Peroxisome Proliferator-activated Receptor Coactivator-1α (PGC-1α) Coactivates the Cardiac-enriched Nuclear Receptors Estrogen-related Receptor-α and -γ

The transcriptional coactivator PPARγ coactivator-1α (PGC-1α) has been characterized as a broad regulator of cellular energy metabolism. Although PGC-1α functions through many transcription factors, the PGC-1α partners identified to date are unlikely to account for all of its biologic actions. The orphan nuclear receptor estrogen-related receptor α (ERRα) was identified in a yeast two-hybrid screen of a cardiac cDNA library as a novel PGC-1α-binding protein. ERRα was implicated previously in regulating the gene encoding medium-chain acyl-CoA dehydrogenase (MCAD), which catalyzes the initial step in mitochondrial fatty acid oxidation. The cardiac perinatal expression pattern of ERRα paralleled that of PGC-1α and MCAD. Adenoviral-mediated ERRα overexpression in primary neonatal cardiac mycoytes induced endogenous MCAD expression. Furthermore, PGC-1α enhanced the transactivation of reporter plasmids containing an estrogen response element or the MCAD gene promoter by ERRα and the related isoform ERRγ. In vitro binding experiments demonstrated that ERRα interacts with PGC-1α via its activation function-2 homology region. Mutagenesis studies revealed that the LXXLL motif at amino acid position 142–146 of PGC-1α (L2), necessary for PGC-1α interactions with other nuclear receptors, is not required for the PGC-1α·ERRα interaction. Rather, ERRα binds PGC-1α primarily through a Leu-rich motif at amino acids 209–213 (Leu-3) and utilizes additional LXXLL-containing domains as accessory binding sites. Thus, the PGC-1α·ERRα interaction is distinct from that of other nuclear receptor PGC-1α partners, including PPARα, hepatocyte nuclear factor-4α, and estrogen receptor α. These results identify ERRα and ERRγ as novel PGC-1α interacting proteins, implicate ERR isoforms in the regulation of mitochondrial energy metabolism, and suggest a potential mechanism whereby PGC-1α selectively binds transcription factor partners.

Cellular energy production is tightly linked to metabolic demand, which is, in turn, dictated by diverse developmental, physiologic, and environmental conditions. The capacity for cellular ATP production is controlled, in part, by the expression levels of nuclear genes involved in mitochondrial oxidative metabolism. Thus, tight regulation of cellular energy metabolism necessitates transduction of diverse signals related to cellular energy demands to the nucleus. Although numerous factors involved in the transcriptional regulation of metabolic gene expression have been identified, the precise pathways involved in the physiologic control of cellular energy metabolism have not been delineated. The recent discovery of PPAR␥ coactivator-1␣ (PGC-1␣), 1 PGC-1␤, and the PGC-1-related protein, a family of inducible transcriptional coactivators responsive to selective physiological stimuli, have provided new insights into the link between extracellular events and the regulation of genes involved in energy metabolism. PGC-1␣, the first member of this novel coactivator family to be identified, was initially characterized as a key regulator of thermogenesis in brown adipose tissue (BAT) and skeletal muscle via its coactivation of the adipose-enriched nuclear receptor, PPAR␥ (1,2). Subsequent studies have revealed a broader role for PGC-1␣ in a variety of cellular energy metabolic processes including mitochondrial biogenesis, mitochondrial fatty acid oxidation (FAO), and gluconeogenesis (2)(3)(4)(5)(6). The function of PGC-1␤ and PGC-1-related protein remain to be defined.
PGC-1␣ is unique from the p160 and p300/cAMP response element-binding protein-binding protein classes of transcriptional coactivators in its tissue-restricted expression pattern, its developmental regulation, and its inducibility by specific physiological stimuli. PGC-1␣ is enriched in tissues reliant on oxidative metabolism for ATP generation (heart, skeletal muscle) or heat (BAT) but is also expressed in liver, brain, and kidney (1). Immediately after birth, PGC-1␣ expression increases in heart coincident with a shift from reliance on glycolysis to mitochondrial FAO as the chief energy source in the adult myocardium (4). PGC-1␣ expression is induced in adult skeletal muscle, BAT, and heart in response to stimuli that increase energy demands. For example, cold exposure leads to a rapid induction of PGC-1␣ gene expression in BAT (1,7). In addition, fasting and short-term exercise induces PGC-1␣ gene expression in heart and skeletal muscle, respectively (4,8,9). Recent studies have shown that PGC-1␣ protein is phosphorylated in response to cytokine stimulation of the p38 mitogenactivated protein kinase pathway resulting in stabilization of the protein (10). Furthermore, we have shown that activation of the p38 pathway enhances ligand-dependent PGC-1␣ coactivation of PPAR␣ (11). Finally, Knutti et al. (12) demonstrated p38-mediated activation of PGC-1␣ via release of a repressor.
The results of recent gain-of-function studies have demonstrated that PGC-1␣ serves as a global regulator of mitochondrial metabolic capacity. Spiegelman and co-workers (2) have shown that overexpression of PGC-1␣ in myogenic cells activates the mitochondrial biogenic program leading to an increase in mitochondrial number and respiration rates. Our laboratory has shown that forced expression of PGC-1␣ in primary cardiac myocytes and in hearts of transgenic mice leads to the transcriptional activation of genes encoding mitchondrial FAO enzymes, such as medium-chain acyl-CoA dehydrogenase (MCAD), and triggers a robust increase in mitochondrial cellular volume density (4). Collectively, these recent studies suggest that PGC-1␣ serves to transduce stimuli linked to physiologic demands to the transcriptional control of mitochondrial functional capacity.
