ABSTRACT
Muscle wasting (or atrophy), defined by reduced myofiber size and muscle strength, occurs in primary neuromuscular diseases and during aging. The progressive loss of muscle mass with aging is known as sarcopenia, and affects 35.4% and 75.5% of women and men over 60 years of age, respectively 1. Mitochondrial dysfunction has been proposed to contribute to progressive muscle wasting 2, 3. Interestingly, substantial levels of bioenergetic deficiency and oxidative stress seem to be insufficient by themselves to intrinsically cause muscle wasting during aging4. This raises the possibility that mitochondria may affect muscle mass by additional mechanisms. Here, we show that chronic adaptations to mitochondrial protein overload cause muscle wasting. Protein overloading, together with conditions that directly and indirectly affect the protein import machinery, are known triggers of mitochondrial Precursor Over-accumulation Stress (mPOS), a newly discovered cellular stress mechanism caused by toxic accumulation of precursor proteins in the cytosol 5–7. Our evidence was obtained from transgenic mice that were generated to have a two-fold increase in the nuclear-encoded Ant1 protein that is involved in ATP/ADP exchange on the inner mitochondrial membrane (IMM). These animals progressively lost muscle mass with age, although their lifespan was unaffected. At two years of age, the skeletal muscle of these mice was severely atrophic. The ANT1-transgenic muscles have a drastically remodeled transcriptome that appears trying to counteract mPOS, by repressing protein synthesis and stimulating proteasomal function, autophagy, lysosomal amplification and Gadd45a-signaling. These processes are all known to promote muscle wasting 8, 9. Thus, chronic proteostatic adaptation to mPOS is a robust mechanism of muscle wasting. These findings may help improve the understanding of how mitochondria contribute to muscle wasting during aging. They may also have direct implications for human diseases associated with ANT1 overexpression10–12.
RESULTS AND DISCUSSION
Our recent studies showed that overloading of mitochondria with carrier proteins leads to the accumulation of unimported mitochondrial proteins in the cytosol, followed by proteostatic stress (or mPOS) and cell death 5. However, it is unknown whether mPOS can occur in vivo and whether it causes pathology. To address this, we prepared transgenic mice overexpressing ANT1, encoding an adenine nucleotide carrier on the IMM. An earlier study reported the use of transgenic mice expressing ANT1 cDNA from the human α-skeletal actin promoter to model Facioscapulohumeral Dystrophy (FSHD)13. These mice did not develop muscle pathology. However, that study failed to examine whether or not the Ant1 protein was actually overexpressed compared to control animals, which was a considerable limitation given that Ant1 is highly abundant in muscle mitochondria14. Thus, whether ANT1 overexpression does, or does not, cause muscle pathology remains undetermined. To address this, we generated ANT1-transgenic mice using the ANT1 genomic locus and its native promoter (Fig. 1a). The resulting hemizygous transgenic mice (hereafter referred to as ANT1Tg/+) are kyphotic, and have reduced body weight, with more pronounced effects in males than in females (Extended Data Fig. 1a-1d). We mainly focused our studies on a specific transgenic line, in which Ant1 levels are increased by only two-fold in the skeletal muscles (Fig. 1b and 1c).
The ANT1Tg/+ mice have a normal body length (Fig. 1d). They have increased spinal curvature (Fig. 1e), decreased home cage activity, and reduced exercise tolerance (Extended Data Fig. 1f–1m), which suggests muscle weakness. Gait abnormalities become noticeable from the age of 6 months and progressively deteriorate during aging (Fig. 1f; Supplemental video 1-3). Parallel to reduced body weight (Fig. 1g), the female and male mice lose lean mass by 24.18% and 26.64% respectively at 6 months, and by 35.01% and 36.28% respectively at 12 months of age (Fig. 1h). Skeletal muscle becomes severely atrophic at two years of age (Fig. 1i), while the overall lifespan is unaffected (Extended Data Fig. 1e). Supporting muscle wasting, the ANT1Tg/+ mice have increased myofiber size variability (Fig. 1j), with an overall reduction in average myofiber diameter of 26.18% and 49.46% at 12 and 20 months of age, respectively (Fig. 1k; Extended Data Fig. 2a and 2b). Round- and angular-shaped myofibers, and mild increase of endomysial connective tissues were observed in the ANT1-transgenic muscles (Fig. 1l and Extended Data Fig. 2c–2f). No myofiber type grouping suggestive of a chronic neuropathic process was observed. Furthermore, the ANT1Tg/+ muscles have increased basophilic stippling, suggesting increased acidic cellular components (Extended Data Fig. 3a). Moth-eaten myofibers were detected in the ANT1Tg/+ muscles, as indicated by poor staining of NADH, succinate dehydrogenase (SDH) and cytochrome c oxidase (COX) (Fig. 2a). The level of the proapoptotic Bax protein is elevated but without an increase in the percentage of central nuclei (Extended Data Fig. 3b–3d). Taken together, we conclude that ANT1-overexpression likely causes progressive muscle wasting mainly because of processes that lead to reduced myofiber size.
