Evidence for a Mitochondrial Regulatory Pathway Defined by Peroxisome Proliferator–Activated Receptor-γ Coactivator-1α, Estrogen-Related Receptor-α, and Mitofusin 2

Various fragments of the 5′ flanking region of the human Mfn2 gene were amplified and subcloned into the pGL3basic reporter gene vector (Promega) or a pGL3 reporter under the control of a minimal thymidine kinase promoter. Expression plasmids for PGC-1α (FLAG-tagged PGC-1α) were a kind gift from Dr. Pere Puigserver (Johns Hopkins Medical School), and plasmid encoding ERRα was a gift from Dr. Diego Haro (University of Barcelona). Site-directed mutagenesis was performed with the QuikChange site-directed mutagenesis kit (Stratagene), following the manufacturer’s instructions. The sequence of the oligonucleotides used is available on request.

10T1/2, L6E9, C2C12, and HeLa cells were maintained in Dulbecco’s modified Eagle’s medium supplemented with 10% fetal bovine serum in subconfluent cultures. Mouse embryonic fibroblasts were cultured in Dulbecco’s modified Eagle’s medium containing 10% calf serum. Myotube differentiation was induced as previously reported (22). Nuclear and mitochondrial fractions were obtained by differential centrifugation of L6E9 cell homogenates (21). Homogenates were also obtained from 10T1/2 and mouse embryonic fibroblast cells. Western blot assays were performed, using specific antibodies against PGC-1α, Mfn2, subunit α from H⁺-F₁-ATP synthase, porin, and actin. Peroxidase-conjugated anti-mouse or anti-rabbit immunoglobulins were used as secondary antibodies, and bands were detected by an enhanced chemiluminescence Western blotting detection analysis system (Amersham).

For reporter gene assays, cells were transfected with FuGene 6 transfection reagent (Roche). Luciferase activity was measured 40 h after transfection, as previously reported (22). Expression of reporter genes was normalized to the number of green fluorescent protein–positive cells, which was measured by flow cytometry.

Adenovirus encoding PGC-1α was a kind gift from Dr. Pere Puigserver (John’s Hopkins University). 10T1/2, mouse embryonic fibroblast, and L6E9 (at differentiation day 3) cells were infected with adenoviruses. Cells were harvested 48 h after infection and various measurements performed (16,21).

Male Wistar rats were subjected to 48 h at 4°C or were injected with the β₃-adrenergic agonist CL-316243 for 1 or 7 days (1 mg · kg⁻ · day⁻). Brown adipose tissue and gastrocnemius muscle were removed, rapidly frozen in liquid nitrogen, and stored at −80°C until RNA extraction. Mfn2^+/− KO mice (C57BL/6J genetic background) were obtained from the California Institute of Technology (Pasadena, CA). Genotypes were determined by PCR from tail genomic DNA (17). Male mice (9–10 weeks old) where housed individually during exposition to cold (4°C). After 48 h of cold exposure, quadriceps muscle was removed. Animals were treated in accordance with the ethics committee in animal experimentation principles and guidelines of the Barcelona Science Park.

A radiolabeled double-stranded oligonucleotide probe corresponding to the human Mfn2 promoter sequence 5′-TTT CCT CAA AGG CGA CTG AAG GGC AG-3′ (−422/−396) was incubated with 3 μl of in vitro–transcribed and –translated ERRα, and electrophoresis was performed as described (23). Competitor-mutated oligonucleotide 5′-TTT CCT CAT ATG CGA CTG AAG GGC AG-3′ differs from wild-type electrophoretic mobility shift assay sequence by the same two-base substitution used in the functional experiments.

HeLa cells were transfected by the calcium phosphate method with a FLAG-tagged PGC-1α expression vector or an irrelevant expression vector. After 48 h, they were cross-linked by formaldehyde, lysed, and subjected to chromatin shearing. Immunoprecipitation was performed with M2 anti-FLAG antibody (Sigma). In other studies, extracts from wild-type HeLa cells were immunoprecipitated with an antibody against human ERRα (PPMX Perseus Proteomics). After chromatin immunoprecipitation, DNA was purified by phenol/chloroform extraction. Input (1% of total immunoprecipitated) and immunoprecipitated DNA were subjected to PCR analysis with primers flanking the ERRα site on the Mfn2 promoter (−505/−325 of the human Mfn2 promoter) and the primers for amplification of the cyclophilin gene.