The regulatory effects of PGC-1␣ are thought to be mediated primarily by its ability to interact with and coactivate numerous nuclear receptors, as well as non-nuclear receptor transcription factors, based on in vitro and cell culture studies (1-3, 6, 12-14). The effects of PGC-1␣ on mitochondrial FAO enzyme gene expression occur, at least in part, through its activation of the nuclear receptor PPAR␣ (3,4). The mitochondrial biogenesis response involves the transcription factors nuclear respiratory factor-1 and nuclear respiratory factor-2 (2). However, not all of the downstream effects of PGC-1␣ on cellular energy metabolism have been ascribed to particular PGC-1␣ transcription factor partners. The role of PGC-1␣ as a "master" regulator of cellular energy metabolism is likely mediated by multiple transcription factors, some of which could be novel. In addition, the mechanisms involved in partner selection by PGC-1␣ in the context of coexpressed transcription factors are also unknown.
In an attempt to identify PGC-1␣ interacting proteins relevant to the postnatal heart, we performed a two-hybrid screen of an adult human heart cDNA library in yeast using PGC-1␣ as "bait." We focused on the adult mammalian heart because of its extraordinary capacity to match mitochondrial energy production with high demands. The orphan nuclear receptor, estrogen-related receptor ␣ (ERR␣; NR3B1), was identified as a novel PGC-1␣ interacting protein. PGC-1␣ enhanced the transcriptional activities of ERR␣ and the related isoform ERR␥. The PGC-1␣⅐ERR␣ complex was shown to directly activate the MCAD promoter, a gene implicated previously as an ERR␣ target. Functional assays and in vitro binding studies demonstrated that ERR␣ binds PGC-1␣ via a novel set of leucine-rich domains compared with other characterized nuclear receptor PGC-1␣ partners. These results identify a novel PGC-1␣ target in heart. In addition, our results suggest that the utilization of distinct binding interfaces within PGC-1␣ provides one mechanism whereby this versatile coactivator may differentially activate multiple downstream effectors in diverse cellular and physiologic contexts.

EXPERIMENTAL PROCEDURES
Mammalian Cell Culture and Transient Transfections CV1, 293, and HepG2 cells were maintained at 37°C and 5% CO 2 in Dulbecco's modified Eagle's medium/10% fetal calf serum. For experiments requiring estradiol, HepG2 cells were plated the day before transfection in Earle's minimum essential medium without phenol red/ 10% stripped fetal calf serum. Transient transfections were performed by the calcium phosphate coprecipitation method as described (15). Reporter plasmids (4 g/ml) were cotransfected with Rous sarcoma virus-␤-galactosidase (0.5 g/ml), expressing the ␤-galactosidase gene driven by the Rous sarcoma virus promoter, to control for transfection efficiency. For cotransfection experiments, mammalian expression vectors (see below) for nuclear receptors, PGC-1␣ constructs, or the corresponding empty vectors were used. Ligands were added immediately following transfection protocol at the concentrations indicated in the figure legends, and cells were collected and assayed 48 h later. Luciferase and ␤-galactosidase activities were measured as described (16).
Ventricular cardiac myocytes were prepared from 1-day-old Harlan Sprague-Dawley rats as described (15). After 24 h cells were infected with adenovirus expressing GFP (Ad-GFP) or ERR␣ (Ad-ERR␣) driven by a cytomegalovirus promoter. The latter construct also expresses GFP from an independent promoter. Infection rate of 90 -95% was achieved by 18 h as assessed by quantitation of GFP-expressing cells using fluorescence microscopy. Whole cell protein extracts were prepared from cells 72 h post-infection. The adenoviral construct, Ad-ERR␣, was constructed by subcloning a SalI/NotI fragment from the pBK-ERR␣ (generously provided by C. Teng) construct containing the full-length human ERR␣ cDNA encoding amino acids 1-422 into the pAd-Trackcytomegalovirus vector. Recombination and propagation of adenovirus expressing ERR␣ was performed as described (17).

Plasmid Constructs
Reporter Plasmids-The MCAD promoter-luciferase plasmids have been described (15,18). The Vit 2 P36.Luc, containing two copies of the vitellogenin estrogen responsive element upstream of the prolactin minimal promoter, was generously provided by S. Adler (Southern Illinois University). The (UAS) 3 .TK.Luc was a gift from D. Moore (Baylor College of Medicine). The (APOCIII) 2 .TK.Luc was constructed by ligation of annealed oligonucleotides (sense strand 5Ј-GATCCTCATC-TCCACTGGTCAGCAGGTGACCTTTGCCCAGCGCCCTGGGA-3Ј) into the BamHI/BglII site upstream of the thymidine kinase promoter of the pGL2-TK.Luc reporter plasmid. The sequence is based on the HNF-4 response element contained in the human apolipoprotein CIII gene promoter (19).
