Transcriptional cross talk between orphan nuclear receptor ERRγ and transmembrane transcription factor ATF6α coordinates endoplasmic reticulum stress response

Orphan nuclear receptor ERRγ is a member of nuclear receptor superfamily that regulates several important cellular processes including hepatic glucose and alcohol metabolism. However, mechanistic understanding of transcriptional regulation of the ERRγ gene remains to be elucidated. Here, we report that activating transcription factor 6α (ATF6α), an endoplasmic reticulum (ER)-membrane–bound basic leucine zipper (bZip) transcription factor, directly regulates ERRγ gene expression in response to ER stress. ATF6α binds to ATF6α responsive element in the ERRγ promoter. The transcriptional coactivator peroxisome proliferator-activated receptor gamma coactivator 1-α (PGC-1α) is required for this transactivation. Chromatin immunoprecipitation (ChIP) assay confirmed the binding of both ATF6α and PGC1α on the ERRγ promoter. ChIP assay demonstrated histone H3 and H4 acetylation occurs at the ATF6α and PGC1α binding site. Of interest, ERRγ along with PGC1α induce ATF6α gene transcription upon ER stress. ERRγ binds to an ERRγ responsive element in the ATF6α promoter. ChIP assay confirmed that both ERRγ and PGC1α bind to a site in the ATF6α promoter that exhibits histone H3 and H4 acetylation. Overall, for the first time our data show a novel pathway of cross talk between nuclear receptors and ER-membrane–bound transcription factors and suggest a positive feed-forward loop regulates ERRγ and ATF6α gene transcription.


