Estrogen-related Receptor β Reduces the Subnuclear Mobility of Estrogen Receptor α and Suppresses Estrogen-dependent Cellular Function*

Background: The role of orphan nuclear receptor ERRβ in estrogen-related pathophysiology is poorly understood. Results: ERRβ repressed the transactivity of ERα and proliferation of MCF-7 breast carcinoma cells through reduction of the intranuclear mobility of ERα. Conclusion: ERRβ affects ERα dynamics and function. Significance: Regulation of ERRβ will provide a potential therapeutic approach for estrogen-related cancer. Estrogen-related receptor (ERR) is a member of the nuclear receptor superfamily that has strong homology with estrogen receptor (ER) α. ERR has three subtypes (α, β, and γ) expressed in estrogen-sensitive organs, including ovary, breast, and brain. No endogenous ligands of ERRs have been identified, but these receptors share a common DNA element with ERα and control estrogen-mediated gene transcription. Recent evidence suggests a role of ERRs in estrogen-related pathophysiology, but the detailed mechanisms of ERR functions in estrogen-related tissues are unclear. Using live-cell imaging with fluorescent protein labeling, we found that only ERRβ among the ERRs exhibits a punctate intranuclear pattern overlapping with ERα following 17β-estradiol (E2)-stimulation. Fluorescence recovery after photobleaching showed significant reduction of the mobility of ligand-activated ERα with co-expression of ERRβ. Fluorescence resonance energy transfer revealed that ERRβ directly interacts with ERα. The N-terminal domain of ERRβ was identified as the region that interacts with ERα. We also found a correlation between punctate cluster formation of ERα and interaction between the receptors. Expression of ERRβ significantly repressed ERα-mediated transactivity, whereas that of other ERR subtypes had no effect on the transactivity of ERα. Consistent with this finding, E2-stimulated proliferation of MCF-7 breast carcinoma cells and bcl-2 expression was significantly inhibited by expression of ERRβ. These results provide strong evidence for a suppressive effect of ERRβ on estrogen signaling through reduction of the intranuclear mobility of ERα. The findings further suggest a unique inhibitory role for ERRβ in estrogen-dependent cellular function such as cancer cell proliferation.

repression of ovarian cancer (24). Based on these findings, ERR␣ and -␥ are potential diagnostic biomarkers for unfavorable and favorable prognoses, respectively, in estrogen-related cancers (25)(26)(27). A clinical study showed that ERR␤ is present in normal breast adipose fibroblasts but is absent in cancerassociated fibroblasts from patients with ER␣-positive breast cancer (28). These reports indicate the importance of ERRs in estrogen-related tumorigenesis, but the role and molecular mechanism of ERR␤ in estrogen signaling are poorly understood.
Imaging of NRs using green fluorescent protein (GFP)-labeling has revealed a close relationship of nuclear events such as protein degradation, gene transcription, and RNA processing with subcellular dynamics (29 -32). ER␣ localizes primarily in the nucleus, and its distribution pattern changes from diffuse to punctate in response to estrogen through attachment with nuclear meshwork components, which are referred to as the nuclear matrix (33)(34)(35).
In this study, we examined the relationship between the subcellular dynamics of ERRs in association with estrogenic stimuli and estrogen-dependent transcriptional control when co-expressed with ER␣. We found a unique contribution of ERR␤ to estrogen signaling, because only ERR␤ among the ERRs labeled with fluorescent proteins showed punctate partitioning in the nucleus in response to 17␤-estradiol (E 2 ) stimulation when coexpressed with ER␣. The intranuclear mobility of ligand-activated ER␣ was reduced by direct interaction with ERR␤. Consistent with this finding, ERR␤ repressed the transactivity of ER␣ in ERE-driven transcription and reduced proliferation of ER␣-positive breast cancer cells. These findings reveal a co-regulatory pathway of estrogen signaling by classical ER␣ and ERR␤, and they suggest a novel hormone-response mechanism of ligand-dependent transregulation mediated by orphan NRs.

