Skeletal muscle DNMT3A plays a necessary role in endurance exercise by regulating oxidative capacity of red muscles

Exercise interventions alter the DNA methylation profile in skeletal muscle, yet little is known about the role of the DNA methylation machinery in exercise capacity. In this study, we found that in oxidative red muscle, DNMT3A expression increases greatly following a bout of endurance exercise. Mice lacking Dnmt3a in skeletal muscle fibers had reduced tolerance to endurance exercise, accompanied by reduced oxidative capacity and reduced mitochondrial counts. Moreover, during exercise, the knockout muscles overproduced reactive oxygen species (ROS), which are major contributors to muscle dysfunction. In mechanistic terms, we demonstrated that Aldh1l1 is a key target of repression by DNMT3A in red muscles. DNMT3A directly regulated the Aldh1l1 transcription by binding to the Aldh1l1 promoter region and altering DNA methylation and histone modification. Enforcing ALDH1L1 expression, leading to elevated NADPH, led to overproduction of ROS by the NADPH oxidase complex (NOX) in myotubes, ultimately resulting in mitochondrial defects. Moreover, both genetic inhibition of ALDH1L1 and pharmacological inhibition of NOX rescued oxidative stress and mitochondrial decline in Dnmt3a-deficient myotubes, confirming the essential role of ALDH1L1-dependent ROS generation as a downstream effector of DNMT3A loss of function. Together, our results reveal that DNMT3A in skeletal muscle plays a pivotal role in endurance exercise by controlling intracellular oxidative stress.


Introduction
Physical exercise maintains health, prevents disease, and even acts as medicine for a wide range of noncommunicable diseases by improving whole-body metabolism and inducing beneficial adaptations in multiple tissues throughout the body [1,2]. Exercise can be classified broadly into two types, endurance and resistance. Endurance exercise (e.g., marathon running and swimming), also called aerobic exercise, is generally characterized by high frequency, long duration, and low power output. By contrast, resistance exercise (e.g., body building and throwing events), which is based on strength, is characterized by activities of low frequency, high resistance, high intensity, and short duration [3,4].
Endurance exercise primarily relies on slow-twitch muscles, such as the soleus, which are also called red muscles due to their high myoglobin content [5]. Red muscles contain more mitochondria than fast-twitch (white) muscles and are more aerobic; consequently, they rely on oxidative metabolism, which favors fatty acids as a fuel source [5]. Accordingly, endurance exercise stimulates mitochondrial biogenesis and the expression of genes involved in mitochondrial respiration and β-oxidation of free fatty acids, driving a phenotypic adaptation toward a more oxidative phenotype [6]. Reduced oxidative potential due to mitochondrial decline or dysfunction is often associated with exercise intolerance and muscle disorders [7- (Supplemental Fig. 1A).