INTRODUCTION
Estrogen-related receptors (ERRs) are members of the NR3B subfamily of nuclear receptors, which include ERRa, ERRb and ERRg. These orphan nuclear receptors regulate transcription via estrogen response elements and the closely related ERR response elements (ERREs) but do not bind endogenous estrogen (1). The ERRs are named owing to the conservation in the structure of their DNA-binding domains with the highly homologous Estrogen Receptor (2). Crystallographic studies indicate that the ERRs along with ERRg are constitutively active without a natural ligand, while several synthetic ligands either stimulate or repress the activity of ERRg by promoting or disrupting ERR-coactivator interactions (3). Among them, GSK5182, a 4-hydroxy tamoxifen analog, is a selective inverse agonist of ERRg relative to other nuclear hormone receptors (4). ERRg is primarily expressed in heart, brain, kidney, pancreas and liver *To whom correspondence should be addressed. Tel: +82 62 530 0503; Fax: +82 62 530 0506; Email: hsc@chonnam.ac.kr tissues (3). The transcriptional activity of the ERR family is dependent on interactions with coactivators, in particular PGC-1a and PGC-1b (5). ERRa and ERRg regulate mitochondrial programs involved in oxidative phosphorylation and a nuclear-encoded mitochondrial genetic network that coordinates the postnatal metabolic transition in the heart (5). We previously reported that hepatic ERRg regulates hepatic gluconeogenesis by directly binding to the Phosphoenolpyruvate carboxykinase and Glucose 6-phosphatase (G6Pase) promoters along with coactivator PGC-1a (6). Previous results from our laboratory also demonstrated that ERRg directly binds to the LIPIN1 promoter along with coactivator PGC-1a to regulate LIPIN1 gene expression, and inhibits hepatic insulin signaling (7), ERRg controls hepatic CB1 receptor-mediated CYP2E1 expression at the transcriptional level and thus contributes to the oxidative liver injury by alcohol (8). Finally, hypoxia induces pyruvate dehydrogenase kinase 4 (PDK4) gene expression through induction of ERRg (9). Although all these reports clearly suggest a key role of ERRg in different cellular processes, its role in endoplasmic reticulum (ER) stress is yet to be determined.
Recently, numerous studies demonstrate the importance of ER stress in the pathogenesis of various liver diseases, including chronic viral hepatitis, insulin resistance, nonalcoholic fatty liver disease, ischemia-reperfusion injury, genetic disorders of protein misfolding and alcoholic liver disease (10)(11)(12). The ER stress response involves the function of three molecular components: protein kinase R-like ER kinase, inositol requiring enzyme-1/ X-box binding protein (XBP)-1 and activating transcription factor 6a (ATF6a) (13). Among these, ATF6a is a member of the ATF/cAMP response element-binding protein basic-leucine zipper family of DNA-binding proteins (14). On induction of ER stress, ATF6a translocates from the ER to the Golgi (15), where it is cleaved by site 1 and 2 proteases (16). Proteolytic cleavage of ATF6a directly induces transcriptional activation of ER chaperones and other enzymes that are essential for protein folding (15)(16)(17)(18). In addition to posttranslational modification of ATF6a, accumulating evidence suggests that ER stressors, including hypoxia and tunicamycin (Tm) upregulate ATF6a mRNA expression, which suggests that an increase in the expression of ATF6a is also important for the ER stress response (16,19). ATF6a has been reported to regulate Glucose-Regulated Protein78 (GRP78) gene expression (20). ATF6a also interacts with serum response factor to regulate seruminduced expression of the c-fos gene (21). One report also suggests that ATF6a concertedly works with coactivator PGC1a to regulate different gene expression (22). Although all these reports clearly suggest a key role of ATF6a in regulation of different cellular factors, its role in regulation of nuclear receptors during ER stress is yet to be determined.
PGC-1a, a member of a small family of coactivators, was identified using yeast two-hybrid assays for peroxisome proliferator-activated receptor g (PPARg)-interacting proteins (23) and is implicated in mitochondrial metabolism, thermogenesis, mitochondrial biogenesis, adipocyte differentiation, gluconeogenesis and glucose uptake (24,25) and to interact with a number of other nuclear receptors such as glucocorticoid receptor (GR) (26), nuclear respiratory factor-1 (NRF-1) (27), hepatocyte nuclear factor 4a (HNF4a) (28), estrogen receptor a (ERa) (29), peroxisome proliferator-activated receptor a (PPARa) (30), retinoid X receptor (RXR) (31) and ERRa (32). PGC-1a is a 798 amino acid multifunctional protein harboring several domains with distinct activities. The aminoterminal domain exhibits a transcriptional activation function (29) that is followed by an overlapping region involved in interactions with nuclear receptors containing two well-characterized NR boxes required for receptor recognition (24,29,31,33). For ligand-dependent interaction with the nuclear receptors at least one of the three NR box motifs of PGC-1a is required (29,31,33). The lysinerich region (residues 214-250) is thought to contain putative nuclear localization signals. PGC-1a targets promoters by interacting directly with numerous DNAbinding transcription factors and then coordinating several biochemical events, including recruitment of chromatin-modifying enzymes such as p300/CREB-binding protein (CBP) and steroid receptor coactivator-1 (SRC-1), interaction with the basal transcription machinery and linking of transcription to RNA splicing (34).
Previous reports suggest different nuclear receptors and many transcription factors cross talk to regulate mammalian gene expression in response to ER stress. ER stressinduced activation of ATF6a decreases insulin gene expression via upregulation of orphan nuclear receptor small heterodimer partner (SHP) (35). One report suggests that nuclear receptor PPARb/d is regulated by ATF4 (36). XBP1 increases ERa transcriptional activity (37,38). ERmembrane-bound cyclic AMP responsive element binding protein-H (CREBH) is regulated by ER stress (39), fatty acids and PPARa (40) and HNF4a (41). Several reports also demonstrate that PPARd activation rescues pancreatic b-cells from palmitate-induced ER stress through enhanced fatty acid oxidation (42). Furthermore, ER stress-induced CHOP (C/EBP Homologous Protein) enhances nuclear factor-kb (NF-kb) signaling via repression of PPARg (43). Therefore, all these previous reports indicate that different nuclear receptors and transcription factors coordinate the mammalian ER stress response under different physiological conditions.
Here, we examined the mechanism of cross talk between a nuclear receptor, ERRg and an ER-membrane-bound bZIP transcription factor, ATF6a. In response to ER stress, expression of both transcription factors increases significantly through reciprocal activation. ATF6a directly binds to the ERRg promoter, and ERRg directly binds to the ATF6a promoter. Of most interest, PGC1a acts as a coactivator in both cases. The physical interaction of PGC1a with both ERRg and ATF6a increases significantly upon ER stress that identifies the importance of PGC1a in this cross talk. We observed a significant increase in histone H3 and H4 acetylation in both promoters upon induction of ER stress by Tm treatment. Knockdown of either factor significantly decreases the expression of the other, suggesting their ability to trans-activate each other. Moreover, our study reveals ERRg is induced earlier that ATF6a in response to ER stress. Together, we present a novel mechanistic pathway that would encourage further study to elucidate the relation between nuclear receptors and ER stress.
Cell culture, transient transfection and luciferase assay AML12, 293T and HeLa cells were obtained from the American Type Culture Collection. Maintenance of cell lines and transient transfection assays were performed using Lipofectamine2000 transfection reagent (Invitrogen) according the manufacturer's instructions as described elsewhere (48). Briefly, cells were transfected with indicated reporter plasmids together with expression vectors encoding various transcription factors or treated with various chemicals. Total cDNA used for each transfection was adjusted to 1 mg/well by adding appropriate amount of empty vector and pCMV-b-gal plasmid was used as an internal control. The luciferase activity was normalized to b-galactosidase activity and expressed as relative luciferase units. The generation of ATF6a-null hepatocyte cell lines from ATF6a-null mice was previously described (18,49).

