Mitochondrial localization of estrogen receptor β

Abstract

Estrogen receptors (ERs) are believed to be ligand-activated transcription factors belonging to the nuclear receptor superfamily, which on ligand binding translocate into the nucleus and activate gene transcription. To date, two ERs have been identified: ERα and ERβ. ERα plays major role in the estrogen-mediated genomic actions in both reproductive and nonreproductive tissue, whereas the function of ERβ is still unclear. In this study, we used immunocytochemistry, immunoblotting, and proteomics to demonstrate that ERβ localizes to the mitochondria. In immunocytochemistry studies, ERβ was detected with two ERβ antibodies and found to colocalize almost exclusively with a mitochondrial marker in rat primary neuron, primary cardiomyocyte, and a murine hippocampal cell line. The colocalization of ERβ and mitochondrial markers was identified by both fluorescence and confocal microscopy. No translocation of ERβ into the nucleus on 17β-estradiol treatment was seen by using immunocytochemistry. Immunoblotting of purified human heart mitochondria showed an intense signal of ERβ, whereas no signals for nuclear and other organelle markers were found. Finally, purified human heart mitochondrial proteins were separated by SDS/PAGE. The 50,000–65,000 M_r band was digested with trypsin and subjected to matrix-assisted laser desorption/ionization mass spectrometric analysis, which revealed seven tryptic fragments that matched with those of ERβ. In summary, this study demonstrated that ERβ is localized to mitochondria, suggesting a role for mitochondrial ERβ in estrogen effects on this important organelle.

Estrogens play an important role in development, growth, and differentiation of both female and male secondary sex characteristics. Estrogen receptors (ERs) were the first identified nuclear receptor family member (1). The first ER, now called ERα, was cloned in 1986 (2, 3). A second ERβ, was identified and cloned a decade later (4, 5). Like other members of the nuclear receptor superfamily, both ERs have a modular structure consisting of distinct functional domains (1). The DNA-binding domain (DBD) enables the receptor to bind its cognate target site consisting of an inverted repeat of two half-sites with the consensus motif AGGTCA spaced by 3 bp, referred to as an estrogen response element (ERE). The ligand-binding domain enables estrogen binding to the receptors. ERs are highly conserved between ERα and ERβ, with >95% homology for the DBD and ≈50% homology for the ligand-binding domain. Less homology is observed for the transactivational domain between ERα and ERβ (5, 6).

Genomic actions of ERα are well described (7). On binding to ERα, estrogens induce a conformational change in the ERα proteins, which is accompanied by the dissociation of the accessory protein, heat shock protein 90, thereby exposing the DBD. In the nucleus, the receptor–ligand complex binds to DNA and modulates gene transcription. This transcriptional/translational activation is comparatively slow and sensitive to cycloheximide and actinomycin D (8). The high homology of ERβ to ERα in their DBD and ligand-binding domain indicates that these receptors may regulate common gene networks and respond to similar ligands. It is clear that ERα plays a major role as a transcription factor in the reproductive tissues for both male and female, and both male and female ERα knockout mice are infertile (9, 10). On the other hand, little change of the transcription pattern in the reproductive tissue in the ERβ knockout mice is seen (11).

Increasing evidence supports nonclassical modes of ER action. Nongenomic actions of estrogens include cellular calcium homeostasis, induction of nitric oxide synthesis, and rapid activation of extracellular signal-regulated protein/mitogen-activated protein kinase pathways (8, 12). A membrane ER that colocalizes with caveolin 1 has been described, which is identical with ERα and involved in the activation of endothelium nitric oxide synthase and extracellular signal-regulated protein (13). Also, a putative plasma membrane-associated ER-X has been proposed, which is associated with estradiol-induced activation of the mitogen-activated protein kinase cascade (14). Similarly, a subpopulation of ERβ has also been localized in the plasma membrane of endothelial cell, which is involved in endothelium nitric oxide synthase signaling (15). Other cytoplasmic organelles have been described as containing ERs. Specific binding of estrogens to sites in the mitochondria has been described (16). Estrogens have consistently been indicated to modulate mitochondrial function, such as ATP production, mitochondrial membrane potential, and calcium concentration, although it is still unclear whether these actions are ER-dependent (17, 18).