Muscle-specific Dnmt3a deficiency decreases the capacity for endurance exercise
The increase in DNMT3A expression in soleus (red) and GA (mixed) but not in EDL (white) suggested that DNMT3A plays an important role in endurance exercise, which primarily relies on red muscles that are largely dependent on aerobic respiration. To address this question, we used muscle creatine kinase (MCK)-Cre to generate muscle-specific knockout mice (MCK-Dnmt3a KO) (Supplemental Figs. 2A, B).
MCK-Dnmt3a KO mice were viable, exhibited normal growth and fertility, and displayed no significant differences in body weight, body composition, energy balance, or muscle mass relative to wild-type littermates harboring two floxed alleles without Cre (WT) (Supplemental Fig. 3).
To assess tolerance to endurance exercise, we employed two different regimens: (1) a low-intensity regimen (Supplemental Fig. 4A) that tested the ability to run steadily at relatively low speed (12 m/min) for an initial 40 min, followed by a gradual increase in speed until exhaustion [43], and (2) a high-intensity regimen (Supplemental Fig. 4B) that rapidly increased the running speed (6 m/min and increased by 2 m/min every 5 min) to a maximal pace of 30 m/min, which persisted until exhaustion [43,44]. The first regimen preferentially tests fatty acid oxidation capacity, whereas the latter focuses on glucose usage efficiency [43,44]. The running capacity of the MCK-Dnmt3a KO mice was greatly impaired: both distance and duration were reduced by 30-40% under both the low-and high-intensity regimens (Figs. 1A, B).
Cardiac muscle [45,46] plays an important role in exercise by maintaining blood flow to muscles [47].
Because the Cre recombinase in the MCK-Cre line is also expressed in cardiac muscle [45,46], we had to consider a cardiac contribution to the exercise incapacity of the KO mice. To investigate this possibility, we generated an independent line of muscle-specific Dnmt3a KO mice, ACTA1-Dnmt3a KO, using human alpha-skeletal actin (ACTA1-Cre-ERT), which drives Cre expression in skeletal muscle in a tamoxifendependent manner, without expression in cardiac muscle (Supplemental Figs. 5A, B). To achieve musclespecific Dnmt3a KO in this model, we injected tamoxifen (50 mg/kg per day × 5 days, intraperitoneal [i.p.]) into 8-week-old Dnmt3af/f/ACTA1-Cre+ mice; their tamoxifen-treated Cre-littermates were used as controls. After 2 weeks of tamoxifen washing, we subjected the mice to the treadmill regimens to test their exercise capacity. As with MCK-Dnmt3a KO mice, ACTA1-Dnmt3a KO mice displayed an exercise intolerance of a similar magnitude in both regimens (Figs. 1C, D), ruling out a cardiac contribution to their exercise incapacity.
During endurance exercise, metabolism shifts its fuel source from carbohydrates to fatty acids in order to spare glucose for the brain and sustain muscle energy [7,48]. Hence, we investigated whether reduced exercise intolerance was associated with altered fuel utilization. In both KO models, blood glucose levels at rest did not differ from those of controls (Supplemental Figs. 6A, D), indicating that the reduced exercise capacity was not simply due to hypoglycemia. Following exercise, however, KO mice exhibited subtly reduced plasma glucose levels (Supplemental Fig. 6A, D) and elevated lactate levels (Supplemental Fig.   6B, E) relative to WT mice, implying that KO mice favor non-oxidative glycolysis. On the other hand, free fatty acid levels were reduced following exercise relative to controls (Supplemental Figs. 6C, F). Notably, voluntary physical activity of the MCK-Dnmt3a KO mice, as assessed using a CLAMS unit (Supplemental Figs. 3E, F), was not altered, eliminating the possibility that reduced exercise capacity was due to impaired voluntary movement.

