lanes 3), and an ERα mutant with inactive AF-2 (ER-TAF1) (lanes 4) were produced together with Fe65 by in vitro transcription/translation reactions in the presence of E2. GST pull-down assays were performed. Precipitated proteins were detected by immunoblotting analyses with anti-ERα (C-terminal, top panel), anti-ERα (N-terminal, middle panel), and anti-Fe65 (lower panel) antibodies. Coomassie blue (C.B.) staining data were included to show that similar amounts of GST and GST-APPct were used for the pull-down assays. (B) HeLa cells were transfected and treated as described in the legend to Fig. 1A, except that 0.3 μg of PRST7-hERα or a deletion mutant was used for transfections. Luciferase activity was determined and normalized with cognate β-galactosidase activity.

Interaction and colocalization between endogenous ERα, Fe65, and APP in mouse brains.

Since full-length APP interacts with Fe65 through its carboxyl-terminal region, it should also interact with ERα in the same way as does Gal4-APPct. Indeed, anti-ERα antibody coprecipitated APP from extracts of HeLa cells cotransfected with ERα, APP, and Fe65 (Fig. 5A). ERα did not interact with APP in the absence of ectopic Fe65 or with APP containing a mutation in its Fe65 binding site (APP*). These data demonstrate the Fe65 dependence of the APP-ERα interaction.

FIG. 5.

Interaction and colocalization of endogenous ERα, Fe65, and APP in brains of APP and presenilin 1 double-transgenic (TG) and nontransgenic (non-TG) mice. (A) HeLa cells were transfected with 2 μg of pLENhERα and 4 μg of HA-Fe65, together with 4 μg of APP or APP mutant (APP*), and treated for 72 h. The interaction between full-length APP and ERα was determined by coimmunoprecipitation. (B to D) Coimmunoprecipitations were performed with homogenized brain tissues of female mice. The levels of endogenous proteins in mouse brain tissues or the precipitates were determined by immunoblotting. Extracts of HeLa cells transfected with control or ERα, Fe65, and APP expression vectors were included as controls. One mouse brain tissue was used per assay, and the data were reproduced twice. (E) Imaging analyses of female mouse brain sections through the area of the hippocampus. (E1) Coronal section through the hippocampus, with high-intensity Fe65 signal. (E2) Confocal images of tissue sections of a selected hippocampus area (see the white square in panel E1) after double staining for endogenous Fe65 (red) and ERα (green). (E3) Confocal images of tissue sections of the same hippocampus area after double staining for endogenous APP (red) and ERα (green).

Although transcriptional activity associated with APPct has been demonstrated nicely in transfection assays, endogenous APPct in neuronal cells is difficult to detect at the protein level, presumably due to its instability after cleavage by γ-secretase (20). Since ERα interacts with full-length APP in cotransfection assays (Fig. 5A), we performed coimmunoprecipitations to determine whether endogenous ERα, Fe65, and APP interact in the brain tissues of control mice and transgenic mice that express APP with the “Sweden” mutation and presenilin 1 with a deletion of exon 9, a mutation that corresponds to a form of early-onset AD (19). This transgenic mouse strain was included in the analyses because it contains increased APP and, presumably, increased APPct due to increased APP as well as γ-secretase activity as a result of increased presenilin 1 expression. As shown in Fig. 5B, anti-Fe65 antibody, but not rabbit IgG, coimmunoprecipitated endogenous ERα and Fe65 from brain homogenates of control mice. In reciprocal analyses, anti-ERα antibody coprecipitated both endogenous Fe65 and APP (Fig. 5C). More Fe65 and APP were coprecipitated from transgenic than from wild-type mice. The levels of Fe65 and ERα expression were comparable among wild-type and transgenic mice, and as expected, the level of APP expression was higher in transgenic than in nontransgenic mice (Fig. 5D). It appears that the interaction between Fe65 and ERα is greater in transgenic than in wild-type mice. The exact reason is unknown, but it is possible that greater APP expression somehow enhances the interaction through feedback.

Colocalization analyses with immunofluorescence, using sections cut through the hippocampus area (Fig. 5E1), were performed to confirm the interaction between ERα, Fe65, and APP in vivo. In this experiment, endogenous Fe65 and ERα (Fig. 5E2) were coexpressed in the majority of the neurons, and confocal imaging showed that the colocalization mainly occurred in the nucleus. In contrast to ERα, which is predominantly localized in the nucleus, APP is predominantly localized in the cytoplasm (Fig. 5E3). However, there were neurons in which cytoplasmic ERα or nuclear APP was detected (Fig. 5E3). In those cells, colocalization with ERα in the cytoplasm was also detected by confocal imaging. Overall, these immunological analyses show that the interaction between ERα, Fe65, and APP demonstrated in cell lines occurs in vivo in mouse brains.