Coimmunoprecipitation assay and western blot analysis
Coimmunoprecipitation (Co-IP) and western blot analyses were performed as described previously (48). For Co-IP from tissue extracts, C57BL/6J mice (n = 5) were maintained ad libitum for desired experimental period and sacrificed. Liver tissue samples were used for Co-IP assay. For western blot analysis, cell lysates were prepared and analyzed as previously described (50).

Confocal microscopy
Confocal microscopy was performed as described elsewhere (48). In brief, the HeLa cells grown on gelatin-coated coverslips were transfected using Lipofectamine2000 transfection reagent (Invitrogen) according to the manufacturer's instructions. At 24 h after transfection, the cells were fixed with 2% formaldehyde, immunostained and subjected to observation by confocal microscopy.

RNA interference
Knockdown of PGC1a was performed using the pSuper vector system (50,51). AML12 cells were transfected with siRNA constructs using Lipofectamine 2000 (Invitrogen) according to the manufacturer's instructions. siRNAtreated cells were analyzed using reverse transcription polymerase chain reaction (RT-PCR) to measure the extent of knockdown.

Reverse transcriptase PCR and quantitative real-time PCR analysis
Total RNA was isolated using the TRIzol reagent (Invitrogen) according to the manufacturer's instructions. The mRNAs of ATF6a and PGC1a were analyzed by RT-PCR or quantitative real-time RT-PCR (qPCR) as indicated. DNA samples from total RNA reverse transcription or from chromatin immunoprecipitation (ChIP) assays served as the templates for qPCR, which were performed with TOPreal SYBR Green PCR Kit (Enzynomics) and the Step One Plus real-time PCR system (Applied Bioscience) in triplicate. mRNA expression levels were normalized to those of b-actin (ACTB). The RT-PCR and qPCR primer sequences are available on request.

ChIP assay
Formaldehyde cross-linking of cells, ChIPs and real-time PCR analyses were performed as described elsewhere (5). After sonication, soluble chromatin was subjected to immunoprecipitation using anti-ATF6a, anti-PGC1a, anti-Acetyl Histone 3, anti-Acetyl Histone 4 and anti-ERRg antibody. DNA was recovered by phenol/chloroform extraction and analyzed by PCR and/or qPCR using primers against relevant promoters.

Statistical analysis
Data are expressed as means ± SEM. Statistical analysis was performed using the two-tailed Student t test. Differences were considered statistically significant at P < 0.05.

ER stress induces ERRc gene expression
ER stress promotes LIPIN2-dependent hepatic insulin resistance (45) and ERRg is a novel transcriptional regulator of LIPIN1, and inhibits hepatic insulin signaling (7). To investigate whether there was any connection between ER stress and ERRg, AML12 (Mouse hepatoma cell line) cells were treated with known ER stress inducers, Tm, Thapsigargin (45) and Brefeldin A (52). A significant increase in the ERRg mRNA level was observed by qRT-PCR, though there was no significant increase in ERRa and ERRb mRNA level ( Figure 1A). Similar results were obtained when western blot analysis was performed after treating AML12 cells with Tm in a timedependent manner ( Figure 1B). To test the effect of Tm in vivo, mice were injected with Tm. A significant increase in ERRg mRNA (Supplementary Figure S1A) and protein ( Figure 1C) was observed in a time-dependent manner reconfirming our in vitro findings. To evaluate the potential role of ER stress on ERRg promoter activity, a transient transfection assay was performed with an ERRg promoter containing reporter. Tm treatment significantly increased ERRg promoter activity ( Figure 1D). Taken together, these initial results indicate that ER stress specifically increases ERRg gene expression in hepatocytes.

ER stress induces ERRc gene expression via ATF6a
The response to ER stress is a coordination of signaling pathways that activates several transcription factors including ATF6a, CREBH, ATF4, XBP1, and CHOP (53). These transcription factors induce different genes involved in numerous biological processes (6,7,44,54,55). To assess whether any of these transcription factors was involved in ER stress mediated induction of ERRg gene expression, XBP1, ATF4, ATF6a, CHOP or CREBH cDNAs were overexpressed in AML12 cells. Interestingly, only ATF6a overexpression significantly increased ERRg mRNA level ( Figure 2A). To further elucidate the role of these factors, transient transfection assays were performed with the ERRg promoter containing reporter along with XBP1, ATF4, ATF6a, CHOP or CREBH expression vectors. Only ATF6a significantly (>10 fold) activated the ERRg promoter, consistent with our previous observation ( Figure 2B). Though the promoter activity was also increased in the presence of XBP1, the increase was negligible compared to ATF6a. To confirm the role of ATF6a, next we checked the effect of adenovirus mediated overexpression of ATF6a (Ad-ATF6a) in AML12 cells. More than 4 fold increase in ERRg protein level was observed in presence of Ad-ATF6a compared to control ( Figure 2C). Similar results were obtained when ATF6a was overexpressed in HepG2 cells (human hepatoma cell line) (Supplementary Figure S1B and S1C). All these results indicated that ER stress induced ERRg gene expression was mediated through ATF6a. To test this hypothesis, endogenous ATF6a was knocked down by Ad-shATF6a in presence of Tm in AML12 cells. As expected, the Tm mediated increase in ERRg protein level was significantly decreased in response to ATF6a knock down ( Figure 2D). To further verify the role of ATF6a, wild-type and Atf6a-deleted immortalized hepatocyte cell lines were treated with Tm. The ERRg protein level ( Figure 2E) was significantly reduced in Atf6a-deleted cells compared to wild-type cells under basal conditions, as well as after Tm treatment. Collectively these results demonstrate that ATF6a mediates the induction of ERRg gene expression by ER stress.