Dnmt3a-KO muscle has reduced oxidative capacity and diminished mitochondrial content
High oxidative potential and mitochondrial content of skeletal muscle is critical for supporting endurance exercise. Succinate dehydrogenase (SDH), located in the inner membrane of the mitochondrion, is responsible for oxidizing succinate to fumarate in the citric acid cycle [49]. Hence, we performed SDH staining to distinguish between oxidative and less-oxidative muscles. In KO animals, both soleus and GA muscles exhibited a dramatic (50%) decline in oxidative potential (Figs. 3A-D). Consistent with this observation, mitochondrial copy number was reduced by 43% and 47% in the red soleus and mixed GA muscles, respectively (Fig. 3E), but was unchanged in the white EDL muscle (Fig. 3E). The magnitude of the decrease in mitochondrial content was similar in the soleus and GA muscles, accounting for their reduced oxidative capacity. Consistent with decreased mitochondrial DNA amount, reduced oxygen consumption rate was noted in KO soleus muscles (Figs. 3F, G). Further, we noted that Dnmt3a knockdown in mature L6 myotubes has the same effects on mitochondrial counts and oxygen consumption rates Myosin heavy chains (MHC) are responsible for the contractility of muscle [50,51] and are thus responsible for how aerobic a muscle is. Red muscles like soleus are predominantly enriched in MHCI fibers and some IIa fibers, whose biophysical attributes confer a slow-twitch phenotype, making the muscle more aerobic.
By contrast, white muscles like EDL are enriched in the faster MHCIIb fibers that are responsible for the fast twitch property. To determine whether the change in oxidative potential is due to a change in fiber type distribution in KO muscle, we performed chromatographic ATPase staining, which is more permissive for myosin heavy chain (MHC) I ATPase than for MHCII ATPase activity [52]. Overall intensity of ATPase staining was diminished in KO soleus muscle, but unchanged in KO GA muscle (Figs. 3K, L), suggesting an overall reduction in the oxidative capacity of both muscle fiber types in soleus, rather than fiber-type skewing. Indeed, we observed no discernable change in the expression of muscle fiber type-specific MHC isoforms between WT and KO muscles (Supplemental Fig. 7). Collectively, our data suggest that DNMT3A is required for the full oxidative capacity of skeletal muscle and is not associated with fiber type determination.
Gene profiling studies identify muscle-specific DNMT3A target genes.
We and others have shown that DNMT3A regulates biological processes by regulating non-overlapping sets of cell type-specific target genes [53][54][55]. To elucidate the underlying molecular basis by which Dnmt3a regulates exercise capacity, we performed RNA-Seq on WT and KO soleus muscle. The transcriptome profiles revealed 26 genes that were upregulated, and 4 that were downregulated in Dnmt3adeficient soleus muscle (Figs. 4A, B). These data are consistent with the expectation that DNMT3A works primarily as a repressor of gene expression.
Because a relatively small number of genes were differentially regulated in the KO tissue, we did not identify any specific biological pathways enriched among the DNMT3A target genes. However, genes involved in the detoxification of reactive oxygen species (e.g., Sod1 and Gpx3) were upregulated in KO muscle, implying that the level of oxidative stress is elevated in KO muscles. Indeed, that was the case: In both MCK-and ACTA1-Dnmt3a KO muscles, ROS levels were significantly elevated at rest relative to controls, and the difference grew after exercise (Supplemental Figs. 8A, B). Importantly, Dnmt3a knockdown in L6 myotubes was sufficient to recapitulate the increase in ROS levels in Dnmt3a deficiency (Supplemental Fig. 8C), suggesting a cell-autonomous role for Dnmt3a in this context. Next, we sought to identify the targets responsible for the increase in ROS production. As a candidate, we turned our attention to Aldh1l1, which encodes aldehyde dehydrogenase 1 family member L1 (ALDH1L1), a cytosolic enzyme involved in folate and one-carbon metabolism. Specifically, ALDH1L1 oxidizes 10formyltetrahydrofolate to tetrahydrofolate, producing NADPH as a byproduct [56] (Fig. 4C). Notably in this regard, NADPH plays a dual role in the regulation of oxidative stress. On the one hand, it is a reducing agent for glutathione, thioredoxins, peroxiredoxins, and glutathione peroxidases, which neutralize ROS [57]. On the other hand, it contributes to ROS generation through the activity of the NADPH oxidase complex (NOX) (Fig. 4C), which is located within the sarcoplasmic reticulum, transverse tubules, and sarcolemma in skeletal muscle fibers [22]. We hypothesized that ALDH1L1-mediated NADPH production feeds into NOX, thereby increasing intracellular ROS in Dnmt3a KO muscle.
Prior to functional study, we studied the regulation of Aldh1l1 by DNMT3A. First, we measured the levels of Aldh1l1 mRNA and ALDH1L1 protein were elevated in KO muscle tissues (Figs. 4D-F). Importantly, ALDH1L1 expression was elevated only in the red soleus and mixed GA muscles, but not in the white EDL muscle (Figs. 4E, F), suggesting that DNMT3A-mediated repression of Aldh1l1 is red muscle-specific. To determine whether Aldh1l1 is a direct target of DNMT3A, we conducted methylated DNA immunoprecipitation (MeDIP)-qPCR analysis of the CpG rich Aldh1l1 promoter regions (Fig. 4G). In KO soleus muscle, DNA methylation was greatly reduced in four of the five regions we tested, including the CpG island (P4), (Fig. 4H). In vivo ChIP assay confirmed strong enrichment of DNMT3A at those differentially methylated regions (Fig. 4I). In gene regulation, the DNA methylation and histone regulation machineries often engage in crosstalk [58]. Therefore, we conducted a ChIP assay for H3K27ac, a histone modification marker of active promoters and enhancers, and detected strong signals at Aldh1l1 promoter regions in KO soleus tissues (Fig. 4J). Collectively, these data demonstrate that DNMT3A directly regulates gene expression of Aldh1l1 by modifying the epigenetic profile at its regulatory regions.