Suppression of the ability of the APPct complex to induce KAI1 expression by estrogens via recruitment of ERα to the promoter.

Since the transcriptional activity of the APPct complex has just recently been realized, only a few target genes are known. These include the gene encoding KAI1, or tetraspanin (2, 32), a cell surface molecule acting as a tumor metastasis suppressor. Therefore, we tested the effects of estrogens and the APPct complex on the activity of the KAI1 gene promoter in ERα-positive N2a neuroblastoma cells. Using a GFP vector as a marker, we first showed that >40% of N2a cells were susceptible to transfection under our conditions (data not shown). We next showed that ectopic expression of APPct, Fe65, and Tip60 increased the expression of KAI1 and that treatment of cells with 17β-estradiol suppressed the induction (Fig. 6A). This experiment showed that the hormone-mediated inactivation of the APPct complex demonstrated in reporter assays can be translated into the suppression of endogenous gene expression in neuronal cells. The estrogen suppression was reversed by cotreatment with the antiestrogen ICI 182,780, showing that the estrogen effect is mediated through the endogenous ER.

FIG. 6.

Induction of KAI1 by the APPct complex and its suppression by 17β-estradiol. (A) N2a cells were transfected with 2 μg of control vector, Gal4-APPct, HA-Fe65, and HA-Tip60, as indicated. After treatment with EOH, 10⁻⁸ M E2, or 10⁻⁸ M E2 plus 10⁻⁷ M ICI 182,780 for 48 h, the expression of KAI1 and β-actin was determined by immunoblot analyses. The β-actin blot was included as a loading control. (B) 293 cells were transfected with 4 μg of HA-Fe65, Myc-Tip60, pLENhERα or control vector, and pCMV-APP or pCMV-APP*. After treatment with EOH or 10⁻⁸ M E2 for 48 h, CHIP assays were performed with the indicated antibodies. (C) Cells were transfected as described for panel B, but with or without HA-Fe65. Forty-eight hours later, the cells were treated with EOH or E2 for 45 min, and CHIP assays were performed. (D) Brain tissues were dissected from female APP and presenilin 1 double-transgenic mice, and CHIP assays were performed with the indicated antibodies.

To test whether ERα is recruited to the promoter of the KAI1 gene, 293 cells were transfected with Fe65, Tip60, and APP together, with or without ERα, and the presence of ERα, Fe65, Tip60, and APP on the promoter was analyzed by CHIP assays. Mutant APP (APP*) that lacks the ability to interact with Fe65 was included as a control. As shown in Fig. 6B, Tip60 was recruited to the promoter of the KAI1 gene in cells expressing wild-type APP but not mutant APP. Interestingly, ERα was recruited to the KAI1 gene promoter, which was enhanced by treatment of cells with 17β-estradiol (Fig. 6B). Because 17β-estradiol decreases total amounts of ERα (data not shown), the estrogen regulation of the recruitment is more profound than what appears in Fig. 6B. Concomitant with the increased ERα recruitment in 17β-estradiol-treated cells, the amounts of Tip60 on the promoter were decreased. Immunoblotting analysis showed that 17β-estradiol did not alter the levels of APP, Fe65, and Tip60 expression in the transfected cells (data not shown). These data suggest that the inhibition of the transcriptional activity of the APPct complex by estrogens is due to the interference of the recruitment of coactivators to promoters by ERα.

Further CHIP analyses showed that treatment with 17β-estradiol for 45 min increased ERα and reduced Tip60 recruitment to the KAI1 gene promoter, suggesting that the estrogen effect does not require long estrogen treatments (Fig. 6C). More importantly, in cells lacking Fe65 expression, ERα was not recruited to the KAI1 gene promoter. This demonstrates that ERα occupancy depends on the presence of the APPct-Fe65 complex. In mouse brains, ERα was recruited to the promoter of the KAI1 gene together with other components of the APPct complex (Fig. 6D), indicating that the CHIP data obtained with transfected proteins truly reflect what happens in vivo in mouse brains.

Estrogen suppression of neuronal cell apoptosis induced by the APPct complex.