EXPERIMENTAL PROCEDURES
Plasmid Construction-Total RNA was extracted from kidney and retina of male Wistar rats using Sepasol RNA I (Nacalai Tesque, Kyoto, Japan) and reverse-transcribed with ReverTra Ace-␣ (Toyobo, Osaka, Japan). Synthesized cDNA encoding ERR genes was amplified by PCR with the Expand High Fidelity PLUS PCR system (Roche Diagnostics). The PCR products were electrophoresed on a 1.5% agarose gel, purified using the QIAquick gel extraction kit (Qiagen, Valencia, CA), and inserted into a pGEM-T easy vector (Promega, Madison, WI). The inserts were cut with EcoRI and subcloned into a pcDNA3.1 vector (Invitrogen), a pEYFP-C1 vector (Clontech), or a pECFP-C1 vector (Clontech) at the EcoRI site. To obtain deletion mutants of pEYFP-ERR␤, the NTD, DBD, or LBD region was removed by inverse PCR using a KOD PLUS mutagenesis kit (Toyobo). The deleted regions of ERR␤ correspond to amino acids 1-92, 115-164, and 252-421, respectively. Inserts of each ERR␤ deletion mutant were subcloned into a pEYFP-C1 vector (Clontech) at the EcoRI site. To create deletion mutants of pECFP-ER␣, the insert of each deletion mutant of pEYFP-ER␣ was subcloned into a pECFP-C1 vector (Clontech) at the EcoRI and XhoI sites. The resulting inserts were confirmed by digestion with restriction enzymes and DNA sequencing. The details of pEC/YFP-ER␣, N-terminal (dN81, dN140, and dN246) and C-terminal (dC341 and dC430) deletion mutants of pEYFP-ER␣, the estrogen-response element-containing luciferase-reporter construct (ERE-Luc), and the ␤-actin promoter-driven ␤-galactosidase expression construct (pAct-␤gal) have been described previously (36 -38).
Transfection-All plasmids were transiently transfected using Lipofectamine and Plus Reagent (Invitrogen). In live-cell imaging and transcription assays using COS-1 cells, the medium was replaced by serum-free Opti-MEM (Invitrogen) containing 1% pen/strep solution to exclude the possibility of any hormonal effects after the transfection and recovery culture. In the cell proliferation assay, the medium was replaced by phenol red-free RPMI 1640 medium supplemented with 10% dextran-coated charcoal-stripped FBS and 1% pen/strep solution after the transfection and recovery culture. For ligand stimulation, cells were treated with 10 Ϫ7 M E 2 (Sigma) at 37°C. The same amount of dimethyl sulfoxide (DMSO) (Nacalai Tesque) was used as control.
Time-lapse Image Acquisition-The day before transfection, 3 ϫ 10 4 COS-1 cells were transferred to a 35-mm glass bottom dish (Matsunami, Osaka, Japan) pre-coated with poly-L-lysine (Sigma). The cells were transfected with 125 ng of full-length or deletion mutants of pEYFP-ERRs and/or 875 ng of full-length or deletion mutants of pECFP-ER␣, as described above. Image acquisition was performed using confocal laser scanning microscopy (Carl Zeiss, Jena, Germany) using an immersion 63ϫ objective lens. The cells were maintained in the on-stage incubation chamber at 37°C under 5% CO 2 /air. Live-cell images were acquired before and 40 min after treatment with 10 Ϫ7 M E 2 or the same amount of vehicle (DMSO). YFP fluorescence was detected using a filter set at 514 nm excitation and 530 -600 nm emission with an HFT 458/514 dichroic mirror. CFP fluorescence was observed using a filter set at 458 nm excitation and 475-525 nm emission with an HFT 458/514 dichroic mirror. Images were analyzed using software in the confocal microscopy system (Carl Zeiss).
FRAP Analysis-Fluorescence recovery after photobleaching (FRAP) analysis was performed using confocal laser microscopy (LSM510 META). COS-1 cells seeded at 3 ϫ 10 4 per 35-mm glass bottom dish pre-coated with poly-L-lysine were transfected with 125 ng of pEYFP-ERR␤ and/or 875 ng of pECFP-ER␣, as described above. After transfection, cells were maintained in serum-free Opti-MEM for 15-24 h and then treated with 10 Ϫ7 M E 2 , ER␣-specific agonist propyl pyrazole triol (PPT, 10 Ϫ9 M) (Tocris Bioscience, Ellisville, MO), 10 Ϫ8 M OHT, or vehicle for 1 h. Fluorescent images (25% of maximum laser power, 512 ϫ 512 pixels, zoom factor 5, scan speed 9) of a single Z section using an immersion 63ϫ objective lens were then taken at time intervals of 1 s after photobleaching at a wavelength of 514 nm for YFP or 458 nm for CFP at maximum laser power for 200 iterations. The time for the first scan after photobleaching was set to 0 s. A photobleach zone (ROI) in randomly selected transfected cells was defined as a circle of 50 pixels (ϳ2.6 m) in diameter. For quantitative analysis, fluorescence intensities of photobleached regions of 25 pixels in diameter were measured at each time point using LSM 510 software. The mean half-recovery time (t1 ⁄ 2 ), the time for fluorescence to recover by 50% due to diffusion, was determined from a fluorescence recovery curve generated from 30 -36 individual cells.
The emission fingerprinting method was used to detect emission spectra. Spectral signatures were captured by stack acquisition with excitation at 458 nm and detection at 10-nm intervals from 456.5 to 552.8 nm using an HFT 458/543 dichroic mirror. The ROI (a circle of 20 pixels) was randomly selected from the transfected cells to obtain emission spectral patterns.
