Mice were maintained on 12-h light/dark cycle and cared for in accordance with the Animal Care and Use Committee protocol in the Novartis Institutes for BioMedical Research animal facility. Body composition was determined using EchoMRI (Echo Medical Systems, Houston, TX). ERRγ heterozygous null mice were obtained from Deltagen (San Mateo, CA) and backcrossed to C57Bl/6 for four generations (7). ERRγ transgenic mice were generated on a C57Bl/6 background using standard procedures. The transgenic construct consists of the 4.8-kb mouse muscle creatine kinase promoter driving a hemagglutinin-tagged VP16-ERRγ fusion or a hemagglutinin-tagged ERRγ gene.

Mouse myoblasts were isolated as described previously (20). Human ERRγ-IRES green fluorescent protein adenovirus was generated by Welgen (Worcester, MA). Following a 48-h differentiation period, myotubes were transduced with green fluorescent protein or ERRγ adenovirus for 48 h. The titers to achieve low and high levels of ERRγ expression were 1 × 10⁸ and 5 × 10⁸ virus particles/ml, respectively. GSK4716 was synthesized as described previously (9) and dissolved in DMSO for cell culture experiments. Cells treated with a corresponding volume of DMSO were used as controls.

For both tissues as well as cell lysates, total RNA was isolated using TRIzol reagent (Invitrogen). For muscle tissue, tissues were pulverized, and the RNeasy® Fibrous Tissue Minikit (Qiagen, Valencia, CA) was utilized for isolation of RNA. Quantitative real time PCR was performed using Assay-on-Demand® primer probes (Applied Biosystems, Foster City, CA) (supplemental Table S1). Mitochondrial DNA levels were assessed as described previously (17).

Tissues or cell lysates were prepared in RIPA buffer and separated using standard techniques. Normalization for loading was done using Ponceau staining. The polyclonal anti-human ERRγ antibody was generated using a 229-amino acid peptide corresponding to the ligand-binding domain (Covance Research Products Inc, Denver, PA). A donkey anti-rabbit IgG horseradish peroxidase-linked secondary antibody (GE Healthcare) and SuperSignal West Dura extended duration ECL substrate (Pierce) were used for detection. Citrate synthase activity was determined using 2 μg of total protein as prepared above, using the citrate synthase assay kit (Sigma).

RNA isolated from myotubes transduced with adenovirus or intact gastrocnemius muscle were used for microarray analysis. RNA was hybridized to mouse 430_2 Affymetrix chip (Affymetrix, Santa Clara, CA) in triplicate.

Muscle samples were rapidly frozen by submerging for 20 s in a chilled isopentane bath wrapped in pre-cooled plastic wrap and placed in a −80 °C freezer for storage until cryo-sectioning. Samples were sectioned at 6 μm in a cryo-microtome and mounted on glass slides immediately prior to staining. Succinate dehydrogenase activity was measured histochemically according to published protocols (21), and muscle fiber size was measured by immunohistochemical staining using a polyclonal rabbit anti-laminin antibody (Sigma) at room temperature for 1 h at a 1:800 dilution. Detection was performed using a rabbit-on-rodent APL polymer (Biocare, Concord, CA) and Vulcan Fast Red chromogen (Biocare).

Stained slides were digitized using an Aperio ScanScope XT slide scanner and analyzed using algorithms from Aperio (color deconvolution for succinate dehydrogenase staining intensity and positive pixel count for total tissue area) as well as a custom watershed algorithm (ASTORIA) developed to segment the muscle fibers from the laminin-stained sections. Area of positive staining was reported as the percent tissue area that stained positive for succinate dehydrogenase, and staining intensity was reported as the total optical density of all positive pixels divided by the total tissue area.

Transmission electron microscopy was performed using standard techniques. Gastrocnemius muscles were dissected unilaterally and placed in a modified Karnovski's fixative. Samples were routinely processed and were embedded in EMbed (Epon) 812 such to maintain a longitudinal orientation of the resulting sections. Thick sections were stained with toluidine blue to identify the best longitudinal block for each animal. Ultrathin sections cut for transmission electron microscopy survey were double-stained with uranyl acetate and lead citrate and examined using a FEI Tecnai G2 BioTwin electron microscope. Photomicrographs were captured using an Olympus-SIS Morada digital camera.

