Abstract Objectives Cachexia is a metabolic disorder and comorbidity with cancer and heart failure. The syndrome impacts more than thirty million people worldwide, accounting for 20% of all cancer deaths. In acute myeloid leukemia, somatic mutations of the metabolic enzyme isocitrate dehydrogenase 1 and 2 cause the production of the oncometabolite D2-hydroxyglutarate (D2-HG). Increased production of D2-HG is associated with heart and skeletal muscle atrophy, but the mechanistic links between metabolic and proteomic remodeling remain poorly understood. Therefore, we assessed how oncometabolic stress by D2-HG activates autophagy and drives skeletal muscle loss. Methods We quantified genomic, metabolomic, and proteomic changes in cultured skeletal muscle cells and mouse models of IDH-mutant leukemia using RNA sequencing, mass spectrometry, and computational modeling. Results D2-HG impairs NADH redox homeostasis in myotubes. Increased NAD+ levels drive activation of nuclear deacetylase Sirt1, which causes deacetylation and activation of LC3, a key regulator of autophagy. Using LC3 mutants, we confirm that deacetylation of LC3 by Sirt1 shifts its distribution from the nucleus into the cytosol, where it can undergo lipidation at pre-autophagic membranes. Sirt1 silencing or p300 overexpression attenuated autophagy activation in myotubes. In vivo, we identified increased muscle atrophy and reduced grip strength in response to D2-HG in male vs. female mice. In male mice, glycolytic intermediates accumulated, and protein expression of oxidative phosphorylation machinery was reduced. In contrast, female animals upregulated the same proteins, attenuating the phenotype in vivo. Network modeling and machine learning algorithms allowed us to identify candidate proteins essential for regulating oncometabolic adaptation in mouse skeletal muscle. Conclusions Our multi-omics approach exposes new metabolic vulnerabilities in response to D2-HG in skeletal muscle and provides a conceptual framework for identifying therapeutic targets in cachexia. Keywords: Oncometabolism, Autophagy, Cachexia, Systems biology Graphical abstract Image 1 [65]Open in a new tab Highlights * • Somatic mutations in IDH1 and 2 can produce the oncometabolite D2-HG in cancer cells. * • Increased production of D2-HG causes skeletal muscle atrophy and activates autophagy. * • Autophagic activation is driven by impaired oxidative metabolism and NADH redox homeostasis. * • Metabolic adaptation and skeletal muscle wasting is sex-dependent. 1. Introduction Cachexia is a multifactorial and systemic multi-organ syndrome that is one of the leading causes of morbidity and mortality in heart failure and cancer [[66]1,[67]2]. About 20–30% of cancer-associated deaths are due to cachexia, which encompasses a diverse range of symptoms, including involuntary weight loss and muscle wasting [[68]1]. The genetic and metabolic basis for developing muscle loss is elusive and limits our ability to generate targeted therapeutic strategies without those side effects. Muscle mass homeostasis is maintained through balancing protein synthesis and degradation rates, including the ubiquitin/proteasome pathway (UPP), autophagy, caspases, cathepsins, and calcium-dependent calpains. Previous studies have demonstrated the involvement of the UPP with increased mRNA expression of muscle-specific E3 ubiquitin ligases, including atrogin-1 or maf box 1 (MafBx1) and muscle-specific ring finger protein 1 (Murf1 or Trim63), during the acute stress response in skeletal muscle [[69]3,[70]4]. Increasingly, the role of autophagic-lysosomal proteolysis has been recognized in driving muscle loss during cachexia. Autophagy contributes to cellular homeostasis by regulating the degradation of proteins and organelles and supplying metabolic intermediates, including glucose and amino acids, to maintain energy provision via ATP. Prolonged autophagic activation can cause skeletal muscle loss during cancer, characterized by increased proteome proteolysis [[71]5,[72]6], impaired oxidative metabolism [[73]7,[74]8], and extensive cellular lipid remodeling [[75]9]. Severe skeletal muscle loss is commonly associated with hematological malignancies, including acute myeloid leukemia (AML). Genomic studies in AML have identified somatic mutations of isocitrate dehydrogenase (IDH) 1 and 2 to drive altered metabolism and tumorigenesis. IDH1 and 2 mutations produce the oncometabolite D-2-hydroxyglutarate (D2-HG), which accumulates in both tumors and the bloodstream [[76][10], [77][11], [78][12], [79][13]]. D2-HG suppressed energy provision and impaired NADH regeneration in the heart through inhibition of α-ketoglutarate dehydrogenase (α-KGDH) [[80]14]. Prolonged exposure to D2-HG was associated with heart and skeletal muscle atrophy in mice [[81]14], prompting the hypothesis that oncometabolic stress mediates autophagy or proteasomal degradation of muscle proteins. Metabolites serve as signals to regulate autophagy [[82]15]. How cancer-specific oncometabolites regulate autophagy and the underlying molecular mechanisms remain largely unknown. Here, we assessed the metabolic, proteomic, and genomic changes in skeletal muscle in response to D2-HG-mediated stress in skeletal muscle cell culture and mouse models. We link gene and protein expression to metabolic changes and reveal the complex interplay between biological processes in response to acute and chronic stress. The NAD^+-dependent sirtuin 1 signaling pathway integrates protein and nutrient signals to regulate autophagy during D2-HG-mediated oncometabolic stress. Targeted metabolomics revealed that changes in autophagy activation impair the metabolism of energy-providing substrates in skeletal muscle cells. Importantly, using an integrative multi-omics approach, we show that these changes induce a system-wide adaptive response in female mice, activating compensatory gene and protein expressions. At the same time, male animals lack the same molecular features. Our study highlights the importance of system-wide profiling to decipher complex biological adaptations in response to oncometabolic stress. 2. Methods 2.1. Animals Animals were fed a standard laboratory chow, LabDiet 5001 (PMI Nutrition International, St. Louis, MO, USA). C57BL/6J Mice were obtained from Jackson Laboratory (Bar Harbor, ME, USA; CAT#000664, RRID: IMSR_JAX:000664), and both male and female mice were used in experiments. All animal procedures were approved by the Animal Care and Use Committee of Cedars Sinai Medical Center and the University of Texas Health Science Center and adhered to NIH Public Health Service guidelines. All studies were conducted at 10–14 weeks of age. Mice were housed in a temperature-controlled, specific pathogen-free, air-conditioned animal house with 14 and 10 h light/dark cycles. Food and water were fed ad libitum. During oncometabolic stress experiments, male and female mice (10 weeks old) were injected daily for 30 days intraperitoneally according to body weight with phosphate-buffered saline (PBS) or D2-HG (250 mg/kg). D2-HG was dissolved in PBS at a concentration of 500 mg/mL. At the end of the treatment protocol, mice were euthanized using pentobarbital (100 mg/kg body weight) by intraperitoneal injection. Skeletal muscle tissue samples were dissected and immediately freeze-clamped using aluminum tongs cooled in liquid nitrogen and stored at −80 °C for further tissue analysis [[83]14]. 2.2. In vivo mouse grip strength testing Male and female mice (10 weeks old) were treated with D2-HG for 30 days. During this time, mice were subjected to weekly strength testing as described by Contet C. et al. [[84]16]. Briefly, grip strength was assessed using seven weights ranging from 20 g to 90 g. Each mouse was held by the middle/base of the tail and lowered to allow the mouse to grasp the first weight (20 g). A stopwatch was started once the mouse grasped the weight pouch with its forepaws. Then, the mouse was raised until the pouch was clear of the table. Each weight had to be held by the animal for 3 s. Weights were tested sequentially, starting with the lowest at 20 g and up to 100 g. Between weights, mice were rested for about 10 s. If a mouse failed to hold a weight, the mouse was rested for 10 s, and the same weight was tried again. The trial was terminated if a mouse failed to hold a weight three times. The final score was calculated as described by Deacon R.M.J. [[85]17]. 