For acceptor photobleaching FRET analysis, the acceptor (YFP) of each transfected cell was photobleached using a 514-nm laser at maximum power for 250 iterations. The photobleached zone was defined as a circle of 30 pixels. Upon image acquisition, all parameters (laser power, detector gain, amplifier offset, and pinhole size) were fixed before and after photobleaching. The emission spectral patterns were obtained, and changes in donor (CFP) fluorescence intensity at 473 nm were analyzed. To visualize the changes of fluorescence intensity before and after photobleaching, the acquired images of the donor (473 nm, CFP) and acceptor (527 nm, YFP) were pseudocolored, with red and blue ranges indicating high and low intensity, respectively.
Double Immunofluorescent Labeling-MCF-7 cells (5 ϫ 10 4 ) were plated on 35-mm glass bottom dishes precoated with poly-L-lysine and cultured in phenol red-free RPMI 1640 medium supplemented with 10% dextran-coated charcoalstripped FBS and 1% pen/strep solution for 24 h. Then the cells were treated with 10 Ϫ7 M E 2 for 30 min at 37°C and fixed with 4% paraformaldehyde in 0.1 M phosphate buffer at room temperature. After fixation, the cells were delipidated in ethanol for 30 min at Ϫ20°C and incubated together with rabbit polyclonal antibody against ER␣ (Millipore) and mouse monoclonal antibody against ERR␤ diluted 1:4,000 in PBS containing 1% normal goat serum (Nichirei Bioscience, Tokyo, Japan) overnight at 4°C. Subsequently, the cells were incubated with Alexa Fluor 488-labeled anti-mouse IgG antibody (Invitrogen) and Alexa Fluor 546-labeled anti-rabbit IgG (Invitrogen) diluted 1:1,000 in PBS for 3 h at room temperature. After embedding in Vectashield (Vector Laboratories), images were collected with LSM510 META confocal laser scanning microscopy using an immersion 63ϫ objective lens and analyzed with software provided with the microscopy system.
Transcription Assay-COS-1 cells were plated at a density of 2.5 ϫ 10 4 per well on a 24-well plate (Corning Glass). The next day, cells were co-transfected with ERE-Luc (250 ng), pcDNA3.1-ER␣ (2.5 ng), and various amounts of pcDNA3.1-ERR (0 -5 ng) vectors, as described above. pAct-␤gal (100 ng) was used as internal control for the transfection efficiency. The total amount of vectors in each well was adjusted to 355 ng with pcDNA3.1 empty vector. After transfection, cells were incubated in Opti-MEM with or without 10 Ϫ7 M E 2 for 24 h and lysed. The same amount of DMSO was used as vehicle. The luciferase and ␤-galactosidase activities of cell lysates were quantified using a Pica Gene luciferase assay kit (Toyo Ink, Tokyo, Japan) and a luminescent ␤-galactosidase detection kit II (Clontech), respectively, on the MicroLumat PLUS microplate luminometer LB96V (Berthold, Bad Wildbad, Germany).
Real Time RT-PCR-MCF-7 cells (5 ϫ 10 4 /well) were seeded on 12-well plates (Corning Glass) and transfected with 250 ng of pcDNA3.1-ERR␤ or pcDNA3.1 empty vector, as described above. The next day, cells were treated with 10 Ϫ7 M E 2 or vehicle for 2 h. Total RNA was isolated using an RNeasy micro kit (Qiagen), and 1 g of the total RNA was reverse-transcribed, as described above. Quantitative PCR was performed using a LightCycler 480 II real time PCR instrument (Roche Diagnostics). The reaction was carried out in duplicate in 96-well plates with LightCycler 480 Probes Master (Roche Diagnostics) and Universal Probe Library Probes (Roche Diagnostics), using the following conditions: preincubation for 10 min at 95°C, followed by 45 cycles for 10 s at 95°C and 25 s at 60°C. The combinations of probes and primers were designed using online software (Probe Finder version 2.50 for human, Roche Diagnostics) as follows: bcl-2 (probe 75), right primer 5Ј-AGT ACC TGA ACC GGC ACC T-3Ј and left primer 5Ј-GCC GTA CAG TTC CAC AAA GG-3Ј; c-myc (probe 66), left primer 5Ј-GCT GCT TAG ACG CTG GAT TT-3Ј and right primer 5Ј-TAA CGT TGA GGG GCA TCG-3Ј; gapdh (probe 60), left primer 5Ј-AGC CAC ATC GCT CAG ACA C-3Ј and right primer 5Ј-GCC CAA TAC GAC CAA ATC C-3Ј. Relative gene expression levels were calculated using the comparative ⌬C p method and normalized to gapdh expression using software provided with the LightCycler 480 II instrument (Roche Diagnostics).
Statistical Analysis-All values were expressed as means Ϯ S.E. Data were analyzed by unpaired t test or by one-way analysis of variance and Bonferroni/Dunn post hoc tests. All analyses were performed with StatView version 5.0 (SAS Institute Inc., Cary, NC). The results were considered significant if the p value was Ͻ 0.05.