Total RNA was extracted by using Trizol (Invitrogen). Mfn2 and cytochrome c oxidase subunit-II (COX-II) mRNAs were detected after hybridization with a 1,200-bp SacI human Mfn2 fragment and a 500-bp PstI fragment as cDNA probes, respectively (16). PGC-1α mRNA was detected by hybridization with a 500-bp EcoRI fragment

We isolated 2 μg of total RNA from rat or mice tissues or cell lines with the Trizol reagent (Invitrogen) and used them to synthesize full-length cDNA, in a 20-μl reaction, using oligo dT and SuperScript II reverse transcriptase (Invitrogen). Real-time PCR was performed with 0.8 μl cDNA (diluted 1:10), specific primers (300 nmol/l for most primers, 150 nmol/l for MEF2A, or 150 nmol/l for β-actin), and 2× SYBR Green PCR Master Mix (Applied Biosystems) in a final volume of 16 μl. PCR involved 40 cycles of denaturation (95°C, 15 s) and amplification (60°C, 1 min). In all cases, a single PCR product with correct base pair size was detected. Mouse PGC-1α and β-actin primers were from Kressler et al. (24), mouse MEF2A primers were from Kim et al. (25), and rat PGC-1α and cytochrome c primers were from Rodgers et al. (26). Sequence of primers is available on request. Quantification of cDNA was performed as reported (27). Results are the means of 5–6 cDNA samples from different observations.

10 T1/2 wild-type or mouse embryonic fibroblasts were infected with adenoviral vectors encoding β-galactosidase, green fluorescent protein, or PGC-1α at different multiplicities of infection, and then, 48 h later, mitochondrial membrane potential was measured with 40 nmol/l DiIC₁(5) (Molecular Probes) for 30 min at 37°C. DiIC₁(5) fluorescence emission after laser excitation was measured, using a Moflo flow cytometer (Summit Version 3.1 software, DakoCytomation) as previously reported (21).

To study the regulatory profile of Mfn2, we analyzed its expression under conditions such as exposure to cold or treatment with β₃-adrenergic agonists that stimulate basal energy expenditure and PGC-1α expression (5,28,29). Exposure to cold for 48 h caused stimulation of Mfn2 mRNA levels in skeletal muscle and brown adipose tissue (1.9- and 2.3-fold stimulation, respectively) (Figs. 1A and B). Under these conditions, the gene expression of the mitochondrially encoded COX-II or cytochrome c were not altered in muscle, whereas β-ATP synthase mRNA was enhanced (Fig. 1B); in brown adipose tissue, COX-II was modestly enhanced in response to cold (35% increase) (Fig. 1B). These results indicate that Mfn2 upregulation was specific. Exposure to cold for 48 h did not increase UCP3 mRNA levels in skeletal muscle (data not shown), which is in keeping with previous reports (30).

Treatment of rats with the β₃-adrenergic agonist CL-316243 for different times also increased Mfn2 expression. Administration of CL-316243 for 24 h caused a 60% increase in Mfn2 expression in brown adipose tissue, which occurred in the absence of changes in COX-II, β-ATP synthase, or cytochrome c mRNAs (Figs. 1A and B). CL-316243 for 1 week caused a marked stimulation of Mfn2 mRNA levels in muscle (2.2-fold), again in the absence of alterations in COX-II, β-ATP synthase, or cytochrome c mRNA and in the presence of enhanced UCP3 mRNA levels (Figs. 1 and data not shown). Treatment of rats with CL-316243 for 24 h did not cause any alteration in the muscle Mfn2 mRNA (data not shown). Cold exposure and treatment with CL-316243 enhanced PGC-1α expression in tissues (Fig. 1B).