2.3. Skeletal muscle tissue histology Skeletal muscle tissue samples were fixed overnight in 10% neutral-buffered formalin and dehydrated in 70% ethanol before paraffin embedding. Muscle cross-sections (5 μm) were cut and stained with hematoxylin and eosin by the Histopathology Laboratory in the Department of Pathology and Laboratory Medicine within the McGovern Medical School at The University of Texas Health Science Center at Houston. Slides were imaged with a Keyence Microscope (Keyence, Itasca, IL, USA). 2.4. Cell lines, culture, and treatments L6 myoblasts (rat skeletal muscle cell line), primary human skeletal muscle myoblasts (HSkMC), and Sol8 myoblasts (mouse muscle cell line) were purchased from American Type Culture Collection (ATCC, Manassas, VA, USA; ATCC CAT#CRL-1458, CAT#PCS-950-010, CAT#CRL-2174). L6Ms and Sol8 were grown in Dulbecco's Modified Eagle Medium (DMEM; Thermo Fisher Scientific, Hampton, NH, USA; CAT#10567022) and HSkMC were grown in Mesenchymal Stem Cell Basal Medium (ATCC, Manassas, VA, USA; ATCC CAT#PCS-500-030) supplemented with 10% (v/v) fetal bovine serum (FBS; Thermo Fisher Scientific, Hampton, NH, USA; CAT#16000044) and penicillin-streptomycin (100 units/mL; Thermo Fisher Scientific, Hampton, NH, USA; CAT#15140163). HSkMC were grown in Mesenchymal Stem Cell Basal Medium supplemented with 10 mmol/L l-glutamine, 5 ng/mL rhEGF, 10 μmol/L dexamethasone, 5 ng/mL rh FGF-b, 25 μg/mL rh insulin and 4% stem cell qualified FBS (ATCC, Manassas, VA, USA; ATCC CAT#PCS-500-030) and penicillin-streptomycin (100 units/mL; Thermo Fisher Scientific, Hampton, NH, USA; CAT#15140163). L6Ms, Sol8, and HSkMCs were cultured at 37 °C and 5% CO[2] in a humidified incubator. Cultures were grown for at least five passages before differentiation or transfection. Myoblasts fuse in culture to form multinucleated myotubes and striated fibers. For differentiation into myotubes, L6Ms, and Sol8 were grown in DMEM supplemented with 1% (v/v) FBS for 72 h to promote cell differentiation and fusion. HSkMC were differentiated into myotubes using the skeletal muscle differentiation tool (ATCC CAT#PCS-950-050) for 72 h. L6, Sol8, and HSkMC myotubes were maintained at 37 °C and 5% CO[2] in a humidified incubator for the duration of the experiments, changing the media every 2–3 days. For autophagy experiments, L6, Sol8 and HSkMC myotubes were treated for 24 h with phosphate-buffered saline (PBS, Sigma Aldrich, St. Louis, MO, USA; CAT#506552), D-2-hydroxyglutarate (D2-HG, 0–1.0 mmol/L; Tocris Bioscience, Minneapolis, MN, USA; CAT#6124), or dimethyl 2-oxoglutarate (DMKG, 1 mmol/L, Sigma Aldrich, St. Louis, MO, USA; CAT#349631). To quantify autophagic flux using western blotting, L6, Sol8, and HSkMC myotubes were treated with PBS (control) or D2HG (0.5 mmol/L or 1 mmol/L) for 24 h. The activation of autophagy was assessed in cells using bafilomycin A1 (BafA1, 200 nmol/L; Sigma Aldrich, St. Louis, MO, USA, CAT#B1793). Myotubes were treated with BafA1 for 2 h before the end of the experimental protocol. 2.5. Plasmid construction and L6M transfection Plasmid ptfLC3 (a gift from Tamotsu Yoshimori, AddGene, Watertown, MA, USA; CAT#21074; [86]http://n2t.net/addgene:21074; RRID: Addgene_21074) was used to generate GFP-tagged LC3-K49R/K51R and LC3-K49Q/K51Q mutants. Site-directed mutagenesis was conducted using QuikChange Lightning site-directed mutagenesis kit (Agilent, Santa Clara, CA, USA; CAT#210515) according to the manufacturer's instructions and using the following primers: LC3-K49R/K51R-F (5′-GAACCTGGTCCTGTCCA -3′) and LC3-K49R/K51R-R (5′-GTCCTGGACAGGACCA-3′) and LC3-K49Q/K51Q-F (5′-GAACTGGGTCTGGTCCA-3′) and LC3-K49Q/K51Q-R (5′- GGACCAGACCCAGTTCC-3′). Clones were confirmed by genomic DNA extraction using the QIAGEN Plasmid Maxi Kit (Qiagen, Hilden, Germany; CAT#12163), PCR amplification of the target loci with Q5® polymerase (New England Biolabs Inc., Ipswich, MA, USA; CAT#M0491S), and Sanger sequencing of PCR products (Genewiz LLC, South Plainfield, NJ, USA). L6Ms overexpressing p300 were generated using plasmid 1245 pCMVb p300 (AddGene, Watertown, MA, USA; CAT#10717, RRID: Addgene_10717). According to the manufacturer's instructions, transfections were conducted using LipofectamineTM 3000 (Invitrogen, Thermo Fisher Scientific, Hampton, NH, USA; CAT#L300008). Briefly, L6Ms were grown to 70–80% confluence in a 6-well plate with DMEM containing 10% (v/v) FBS and penicillin-streptomycin (100 U/mL). Plasmid DNA-lipid complexes were in sterile OptiMEM (Gibco, Thermo Fisher Scientific, Hampton, NH, USA; CAT#51985091) by diluting DNA (0.1 μg/μL) and 5% Lipofectamine™ 3000 (v/v) separately. The diluted Lipofectamine™ was then added to the DNA (1:1 ratio), mixed by gentle pipetting, and incubated for 20 min at room temperature. The cell culture medium was removed from each well and replaced with a mix of serum-free DMEM and OptiMEM in a ratio of 1:0.8 (v/v) for the duration of the transfection. The DNA-lipid complexes were added to each well, and L6Ms were cultured for 18 h at 37 °C and 5% CO[2]. Then, the transfection medium was removed and replaced with 2 mL DMEM containing 10% FBS (v/v). After 24, the cell culture medium was replaced with phenol red-free DMEM (Thermo Fisher Scientific, Hampton, NH, USA; CAT#11054020) containing 5 mmol/L glucose, 0.5 mmol/L glutamine, and 0.5 mmol/L pyruvate. Negative controls for each experiment consisted of cultures treated with non-targeting plasmid DNA. Sirt1 silencing was conducted using ON-TARGET plus siRNA containing SMARTpool of three siRNAs (Dharmacon, Horizon Discovery, Cambridge, United Kingdom; CAT# L-049440-00-0005). Exponentially growing cells in a 6-well plate format were transfected with 100 μM of siRNA complex prepared in DharmaFECT according to the manufacturer's instructions. Cells were incubated for an additional 72 h, and the silencing of proteins was confirmed by western blot. Following 48 h of Sirt1 silencing, cells were co-transfected with ptfLC3 plasmid and treated as described above. 2.6. Cell imaging and fluorescent puncta counting Cell treatments, microscopy, and puncta counting were conducted by six different researchers (two per task). L6Ms were imaged 24 h after treatments. Cell culture media was removed and replaced by fresh phenol red-free DMEM containing 5 mmol/L glucose, 0.5 mmol/L glutamine, and 0.5 mmol/L pyruvate. For lysosome imaging, L6Ms were live-stained by adding 50 nmol/L Lysotracker-red DND-99 dye (Invitrogen, Waltham, MA, USA; CAT#L-7528) to the cell culture medium at 37 °C for 30 min. Cells were then washed twice with Dulbeccos's phosphate-buffered saline (PBS; Gibco, Thermo Fisher Scientific, Hampton, NH, USA; CAT#14190-136) and 2 mL of fresh phenol red-free DMEM containing 5 mmol/L glucose, 0.5 mmol/L glutamine and 0.5 mmol/L pyruvate was added. Image acquisition was carried out in an inverted Olympus IX81 microscope equipped with a ProScan II (Prior Scientific) motorized stage, a Lambda 10-3 filter wheel system, a Lumen 200 (Prior Scientific) fluorescence illumination system, a 60× DIC objective, and a high-resolution Hamamatsu [87]C10600 CCD camera. The GFP (Em = 472 ± 18 nm, Ex = 520 ± 18 nm) and TRITC (Em = 545 ± 20 nm, Ex = 593 ± 20 nm) filter sets were used to visualize LC3-GFP and Lysotracker-red staining, respectively. The exposure time was 2–5 s for LC3-GFP and 0.2–0.5 s for Lysotracker-red staining. All in vitro assays were conducted after treatments for 24 h with D2-HG. The image recordings were visualized and quantified in Fiji [[88]18]. Recordings allowed for counting cells, and within those cells, the individual spots were counted manually by three independent users. Images were collected at similar magnification levels. The ratio of the total number of spots over the total number of cells was then calculated. 