All the fusion proteins were mainly distributed in the nucleus (Fig. 1B). Diffusely distributed CFP-ER␣ redistributed into discrete clusters after E 2 treatment for 40 min, as shown previously (37), although none of the YFP-ERRs showed cytological changes in nuclear distribution after E 2 treatment (Fig. 1B). This result is consistent with the finding that ERRs do not bind to endogenous steroids (4). To examine the effect of ER␣ coexpression on the distribution of ERRs, co-transfection of pECFP-ER␣ and each pEYFP-ERR subtype was performed, and expression patterns were observed after E 2 stimulation. Interestingly, only YFP-ERR␤ showed a dot-like distribution pattern that overlapped with that of CFP-ER␣ (Fig. 1D). YFP-ERR␣ and -␥ were unchanged, and CFP-ER␣ showed a clear punctate expression pattern after E 2 stimulation (Fig. 1, C and E).
Endogenous ER␣ and ERR␤ Overlap with Each Other in the Nucleus of MCF-7 Cells-Because subnuclear overlap of discrete clusters of ER␣ and ERR␤ was observed in transfected cells, we next examined whether this phenomenon occurs for endogenous ER␣ and ERR␤. MCF-7 breast carcinoma cells were treated with E 2 for 30 min and stained for ER␣ and ERR␤ by immunofluorescent labeling. Fluorescent images were acquired, and punctate patterns of overlapping endogenous ER␣ and ERR␤ were observed ( Fig. 2A). Using confocal laser scanning microscopy, curves approximating the fluorescence intensity along the line crossing the nuclear area were plotted (Fig. 2B). The intensive amplitudes for ER␣ and ERR␤ reflect their discrete distributions. The peak fluorescence intensities of ER␣ (ϳ24%) and ERR␤ (ϳ29%) also overlapped with each other, indicating co-localization of the two receptors within the punctate clusters (Fig. 2B, arrowheads). coIP analysis also confirmed a protein-protein interaction of endogenous ER␣ and ERR␤ in MCF-7 cells treated with E 2 for 1 h (Fig. 2C).
ERR␤ Reduces the Intranuclear Mobility of ER␣ Following E 2 Stimulation-Several nuclear receptors, including ER␣, show ligand-dependent reduced intranuclear mobility (34,35,38,42). Because YFP-ERR␤ showed discrete clusters only when co-expressed with CFP-ER␣, we examined whether both receptors had decreased intranuclear mobility using FRAP analyses, with a view to examine an interaction between the two receptors.
In the absence of E 2 , bleach zones of CFP-ER␣ were not detected regardless of the presence of YFP-ERR␤ because of the extreme mobility of unliganded CFP-ER␣ (Fig. 3, A and C), in agreement with a previous report (42). In the presence of E 2 , YFP-ERR␤ significantly reduced the intranuclear mobility of CFP-ER␣ (Fig. 3, B and D-F). The mobility of YFP-ERR␤ was also decreased by ligand-activated CFP-ER␣ (Fig. 4, D-F). In the presence of the ER␣-selective agonist PPT, YFP-ERR␤ significantly reduced the CFP-ER␣ mobility (Fig. 3, E and F). Con-sistent with this finding, PPT-stimulated CFP-ER␣ significantly reduced the mobility of YFP-ERR␤, suggesting an interaction between ER␣ and ERR␤ (Fig. 4, E and F). This interaction appeared to be agonist-dependent because the mobility of YFP-ERR␤ was not reduced in the presence of CFP-ER␣ under ligand-free conditions, and it was reduced only in the presence of CFP-ER␣ stimulated by E 2 or PPT (Fig. 4, C-F). Interestingly, in the presence of anti-estrogen OHT, YFP-ERR␤ did not affect the mobility of CFP-ER␣ (Fig. 3F). Therefore, the mobility reduction of YFP-ERR␤ seen in the presence of CFP-ER␣ and OHT (Fig. 4F) is probably due to binding of YFP-ERR␤ with OHT.
A protein-protein interaction between E 2 -activated ER␣ and ERR␤ was also shown by coIP using a specific antibody against ER␣ or ERR␤ following co-transfection of pcDNA3.1-ER␣ and pcDNA3.1-ERR␤ expression vectors in COS-1 cells (Fig. 4G).
ERR␤ Directly Interacts with ER␣ in Live Cells-We further analyzed whether ER␣ and ERR␤ interact directly using FRET microscopy. COS-1 cells were co-transfected with pECFP-C1 and pEYFP-C1, with pECFP-ERR␤ and pEYFP-C1, or with pECFP-ERR␤ and pEYFP-ER␣. After E 2 treatment for 1 h, images were captured using confocal laser scanning microscopy. In the emission spectral analysis, ROIs of cells co-expressing CFP-ERR␤ and YFP-ER␣ produced FRET signals that exhibited a prominent peak at 527 nm, whereas ROIs of cells co-expressing CFP and YFP or co-expressing CFP-ERR␤ and YFP did not show FRET (Fig. 5A).