Mammalian Expression Plasmids-The mammalian expression vector pcDNA3.1-Myc/His.PGC-1␣ has been described elsewhere (3). Sitedirected mutations in PGC-1␣ were introduced by a PCR-based strategy (Quickchange Mutagenesis kit; Stratagene) using the Myc/His.PGC-1␣ as template for single-site mutants. The double-and triple-site mutations were made using the single-and double-mutant templates, respectively. The PGC-1␣ deletion series, PGC 338 , PGC 285 , and PGC 120 , have been described (3). Additional FLAG-tagged PGC-1␣ deletion constructs were generated by PCR introducing an in-frame BglII at the start codon and a stop codon at 273, 213, 208, or 191. PCR fragments were then subcloned (BglII/XhoI) into cytomegalovirus promoter-Tag1 (Stratagene) to fuse a cassette encoding a FLAG epitope to 5Ј end of the PGC-1␣ sequence. The PGC-1␣ constructs were subsequently cloned into the NotI site of the pcDNA3.1 mammalian vector for cotransfection studies. The pcDNA3.1-FLAG-ERR␣ full-length and deletion constructs were generated by the same procedure described above for PGC-1␣ using PCR to introduce a stop codon at 403, 359, or 209. The Gal4-ERR␣ and Gal4-ERR␣ 403 were generated by subcloning the BamHI fragment from the corresponding pcDNA3.1 construct into the pCMX-Gal4 plasmid (a kind gift from D. Moore). The Gal4-PPAR␣ has been described (3). The pSG5-hemagglutinin-ERR␥, the pBK-Rous sarcoma virus-ER␣, and the pMT-HNF-4 expression vectors were kind gifts from M. Stallcup (University of Southern California), S. Adler, and J. Ladias (Harvard Medical School), respectively.

Northern Blot Analysis
Total cellular RNA isolation and blotting was performed as described (16). Blots were hybridized with radiolabeled probes derived from the following cDNA mouse clones: MCAD, ERR␥, PPAR␣, and PGC-1␣. In addition, human ERR␣, rat M-CPT I, and universal actin probes were used.

Immunoblotting
Protein extracts were resolved by SDS-PAGE (7.5%). Transfer and detection were performed as described (20). Immunodetection of ERR␣ and COUP-TF were performed using polyclonal anti-ERR␣ or anti-COUP-TF antibodies generously provided by V. Giguere (McGill University) and M.-J. Tsai (Baylor College of Medicine), respectively. MCAD was detected using the anti-MCAD antibody described previously.

GST Pull Down Assays
In vitro protein-protein interaction assays have been described previously (3). 35 S-Labeled proteins were synthesized in the TNT T7 quickcoupled in vitro transcription/translation system (Promega). In pull down reactions, 50 l of a 50% slurry of GST fusion protein bound to glutathione-Sepharose was incubated with 10 l of 35 S-labeled protein in 500 l of binding buffer (20 mM Tris, 7.5, 100 mM KCl, 0.1 mM EDTA, 0.05% Nonidet P-40, 10% glycerol, 1 mg/ml bovine serum albumin, 0.5 mM phenylmethylsulfonyl fluoride, and 1ϫ Complete (Roche Molecular Biochemicals)) for 1 h at 4°C. The beads were pelleted and washed five times with cold binding buffer. SDS-PAGE reducing buffer was added to the beads, samples were boiled for 5 min, and the eluted proteins were analyzed by SDS-PAGE. The gels were fixed and dried, and band intensities were quantified by phosphorimage analysis using the Bio-Rad GS 525 molecular imaging system.

Identification of ERR␣ as a PGC-1␣ Interacting Partner-
Using a yeast two-hybrid approach, a human adult cardiac cDNA library was screened to identify proteins that interacted with the region of PGC-1␣ encompassing amino acids (aa) 122-403. This region of the PGC-1␣ protein contains all of the domains known to mediate interactions between PGC-1␣ and transcription factors partners identified to date (21). In a screen of 4 ϫ 10 7 transformants, ϳ40% of the clones encoded the orphan nuclear receptor, human ERR␣. Importantly, two groups of unique clones encoding human ERR␣ were identified (Fig. 1A). Group 1 encodes aa 206 -422 of ERR␣, which in-cludes the ligand binding domain (LBD). Group 2 encodes aa 121-422, which contains a portion of the DNA binding domain (DBD) and the hinge region, in addition to the LBD.
To demonstrate that PGC-1␣ interacts directly with ERR␣, in vitro GST pull down assays were performed using a GST-ERR␣ fusion protein produced in bacteria and full-length 35 S-PGC-1␣ generated by in vitro translation. Conversely, a PGC-1␣-GST fusion construct, containing aa 1-338 (GST-PGC 338 ), was incubated with 35 S-ERR␣. An interaction was observed between PGC-1␣ and ERR␣ in both configurations, whereas no significant binding occurred between the 35 S-labeled proteins and GST alone, confirming the specificity of the interaction (Fig. 1B). Given the structural homology between ERR␣ and the related isoform, ERR␥, and the observation that both are cardiac-enriched (22)(23)(24)(25), we also examined whether PGC-1␣ could bind ERR␥. Using GST-PGC 338 , a specific interaction between PGC-1␣ and ERR␥ was demonstrated, thereby extending our findings to encompass multiple ERR isoforms (Fig. 1B). These results demonstrate that PGC-1␣ physically interacts with ERR␣ and ERR␥ and that the interacting region is contained within amino acids 1-338 of the PGC-1␣ molecule.