To determine whether inhibition of APPct activity by estrogen had biological consequences, we performed the following experiment. N2a cells were transfected with GFP, APPct, Fe65, and either Tip60 or vector alone. The transfected cells were then treated with ethanol or 17β-estradiol, and the number of GFP-positive cells in each well was counted. Ectopic expression of APPct and Fe65 reduced the number of N2a cells, and coexpression of Tip60 slightly accentuated this effect (Fig. 7A). Treatment of cells with 17β-estradiol prevented the decrease in cell number induced by the APPct-Fe65-Tip60 complex.

FIG.7.

Estrogen protection of cells from APPct complex-induced apoptosis and its dependency on the ER. (A) N2a cells were transfected with 0.3 μg of pEGFP, Gal4-APPct, and HA-Fe65 together, with or without Myc-Tip60. After treatment with EOH or 10⁻⁸ M E2 for 72 h, the number of transfected (green) cells in each well was counted. Triplicate samples were analyzed for each data point, and the data were reproduced. (B) N2a cells transfected and treated as described for panel A were fixed and stained with DAPI. Representative micrographs are shown. (C) N2a cells were transfected and treated as described for panel A. The apoptotic index of transfected cells was determined. Triplicate samples were analyzed per data point, and the graph represents three independent experiments. (D) Control (upper panels) or differentiated (lower panels) N2a cells were transfected with GFP and stained with DAPI. Neurofilament (green) was visualized by immunofluorescence with anti-MAP2 antibody. cAMP, cyclic AMP. (E) Differentiated N2a cells were transfected with 0.3 μg of APPct, Fe65, and Tip60 and treated with EOH or E2. The apoptotic index of GFP-positive cells was determined as described for panel C. (F) N2a cells were transfected with 4 μg of plasmids, as indicated. Forty-eight hours later, cells were treated for 45 min, and caspase-3 activity was determined. Triplicate samples were analyzed for each data point, and the data were reproduced. (G) HeLa cells were transfected and treated as described for panel A. The only difference is that 0.3 μg of pLENhERα or a control vector was included in the transfection mix. The number of transfected cells was determined. (H) Cells were transfected as described for panel A and stained with DAPI. Representative micrographs are shown. (I) HeLa cells were transfected and treated as described for panel A. The apoptotic index of transfected cells was determined.

To determine whether the decrease in the number of GFP-positive cells by the APPct complex involves apoptosis, transfected N2a cells were fixed and stained with DAPI, and the nuclear morphology of green cells was examined for features of apoptosis under a fluorescence microscope. As shown by representative micrographs in Fig. 7B, cells transfected with empty vector displayed a normal morphology similar to that of surrounding nontransfected cells. Cells transfected with APPct, Fe65, and Tip60 and treated with ethanol frequently displayed nuclear condensation or fragmentation, an apoptotic morphology. In contrast, cells transfected with APPct, Fe65, and Tip60 and treated with 17β-estradiol showed a normal morphology similar to that of cells transfected with empty vector. The data suggest that apoptosis is at least part of the mechanism responsible for the decrease in cell number induced by the APPct complex.

The apoptotic index was then determined by calculating the percentage of apoptotic cells in 300 randomly selected green cells. As shown in Fig. 7C, apoptotic indices for EOH-treated cells were 8% for control cells (vector only), 20% for cells receiving APPct and Fe65, and 28% for cells receiving APPct, Fe65, and Tip60. Apoptotic indices for 17β-estradiol-treated cells were similar to those for control cells, regardless of whether they expressed APPct and Fe65 or APPct, Fe65, and Tip60. These analyses show that the APPct complex induces the apoptosis of N2a cells and that this function is suppressed by estrogen treatment.

To test whether the estrogen suppression of apoptosis induced by the APPct complex also occurs in differentiated neurons, N2a cells were induced to differentiate by serum withdrawal in combination with cyclic AMP treatment (Fig. 7D). Cells were then transfected with APPct, Fe65, and Tip60 and assayed for apoptosis in the presence of 17β-estradiol or vehicle. In the absence of 17β-estradiol, Fe65 and Tip60 did not induce the apoptosis of differentiated N2a cells, but APPct was slightly apoptotic; 17β-estradiol treatment had little effect on the basal level of apoptosis (Fig. 7E). Transfection of APPct with Fe65 and with both Fe65 and Tip60 increased the percentage of apoptotic cells four- and fivefold, respectively. Treatment of cells with 17β-estradiol prevented apoptosis induced by APPct, Fe65, and Tip60. Consistent with the effect of estrogens on apoptosis induced by the APPct complex, caspase-3 activation by the APPct complex was also inhibited by 17β-estradiol (Fig. 7F). The effect was observed in cells treated with the hormone for either 45 min (Fig. 7F) or 48 h (data not shown). The caspase-3 analyses show that the apoptosis induced by the APPct complex is likely through the intrinsic apoptotic pathway that involves caspases. The suppression by a short estrogen treatment indicates that the estrogen effect is rather direct. Overall, these analyses show that the induction of apoptosis by the APPct complex and the protection by estrogens also occur in differentiated neurons.