2.7. Mitochondrial respirometry Oxygen consumption of intact L6 myotubes was measured using a Seahorse XF96 Extracellular Flux Analyzer. Cells were seeded at a density of 2 × 10^4 cells/well in a 96-well XF microplate cultured and differentiated with DMEM (Thermo Fisher Scientific, Hampton, NH, USA; CAT#10567022) containing 10% (v/v) FBS (Thermo Fisher Scientific, Hampton, NH, USA; CAT#16000044). One day before initiation of measurements, the growth medium was replaced with XF-compatible DMEM (CAT#A14430-01, Thermo Fisher Scientific, Hampton, NH, USA) supplemented with physiologic concentrations of glucose (5.5 mmol/L; Sigma Aldrich, St. Louis, MO, USA; CAT#G8270), glutamine (0.5 mmol/L; Thermo Fisher Scientific, Hampton, NH, USA; CAT#25030081) and pyruvate (0.1 mmol/L; Thermo Fisher Scientific, Hampton, NH, USA; CAT#11360070). Cells were incubated at 37 °C without CO[2] for 1 h to allow pre-equilibration with the assay medium. Oligomycin A (Sigma Aldrich, St. Louis, MO, USA; CAT#75351), Carbonyl cyanide 4-(trifluoromethoxy)phenylhydrazone (FCCP, Sigma Aldrich, St. Louis, MO, USA; CAT#C2920), rotenone/antimycin A (Sigma Aldrich, St. Louis, MO, USA; CAT#R8875 and CAT#A8674) were loaded into the Seahorse XF96 cartridge injector ports. The final concentrations for reagents were as follows: 1 μM oligomycin A, 8 μM FCCP, 0.5 μM rotenone/antimycin A. Oxygen consumption rate (OCR) and extracellular acidification rate (ECAR) were detected under basal conditions followed by sequential addition of oligomycin, FCCP, as well as rotenone/antimycin A. 2.8. Western blotting and immunoprecipitation L6 myotubes were treated with PBS (control) or D2HG (0.5 mmol/L or 1 mmol/L) for 24 h. The activation of autophagy was assessed in cells using bafilomycin A1 (BafA1, 200 nmol/L; Sigma Aldrich, St. Louis, MO, USA, CAT#B1793). L6Ms were treated with BafA1 for 2 h before the end of the experimental protocol. Proteins were extracted from tissue and cell samples using RIPA lysis and extraction buffer (Thermo Fisher Scientific, Hampton, NH, USA, CAT#89900) per the manufacturer's instructions. Tissue and cell homogenates for western blotting were prepared in the presence of phosphatase inhibitors (CAT#P5726 and CAT#P0044; Sigma-Aldrich, St. Louis, MO, USA) and protease inhibitors (Complete™ Protease Inhibitor Cocktail, Sigma-Aldrich, St. Louis, MO, USA; CAT#11697498001). Proteins were separated on 4–20% SDS-PAGE gels, transferred to PVDF membranes and probed with antibodies from Cell Signaling Technology (CS, Danvers, MA, USA) against LC3B (CS, CAT#3868), Sirt1 (CS, CAT#8469), SQSTM1/p62 (CS, CAT#23214), Acetylated-lysine antibody (CS, CAT#9441), p300 (CS, CAT#86377), GAPDH (CS, CAT#5174), normal rabbit IgG (CS, CAT#2729), AMPK (CS, CAT#5831), Phospho-AMPK (CS, CAT#50081), mTOR (CS, CAT#2972), Phospho-mTOR (CS, CAT#2971). Secondary antibodies conjugated to HRP were used following incubation with primary antibodies: anti-rabbit antibody (CS, CAT#7074) and anti-mouse antibody (CS, CAT#7076). Levels of proteins were detected by immunoblotting using horseradish peroxidase-conjugated secondary antibodies and chemiluminescence. For immunoprecipitation experiments, cultured L6Ms were washed with ice-cold PBS twice and lysed in hypotonic/digitonin buffer containing 1,4-Piperazinediethanesulfonic acid (PIPES, 20 mmol/L, pH 7.2; Sigma-Aldrich, St. Louis, MO, USA; CAT#P6757), Ethylenedinitrilo)tetraacetic acid (EDTA, 5 mmol/L; Sigma-Aldrich, St. Louis, MO, USA; CAT#E9884), MgCl[2] (3 mmol/L; Sigma-Aldrich, St. Louis, MO, USA; CAT#M8266), glycerophosphate (10 mmol/L; Sigma-Aldrich, St. Louis, MO, USA; CAT#G9422), pyrophosphate (10 mmol/L; Sigma-Aldrich, St. Louis, MO, USA; CAT#P8010), digitonin (0.02%; Sigma-Aldrich, St. Louis, MO, USA; CAT#D141), as well as protease and phosphatase inhibitors (see above). After 40 min nutation at 4 °C, samples were centrifuged at 20,000×g for 7 min, and the supernatant was transferred into new microcentrifuge tubes. LC3, Sirt1, or acetylated proteins were immunoprecipitated using the monoclonal antibody described above. Cell lysates were incubated (300–500 μg of total protein) with antibodies at 4 °C overnight. Immunocomplexes were then incubated for 2 h with protein A/G PLUS-agarose beads (30 μL of 50 % slurry; Thermo Fisher Scientific, Hampton, NH, USA; CAT#20423). Samples were washed with ice-cold lysis buffer four times, and beads were boiled in 4× Lithium dodecyl sulfate buffer (∼30 μL; Thermo Fisher Scientific, Hampton, NH, USA; CAT#B0008) for 15 min at 37 °C to elute captured protein. Samples were then subjected to Western blotting. Signals were quantified using NIH ImageJ software ([89]http://imagej.nih.gov/ij; Bethesda, MA, USA). Bands were normalized to that of loading control obtained from the same gel (GAPDH for whole cell lysate and target molecule of immunoprecipitation for the study of LC3 and Sirt1 interaction), and a percentage of control was obtained. 2.9. Real time-quantitative PCR (RT-qPCR) RNA was extracted from L6Ms using the QIAGEN RNeasy Fibrous Tissue Mini Kit (Qiagen, Hilden, Germany; CAT#74704). According to the manufacturer's instructions, RNA concentration was determined using a Qubit 4.0 fluorometer (Life Technologies) and Qubit™ RNA BR Assay Kit (ThermoFisher Scientific, Waltham, MA; CAT#[90]Q10210). The quality of each RNA sample was checked using an Agilent 4200 Tape Station with Agilent High Sensitivity RNA ScreenTape (Agilent Technologies, Santa Clara, CA, USA; CAT#5067-5579) according to the manufacturer's instructions. cDNAs were synthesized using the iScript gDNA Clear cDNA Synthesis Kit (Bio-Rad, Hercules, CA, USA; CAT#172-5035) from 500 ng of total RNAs. Quantitative PCRs were conducted by mixing cDNA samples (final concentration per well 10 ng) with iTaq Universal SYBR® Green Supermix (Bio-Rad, Hercules, CA, USA; CAT#172-5124) and the appropriate primers according to the manufacturer's instructions. The following validated primers from Bio-Rad (CAT#100-25636) were used in this study: Beclin1 (qRnoCID00038), Trim63 (qRnoCED0003), p62/Sqstm1 (qRnoCED0006), Microtubule-associated proteins 1A/1B light chain 3B (Map1/LC3B; qRnoCED0002031), atrogin-1 (MAFbx; qRnoCID0008629). Pre-designed primers for Peptidylprolyl isomerase A (cyclophilin A; Sigma-Aldrich; CAT#KSPQ12012) were used for gene expression analysis. Quantitative PCR was conducted, including an RNA quality control (Bio-Rad, Hercules, CA, USA; CAT#10044157) and genomic DNA control (CAT#10044158, Bio-Rad, Hercules, CA, USA) for each sample. Real-time PCR was conducted using a CFX384 Touch Real-Time PCR Detection System (Bio-Rad, Hercules, CA, USA). The cycling protocol was as follows: activation at 97 °C for 2 min, followed by 40 cycles of denaturation at 95 °C for 5 s, and annealing/extension at 60 °C for 30 s. At the end of the protocol, an additional Melt curve step was added at 65–95 °C with 0.5 °C increments for 5 s/step. 2.10. RNA sequencing RNA extraction and library preparation: RNA-sequencing libraries were prepared using the Illumina TruSeq Stranded mRNA Sample Preparation Kit (Illumina, San Diego, CA; CAT#20020594). RNA was extracted from 20 to 30 mg skeletal muscle tissue using the QIAGEN RNeasy Fibrous Tissue Mini Kit (Qiagen, Hilden, Germany; CAT#74704). According to the manufacturer's instructions, RNA concentration was determined using a Qubit 4.0 fluorometer (Life Technologies) and Qubit™ RNA HS Assay Kit (Thermo Fisher Scientific, Waltham, MA; CAT#[91]Q32852). The quality of each RNA sample was checked using an Agilent 4200 Tape Station with Agilent High Sensitivity RNA ScreenTape (Agilent Technologies, Santa Clara, CA, USA; CAT#5067-5579) according to the manufacturer's instructions. For library preparation, poly-A RNA was purified from mRNA using oligo-dT magnetic beads. mRNA was fragmented, and cDNA was synthesized. A-tailing was conducted to enable subsequent ligation of paired-end sequencing adapters. Library quality was determined using an Agilent 4200 Tape Station with Agilent High Sensitivity D5000 DNA ScreenTape (CAT#5067-5592, Agilent Technologies, Santa Clara, CA, USA) to ensure correct fragment sizes between 250 and 300 bp. RNA sequencing and data processing: Libraries were clustered onto a HiSeq patterned flow cell using a cBot2 system (Illumina, Inc., San Diego, CA, USA) and HiSeq 3000/4000 PE Cluster Kit (Illumina, Inc., San Diego, CA, USA; CAT#PE-410-1001). The clustered flow cell was then run on a HiSeq 4000 Sequencing System (Illumina, Inc., San Diego, CA, USA) using a HiSeq 3000/4000 SBS 300-cycle kit (Illumina, Inc., San Diego, CA, USA; CAT#FC-410-1003) for 2 × 150 paired-end reads. All instrument preparation and run parameters were carried out or selected according to manufacturer specifications. Raw reads from the sequencer were demultiplexed and converted to fastq format using bcl2fastq v2.19.1.403 (Illumina, Inc., San Diego, CA, USA), optioned to allow for 0 mismatch in the barcodes. 