In transfected cells, ERβ localizes to the nucleus (19), whereas the nuclear localization of ERβ in nontransfected cells is rarely reported. Further, ERβ is a poor transcriptional factor (20–24). The nuclear targeting of ERα is regulated cooperatively by multiple signals located in its hinge region (25). However, the hinge domain is one of the least conserved regions in ERβ compared with ERα (4, 5). The unique structural characteristics of the hinge domain of ERβ may lead to the differential intracellular targets of the receptor. In this study, we used immunocytochemistry, immunoblotting, and MS to demonstrate that ERβ is localized to mitochondria. Our data establish this ERβ localization in a variety of cell types, suggesting that estrogens can directly affect mitochondrial function through ERβ.

Materials and Methods

Chemicals and Reagents. 17β-Estradiol was obtained from Steraloids (Wilton, NH). Tissue culture material, Laemmli sample buffer, human recombinant ERβ (hrERβ, long form), DTT, Coomassie brilliant blue, 20% Tris-glycine gel, and SeeBlue Plus 2 protein standard mixture were obtained from Invitrogen. Charcoal-stripped FBS was from HyClone. MitoTracker Red, SlowFade Light Antifade reagent, and Alexa Fluor 488 goat anti-rabbit IgG were from Molecular Probes. ERβ (H-150) (Santa Cruz Biotechnology) is a rabbit polyclonal antibody raised against a recombinant protein corresponding to amino acids 1–150 mapping at the N terminus of ERβ of human origin. It reacts with ERβ of mouse, rat, and human origin by Western blotting, immunoprecipitation, and immunohistochemistry, and it is not cross-reactive with ERα. Z8P (Zymed) is an epitope-affinity-purified rabbit antiserum raised against an 18-aa synthetic peptide (468–485, CSTEDSKSKEGSQNLQSQ) derived from the C terminus of mouse ERβ protein. HPLC-grade acetonitrile, acetic acid, methanol, trifluoroacetic acid, and Optima water were purchased from Fisher Scientific. Sequencing-grade, modified trypsin along with the stability-optimized dilution buffer were obtained from Promega. Ultrapure SDS was purchased from Schwarz/Mann. Ammonium bicarbonate and α-cyano-4-hydroxycinnamic acid were obtained from Sigma.

For immunocytochemistry studies 17β-estradiol was initially dissolved in DMSO and diluted in DMEM media to the final concentration of 10 nM with final DMSO concentration <0.001%. For 17β-estradiol treatment, cells were treated with 17β-estradiol at 10 nM or 0.001% DMSO for 30 min, respectively. For mitochondrial labeling, cells were treated with 100 nM MitoTracker Red for 30 min in growth media. Then the media were removed, and immunostaining was done according to the manufacturers' protocol.

Cell Culture. Primary cerebral cortical and hippocampal neurons. Sprague–Dawley rat embryos (18 days old; from Charles River Breeding Laboratories) were externalized under halothane anesthesia. The cerebral cortex and hippocampus were dissected and harvested in 2 ml of preparation medium (DMEM, 4.5 g/liter glucose/100 units/ml penicillin/100 μg/ml streptomycin). The cortex and hippocampus were treated with trypsin. The tissue was washed three times with washing medium (Hanks' medium, 4.5 g/liter glucose/100 units/ml penicillin/100 μg/ml streptomycin), and individual cells were isolated by trituration 10 times with three different sizes of fire-polished Pasteur pipettes. The cells were harvested in seeding medium (DMEM, 4.5 g/liter glucose/100 units/ml penicillin/100 μg/ml streptomycin/2 mM glutamine/19% horse serum) and filtered through a 40-μm filter. The cerebral cortical cells and hippocampal cells were seeded in eight-well poly-l-lysine-treated chamber slide at the density of 20,000 cells per well. The cells were incubated in neurobasal medium (DMEM, 4.5 g/liter glucose/100 units/ml penicillin/100 μg/ml streptomycin/2 mM glutamine, B27) in normal cell culture conditions. The cells were subjected to immunocytochemistry staining at day 7.

