Extracellular Signal-regulated Kinase 8 (ERK8) Controls Estrogen-related Receptor α (ERRα) Cellular Localization and Inhibits Its Transcriptional Activity*

ERK8 (MAPK15) is a large MAP kinase already implicated in the regulation of the functions of different nuclear receptors and in cellular proliferation and transformation. Here, we identify ERRα as a novel ERK8-interacting protein. As a consequence of such interaction, ERK8 induces CRM1-dependent translocation of ERRα to the cytoplasm and inhibits its transcriptional activity. Also, we identify in ERK8 two LXXLL motifs, typical of agonist-bound nuclear receptor corepressors, as necessary features for this MAP kinase to interact with ERRα and to regulate its cellular localization and transcriptional activity. Ultimately, we demonstrate that ERK8 is able to counteract, in immortalized human mammary cells, ERRα activation induced by the EGF receptor pathway, often deregulated in breast cancer. Altogether, these results reveal a novel function for ERK8 as a bona fide ERRα corepressor, involved in control of its cellular localization by nuclear exclusion, and suggest a key role for this MAP kinase in the regulation of the biological activities of this nuclear receptor.

Mitogen-activated protein (MAP) kinases are a family of proline-directed serine/threonine kinases expressed in all eukaryotic cells, from yeast to human, and involved in key signaling pathways regulating cell proliferation, differentiation, apoptosis, and stress response (1,2). ERK8 is the last identified member of the MAP kinase family. Along with a typical MAP kinase domain, it possesses a peculiarly long, unique C-terminal domain (3). Its activity can be modulated by serum, DNA damage, and activated human oncogenes such as BCR/ABL and RET/PTC3 (3)(4)(5). Still, although ERK8 has the classical signature Thr-Glu-Tyr activation motif of signal-dependent kinases such as ERK1, ERK2, and ERK5, it is not significantly activated by many extracellular stimuli that typically impinge on MAP kinases, and no MAP kinase kinase (MAPKK) has been identified that regulates its activity (6). 2 Interestingly, ERK8 has been recently involved in the transformation of human colon cancer cells (7) and in the maintenance of genomic integrity, by inhibiting proliferating cell nuclear antigen (PCNA) degradation (8). Recent data have also shown the ability of ERK8 to reduce the activity of nuclear receptors such as androgen and glucocorticoid receptors (9,10). Moreover, ERK8 strongly enhances ubiquitin-dependent degradation of estrogen receptor ␣ (ER␣), 3 and loss of its expression has been correlated to breast cancer progression and increased ER␣ levels in mammary tumors (9). However, its targets and downstream effectors still remain almost completely unknown. Likewise, its biological functions have yet to be unraveled. ERR␣ (NR3B1) was the first "orphan" nuclear receptor to be identified, on the basis of its similarity to ER␣ (11). It is a key regulator of mitochondrial biogenesis (12), energy metabolism (13), and bone formation and maintenance (14). ERR␣ binds to DNA and regulates transcription through specific ERR response elements (ERREs), also named SF-1 response elements (SFREs), as well as through classical estrogen receptor response elements (EREs) (15,16). The natural ligand of ERs, 17␤-estradiol (E 2 ), does not bind to ERR␣ (17). However, based on the ability of this steroid receptor to compete with ER in the binding to EREs and to heterodimerize with it, ERR␣ may play a role in the response of some genes to estrogen, thus leading to a possible overlap between ERR␣ and ERs transcriptional targets (15).
No endogenous ligand of ERR␣ has been reported to date. This observation, along with structural data showing that the putative ligand-binding pocket of ERR␣ is already locked in an agonist-bound conformation, led to considering ERR␣ as a "true orphan," constitutively active, nuclear receptor (18). However, despite the lack of natural ligands, the transcriptional activity of ERR␣ can be finely modulated by coregulatory "proteic ligands" such as coactivators and corepressors (19). As a matter of fact, such coregulators assume a huge importance in the modulation of orphan nuclear receptors, due to the lack of regulation by natural ligands. Indeed, the regulation of ERR␣ activity heavily relies on peroxisome proliferator-activated * This work was supported by a "Start-up grant" from Regione Toscana and by a grant from the Italian Association for Cancer Research (to M. C.). □ S The on-line version of this article (available at http://www.jbc.org) contains supplemental Fig. S1. 1  receptor-␥ coactivator-1␣ (PGC1␣), a powerful coactivator that controls its role in cellular functions such as oxidative phosphorylation, mitochondrial biogenesis, and respiration (20,21).
Evidence of ERR␣ roles in human malignancies, especially in breast cancer, is rapidly accumulating (22)(23)(24)(25)(26). Moreover, the EGF receptor and ErbB2 signaling pathways regulate ERR␣ transcriptional activity, leading to its hyperphosphorylation and enhancing its DNA binding (27,28). Therefore the possibility to modulate ERR␣ activity is currently regarded as a valuable approach to target breast cancer as well as other aggressive human malignancies (23)(24)(25).

EXPERIMENTAL PROCEDURES
Antibodies and Western Blot Analysis-The following primary antibodies were used for Western blot experiments: HA (Covance), Lamin A/C, IB␣, ERK2, PGC1␣ (Santa Cruz Biotechnology), ␤-actin (Sigma), and ERK8 (Kinasource). HA (Santa Cruz Biotechnology) was used as the primary antibody for confocal microscopy experiments. The following primary antibodies were used for Western blot and confocal microscopy experiments: ERR␣ (Epitomics) and ERK8 (custom preparation). AU1 (Covance) was used as the primary antibody for Western blot, immunoprecipitation, and confocal microscopy experiments. Rabbit preimmune serum (Santa Cruz Biotechnology) was used as a negative control in immunoprecipitation experiments. The following secondary antibodies were used for Western blot experiments: HRP-conjugated anti-mouse IgG, HRP-conjugated anti-rabbit IgG (Santa Cruz Biotechnology), and HRP-conjugated anti-sheep IgG (Calbiochem). For Western blot analysis, proteins derived from total lysates, immunoprecipitations, or affinity precipitations were loaded on SDS-PAGE, transferred to Immobilon-P PVDF membrane (Millipore), probed with appropriate antibodies, and revealed by enhanced chemiluminescence detection (ECL Plus; GE Healthcare). Bacterially expressed proteins were loaded on SDS-PAGE gels, stained with SimplyBlue SafeStain (Invitrogen), and revealed using the Odyssey Infrared Imaging System (Li-Cor Biosciences).