Because FRET was clearly observed, we performed acceptor photobleaching FRET microscopy. This method is based on the increase of donor fluorescence intensity when the acceptor fluorophore is abolished in a system showing FRET. The emission spectra in Fig. 5B indicate that the donor intensity at 473 nm (CFP-ERR␤) was markedly increased after photobleaching of acceptor (YFP-ER␣) at 514 nm. Pseudocolored images at 473 and 514 nm were acquired before and after photobleaching (Fig. 5C). After photobleaching, the intensity of YFP-ER␣ decreased (Fig. 5C, lower), whereas the intensity of CFP-ERR␤ increased (Fig. 5C, upper) in the bleach zone (Fig. 5C, arrowheads). In contrast, the intensities of CFP-ERR␤ and YFP-ER␣ in non-bleach zones did not change from before to after photobleaching (Fig. 5C, arrows). Quantification of acceptor photobleaching showed significant enhancement of fluorescence intensity (at 473 nm) of ROIs from cells co-expressing CFP-ERR␤ and YFP-ER␣ (Fig. 5D). In contrast, the intensity of ROIs from cells co-expressing CFP and YFP or co-expressing CFP-ERR␤ and YFP did not change after photobleaching, indicating that CFP and YFP, and ERR␤ and YFP did not interact with each other and did not produce FRET (Fig. 5D). From these findings, we conclude that ER␣ and ERR␤ interact directly with each other in live cells.  (37) demonstrated that the ligand-induced clustering activity of ER␣ depends on the latter part of AF-1 within the NTD, the DBD, or AF-2 within the LBD. COS-1 cells were co-transfected with the deletion mutants of pECFP-ER␣ and full-length pEYFP-ERR␤, and live-cell imaging was performed. Protein expression of the deletion mutants was confirmed by Western blotting from total lysates of COS-1 cells transfected with each expression plasmid (Fig. 6, B and C).
When co-expressed with ERR␤, dN81 and dN140 exhibited dot-like expression patterns following E 2 stimulation, whereas dN246, dC341, and dC430 did not do so (Fig. 6, E-I). dN140 itself does not have clustering activity in response to E 2 (Fig.  6A). Therefore, these results show that ERR␤ can recover the clustering activity of the latter part of AF-1 of ER␣. Similarly, coIP showed that dN81 and dN140 interact with ERR␤, whereas dN246, dC341, and dC430 did not show this interaction, suggesting that cluster formation overlapping with each other is critical for the interaction between the two receptors (Fig. 6J). Because dN246, dC341, and dC430 lack the activity to bind with DNA or ligand, or to dimerize, these activities may be required for the interaction with ERR␤.
ERR␤ Interacts with ER␣ through the N-terminal Domain-To identify which domain of ERR␤ is required for interaction with ER␣, deletion mutants of ERR␤ tagged with YFP were generated. The schematic structures of these deletion mutants are shown in Fig. 7A. Protein expression of the mutants was confirmed by Western blotting from total lysates of COS-1 cells transfected with each expression plasmid (Fig. 7B).
pECFP-ER␣ and expression plasmids of the deletion mutants were co-transfected into COS-1 cells, and live-cell imaging was performed. After E 2 stimulation, dNTD did not form discrete clusters that overlapped with CFP-ER␣, although other deletion mutants of dDBD and dLBD exhibited dot-like expression patterns that overlapped with CFP-ER␣ (Fig. 7, C-F). In coIP, binding with ER␣ was undetected for dNTD, whereas dDBD and dLBD bound to ER␣ (Fig. 7G). These results show that the NTD of ERR␤ interacts with clustered ER␣.
ERR␤ Represses the Transcriptional Activity of ER␣-Because the intranuclear mobility of ER␣ was decreased by direct interaction with ERR␤ (Figs. 3 and 5), we examined whether the transcriptional activity of ER␣ was affected by ERR␤ using an ERE-driven luciferase reporter assay. COS-1 cells were cotransfected with pcDNA3.1-ER␣ and various amounts of pcDNA3.1-ERR␤ and were incubated in the presence or absence of E 2 for 20 h. Then normalized luciferase activity was determined. As shown in Fig. 8, A-C, ERR␤ reduced the ER␣mediated transcriptional activity of ERE in a dose-dependent manner, whereas ERR␣ and -␥ appeared to have no effect on the transactivity of ER␣. These results suggest that the punctate expression pattern of ERR␤ and the correlated reduced mobility of ER␣ are associated with transcriptional suppression (Figs. 1-3).
ERR␤ Inhibits Estrogen-dependent Cellular Proliferation of MCF-7 Breast Cancer Cells-We further speculated that ERR␤ has the potential to reduce estrogen-dependent cellular func-

FIGURE 2. Endogenous ER␣ and ERR␤ exhibit discrete overlapping clusters and interact in MCF-7 cells. A, MCF-7 cells were treated with E2 for 30 min and subjected to immunofluorescent staining with anti-ER␣ (green) and anti-ERR␤ (red). Yellow dots represent overlap of ER␣ and ERR␤ in the nucleus (arrowheads).