The Expression of ERR␣ Parallels That of PGC-1␣ and Several of Its Target Genes during Cardiac Development-The identification of ERR␣ as a PGC-1␣ interacting protein was intriguing for several reasons. First, we and others (22,23) have identified a likely role for ERR␣ in the transcriptional regulation of the gene encoding MCAD, which catalyzes the initial step in the mitochondrial FAO spiral. PGC-1␣ coactivates genes involved in mitochondrial FAO and biogenesis (2)(3)(4). Second, the expression of ERR␣ is enriched in heart and BAT, a pattern similar to that of PGC-1␣ (23). As an initial step to determine whether the ERR␣ interaction with PGC-1␣ was of potential biological relevance, the expression patterns of each were compared in hearts from different developmental stages with varying degrees of mitochondrial oxidative capacity. Consistent with its role in the transcriptional regulation of mitochondrial oxidative metabolism, cardiac PGC-1␣ expression is induced following birth, during a period of intense mitochondrial biogenesis, coincident with the shift toward reliance on FAO as the chief source of energy (4). Therefore, the cardiac perinatal-to-adult expression profile of the ERR␣ gene was compared with that of PGC-1␣. Northern analysis revealed that the ERR␣ transcript was in relatively low abundance before birth but displayed a postnatal day 1 spike followed by a gradual increase to its maximum expression levels in the adult mouse heart (Fig. 2A). This pattern paralleled that of PGC-1␣, the putative ERR␣ target gene MCAD, and PPAR␣, a known regulator of postnatal cardiac energy metabolism (16,18). Levels of ERR␣ protein correlated with mRNA abundance during the cardiac metabolic transition (Fig. 2B). In contrast, expression of the COUP-TF, a known antagonist of nuclear receptor signaling and MCAD expression (26,27), was expressed in a reciprocal pattern, falling to low levels in the adult heart. These data demonstrate that the expression of ERR␣ parallels that of PGC-1␣ and downstream targets involved in mitochondrial energy production during perinatal cardiac development.
Forced Overexpression of ERR␣ Induces MCAD Gene Expression-Because ERR␣ and FAO enzyme genes appeared to be coordinately expressed in heart, we next wanted to determine whether ERR␣ could directly regulate the expression of the endogenous MCAD gene. For these studies, ERR␣ was overexpressed in cardiac myocytes using adenoviral constructs expressing either GFP alone (to monitor infection efficiency) or both GFP and ERR␣ from independent promoters. Analysis of whole cell extracts by immunoblotting revealed that MCAD expression was significantly induced in the ERR␣-infected cells The transcription factor binding domain (aa 122-403) was used as "bait" in a yeast two-hybrid screen of a human adult cardiac cDNA library. Bottom, schematic of the hERR␣ protein, highlighting the conserved DBD and LBD. Two distinct populations of ERR␣ clones, both containing the LBD, were isolated in the screen. The numbers in parentheses indicated the number of clones isolated within each group. B, bacterially expressed GST-PGC338 (top) or GST-ERR␣ (middle) was used in GST pull down assays with 35 S-labeled full-length ERR␣ or PGC-1␣, respectively. Bottom, binding of the ERR␥ isoform with PGC-1␣ was assessed using GST-PGC338 with 35 S-labeled ERR␥. GST alone was used to control for nonspecific binding of 35 S-labeled proteins. 20% of the radiolabeled protein used in the binding reactions (20% Input) was run in parallel lanes for comparison. compared with GFP controls (Fig. 2C). No further increase in MCAD expression was observed upon coexpression of ERR␣ and PGC-1␣ (data not shown) likely because of the presence of endogenous ERR␣ coactivators in these cells. These results strongly suggest that the parallel expression pattern of ERR␣ with FAO enzyme genes reflects a biologically relevant role for ERR␣ in the regulation of mitochondrial fatty acid metabolism.
The PGC-1␣⅐ERR Complex Directly Activates Transcription through a Consensus ERE and through the NRRE-1 of the MCAD Gene Promoter-To determine whether PGC-1␣ can function as a coactivator of ERR␣ and ERR␥, transient transfections were first performed using the ERR␣ responsive Vit 2 P36.Luc reporter construct, containing two copies of the estrogen receptor response element derived from the vitellogenin promoter. No effect on reporter activity was observed upon cotransfection with either PGC-1␣ or ERR␣ alone (Fig. 3A). However, a marked activation of Vit 2 P36.Luc activity (Ͼ22fold) was observed when ERR␣ and PGC-1␣ were coexpressed.
Based on the results of previous studies and the data shown above, ERR␣ is predicted to regulate MCAD gene transcription in BAT and heart (22,23). However, despite the observation that ERR␣ binds a complex NRRE-1 within the MCAD gene promoter, transient transfection studies in a variety of mammalian cell lines have failed to reveal a direct regulatory effect of ERR␣ on MCAD gene transcription. Indeed, studies of ERR␣-mediated activation of most target genes identified to date have demonstrated relatively modest transactivation potency suggesting that an ERR␣ coactivator is absent or is expressed at low levels in the cell lines used to evaluate ERR␣ transactivating properties. Given that most mammalian cell lines are devoid of PGC-1␣, we hypothesized that PGC-1␣ might be a relevant coactivator for the ERR␣-mediated regulation of MCAD gene transcription. Accordingly, the regulatory effects of ERR␣ on the MCAD promoter were evaluated in the absence or presence of overexpressed PGC-1␣ (Fig. 3B). For these experiments, a luciferase reporter construct driven by a segment of the human MCAD gene promoter containing the ERR␣ binding site, NRRE-1, was used (376.MCAD.Luc). As expected, 376.MCAD.Luc activity was not affected by either factor alone. However, overexpression of both ERR␣ and PGC-1␣ induced a synergistic activation (Ͼ8-fold) of Ϫ376.MCAD.Luc (Fig. 3B). When the same experiment was performed with an MCAD gene promoter-reporter construct containing an inactivated NRRE-1 (NRREmut.MCAD.Luc) (15), the PGC-1␣⅐ERR␣-mediated activation was abolished (data not shown). Collectively, these results demonstrate that PGC-1␣ coactivates ERR␣ and ERR␥ and that the PGC-1␣⅐ERR␣ complex transcriptionally regulates at least one gene target involved in cardiac mitochondrial energy metabolism.