Expression Vectors-For two-hybrid screening, a DNA fragment encoding for the C-terminal domain of ERK8 from amino acids 267 to 544 was cloned in the pGBKT7 vector (Clontech). For bacterial expression of the C-terminal domain of the ERR␣ protein, the pGEX-4T3-ERR␣_C-terminal plasmid was prepared by cloning a DNA fragment encoding for the murine ERR␣ C-terminal domain, specifically contained in the twohybrid clone, in the pGEX-4T3 vector. The pCEFL HA ERK8 expression vector has been previously described (4). The pCEFL EGFP ERK8 expression vector was generated by subcloning the ERK8 cDNA, obtained by restriction enzyme digestion from pCEFL HA ERK8, into the pCEFL EGFP vector. The pCEFL HA ERK8 3LA1, pCEFL HA ERK8 3LA2, and pCEFL HA ERK8 3LA1-2 expression vectors were generated with the QuikChange site-directed mutagenesis kit (Stratagene), using pCEFL HA ERK8 as a template. The pCEFL AU1 ERR␣ expression vector was generated by subcloning the ERR␣ cDNA, obtained by PCR from the pBlueScript ERR␣ vector (Clone ID 30336569) from the IMAGE Consortium, into the pCEFL AU1 vector. The pCEFL GST ERR␣ expression vector was generated by subcloning the ERR␣ cDNA, obtained by restriction enzyme digestion from pCEFL AU1 ERR␣, into the pCEFL GST vector. The pCDNA3 HA JNK expression vector has been previously described (29). The pCEFL AU1 ERR␤ and pCEFL AU1 ERR␥ expression vectors were generated by subcloning the ERR␤ and ERR␥ cDNAs, obtained by PCR from pBlueScriptR ERR␤ (Clone ID 30344716) and pENTR223.1 ERR␥ (Clone ID 100015441) from Open Biosystems, into the pCEFL AU1 vector. The ERRE_Luc firefly luciferase reporter vector is a kind gift from J. M. Vanacker (16). The pcDNA4 PGC1␣ (Addgene plasmid 10974), pBABE5Ј EGFR L858R (Addgene plasmid 11012), and pBABE5Ј EGFR D770_N771 insNPG (Addgene plasmid 11016) expression vectors were obtained from Addgene (addgene.org). The identity and integrity of all vectors was confirmed by DNA sequencing.
Cell Culture and Transfection-HEK293T and HeLa cells were maintained in Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum, 2 mM L-glutamine, and 100 units/ml of penicillin-streptomycin at 37°C in an atmosphere of 5% CO 2 /air. MCF10A cells were maintained in DMEM/Ham's F-12 medium supplemented with 5% horse serum, 2 mM L-glutamine, 100 units/ml of penicillin-streptomycin, 10 ng/ml of epidermal growth factor (EGF), 0.5 g/ml of hydrocortisone, 100 ng/ml of cholera toxin, and 10 g/ml of insulin at 37°C in an atmosphere of 5% CO 2 /air. To generate the MCF10A cell lines stably expressing EGFP and HA ERK8, MCF10A cells were transfected with the pCEFL EGFP and pCEFL HA ERK8 expression vectors, respectively, and subjected to selection with G-418 sulfate for 2 weeks.
For Western blot and immunoprecipitation experiments, 1 ϫ 10 6 cells were seeded in 6-cm plates and transfected with 1 g of the different expression vectors using Lipofectamine LTX (Invitrogen). For luciferase assays, 1 ϫ 10 5 cells were seeded in 12-well plates and transfected with 50 to 100 ng of the ERRE_ Luc firefly luciferase reporter vector and 500 ng of the different expression vectors using Polyfect (Qiagen) for HeLa cells and Lipofectamine LTX (Invitrogen) for MCF10A cells. All experiments were performed 24 h after transfection. For confocal microscopy experiments, 2.5 ϫ 10 4 cells were seeded on coverslips placed onto 12-well plates. For 293T cells, coverslips were pre-coated with polylysine (Sigma). Each sample was transfected with 200 to 500 ng of each plasmid using Lipofectamine LTX (Invitrogen).
Yeast Two-hybrid Screening-To isolate potential ERK8 interacting proteins, we used a region comprising its C-terminal 277 amino acids (amino acids 267-544) as a bait to perform a yeast two-hybrid screening. The cDNA corresponding to such a region was cloned in the pGBKT7 plasmid (Clontech), inframe with the yeast GAL4 DNA-binding domain, by PCR amplification followed by enzymatic digestion. Thanks to the presence of a MYC epitope in the resulting fusion protein (MYC-GAL4-ERK8 C-term), we confirmed the expression of such a protein in yeast. Before starting the screening we also confirmed that our bait was not able to activate, by itself, the transcription of reporter genes. In addition, we confirmed that our bait could not interact directly with the GAL4 transactivation domain contained in the plasmid employed to engineer the ERK8 as an ERR␣ Corepressor cDNA library used for the screening. In this regard, we used a Clontech mouse brain cDNA library (Matchmaker GAL4 twohybrid system 3).