Primary cardiomyocyte. Primary cardiomyocyte cultures were prepared from 2- to 4-day-old Sprague–Dawley rats. Cultures were maintained in medium 199 supplemented with 10% FBS (26).

Murine hippocampus cell line. HT-22 cells (gift from D. Schubert, Salk Institute, San Diego), which are an immortalized murine hippocampal cell line, were maintained in DMEM media supplemented with 10% charcoal-stripped FBS and 20 μg/ml gentamycin at 37°C in a humid atmosphere with 5% CO₂. HT-22 cells (passages 18–25) were seeded into eight-well chamber slides at a density of 9,000 cells per well.

Immunofluorescence Staining and Microscopy. Monolayer cells were washed with PBS (pH 7.4) and fixed with cold methanol for 15 min at –20°C. Cells were rinsed several times in PBS and incubated in ice-cold 0.2% Triton X-100 for 10 min to permeabilize the cells. Nonspecific sites were blocked for 1 h at room temperature with 5% normal goat serum and 5% BSA in PBS. Cells were then incubated with an ERβ antibody (H-150 or Z8P) at 1:50 dilution at 4°C overnight. The sections were washed for 30 min in PBS, then incubated with Alexa Fluor 488 goat anti-rabbit IgG (1:200) in 5% normal goat serum with 5% BSA in PBS for 1 h at room temperature. After washing in PBS for 30 min the cells were stained with 100 μM 4′,6-diamidino-2-phenylindole (DAPI) for 5 min. Cells were mounted with SlowFade Light Antifade reagent and covered with a coverslip. Three controls were included in each experiment, in which we omitted the primary or secondary antibody or MitoTracker Red. Samples were analyzed with either an Olympus microscope with appropriate excitation/emission filter pairs or a Zeiss LSM confocal microscope.

Purification of Mitochondria from Human Heart and Immunoblots of ERβ. Human heart mitochondria were isolated by differential centrifugation from three donor hearts (obtained from Analytical Biological Services, Wilmington, DE). The donors were between 16 and 64 years of age and showed no indication of cardiovascular disease. Mitochondria (40 mg total) were further purified by metrizamide gradient centrifugation (27), and their integrity and purity were assessed by Western blot analysis for several proteins, including actin, dynamin, prohibitin, and the unique endoplasmic reticulum-resident C-terminal sequence (KDEL), which is required for the retention of proteins in the endoplasmic reticulum (28). Mitochondrial preparations were not made from any other cell types in this study.

The extracted human heart mitochondria were combined in Laemmli buffer with β-mercaptoethanol and boiled for 5 min. Of the mitochondrial samples 30 μg were separated by 10% Tris-glycine polycrylamide gel (Gradipore, Frenchs Forest, Australia) and then transferred to a nitrocellulose membrane (Millipore). Lanes containing biotinylated protein standards (Cell Signaling Technology, Beverly, MA) or Kaleidoscope prestained standards (Bio-Rad) were used to evaluate the size of the bands detected with ERβ. As a positive control, rat cerebral lysate or hrERβ (long form) was assessed. The membranes were blocked for 1 h with 5% nonfat dry milk in PBS and were incubated overnight at 4°C with ERβ antibody (H-150) at a dilution of 1:1,000. The membranes were repeatedly washed with PBS before incubation with a secondary antibody (horseradish peroxidase-conjugated goat anti-rabbit IgG, Bio-Rad) at 1:5,000 in PBS. The blots were developed with an enhanced chemiluminescent kit (Pierce).