Because N2a cells express ERα, it is impossible to tell whether the suppressive effect of estrogens on apoptosis induced by the APPct complex is mediated through the ER. To address the issue of ER dependency, ER-negative HeLa cells were transfected with or without ERα, and the effect of estrogens on apoptosis induced by the APPct complex was examined. As shown in Fig. 7G, 17β-estradiol had no effect on the number of transfected cells in the absence of ectopic ERα expression. After cotransfection with ERα, 17β-estradiol blocked both the decrease in cell number (Fig. 7G) and the increase in apoptotic index (Fig. 7H and I) induced by the APPct complex. These findings demonstrate that the estrogen effect on apoptosis induced by the APPct complex is mediated through the ER.

Essential role of complex formation between ERα and Fe65 in mediating estrogen suppression of the activity and function of the APPct complex.

Our analyses so far have shown that the estrogen suppression of the transcriptional and biological activities of the APPct complex correlates very well with the ability of the ERα to form a complex with Fe65, suggesting a cause-and-effect relationship. However, the data have not ruled out the possibility that the estrogen suppression is indirectly mediated through an estrogen target gene whose expression is regulated by the activated ERα through its binding to an ERE. A mutant ERα that does not bind Fe65 but is active on an ERE-based reporter gene would be helpful to distinguish between these two possibilities.

Previous studies have shown that mutation of tyrosine-537 to alanine produces a mutant ERα that is constitutively active on ERE-based reporter genes (49), a phenotype we confirmed in reporter analyses (Fig. 8A). Because tyrosine-537 is located in the AF-2 region that is critical for Fe65 interaction (Fig. 4), we tested the ability of this mutant ERα to bind Fe65 and inhibit the activity of the APPct complex. As shown in Fig. 8, this mutant ERα failed to interact with Fe65 (Fig. 8B) and to suppress the ability of the APPct complex to induce transcriptional activation (Fig. 8C) or apoptosis (Fig. 8D). As controls, the wild-type and mutant ERα proteins were expressed to comparable levels, and they both did not decrease the level of ectopic expression of APPct or Fe65 (data not shown). These analyses link the inhibitory effect of estrogens on the APPct complex to the ability of ERα to bind Fe65 and separate the effect from the classic nuclear actions of the hormone on transcription mediated through EREs.

FIG. 8.

Effect of tyrosine-537 mutation on ERα-mediated suppression of the transcriptional activity of the Fe65 complex and its ability to induce apoptosis. (A) HeLa cells were transfected with 0.5 μg of EREe1bLuc, 0.5 μg of pLENβgal, and 0.3 μg of pLENhERα or pLENhERαY537A, as indicated. Cells were treated with 10⁻⁸ M E2 or EOH for 48 h, and the luciferase activity was determined. (B) HeLa cells were transfected with 4 μg of HA-Fe65 and 2 μg of pLENhERα or pLENhERαY537A and treated with 10⁻⁸ M E2 for 72 h. The interaction between Fe65 and wild-type or mutant ERα was determined by coimmunoprecipitation. (C) HeLa cells were transfected with 0.5 μg 3×17merLuc, 0.1 μg pLENβGal, 0.3 μg Gal4-APPct* or Gal4-APPct, 0.3 μg HA-Fe65, and 0.3 μg pLENhERα or ERαY537A, as indicated. Cells were treated with 10⁻⁸ M E2 or EOH for 48 h, and luciferase activity was determined. (D) HeLa cells were transfected with 0.3 μg of pEGFP, APPct, Fe65, ERα, or ERαY537A, as indicated, and treated with EOH or E2. The apoptotic index of GFP-positive cells was determined. (E) Estrogen action through “tethering” of the ER to the Fe65-APP complex. APP forms a complex with Fe65 on the plasma membrane through its carboxyl terminus. After cleavage by γ-secretase, Fe65 moves to the nucleus, recruits Tip60, and induces neuronal cell apoptosis by activating the transcription of target genes. The activated ER binds to the Fe65 transcriptional complex and suppresses gene activation and cell death by decreasing the amount of Tip60 coactivator in the complex. Using Fe65 as an adaptor, the ER may also interact with full-length APP in the cytoplasm to regulate its processing by proteolytic enzymes.