ALDH1L1 drives the increase in ROS level and mitochondrial dysfunction in KO soleus muscle
To determine whether elevated ALDH1L1 expression is responsible for NADPH-dependent generation of ROS, we compared NADPH levels in WT and KO muscle. Indeed, NADPH levels were significantly increased in KO muscle (Fig. 5A). To assess whether this resulted in increased flux into NOX, we measured the NADPH consumption rate [59,60] and detected elevated NOX activity in KO tissues (Fig. 5B). Next, to evaluate the functional significance of ALDH1L1, we performed an ALDH1L1 gain-of-function study in L6 rat myotubes in the presence and absence of apocynin, a specific NOX inhibitor [61]. Overexpression of ALDH1L1 had no effect on myogenesis (Supplemental Fig. 9), but remarkably, it was sufficient to recapitulate the redox changes in Dnmt3a-KO muscles, including the increases in the levels of NADPH Importantly, all of these changes associated with ALDH1L1 overexpression were largely reversed by treatment with apocynin (Figs. 5C-H). To prove that ALDH1L1 is essential for the phenotype of the Dnmt3a loss-of-function model, we knocked out Aldh1l1 in Dnmt3a knockdown L6 myotubes, which resulted in dramatic rescue of the oxidative stress and mitochondrial decline caused by Dnmt3a deficiency (Figs. 5I-M).
Collectively, these data suggest that changes in promoter methylation are associated with alterations in gene expression following aerobic exercise. By the same token, genome-wide studies have reported that both acute and chronic exercise interventions produce profound changes in CpG methylation [29][30][31][32][33][34][35][36][37]. Some of the alterations are concordant with changes in gene expression [29,34,37,40]. Despite the apparent connection between altered DNA methylation and exercise, the underlying roles of DNMTs in exercise performance remain unclear. In this study, using two independent genetic models of muscle-specific Dnmt3a-KO mice, we showed that DNMT3A plays an essential role in endurance exercise.
A key link between exercise and control of mitochondrial biogenesis was revealed by the observation that PGC-1α expression is transiently induced in skeletal muscle following an acute bout of exercise [82]. Since that discovery, a great deal of research effort has been devoted to elucidating the role of PGC1A in skeletal muscle mitochondria biology and exercise. For example, transgenic expression of PGC1A increases mitochondrial content and function, increases the abundance of oxidative type I muscle fibers, and decreases muscle fatigue. However, loss of PGC1A has only a mild effect on exercise capacity and does not alter fiber-type composition in muscle [83] or affect training-induced increases in the expression of genes involved in oxidative phosphorylation [84]. This suggests that PGC1A is sufficient, but not necessary, to mediate metabolic adaptations in response to exercise. In this study, we found that DNMT3A is required for mitochondria and metabolic adaptations in red skeletal muscle. Because we did not detect obvious changes in the expression of PGC1A (not shown), we speculate that the role of DNMT3A in the regulation of mitochondrial function and mass is likely to be PGC1A-independent.
Overproduction of ROS induced by unaccustomed, exhaustive exercise training or other stresses can lead to oxidative stress-related tissue damage and reduced contractility [85,86]. NADPH oxidases are major contributors to ROS production [87,88]. Previous work showed that physical stretching can increase the activity of NADPH oxidase, especially NOX2, to produce ROS in microtubule-dependent process [89].
That study described mechanotransduction-dependent activation of NADPH in cardiac muscle; here, by contrast, we identified ALDH1L1-dependent activation of NADPH oxidase in skeletal muscles. Our finding that inhibition of NADPH oxidase rescued both oxidative stress and decline in mitochondrial content raises the possibility of repurposing inhibitors to improve exercise trainability.
Although we focused on Aldh1l1, DNMT3A has other target genes worthy of additional attention. Indeed, candidate genes revealed by our transcriptomics analysis suggest that several secretory factors involved in fibrosis and inflammation are up-regulated in Dnmt3a-deficient muscles. For example, Spp1 encodes osteopontin (OPN), also referred to as bone sialoprotein 1 (BSP-1), which contributes to fibrosis, inflammation, and collagen deposition in aged animals and models of muscle dystrophy, e.g., Mdx mice [90,91]. Additionally, we noted that Ccl21c and Ccl21d, both of which encode the proinflammatory chemokine CCL21, were upregulated in Dnmt3a-deficient muscles. Notably in this regard, both OPN and CCL21 promote migration of immune cells such as macrophages, neutrophils, or T cells [91][92][93].
Therefore, it is conceivable that elevated expression of these factors may slow down recovery from myopathy in Dnmt3a-deficient muscle following exercise or muscle damage. Future studies should address these possibilities.
Here, we highlighted the surprising role of DNMT3A in endurance exercise, skeletal muscle mitochondrial biology, and energy utilization. In mechanistic terms, we revealed that ALDH1L1 serves as a novel molecular link that contributes to oxidative stress and mitochondrial dysfunction following the loss of Dnmt3a in red muscle. This is of great importance from the standpoint of exercise physiology, as physical activity is strongly encouraged as a key strategy for preventing and treating a wide range of human diseases.
Understanding the epigenetic and molecular basis of exercise tolerance will help us to address several critical health issues that arise due to lack of exercise.