For the screening, we co-transformed 500 g of the pGBKT7 ERK8_C-term bait and 1 mg of library DNA in yeast strain PJ69-4A. Protein-protein interactions were assessed by streaking transformants on selective medium lacking Leu, Trp, and His with addition of 3-10 mM 3-aminotriazole. Potential interactors were sequenced from the pACT2 plasmid by using the Matchmaker 5Ј and Matchmaker 3Ј primers (Clontech).
Bacterial Expression of GST Fusion Proteins-The BL21 Lys strain of Escherichia coli was transformed with the pGEX-4T3 vector alone or encoding for the mouse ERR␣ C-terminal fusion protein. Bacterially expressed GST and GST fusion protein were purified as previously described (30).
Immunofluorescence, Confocal Microscopy, Intensitometric Analysis of Fluorescence, and Count of ERR␣-positive Nuclei-Twenty-four hours after transfection, cells were washed with PBS, then fixed with 4% paraformaldehyde in PBS for 20 min and permeabilized and blocked with a solution of 0.075% saponin (Sigma) and 0.2% gelatin (Sigma) in PBS for 20 min. Cells were incubated with appropriate primary antibodies for 1 h, washed three times with PBS, incubated with appropriate Cy2conjugated and TRITC-conjugated secondary antibodies (Jackson ImmunoResearch Laboratories), and then washed again three times with PBS. Nuclei were stained with a 15 M solution of 4Ј,6-diamidino-2-phenylindole (DAPI) (Sigma) in PBS for 3 min. Coverslips were mounted in Fluorescence Mounting Medium (Dako). Samples were visualized on a TSC SP5 confocal microscope (Leica) adapted to an inverted LEICA DMI 6000CS microscope and equipped with an oil immersion Pla-nApo ϫ63 1.4 NA objective. Images were acquired using LAS AF acquisition software (Leica). Intensitometric analysis of fluorescence was performed using the Quantitation Module of Volocity software (PerkinElmer Life Science). For the count of ERR␣-positive nuclei, the total number of ERR␣-positive cells and the number of cells with nuclear ERR␣ staining were determined in 20 random fields; the results were expressed as percentages of the ratio between ERR␣-positive nuclei and ERR␣-positive cells. In samples co-transfected with ERR␣ and ERK8, only cells expressing both proteins were considered.
Luciferase Assays-HeLa cells were transfected with 50 ng of the ERRE_Luc firefly luciferase reporter vector and 500 ng of different expression vectors (unless otherwise indicated). MCF10A cells were transfected with 100 ng of the SFRE_Luc firefly luciferase reporter vector and 500 ng of different expression vectors. Twenty-four hours after transfection, cells were lysed in Passive Lysis Buffer (Promega) and luciferase activity in the cellular lysates was assessed on a Glomax 20/20 luminometer (Promega) using the Luciferase Assay System (Promega). Results were normalized for total protein content. All luciferase results represent the normalized average Ϯ S.D. of at least two independent transfections. All samples were read in triplicate.
Expression Analysis of ERR␣ Target Genes-Total RNA was purified using TRIzol Reagent (Invitrogen). Reverse transcrip-
Subcellular Fractionation-Cytoplasmic and nuclear fractions were obtained by the Subcellular Protein Fractionation Kit (Thermo Scientific), according to the manufacturer's instructions.

RESULTS
ERK8 Interacts with ERR␣ in Vitro and in Vivo-We carried out a yeast-based two-hybrid screening to identify novel ERK8 interacting proteins. As ERK8 is characterized by a quite peculiar, long C-terminal domain, we decided to use this domain as bait for this screening. We chose to screen a mouse brain library, based on the evidence that ERK8 is highly expressed in the nervous system (3). Among the positive clones, we found multiple clones encoding for ERR␣. One of these was confirmed to be devoid of autoactivation (Fig. 1A). Next, we performed pulldown experiments using full-length, HA-tagged ERK8 from a 293T cellular lysate and bacterially expressed, GST-tagged C-terminal domain of ERR␣. As shown in Fig. 1B, the ERR␣ fragment corresponding to the clone identified by the two-hybrid screening readily interacted, in vitro, with the ERK8 protein. To determine whether ERK8 is able to interact also in vivo with ERR␣, we performed co-immunoprecipitation experiments in 293T cells, co-transfecting EGFP-ERK8 with fulllength, AU1-tagged, human ERR␣. As shown in Fig. 1C, ERK8 evidently co-immunoprecipitated in vivo with ERR␣. Altogether, these results indicate a physical interaction between ERK8 and ERR␣.
ERK8 Promotes Re-localization of Nuclear ERR␣ to the Cytoplasm-To confirm the in vivo interaction of ERR␣ and ERK8, we next decided to investigate the cellular localization of these two proteins. Information about ERK8 subcellular localization is still limited. Therefore, we first sought to determine its subcellular localization in 293T cells, our experimental model. In these settings, ERK8 was mostly localized to the cytoplasm, whereas a much lower signal appeared in the nucleus ( Fig. 2A, left panels). In turn, in the same model, we confirmed the already described (31,32) predominantly nuclear localization of ERR␣ (Fig. 2B, left panels). To further support the aforementioned evidence and obtain quantitative information about the subcellular localization of ERK8 and ERR␣, we measured fluorescence intensity in the nucleus and cytoplasm, as shown in the representative images in the left panels of Fig. 2, A and B, respectively. As expected, ERK8 was mainly cytoplasmic ( Fig.  2A, right panel), whereas ERR␣ was mostly nuclear (Fig. 2B, right panel). Still, co-transfecting ERK8 and ERR␣, we observed that they co-localized to the cytoplasm only in cells expressing both proteins, whereas ERR␣ maintained its nuclear localization in cells not expressing ERK8 (Fig. 2C, left panels). As a control, the overexpression of another MAP kinase, JNK, together with ERR␣ did not affect subcellular localization of this nuclear receptor (Fig. 2D, left panels). As an additional confirmation to these observations and to obtain quantitative data, we counted the number of ERR␣-positive nuclei from cells expressing only ERR␣ or both ERR␣ and ERK8. ERK8 coexpression led to an ϳ80% reduction in the number of cells with ERR␣-positive nuclei (Fig. 2C, right panel). As a control, no decrease in the number of cells with ERR␣-positive nuclei occurred following JNK co-expression (Fig. 2D, right panel).