ATF6a induces ERRc gene transcription via an ATF6aRE
Next, we attempted to ascertain the molecular mechanism of ATF6a-mediated ERRg gene induction. It was previously reported that PGC1a acts as a coactivator of ATF6a (22). To test whether PGC1a along with ATF6a has any role in the induction of ERRg, transient transfection assays were performed with mouse ERRg promoter containing reporter and ATF6a and PGC1a expression vectors. ATF6a significantly increased ERRg promoter activity, and this activity was further increased in the presence of PGC1a (>15-fold) ( Figure 3A). Similar results were obtained for the human ERRg promoter (Supplementary Figure S1D). To identify the DNA sequence conferring ATF6a-mediated ER stress effect on the ERRg promoter, a series of deletion constructs was analyzed. Deletion of the ERRg promoter sequence from 1.5 to 1.253 kb drastically decreased the promoter activity conferred by ATF6a, suggesting that the region from 1.5 to 1.253 bp conferred the activation of ERRg promoter ( Figure 3B). It was previously reported that ATF6a binds to a consensus sequence (G)(G)TGACGTG(G/A) (17). We aligned this sequence with ERRg promoter sequence, and although we could not find a perfect match to the consensus sequence, TTTGACTGAG spanning region 1.5-1.253 kb was found. To test whether TGAC may be the core sequence critical for ATF6a binding, transient transfection assays were performed using wild-type and TGAC-mutant reporters with Tm and ATF6a. This mutant reporter did not show any significant response to either Tm treatment or ATF6a cotransfection ( Figure 3C). Next, ChIP assay was performed to monitor the effect of Tm on ATF6a and PGC1a recruitment to the endogenous ERRg gene promoter. Under basal conditions, both ATF6a and PGC1a occupied the ERRg promoter. However, Tm treatment significantly augmented ATF6a and PGC1a occupancy on the ERRg promoter ( Figure 3D). To further confirm the binding site, ChIP assay was performed with the wild-type and TGAC-mutant ERRg promoter. The results demonstrated that ATF6a and PGC1a were present in ERRg promoter and Tm further induced ATF6a and PGC1a binding to ERRg chromatin. As expected, no binding was observed in the TGAC mutant ERRg promoter ( Figure 3E), therefore suggesting that Tm treatment activates the ERRg gene transcription via enhancing ATF6a and PGC1a binding to the promoter. The ChIP assay results provide critical in vivo evidence that the Tm/ATF6a-PGC1a signaling pathway increases ERRg gene transcription. Because gene activation is often associated with increased histone acetylation (57), to determine whether Tm treatment results in increased template-associated histone (H3 and/or H4) acetylation of the ERRg gene promoter, ChIP assay was performed ( Figure 3F). Tm treatment as well as adenoviral overexpression of ATF6a or PGC1a increased acetylation of H3 (Ac-H3) and H4 (Ac-H4) on the ATF6a-responsive region of ERRg promoter, whereas knockdown of endogenous ATF6a or endogenous PGC1a significantly reduced the histone (H3 and/or H4) acetylation. Overall these results demonstrate that ER stress augments the binding of ATF6a and PGC1a to the ERRg promoter and increases template-associated histone (H3 and H4) acetylation to facilitate ERRg gene transcription.