The effects caused by cold or CL-316243 on Mfn2 expression were not mimicked by incubation of C2C12 muscle cells with analogs of cAMP (8-bromo-cAMP) or with a β-adrenergic agonist (isoproterenol) (data not shown), indicating that the stimulation is independent of cAMP.

We next examined whether PGC-1α was responsible for the Mfn2 induction under conditions of enhanced energy expenditure. PGC-1α shows a low expression in cultured muscle cells (6); thus, we analyzed the effect of PGC-1α overexpression in differentiated L6E9 myotubes. To this end, myotubes were infected with adenoviruses encoding PGC-1α or β-galactosidase (used as a control), and total RNA or mitochondrial or nuclear fractions were obtained. Mfn2 mRNA levels were markedly enhanced (near fourfold) in response to PGC-1α expression (Fig. 2A); under these conditions, genes induced by PGC-1α, such as β-ATP synthase, COX-II, or cytochrome c, were more moderately enhanced (60, 20, and 20% of increase, respectively) (Figs. 2A and data not shown). Mfn2 protein was also markedly increased in response to PGC-1α expression (3.6-fold increase), and this occurred in the presence of moderate changes in the α subunit of H⁺-F₁-ATP synthase (1.5-fold increase) (Fig. 2B).

To determine whether the effect of PGC-1α expression was a consequence of transcriptional regulation, L6E9, 10T1/2, or HeLa cells were transiently transfected with a fragment of the human Mfn2 gene promoter (−1982/+45) fused to the reporter gene luciferase. Cells transfected with the reporter gene showed low levels of luciferase activity, and cotransfection with PGC-1α caused a fourfold increase in activity (Fig. 2C). This increase was caused by the presence of specific sequences in the Mfn2 promoter, and cotransfection of PGC-1α and pGL3-basal promoter containing no promoter sequences to drive the luciferase reporter had no stimulatory effect (data not shown).

To determine the cis-elements involved in the effects of PGC-1α on the transcriptional activity of the Mfn2 promoter, 5′ deletion constructs of the Mfn2 promoter fused to the luciferase reporter gene were generated and transiently transfected in 10T1/2 cells (Fig. 3A). Deletion from −1982 to −532 of the Mfn2 promoter did not modify the effect of PGC-1α on luciferase activity (Fig. 3A), and deletion from −532 to −229 cancelled the effect of this coactivator (Fig. 3A). Next, distinct fragments of the promoter were cloned at 5′ of the thymidine kinase promoter and fused to the luciferase reporter. The fragment −532/−229 was sufficient to permit the response to PGC-1α (Fig. 3B), and analysis of deletion fragments revealed that −532/−352 and −459/−352 retained the capacity to respond to PGC-1α (Fig. 3B). A smaller fragment −432/−352 also responded to this coactivator by enhancing luciferase activity, but the fragment −407/−352 did not (Fig. 3C), indicating that the region −432/−407 is critical for conferring the capacity to respond to PGC-1α.

To identify the specific element that conferred sensitivity to PGC-1α, we analyzed the sequences corresponding to the −432/−352 fragment in human and mouse Mfn2 genes. This fragment was highly conserved, especially the region encompassing −432/−392 (Fig. 4A). Visual analysis of elements revealed the putative presence of three binding boxes for nuclear hormone receptors conserved in human and mouse Mfn2 promoters (Fig. 4A). Based on these observations, the three boxes were mutated in the −459/−352–thymidine kinase–luciferase construct, and functional analysis was performed in 10T1/2 cells. Cancellation of the middle box (box 2) completely blocked the response to PGC-1α, whereas mutation in box 1 did not (Fig. 4B). Mutation of box 3 or double mutation in boxes 1 and 3 reduced the stimulatory effect of PGC-1α compared with the wild-type construct (Fig. 4B). These results indicate that box 2 is sufficient for activity induced by PGC-1α and that the maximal activity requires the integrity of a second box, most likely box 3. This suggests the operation of a nuclear hormone receptor endogenously expressed in 10T1/2 cells with activity as a monomer. In this regard, transient transfection of cells with thyroid hormone receptor TRα1 in the presence of T₃ did not enhance the transcriptional activity of the Mfn2 promoter (data not shown).