ERR␣, ERR␤, and ERR␥ belong to the same subfamily of nuclear receptors and share several features, such as the lack of an endogenous ligand and the ability to bind the same DNA consensus sequence (33). Nevertheless, co-expression with ERK8 did not interfere with nuclear localization of ERR␤ and ERR␥, showing a clear difference in their behavior compared with ERR␣ (supplemental Fig. S1A). These observations therefore confirm our data showing a direct interaction between ERR␣ and ERK8 and strongly suggest a specific role for ERK8 in selectively determining the subcellular localization of ERR␣.
ERK8 Re-localizes Nuclear ERR␣ to the Cytoplasm with a CRM1-dependent Mechanism-The antifungal compound leptomycin B (LMB) specifically blocks the nuclear export of proteins by preventing their association with the CRM1 export receptor (34). Indeed, the use of LMB already implicated this karyopherin in the export of different nuclear receptors. One such example is ER␣, whose nuclear export is inhibited by LMB in breast cancer cell lines (35,36).
As no information is yet available about ERK8 nucleocytoplasmic transport, we first determined the behavior of this protein upon LMB treatment in HeLa cells, often used to investigate CRM1 activity. As described in Fig. 3A, left panels, LMB induced nuclear compartmentalization of the otherwise predominantly cytoplasmic ERK8 protein, suggesting the existence of an active, CRM1-dependent mechanism of ERK8 nuclear export. As expected, LMB did not perturb the already nuclear localization of ERR␣ (Fig. 3B, left  panels).
We next wished to determine whether CRM1 mediates ERK8-dependent nuclear export of ERR␣. As depicted in Fig.

ERK8 as an ERR␣ Corepressor
3C, left panels, upper row, ERR␣ was predominantly localized in the nucleus in HeLa cells not expressing ERK8, whereas cells co-expressing this MAP kinase showed ERR␣ localization to the cytoplasm, in line with our previous observations in 293T cells. However, treating HeLa cells with LMB, ERR␣ localization was completely restrained to the nucleus, despite ERK8 co-expression (Fig. 3C, left panels, lower row), definitively suggesting CRM1 involvement in ERR␣ nuclear exclusion induced by ERK8. As an additional confirmation to these observations and to obtain quantitative data, we counted the number of ERK8-positive nuclei from cells expressing only ERK8 (Fig. 3A,  right panel), of ERR␣-positive nuclei from cells expressing only ERR␣ (Fig. 3B, right panel) and of ERR␣-positive nuclei from cells expressing both ERR␣ and ERK8 (Fig. 3C, right panel), in the presence or absence of LMB. As expected, in all cases LMB treatment led to an almost complete re-localization of both ERK8 and ERR␣ to the nucleus. Altogether, these data therefore allow us to conclude that ERK8 re-localizes ERR␣ to the cytoplasm through a CRM1-dependent mechanism.
ERK8 Inhibits ERR␣ Transcriptional Activity-Based on the previously described data and on ERK8 ability to control the activity of different nuclear receptors (9, 10), we next asked whether this MAP kinase was able to modulate ERR␣ transcriptional activity. It is known that ERR␣ and ER␣ can interact in vitro and induce transcription through both the classical ERE and the ERR␣ response element (ERRE/SFRE) (15,33). To avoid potential biases due to cross-talk between ERR␣ and ER␣, we therefore decided to use the ER␣-negative HeLa cells, a typical model system used for functional studies on ERR␣ transcriptional activity (37,38).
To investigate the transcriptional activity of this nuclear receptor, we used a firefly luciferase reporter vector, ERRE_Luc, in which the luciferase gene is under control of a minimal promoter harboring three ERRE/SFRE repeats (16). As expected (16), the activity of the reporter was dependent on ERR␣ expression (Fig. 4A). Next, we studied the effect of ERK8 on ERR␣ transcriptional activity by co-transfecting, together with the ERRE_Luc reporter, an ERR␣ expression vector alone or in combination with the ERK8 expression vector. Luciferase activity increased ϳ3-fold in the presence of ERR␣ but, remarkably, it returned to basal levels when ERR␣ and ERK8 were co-expressed (Fig. 4B), therefore indicating a corepressive role for ERK8 on ERR␣ transcriptional activity. As a control, the expression of ERK8 alone had no effect on the reporter (Fig.  4B). Furthermore, overexpression of other MAP kinases such as JNK could not counteract ERR␣-dependent induction of the ERRE_Luc reporter (Fig. 4C). Although ERR␤ was a very poor inducer of the reporter (39), ERR␤ and ERR␥ activities were not affected by ERK8 in HeLa cells (supplemental Fig. S1B), confirming the specificity of the ERK8 effect on ERR␣.
Due to the lack of a natural agonist, the induction of ERR␣ transcriptional activity heavily relies on "proteic coactivators," the best characterized of which is PGC1␣ (40). Therefore, we sought to determine whether ERK8 could counteract PGC1␣dependent ERR␣ activation as well. Fig. 4D clearly shows that, despite the remarkable induction of ERR␣ transcriptional activity by PGC1␣, ERK8 was still able to inhibit such activation.