We found that Ant1 overexpression has a moderate effect on mitochondrial function. The assembly of respiratory complexes and supercomplexes is not noticeably affected (Fig. 2b and 2c). The state-3 and state-4 respiratory rates are only moderately reduced (Fig. 2d). The overall respiratory control ratio is unaffected, suggesting that moderate ANT1-overexpression is insufficient to uncouple mitochondrial respiration (Fig. 2e). Previous studies showed that more severe bioenergetic deficiencies are not sufficient to induce muscle wasting during aging, raising the possibility that additional factors likely cause ANT1-induced muscle pathology.
To understand how ANT1 induces muscle wasting, we analyzed the transcriptome of ANT1Tg/+ skeletal muscles. We identified 4,947 genes expressed at reliable levels (Fig. 3a), among which 1,016 genes were differentially expressed in the ANT1Tg/+ versus wild-type (WT) muscles (two-way ANOVA, q < 0.05; Supplemental Table 1). Among these genes, 285 genes were upregulated and 78 genes down-regulated at the ±1.5 fold level, and 38 genes were up-regulated and 6 down-regulated at the ±5 fold level (Fig. 3b). Interestingly, out of the top 20 genes previously identified as upregulated in sarcopenia15, we found 6 are activated in the ANT1Tg/+ muscles (Fig. 3c; Supplemental Table 2), suggesting that the ANT1-overexpressing and aging muscles share at least some overlapping stresses. Consistent with our evidence of a mild bioenergetic defect, genes involved in oxidative phosphorylation and anti-oxidant defense were not over-represented among the significantly upregulated genes. Instead, genes for Integrated Stress Response (ISR, Fig. 3d), One-Carbon Metabolism and the Starvation Response are upregulated (Supplemental Table 3). These transcriptional signatures have previously been reported in mitochondrial disease models16–18. Phosphorylation of eIF2α, the central player in the ISR, was increased in ANT1Tg/+ muscles (Fig. 3e). This eIF2α phosphorylation is likely responsible for the upregulation of ATF4, ATF5, DDIT3 and NUPR1 (Fig. 3d), by promoting cap-independent translation of these transcriptional factors followed by a feed-forward loop of transcriptional activation19. We found that eIF2α phosphorylation is elevated in aged muscles not only in ANT1Tg/+ but also wild-type mice (Extended Data Fig. 4). Phosphorylated eIF2α globally attenuates cap-dependent translation in response to various stresses including mitochondrial damage20. Thus, our data support a global proteostatic stress state in ANT1Tg/+ muscles, and perhaps also in the physiologically aged muscles. Additionally, we observed the upregulation of genes for protein import and mitochondrial chaperones (Fig. 3g). This is reminiscent of the mitochondrial Compromised Protein import Response (mitoCPR) and the mitochondrial Unfolded Protein Response (mtUPR) mechanisms, which are important for improving protein import and mitochondrial homeostasis 21, 22. These data show that ANT1-overexpression induces both protein import and proteostatic stresses in mitochondria, and proteostatic stress in the cytosol (mPOS).