ER stress induces ATF6a gene expression via ERRc
Previous reports suggest different nuclear receptors and transcription factors regulating mammalian ER stress response cross talk with each other under conditions of ER (38,40,43,56). In addition, the nuclear receptor HNF4a regulates transcription of CREBH (38).
To investigate whether ERRg regulates any transcription factors regulating mammalian ER stress response, ERRg was overexpressed by adenovirus (Ad-ERRg) in AML12 cells. Surprisingly, a >10-fold increase in ATF6a mRNA level was observed, and although XBP1, ATF4 and CREBH mRNA levels also increased to some extent, the increase was insignificant compared with ATF6a ( Figure 4A). Similar results were obtained when ERRg was overexpressed in HepG2 cells (Supplementary Figure S3A). Because Tm induces transcription of ATF6a (16) and ERRg (Figure 1), we examined the time-course of ERRg and ATF6a induction in response to ER stress in AML12 cells. Tm induced expression of ERRg within 1 h, whereas ATF6a optimal expression took almost 3 h. Thus, ERRg is apparently induced before ATF6a in response to ER stress ( Figure 4B). Furthermore, overexpression of ERRg in AML12 cells ( Figure 4C) and in mouse liver tissue ( Figure 4D) resulted in almost 4-and 4.5-fold increase in ATF6a active form (ATF6a-N), respectively. Similar results were obtained when ERRg was overexpressed in HepG2 cells (Supplementary Figure S3B). As we noticed an early gene induction of ERRg compared with ATF6a upon ER stress ( Figure 4B) and overexpression of ERRg increased ATF6a expression both in vivo and in vitro ( Figure 4A, C and D), we speculated that ERRg could be responsible for the Tm-mediated increase in ATF6a gene expression. To verify this, we knocked down endogenous ERRg by Ad-shERRg in AML12 cells. As expected, the Tm-mediated increase in ATF6a-N protein level was significantly reduced after ERRg knockdown ( Figure 4E). Similar 293T cells were transfected with wild-type or ATF6aRE-mutant ERRg promoter along with ATF6a plasmid DNAs or treated with Tm (5 mg/ml). Data are representative of three independently performed experiments and shown as mean ± SD; *P < 0.05 and **P < 0.005 using Student's t-test.
(D) ChIP assay to detect the binding of ATF6a and PGC1a to the endogenous ERRg promoter by semiquantitative PCR. AML12 cells were treated with DMSO or Tm for 12 h. After completion of the treatment, Chromatin fragments were prepared and immunoprecipitated with ATF6a, PGC1a or IgG control antibodies. DNA fragments covering À1460 to À1325 and À428 to À264 elements on the ERRg promoter were PCR-amplified. Ten percent of the soluble chromatin was used as input. Data are representative of three individually performed experiments. (E) AML12 cells were transfected with wild-type or TGAC-mutant ERRg promoter. Following transfection, cells were treated with DMSO or Tm (5 mg/ml) for 12 h. Soluble chromatin was prepared and immunoprecipitated with antibody against ATF6a, PGC1a or IgG only as indicated. Ten percent of the soluble chromatin was used as input. Semiquantitative PCR (left panel) and qPCR (right panel) was performed to determine and quantify the binding of ATF6a and PGC1a to transfected ERRg promoter. Data are representative of three individually performed experiments. *P < 0.05 and **P < 0.005 using Student's t-test. ND, not detectable. (F) ChIP assay for detection of histone acetylation at the ATF6a/PGC1a binding site in the endogenous ERRg promoter under the indicated conditions in AML12 cells. Chromatin fragments were prepared and immunoprecipitated with Acetyl-Histone 3 and Acetyl-Histone 4 antibodies. DNA fragments covering À1460 to À1325 elements in the ERRg promoter were qPCR-amplified as described in the 'Materials and Methods' section. Data are representative of three independently performed experiments and shown as mean ± SD; *P < 0.05 and **P < 0.005 using Student's t-test.
results were observed for ATF6a mRNA level (Supplementary Figure S3C). To further confirm this, GSK5182, an inverse agonist of ERRg (6), which specifically binds to ERRg and inhibits transcriptional activity of ERRg, was used. In agreement with the previous results, we noticed that GSK5182 treatment substantially reduced the protein level of ATF6a-N in mouse liver ( Figure 4F). Overall we demonstrate that during ER stress, ATF6a mediates the increase in ERRg gene expression, ERRg in turn mediates the increase in ATF6a gene expression.