ERR␣ is phosphorylated in vivo on multiple sites (27,28). It has been recently reported that some of these phosphorylation events are responsive to epidermal growth factor receptor (EGFR) stimulation and are able to enhance ERR␣ binding to DNA and, therefore, its transcriptional activity (28). As HeLa cells express high EGFR levels (Fig. 4E), we tested the ability of ERK8 to interfere with the activity of ERR␣ induced by the EGF signaling pathway in this experimental model. As shown in Fig.  4F, ERK8 retained its corepressive function on ERR␣ even when this nuclear receptor was activated by EGF. Similarly, ERR␣ stimulation induced by two human-activated EGFR oncogenic mutants (EGFR L858R and EGFR D770_N771 insNPG) (41) was also completely abolished by ERK8 expression (Fig. 4F). Altogether, our data therefore demonstrate that ERK8 is a bona fide ERR␣ corepressor, exerting such function in a highly specific manner on ERR␣, both on its basal activity and in situations in which this nuclear receptor is induced by proteic coactivators, such as PGC1␣, or by well characterized stimuli, such as activation of the EGF signaling pathway.
ERK8 Modulates the Expression of ERR␣ Target Genes-Based on evidence that ERK8 inhibits ERR␣ transcriptional activity in a luciferase reporter system, we sought to study the ability of ERK8 to control the endogenous expression of ERR␣ transcriptional targets as well. To this purpose, ERK8 expression was silenced in HeLa cells using a validated (5) siRNA to score, by real time PCR, the effects of ERK8 depletion on the expression levels of two well established ERR␣ target genes, CYP11A1 and CYP27A1 (42). In these conditions, whereas ERK8 mRNA levels underwent a 50% reduction (Fig. 5A) we observed a ϳ2and ϳ2.5-fold increase in CYP11A1 (Fig. 5B) and CYP27A1 mRNA expression (Fig. 5C), respectively. As a control, ERK8 mRNA depletion had no effect on the expression of a gene whose expression does not depend on ERR␣, namely ERK2 (Fig. 5D). These data therefore support our previous  MARCH  ERK8 Binds to ERR␣ through LXXLL Motifs Typical of Agonist-bound Nuclear Receptor Corepressors-Interactions of nuclear receptors with coregulatory proteins are mediated by conserved motifs, well defined for both coactivators and corepressors. Classical corepressors, which prevent unliganded nuclear receptors from being activated by ligands, are characterized by the "corepressor nuclear receptor" (CoRNR) box, typically having a sequence of (L/V)XX(I/V)I or alternatively LXXX(I/L)XXX(I/L) (where L is leucine, V is valine, I is isoleucine, and X is any amino acid) (43). Ligand binding favors the displacement of this kind of corepressors and the recruitment of coactivators, characterized instead by the LXXLL box (44). Still, a different family of corepressors that are able to inhibit active, ligand-bound, nuclear receptors through an LXXLL box, hence named "agonist-bound nuclear receptor corepressors," has been recently described (45).

ERK8 as an ERR␣ Corepressor
In silico analysis of the ERK8 protein sequence led us to identify two putative LXXLL boxes starting at amino acids 265 and 281, respectively (referred to as LXXLL Box 1 and LXXLL Box 2 herein), in the C-terminal domain of ERK8. Of note, the two putative LXXLL boxes appeared to be very well conserved among different species (Fig. 6A), the conservation of this particular structural motif being suggestive of its importance in the function of ERK8, otherwise exhibiting very low homology with its orthologs, especially in its C-terminal domain (3). Hence, to investigate whether the two LXXLL boxes are involved in the interaction with ERR␣, we generated ERK8 mutants in which we substituted leucine residues with alanines in LXXLL Box 1 (ERK8 3LA1 mutant), in LXXLL Box 2 (ERK8 3LA2), and in both (ERK8 3LA1-2), respectively. As shown in Fig. 6B, interaction of these mutants with ERR␣ was barely detectable. Next, based on the ability of ERK8 to prevent nuclear localization of ERR␣ (see above), we wanted to determine whether the LXXLL mutants, whose interaction with ERR␣ is hindered, could still affect the subcellular localization of this nuclear receptor. As shown in Fig. 6C, the ERK8 LXXLL mutants, whereas showing the same localization of wild-type ERK8 (middle row), lost the ability to re-localize ERR␣ to the cytoplasm. Biochemical fractionation in Fig. 6D also confirmed the increase of ERR␣ in the cytoplasm in the presence of wild-type ERK8 as well as the inability of the ERK8 LXXLL mutants to efficiently promote the nuclear export of ERR␣. Accordingly, 293T cells transfected with the nuclear receptor showed a more modest decrease in the number of ERR␣-positive nuclei when co-expressing the ERK8 LXXLL mutants, compared with cells co-expressing the wild-type ERK8 (Fig. 6E). Next, to ascertain whether the interaction between these two proteins is necessary for ERK8 corepressive function, we tested the ERK8 LXXLL mutants for their ability to inhibit ERR␣ transcriptional activity. We therefore co-transfected the ERRE_Luc reporter vector along with ERR␣ and either wild-type ERK8 or the different mutants. Fig. 6F clearly shows that all ERK8 mutants lost the ability to repress ERR␣ transcriptional activity, thus supporting the conclusion that the LXXLL boxes are essential for ERK8 corepressive function.
Altogether, our data identify two LXXLL motifs contained in the ERK8 C-terminal domain as necessary features for ERK8 to interact with ERR␣, to regulate its subcellular localization and to inhibit its transcriptional activity. Based on previous data indicating ERR␣ as a constitutively active nuclear receptor (18), such features also suggest ERK8 as a new member of the family of agonist-bound nuclear receptor corepressors, unique in its ability to control the functions of orphan members of this family of proteins, whose structure is already locked in a transcriptionally active conformation, even in the absence of a natural ligand.