Animals
Animal Care Mice were maintained under a 12-hr light /12-hr dark cycle at constant temperature (23°C) with free access to food and water. All mice were extensively back-crossed onto a C57Bl/6J background. All animal work was approved by UC Berkeley and the University of IOWA ACUC. In vivo assays were done with 7-to 20-week-old littermate male mice.

Measurement of exercise capacity
All mice were acclimated to the treadmill prior to the exercise test session. For each session, food was removed 2 hr before exercise. Acclimation began at a low speed of 5 to 8 meters per minute (m/min) for a total of 10 min on Day 1 and was increased to 5 to 10 m/min for a total of 10 min on Day 2. The experiments were performed on Day 3. For the low intensity treadmill test, the treadmill began at a rate of 12 m/min for 40 min. After 40 min, the treadmill speed was increased at a rate of 1 m/min every 10 min for a total of 30 min, and then increased at the rate of 1 m/min every 5 min until the mice were exhausted. The high intensity treadmill test was conducted on the same open-field six-lane treadmill set at a 10% incline. Following a 5-min 0 m/min acclimation period, the speed was raised to 6 m/min and increased by 2 m/min every 5 min to a maximal pace of 30 m/min until exhaustion. Mice were considered exhausted when they were unable to respond to continued prodding with a soft brush.

Grip strength assessment
Grip strength was measured either with a wire or inverted wire mesh grid. For wire hanging, each mouse was placed with their forelimbs on a 2 mm wire that was tied from end to end. The hang time was measured for each mouse until they dropped down. For the string test, each mouse was placed with their forelimbs on a 2 mm wire that was tied from end to end. The time taken for mice to hold on to the wire with their hindlimbs was noted. Mice that failed to hold on with hind limbs were not included in the calculation. For the cage top test, mice were placed on top of an elevated wire mesh grid, the screen was inverted, and the animals were timed until they let go of the grid.

RNA extraction and quantitative PCR
Total RNA was extracted from tissues using TRIzol reagent according to the manufacturer's instructions. cDNA was reverse transcribed from 1 μg of RNA using the cDNA Reverse Transcription Kit (Applied Biosystems). Quantitative PCR (qPCR) was performed with SYBR green qPCR master mix (AccuPower 2X, Bioneer) using a CFX96 Touch (Bio Rad). The relative amount of mRNA normalized to cyclophilin B was calculated using the delta-delta method. Primer sequences are listed in Supplemental Table 1.

Western blot analysis and antibodies.
Whole-cell protein lysates were prepared according to the manufacturer's protocol using RIPA lysis buffer and protease inhibitor cocktail. Proteins were size fractionated by SDS-PAGE and then transferred to polyvinylidene difluoride membranes. After blocking with 5% nonfat dried milk in TBS-Tween (0.25%), the membranes were incubated with the appropriate primary antibodies against antiDNMT3A antibodies (#2160S). The loading control included anti-GAPDH (#2118) and anti-Histone H3 (#14269). Immunoblots were quantified by ImageJ.