More importantly, RNA-Seq also revealed other proteostatic responses in ANT1Tg/+ muscles in addition to eIF2α phosphorylation, with some of them being downstream targets of the ISR. First, expression of several cytosolic chaperones and the heat shock transcriptional factor 1 (HSF1) are increased (Extended Data Fig. 5a). Secondly, we found that the phosphorylation of Rps6 kinase (S6K) is not decreased in ANT1Tg/+ muscles, suggesting that mTOR-signaling is not attenuated (Extended Data Fig. 5b & 5c). Interestingly, the transcription of EIF4EBP1 (or 4EBP1), a downstream target of the ISR23, is increased (Fig. 4a). Consequently, both 4EBP1 and its phosphorylated form, P-4EBP1, are increased (Fig. 4b–4d). The non-phosphorylated 4EBP is a repressor of protein translation. Thirdly, we found that the overall levels of ubiquitinated proteins, and p62 that promotes the formation and removal of ubiquitinated and aggregated proteins, are not increased in the ANT1Tg/+ muscles (Extended Data Fig. 5d–5f). On the other hand, genes encoding proteasomal subunits, and NFE2L1 and NFE2L2 that activate the proteasomal genes, are transcriptionally upregulated (Extended Data Fig. 5g & 5h). Accordingly, proteasome-associated trypsin-like activity is increased in the ANT1Tg/+ muscles (Extended Data Fig.5i). The data support an activation of proteasomal function, which helps preventing the accumulation of ubiquitinated proteins. Fourthly, we found that numerous genes involved in autophagy and intracellular vesicle trafficking are upregulated in ANT1Tg/+ muscles (Fig. 4e & 4f). This includes SESN2 that is directly activated by the ISR24.
Finally, we found that the transcription of CTSL, encoding the lysosomal cathepsin L protease, is upregulated (Fig. 4g). In parallel, we observed a strong amplification of Lamp2-possitive lysosomes and/or lysosome-derived structures in the ANT1Tg/+ muscles (Fig. 4h; Extended Data Fig. 6; Extended Data video 4-5). Transmission electron microscopy revealed many autophagy-related structures of varying morphologies that are absent in WT muscles. Notably, aggresome-like structures were detected in the cytosol of ANT1Tg/+ muscles that contain electron-dense material resembling protein aggregates (Fig. 4i). Also detected were multivesicular vacuoles, multilamellar vesicles, flattened multilamellar membranes, and mitophagic and glycophagic bodies (Extended Data Fig. 7 & 8). These observations support hyper-activation of autophagic and lysosomal activities. Collectively, the data strongly support the model that ANT1 overexpression induces cytosolic protein misfolding/aggregation and mPOS. Muscle cells respond to mPOS by activating multiple proteostatic pathways to rebalance cytosolic proteostasis.
The observation that moderate overloading of the mitochondrial adenine nucleotide carrier is sufficient to cause mPOS was independently confirmed in yeast. We generated a yeast strain that overexpresses AAC2, the mammalian homolog of ANT1, by 2.2 fold (Extended Data Fig. 9a). Like in ANT1Tg/+ muscles, AAC2-overexpression reduces mitochondrial respiration (Extended Data Fig. 9b & 9c). AAC2-overexpression also reduces cell growth. More importantly, cell growth is inhibited even on the fermentable carbon sources when combined with the disruption of YME1, PHB1 and PHB2 (Extended Data Fig. 9d). YME1 encodes the i-AAA protein quality control protease on the IMM. Phb1 and Phb2 form the prohibitin complex required for the stabilization of newly synthesized IMM proteins. Disruption of YME1, PHB1 and PHB2 is known to cause mPOS6. The genetic interactions suggest that AAC2-overexpression induces mPOS, which synergizes with yme1Δ, phb1Δ, and phb2Δ to inhibit cell growth. We found that aac2 alleles with transport-defective mutations also inhibit the growth of yme1Δ cells (Extended Data Fig. 9e). mPOS induction by AAC2-overexpression is therefore independent of nucleotide transport. Consistent with mPOS, AAC2-overexpressing cells cannot tolerate a further proteostatic challenge in the cytosol instigated by the expression of the aggregation-prone human huntingtin protein, Htt(25Q) (Extended Data Fig. 9f). We speculate that Aac2 overloading partially saturates the cytosolic chaperoning system and/or the protein import machinery, which globally reduces mitochondrial protein import and ultimately causes mPOS (Extended Data Fig. 9g). Likewise, moderate ANT1-overexpression in mice may also cause mPOS in the ANT1Tg/+ muscles by the same mechanism.