C57Bl/6 mice (age 15 weeks) were trained for a period of 8 days by running on a treadmill (10 m/min, zero degree incline) for 2 h each day. Tissues were isolated immediately after the last bout of exercise training. To determine exercise capacity, 4-month old male mice were placed on a 6-lane treadmill (Columbus Instruments, Columbus, OH) and run with a fixed upward slope of 10°. The speed during the 1st h was 10 m/min and increased 2 m/min every 15 min thereafter. Work, peak VO₂, and respiratory exchange ratios were assessed as described previously (17).

Statistical analysis was performed using a two-tailed Student's t test or analysis of variance as appropriate and was considered statistically significant if p < 0.05. For analysis of microarray data, the Probe Logarithmic Intensity Error (22) method was used to normalize the raw data and generate the summarized probe-set level gene expression values. A moderated t test (23) was applied to identify the significantly altered genes using the “limma” package from Bioconductor. p values were adjusted using the BH (24) method to control for false discovery rate in multiple testing. Functional enrichment analyses were performed at DAVID (Data base for Annotation, Visualization and Integrated Discovery) (25).

Exercise is a physiological intervention that results in the dynamic induction of mitochondrial biogenesis in skeletal muscle. We showed that both PGC-1α and ERRα were elevated with our exercise protocol (Fig. 1, A and B), as reported previously (26–28). Interestingly, ERRγ expression is induced by exercise (Fig. 1C). The expression of ERRγ transcript was found to correlate with the oxidative potential of skeletal muscle; it is expressed at high levels in the soleus (type I, oxidative muscle) and at low levels in the gastrocnemius (type I and II mixed fiber muscle) and extensor digitorum longus (type II, glycolytic muscle) (Fig. 1D).

Primary mouse myotubes were transduced with green fluorescent protein or GFP-IRES-ERRγ using two concentrations of adenovirus per treatment. The levels of ERRγ protein obtained are shown in supplemental Fig. 1A. Quantitative real time PCR demonstrated that induction of ERRγ in myotubes resulted in a down-regulation of ERRα and PGC-1α (supplemental Fig. 1B), suggesting the presence of a negative feedback mechanism to regulate levels of mitochondrial biogenesis. Additionally, genes of fatty acid metabolism, including carnitine palmitoyltransferase 1b (Cpt1b) and fatty acid-binding protein 3 (Fabp3), were also increased, whereas those of oxidative phosphorylation were mildly up-regulated (supplemental Fig. 1B). Interestingly, one of the most highly up-regulated genes is Epas1 (encoding endothelial PAS domain protein 1). Citrate synthase activity was increased with ERRγ expression (supplemental Fig. 1C). These data indicate that the expression of ERRγ results in a coordinate increase in mitochondrial function in primary mouse myotubes. Gene set enrichment analysis of microarray data of the ERRγ-transduced cells indicated a coordinated increase in genes of the mitochondrial electron transport chain, tricarboxylic acid cycle, and fatty acid oxidation (supplemental Fig. 1D) as reported previously in cardiac muscle (6). In addition, genes encoding pathways for muscle calcium handling and contractile proteins for slow muscle were also found to be up-regulated by ERRγ.

