PGC-1α regulates a HIF2α-dependent switch in skeletal muscle fiber types

PGC-1α Regulates Expression of HIF2α.

Before this study, the entire spectrum of transcription factors that are regulated by PGC-1α in any particular tissue had not been described. To elucidate the composition of the PGC-1α transcriptome in skeletal muscle, we expressed GFP or PGC-1α in differentiated myotubes and compared the expression of all known transcription factors, coactivators, and repressors (or proteins containing a motif associated with these functions) in a high-throughput qPCR-based assay we recently described (Quanttrx) (28). Of the ∼1,800 genes represented in this database, >300 factors were increased in response to PGC-1α expression; HIF2α showed one of the greatest quantitative changes (∼300-fold; Dataset S1). Although many known targets of PGC1α action were also increased, such as ERRα and PPARα, the RNA expression of HIF1α was not affected (Fig. 1A). This result is in agreement with earlier work (6). The PGC-1α–mediated increase in HIF2α mRNA expression did not cause a detectable increase in HIF2α protein under normoxic conditions (21% O₂); however, upon PHD inactivation and HIF stabilization with 2,2 dipyridyl, HIF2α protein was markedly elevated in the PGC-1α–expressing myotubes, compared with GFP controls (Fig. 1B). This result suggests that HIF2α activity may not be dramatically altered by elevated PGC-1α alone and may require a secondary stabilization event to display maximal transcriptional activity. However, because intramuscular oxygen tensions are closer to 6–7% even in the sedentary state, this likely provides a more permissive environment for HIFα stabilization.

To determine whether PGC-1α can regulate HIF2α expression in vivo, we used several genetic gain- and loss-of-function models and also challenged mice with a number of stimuli known to induce PGC-1α. HIF2α mRNA expression is increased 2.5-fold in the gastrocnemius of mice with a muscle-specific transgenic expression of PGC-1α (Fig. 1C) (32). Additionally, overnight wheel running, an exercise protocol known to induce PGC-1α, increased the expression of HIF2α approximately twofold, whereas it induced PGC1α approximately sevenfold (Fig. 1D). Conversely, HIF2α expression is decreased 27% in the gastrocnemius of muscle-specific PGC-1α knockout (MKO) animals (Fig. 1E). PGC-1α is also known to be potently regulated by β-adrenergic receptor activation during exercise and this effect can be mimicked by injection of the β2-receptor agonist clenbuterol (33). Intraperitoneal administration of clenbuterol promoted a rapid induction of PGC-1α (Fig. 2A) and also stimulated an elevation of HIF2α expression; this effect was completely abolished in the PGC-1α MKO mice (Fig. 2A). Taken together, these data indicate that HIF2α mRNA expression is induced in response to pharmacological and physiological stimuli in muscle and that this inducible regulation is completely dependent on PGC-1α.

PGC-1α Regulation of HIF2α Is Partially Muscle Selective.

PGC-1α is highly expressed in tissues with greatest energy demands, such as brown fat, skeletal muscle, heart, and kidney (Fig. S1). Within skeletal muscle, PGC-1α is most abundant within oxidative muscle, such as the soleus, and is expressed at lower levels in “whiter” muscles such as the gastrocnemius and quadriceps (Fig. S1). Consistent with this, HIF2α expression correlates positively with the expression of PGC-1α in a majority of tissues, being most abundantly expressed in highly metabolic tissues including brown fat, soleus, heart, and kidney (Fig. S1). Somewhat paradoxically, HIF2α mRNA expression is highest in the lung, a tissue with high oxygen tensions that is unlikely to allow for HIF2α stabilization (Fig. S1).

Because HIF2α expression is correlated with the expression of PGC-1α in many tissues, we asked whether PGC-1α could regulate the expression of HIF2α in a variety of cell types. As previously shown, PGC-1α robustly up-regulated HIF2α expression in primary myotubes and significantly increased HIF2α expression in immortalized C2C12 myotubes (fivefold) (Fig. 2B). Forced expression of PGC-1α in primary hepatocytes, primary renal proximal tubular cells (RPTCs), and primary brown adipocytes resulted in very modest increases HIF2α expression (two- to threefold), whereas HIF2α expression remained unchanged in 3T3-L1 adipocytes (Fig. 2B). These data suggest that robust regulation of HIF2α by PGC1α is at least partially muscle-cell selective.