2.11. Metabolomics by LC and ion chromatography (IC)-MS The relative abundance of polar metabolites in mouse skeletal muscle (M. gastrocnemius) and L6 myocyte samples were analyzed from metabolite extracts by ultra-high-resolution mass spectrometry (HRMS). For tissue samples, approximately 10–20 mg of skeletal muscle were pulverized on liquid nitrogen and then homogenized with Precellys Tissue Homogenizer (Bertin Instruments, Rockville, MD, USA; CAT#P000062-PEVO0-A). Approximately 0.5 × 10^6 cells from cultured L6 myotubes were flash-frozen in liquid nitrogen and then homogenized for metabolomics. Metabolites were extracted using 1 mL ice-cold 0.1% Ammonium hydroxide in 80/20 (v/v) methanol/water. Extracts were centrifuged at 17,000 g for 5 min at 4 °C, and supernatants were transferred to clean tubes, followed by evaporation to dryness under nitrogen. Dried extracts were reconstituted in deionized water before MS, and 5 μL were injected for analysis by ion chromatography (IC)-MS. IC mobile phase A (MPA; weak) was water, and mobile phase B (MPB; strong) contained 100 mM KOH. A Thermo Scientific Dionex ICS-5000+ system included a Thermo IonPac AS11 column (4 μm particle size, 250 × 2 mm) with a column compartment kept at 30 °C. The autosampler tray was chilled to 4 °C. The mobile phase flow rate was 360 μL/min, and the gradient elution program was: 0–5 min, 1% MPB; 5–25 min, 1–35% MPB; 25–39 min, 35–99% MPB; 39–49 min, 99% MPB; 49-50, 99-1% MPB. The total run time was 50 min. Methanol was delivered by an external pump and combined with the eluent via a low dead volume mixing tee to assist desolvation and increase sensitivity. Data were acquired using a Thermo Orbitrap Fusion Tribrid Mass Spectrometer under ESI negative ionization mode. For amino acid analysis, samples were diluted in 90/10 acetonitrile/water containing 1% formic acid, and then 15 μL was injected for analysis by liquid chromatography (LC)-MS. LC mobile phase A (MPA; weak) was acetonitrile containing 1% formic acid, and mobile phase B (MPB; strong) was water containing 50 mM ammonium formate. A Thermo Vanquish LC system included an Imtakt Intrada Amino Acid column (3 μm particle size, 150 × 2.1 mm) with a column compartment kept at 30 °C. The autosampler tray was chilled to 4 °C. The mobile phase flow rate was 300 μL/min, and the gradient elution program was: 0–5 min, 15% MPB; 5–20 min, 15–30% MPB; 20–30 min, 30–95% MPB; 30–40 min, 95% MPB; 40–41 min, 95-15% MPB; 41–50 min, 15% MPB. The total run time was 50 min. Data were acquired using a Thermo Orbitrap Fusion Tribrid Mass Spectrometer under ESI positive ionization mode at a resolution of 240,000. Raw data files were imported to TraceFinder™ 5.1 SP1 (Thermo Fisher Waltham, MA, USA; CAT#OPTON-31001) for peak identification and quantification. Metabolite abundances were normalized by DNA concentrations. 2.12. Detection and quantification of D2-HG using LC-MS/MS Metabolites from freeze-clamped skeletal muscle tissue samples and L6Ms were extracted for targeted quantification of D2-HG. Samples were homogenized with 500 μl of cold 40% acetonitrile, 40% methanol, and 20% water, vortexed vigorously for 15 min at 4 °C, and spun at 10,000×g for 10 min at 4 °C. Supernatants were removed and dried to completion in a SpeedVac, and resuspended using 20 μl of methanol, followed by vortexing, and subsequently 80 μl of water was added along with final vigorous vortexing. Standards of [U-^13C]-D2-HG sodium salt (Cambridge Isotope Laboratories, Tewksbury, MA, USA; CAT#CLM-10351-PK) and unlabeled D2-HG (biotechne, R&D Systems, Minneapolis, MN, USA; CAT#6124) at 40 mg/mL were prepared in water and serially diluted to generate a standard curve to assess linearity and for quantitation. Serially diluted samples were dried to completion in a SpeedVac and resuspended similarly to the samples as described above. Cell metabolite extractions were analyzed by injecting 4 μl of sample volume and using multiple reaction monitoring LC-MS with an Agilent 6470A triple quadrupole mass spectrometer. This instrument operated in negative mode and connected to an Agilent 1290 ultra-high-performance liquid chromatography (UHPLC) system (Agilent Technologies, Santa Clara, CA). All mobile phase solvents consisted of HPLC or LC-MS grade reagents. Buffer A: water with 3% methanol, 10 mM tributylamine, and 15 mM glacial acetic acid. Buffers B: isopropyl alcohol. Buffer C: methanol with 10 nM tributylamine and 15 mM glacial acetic acid. Buffer D: acetonitrile. Finally, the analytical column used was an Agilent ZORBAX RRHD Extend-C18 1.8 μm 2.1 × 150 mm coupled with a ZORBAX Extend Fast Guard column for ultra-high-performance liquid chromatography Extend-C18, 2.1 mm, 1.8 μm. The resulting chromatograms were visualized in Agilent MassHunter Quantitative Analysis for QQQ (Agilent Technologies). According to our method, D2-HG eluted with a retention time of approximately 14 min and was identified by monitoring the transition of m/z 147.0 for precursor ion to m/z 129.1 and 101.2 for product ions. The final peak areas were manually checked for consistency and proper integration. Samples were normalized by internal standards and DNA concentration. 2.13. Untargeted proteomics by LC-MS/MS Tissue was fractionated into cytosolic-, myofilament-, and insoluble-enriched fractions by the “IN sequence” method as described previously [[92]19]. The skeletal muscle tissue was cryo-homogenized at an approximate ratio of tissue weight to homogenization buffer volume at 1:4 (v/v). Tissue sample were separated into cytosolic-, myofilament-, and insoluble-enriched fractions using a homogenization buffer containing 250 mmol/L HEPES-NaOH (Sigma Aldrich, St. Louis, MO, USA; CAT#H3375), EDTA (2.5 mmol/L, pH 8.0; Sigma Aldrich, St. Louis, MO, USA; CAT#E9884), a homogenization buffer containing 1% trifluoroacetic acid (Sigma-Aldrich, St. Louis, MO, USA; CAT#299537), and a homogenization buffer containing 2% sodium dodecyl sulfate (SDS; Sigma Aldrich, St. Louis, MO, USA; CAT#L6026) respectively. Protein concentration was determined by Pierce™ BCA Protein Assay Kit (Thermo Fisher Scientific, Waltham, MA, USA; CAT#23225). Two hundred μg of each fraction (total protein quantification) were reduced using 1 mmol/L Tris(2-carboxyethyl)phosphine hydrochloride (TCEP, Sigma Aldrich, St. Louis, MO, USA; CAT#C4706). Samples were cleaned and alkylated using filter-aided sample preparation (FASP 10 kDA). Samples were digested for 15–18h at 37 °C using ultra-grade Trypsin (Promega, Madison, WI, USA; CAT#V5111) at a 1:100 enzyme: protein ratio. Samples were desalted using Oasis HLB 96-well plates (30 μm and 5 mg sorbent; Waters, Milford, MA, USA; CAT#18600309), vacuum-dried, and stored at −80 °C until analysis. For all methods, peptides were analyzed in two replicate runs by liquid chromatography-tandem mass spectrometry (LC-MS/MS) on a Dionex Ultimate 3000 NanoLC connected to an Orbitrap Fusion™ Lumos™ Tribrid™ Mass Spectrometer (ThermoFisher Scientific, Waltham, MA, USA) equipped with an EasySpray ion source as previously reported [[93]20]. (A) Analysis of nano-IP and gel pieces using short gradient: Lumos™, USAPeptides were loaded onto a PepMap RSLC C18 column (2 μm, 100 Å, 150 μm i.d. x 15 cm, ThermoFisher Scientific, Waltham, MA, USA) using a flow rate of 1.4 μL/min for 7 min at 1% B (mobile phase A was 0.1% formic acid in water and mobile phase B was 0.1 % formic acid in acetonitrile) after which point they were separated with a linear gradient of 5–35%B for 11 min, 35-85%B for 3 min, holding at 85%B for 5 min and re-equilibrating at 1%B for 17 min. The nano-source capillary temperature was set to 300 °C, and the spray voltage was set to 1.8 kV. MS1 scans were acquired in the Orbitrap at a resolution of 120,000 Hz in the mass range of 400–1600 m/z. For MS1 scans, the AGC (Automatic Gain Control) target was set to 4 × 10^5 ions with a max fill time of 50 ms. MS2 spectra were acquired using the TopSpeed method with a total cycle time of 3 s, an AGC target of 1 × 104, and a max fill time of 100 ms, with an isolation width of 1.6 Da in the quadrupole. Precursor ions were fragmented using HCD (Higher-energy collisional dissociation) with a normalized collision energy of 30% and analyzed using rapid scan rates in the ion trap. Monoisotopic precursor selection was enabled, and only MS1 signals exceeding 5000 counts triggered the MS2 scans with +1, and unassigned charge states were not selected for MS2 analysis. Dynamic exclusion was enabled with a repeat count of 1 and an exclusion duration of 15 s. Raw data was searched using the most recent version of the Uniprot rat database with the Sorcerer-Sequest™ search engine (Sagen-N Research Inc., Milpitas, CA, USA) with the following search parameters: full Trypsin cleavage, static modification of +57 Da on Cysteine, variable modification of +16 Da on Methionine (Oxidation), +54 Da on Arginine (Methylglyoxal-Hydroimidazolone 1, MG-H1), and +72 on Lysine (Carboxyethyl), with a MS1 error of 10 ppm, and MS2 error of 1 Da. Data was