Immunoprecipitation
For immunoprecipitation, tissues were lysed with 300 μl lysis buffer, containing 50 mM Tris-HCl (pH7.5), 150 mM NaCl, 1% Triton X-100, and protease inhibitor cocktail for 30 min on rotation. The lysates were centrifuged at 13,000 rpm for 10 min at 4°C, then supernatants were incubated with antiDNMT3A antibodies (#2160S) overnight then with protein A/G PLUS-Agarose (#sc-2003) at 4°C for 2 hr. The beads were washed four times with lysis buffer. Bound proteins were eluted by boiling in SDS sample buffer and processed for western blot.

Genomic DNA extraction
Snap-frozen tissues were treated with proteinase K overnight at 56°C. DNA was extracted by standard phenol chloroform-ethanol precipitation and eluted in DNase-free water.

Histological analysis of skeletal and cardiac muscle fibers
Harvested skeletal muscles were immediately embedded in T.F.M. compound (Tissue-Tek) and snap frozen using a Stand-Alone Gentle Jane (Instrumedics Inc.). We then prepared 10 μm sections from the muscle mid-belly using a Leica cryostat. Hematoxylin and eosin (H&E) staining was performed following fixation in ice-cold zinc formalin (Anatech Ltd. #175) for 60 min. Image analysis was performed using ImageJ software. For ATPase stains, slow-type fibers were dark, whereas fast-type fibers were lightly stained following preincubation at pH 4.35. Fiber type was identified on the basis of color for each myofiber. For Sirius Red staining, cardiac sections were then incubated in Bouin's' solution (5% acetic acid, 9% formaldehyde, and 0.9% picric acid) at room temperature for 1 hr. Next, after washing, slides were incubated in 0.1% Fast Green (Fisher, F-99) for 3 min, then in 0.1% Sirius Red (Direct red 80, Sigma, 0-03035) for 2 min. After three washes, slides were dehydrated with ethanol and xylene based on standard procedures.

Mitochondrial DNA copy number
Mitochondrial DNA (mtDNA) copy number was used as a marker for mitochondrial density. Briefly, after isolating genomic DNA from muscle tissues, mtDNA was quantified using quantitative RT-PCR. The mitochondrial DNA copy number was calculated from the ratio of COX II (mitochondrial-encoded gene)/cyclophilin A (nuclear-encoded gene).

Measurement of plasma parameters
Mice were either kept at resting or allowed to run for 50 min on the low-intensity regime. Following this, blood was collected from the tail tip for WT and MCK-Dnmt3a KO mice from both groups. Blood plasma was separated by centrifugation for 15 min at 4,000 rpm. Plasma nonesterified fatty acids were measured using the NEFA-HR2 Assay Kit (Wako Diagnostics). Lactate levels in the plasma were measured using a Lactate Colorimetric Assay Kit (Abcam).

Reactive oxygen species generation
Intracellular accumulation of reactive oxygen species (ROS) was monitored using OxiSelect™ Hydrogen Peroxide/Peroxidase Assay Kit (Cell Biolabs, Inc., San Diego, CA). To investigate ROS generation, the muscle tissues were homogenized in 1× assay buffer provided from the kit. The lysates were assayed according to the manufacturer's procedure.

RNA-Seq library generation and analysis
RNA samples were extracted using the RNeasy Mini kit (Qiagen, 74104), and the quality of total RNA was assessed by the 2100 Bioanalyzer (Agilent) and agarose gel electrophoresis. Libraries were prepared using the BGI Library Preparation Kit, and sequencing was performed on the BGISEQ (BGI, China). RNA-Seq reads were aligned to the UCSC mm10 genome using HISAT2 (Hierarchical Indexing for Spliced Alignment of Transcripts [94]), and mapping was done using Bowtie2 [94]. Differentially regulated genes were calculated using DEseq2 [95].