PGC-1α transactivates oxidative phosphorylation gene transcription through binding to ERRα (13,14). Based on this observation, 10T1/2 cells were transfected with −1982/+45-luciferase plasmids either in the wild-type form or mutated in box 2, in the absence or presence of PGC-1α and/or ERRα. In the wild-type promoter, ERRα stimulated luciferase activity, and the cotransfection of PGC-1α and ERRα caused a synergic effect (Fig. 4C). Both actions were suppressed when analyzed in the promoter mutated in box 2 (Fig. 4C). Similar data were obtained when cells were transfected with the −459/−352–thymidine kinase–luciferase plasmid (data not shown).

To determine whether ERRα binds to the Mfn2 promoter, a DNA fragment encompassing the sequence 422/396 was labeled, and electrophoretic mobility shift assays were performed in the presence of ERRα synthesized in vitro (Fig. 5A). A retardation band was specifically detected, and it was competed by a 10- or 50-fold excess of unlabeled oligonucleotide and only partially competed by a mutated form of excess oligonucleotide (Fig. 5A). To confirm that ERRα or PGC-1α bind to the Mfn2 promoter in vivo, chromatin immunoprecipitation assays were also performed. To this end, extracts from HeLa cells, which endogenously express ERRα (31), were immunoprecipitated with an anti-ERRα antibody, and a fragment of the promoter containing the ERRα-binding element was PCR amplified (−505/−325). Immunoprecipitates specifically amplified the Mfn2 promoter, thereby indicating the binding of ERRα in vivo (Fig. 5B). In addition, cells were transfected with FLAG–PGC-1α, extracts were immunoprecipitated with an anti-FLAG antibody, and a fragment of the promoter containing the ERRα-binding element was amplified by PCR. Immunoprecipitates specifically amplified the Mfn2 promoter, indicating the binding of PGC-1α in vivo (Fig. 5C). In contrast, immunoprecipitates did not amplify a fragment of the cyclophilin gene, used as a negative control (Fig. 5C).

To determine whether some of the effects of PGC-1α depend on Mfn2, we studied the impact of Mfn2 loss of function. To this end, we analyzed two different cell models: 10T1/2 cells stably transfected with an antisense form of Mfn2, which show a 50% reduction in Mfn2 (16), and mouse embryonic fibroblasts obtained from Mfn2 KO embryos (17). Infection of wild-type 10T1/2 cells with an adenoviral vector encoding PGC-1α caused a specific and marked enhancement of mitochondrial membrane potential (Fig. 6A). 10T1/2 cells stably transfected with an antisense form of Mfn2 showed a lower PGC-1α–stimulated mitochondrial membrane potential (Fig. 6A). Similarly, the stimulatory effect of PGC-1α on mitochondrial membrane potential was reduced in Mfn2^−/− mouse embryonic fibroblast cells (Fig. 6B). The mitochondrial membrane potential that we measured in cells is a reflection of the number of mitochondria per cell and their energization state. Because Mfn2 loss of function did not completely block the effect of PGC-1α on mitochondrial membrane potential, we next determined the effect on the abundance of porin, a marker of mitochondrial mass. PGC-1α increased the total expression of porin in 10T1/2 and mouse embryonic fibroblast cells (Figs. 6C and D), and Mfn2 loss of function did not alter this effect (Figs. 6C and D).

Based on the observation that cold enhances PGC-1α in skeletal muscle, we next aimed to determine whether the regulatory pathway defined by PGC-1α and Mfn2 may be altered when only one allele of Mfn2 was present. To this end, wild-type and heterozygous KO mice for the Mfn2 gene (+/−) were subjected or not to cold for 48 h and PGC-1α and Mfn2 mRNA levels were measured in skeletal muscle. Under basal conditions, Mfn2 expression was reduced by near 50% in skeletal muscle from heterozygous mice (Figs. 7A and B), whereas COX-II mRNA levels were normal (Figs. 7A and B). Under these conditions, the expression of Mfn2 protein in skeletal muscle extracts from Mfn2^+/− KO mice remained unaltered (data not shown). Exposure to cold caused a 60% increase in Mfn2 mRNA levels in wild-type mice and a greater increase was detected in muscle from Mfn2^+/− KO mice (threefold increase) (Figs. 7A and B). Cold did not cause any increase in COX-II mRNA levels in muscle (Figs. 7A and B).