ERRc regulates ATF6a transcription via an ERRE
Next, we focused on the molecular mechanism of how ERRg overexpression led to the increase in ATF6a protein. A transient transfection assay was performed with the ATF6a promoter along with ERRg and PGC1a expression vectors, as PGC1a acts as a coactivator of ERRg (6). ERRg significantly increased ATF6a promoter activity, and this was further augmented in presence of PGC1a ( Figure 5A). Similar results were obtained for the human ATF6a promoter (Supplementary Figure S3D). To identify the DNA sequence conferring ERRg-mediated ER stress effect on the ATF6a promoter, we used a series of deletion constructs of ATF6a promoter for transient transfection assay. Deletion of the ATF6a promoter sequence from 2.6 to 2.3 kb drastically decreased the promoter activity conferred by ERRg, suggesting that the region from 2.6 to 2.3 kb conferred the activation of the ATF6a promoter ( Figure 5B). It was previously reported that ERRg binds to a sequence AGGTCA (7). A close investigation of the ATF6a promoter revealed a close sequence, AGGTCC, in between region 2.6 kb and 2.3 kb. To check whether the sequence AGGTCC mediates Tm-or ERRg-induced activation of ATF6a, transient transfection assays were performed using wild-type and AGGTCC-mutant reporters with Tm treatment or ERRg expression vector. This mutant reporter did not respond significantly to either Tm treatment or ERRg coexpression ( Figure 5C). Next we performed ChIP assay to detect whether endogenous ERRg or PGC1a binds to the ATF6a promoter upon Tm treatment. Tm treatment significantly increased ERRg and PGC1a occupancy to the ATF6a promoter compared with control cells ( Figure 5D). To further confirm the binding site, ChIP assay was performed with the wild-type and mutant ATF6a promoter. ChIP assay results demonstrated that ERRg and PGC1a were present in ATF6a promoter, and Tm treatment further induced ERRg and PGC1a binding to the ATF6a promoter. We could not detect any binding for AGGTCC mutant ATF6a promoters ( Figure 5E), indicating that Tm activates the ATF6a gene transcription via enhancing ERRg and PGC1a binding to the promoter. We also observed a significant increase in acetylation of H3 (Ac-H3) and H4 (Ac-H4) on ATF6a promoter in response to Tm treatment as well as adenoviral overexpression of ERRg or (D) ChIP assay shows the binding of ERRg and PGC1a to the endogenous ATF6a promoter by semiquantitative PCR. AML12 cells were treated with DMSO or Tm for 12 h. After completion of the treatment, chromatin fragments were prepared and immunoprecipitated with ERRg, PGC1a or IgG control antibodies. DNA fragments covering À2482 to À2375 and À732 to À551 elements on the ERRg promoter were PCRamplified. Ten percent of the soluble chromatin was used as input. Data are representative of three individually performed experiments. (E) AML12 cells were transfected with wild-type or AGGTCC-mutant ATF6a promoter. Following transfection, cells were treated with DMSO or Tm (5 mg/ml) for 12 h. Soluble chromatin was prepared and immunoprecipitated with antibody against ERRg, PGC1a or IgG only as indicated. Ten percent of the soluble chromatin was used as input. Semiquantitative PCR (left panel) and qPCR (right panel) were performed to determine and quantify the binding of ERRg and PGC1a to transfected ATF6a promoter. Data are representative of three individually performed experiments. *P < 0.05 and **P < 0.005 using Student's t-test. ND, not detectable. (F) ChIP assay for detection of histone acetylation at the ERRg/PGC1a binding site under the indicated conditions in AML12 cells. Chromatin fragments were prepared and immunoprecipitated with Acetyl-Histone 3 and Acetyl-Histone 4 antibodies. DNA fragments covering À2482 to À2375 element on ATF6a promoter were qPCR-amplified as described in the 'Materials and Methods' section. Data are representative of three independently performed experiments and shown as mean ± SD; *P < 0.05 and **P < 0.005 using Student's t-test.
PGC1a, whereas knockdown of endogenous ERRg or PGC1a resulted in significant decrease in acetylation of histone even after Tm treatment ( Figure 5F). As a whole, these results describe that ERRg mediates induction of the ATF6a gene on ER stress.

PGC1a regulates transcriptional cross talk of ATF6a and ERRc
Next, to elucidate the role of PGC1a in this ATF6a-ERRg cross talk in more detail, a transient transfection assay was performed with ERRg and the ATF6a promoter ( Figure 6A left and right panel, respectively). Tm treatment significantly increased both ERRg and ATF6a promoter activity but this activation was severely compromised when endogenous PGC1a was knocked down. Next, Ad-ATF6a along with Ad-PGC1a significantly increased the ERRg protein level but this protein level was significantly reduced on knockdown of endogenous PGC1a in AML12 cells ( Figure 6B). Similar results were observed on overexpression of ERRg and PGC1a in AML12 cells. Ad-ERRg along with Ad-PGC1a significantly increased the ATF6a-N protein level but this protein level was significantly reduced on knockdown of endogenous PGC1a ( Figure 6C). Next, we tested whether PGC1a physically interacts with ERRg and ATF6a. In absence of ER stress, PGC1a interacted with both ERRg and ATF6a, but this interaction was significantly enhanced for both ERRg and ATF6a in presence of ER stress ( Figure 6D left and right panel, respectively). To investigate whether PGC1a-ERRg and PGC1a-ATF6a were co-localized in the same subcellular compartment, confocal microscopy was performed in HeLa cells ( Figure 6E upper and lower panel). Our results demonstrated that both ERRg and ATF6a co-localized with PGC1a in the nucleus as can be evidenced from the merged image. Overall we demonstrate that PGC1a is the key element in the cross talk between ERRg and ATF6a.