In the ANT1Tg/+ muscles, we found that the protein content is significantly reduced (Fig. 4k). We propose that chronic proteostatic adaptations to mPOS, which include increased 4EBP expression and eIF2α phosphorylation, and the activation of proteasomal, autophagic and lysosomal functions, converge to cause an imbalance between protein synthesis and degradation (Fig. 4l). Unbalanced protein synthesis and degradation is a known driver of myofiber shrinkage and muscle wasting8. In addition to these pro-atrophic pathways, our RNA-Seq analysis also showed that GADD45A and CDKN1A, two downstream targets of the ISR 25, are upregulated in the ANT1Tg/+ muscles (Fig. 4j). GADD45A and CDKN1A encode the growth arrest and DNA-damage-inducible protein α, and the cyclin-dependent kinase inhibitor 1A (or p21) respectively. Both proteins promote muscle wasting 9, 26. On the other hand, FBXO32 and TRIM63, encoding the MAFBx/atrogin-1 and MuRF1 ubiquitin ligases that are upregulated in acute models of muscle atrophy8, are instead down-regulated in ANT1Tg/+ muscles (Extended Data Fig. 5j). Thus, ANT1-induced progressive muscle wasting is mechanistically linked to the chronic activation of adaptive proteostatic programs without the involvement of FBXO32 and TRIM63.
In summary, we found that moderate ANT1-overexpression is sufficient to cause progressive muscle wasting. Mitochondrial protein import appears to be readily saturable in the skeletal muscle. Chronic proteostatic adaptation to mitochondrial protein overloading and the resulting mPOS appears to be a robust mechanism of muscle wasting. Protein import is an elaborate process 27. It is impaired during aging and in those muscle diseases that directly or indirectly affect the import machinery 28, 29. Our finding could therefore help the understanding of how mitochondrial damage during aging and in import-related diseases affects muscle mass and function. In addition, our work may also have implications for FSHD and dilated cardiomyopathy that are associated with ANT1 activation 10–12. FSHD is primarily linked to the ectopic expression in skeletal muscle of DUX4, a gene that encodes a transcription factor not normally expressed in somatic tissue 30. In light of our study, it might be worth reexamining whether or not ANT1 overexpression contributes to or modifies the course of this disease.
Supplemental Information
contains supplementary table 1-3 and video 1-5.
Author Contributions
X.W., X.J.C. and R.T. performed the experiments. F.M. processed and analyzed the RNA-Seq data. X.W. and X.J.C. wrote the manuscript.
The authors declare no competing financial interests.
METHODS
Antibodies for western blot and immunofluorescence analysis
Antibodies used in this study included the Total OXPHOS Human WB Antibody Cocktail (#ab1104a1, Abcam), anti-Ant1 (#SAB2108761, Sigma), anti-Bax (#ab32503, Abcam), anti-eIF2α (#9722, Cell Signaling), anti-phospho-eIF2α (Ser51) (#3597, Cell Signaling), anti-4EBP1 (#9644, Cell Signaling), anti-Lamp2 (#GL2A7, Developmental Studies Hybridoma Bank, University of Iowa), anti-phospho-4EBP1 (#2855, Cell Signaling), anti-p62 (#23214, Cell Signaling), anti-rpS6 (#ab40820, Abcam), anti-phospho-rpS6 (#ab65748, Abcam), anti-ubiquitin (#701339, Thermo Scientific). Total proteins were stained with REVERT Total Protein Stain (#926-11011, LI-COR). For western blot, band intensities were quantified by the LI-COR Odyssey Fc Imaging system and normalized against total protein staining.
Yeast strains and growth media
All the Saccharomyces cerevisiae strains used are derived from the BY4741/BY4742 background. The GAL10-AAC2 cassette was integrated into the chromosome and AAC2 expression was induced in complete medium containing 2% galactose (Gal) and 2% raffinose (Raf). To test AAC2 toxicity, the wild-type and mutant AAC2 alleles were expressed from the multicopy vector pRS425 (LEU2). The plasmids were introduced into yme1Δ cells by LiAc-mediated transformation. The transformed cells were plated onto minimal glucose medium lacking leucine and selected for Leu+ colonies after incubating at 30 ° C for three days. To test the synthetic growth defect between AAC2 and HTT(25Q), the strain BY/GAL-AAC2 (GAL10-AAC2) was transformed with a URA3-based plasmid expressing HTT(25Q) under the control of the GAL10 promoter. The Ura+ transformants were selected on minimal glucose medium lacking uracil, and were then tested on minimal galactose medium lacking uracil for growth at 25°C for four days.