The PGC-1␣⅐ERR␣ Interaction Maps to the AF-2 Region of ERR␣-Generally, nuclear receptor-coactivator interactions involve domains within the receptor LBD. Indeed, the smallest ERR␣ clone isolated in our yeast two-hybrid screen contained only the LBD (Fig. 1A, Group 1). Therefore, to evaluate the role of the LBD in the PGC-1␣⅐ERR␣ interaction, a series of ERR␣ C-terminal deletion mutants were generated and evaluated using GST pull downs and transient transfection assays. As shown above, full-length ERR␣ bound strongly with GST-PGC 338 . In contrast, deletion of the C-terminal 19 aa, containing the consensus AF-2 domain, resulted in an ϳ80% reduction in binding activity (Fig. 4A). No further decrease was observed when subsequent deletions were made to residue 209, removing all but the N-terminal 12 aa of the LBD. Considering the clones isolated in the yeast two-hybrid library screen (Fig. 1A), the 209 -422 region of the LBD is likely sufficient to interact with ERR␣ (Group 1 clone). The small degree of residual binding observed in the ERR 209 mutant suggests that a low affinity binding region exists within residues 1-209 of the ERR␣ protein, possibly in the hinge region, which was contained in half of the clones isolated in the library screen.
Given that deletion of the AF-2 region in ERR 403 had the most profound influence on binding, the effect of an AF-2 deletion on PGC-1␣-mediated ERR␣ coactivation was assessed using a Gal4 system in which the entire ERR␣ protein was fused to the yeast Gal4 DNA binding domain (Gal4-DBD) (Fig.  4B). Using this system, PGC-1␣⅐ERR␣ interactions are assessed in mammalian cells without potential background from endogenous PGC-1␣ interacting partners. Compared with the Gal4-DBD control, Gal4-ERR␣ repressed activity of the (UAS) 3 .TK.Luc reporter by 65%. Addition of PGC-1␣ coactivated Gal4-ERR␣ modestly (3-fold), essentially raising activity back to control levels. Interestingly, deletion of the AF-2 region FIG. 2. PGC-1␣ and ERR␣ are coordinately expressed with FAO enzyme genes, and ERR␣ induces endogenous MCAD expression. A, the results of Northern analyses performed with total RNA (15 g) from hearts of mice at different stages of development. B, immunoblot analyses using ␣-ERR␣ and ␣-COUP-TF antibodies with nuclear protein extracts prepared from hearts of mice at the indicated developmental stages. ed, embryonic day; pd, post-natal day. C, autoradiograph representing immunoblot analyses of whole cell extracts (WCE) from primary neonatal cardiac myocytes that were either uninfected (C) or infected with adenovirus expression GFP alone (GFP) or GFP and ERR␣ (ERR␣). The blot was hybridized sequentially with antibodies to ERR␣ or MCAD as indicated.
resulted in a significant increase in baseline activity 2.5-fold above the control (Gal4-DBD) levels. No further increase in activity was observed upon cotransfection of PGC-1␣. These data support the GST pull down results localizing the major PGC-1␣ interaction domain to the consensus AF-2 region of ERR␣. The results also suggest that the coactivating effects of PGC-1␣ on ERR␣ in this assay system involves, at least in part, displacement of a repressor bound to the AF-2 region of the ERR␣ molecule.
The ERR␣ Interacting Domain Maps to a Region of the PGC-1␣ Protein Containing an Inverted LXXLL Motif-Next, in vitro mapping experiments were performed to identify the PGC-1␣ domain essential for its interaction with ERR␣. A number of C-terminal truncated PGC-1␣ mutants were analyzed for their ability to interact with GST-ERR␣ (Fig. 5A). No decrease in binding affinity was observed with either PGC 338 or PGC 285 compared with the full-length protein. However, a significant drop was observed when the region of PGC-1␣ encompassing aa 191-285 was removed. This region contains L3, one of three consensus LXXLL motifs (L1, L2, L3; see Fig. 1A and Fig. 5A) within the PGC-1␣ protein (12). The LXXLL motifs of coactivator proteins are predicted to be embedded within short ␣-helical stretches. Typically one or more LXXLL ␣-helical domains within a coactivator protein are involved in AF-2-dependent interactions with nuclear receptors. The LXXLL motifs of PGC-1␣ are contained within the region encompassing aa 86 -213 of the protein (Fig. 1A). To date, only L2 has been shown to be necessary for PGC-1␣/nuclear receptor interactions (3,6,(12)(13)(14)28). Surprisingly, deletion of the region encompassing aa 191-285 of PGC-1␣, which contains L3 (aa 209 -213), resulted in a dramatic loss of ERR␣ binding (Fig.  5A). Deletion of PGC-1␣ amino acids 191-120, which contains L2, abolished the residual low affinity interaction.