The muscle creatine kinase promoter was used to generate muscle-specific ERRγ transgenic mice (Fig. 2A, top panel). We analyzed two independent founder lines, one where the ERRγ was expressed as a fusion with the VP16 viral transactivator and the other expressing the “native” version of the protein. Expression of transcriptional factors as VP16 fusions is a strategy often utilized in vivo to ascertain effects elicited by increasing intrinsic activity of the factor beyond that achieved by enhanced protein levels alone (29, 30). Fig. 2A, lower left panel, demonstrates expression of the VP16ERRγ transgenic protein in founder line 431. The transgene was expressed at equal levels in gastrocnemius, soleus, and tibialis anterior muscle but was not expressed in the heart (supplemental Fig. 2A). We observed that VP1ERRγ transgenic muscles were noticeably redder and all hind limb muscles exhibited the same color as the soleus muscle (Fig. 2A, lower right panel). The animals had a similar body weight and body composition as compared with the control littermates (supplemental Fig. 2, B and C). The tibialis anterior (glycolytic type II fiber type muscle) and gastrocnemius (mixed fiber type muscle) muscles were smaller in the transgenics (Fig. 2B), although the mass of an oxidative type I fiber muscle such as soleus was not decreased. The transgenic line 339, expressing increased native ERRγ protein in muscle (supplemental Fig. 3A), exhibited a similar decrease in muscle weight of glycolytic and mixed fiber muscles but no change in soleus muscle, in the absence of any body weight changes (supplemental Fig. 3, B–E).

Individual fiber size was determined in the entire muscle cross-section by laminin staining. The fiber size distribution for the VP16ERRγ transgenic muscle was left-shifted, indicating a smaller muscle fiber size as evidenced by the histogram in Fig. 2C. Analysis of mitochondrial DNA levels in gastrocnemius muscle demonstrated a trend toward an increased content in the transgenic mice (Fig. 2D, p = 0.069). Electron microscopy of gastrocnemius muscle demonstrated an increase in numbers of large mitochondria, characterized by an enlargement of mitochondrial surface area in the transgenics (Fig. 2E). These large mitochondria were more pronounced in the subsarcolemmal region as opposed to the intramyofibrillar space (Fig. 2E, High, lower panels).

The VP16ERRγ transgenics did not show any alterations in energy expenditure, activity, or respiratory exchange ratios in the basal state (data not shown). When challenged with an endurance treadmill test, the VP16ERRγ transgenic mice demonstrated a significant increase in the work performed and distance traveled compared with the wild type controls (Fig. 2, F and G). In a peak VO₂, the VP16ERRγ transgenic mice surpassed their control littermates (Fig. 2H). Interestingly, the respiratory exchange ratio values in the transgenic mice were lower during this exercise challenge, indicating a greater preference for fat oxidation (supplemental Fig. 2D). The respiratory exchange ratio values for the transgenics do not increase beyond a value of 1, implying that the transgenic mice are protected from the lactic acid-induced hyperventilation seen under maximal exercise. The non-VP16-tagged ERRγ transgenic mice also demonstrated an increase in distance run during an exercise capacity challenge and decreased peak oxidative capacity (Fig. 2, I–K).

Analysis of succinate dehydrogenase enzyme activity in the muscle cross-sections demonstrated a robust increase in the succinate dehydrogenase positive area as well as intensity in the transgenic muscle (Fig. 3, A and B). Succinate dehydrogenase is situated in the mitochondrial inner membrane and is a component of the electron transport chain as well as the tricarboxylic acid cycle. Aconitase, like citrate synthase, is a mitochondrial matrix enzyme that is a component of the tricarboxylic acid cycle. The activities of both these enzymes were robustly elevated in the transgenic gastrocnemius muscle (Fig. 3, C and D). These data indicate that muscle mitochondrial activity in the ERRγ transgenic mice is strongly up-regulated.