Identification of ERβ in Mitochondria by MS. For SDS/PAGE separation, the mitochondrial fraction was combined with 2× Laemmli sample buffer plus DTT (0.125 M Tris·HCl/4% SDS/40% glycerol/0.1% bromophenol blue, pH 6.8/9 mM DTT) and heated at 90°C for 10 min. Mitochondrial protein (60 μg) was then loaded into each well of a 4–20% gradient Tris-glycine gel for protein separation. For molecular weight calibration, 10 μl of SeeBlue Plus 2 protein standard mixture was added to the first well. After protein separation, the gel was stained overnight with 0.1% Coomassie brilliant blue R-250 (45% methanol/10% acetic acid). The gel was then placed in 5% acetic acid/20% methanol (vol/vol) to remove excess Coomassie brilliant blue that remained in the gel. A band that corresponded to molecular weights between 50,000 and 65,000 was excised from the gel, and the proteins were digested in-gel with trypsin as described by the Howard Hughes Medical Institute/Keck Facility at Yale University (http://info.med.yale.edu/wmkeck/prochem/geldig3.htm). The digested mitochondrial proteins from the gel were purified with ZipTip microcolumns (Millipore) for matrix-assisted laser desorption ionization (MALDI)–time-of-flight (TOF) MS. The tryptic peptides were concentrated onto a C18 ZipTip microcolumn, washed several times with 0.1% trifluoroacetic acid, and eluted off the column onto the MALDI plate with 1 μl of matrix solution. The matrix solution used was prepared by dissolving 10 mg of α-cyano-4-hydroxycinnamic acid in 1 ml of 60% acetonitrile/0.1% trifluoroacetic acid. MALDI mass spectra were acquired on a Voyager DE-Pro MALDI-TOF mass spectrometer (Applied BioSystems) operated in reflector mode. Ions were accelerated by 20 kV after an extraction delay time of 200 ns. Grid and guide wire voltages were adjusted to 72% and 0.01% of the acceleration voltage value, respectively. MALDI-TOF/MS values were a signal average of 50–100 laser shots and were mass-calibrated by using a four-point external calibration or internally calibrated (when applicable) to minor trypsin autolysis peaks at m/z 842 and 2,211. Peptide mass values obtained by the MALDI-TOF/MS analysis of in-gel tryptic digests were input into the ms-fit search program of Protein Prospector (http://prospector.ucsf.edu). The National Center for Biotechnology Information (NCBI) nonredundant (nr) database was used in the appropriate program search options for protein identification.

Results

Mitochondrial Localization of ERβ in Primary Neurons. To determine the purity of the neuronal culture, primary cortical and hippocampal neurons were stained with mitogen-activated protein 2, glial fibrillary acidic protein, and DAPI. A specific neuronal marker (mitogen-activated protein 2 staining) was seen in cells with a neuronal morphology, whereas no evidence of glia (glial fibrillary acidic protein staining) was seen (data not shown). MitoTracker Red is a mitochondrion-selective dye that is well retained during cell fixation. Primary hippocampal neurons were costained with ERβ (H150) and MitoTracker Red and were evaluated by using fluorescence microscopy. ERβ staining was predominantly in the cytosol and presented a punctuate distribution, similar to that of mitochondria. In primary hippocampal neurons, ERβ staining colocalized with MitoTracker Red. Other, yet unidentified cytosolic components were also stained with the antibodies used in these studies (Fig. 1). For primary cortical neurons, a similar colocalization of these two markers was seen (data not shown). To be certain that the two colors generated from the same mitochondria were not imaging artifacts caused by reciprocal detection of light emitted by MitoTracker Red or Alexa Fluor 488 (bleed-through), we omitted either MitoTracker Red or Alexa Fluor 488 conjugated secondary antibody, then performed immunocytochemistry. No signal was detected when cells were stained with MitoTracker Red and imaged for Alexa Fluor 488 and vice versa.