The mutant PGC-1␣ proteins were evaluated in functional assays using Gal4-ERR␣ as a target in transient transfections.
As shown in Fig. 5B, the relative activity of each PGC-1␣ truncation mutant paralleled its binding efficiency. Specifically, full-length PGC-1␣ displayed 4-fold activation, whereas the PGC 338 and PGC 285 mutants activated Gal4-ERR␣ to a greater degree than wild-type PGC-1␣. This enhanced activity of truncated PGC-1␣ proteins has been described previously (1,3) and is thought to be because of removal of an autoinhibitory domain located in the C-terminus of PGC-1␣. As predicted by the binding studies, subsequent removal of the 285 to 191 region eliminated PGC-1␣-mediated coactivation of Gal4-ERR␣ (Fig. 5B).
The functional mapping and binding studies suggested that the region of PGC-1␣ containing L3 was necessary for its interaction with ERR␣. These findings were surprising, because the interaction between PGC-1␣ and other nuclear receptors characterized to date have mapped to the L2 motif at aa 142-146 (Fig. 5C). A second point of interest is that two p38 mitogen-activated protein kinase (p38-MAPK) phosphorylation sites are located between residues 285 and 191 at Thr-262 and Ser-265 of PGC-1␣ (see Fig. 5C and Refs. 10 and 12). Therefore, the activities of additional PGC-1␣ truncation mutants were analyzed to determine whether the p38-MAPK sites or L3 influenced coactivation ERR␣ by PGC-1␣. As observed with PGC 285 , the PGC 273 and PGC 213 mutants displayed strong activation of Gal4-ERR␣ (Fig. 5C). This suggests that the phosphorylation sites at 262/265 are not essential for the functional interaction of PGC-1␣ with ERR␣. Further removal of the five residues corresponding to the L3 motif, an inverted LXXLL, resulted in nearly a complete loss of PGC-1␣ activity with ERR␣ (Fig. 5C). Parallel analyses of these mutant PGC-1␣ constructs performed with the nuclear receptors, PPAR␣ and ER␣, which have been shown to interact with PGC-1␣ via L2, demonstrated that all of the mutants were fully active with these receptors (data not shown). The binding activity of the PGC-1␣ mutants with GST-ERR␣ paralleled their activity in the functional assays, displaying a significant reduction in binding affinity for ERR␣ upon deletion of the L3 motif (Fig.  5D). Collectively, these results indicate that PGC-1␣-mediated coactivation of ERR␣ involves L3 but not the p38-MAPK phosphorylation sites. The results identify a unique nuclear receptor interaction site within the PGC-1␣ molecule.
The PGC-1␣⅐ERR␣ Interaction Interface Is Distinct from That of Other Nuclear Receptor PGC-1␣ Partners-The results shown above suggest that the PGC-1␣⅐ERR␣ interaction involves the L3 site, a novel PGC-1␣/nuclear receptor interface. However, a cooperative role for the other LXXLL motifs was not excluded in the above analysis. Furthermore, it is important to evaluate the role of the L3 site in the context of the full-length PGC-1␣ protein. Accordingly, the leucine residues in L1, L2, and L3 were substituted with phenylalanines in the full-length PGC-1␣ protein (Fig. 6A). The binding and functional activities of these PGC-1␣ mutants were assessed with ERR␣ (Fig. 6). As predicted, the single L1 (mL1) and L2 (mL2) PGC-1␣ mutants bound GST-ERR␣ with the same affinity as wild-type PGC-1␣ and were fully active with ERR␣ on the Vit 2 P36.Luc reporter in 293 cells (Fig. 6). Unexpectedly, ERR␣ coactivation by the PGC-1␣ mutant harboring substitutions in L3 remained intact (Fig. 6B). However, disruption of L3, in combination with either of the other two sites, markedly re-

FIG. 4. PGC-1␣ interacts with ERR␣ through the consensus AF-2 domain.
A, top, schematic of ERR␣ C-terminal truncation mutants to used to localize the PGC-1␣ interacting domain within the LBD. All constructs begin with aa 1; mutant designations indicate their C-terminal ends. Bottom, GST pull downs were performed as described in Fig. 1B using GST alone or GST-PGC338 with the radiolabeled mutant proteins. wt, wild-type. B, transient cotransfections using the (UAS) 3 .TK.Luc heterologous reporter (4 g) were performed in CV1 cells with 0.5 g of either the Gal4-ERR␣ or Gal4-ERR␣ 403 fusion construct. Empty vector (Ϫ) or PGC-1␣ expression vector (ϩ) were cotransfected with Gal4 constructs, as indicated. Bars represent mean (Ϯ S.E.) ␤-galactosidase-corrected RLU, normalized (ϭ 1.0) to the activity of the reporter cotransfected with Gal4-DBD. The Gal4-DBD control was unaffected by PGC-1␣ cotransfection (data not shown). Data represent three trials performed in triplicate. The asterisk indicates the mean values are significantly different (p Ͻ 0.05) from the Gal4-ERR␣ value. n.s., not significant.
duced the ability of PGC-1␣ to bind and coactivate ERR␣. These data suggest that although the L3 motif is sufficient to mediate the functional interaction of PGC-1␣ with ERR␣ (PGCmL1/2 mutant displays full activity) and is likely the primary binding site for ERR␣, L3 is not absolutely required when multiple alternate LXXLL motifs are available on PGC-1␣. Hence, the alternate sites (L1 and L2) may serve a low affinity accessory function in the PGC-1␣ interaction with ERR␣.