Measurement of NADPH
The NADPH measurement was performed on GA/Soleus extracts using the NADP/NADPH assay kit (Cat#K347, BioVision) according to the manufacturer's instructions. Briefly, ~20 mg samples were extracted in 400 µL of the given extraction buffer, and 50 µL was processed following instructions of the kit. OD450 measurements were made on a plate-reader (SpectraMAX i3 Plate reader) at 25°C, and the data was calculated using a standard curve.

Measurement of NADPH oxidase activity NOX activity was measured by accessing oxidation of NADPH through Continuous Spectrophotometric
Rate Determination [59,60]. Briefly, samples of tissue or cells were extracted in 500 µL and 200µL of potassium phosphate buffer, respectively. The tissues and cells were first homogenized and then sonicated. The homogenate was centrifuged at top speed for 10 min, and the supernatant was used for reading. Oxidation of NADPH was monitored at 340 nm on the SpectraMAX i3 Plate Reader at 30°C [59,60].

MeDIP-qPCR
Genomic DNA was sheared using a Covaris S220 to an average of 200-600 bp. 600 ng of denatured DNA was incubated with 2 mg of 5-methylcytosine Cell culture L6 rat myoblasts were obtained from UCSF and were cultured in Dulbecco's modified Eagle's medium (DMEM) supplemented with 10% fetal bovine serum. Differentiation was carried out in Dulbecco's modified Eagle's medium (DMEM) supplemented with 10% horse serum, 100 U/ml penicillin, and 100 μg/ml streptomycin. Myocytes were maintained in a humidified incubator under an atmosphere of 5% CO2 at 37°C. To generate lentivirus particles, lentiviral constructs were co-transfected with pMD2.G-and psPAX2-expressing plasmids into 293T cells. After 48 h, the virus-containing supernatant was collected, filtered through 0.45-mm filters, and added to mature L6 myotubes for 48 h along with 8 mg/ml polybrene. Transduction efficiency was determined by comparing to cells transduced in parallel with a GFP-expressing lentivirus, and myotubes were treated with 100 μM Apocynin (Tocris, Cat. No. 4663) for 24 hours.

Metabolic cages
Indirect calorimetry and food intake, as well as locomotor activity, were measured using the Comprehensive Lab Animal Monitoring System (CLAMS) (Columbus Instruments). The calorimetry system is an opencircuit system that determines O2 consumption, CO2 production, and RER. Data were collected after 3 hr of adaptation in acclimated singly housed mice.

Statistical Analysis
Data from independent experiments are presented as the mean ± SE. Statistical differences were assessed using PRISM4 software (GraphPad). Unpaired two-tailed Student's t tests and two-way ANOVA were used. p < 0.05 was considered statistically significant.
Aldh1l1-P5_R CCTCCAGGCAGAGAAGAGG    B) and GA (C, D) muscles after a bout of low-intensity exercise for 50 min (10X, 20X magnifications), and the staining intensity was quantified using ImageJ (n = 2, p < 0.05, Student's t-test, mean ± s.e.m.). (E) Genomic DNA was extracted from various muscle types from WT and KO mice after a bout of lowintensity exercise for 50 min. Mitochondrial DNA copy-number was calculated from the ratio of mitochondria-encoded COXII to nuclear-encoded cyclophilin A (n = 6, p < 0.05, Student's t-test, mean ± s.e.m.). (F, G) Mitochondrial respiration was measured in WT and KO soleus tissue after a bout of lowintensity exercise for 50 min under basal conditions and in response to 1.5 mM oligomycin (complex V inhibitor), 4 mM FCCP (uncoupler), or 2 mM rotenone (complex I inhibitor) (n = 3, p < 0.05, Student's ttest, mean ± s.e.m.). (H-J) L6 myotubes were transduced with lentiviral expression plasmids for Flag-ALDH1L1 and GFP and mitochondrial DNA copy number and mitochondrial respiration was was measured from these cells (n = 6, p<0.05, Student's t-test, mean ±s.e.m.). (K, L) ATPase staining was conducted in soleus (K) and GA muscles (L), (type I: dark blue; type IIa: lightest blue; type IIb: light blue, 40X magnification). H3K27ac ChIP-PCR was conducted in WT and KO soleus muscles (n = 3, p < 0.05, Student's t-test, mean ± s.e.m.).