Fig. 1.

Fluorescence microscopy localization of ERβ to mitochondria in primary rat hippocampal neurons. (A) ERβ stained by using ERβ antibody, H150 (green). (B) Mitochondria stained with MitoTracker Red (red) and nuclei stained with DAPI (blue). (C) Merged image of ERβ and mitochondria (yellow).

Mitochondrial Localization of ERβ in Primary Cardiomyocytes. To determine whether this apparent mitochondrial localization of ERβ was limited to neurons, colocalization of ERβ and mitochondrial markers was assessed in primary cardiomyocytes. In primary cardiomyocytes, ERβ (H-150) staining exhibited a punctuate cytoplasmic distribution, which colocalized with MitoTracker Red (Fig. 2A). A second antibody, Z8P, which detects 18 aa in the C terminus of ERβ, showed similar punctuate staining that also colocalized with MitoTracker Red (Fig. 2B).

Fig. 2.

Fluorescence microscopy localization of ERβ to mitochondria in rat primary cardiomyocytes. (Aa) ERβ stained by using ERβ antibody, H150 (green). (Ab) Mitochondria stained with MitoTracker Red (red) and nuclei stained with DAPI (blue). (Ac) Merged image of ERβ and mitochondria (yellow). (Ba)ERβ stained by using ERβ antibody, Z8P (green). (Bb) Mitochondria stained with MitoTracker Red (red) and nuclei stained with DAPI (blue). (Bc) Merged image of ERβ and mitochondria (yellow).

Mitochondrial Localization of ERβ in HT-22 Cells by Confocal Microscopy. In HT-22 cells, ERβ H-150 immunostaining exhibited a granular cytoplasmic distribution regardless of the fixative used (acetone/methanol or paraformaldehyde). The staining of the ERβ coincided precisely with MitoTracker Red staining (Fig. 3). Moreover, no translocation of ERβ into nucleus on 17β-estradiol treatment for 0.5 h was seen (Fig. 3). Further, no ERβ signal was seen in the cells in which the primary antibody was omitted (data not shown).

Fig. 3.

Confocal microscopy of the colocalization of ERβ (H-150) with MitoTracker Red in HT-22 cells. ERβ was stained with ERβ (H150) as green, mitochondria with MitoTracker Red as red, and the merged image of ERβ and mitochondria as yellow. Cells were treated with DMSO (Left)or17β-estradiol (17-beta-E2, 10 nM) (Right) for 0.5 h. No nuclear translocation of ERβ was indicated after 17β-estradiol treatment. No ERβ staining was seen when H150 was omitted (data not shown).

Immunoblot of ERβ in Purified Mitochondria from Human Heart. To confirm our immunocytochemistry study, immunoblots of ERβ were performed with purified human heart mitochondrial protein. The nuclear marker (histone H1) was not observed, intense staining for the mitochondrial enzyme manganese superoxide dismutase (MnSOD) was evident, and a band with a molecular weight of 60,000 that reacted with the ERβ antibody, H-150, was seen in the mitochondrial lysate (Fig. 4A). Rat cerebral lysate was used as positive control. MnSOD and histone H1 were seen in the cerebral lysate (Fig. 4A). As an additional positive control, we used full-length hrERβ, and a band with the molecular weight of ≈60,000 that stained with ERβ H150 was seen for both human heart homogenate and purified human heart mitochondria (Fig. 4B). ERβ was more concentrated in the mitochondrial preparation than in the whole lysate. In a separate experiment, immunoblot of ERβ was performed by using purified human heart mitochondrial protein from a different supplier (Molecular Probes), and the same results were obtained (data not shown).

Fig. 4.