ERRα Is Required for the PGC-1α–Mediated Increase in HIF2α Expression.

PGC-1α is a key mediator of mitochondrial biogenesis and angiogenesis in response to exercise in skeletal muscle; both the mitochondrial energetics and angiogenic response to PGC-1α are completely dependent on PGC-1α binding to ERRα (7, 34). We investigated whether ERRα may also play a role in the PGC-1α–dependent regulation of HIF2α. Primary ERRα KO myotubes show a dramatic reduction in the expression of HIF2α in the basal state, suggesting that ERRα may indeed regulate the expression of HIF2α (Fig. 3A). Additionally, adenoviral expression of ERRα results in a dramatic increase in HIF2α expression, whereas the forced expression of ERRγ has a much more mild effect (Fig. 3B) (8-fold vs. 36-fold).

To determine whether the PGC-1α–mediated induction of HIF2α requires ERRα, we first used an ERRα-specific inverse agonist, XCT-790. Pretreating myotubes with XCT-790 (10 μM) before infection with PGC-1α blocked the up-regulation of HIF2α mRNA (Fig. 3C). In addition, the genetic loss of ERRα completely abolished the PGC-1α–mediated increases in HIF2α expression (Fig. 3D). This result demonstrates an absolute requirement for ERRα in this process, at least in primary cell culture. Interestingly, the increase in HIF2α in response to the forced expression of ERRα was significantly blunted in PGC-1α^−/− myotubes, suggesting that ERRα also requires coactivation by PGC-1α to influence the expression of HIF2α (Fig. 3E).

PGC-1α Regulation of HIF2α Is Modulated by SIRT1.

SIRT1 is a NAD⁺-dependent protein deacetylase that has been implicated in regulating whole body metabolism (35). SIRT1 has been shown to deacetylate and active PGC-1α to enhance the regulation of PGC-1α gene programs (36). In addition, SIRT1 has also recently been shown to associate with, deacetylate, and activate HIF2α (37). To examine the involvement of SIRT1 in the PGC1α/HIF2α axis, we pretreated myotubes with a specific SIRT1 inhibitor, EX-527, or a noncompetitive inhibitor of NAMPT, FK-866, known to reduce intracellular concentrations of NAD⁺ and reduce SIRT1 activity (38, 39). Pretreatment with either EX-527 or FK-866 blocked the PGC-1α–mediated up-regulation of HIF2α (Fig. 3F). Conversely, we explored whether elevated SIRT1 expression and activity in vivo was sufficient to enhance the PGC-1α regulation of HIF2α mRNA. Wild-type littermate controls and whole-body SIRT1 transgenic mice (approximately threefold overexpression, Fig. 3G) were injected i.p. with clenbuterol to promote a PGC-1α–dependent increase in HIF2α in the gastrocnemius (Fig. 3H). Up-regulation of SIRT1 resulted in a marked potentiation of HIF2α expression in response to clenbuterol (Fig. 3H). Taken together, these data suggest that SIRT1 enhances the activity of PGC-1α and the subsequent expression and activity of HIF2α.

HIF2α Coordinates a Slow-Type Muscle Fiber Switch.

Because HIF2α is a robust transcriptional target of PGC-1α and ERRα, we next sought to explore the function of HIF2α in skeletal muscle. Because a role for HIF2α has not been described in skeletal muscle, we took an unbiased approach to characterizing HIF2α-regulated gene programs. Primary myotubes were virally transduced with either a GFP or a HIF2α protein, in which key proline residues have been mutated to alanine, allowing for HIF2α stabilization and accumulation in normoxia (∼2,000-fold overexpression of HIF2α) (40). RNAs were then isolated and subjected to Affimetrix analysis (Dataset S2). All genes that were altered >1.5-fold were included in an unbiased Ingenuity pathway analysis (Fig. S2). Interestingly, one of the most significant biological pathways identified by this analysis was skeletal muscle system development and function (Fig. S2). Gene sets involved in cardiovascular disease and skeletal and muscular disorders were scored as significant, suggesting that HIF2α may indeed play an important role in skeletal muscle development and function (Fig. S2). Further analysis of gene sets involved in skeletal muscle function and development revealed a substantially overrepresented set of myosin heavy chain family members that are dramatically altered by HIF2α (Fig. 4A). Interestingly, MyoVIIa (MyoHCI), a key part of slow twitch type I muscle fiber function, was dramatically increased by HIF2α. The second most down-regulated gene on the array was MyoIV (MyoHCIIb), a gene whose expression is characteristic of fast-twitch type IIb fibers (Fig. 4A and Dataset S2). These data suggest HIF2α may be a potential regulator of slow fiber-type determination, by increasing the slow twitch myosin heavy chain muscle program and suppressing the expression of fast twitch markers.