Generation of ANT1-transgenic mice
The Ant1 transgene was prepared by recombineering according to Lee et al.31. Briefly, genomic sequences corresponding to Ant1 5’ upstream and 3’ downstream genomic sequence were prepared by PCR using primer pairs ret5F (5’-GTCGAATTC GTATATAAATAAATAAAAGAAAG) and ret5R (5’-GTCAGACGTC CAATGTTGCTACTTAAACACTCTTG) and ret3F (5’ GTCA GACGTC CCTTGAGAACTAACACAGAGCAG) and ret3R (5’ GTCAAAGCTT GGTGATATGGGGACAGGAAGGAG). The 5’ and 3’ mini-arms were digested with EcoRI and AatII, and HindIII and AatII, respectively, and then inserted into pSK+. The mini-arm vector was then digested with AatII, treated with phosphatase, and then electroporated into EL350 together with a BAC clone, RP24-108A1, which contains the entire Ant1 genomic sequence, to retrieve Ant1 genomic sequence from the BAC by gap repair. The retrieved Ant1 genomic sequence is 13.7 kb in size containing all 4 exons together with 4.2 kb and 4.96 kb of 5’ up- and 3’ down-stream sequences, respectively. The Ant1 transgene fragment was released from the plasmid by NotI and KpnI digestion, fractionated by agarose gel electrophoresis, column purified and resuspended in 10mM Tris, pH8.0.
ANT1-transgenic mice were generated according to Nagy et al.32. Briefly, four to six week old C57BL6j females were superovulated by first adminstration of 5 IU of pregnant mare serum (PMS) and 44 to 46 hour later with 5 IU human chorionic gonadotropins (HCG). The females are then mated with C57BL6j stud males. Fertilized one-cell embryos were collected from the oviduct the next day and maintained in KSOM medium in an air/CO2 incubator at 37°C until use. Purified Ant1 genomic fragment were diluted to approximately 1 ng/μl and then microinjected into the pronuclei of the one-cell embryos. Injected embryos were implanted into the oviducts of 0.5 days post coitium pseudopregnant females for subsequent development in utero. Transgenic mice were identified by PCR using genomic DNA prepared from ear notch as template. Two primer pairs specific to the 5’ and 3’ end of the Ant1 transgene were used for genotyping. Primer pair BHtoRI (5’-GGATCCCCCGGGCTGCAGGAATTC) and Ant R5 (5’-CAATGTTGCTACTTAAACACTCTTG), corresponds to the 5’ will identify a fragment of 515 bp, and a second pair of primers, XhotoHind (5’-GAGGTCGACGGTATCGATAAGCTTG) and AntR3F (5’-CTTTCCTGGACCCCTGTAAGCTTG) will detect a fragment of 432 bp specific to the 3’ end of the transgene. Founder animals were then mated with C57BL6NTac mice to establish and maintain the transgenic line. All the animal experiments have been approved by the Institutional Animal Care and Use Committee (IACUC) of the State University of New York Upstate Medical University. The experiments were performed using age-matched and mostly littermate controls. The animals were fed with a regular chow diet ad libitum and were housed at an ambient temperature. Food and water were placed at a low and reachable position for the ANT1-transgenic mice over one-year-old.
Lean mass, home cage activity and exercise endurance tests
Lean mass and home cage activity were measured by quantitative magnetic resonance and Promethion metabolic caging respectively at Vanderbilt University Mouse Metabolic Phenotyping Center. Exercise endurance was measured with an Exer 3/6 Treadmill (Columbus Instruments) by determining the maximum distance and speed of the animals before exhaustion. The animals were trained for at least two sessions before the tests.
Mitochondrial isolation and bioenergetic assay
Unanesthetized mice were decapitated using a guillotine. Muscle mitochondria were isolated according to Garcia-Cazabin et al. 33. Oxygen consumption rates were measured using an Oxygraph Plus oxygen electrode (Hansatech Instruments Ltd), with 150 μg mitochondria, 5 mM glutamate, 2.5 mM malate, 150 μM ADP. Oligomycin (5 μg/ml) was added to established state-4 respiration. Respiratory control ratio was established by dividing state-3 respiratory rate with oligomycin-inhibited respiratory rate.
Muscle histology and immunohistochemistry
Standard procedures were followed for H&E, Trichrome, NADH, SDH and COX activity staining, and for the determination of lesser diameter and variability coefficient of muscle fibers 34. For immunofluorescence microscopy, muscle sections of 10 μm thickness were fixed and permeabilized with 100% methanol at −20°C for 15 minutes, rinsed in PBS, treated in blocking buffer (1XPBS/5% Normal Goat Serum/0.3% Triton X-100) for one hour at room temperature, followed by incubation with anti-Lamp2 antibody at 4 ° C overnight. The specimens were then washed in 1XPBS/0.1% Tween 20 before being probed with an Alexa Fluor 488-conjugated anti-rat IgG (H+L) antibody for one hour at room temperature. After washing with 1XPBS/0.1% Tween 20, the tissue samples were mounted with ProLong Diamond Antifade Mountant with DAPI (#P36962, Invitrogen), and were visualized using a Leica SP8 confocal microscope.