ERK8 Negatively Regulates Endogenous ERR␣ Activity and Controls Its Cellular Localization, in Human Breast Epithelial
Cells-Next, we decided to evaluate the effect of ERK8 on endogenous ERR␣, in a biologically relevant cellular system, the human immortalized mammary epithelial MCF10A cell line. These cells are in fact often used to recapitulate differentiation and transformation of the human breast epithelium (46,47), a tissue whose malignant proliferation is indeed affected by ERR␣ B, AU1-ERR␣-induced ERRE_Luc luciferase reporter activity is inhibited by HA-ERK8; 20 l of the same cellular lysate was subjected to Western blot with anti-ERR␣ and anti-ERK8 antibodies. C, AU1-ERR␣-induced ERRE_Luc luciferase reporter activity is not inhibited by HA-JNK; 20 l of the same cellular lysate was subjected to Western blot with anti-ERR␣ and anti-HA antibodies. D, HA-ERK8 inhibits AU1-ERR␣-induced ERRE_Luc luciferase reporter activity following stimulation of AU1-ERR␣ with myc-tagged PGC1␣; 20 l of the same cellular lysate was subjected to Western blot with anti-ERR␣, anti-ERK8, and anti-PGC1␣ antibodies. E, expression levels of endogenous EGFR. Cellular lysates (30 g) from 293T, HeLa, and MCF10A cells were subjected to SDS-PAGE followed by Western blot with anti-EGFR and anti-Actin antibodies. F, HA-ERK8 inhibits AU1-ERR␣-induced ERRE_Luc luciferase reporter activity following stimulation of AU1-ERR␣ with EGF (100 g/ml, 4 h) or transfection with two constitutively active mutants of EGFR (EGFR L858R and EGFR D770_N771 insNPG); 20 l of the same cellular lysate was subjected to Western blot with anti-ERR␣ and anti-ERK8 antibodies. RLU, relative light units.  (22,23,26). Importantly, MCF10A cells express high levels of ERR␣ but no ER␣ (Fig. 7A and Ref. 9), allowing us to avoid possible biases due to cross-talk between ERR␣ and ER␣ (see above). Therefore, we first confirmed in these mammary epithelial cells our findings obtained in 293T and HeLa, by showing that ERK8 induced the cytoplasmic re-localization (Fig. 7B) and inhibited the transcriptional activity (Fig. 7C) of ectopically expressed ERR␣. Next, to determine whether our reporter assay was able to detect the signal generated by the endogenous ERR␣, we treated ERRE_Luc-transfected MCF10A cells with the ERR␣-specific inverse agonist XCT790. As a result of drug action on both ERR␣ transactivation potential and protein levels (25), the activity of the luciferase reporter underwent a marked dose-dependent decrease, thus indicating that the basal ERRE_Luc luciferase signal observed in MCF10A is mostly due to endogenous ERR␣ (Fig. 7D). Based on this information, we observed a highly significant decrease in the ERRE_Luc reporter activity in MCF10A cells transfected with ERK8, indicating that ERK8 can indeed inhibit the transcriptional activity of endogenous ERR␣ (Fig. 7E). Next, based on the ability of ERK8 to prevent nuclear localization of exogenously expressed ERR␣, we wanted to determine whether this MAP kinase could also affect the subcellular localization of the endogenous nuclear receptor. Indeed, as shown in Fig. 7F, only MCF10A cells expressing ERK8 showed cytoplasmic localization of the ERR␣ protein. In line with this observation, cell fractionation experiments showed a clear increase of ERR␣ levels in the cytoplasmic fraction of MCF10A cells stably expressing the ERK8 protein (Fig. 7G, left panels). As MCF10A cells also express endogenous ERK8 (8,9), we next confirmed the interaction between endogenous ERK8 and ERR␣ proteins. Indeed, in line with our results on ectopically expressed proteins (Figs. 1C and 6B), endogenous ERK8 evidently co-immunoprecipitated in vivo with endogenous ERR␣ (Fig. 7H). Ultimately, we set up to evaluate the ability of ERK8 to counteract activation of endogenous ERR␣ by stimuli relevant for breast cancer, e.g. activation of the EGFR pathway (48). To this aim, as MCF10A cells express very low levels of EGFR (see Fig. 4E), we decided to overexpress in these cells the constitutively active EGFR mutants, EGFR L858R and EGFR D770_N771 insNPG, which, although typically found in non-small cells lung carcinomas (41), well recapitulate activation of the EGFR pathway in human cancer. In agreement with our findings in HeLa cells (see Fig. 4F), activation of the EGFR signaling pathway in MCF10A cells increased the activity of endogenous ERR␣, whereas ERK8 retained its repressory effect even on the endogenous nuclear receptor stimulated by EGFR-derived activated oncogenes (Fig. 7I). These results therefore propose ERK8 as a corepressor of endogenous ERR␣, with the ability to promote its cytoplasmic localization and counteract its activation by positive cellular stimuli often deregulated in human cancers.

DISCUSSION
In this paper we demonstrate that ERK8, a member of the MAP kinase family of proteins, is a bona fide ERR␣ corepressor, able to modulate the subcellular localization and activity of this nuclear receptor. Indeed, our data support a model in which ERK8 binds to ERR␣, inhibits both its basal and stimulated (by PGC1␣ and the EGFR pathway) activity, and induces its localization to the cytoplasm (Fig. 8).