Identification of ERβ in mitochondrial preparation. (A) MnSOD and histone H1 were used as a mitochondrial and a nuclear marker, respectively. Strong signals for ERβ and MnSOD were evident, but no nuclear (histone H1) signal was seen in the mitochondrial lysate (lane a). In cerebral lysate (lane b), ERβ, MnSOD, and histone H1 were detected. (B) Full-length hrERβ was used as positive control (lane b). A band with the same molecular weight as hrERβ was seen in both human heart homogenate (lane c) and purified human heart mitochondria (lane d). Biotinylated protein standards (lane a) were used as molecular weight markers. M_r values of 80,000, 60,000, and 50,000 are shown.

Identification of ERβ in Human Heart Mitochondria by MS. Gel electrophoresis monitored by in-gel protease (trypsin) digestion and MALDI-TOF/MS and sequence database searching were used for protein identification (29, 30). Identification of membrane or membrane-bound proteins such as receptors, transport channels, and ectoenzymes is difficult by using 2D gel electrophoresis. To circumvent this shortcoming, we used a 1D (SDS/PAGE) separation before trypsin digestion and MS. A section of the gel containing protein with an apparent relative molecular weight (M_r) of 50,000–65,000 and, therefore, expected to contain the ERβ, was excised, digested in-gel with trypsin, and analyzed by MALDI-TOF/MS. As indicated in Fig. 5, several tryptic peptides were generated and were easily resolved for database search. Consistent with our immunoblots, numerous intense tryptic fragments of ATP synthase were found in our mitochondrial lysate, which indicated high purity of the mitochondrial preparation. In addition, monoisotopic masses input into ms-fit also matched to several tryptic peptides derived from human ERβ proteins, as shown in Tables 1 and 2. The sequences matched include both N-terminal and C-terminal epitopes to which the ERβ antibodies, H150 and Z8P, respectively, are directed. H150 is directed to the fragment of N-terminal 150 amino acids, and we found fragments of 1–18 and 59–68. Z8P is directed to the fragment of C-terminal 18 aa (CSTEDSKSKEGSQNLQSQ), and SKEGSQNPQSQ was the tryptic peptide present. The DBD (147–154), ligand-binding domain (339–348), and hinge domain (179–191) sequences were also found in the matched sequences. The matched fragments were in fairly large amounts, indicating that these fragments are not likely the result of contamination from the nucleus. As shown in Table 2, eight sequences that matched with ERβ isoform 3 were found in the mitochondrial lysate.

Fig. 5.

MALDI-TOF mass spectrum obtained from the M_r of 50,000–65,000 SDS/PAGE band of the isolated human heart mitochondria after in-gel digestion with trypsin. The major proteins identified by the database search are indicated in the spectrum, and the tryptic peptides that matched to those derived from sequence information of the ER proteins in the National Center for Biotechnology Information database are presented in Tables 1 and 2.

Table 1. MS identification of ERβ from purified mitochondrial protein.

m/z submitted	M + H+ matched	ΔMr	Peptide sequence
993.5115	993.4722	0.039	(147-154) SCQACRLR, cysteine alkylation
1165.5804	1165.6329	-0.053	(59-68) SLEHTLPVNR
1189.6725	1189.5449	0.13	(467-477) SKEGSQNPQSQ
1215.6828	1215.5428	0.14	(181-191) SADEQLHCAGK, cysteine alkylation
1260.6996	1260.6775	0.022	(339-348) LQHKEYLCVK
1442.7152	1442.6810	0.034	(179-191) QRSADEQLHCAGK
1922.0015	1921.9078	0.094	(1-18) MNYSIPSNVTNLEGGPGR, ox. methionine

Peptide fragments listed match the National Center for Biotechnology Information (NCBI) nonredundant (nr) database record numbers gi_7441774 and gi_1518263 for ERβ; M_r = 53,384.