post-processed using Scaffold 4 [94]www.proteomesoftware.com, Proteome Software, Inc., Portland, OR, USA). (B) Discovery proteomics using long gradient: Lumos™Peptides were loaded onto a PepMap RSLC C18 column (2 μm, 100 Å, 150 μm i.d. x 25 cm, ThermoFisher Scientific, Waltham, MA, USA) using a flow rate of 1.4 μL/min for 7 min at 1% B (mobile phase A was 0.1% formic acid in water and mobile phase B was 0.1 % formic acid in acetonitrile) after which point they were separated with a linear gradient of 5–20%B for 45 min, 20–35%B for 15 min, 35–85%B for 3 min, holding at 85%B for 5 min and re-equilibrating at 1%B for 5 min. Each sample was followed by a blank injection to clean the column and re-equilibrate at 1%B. The nano-source capillary temperature was set to 300 °C, and the spray voltage was set to 1.8 kV. MS1 scans were acquired in the Orbitrap at a resolution of 240,000 Hz with a mass range of 400–1600 m/z with advanced peak detection enabled. For MS1 scans, the AGC target was set to 4 × 10^5 ions with a max fill time of 50 ms. MS2 spectra were acquired using the TopSpeed method with a total cycle time of 3 s, an AGC target of 1 × 10^4, and a max fill time of 50 ms, with an isolation width of 1.6 Da in the quadrupole. Precursor ions were fragmented using HCD with a normalized collision energy of 30% and analyzed using rapid scan rates in the ion trap. Monoisotopic precursor selection was enabled, and only MS1 signals exceeding 5000 counts triggered the MS2 scans, with +1 and unassigned charge states not being selected for MS2 analysis. Dynamic exclusion was enabled with a repeat count of 1 and an exclusion duration of 15 s. Raw data was searched using the most recent version of the Uniprot rat database with the Sorcerer-Sequest™ search engine (Sage-N Research Inc., Milpitas, CA, USA) with the following search parameters: full Trypsin cleavage, static modification of +57 Da on Cysteine, variable modification of +16 Da on Methionine (Oxidation), +54 Da on Arginine (Methylglyoxal-Hydroimidazolone 1, MG-H1), and +72 on Lysine (Carboxyethyl), with a MS1 error of 10 ppm, and MS2 error of 1 Da. Data was post-processed using Scaffold 4 ([95]www.proteomesoftware.com, Proteome Software, Inc., Portland, OR, USA). Data processing for proteomics analysis: Raw MS/MS data files were converted to mzXML format using MSconvert (v.3.0.6002) from ProteoWizard [[96]21] for peak list generation and used for database searching using two engines, X! TANDEM Spectrum Modeler (v2013.06.15.1) [[97]22] and Comet (v2014.02 rev.2) [[98]23]. The dataset was searched against the concatenated target/decoy mouse UniProt Knowledgebase (34018 proteins and decoys, December 06, 2018) [[99]24,[100]25]. The uniqueness of identified sequences assigned to non-reviewed proteins was validated by manual Blastp search, thus filtering multiple assigned peptides. Protein isoforms were only reported if a peptide comprising an amino acid sequence unique to the isoform was identified. Protein quantification was done by averaging the raw peptide intensity among technical replicates and summing it among cellular sub-fractions. The peptide signal intensity was normalized to the median of the overall sample signal intensity, and protein level abundance inference was calculated using the linear mixed effects model built into the open sources MSstats (v3.2.2). 2.14. Statistical analysis Statistical analysis was conducted using RStudio Desktop (v1.2.5042 for Linux, Boston, MA, USA) [[101]26], Bioconductor [[102]27,[103]28], and GraphPad Prism (version 8.1.2 for MacOS, GraphPad Software, La Jolla California USA, [104]www.graphpad.com). Indicated sample sizes (n) represent individual tissue samples. For GFP-puncta counting, sample size (n) represents the number of cells analyzed from three or more independent experiments. Sample size and power calculations for in vitro and in vivo were based on Snedecor [[105]29] and GPower [[106]30] (version 3.1.9.2 for windows). The type 1 error and power were considered at 5% (P-value of 0.05) and 80%, respectively. The expected difference in the mean between groups was 50–30%, and the standard deviation was 25–12.5%. Unsupervised hierarchical clustering: Clustering was conducted using the R package ‘pheatmap’ (v1.0.12) after log2-transformation and z-score scaling of the data. The z-score is defined as follows: [MATH: Z=xμ< /mrow>σ :MATH] where [MATH: x :MATH] denotes the observed value, [MATH: μ :MATH] the mean of the sample, and [MATH: σ :MATH] the standard deviation of the sample. The similarity between groups was assessed using an Euclidean distance, and the number of clusters was determined using the k-means algorithm. We applied the ‘Elbow’ method to determine the optimal number of clusters. Linear models to identify multi-omics changes in response to D2-HG: Linear models adjusted for sex and batch information were computed in R using the ‘limma’ package (v3.40.06). Linear models were applied to log2-transformed data. A comparison between groups was conducted by one-way ANOVA (two-sided) and Tukey's posthoc test to identify significance. Normality was tested using a Shapiro-Wilk test. P-values were corrected for multiple hypotheses using the Benjamini-Hochberg method, and samples with a false discovery rate (FDR) below 0.01 (RNA-sequencing) and 0.05 (proteomics and metabolomics) were considered statistically significant. Pathway enrichment analysis: We used the STRING database ([107]https://string-db.org) [[108]31] to identify enriched pathways using differentially expressed genes and proteins from RNA-sequencing and MS-based proteomics, respectively. The significance of pathways was determined by the hypergeometric test (one-sided) in IPA and a permutation-based, non-parametric test in STRING. P-values were corrected for multiple hypotheses using the Benjamini-Hochberg methods, and pathways with FDR below 0.01 (RNA-sequencing) and 0.05 (proteins and metabolites) were considered significant. Protein–Protein interaction (PPI) network analysis: Functional interactions between proteins were determined using the STRING database ([109]https://string-db.org) [[110]31] and the STRING plugin tool for Cytoscape (v3.8.0)^32 was used to visualize the PPI networks. In the network, the nodes correspond to the proteins, and the edges represent the interactions. Metabolite network generation: Metabolite interaction networks were generated using the MetaMapp tool ([111]http://metamapp.fiehnlab.ucdavis.edu) [[112]33]. Metabolites from MS-based metabolomics data were annotated to PubChem CIDs, KEGG IDs, and SMILES codes. The pair-wise Tanimoto chemical similarity coefficients were generated using the MetaMapp tool using a threshold of 0.7^33. Cytoscape was used to visualize the differential statistics output on network graphs. Statistical results (P-value <0.05) were mapped as color, and fold changes were mapped as node size. Network generation from multi-omics data. Differential expression analysis between groups was conducted using the linear models described above. We assigned proteome measurements to metabolites using UniProt Knowledgebase and KEGG [[113][34], [114][35], [115][36]]. Pairwise Spearman's rank correlation and Pearson's correlation coefficient were calculated between metabolites and proteins using the R base function ‘cor,’ and the network was plotted with Cytoscape (v3.8.0)^32. In the network, each node represents a metabolite or protein, and an edge between two nodes represents a metabolite–metabolite, protein–protein, or metabolite–protein interaction. PPIs were determined using the STRING database [[116]31]. 