Dependence of both ERRc and ATF6a on each other
Finally, we examined the regulation of target genes of both ERRg and ATF6a, as our results (Figures 2-5) demonstrated that ATF6a regulates ERRg gene expression on ER stress and vice versa. As PDK4 is an ERRg target gene (9), we examined whether ectopic expression of ATF6a had any role in PDK4 gene expression. Overexpression of ATF6a significantly increased PDK4 expression, and this increase was significantly attenuated on endogenous ERRg knockdown in AML12 cells ( Figure 7A). To provide insight into the mechanism, we mutated the ERRRE in the PDK4 gene promoter. Where both Tm and ATF6a activated the wild-type PDK4 promoter, on mutation of the ERRRE, neither Tm nor ATF6a could activate the PDK4 promoter ( Figure 7B). Likewise, as GRP78 is a target gene of ATF6a (58), we analyzed whether ectopic expression of ERRg had any regulatory role in GRP78 gene expression.
Overexpression of Ad-ERRg significantly increased GRP78 protein, but this increase was significantly compromised on knockdown of endogenous ATF6a by Ad-shATF6a in AML12 cells ( Figure 7C). Mutation of the ATF6aRE on the GRP78 promoter destroyed activation by either Tm or ERRg ( Figure 7D). Taken together, we demonstrate that Tm/ATF6a can induce ERRg target genes by regulating ERRg itself and Tm/ ERRg can induce ATF6a target genes by regulating ATF6a gene expression.