Recent work has already revealed that ERK8 negatively regulates the activity of androgen and glucocorticoid receptors by interacting with the HIC-5 coactivator (10), and that the product of its murine ortholog gene enhances the degradation of ER␣ (9). However, differently from what was demonstrated for these nuclear receptors, we show a direct interaction between ERK8 and ERR␣. Also, unlike ER␣, we observed no decrease in the protein levels of ERR␣ in the presence of ERK8. Therefore, it appears that ERK8 is a multifunctional corepressor for different nuclear receptors. Still, its corepressive activity is certainly not aspecific, as we demonstrate that it is able to discriminate among very similar proteins such as different members of the ERR subfamily. In this context, it is worth noting that ERK8 is able to influence the expression of ERR␣ target genes such as CYP11A1 and CYP27A1 (42). Interestingly, CYP11A1 encodes for the cholesterol side chain cleavage enzyme (P450scc), which mediates the initial and rate-limiting step in steroidogenesis, converting cholesterol to pregnenolone, the precursor of androgens, estrogens, and progesterone, and which has been heavily associated to prostate (49), breast (50), and endometrial (51) cancer risk. On the other hand, CYP27A1 encodes for sterol 27-hydroxylase, a cytochrome P450 family member implicated in the metabolism of vitamin D. Overall, the presence of ERK8 may therefore coordinately modulate, by multiple and different mechanisms, the activity of several nuclear receptors, at multiple levels, possibly in response to specific cellular stimuli.
Coregulatory proteins are key modulators of nuclear receptor-dependent transcription of target genes (52). In nuclear receptors, in fact, helix 12 (also referred to as activation func-

ERK8 as an ERR␣ Corepressor
tion 2) is critical in determining which kind of coregulator docks. Inactive nuclear receptors, unbound to ligands, have helix 12/activation function 2 in a conformation that provides access to corepressors interacting via the so-called corepressor nuclear receptor box (52), typically having a sequence of (L/V)XX(I/V)I or LXXX(I/L)XXX(I/L) (53). Upon agonist binding, helix 12/activation function 2 is repositioned in a conformation that no longer allows interaction of the corepressor nuclear receptor box with the corepressor-binding area, although favoring the interaction with coactivator boxes, typically having an LXXLL sequence (52). Therefore, typical corepressors lack the structural means to repress agonist-bound, active nuclear receptors. This model was initially challenged by the discovery of receptor interacting protein 140 (RIP140), which has the structural features of a coactivator, and included LXXLL sequences determining the interaction with the receptors, but acts as a corepressor (45). The discovery of a few more such coregulators gave rise to a small family of agonist-dependent corepressors whose interaction with activated nuclear receptors depends on LXXLL boxes (54). Moreover, ERR␣, even if not bound to any ligand, is constitutively active and arranged in a structure typical of ligand-bound receptors (18). It is therefore not surprising for us to find two functional LXXLL motifs in ERK8 mediating its corepressive functions on ERR␣. Together, these data led us to the conclusion that ERK8 is the first representative of a new class of corepressors that are able to control, through LXXLL boxes, the functions of orphan nuclear receptors whose structure is already locked in a transcriptionally active conformation, even in the absence of a natural ligand. In this context it is important to note that the two LXXLL boxes of ERK8 are perfectly conserved in mammals and birds, five of the six leucines are conserved in zebrafish, and the LXXLL box 2 is perfectly conserved down to Drosophila melanogaster (data not shown). Such remarkable conservation FIGURE 7. ERK8-dependent cytoplasmic re-localization of endogenous ERR␣ and inhibition of its transcriptional activity, in MCF10A cells. A, expression levels of endogenous ERR␣ in our experimental cellular models. Total cellular lysates (30 g) from 293T, HeLa, and MCF10A cells were subjected to SDS-PAGE followed by Western blot (WB) with anti-ERR␣ and anti-Actin antibodies. B, confocal microscopy images of MCF10A cells co-transfected with AU1-ERR␣ (500 ng) and HA-ERK8 (1 g), then incubated with anti-AU1 and anti-ERK8 antibodies and labeled with TRITC-and Cy2-conjugated secondary antibodies. C, ERR␣-induced ERRE_Luc luciferase reporter activity is inhibited by HA-ERK8 in MCF10A cells; 20 l of the same cellular lysate was subjected to Western blot with anti-ERR␣ and anti-ERK8 antibodies. D, ERRE_Luc basal activity in MCF10A cells is due to the transactivation potential of endogenous ERR␣. MCF10A cells were transfected with the ERRE_Luc luciferase reporter vector and treated with 10 or 20 M XCT790; after 24 h, cells were lysed and assayed for luciferase activity; 20 l of the same cellular lysate was subjected to Western blot with anti-ERR␣ antibody. E, HA-ERK8 inhibits ERRE_Luc luciferase reporter activity induced by endogenous ERR␣ in MCF10A cells (***, p Ͻ 0.001 according to unpaired Student's t test); 20 l of the same cellular lysates was subjected to Western blot with anti-ERR␣ and anti-ERK8 antibodies; F, confocal microscopy images of MCF10A cells transfected with HA-ERK8 (1.5 g), then incubated with anti-ERR␣ and anti-HA antibodies and labeled with TRITC-and Cy2-conjugated secondary antibodies and DAPI for nuclear staining. G, ERK8 expression increases the amount of ERR␣ in the cytoplasmic fraction of MCF10A cells; MCF10A stably expressing EGFP or HA-ERK8 (5 ϫ 10 6 cells) were subjected to fractionation using the Subcellular Protein Fractionation Kit (Thermo Scientific); lysates were subjected to SDS-PAGE followed by Western blot with anti-ERR␣ (top), anti-ERK8 (middle top), anti-IkB␣ (middle bottom), and anti-Lamin A/C (bottom) antibodies; H, interaction between endogenous ERK8 and ERR␣ proteins; total cellular lysate (1 mg) from MCF10A cells was subjected to immunoprecipitation with anti-ERR␣ antibody or with rabbit preimmune serum (4 h at 4°C), then subjected to SDS-PAGE and blotted with anti-ERK8 antibody (Kinasource) (top). Total cellular lysates (50 g) were blotted with anti-ERK8 (middle) and anti-ERR␣ (bottom) antibodies for normalization purposes. I, HA-ERK8 inhibits ERRE_Luc luciferase reporter activity induced by endogenous ERR␣ following stimulation of the nuclear receptor with two constitutively active mutants of EGFR (EGFR L858R and EGFR D770_N771 insNPG); 20 l of the same cellular lysates was subjected to Western blot with anti-ERR␣ and anti-ERK8 antibodies. ERR␣ is a constitutively active orphan nuclear receptor whose activity can be further increased by proteic coactivators (PGC1␣) (40) or EGFR-dependent signaling (27,28). ERK8 acts as a bona fide ERR␣-specific corepressor interacting with the nuclear receptor through its LXXLL motifs and promoting its re-localization from the nucleus to the cytoplasm. MARCH 11, 2011 • VOLUME 286 • NUMBER 10 appears even more relevant when compared with the otherwise extremely low conservation score of the ERK8 protein throughout evolution.