The results shown above suggest that the binding of certain nuclear receptors with PGC-1␣ involves distinct combinations of Leu-rich (LXXLL) interacting motifs. To explore this possibility further, the LXXLL motif mutants were also tested for their ability to coactivate ERR␥, ER␣, and HNF-4␣. Interestingly, the pattern of ERR␥ coactivation by the PGC-1␣ mutants was somewhat different from that of ERR␣ (Fig. 6B). Whereas ERR␣ appeared to primarily interact with the L3 motif of PGC-1␣, ERR␥ displayed a dependence on both L2 and L3. Only when both sites were simultaneously eliminated was PGC-1␣ activation of ERR␥ abolished indicating that ERR␥ binds either L2 or L3 interchangeably. ER␣ and HNF-4␣ are PGC-1␣ targets reported previously to interact with PGC-1␣ via the L2 motif. Consistent with their L2 dependence, disruption of the L2 motif alone or in any combination with the other sites completely abrogated PGC-1␣ enhancement of receptormediated activation (Fig. 6B). Importantly, the mL1/3 mutant, which was inactive with ERR␣, displayed full activity with the L2 binding class of receptors. Collectively, these data confirm that ERR␣ and ERR␥ define distinct classes among nuclear receptor PGC-1␣ partners with regard to binding site specificity.

DISCUSSION
The data presented here provide several lines of evidence that the PGC-1␣⅐ERR␣ interaction represents a functional transcriptional complex involved in the regulation of cardiac and skeletal muscle metabolism. First, multiple independent ERR␣ clones were isolated as PGC-1␣ interacting proteins using a two-hybrid screen. The interaction of PGC-1␣ with ERR␣ and -␥ isoforms was verified using in vitro binding assays. Specific binding domains were mapped within the PGC-1␣⅐ERR␣ binding interface on both ERR␣ and PGC-1␣, thereby indicating the specificity of the interaction. Interestingly, the ERR␣ binding domain within the PGC-1␣ molecule is unique compared with other transcription factor binding sites. Second, PGC-1␣ coactivated both ERR␣ and ERR␥ isoforms in heterologous promoter-reporter assays. Third, involvement of PGC-1␣⅐ERR␣ in regulating cellular metabolism is suggested by parallel tissue-specific and developmental expression profiles of ERR isoforms, PGC-1␣, and mitochondrial FAO enzyme C-terminal ends. Binding assays were performed as described in the legend for Fig. 1B  Mapping studies of the PGC-1␣⅐ERR␣ binding interface revealed a novel LXXLL-type nuclear receptor binding motif on PGC-1␣. Conserved LXXLL motifs within coactivator proteins have been shown to mediate AF-2 domain-dependent interactions with nuclear receptors. The Leu-rich motif adopts an ␣-helical conformation that fits into a hydrophobic binding pocket formed by several helices of the receptor LBD. PGC-1␣ contains three potential LXXLL motifs (L1-L3) although only one (L2) has been shown previously to play a major role in binding nuclear receptors. Our mutagenesis studies demonstrated that the interaction of ERR␣ with PGC-1␣ involves the L3 motif. Identification of an LXXLL motif as the ERR␣ binding site is consistent with the observation that the ERR␣ AF-2 domain is required for the interaction with PGC-1␣. However, this is the first example in which the L3 motif of PGC-1␣ was found to mediate the interaction of PGC-1␣ with a nuclear receptor beyond a role as an accessory site for GR interaction with PGC-1␣ (12). Our results indicate that the L3 motif functions as the primary site of interaction and that it is sufficient to bind ERR␣. This proposed mechanism is distinct from most other NR interactions with PGC-1␣, as demonstrated by the observation that the interaction of PGC-1␣ with ER␣ and HNF-4 required the L2 site.
Although LXXLL defines the consensus signature motif through which coactivator proteins interact with nuclear receptors, numerous studies have established the importance of residues flanking the Leu-rich sequence in determining receptor selectivity and binding affinity (29,30). Chang et al. (29) recently defined three distinct classes of LXXLL domains based upon conserved flanking residues. According to this scheme, the PGC-1␣ L2 site is a class III binding site determined in part by the presence of Leu and Ser residues positioned immediately upstream of the first Leu of the LXXLL motif. The L3 motif, through which ERR␣ primarily interacts, appears to be an inverted LXXLL and, therefore, does not readily conform to this classification scheme. However, recent crystal structure studies performed with ER␣ demonstrated that the coactivator binding pocket of ER␣ could recognize an LXXL motif within an NR box derived from the transcription intermediary factor 2 coactivator (31). Interestingly, despite the presence of a consensus LXXLL motif in the transcription intermediary factor 2 NR box, constraints placed by basic residues N-terminal to the ␣-helical motif shifted the binding site by one residue changing the recognition motif to LXXYL. The L3 motif of PGC-1␣ does, in fact, conform to an LXXYL consensus. Furthermore, a series of basic residues lie upstream of the L3 LXXYL motif that may contribute to its recognition by ERR␣ and ERR␥. Finally, consistent with these findings, the L1 motif (LLavL) does not match either of these consensus sequences and has not been identified as a high affinity binding site in any nuclear receptor AF-2-dependent interaction with PGC-1␣ (12) (present study).