Levels of PGC-1α mRNA were similar in muscle in both control and Mfn2^+/− KO mice under basal conditions (Fig. 7C). However, exposure to cold caused a much greater increase in PGC-1α expression in muscles from heterozygous Mfn2 KO mice (70% increase) than in control mice (30% increase) (Fig. 7C).

Mfn2 is a protein that shows multiple functions, and some of them are of regulatory nature. Mfn2 is crucial in mitochondrial fusion, and loss of function greatly reduces the extent of the mitochondrial network in several cell types (16,17). Mfn2 also inhibits proliferation in smooth muscle cells (32). In addition, Mfn2 modulates mitochondrial metabolism by regulating mitochondrial membrane potential, fuel oxidation, and the oxidative phosphorylation system (16,21). Thus, in myoblasts with a limited oxidative capacity, Mfn2 gain of function causes an increased rate of glucose oxidation and a parallel increase in mitochondrial membrane potential, a consequence of augmented pyruvate oxidation, Krebs cycle, and oxidative phosphorylation activities in mitochondria (21). In addition, Mfn2 repression decreases the oxidation rates of glucose, pyruvate, and palmitate and reduces mitochondrial membrane potential in myotubes (21). As to the mechanisms involved, the alterations in Mfn2 expression cause a parallel change in the expression of subunits of complexes I, III, and V and do not modify mitochondrial biogenesis (21). In all, available evidence indicates that Mfn2 plays a singular regulatory role in mitochondrial metabolism.

Our data indicate the existence of a regulatory pathway that drives mitochondrial metabolism and is defined by PGC-1α, ERRα, and Mfn2. The pathway is characterized by a stimulatory action of PGC-1α on the transcription of Mfn2, via coactivation of ERRα. This is supported by the following experimental evidence: 1) PGC-1α activates Mfn2 expression in cells, 2) the mechanisms by which PGC-1α stimulates this expression are dependent on an intact ERRα binding in the Mfn2 promoter, 3) Mfn2 regulates mitochondrial metabolism, and 4) PGC-1α action explains the stimulatory effect of cold exposure or treatment with the β₃-adrenergic agonist CL-316243 on Mfn2 expression in muscle and brown adipose tissue.

In addition, we have found that cold-exposed Mfn2^+/− KO mice elevate Mfn2 expression through a greater induction of PGC-1α expression, which suggests a possible Mfn2-induced homeostatic process that is aimed at regulating PGC-1α. The mechanism does not involve a direct effect of Mfn2 because RNA interference–induced repression of Mfn2 did not alter PGC-1α expression in C2C12 muscle cells (M.L., A.Z., unpublished observations).

Mfn2 is a key target of the nuclear coactivator PGC-1α. Previous studies show that the upregulation of PGC-1α enhances total mitochondrial membrane potential in cells by increasing mitochondrial number and also by energization of mitochondria (6,10,33). Mfn2 also enhances mitochondrial membrane potential, and some observations support that Mfn2 stimulates mitochondrial proton leak; the effects of Mfn2 are independent of the mitochondrial mass (16,21). These data support the view that Mfn2 and PGC-1α share common effects in mitochondria. In this report we also provide evidence that the maintenance of a normal expression of Mfn2 is critical for the stimulatory effect of PGC-1α on mitochondrial membrane potential; in contrast, the effects of PGC-1α on mitochondrial biogenesis are independent of Mfn2. These data suggest that the effects of PGC-1α on mitochondrial energization may require or may be mediated by Mfn2. Based on the reported biological roles of Mfn2 (17,18,21,32,34), we also propose that PGC-1α may regulate mitochondrial fusion/fission events and cell proliferation in cells.