Regulation of a Slow Twitch Muscle Fiber Gene Program Is HIF2α Dependent.

Because HIF1α and HIF2α regulate a wide overlapping spectrum of genes, we profiled muscle fiber-type genes in response to normoxia-stabile mutants of HIF1α and HIF2α (Fig. 4B). Both HIFα isoforms robustly activated the angiogenic program, as exemplified by increased VEGF-A expression; HIF1α demonstrated a greater potency than HIF2α in these experiments (Fig. 4B). Importantly, expression of genes characteristic of skeletal muscle fiber-type switching appears to be a specific response to HIF2α. mRNAs for several genes characteristic of slow-twitch muscle fibers such as MyoHCI, Myoglobin, Calmodulin2, and Troponin I are increased at the mRNA level by HIF2α, but are largely unchanged in response to HIF1α (Fig. 4B). Additionally, HIF2α but not HIF1α dramatically represses the expression of MyoHCIIb, a gene whose expression serves as a marker of ultrafast twitch (type IIb) muscle fibers (Fig. 4B). HIF2α also caused a mild repression of mRNA for the intermediate fiber-type marker, MyoHCIIa, but has no affect MyoHCIIx expression (Fig. 4B). These data suggest that the gene program of the slow-twitch fiber type is specifically regulated by HIF2α and remains largely unaffected by HIF1α. This result defines an important function for HIF2α.

We next asked whether the well-studied effect of PGC-1α in causing a switch to the oxidative type I and type IIa fibers required HIF2α. To determine HIF2α dependence, primary myoblasts were isolated and cultured from floxed HIF2α mice (HIF2α^fl/fl); the HIF2α gene was excised using an adenoviral Cre-recombinase and HIF2α^fl/fl myotubes or HIF2α^fl/fl-CRE myotubes were treated with or without adenoviral PGC-1α in the presence of absence of 2,2 dipyridyl and a subquent analysis of fiber-type genes was performed by qPCR (Fig. 4C). Under conditions where HIF2α had been stabilized, the forced expression of PGC-1α stimulated mRNA levels for MyoHCI, Myoglobin, and Calmodulin2 and suppressed MyoHCIIb. Importantly, this response was essentially abrogated in myotubes lacking HIF2α, suggesting that the gene program of PGC-1α/ERRα–mediated slow-twitch fiber-type transition is absolutely dependent on the presence of functional HIF2α (Fig. 4C). Additional HIF2α target genes induced by PGC-1α and 2,2 dipyridyl were also regulated in a similar manner. Interestingly, the expression of HIF2α does not appear to be directly regulated by ERRα as there are no ERR concensus sites in HIF2α promoter regions. Unlike many transcription factors that are regulated by PGC-1α, HIF2 is not coactivated by PGC-1α (Fig. S3).

HIF2α Regulates Fiber-Type Determination in Vivo.