The basis for the unique interaction of ERR␣ with PGC-1␣ compared with other L2-dependent PGC-1␣ nuclear receptor partners is unknown but is presumably related to structural differences within the nuclear receptor LBDs. The nuclear receptor interface with PGC-1␣ has not been precisely defined. However, p160 coactivator binding sites have been mapped for a number of receptors, including ER␣, PPAR␥, and retinoid X receptor ␣ (31)(32)(33). ER␣ interacts with NR boxes within transcription intermediary factor 2 via direct contacts between the LXXLL ␣-helix and ER␣ residues Glu-542, which resides within LBD helix 12, and Lys-362, which resides within helix 3. Both of these residues are conserved in ERR isoforms, suggesting that other structural differences within the LBD of ERR␣ and ERR␥ account for their differential binding with PGC-1␣. Modeling of the ERR␣ LBD based on homology with ER␣ reveals that 16 of 19 residues involved in ligand binding to ER␣ are either identical or conservative mismatches in ERR␣ (34). Of the three distinct residues, Phe-329 was shown to be essential for the constitutively active conformation adopted by the LBD of ERR␣ (34). Interestingly, the same residues are also distinct between ERR␣ and ERR␥. These findings suggest that, although these residues are not directly involved with coactivator binding, the amino acid differences may contribute to slight differences in LBD conformation among these receptors to influence the binding interface with PGC-1␣. Lastly, a number of studies have described allosteric effects of DNA response elements on the conformation of bound receptors (35)(36)(37)(38). Recent findings suggest that the specificity of ERR isoforms for regulating particular target promoters may be because of the differential effects of various DNA response elements on ERR/ coactivator or corepressor interactions (39).
PGC-1␣ has been shown to coactivate a variety of transcription factors involved in multiple cellular metabolic pathways. The distinct binding interface between PGC-1␣ with ERR␣ compared with other nuclear receptors provides a potential mechanism by which PGC-1␣ could distinguish among available partners. Coactivator selection of a nuclear receptor partner is obviously dictated by a number of factors, including relative expression levels, differential binding of corepressors, and LBD conformation. The ability of PGC-1␣ to utilize distinct binding sites with different partners provides an additional basis for receptor selection. According to this model, a posttranslational event, such as phosphorylation, could influence the accessibility of specific nuclear receptor binding sites within the PGC-1␣ molecule. Hence, when the L2 site is inaccessible, the branch of the PGC-1␣ regulatory network mediated by L2-dependent receptors would be inactive whereas PGC-1␣ is recruited to alternate partners, such as ERR␣. Such a mechanism would allow dynamic control of PGC-1␣-mediated activation of related metabolic gene targets downstream of its numerous partners.
The ERR isoforms (␣, ␤, and ␥) comprise a subfamily of nuclear receptor transcription factors involved in transcriptional activation or repression of target genes (40,41). Although ERR␣ was the first orphan nuclear receptor described within the mammalian nuclear receptor superfamily, little is known about its biological function. Some proposed models have focused on the potential for ERR␣ modulation of estrogen signaling. ERR␣ and ER␣ share some common targets, and ERR␣ has been shown to modulate ER␣-mediated transcriptional responses (42)(43)(44). ERR␣ has also been implicated in the regulation of cellular differentiation and metabolism (22,23). Regarding the latter, several lines of evidence suggest that ERR␣ serves to regulate PGC-1␣ target genes involved in mitochondrial metabolism. First, ERR␣ and ERR␥ expression is enriched in tissues that utilize mitochondrial FAO as a primary source for energy or heat generation, such as heart and BAT. We also found that ERR␣ and ␥ expression parallels FAO capacity among skeletal muscle types. 2 ERR isoforms and PGC-1␣ are highly expressed in rodent skeletal muscles comprised of slow-twitch oxidative fibers compared with fasttwitch fiber types in which these factors are nearly undetectable. Second, the coordinated developmental pattern of ERR␣ and PGC-1␣ expression suggests a role in metabolic maturation. ERR␣ transcript and protein levels are low in the prenatal stages in heart and BAT (see Fig. 2 and Ref. 3) when glycolytic metabolism serves as the chief source of energy. Following birth, ERR␣ expression is induced coincident with increased reliance on fat oxidation for energy generation. Finally, our current results, as well as previous studies (22,23), have shown that the MCAD gene is a specific ERR␣ target. ERR␣ directly binds the MCAD gene promoter via a NRRE-1, which is essential for the tissue-and developmental-specific expression profile of MCAD (22,23). However, prior to this report, evidence showing direct ERR␣-mediated regulation of MCAD gene transcription was lacking. The results of the ERR␣ overexpression and cotransfection studies demonstrate that MCAD is an endogenous target for ERR␣ and that in the presence of PGC-1␣ ERR␣ transactivates the MCAD promoter. Recently, numerous studies have shown that PGC-1␣ plays a critical role in regulating cardiac mitochondrial biogenesis and mitochondrial oxidative metabolism during the postnatal period. We propose that ERR␣ and ERR␥ serve as one component of this broad regulatory cascade.