Regarding the mechanisms by which PGC-1α stimulates Mfn2 expression, we demonstrate the operation of a transcriptional mechanism that involves the activation of the Mfn2 promoter, which in turn requires the integrity of ERRα-binding elements located at −413/−398. This conclusion is based on a number of observations, namely 1) ERRα activates the transcriptional activity of the Mfn2 promoter, and the effects are synergic with those of PGC-1α; 2) cancellation of ERRα-binding elements block or diminish the effect of PGC-1α on promoter activity; 3) ERRα binds in vitro to the ERRα-binding element −410/−405 and in vivo to a Mfn2 promoter fragment centered around the ERRα-binding element; and 4) PGC-1α binds in vivo the Mfn2 promoter in a region centered around the ERRα-binding element. Thus, Mfn2 can be added to the list of genes recently reported to be regulated by the PGC-1α/ERRα pathway, such as GA repeat-binding protein-α, ERRα, mtTFA, medium-chain acyl-CoA dehydrogenase, ATP synthase, cytochrome c oxidase 5b, isocitrate dehydrogenase, TIM22 (translocase of inner mitochondrial membrane 22), or carnitine/acylcarnitine translocase (13,14,35).

ERRα is induced by PGC-1α in cells (36,37) and is also induced in brown adipose tissue and skeletal muscle after 6 h of cold exposure (36). Based on these observations and on the role of ERRα in Mfn2 gene transcription, we propose that ERRα participates in the regulatory pathway that connects PGC-1α and Mfn2 and regulates mitochondrial function.

Exposure to cold temperature or treatment with the β₃-adrenergic agonist CL-316243 induces PGC-1α, which in turn regulates mitochondrial biogenesis and mitochondrial metabolism in both skeletal muscle and brown adipose tissue (5,28,29). Our data indicate that Mfn2 gene expression is also upregulated in skeletal muscle and brown adipose tissue by cold exposure or by treatment with the β₃-adrenergic agonist CL-316243. The stimulatory effect of cold on Mfn2 expression was similar in skeletal muscle and in brown adipose tissue. However, a distinct profile was detected in brown adipose tissue and skeletal muscle in response to CL-316243 treatment. Although 1 day of treatment was enough to induce Mfn2 in brown adipose tissue, a rapid response was not detected in skeletal muscle, and induction of expression was found after 7 days of treatment. Our observations suggest that the effects of the β₃-adrenergic agonist CL-316243 in skeletal muscle are indirect and independent of cAMP. The stimulation of Mfn2 gene expression in skeletal muscle and brown adipose tissue under conditions of enhanced energy expenditure may play a relevant role in the adaptive regulation of the mitochondrial metabolism aimed at maintaining energy homeostasis.

The regulatory pathway constituted by PGC-1α, ERRα, and Mfn2 may be altered in insulin-resistant conditions and particularly in type 2 diabetes. Thus, it has been reported that Mfn2 and PGC-1α expression is deficient in type 2 diabetic subjects (20,38). In addition, amelioration of insulin sensitivity caused by weight loss in morbidly obese subjects or by acute exercise are associated with increased Mfn2 expression in skeletal muscle (27,39). These data may help to explain the molecular basis for the alterations in mitochondrial function associated with insulin-resistant conditions and type 2 diabetes.

This study was supported by research grants from the Ministerio de Ciencia y Tecnología (SAF 2005-00445), Grant 2005 SGR00947 from the Generalitat de Catalunya, the Fundació Marató de TV3 (300720), and the Instituto de Salud Carlos III RCMN (C03/08), RGDM (G03/212), and RGTO (G03/028). F.X.S. was the recipient of a predoctoral fellowship from the Instituto Danone and from the University of Barcelona, Barcelona, Spain. M.L. is the recipient of a predoctoral fellowship from the Ministerio de Educación y Cultura, Spain.

We thank Tanya Yates for her editorial support, Dr. Diego Haro (University of Barcelona) for his scientific advice, and Drs. Luc Marti, Sara Pich, and Marta Camps for their help in animal studies.