3. Results 3.1. Inhibition of α-KGDH causes metabolic remodeling in skeletal muscle cells The oncometabolite D2-HG can inhibit α-KG-dependent enzymes, including α-KGDH, leading to decreased Krebs cycle flux [[117]14], impairing energy provision, and the availability of substrates for macromolecule synthesis in skeletal muscle. To identify metabolites that differentially accumulate in skeletal muscle cells in response to oncometabolic stress, we conducted targeted liquid chromatography and mass spectrometry (LC-MS/MS) on L6 myotubes. These myotubes are derived from rat L6 myocytes (L6Ms, rat skeletal muscle cell line) and consist of fused myoblasts, which retrain the ability to contract [[118]37]. We cultured L6 myotubes with phosphate-buffered saline (PBS, control), D2-HG (1.0 mmol/L), or dimethyl alpha-ketoglutarate (DMKG, 1.0 mmol/L) for 24 h in defined nutrient-rich media ([119]Figure 1A). Increased D2-HG levels were only detectable in D2-HG-treated L6 myotubes ([120]Supplementary Fig. 1A). DMKG is a membrane-permeable ester of α-KG that is cleaved to α-KG in the cytoplasm [[121]38]. DKMG is an additional control to identify whether metabolic effects caused by D2-HG are driven by changes in the availability of α-KG or potentially other metabolites. Principal component analysis (PCA) revealed a clear separation between treatment and control groups ([122]Figure 1B), indicating that D2-HG and DMKG treatment promoted considerable metabolic remodeling in L6 myotubes. D2-HG treatment caused the accumulation of Krebs cycle intermediates, including succinate and isocitrate ([123]Figure 1C). Further, D2-HG impaired NADH redox homeostasis as evidenced by increased NAD^+ levels and NAD^+/NADH redox ratio, whereas DMKG did not affect these metabolites ([124]Figure 1C and [125]Supplementary Fig. 1B). We integrated the targeted metabolomic data into pathway enrichment analysis to assess metabolic remodeling based on pathways and chemical similarity using MetaMapp [[126]33] and Cytoscape [[127]32] ([128]Figure 1D). The resulting network comprises 101 metabolites with 615 metabolite–metabolite interactions (MMIs). We identified the enrichment of metabolites in amino acid and glucose metabolism, the pentose phosphate pathway (PPP), and redox homeostasis. These findings are consistent with previous studies in heart tissue suggesting that cancer cells producing D2-HG have a systemic metabolic impact on muscle tissue [[129]14]. Figure 1. [130]Figure 1 [131]Open in a new tab D2-HG causes metabolic remodeling in differentiated L6Ms. (A) Schematic of in vitro experimental workflow and protocol. (B) PCA of targeted metabolomics using LC-MS/MS in L6 myotubes treated with phosphate-buffered saline (Cnt), D2-HG (1 mmol/L), or DMKG (1 mmol/L) for 24 h. Each data point represents a biological replicate. Data is log2 normalized. (C) Heatmap showing unsupervised hierarchical clustering of metabolites that are altered (P-value<0.05, FDR = 10%) in the metabolomics dataset used in (B). The color-coded z-score illustrates the increase or decrease in metabolite concentrations upon D2-HG treatment. (D) Network visualization of targeted MS-based metabolomics data. Nodes represent metabolites, and edges represent metabolite–metabolite interactions. Nodes are color-coded by P-value, and their size represents the median fold change relative to the untreated sample. (E) Oxygen consumption rate (OCR) of L6 myotubes in response to D2-HG (0–1 mmol/L) was determined following sequential addition of oligomycin (OM, 1 μmol/L) to measure ATP-linked OCR, Carbonyl cyanide 4-(trifluoromethoxy)phenylhydrazone (FCCP, 0.2 μmol/L) to determine maximal respiration, and rotenone and antimycin (R/A, 1 μmol/L each) to determine the non-mitochondrial respiration as indicated. ATP-linked OCR and spare capacity were determined in L6 myotubes treated with the indicated dose of D2-HG (0–1 mmol/L) (n = 7 technical and biological replicates per group). Data are mean ± s.d. ∗P-value<0.05, ∗∗∗∗P-value<0.0001. Statistical analysis using two-way ANOVA with post-hoc Tukey's multiple comparisons test. NAD^+ is a critical co-factor and reducing equivalent for various metabolic processes. Alteration in the NAD^+/NADH redox potential can directly affect the function of enzymatic reactions and indicate functional impairment of mitochondria. Thus, we reasoned that if α-KG levels are elevated in L6 myotubes as a response to impaired Krebs cycle flux, the increase in NAD^+ levels might be accompanied by a decreased mitochondrial function and ATP provision. Using Agilent Seahorse XF Cell Mito Stress Tests, we confirmed that both D2-HG and DKMG treatment decreases the ATP-linked oxygen consumption rate (OCR) and increases the mitochondrial spare capacity in L6 myotubes in a concentration-dependent manner ([132]Figure 1E and [133]Supplementary Fig. 1C). Intriguingly, we did not observe changes in mitochondrial function or ATP-related OCR at D2-HG concentration between 0.06 and 0.125 mmol/L ([134]Supplementary Figs. 1C and D). These observations are consistent with previous reports that D2-HG or L2-HG may accumulate under physiological conditions without a notable effect on enzymatic activities in certain cancer cell lines and tumors [[135][39], [136][40], [137][41]]. These data suggest that D2-HG impairs oxidative metabolism and NAD^+ redox homeostasis in L6 myotubes. 3.2. Inhibition of α-KGDH by D2-HG increases autophagic flux in myotubes Recent studies indicate that increased α-KG levels disrupted NAD^+ redox homeostasis or impaired ATP provision during nutrient starvation-induced autophagic flux [[138]15,[139]42]. Thus, we reasoned that an increase in autophagy might accompany the observed metabolic changes in D2-HG-treated myotubes. To explore this possibility, we assessed the initiation and flux of autophagosome formation in myotubes from three species (human, mouse, and rat) by quantifying microtubule-associated protein 1 light chain 3-II (LC3-II) using western blotting. L6 myotubes (rat), Sol8 myotubes (mouse), and human skeletal muscle cell (HSkMC) myotubes were cultured in defined nutrient-rich media and treated with or without D2-HG (1.0 mmol/L). The dynamic turnover and delivery to lysosomes balance the formation of autophagosomes. Therefore, bafilomycin A1 (BafA1), a selective vacuolar H^+ ATPase (V-ATPase) inhibitor, was used to measure autophagic flux. BafA1 inhibits the maturation of autophagosomes by blocking the fusion between autophagosomes and lysosomes, thus preventing LC3-II degradation [[140]43]. We found that D2-HG increases the lipidation of LC3 with and without bafilomycin A1 (BafA1, 200 nmol/L) in L6 myotubes ([141]Figure 2A,B), as well as HSkMC and Sol8 myotubes ([142]Supplementary Figs. 2A and 2B). In contrast, the supplementation of DMKG in L6 myotubes maintained low autophagy levels (LC3-II and p62 expression) both with and without BafA1 conditions in vitro ([143]Supplementary Figs. 2C and 2D). Similarly, the number of autophagic puncta per cell was measured through live-cell fluorescent microscopy in differentiated L6 myotubes expressing green fluorescent protein (GFP) fused with LC3 (see Methods for details). The fusion between autophagosomes and lysosomes can be visualized using LC3 and LysoTracker (see Methods) as markers for autophagosomes and lysosomes. In untreated L6Ms (PBS; control), GFP-LC3 puncta were diffusely localized in the nucleus and cytosol ([144]Figure 2C and [145]Figure 2D). As expected, BafA1 treatment caused a significant increase in GFP-LC3 puncta formation in the cytosol and autolysosome formation in L6Ms incubated in defined growth media without D2-HG, reflecting the basal level of autophagy activation ([146]Figure 2C and [147]Figure 2D). In contrast, GFP-tagged LC3 in L6 myotubes treated with D2-HG (1 mmol/L) resided primarily in the cytoplasm with multiple cytoplasmic puncta. Most GFP-LC3 colocalized with LysoTracker, consistent with increased autophagic flux ([148]Figure 2C and [149]Figure 2D). We also measured the expression of critical genes involved in autophagy: Beclin1, LC3, MafBx1, Murf1, and protein 62 (p62) were measured over 24 h. Gene expression of both LC3 and Murf1 increased within 12 h of treatment in vitro, reaching a plateau after 16 h and significantly increasing at 24 h ([150]Figure 2E and [151]Figure 2F). In contrast, MafBx1 mRNA levels remained unchanged between experimental groups during the 24 h experimental protocol ([152]Supplementary Fig. 2E). Beclin1 mRNA levels increased after 2 h in D2-HG-treated cells ([153]Supplementary Fig. 2F). In comparison, p62 expression decreased after 8 h before returning to the baseline level ([154]Supplementary Fig. 2G). These results demonstrate that D2-HG increases autophagic flux in skeletal muscle cells. Figure 2. [155]Figure 2 [156]Open in a new tab Metabolic remodeling promotes the activation of autophagy. (A and B) Representative western blots of LC3-I, LC3-II, and the cytosolic marker GAPDH in total cell extracts from L6 myotubes treated for 24 h with phosphate-buffered saline (PBS, control) or D2-HG (1 mmol/L). Myotubes were treated with or without bafilomycin A (BafA1, 200 nmol/L) for 2 h at the end of the 24 h culture. Densitometry analysis is shown below the representative blot (n = 3 independent experiments). Data are mean ± s.d. ∗P-value<0.05, ∗∗P-value<0.01, ∗∗∗P-value<0.001. Statistical analysis using two-way ANOVA with post-hoc Tukey's multiple comparisons test. (C–D) Representative live-cell images (C) of