DISCUSSION

While the molecular mechanism underlying estrogen action in the reproductive system is well defined, little is known about estrogen action in neuronal cells. The present study reveals a novel molecular mechanism by which physiological levels of estrogens protect neuronal cells from apoptosis triggered by γ-secretase-mediated cleavage of APP, a pathological process characteristic of neurological disorders such as AD. This mechanism involves the “tethering” of ERα to the Fe65 transcriptional complex on gene promoters (Fig. 8E). In contrast to the typical genomic action of estrogens mediated through ERα sitting on ERE-containing promoters, which uses helix-12 as a main activator to recruit coactivators, this novel mode of estrogen action involves the use of the helix-12 region as an interaction motif to bind an adaptor molecule, leading to a decrease in the recruitment of a transcriptional coactivator. Multiple lines of evidence are presented here to support this mechanism. First, 17β-estradiol inhibited the transcriptional activity of the APPct complex, and the endogenous ERs in Ishikawa cells were sufficient to mediate this action. Second, ERα formed a complex with APPct and full-length APP via Fe65 as an adaptor both in vitro and in vivo. Complex formation was shown for both ectopic proteins and endogenous proteins in the mouse brain, and the amounts of complex were greater in brain tissues of transgenic mice expressing mutant presenilin 1 and APP than in brain tissues of wild-type mice. Third, the transcriptional inhibition observed in reporter assays translated into estrogen suppression of the expression of endogenous genes induced by the APPct complex (for example, the KAI1 gene). CHIP assays detected ERα bound to the KAI1 gene promoter, which was increased by 17β-estradiol treatment. Fourth, apoptotic assays showed that 17β-estradiol impaired the ability of the APPct complex to induce apoptosis in transformed and differentiated neuronal cells and that such an impairment depended on ERα. Finally, the alanine-537 ER mutant was constitutively active on an ERE-based reporter but was unable to bind Fe65 or mediate the suppressive effects of estrogens on the APPct complex.

The analysis with amino-terminally truncated ER shown in Fig. 4 suggested that the N terminus of ERα is expendable, but it did not rule out the potential involvement of AF-1 in mediating the inhibition of the APPct complex by estrogens in the context of the full-length receptor. Similarly, the analysis with the AF-2 mutants did not prove that AF-2 is essential for inhibition because the mutant receptor failed to interact with Fe65. The accurate assessment of the differential involvement of AFs is hindered by the lack of full-length ER mutants that are capable of binding to Fe65 but selectively deficient in the AFs. It is also important that the reduction in the activity of the APPct complex caused by ectopically expressed ERα in the absence of estrogens (Fig. 1) does not necessarily mean that the inactive receptor has the ability to suppress the activity of the APPct complex. Ectopically expressed ERα is known to display constitutive activity in transient reporter assay systems, presumably due to estrogenic compounds present in the medium or produced by the cells. Furthermore, many nonsteroid compounds that do not bind ERα cause ligand-independent activation of the receptor (41). Interestingly, one such compound is dopamine (36, 41). It will be interesting to find out whether dopamine or other neurotransmitters protect neurons from APPct-induced apoptosis by stimulating complex formation between Fe65 and ligand-independently activated ERα.

Consistent with the constitutive effect of ectopic ERα on APPct activity, complex formation also occurred between ectopically expressed ERα and Fe65 in the absence of estrogens. The inhibition of APPct activity by 17β-estradiol is also consistent with the enhancement of complex formation by the hormone. Even though binding assays did not detect an increase in the amount of ERα protein present in anti-HA-Fe65 immunoprecipitates after estrogen treatment, the specific interaction per molecule of ERα was clearly increased by estrogens. This is because the total level of ERα protein is decreased by estrogens (Fig. 1B and 2B), likely due to degradation of ERα by the ubiquitin-proteasome pathway during estrogen-induced transcriptional activation (30, 35). Consistent with the above assessment, complex formation between endogenous ERα and Fe65 in mouse brains is most likely estrogen regulated. This is suggested by the reduced complex formation detected in male and old mice (data not shown), as they contain fewer endogenous estrogens than do female and young mice.