DISCUSSION
ER stress activates the unfolded protein response to generate multiple transcription factors that function in different cellular phenomena, including chromatin remodeling (35,45,10,11,59). In relation to coactivation of both ERRg and ATF6a by ER stress, the present study provides direct evidence for ER stress in induction of ERRg via ATF6a, but also for a newly recognized function of ERRg in transcriptional regulation of ATF6a in response to ER stress. Moreover, physical and functional interactions of both ERRg and ATF6a with coactivator provide a mechanistic basis for the ER stress-mediated induction of ERRg and ATF6a gene expression and suggest the importance of chromatin remodeling during subsequent transcriptional events. Our investigation of transcriptional cross talk between the nuclear receptor ERRg and the ER-membrane-bound bZIP transcription factor ATF6a was motivated by findings that demonstrate many nuclear receptors either regulate specific transcription factors during ER stress or are regulated by them (41,35). Here, we investigated the role of ER stress in ERRg gene expression and subsequently how ERRg regulates ER stress-induced transcription factors. We observed a significant enhancement in ERRg gene expression in response to the ER stressor Tm (Figure 1) that further establishes the interconnection between nuclear receptors and ER stress. Several previous studies show that X box binding protein 1 (XBP-1), which is similar to ATF6a, regulates ERa transcriptional activity through large-scale chromatin unfolding (37,38). Our findings suggest that ATF6a mediates the effect of ER stress on ERRg (Figure 2A-D). These findings were further supported by a significantly low ERRg protein level in ATF6a-null cells under basal conditions, as well as on Tm treatment. Previously, it was reported that ATF6a activates target genes through direct binding to an ATF6aRE in the promoters of target genes (17). This led us identify one ATF6a binding site in the ERRg promoter. ChIP assay further confirmed ATF6a binding to the ERRg promoter in response to ER stress. It has been reported that PGC1a is closely associated with the transcriptional activity of ATF6a (22). We also found that PGC1a plays an important role in ATF6a-mediated regulation of ERRg gene expression. Using an in vivo ChIP assay we demonstrated significant recruitment of PGC1a to ATF6aREs in the ERRg promoter during ER stress. Gene repression is often associated with decreased histone acetylation (56). Chromatin remodeling also occurs during coactivator gene transcription (60). In agreement with previous reports, we observed a significant increase in template-associated histone H3 and H4 acetylation at the ATF6aRE in the ERRg promoter ( Figure 3F).
PDK4 is a member of the pyruvate dehydrogenase kinase superfamily (PDK1, À2, À3, À4) that regulates glucose metabolism. ERRg induces PDK4 expression (61). Hypoxia and/or nutrient deprivation cause ER stress to induce PDK4 transcription through induction of ERRg (9). ER stress potentiates hepatic insulin resistance (62), and insulin suppresses PDK4 expression (63). Therefore, during ER stress, defective insulin signaling might induce PDK4 expression. In accordance with these previous findings, we observed that Tm increased activity of the PDK4 promoter ( Figure 7B). Moreover, overexpression of ATF6a significantly induced PDK4 expression, which was significantly attenuated by knockdown of endogenous ERRg. The findings show that ATF6a induces ERRg to activate ERRg target genes ( Figure 7A). Our results that link transactivation of ATF6a to ERRg gene expression, along previous evidence supporting an interconnection between ER stress and nuclear receptor signaling (35,40), raise the possibility (discussed further below) that ATF6a may serve as a key mediator in ER stressinduced ERRg gene expression.
CREBH, an ER-membrane-bound transcription factor that is similar to ATF6a in structure and mode of activation, is regulated by nuclear receptor PPARa (40), HNF4a (41) and GR (64). These findings along with a previous report (35) suggest a probable bidirectional regulatory pathway between transcription factors that regulate the mammalian ER stress response and nuclear receptors. To our surprise, overexpression of ERRg significantly increased ATF6a gene expression. This result demonstrates a cross talk between ERRg and ATF6a. In accordance with our observations ( Figure 4A-C), the Tm-mediated increase in ATF6a protein was significantly decreased on either GSK5182 treatment or knockdown of endogenous ERRg, supporting an apparent regulatory role of ERRg for ATF6a expression during ER stress. It was previously reported that ERRg binds to the sequence motif AGGTCA (17). A closer investigation of the ATF6a promoter revealed the existence of a probable ERRRE. Mutation of this site blocked promoter activation by Tm, demonstrating the importance of the ERRRE in ATF6a promoter function. Chip assay further confirmed evidence for ERRg binding to the ATF6a promoter upon ER stress. PGC1a is closely associated with the transcriptional activity of ERRg (3). Several lines of evidence indicate that transcriptional regulation of PDK4 expression by PGC1a is mediated by ERRa or ERRg (61). In accordance with these previous reports, ER stress induced occupancy of PGC1a at the ERRRE in the ATF6a promoter, indicating PGC1a is a coactivator for ERRg. Chromatin remodeling was also identified as a significant increase in the acetylation of histone H3 and H4 at the ATF6a promoter on Tm treatment or on ERRg overexpression. The findings demonstrate chromatin remodeling during gene transcription as reported previously (57,60). BiP/GRP78 is a Ca 2+ -dependent ER chaperone that is induced by ER stress (21). Proteolytic cleavage and activation of ATF6a in response to ER stress upregulates GRP78 transcription (58). GRP78 participates in protein folding, transport and degradation upon ER stresses (65). A significant increase in GRP78 expression was observed upon ERRg overexpression, and this increase was significantly attenuated by knockdown of endogenous ATF6a, suggesting that ERRg induces ATF6a to activate transcription of the ATF6a target gene GRP78 ( Figure 7C). Overall, our current findings reveal a novel molecular mechanism of reciprocal transcriptional activation used by ERRg and ATF6a and provide evidence for a new role of both ERRg and ATF6a in working in concert with PGC1a to affect chromatin remodeling in target genes. Both ATF6a and ERRg are implicated in numerous biological events. Previously, induction of liver steatosis and lipid droplet formation in ATF6a-knockout mice was reported (66,67). ATF6a is required for adipogenesis (68). ATF6a-null mice were reported to be glucose intolerant owing to pancreatic b-cell failure on a high-fat diet (69). The unfolded protein response mediates adaptation to exercise in skeletal muscle through a PGC-1a/ATF6a complex (22). ERRg also plays critical role in cellular physiology. ERRg modulates energy metabolism target genes in human trophoblasts (70). Previously our laboratory reported that ERRg regulates hepatic gluconeogenesis (6) and LIPIN1, a gene involved in lipid metabolism (7). Glucosamine-induced ER stress causes insulin resistance in both human and rat skeletal muscle and impairs GLUT4 production and insulin-induced glucose uptake via an ATF6a-dependent decrease of the GLUT4 regulators MEF2A and PGC1a. Inhibition of ATF6a is sufficient to completely prevent glucosamineinduced inhibition of GLUT4, MEF2A and PGC1a in skeletal muscle cells (71). Moreover, viral infection induces ER stress, and hence induces both GRP78/BiP and ATF6a to facilitate protein folding during viral maturation (72). For example, the final assembly of rotavirus particles that cause severe diarrhea among infants and young children takes place in the ER. Protein disulfide isomerase, GRP78, calnexin and calreticulin are protein chaperones of the ER that are involved in the quality control of rotavirus morphogenesis. Cells with reduced expression of these chaperones exhibit defective maturation of rotavirus (73). All these chaperones are positively regulated at the transcriptional level by ATF6a (74,75). Therefore, inhibition of ATF6a might be an effective therapeutic target against rotavirus infection.
Here, we provide a previously unknown mechanism of regulatory pathway for ERRg and ATF6a. We hypothesize that there are two parts of this regulatory loop ( Figure 7E). In response to ER stress, the expression of both ERRg and ATF6a increases, although ERRg transcription is induced earlier than ATF6a ( Figure 4B). On one hand, ATF6a along with coactivator mediates the increase in ERRg transcription; on the other hand, ERRg in association with coactivator mediates the increase in ATF6a transcription. Overall, our current findings provide insight into a novel chromatin remodeling strategy used by ERRg and ATF6a. From a broader perspective, it may be relevant for different important biological processes in which both ERRg and ATF6a play well-documented key regulatory roles.