ERK8 as an ERR␣ Corepressor
Subcellular localization and dynamic movements between cellular compartments are important mechanisms used by cells to regulate the activity of different transcription factors (55). As concerns nuclear receptors, several groups have demonstrated that they continuously shuttle between the cytoplasm and the nucleus, as a consequence of a fine balance between operational strengths of nuclear localization signals and nuclear export signals (55,56). The subcellular compartmentalization and the timely and efficient localization of these transcription factors are usually modulated through the binding of specific ligands, such as steroid or thyroid hormones (55). In principle, all nuclear receptors, when bound to their natural ligands, translocate to the nucleus and become transcriptionally active, being the nuclear localization necessary for their direct action on target genes. On the other hand, unliganded, inactive receptors appear to have a more diverse localization pattern, from predominantly cytoplasmic to uniformly distributed between the cytoplasm and nucleus (57)(58)(59) to predominantly nuclear (60 -62). Moreover, steroid receptor localization has also been extended to mitochondria (63,64) and plasma membrane (65)(66)(67). Interestingly, some nuclear receptors, such as the ER␣, have even higher complexity, requiring a cytoplasmic localization, even when ligand-bound and active, to exert specific "nongenomic" functions alternative to direct promoter binding and necessary for cellular transformation (35). As mutations of ERK8 that prevent interaction also abolish its corepressive functions and ERR␣ cytoplasmic localization, we can state that direct interaction is a prerequisite for inhibition of ERR␣ activity as well as for its nuclear exclusion. However, the relationship between ERR␣ activity and cytoplasmic localization needs further investigation. Indeed, whereas ERR␣ nuclear exclusion prevents the interaction of this nuclear receptor with the promoters of its target genes and therefore may participate to the inhibition of its constitutive transcriptional activity, we cannot exclude that ERR␣ may also control, in the cytoplasm, alternative, non-genomic functions contributing to its biological activity, as described for ER␣ (35,68,69) and progesterone receptor (70). In this context, it is important to note that a naturally occurring short form of the metastatic tumor antigen 1 (MTA1s) is a characterized ER␣ corepressor that sequesters this steroid receptor to the cytoplasm through an LXXLL motif, enhancing non-genomic functions such as ERK2 activation (71). The analogies between the short form of the metastatic tumor antigen 1-ER␣ and the ERK8-ERR␣ situations may therefore suggest that ERK8, whereas negatively regulating ERR␣ transcriptional activity, may control as yet to be described non-genomic functions of this protein by regulating its nucleocytoplasmic shuttling. Further work will be required to establish the precise functions exerted by ERR␣ in the cytoplasm as well as the role of ERK8 in controlling such activities.
In addition to its physiological functions in cellular metabolism and in responses to stressors requiring shifts in energy production and utilization (12,13,72,73), an increasingly important role for ERR␣ in human malignancies is emerging.
Silencing of ERR␣ and the use of specific pharmacological inhibitors suggest a critical role for this nuclear receptor in the growth of highly aggressive forms of breast cancer (22)(23)(24)(25). Moreover, it is known that ERR␣ and ER␣ can interact in vitro and regulate transcription through the classical ERE (15,33), and that ERR␣ behaves as a repressor or an activator of ER-regulated transcription in a cell type-dependent manner (74). Hence, it has been proposed that, whereas ERR␣ tightly regulates estrogen responsiveness in normal breast cells, it may functionally replace ER in ER-negative breast tumors, thereby constitutively activating ERE-regulated transcription (74). Thus, the conversion of ERR␣ from a repressor to an activator may be a critical step in the progression of breast tumors to a hormone-independent phenotype, suggesting that ERK8 may provide opportunities for novel therapeutic approaches for the treatments of these tumors.
Evidences are accumulating about the involvement of ERK8 in human cancer (7,9). In the context of breast cancer, in particular, high ERK8 expression has been demonstrated in normal human mammary cells, whereas loss of its expression has been correlated with breast cancer progression and increased ER␣ levels (9). Our findings therefore suggest that, in breast, ERK8 may participate in maintaining tissue homeostasis not only by regulating ER␣ protein degradation (9) but also by controlling the activity and cellular localization of ERR␣, whose role in cellular metabolism and transformation has now been well established (28,75,76). Hence, this MAP kinase might regulate estrogen transcriptional targets both directly, by enhancing ER␣ degradation, and indirectly, by modulating the physical and functional interactions of this nuclear receptor with ERR␣. Based on the well established role of estrogens in breast cancer progression, through the expression of cellular proto-oncogenes (e.g. c-MYC) as well as extracellular matrix molecules (e.g. syndecan-2 and metalloproteinase-9) (77,78), the ability of ERK8 to control and integrate signals from ER␣ and ERR␣ receptors in normal and transformed mammary cells warrants further investigation.