All images were captured using confocal laser scanning microscopy. B, fluorescence intensities of ER␣ and ERR␤ along the white arrow in the left panel are plotted with green (ER␣) and red (ERR␤) curves, respectively. Yellow arrowheads are the positions where the fluorescence peaks of ER␣ and ERR␤ overlap. C, protein-protein interaction between endogenous ER␣ and ERR␤. MCF-7 cells were treated with E2 for 1 h, and total cell lysates were subjected to coIP. Lysates without immunoprecipitation (Cell lysate) and lysates immunoprecipitated without antibody (Ab(-)) were loaded as controls. coIP, co-immunoprecipitation; E2, 17␤-estradiol; WB, Western blot; Ab, antibody; bar, 10 m.
tion. To examine this issue, the effect of ERR␤ on proliferation of MCF-7 cells, an estrogen-sensitive breast carcinoma cell line, was analyzed. Expression of ER␣ was confirmed by Western blotting from a total lysate of MCF-7 cells. The total cell lysate of COS-1 cells transfected with pcDNA3.1-ER␣ was used as a positive control (Fig. 9A). MCF-7 cells transfected with pcDNA3.1-ERR␤ were incubated with or without E 2 . Cell proliferation was determined using a WST-8 assay at 3 and 5 days after transfection. Proliferation of MCF-7 cells stimulated with E 2 was significantly reduced by expression of ERR␤ at 3 and 5 days after transfection (Fig. 9B). ERR␤ had no effect on proliferation in E 2 -untreated groups (Fig. 9B). These results show that ERR␤-mediated reduction of ER␣ mobility results in inhibition of estrogen-dependent cellular function. To examine the effect of ERR␤ on estrogen-insensitive activity, TamR cells, which exhibit estrogen-independent proliferation, were established from MCF-7 cells (data not shown) (39,43). Expression of ERR␤ did not affect the estrogen-independent proliferation of TamR cells at 3 and 5 days after transfection (Fig. 9C), suggesting that suppression of estrogen signaling by interaction of ER␣ and ERR␤ is ineffective in estrogen-irresponsive cells.
ERR␤ Selectively Suppresses Estrogen-dependent Gene Expression-Several genes, including bcl-2 and c-myc, are regulated in an estrogen-dependent manner in breast cancer cells (44) and exert anti-apoptotic and oncogenic effects, respectively. Because a growth inhibitory effect of ERR␤ was observed in E 2 -stimulated MCF-7 cells (Fig. 9B), we examined whether estrogen-dependent gene expression is affected by expression of ERR␤. After treatment with E 2 or vehicle for 2 h, mRNA expression for bcl-2 and c-myc in MCF-7 cells transiently transfected with pcDNA3.1-ERR␤ or pcDNA3.1 empty vector was analyzed by real time RT-PCR. The E 2 -activated expression of bcl-2 mRNA was significantly repressed by ERR␤ (Fig. 9D), whereas the expression level of c-myc mRNA was not affected by expression of ERR␤ in E 2 -treated and untreated cells (Fig. 9E).

DISCUSSION
Ligand-activated ER␣ recruits other cofactors to organize transcription machinery associated with the nuclear matrix (33,34,45). ER␣ is extremely mobile in a ligand-free condition, whereas the mobility is reduced following ligand binding due to attachment to the nuclear matrix (35,46). This step is crucial for ligand-dependent biological processes common among steroid hormone receptors (47).
The present FRAP and coIP analyses demonstrated that interaction with ERR␤ further slows the mobility of agoniststimulated ER␣, and the transcription assay revealed ERR␤induced transcriptional repression of ER␣ in the presence of E 2 . Transcriptional repression of ER␣ concomitant with further mobility reduction is also observed in the co-expression of scaf- fold attachment factor B1/2, which mediates binding with the nuclear matrix (38). This report and our present findings support the idea that attachment of ER␣ with protein components results in transcriptional repression. However, the molecular mechanism underlying the correlation between transactivity and nuclear mobility of ER␣ is unclear. Alteration of subnuclear mobility of ER␣ affects its contact frequency with response elements and may reflect exchange of transcriptional co-regulators (48 -52). Therefore, the mechanism of action of ERR␤ may involve prevention of access of ER␣ to the ERE or, alternatively, co-regulator exchange from co-activators to co-repressors by binding with ERR␤. Further studies are required to identify how ERR␤ represses the transactivity of ER␣ by reducing its intranuclear mobility.
Interestingly, in the presence of the anti-estrogen OHT, the mobility of ER␣ was not affected by co-expression with ERR␤, whereas OHT significantly reduced the mobility of ERR␤ when co-expressed with ER␣. These results suggest that, in contrast to E 2 or PPT, OHT-bound ER␣ did not interact with ERR␤, and ERR␤ mobility was reduced by binding with OHT independent of ER␣. These data support the idea that recruited cofactors of ER␣ differ if the bound ligand is an agonist or antagonist (53,54).