Table 2. MS identification of ERβ3 from purified mitochondrial protein.

m/z submitted	M + H+ matched	ΔMr	Peptide sequence
993.5115	993.4722	0.039	(147-154) SCQACRLR, cysteine alkylation
1122.5952	1122.5002	0.095	(504-512) SFEACQQPR, cysteine alkylation
1165.5804	1165.6329	-0.053	(59-68) SLEHTLPVNR
1194.5684	1194.5213	0.047	(467-477) SFEACQQPRE
1215.6828	1215.5428	0.14	(181-191) SADEQLHCAGK, cysteine alkylation
1260.6996	1260.6775	0.022	(339-348) LQHKEYLCVK
1442.7152	1442.6810	0.034	(179-191) QRSADEQLHCAGK
1575.8009	1575.7888	0.012	(477-489) LFMLREASCHGVR, cysteine alkylation

Peptide fragments listed match the National Center for Biotechnology Information (NCBI) nonredundant (nr) database record number gi_3091286 for ERβ3; M_r = 57,519.

Discussion

Four major observations are reported in this study. First, ERβ is localized to the mitochondria in several cell types. Second, this mitochondrial localization is independent of the differentiation state of cells, because it is seen in differentiated primary neurons and cardiac myocytes and in the undifferentiated HT22 cells and human lens epithelial cells (unpublished observations). Third, ERβ does not translocate to the nucleus on exposure to its natural ligand, 17β-estradiol. Finally, the mitochondrial localization can be demonstrated by using immunocytochemistry, immunoblotting, and MS. In MS, a large number of ERβ specific epitopes were identified. Collectively, these results indicate that ERβ is a component of the mitochondria and not primarily a nuclear transcriptional factor.

The identification of ERβ in mitochondria by using antibodies directed against both C-terminal and N-terminal amino acids in sequences that subsequently were identified by MS of purified mitochondria strongly support the conclusion that ERβ is a mitochondrial protein. This observation may explain previous reports that ERβ is a poor transcriptional factor. Studies with mice carrying disrupted ER genes indicated that ERα mediates the major proliferative effects of estrogen. In ERα knockout mice, a significant transcription pattern change is evident, such as hypotrophy of the uterus, ovary, testis, mammal gland, and vagina (9, 10). In contrast, both male and female ERβ knockout mice show little change in their reproductive phenotype (11), suggesting that ERβ is not a major transcriptional factor in reproductive tissues.

Other evidence also suggests that ERβ is a poor transcription factor. ERs regulate gene expression in two ways. ERs activate transcription after binding to ERE in the promoter region of estrogen-response genes. An alternative pathway has been reported in which ERs appear to be able to stimulate transcription from promoters that contain AP-1, Sp1, and cAMP response elements sites. With regard to the ability to activate transcription with constructs that contain ERE, ERβ is much weaker than ERα in most cell systems tested (20–24). Further, 17β-estradiol binding to ERα but not ERβ activates AP1 (31). Similarly, at Sp1 sites, ERα activates and ERβ is nearly inactive, although both ERs physically interact with Sp1 (32). With an in vitro chromatin assembly and transcription system, ERα is a much more potent transcriptional activator than ERβ (33). Induction of cyclin D1 gene transcription by ERs plays an important role in estrogen-mediated proliferation. Different action of ERα and ERβ on cyclin D1 gene expression has been demonstrated. ERα activates cyclin D1 gene expression on binding to 17β-estradiol, whereas ERβ does not (34).

We tested the possibility that the mitochondrial localization of ERβ was the result of artifacts of our experimental procedure. The possibility that the mitochondrial preparations used were contaminated by other cellular organelles was addressed by using two stringent methods for the isolation of mitochondria and the further purification of mitochondrial proteins. This method yields protein fractions that are devoid of cytoskeleton marker proteins (actin and dynamin), the endoplasmic reticulum marker (KDEL), and the nuclear marker (histone), but contained abundant prohibitin, a mitochondrial marker. In this purified mitochondrial protein preparation, an M_r ≈60,000 ERβ antibody-interacting protein consistent with the long form of ERβ was demonstrated by immunoblotting with two purified human heart preparations, one made by us and the other, a commercially available human heart mitochondrial protein (Molecular Probes). Further, seven fragments of ERβ and eight fragments of ERβ-3 were identified by MS. Additionally, several identified fragments of ERβ contain the epitopes to which the two ERβ antibodies used are directed.