Although the data in Fig. 4 and 8 show that complex formation between ERα and Fe65 is required for suppression of the activity of the APPct complex by estrogens, complex formation does not always lead to suppression. Complex formation also occurs in cells treated with tamoxifen, which increases rather than decreases the activity of the APPct transcriptional complex (data not shown). It is reasonable to conclude that complex formation only provides a platform for the subsequent ERα-mediated changes in the recruitment of transcriptional cofactors to the promoter, a process dependent on the properties of the ligand. This idea is supported by the decreased recruitment of the Tip60 coactivator to the KAI1 gene promoter caused by 17β-estradiol, which was associated with a concomitant increase in the recruitment of ERα protein to the promoter (Fig. 6B and C). Overall, the data suggest that complex formation between ERα and APPct mediated through Fe65 is required but may not be sufficient for the suppression of the transcriptional activity of the APPct complex and that the promoter context-specific displacement of transcriptional cofactors induced by ligands might be an additional determinant.

Fe65 belongs to a family of WW domain proteins, among which many have been shown to be involved in transcriptional regulation (42, 43). For example, the Yes kinase-associated protein binds to the polyoma enhancer binding protein 2 transcription factor through its WW domain and serves as a coactivator (52), and PQBP-1, a polyglutamine tract binding protein with a WW domain, inhibits transcription activation by Brn-2 (48). In addition, the WW domains of Ess1/Pin1 PPIase (43), Yes kinase-associated protein (12), and Rsp5/Nedd4 ubiquitin ligase (5) bind the carboxyl-terminal domain of RNA polymerase II, suggesting a fundamental role of WW domains in transcription. It will be interesting to find out whether the interaction between ERα and Fe65 can be extended to other WW proteins and whether the interaction represents a common means for transcriptional regulation by activated ERα.

Previous studies have shown that estrogens control the activation of kinases involved in neuronal death (15) or survival (1) through nongenomic actions or function as antioxidants (6). Some of these activities were demonstrated with high concentrations of estrogens, and the ER was found to not always be required. The suppression of APPct complex-induced apoptosis was clearly ER dependent and happened with 17β-estradiol at physiological concentrations. Moreover, the demonstration of the existence of ERα, Fe65, and APP in the same protein complex suggests a role of estrogens in regulating the pathogenesis of AD that is more direct than what earlier studies may have suggested. Because full-length APP formed a complex with the receptor via Fe65, it is possible that other biological processes involving APP are also regulated by estrogens. For example, estrogens have been shown to regulate the nonamyloidogenic processing of APP by α-secretase, which leads to the suppression of Aβ production by γ-secretase (33). This effect could also be mediated through complex formation between ERα and Fe65 that takes place outside the nucleus. Our confocal immunofluorescence analyses of mouse brain sections detected cytoplasmic signals of ERα in certain neurons (Fig. 5E), suggesting that it is possible for ERα to interact with APP outside the nucleus. It has also been shown that estrogens suppress the transport of proteins such as NF-κB (13) and glucocorticoid receptor (16) from the cytoplasm to the nucleus. The cytoplasmic APP-Fe65-ERα complex may also regulate the intracellular transport of neuronal proteins.

In contrast to animal experiments that consistently demonstrated a neuroprotective activity for estrogens, the Women's Heath Initiative (WHI) studies have indicated that estrogens, alone or in combination with synthetic progesterone, exert no significant benefit for AD and may even have an adverse effect (10, 40). Because the ages of women in the WHI study are about 65 years or older and the age of menopause is from 45 to 55, it is arguable that the use of estrogens in older women may exacerbate existing problems. Literature information has suggested that estrogens may protect women against the development of AD if applied early but are of little or no value once the damage has been done, i.e., after the disease has progressed to a clinically detectable state (18). Since the current study shows that estrogens suppress the transcriptional activation of genes by the APPct complex, the suppression obviously happens at early stages of the apoptotic process. The findings from the present study are thus more relevant to the preventive effect of estrogens on AD than to the cure of preexisting AD by estrogens. Continued investigation of the biological significance of the ERα and APP interaction may provide a better understanding of the opposite effects of estrogens observed in the WHI studies and in animal models. Along these lines, it is important that the expression of KAI1 is controlled by NF-κB and that the effect of the APPct complex on the expression of KAI1 is mediated through the NF-κB site (2). Like the ER, NF-κB plays complex roles in AD. It is antiapoptotic, so its transcriptional activity may be needed for neuronal survival. On the other hand, NF-κB is a proinflammatory transcriptional factor that may facilitate the progression of AD. It is important to identify the subset of NF-κB target genes suppressed by the APPct complex and those genes whose suppression by the complex is relieved by estrogens.