To determine the relative contribution of HIF2α to fiber-type determination in vivo, we created a muscle-specific knockout of HIF2α by excising both alleles of HIF2α with a myogenin-driven Cre-recombinase (HIF2α-MKO). An analysis of the gene program of fiber types was subsequently performed by qPCR on muscles from 12-wk-old mice. Genetic ablation of HIF2α (∼90% reduction in soleus, Fig. 5A), promoted a shift toward fast twitch myosin heavy chain gene expression as exemplified by marked increases in the expression of MyoHCIIb and reductions in MyoHCI, calmodulin2, and TpnI (Fig. 5B). As anticipated, we also observed a significant decrease in the expression of angiogenic genes such as VEGF-A and genes associated with the antioxidant defense system such as SOD2 (Fig. 5B). To further examine the role of HIF2α in the regulation of physiological fiber-type determination, a histological fiber-type analysis was performed. Mice lacking HIF2α showed dramatic reduction in the histological presence of the slow-twitch, oxidative myosin heavy chain MyHCI, accompanied by an up-regulation of the ultrafast-twitch MyHCIIb in the soleus (Fig. 5C). Consistent with the gene expression data, no change in the expression of MyHCIIa was noted (Fig. 5C). Taken together, these data suggest an important role for HIF2α in the determination of muscle fiber type in vivo and provide additional clues regarding the function of HIF2α in skeletal muscle (Fig. S4).

Reagents.

All primers were purchased from Integrated DNA Technologies. Antibodies toward HIF2α were purchased from Novus Bio (NB100-122), whereas IHC antibodies for fiber-type analysis were purchased from the Developmental Studies Hybridoma Antibody Core at the University of Iowa (SC-71, BF-F3, and A4.840). EX-527 and FK-866 were purchased from Tocris. All other reagents were purchased from Sigma.

Animal Experimentation.

Muscle-specific PGC-1α transgenic and knockout mice were generated and maintained as previously described (6, 7, 32, 51, 52). Generation of HIF2α muscle-specific knockout mice was accomplished by crossing mice with a floxed HIF2α allele (24) with animals transgenically expressing Cre-recombinase under the control of the myogenin promoter and the MEF2C enhancer (a generous gift from R. Bassel-Duby and E. N. Olson, University of Texas Southwestern, Dallas, TX) (53). All experimental cohorts were compared with littermate controls and experiments were performed on mice between 12 and 14 wk of age. For exercise experiments, adult C57/bl6 mice (Jackson Laboratories) (6 wk old) were housed individually and subjected to one bout ad libitum running on in-cage voluntary running wheels (Mini Mitter, a Respironics company). The mice ran at least 8 km on the basis of data collected through electronic monitoring and processed with the VitalView Data Acquisition System. Clenbuterol-injected animals were injected i.p. with a dose of 1 mg/kg. Animals were euthanized and tissue samples were collected 6 h after clenbuterol injection. All experiments were performed in accordance with the Animal Facility Institutional Animal Care and Use Committee regulations.

Quanttrx Analysis.

Genome-wide transcription factor profiling (Quanttrx) was performed according to the method of Gupta et al., where 10 μg of RNA was isolated and reverse transcribed for each treatment group. Experiments were performed in triplicate and analyzed by the ΔΔCt method and normalized to several housekeeping genes (54).

Cell Culture.

Primary cell culture was performed as previously described. ERR^−/− myotubes were a generous gift from Zhidan Wu, Novartis Pharmaceuticals (Cambridge, MA). EPAS1^fl/fl myotubes were isolated from 7-d-old litters from EPAS1fl/fl homozygotes from Jackson Laboratories. PGC-1α adenovirus is as previously described. The ERRα and ERRγ adenoviruses were a generous gift from Don McDonell at Duke University (Durham, NC). The stable HIF2α adenovirus was a generous gift from Stephen Lee at University of Ottawa (Ottawa, ON, Canada).

Gene Expression Analysis.

Total RNA was isolated from tissues using the TRIzol reagent (Invitrogen) and prepped using the Qiagen RNeasy kit according to the manufacturer's instructions. RNA was reverse transcribed and analyzed using an Applied Biosystems Real-Time PCR System 7300, using the ΔΔCt method. Relative gene expression was normalized to 18S RNA levels. Affimetrix experiments were performed by the Dana-Farber microarray core facility and subsequent data analysis was conducted with dChip and Ingenuity pathway analysis software suites.

Statistical Analysis.

All results are expressed as means, and error bars depict SEs. A two-tailed Student's t test was used to determine P values. Mulivariate statistics were analyzed using one-way ANOVA with Student Newman–Keuls post hoc analysis. Statistical significance was defined as P < 0.05. For all cell-based experiments, at least n = 3 was reached. For animal experiments, at least four mice per group were used.