We also tested the possibility that the colocalization of ERβ antibody binding with the mitochondrial marker MitoTracker Red was an artifact of bleeding through of emission light from one fluorophore into the detection range of the other. Staining of cells with only one fluorophore did not result in detection of light when the filter was set for the other fluorophore. Thus, the observed colocalization of ERβ with a mitochondrial marker is not due to filter band-pass limitations and is seen in multiple neuronal and nonneuronal cell types.

We assessed the possibility that the epitopes of ERβ detected in mitochondria were from recently described ER-related proteins (ERRs). ERRs belong to orphan nuclear receptors. Three ERR (ERR1, ERR2, and ERR3) have been identified based on their sequence homology with the DBD of ERs (35, 36). The two ERβ antibodies we used in this study are against N-terminal and C-terminal sequences, which share low homology with ERα and ERRs. We performed a blast search of GenBank with the epitopes of the two ERβ antibodies. No hit was found in the ERR sequences. Therefore, it is unlikely that the antibodies we used in the immunocytochemistry and immunoblotting detected ERRs. A blast search of both N-terminal and C-terminal sequences identified by our MS consistently showed no match in the ERR sequences (data not shown).

Mitochondria produce most of the cell's ATP by oxidative phosphorylation and generate most of the endogenous oxygen radicals as a toxic by-product (37). In addition, mitochondria are central in the regulation of apoptosis, calcium homeostasis, and cytoplasmic redox state (38, 39). Estrogens have long been recognized as antioxidants, and recent studies have shown that estrogens are also potent neuroprotective agents (40–42). Several studies have also shown that estrogens may exert direct or indirect effects on mitochondrial function. Estradiol can protect against ATP depletion, mitochondrial membrane potential decline, and the generation of reactive oxygen species induced by 3-nitroproprionic acid (17). 17β-Estradiol can stabilize mitochondrial function against actions of mutant presenilin 1 (43) and modulate mitochondrial calcium influx (18). This study raises the possibility that mitochondrial ERβ may mediate some or all of these actions of estrogen on mitochondrial function.

Mitochondrial genes are potential sites of primary action of steroid hormones, and estrogens could directly modulate mitochondrial gene expression (44). The 16-kb mitochondrial genome encodes only 13 of the >100 proteins involved in oxidative phosphorylation, and little is known about the regulation of mitochondrial gene expression (45). The ERE is characterized by a 15-nt motif that consists of two hexads in a palindromic configuration that is separated by 3 nt (1). The mitochondrial genome contains sequences similar to ERE, which are represented mostly as half-palindromes (44), raising the possibility that ERβ could modulate mitochondrial gene expression through binding to the similar ERE sequences in the mitochondrial genome.

In this study, evidence from immunocytochemistry, immunoblot, and proteometry suggests that ERβ is a mitochondrial component. However, other yet unidentified cytosolic components were also stained with the antibodies used in differentiated primary neurons, but not in differentiated primary cardiac myocytes or undifferentiated hippocampal HT-22 cells. Inasmuch as this study was not designed to determine the effects of the state of differentiation of an individual cell type on the distribution of ERβ, subsequent studies are needed to address this issue.

In summary, our studies indicate that, in a variety of cell types, ERβ is a mitochondrial protein rather than a nuclear receptor. Localization of ERβ to mitochondria suggests that ERβ could mediate estrogen's effects at this organelle, including its ability to modulate calcium influx, ATP production, apoptosis, and free radical species generation.