Gastrocnemius muscle isolated from ERRγ transgenic mice exhibited decreased expression of MHC IIb and a concomitant increase in expression of MHC IId and MHC IIa and a trend toward an increase in slow MHC I, which did not reach statistical significance (Fig. 3E). This is indicative of a muscle fiber-type conversion from fast type II toward slow type I muscle fibers. Furthermore, we observed increased expression of myoglobin, a gene essential for the facilitated oxygen transport required by oxidative skeletal muscle (Fig. 3E). Candidate gene analysis shows decreases in Ppargc1a and Esrra but an increase in ERRβ, fatty acid oxidation genes (Lpl and Fabp3), and Epas1 (Fig. 3F). The down-regulation of Ppargc1a and Esrra suggests that there exist feedback regulatory loops for the control of mitochondrial function. Global gene expression was assessed in an unbiased fashion using microarrays, followed by gene set enrichment analysis. The top regulated pathways by their GO categories in the transgenic mice are described in Tables 1 and 2. Genes representing pathways of lipid metabolism, angiogenesis, and muscle calcium handling and contractility were robustly up-regulated in the transgenic muscle, whereas genes of the electron transport chain showed a smaller magnitude of up-regulation. Among the gene pathways that were down-regulated in transgenic muscle were genes representing fast twitch myosins and calcineurin Aα (Fig. 3G). Gene expression analyses using microarray in gastrocnemius muscle of the native ERRγ transgenics demonstrated similar patterns of changes in this muscle as in the VP16ERRγ muscle (supplemental Fig. 3F).

We analyzed our microarray data in context of the gene signatures observed in two distinct muscle fiber types (Fig. 4A, left panel) that were generated based on a published data set described for tibialis anterior muscle, a representative of type II fast twitch glycolytic muscle, and soleus muscle, which is a slow type I fiber oxidative muscle NCBI GEO accession number GSE10347) (31). Classically, the definition of the skeletal muscle fiber type is based on MHC profile distribution, as well as determination of ATPase activity (32). Interestingly, we observed that the majority of gene signature up-regulated in soleus muscle is also induced in the ERRγ transgenic mice; conversely, the gene pathways decreased in soleus but increased in the tibialis are decreased in the ERRγ transgenic muscle (Fig. 4A, right panel). Gene pathways over-represented in the soleus and transgenic muscle were those of fatty acid oxidation, oxidoreductases, muscle fiber type, angiogenesis, and extracellular matrix components. Gene pathways representing glucose metabolism, fast twitch muscle, and calmodulin binding were lower in soleus as compared with the tibialis muscle and were similarly lower in the transgenic muscle as compared with the wild type. Statistical analysis of this cross-comparison yielded a high a degree of significance (Fig. 4B). Thus, we can closely overlay the molecular genetic signature of ERRγ onto that of a pure oxidative muscle.

Tissues from ERRγ heterozygous mice displayed a 50% reduction in ERRγ mRNA (Fig. 5A) and protein (Fig. 5B). No significant differences in muscle, heart, and total body weights were observed in these mice (data not shown). These mice did not show any alterations in energy expenditure, activity, or respiratory exchange ratios in the basal state (data not shown). The ERRγ heterozygotes ran a shorter distance and performed less work during an exercise capacity test (Fig. 5, C and D). These mice attained a reduced maximum speed (16.71 ± 0.24 versus 21.22 ± 0.29 m/min) compared with their wild type controls. Furthermore, ERRγ heterozygotes demonstrated a lower peak VO₂ measurement (Fig. 5E). Citrate synthase levels trended to be lower, although the difference did not reach statistical significance (Fig. 5F). Gastrocnemius muscle demonstrated no changes in levels of Ppargc1a and Esrra to compensate for the partial loss of ERRγ in the heterozygous mice, although Esrrb levels were reduced. Additionally, fatty acid uptake and oxidation genes such as Lpl and Cpt1b were decreased in the muscle of these mice, indicating that these animals may have impairment in the utilization of fatty acids as fuel.

Treatment of primary mouse myotubes with GSK4716, an ERRβ/γ agonist (9), resulted in a concerted increase in the expression levels of Ppargc1a, Ppargc1b, and the Esrr genes (Fig. 6A). Furthermore, Cpt1b, Atp5b, and Idh3, genes in key mitochondrial pathways, were also induced by GSK4716 (Fig. 6A). The concentration of the compound used in our study is similar to that observed for the activation of both ERRβ and ERRγ in the Gal4 transactivation assay (9). Additionally, GSK4716 increased citrate synthase activity (Fig. 6B) and cytochrome c protein levels (Fig. 6C). These results indicate that the activation of ERRβ/γ transcriptional activity with a small molecule agonist can produce a coordinate increase in mitochondrial gene expression and function.