differentiated L6Ms transfected with a plasmid encoding a GFP-tagged LC3 (green) cultured with PBS, D2-HG (1 mmol/L) or BafA1 (200 nmol/L), and co-stained with LysoTracker (red). Puncta were counted for at least 100 cells per condition (D). Scale bars represent 10 μm. Data are mean ± s.d. ∗∗P-value<0.01, ∗∗∗P-value<0.001. Statistical analysis using two-way ANOVA with post-hoc Tukey's multiple comparisons test. (E–F) Gene expression of Map1-LC3B (E) and Murf1 (F) from L6 myotubes treated for 24 h with phosphate-buffered saline (PBS, control) or D2-HG (1 mmol/L) (n = 8–10 independent experiments per group). Data are mean ± s.d. ∗∗P-value<0.01, ∗∗∗P-value<0.001, ∗∗∗∗P-value<0.0001. Statistical analysis using two-way ANOVA with post-hoc Tukey's multiple comparisons test. 3.3. Deacetylation of LC3 is driven by Sirt1 activation The redistribution of LC3 into the cytosol and subsequent lipidation is regulated by proteins sensing ATP and NAD^+ levels. This raises the possibility that autophagy activation in D2-HG-treated myotubes is driven by metabolic adaptation. Autophagy can be activated through AMP-activated protein kinase (AMPK) and the mammalian target of rapamycin (mTOR), which are crucial cellular energy sensor proteins regulated by ATP levels in the cell. The total and phosphorylated protein expression of AMPK and mTOR were not increased after treatment with D2-HG or BafA1 (P-value>0.05 for all groups; [157]Supplementary Figs. 3A and B), suggesting that autophagy is activated by another mechanism in this model. A second possible mechanism could revolve around NAD-dependent deacetylase sirtuin-1 (Sirt1), as previous studies have shown that during starvation, the Sirt1 can activate autophagy through de-acetylation of LC3 lysine residues at position 49 and 51 ([158]Figure 3A) [[159]44]. In this model, D2-HG increases the co-localization of LC3 and Sirt1, which, in turn, decreases the acetylation of LC3 in L6 myotubes ([160]Figure 3B). Therefore, we developed single and double mutants of LC3 with replacement of lysine at positions 49 and 51 by either arginine (LC3-K49R, LC3-K51R and LC3-K49R-K51R) or glutamine (LC3-K49Q, LC3-K51Q and LC3-K49Q-K51Q). These mutations allowed to mimic a decrease (K to R) or an increase (K to Q) in the acetylation level of LC3 ^44. GFP-tagged single and double mutants were transfected into L6Ms and further differentiated into myotubes to assess their acetylation after treatment with or without D2-HG (1.0 mmol/L) for 24 h. Lysine-to-arginine single replacement at position 49 or 51 reduced LC3 acetylation, and the double mutant (LC3-K49R-K51R) showed the lowest acetylation ([161]Figure 3C). Upon treatment with D2-HG, we detected increased deacetylation of LC3 in L6 myotubes expressing LC3-K49R or LC3-K51R. LC3 acetylation was almost completely abolished in LC3-K49R-K51R mutants and associated with increased LC3 lipidation ([162]Figure 3C). Conversely, we found increased acetylation of LC3 in L6 myotubes expressing GFP-tagged LC3-K to Q mutants treated with and without D2-HG ([163]Supplementary Figs. 3C and 3D). In all mutants, LC3 acetylation was decreased in D2-HG cells compared to untreated cells, indicating that the oncometabolite D2-HG increases the flux of deacetylation in vitro ([164]Supplementary Figure 3D). We then measured the formation of GFP-LC3 containing puncta in GFP-tagged LC3-KR mutants through live-cell fluorescent microscopy. Untreated L6 myotubes expressing LC3-K48R, LC3-K51R, or LC3-K48R-K51R showed a nuclear and cytoplasmic distribution like wild-type (WT) LC3, which was also reflected in the amount of GFP-LC3 containing puncta per cell ([165]Figure 3D). In contrast, treatment with D2-HG caused 2–3 -fold increased formation of GFP-LC3 puncta in the cytosol ([166]Figure 3D). Our findings indicate that the deacetylation of LC3 is the primary driver of autophagy activation in D2-HG treated L6 myotubes. Next, we tested whether the activation of autophagy is attenuated by increasing the deacetylation of LC3. We used two strategies to test this hypothesis. First, we silenced the expression of Sirt1 using siRNA and non-targeting negative controls to assess whether Sirt1 is solely driving the lipidation of LC3. Secondly, we overexpressed the acetyltransferase p300 in L6 myotubes using plasmid to test whether increased protein acetylation would counteract the effect of D2-HG. Previous studies have shown that p300 is a primary regulator of autophagy through protein acetylation during nutrient limitation [[167]45]. At baseline, protein expression of both Sirt1 and p300 was increased in response to D2-HG ([168]Figure 3E). Silencing Sirt1 attenuated both the total LC3-II level and the LC3-II to LC3-I ratio compared to WT conditions within 24 h ([169]Figure 3F and [170]Figure 3G). Nonetheless, the total LC3-II level (P-value = 0.0004; q-value = 0.0008) and LC3-II to LC3-I ratio (P-value = 0.02; q-value = 0.023) remained significantly higher than WT conditions. These data suggest that other sirtuin isoforms may contribute to autophagy activation in the presence of D2-HG. In contrast, p300 overexpression fully attenuated total LC3-II levels (P-value = 0.5; q-value = 0.7) and LC3-II to LC3-I ratios (P-value = 0.6; q-value = 0.7) within 24 h ([171]Figure 3H and [172]Figure 3I), thus counteracting the effect of D2-HG. Our data indicate that NAD^+ redox changes and the deacetylation of LC3 drive autophagy in response to D2-HG treatment in vitro. Figure 3. [173]Figure 3 [174]Open in a new tab Sirt1 drives the deacetylation of LC3 in L6Ms. (A) Schematic of Sirt1-mediated deacetylation of LC3. (B) Representative western blotting depicting co-immunoprecipitation of LC3 and Sirt1 in differentiated L6Ms treated with or without D2-HG (1 mmol/L) for 24 h. Total expression of LC3-I, LC3-II, and GAPDH are depicted from 5% input of the co-immunoprecipitation sample. Images are representative of n = 3 experiments. (C) LC3 was immunoprecipitated from L6 myotubes expressing wild-type LC3 and mutant LC3 with lysine to arginine replacement at positions 49 and 51, respectively. The degree of acetylation was assessed using a pan-acetyl-lysine antibody. Wild-type and mutant LC3-expressing cells were treated with or without D2-HG (1 mmol/L) for 24 h. (D) Representative live-cell images of differentiated L6Ms transfected with a plasmid encoding a GFP-tagged LC3 (green) in wild-type LC3 and mutant LC3 with a lysine to arginine replacement at positions 49 and 51. Cells were treated with or without D2-HG (1 mmol/L) for 24 h, and puncta were counted for at least 100 cells per condition. Scale bars represent 10 μm. Data are mean ± s.d. ∗∗∗P-value<0.001, ∗∗∗∗P-value<0.0001. (E) Representative western blotting of Sirt1, p300, and LC3-I and LC3-II expression in L6 myotubes. (F and G) Representative western blotting (F) and densitometry (G) analysis of L6 myotubes cultured with or without D2-HG following Sirt1 silencing using siRNA and non-targeting (NT)-RNA. Data are mean ± s.d. ∗P-value<0.05, ∗∗P-value<0.01, ∗∗∗P-value<0.001, ∗∗∗∗P-value<0.0001 (n = 3 per group). (H and I) Representative western blotting (H) and densitometry (I) analysis of L6 myotubes cultured with or without D2-HG following p300 overexpression and non-targeting (NT)-cDNA (control). Data are mean ± s.d. ∗P-value<0.05, ∗∗P-value<0.01, ∗∗∗P-value<0.001, ∗∗∗∗P-value<0.0001 (n = 4 per group). All P-values were determined by one-way analysis of variance (ANOVA) followed by Šídák's multiple comparisons test. 3.4. D2-HG activates autophagy and promotes skeletal muscle wasting in vivo Syngeneic mouse models using C26 colon carcinoma, or Lewis lung carcinoma cells, have been successfully applied to study skeletal muscle remodeling and cachexia in response to rapidly growing tumors [[175][46], [176][47], [177][48]]. These models demonstrated that cancer cells release various compounds (e.g., DNA, proteins, metabolites) that have a systemic impact on the body. To understand whether overproduction of D2-HG alone activates autophagy and promotes skeletal muscle wasting in vivo, we treated wild-type (WT) male and female mice (four animals per group; 10 weeks old) for 30 days with either vehicle (PBS) or D2-HG (250 mg/kg body weight) through daily intraperitoneal injections (IP) ([178]Supplementary Fig. 4A) [[179]14]. We observed increased body weight in both males and females, which directly correlated with the overall growth of the animals ([180]Supplementary Fig. 4B). As expected, we observed a reduction in M. gastrocnemius weight (both left and right muscles) in male animals