Electron microscopy
Fresh muscle samples were fixed in 4% glutaraldehyde/0.1 M cacodylate buffer, pH7.2 at room temperature for two hours, and postfixed with 1% osmium tetroxide/0.1 M cacodylate buffer, pH7.2 at room temperature for one hour. The specimens were dehydrated in 50%, 70%, 90%, 100% ethanol and propylene oxide, before being embedded in Luft’s Araldite 502 embedding medium (Electron Microscopy Sciences, Hatfield, PA) and cut into thin sections. The ultrathin sections were then stained with ethanolic uranyl acetate and Reynold’s lead citrate (Polysciences). The samples were examined with a JOEL JEM1400 transmission electron microscope and images were acquired with a Gaten DAT-832 Orius camera.
RNA-Seq analysis
Total RNA was extracted from snap-frozen quadriceps muscles (n=4/genotype/sex) using RNeasy mini kit (Qiagen). The quality of total RNA was validated by Bioanalyzer (Agilent Technologies). Approximately 1 µg of RNA per sample was used to construct the cDNA library using TruSeq stranded mRNA library prep kit (Illumina). The cDNA libraries were quantitated using KAPA library quantification kit for Illumina platforms (Kapa Biosystems). The individual indexed libraries were diluted to 4 nM and pooled in equal quantity, denatured before loading onto the Illumina NextSeq 500. The sequencing was run as paired-end reads (2 × 75 bp per read) with a targeted depth of 60 million paired-end reads per sample. After sequencing, reads were aligned to build mm10 of the mouse genome using Bowtie2. PartekFlow (Partek, Inc., St. Louis, MO) was used for quantification to RefSeq89 with the EM algorithm and normalization was performed with the Reads Per Kilobase per Million mapped reads (RPKM) method. A two-way (genotype x sex) analysis of variance (ANOVA) was initially used to test for possible differences in expression, with the q value set at q < 0.05 for false discovery rate (FDR) correction. This was followed by separate one-way ANOVAs for males and females with the same FDR threshold applied in each case.
Determination of protein/DNA ratio
DNA and proteins were extracted from approximately 30 of quadriceps muscles using AllPrep DNA/RNA/Protein Mini kit (#80004, Qiagen). DNA concentrations were determined by Thermo Scientific NanoDrop 2000c spectrophotometer. Protein concentrations were determined by Bradford protein assay (#5000006, Bio-Rad). Relative protein contents in the muscles were calculated after normalizing by total amount of DNA.
Proteasomal activity assay
Quadriceps muscle samples (1 mg) were sonicated in PBS plus EDTA (5 mM, pH 7.4) buffer three times for 5 seconds with 25 seconds intervals on ice. After the removal of cell debris by centrifugation, protein concentration of the soluble fractions was determined by Bradford assay. 10 μg of the muscle lysates were used for determining proteasome-associated chymotrypsin-like activity with the Proteasome-Glo™ Assay Systems (#G8531, Promega). Luminescence signals were detected with the SpectraMax i3x Multi-Mode Microplate Reader (Molecular Devices) between 10–30 minutes at room temperature after the addition of luminogenic substrates. The chymotrypsin-like activity were calculated by subtracting MG132 (50 μM)-inhibited activities from total activity.
Data deposition
All raw FastQ files and processed read alignment count data from the RNA-Seq experiments have been deposited into the NCBI Gene Expression Omnibus/Sequence Read Archive (accession number: GSE135584).
Acknowledgments
We thank David Mitchell for help on electron microscopy, Siu-Pok Yee (University of Connecticut) for the generation of the ANT1-transgenic mice, Don Henderson (University of Rochester) and Christopher Turner for muscle histology, Jushuo Wang for confocal microscopy, the Chen laboratory members for comments on the manuscript, SUNY Upstate Medical University Department of Laboratory Animal Resources for animal care, and Vanderbilt University Mouse Metabolic Phenotyping Center (MMPC) for mouse phenotyping. This work was supported by the NIH grants AG061204 and AG047400 to X.J.C..