The discrete cluster formation activity of ER␣ depends on binding with the nuclear matrix (46). The dN81 ER␣ mutant retains clustering activity, but others, including dN140, dN246, dC341, and dC430, do not show this activity (37). These results confirmed that dN81 and dN140 possess the ability to form discrete clusters overlapping with ERR␤, whereas dN246, dC341, and dC430 do not have this ability (Fig. 4). Therefore, the redistribution activity of ER␣ associated with ERR␤ overlapping is retained within amino acids 140 -246, which include the latter part of AF-1, or amino acids 341-600, which include LBD and AF-2. Interestingly, co-localization with intact ER␣ (37) or even ERR␤ (Fig. 4F) recovered the clustering activity of dN140 following E2 stimulation. These findings suggest that the presence of AF-1 of ER␣ or ERR␤ is sufficient to form a discrete cluster, although AF-1 in the two receptors has little similarity (55). This also indicates that ERR␤ is not just passively recruited by ER␣ but helps with the clustering activity of ER␣, despite ERR␤ not responding to E2 (7). As expected, ER␣ deletion mutants lacking AF-2, including dC341 and dC430, did not form clusters in the presence or absence of ERR␤, because they fail to bind with E 2 .
ER␣ and -␤ activate transcription following ligand binding as homodimers and a heterodimer (56). Dimerization is mediated by the dimer interface within the LBD (57). In contrast, ERRs stimulate transcription of ERE-containing genes as both monomers and homo/heterodimers (58 -60). Because the dimer interface of ERR␤ is located within the LBD, we first examined whether the binding domain of ERR␤ with ER␣ is located in the LBD. However, contrary to our expectation, the potential binding site was determined to be within the NTD by live-cell imaging and coIP using deletion mutants of ERR␤. In contrast, the interaction has been predicted to depend on the D-domain and LBD by computer modeling (61). The reason for the different findings may be the contrast between computer-assisted threedimensional modeling and live cell-imaging/coIP using deletion mutants of ERR␤. In vivo experiments are affected by many conditions, such as cell context, cofactors of receptors, and other intracellular components, in comparison with an in silico study. FRET microscopy, including emission-spectral analysis and acceptor photobleaching, showed a direct interaction between ER␣ and ERR␤ in live cells. Therefore, we suggest that the NTD of ERR␤ provides a direct binding interface for ER␣.
The NTD of NRs, including ERRs, possesses an AF-1 region that is unfolded and functions independently of ligand stimulation. Several reports suggest that this unfolded AF-1 structure is essential for recruitment of transcriptional co-regulators (62,63). Thus, we propose a model in which AF-1 within the NTD of ERR␤ recognizes ligand-activated ER␣ as a transcriptional cofactor, and their interaction causes transcriptional repression of ER␣. Many transcriptional co-regulators such as SRC-1 and RIP140 possess the LXXLL motif for interaction with NRs, including ER␣ (64). ERR␤ contains this signature motif within the LBD, rather than the NTD. However, the NTD of ERR␤ has at least four signature (I/L)XX(L/H/I) motifs at positions 27-30, 30 -33, 49 -52, and 95-98 (NCBI Protein Database under NCBI accession number NP_001008516.2) that are responsible for co-repressor protein interactions with NRs, including ER␣ (53,65). This sequence is referred to as a co-repressor-nuclear receptor box and may function as a binding surface of ER␣ that causes transrepression, as confirmed for co-repressors. ER␣ and ER␤ have been suggested to bind as a heterodimer in response to estrogen (37,57,66). ERR␤ can also form a heterodimer with ERR␣ that leads to transactivation of ERE and can interact with ER␤, with the resulting modulation of breast cancer-related genes (61,67). Given that the dimer interface of ERR␤ is located within its LBD, while ERR␤ interacts with ER␣ through the NTD, a more complicated complex (trimer or tetramer) of ERs and ERRs centered on ERR␤ may be possible, suggesting multiple signaling pathways of estrogen action.
ERR␤ includes three splicing variants as follows: hERR␤2, hERR␤2-⌬10, and short-form hERR␤ (55). Short-form hERR␤ is a widely expressed human homolog of mouse and rat ERR␤, whereas expression of hERR␤2 and hERR␤2-⌬10 is restricted to testis and kidney (55). Short-form ERR␤ cloned from rat tissue was used in this study, which makes it highly probable that our findings can be extrapolated to human pathophysiol- ogy. Indeed, short-form ERR␤ reduces proliferation of Ishikawa cells, an ER␣ positive-endometrial adenocarcinoma cell line (68), supporting our functional studies on ERR␤, as discussed below.
ERR␤-specific transcriptional repression of ERE-driven luciferase activity mediated by ER␣ further prompted us to analyze the estrogen-dependent cellular functions of ERR␤. Our data indicated that ectopic expression of ERR␤ significantly reduced E 2 -activated proliferation of estrogen-sensitive MCF-7 cells. In contrast, estrogen-independent proliferation of TamR cells was not affected by expression of ERR␤. In addition, E 2 -activated bcl-2 expression was repressed by expression of ERR␤ in MCF-7 cells. Taken together, cellular processes mediated through endogenous ER␣ can be inhibited by ERR␤ following estrogenic stimulation, and ERR␤ is functionally integrated into estrogen signaling that controls cellular properties. Because BCL-2 protein has an anti-apoptotic function (44), at least one of the tumor-suppressive effects of ERR␤ occurs through a pro-apoptotic action. However, E 2 -stimulated c-myc expression was not affected by expression of ERR␤. Therefore,  . The total amount of expression vector in each well was adjusted to 5 ng with pcDNA3.1 empty vector. The cells were incubated with E 2 (shaded bars) or without E 2 (white bars) for 24 h, and luciferase activity was determined by normalization using a ␤-galactosidase activity. Note that ERR␤ significantly repressed the transcriptional activity of ER␣-mediated transactivity of ERE in a dose-dependent manner. Data are shown as mean Ϯ S.E. (n ϭ 6). ns, not significantly different; **, p Ͻ 0.01; #, p Ͻ 0.01 versus E 2 (Ϫ) control; ERE-Luc, estrogen-response element-driven luciferase reporter; pAct-␤gal, ␤-actin promoter driven ␤-galactosidase expression construct; E2, 17␤-estradiol. some target genes of ER␣ are affected, and some are unaffected by ERR␤ following estrogenic stimulation.
Several studies, including clinical findings, suggest an inverse association between the ERR␤ expression level and estrogendependent tumor progression (25,26,28). Our findings support this idea, but a variety of mechanisms has been proposed in other reports. A recent report suggests that ERR␤ expression results in a better prognosis in breast cancer patients with reduced BCA2 and an increased FST level by controlling these genes (61). In addition, ERR␤ exerts a carcinostatic effect through activation of the p21 tumor suppressor gene and cell cycle regulation, which are not linked to anti-estrogenic action (69,70). In fact, our proliferation assay in MRK-nu-1 triplenegative breast cancer cells indicated a growth inhibitory action of ERR␤ independent of estrogen signaling (data not shown). In contrast, as described above, the basal growth activity of E 2 -untreated MCF-7 cells or E 2 -independent proliferation of TamR cells was not affected by expression of ERR␤. Thus, the carcinostatic effect of ERR␤ through an estrogen-independent pathway is cell line-or condition-specific.
Overall, our findings provide new understanding of ERR␤ effects on estrogenic action in association with intranuclear mobility of ER␣. The results also show the pathological and therapeutic importance of ERR␤ in estrogen-dependent oncodevelopment and tumorigenesis. Hormone-dependent gene transcription is regulated by a complex network of interactions with other proteins in the transcriptional machinery. An orphan NR such as ERR␤ may be a potential regulator behind this exquisite control of gene transcription, underlying modulation of sensitivity to hormones and selection of target genes within each organ or cell. Thus, we propose a model of a hormone response mechanism based on subnuclear trafficking, i.e. orphan NR-mediated transregulation initiated by steroid hormones.  1-ER␣ (middle), or MCF-7 breast carcinoma cells (right) were subjected to Western blotting using a specific antibody against ER␣. B, MCF-7 cells were transfected with 0 -5 ng of pcDNA3.1-ERR␤ as indicated. The total amount of expression vector was adjusted to 5 ng with pcDNA3.1 empty vector. The cells were incubated with E 2 (right) or without E 2 (left) for 3 and 5 days after transfection. Cell proliferation activity was then examined with a WST-8 assay using CCK-8 by analyzing the optical density at wavelength of 450 nm. Data are shown as mean Ϯ S.E. (n ϭ 4). C, TamR cells were transfected with 0 -5 ng of pcDNA3.1-ERR␤, as described above, and incubated in the presence of OHT for 3 and 5 days. Cell proliferation was then examined as described above. Note that expression of ERR␤ did not affect proliferation of TamR cells. Data are shown as mean Ϯ S.E. (n ϭ 6). D and E, ERR␤ selectively suppresses estrogen-dependent gene expression. MCF-7 cells were transiently transfected with pcDNA3.1-ERR␤ (250 ng) or the same amount of pcDNA3.1 empty vector and treated with E 2 or vehicle for 2 h. Total RNA from the cells was subjected to real time RT-PCR for quantification of the mRNA expression levels of bcl-2 (D) and c-myc (E). Each value was normalized to the gapdh expression level and is shown as fold activation relative to the value of a vehicle-treated group transfected with an empty vector. Data are shown as mean Ϯ S.E. (n ϭ 6). *, p Ͻ 0.05; **, p Ͻ 0.01; ns, not significantly different; E2, 17␤-estradiol; OHT, 4-hydroxytamoxifen; TamR, tamoxifen resistant.