with D2-HG compared to placebo treatment after 30 days of treatment ([181]Supplementary Fig. 4C). Correspondingly, we found increased D2-HG levels in muscle tissue from male animals ([182]Supplementary Fig. 4D). In contrast, both M. gastrocnemius weight and D2-HG levels were unchanged in female mice ([183]Supplementary Figs. 4C and 4D). To assess the functionality of mice skeletal muscle, grip strength was determined in mice by conducting a series of weekly in vivo weight-lifting tests during the 30-day treatment protocol [[184]16,[185]17]. Mice lifted weights ranging from 20 to 90 g for 3 s ([186]Figure 4A, see Methods for details). In untreated control animals (male and female), grip strength increased in correlation with the animals' overall growth ([187]Figure 4B). Importantly, this growth-dependent increase in grip strength was absent in both males and females when treated with D2-HG ([188]Figure 4B). After four weeks of treatment with D2-HG, the grip strength was significantly reduced in both male and female animals (male, P-value = 0.019; female, P-value = 0.0207). Histological analysis of M. gastrocnemius sample ([189]Figure 4C) revealed that the reduction in muscle mass in D2-HG treated male mice was caused by a decrease in the myofiber nuclei number ([190]Figure 4D). Correspondingly, the LC3-II expression in M. gastrocnemius samples from male animals was reduced in D2-HG treated animals. Correspondingly, we observed an increased p62 and LC3-II expression in M. gastrocnemius samples from male animals treated with D2-HG ([191]Figure 4E). In contrast, in female animals, p62 expression was not changed in response to D2-HG treatment, while both LC3-II expression and LC3-II to LC3-I ratio was increased ([192]Figure 4F). These findings show that prolonged treatment with D2-HG upregulates of LC3-mediated autophagy and p62 aggregation. Grip strength reduction is observed in male and female animals, with males showing earlier signs of skeletal muscle loss than their female counterparts. Figure 4. [193]Figure 4 [194]Open in a new tab In vivo changes in response to D2-HG. (A) Experimental set-up to measure grip strength in mice. Mice are lifting weights ranging from 20 g to 90 g. (B) Weight-lifting scores in male and female mice treated with or without D2-HG (250 mg/kg body weight) throughout the experiment (4 weeks). Scores represent the grip strength in mice and are normalized by body weight (n = 4–5 mice per group and sex). Data are mean [MATH: ± :MATH] s.d. ∗∗ q-value<0.01, ∗∗∗q-value<0.001. Statistical Analysis using an unpaired t-test (two-step method by Benjamini, Krieger, and Yekutieli) with a false discovery rate (FDR) of 5%. (C–D) Representative H&E images of M. gastrocnemius (C) and quantification of nuclei per cross-sectional area (D) from skeletal muscle samples in male and female mice treated with or without D2-HG (250 mg/kg body weight). scale bar = 100 μm; n = 4 per group. Data are mean [MATH: ± :MATH] s.d. Statistical Analysis using an unpaired t-test (two-step method by Benjamini, Krieger, and Yekutieli) with a false discovery rate (FDR) of 5%.∗∗q-value<0.01, ∗∗∗q-value<0.001. (E and F) Quantification of p62, LC3-I, and LC3-II in skeletal muscle tissue from male mice (n = 6 mice per group) (E) and female mice (n = 3 mice per group) (F) treated with or without D2-HG. Densitometry was normalized to the total protein expression. Data are mean [MATH: ± :MATH] s.d. Statistical Analysis using an unpaired t-test with Welch's correction and two-step method by Benjamini, Krieger, and Yekutieli with a false discovery rate (FDR) of 1%. ∗q-value<0.05,∗∗∗q-value<0.001, ∗∗∗∗q-value<0.0001. 3.5. Transcriptional and post-transcriptional regulation of oncometabolic stress To investigate the profile and extent of transcriptional, post-transcriptional, and metabolic events upon oncometabolic stress by D2-HG, our animal studies were complemented by applying an integrated systems biology-wide approach ([195]Supplementary Fig. 4A). Male and female mice were treated daily for 30-days with vehicle (PBS) or D2-HG (450 mg/kg body weight, four animals per group). In-depth multi-omics profiling was conducted on each skeletal muscle tissue sample (M. gastrocnemius) with RNA sequencing (RNA-seq; transcriptomics) and LC-MS/MS for proteomics and targeted metabolomics. Proteins were sub-fractionated into the highly abundant myofilamentous, cytosolic, and insoluble membrane proteins [[196]19] before analysis by LC-MS/MS for identification and relative quantification (see [197]Methods for details). We identified 46,079 RNA-Seq transcripts, quantified 2,153 proteins from untargeted analysis using LC-MS/MS, and quantified 95 metabolites from targeted analysis using LC-MS/MS. PCA analysis of gene, protein, and metabolite levels showed a clear separation between treatment and control groups ([198]Figure 5A). In total, we identified 1,976 differentially expressed genes (FDR<1%, [199]Supplementary Fig. 5A) and 170 differentially expressed proteins (FDR<5%, [200]Supplementary Fig. 5A). Using the Search Tool for the Retrieval of Interacting Genes/Proteins database (STRING) [[201]31], we identified 21 enriched pathways between gene and protein expression (FDR < 0.01) ([202]Figure 5B). Functional enrichment analysis showed that proteins in these clusters are part of cellular protein metabolic processes, regulation of chromatin assembly or disassembly, mitochondrial ATP provision, and muscle contraction ([203]Figure 5B). Next, we reconstructed a protein–protein interaction (PPI) network of significantly regulated proteins using STRING [[204]31]. The network consists of 70 proteins (nodes) and 170 protein–protein interactions (edges) ([205]Figure 5C). D2-HG treatment differentially affected proteins that enriched in four main clusters: (1) mitochondrial respiratory chain complex assembly, (2) cell protein metabolic process, (3) muscle contraction, and (4) chromatin assembly and disassembly. Normalized counts from RNA-seq could be matched with protein MS intensities for 1,346 genes independent of sex. Plotting individual RNA and protein ratios (D2-HG vs. PBS) revealed several proteins with differential expression at both the transcript and protein level ([206]Supplementary Fig. 5C). RNA and protein abundance ratios correlated only poorly (Pearson's correlation coefficient, r = 8.9), indicating translational, miRNA or post-transcriptional regulation of many biological processes in response to oncometabolic stress in skeletal muscle. Intriguingly, several proteins that are part of autophagy regulation, lipid remodeling and known regulators of mitochondrial function, including the NADH dehydrogenase 1 alpha subcomplex subunit 13 (encoded by Ndufa 13), the quinone oxidoreductase (encoded by Cryz) and D-beta-hydroxybutyrate dehydrogenase (encoded by Bdh1), were only regulated at the protein level ([207]Supplementary Fig. 5C). Likewise, unsupervised hierarchical clustering of targeted metabolomics revealed four main clusters comprising 15 metabolic pathways using the Reactome pathway knowledgebase [[208]49] ([209]Supplementary Fig. 6A). Broadly, metabolites were enriched in pathways that are part of the pyrimidine metabolism, nucleotide metabolism, energy substrate metabolism, and DNA replication ([210]Supplementary Fig. 6B). We identified 11 significantly altered metabolite in response to oncometabolic stress and across sex (FDR<5%) ([211]Supplementary Fig. 6C). Together, these findings elucidate the adaptation to D2-HG treatment in vivo through the remodeling of chromatin to active, targeted gene programs, followed by protein expression. Figure 5. [212]Figure 5 [213]Open in a new tab Multi-omics changes in response to D2-HG. (A) PCA of RNA-sequencing (RNA-seq) and MS-based proteomics and metabolomics of skeletal muscle tissue from male (m) and female (f) mice treated with or without D2-HG (250 mg/kg body weight). Each data point represents a biological replicate. n = 4–5 mice per group and sex. Data is log2 normalized. (B) Enrichment analysis of significantly expressed proteins from MS-based proteomics using Gene Ontology (GO) annotations. The top-13 enriched terms after redundancy filtering were visualized according to negative log10 transformed FDR-values. The size of each filled circle depicts the number of enriched genes. (C) STRING protein–protein analysis of significantly expressed proteins from MS-based proteomics. Nodes represent proteins, and edges represent protein–protein interactions. Functional analysis of clusters obtained by Markov clustering. GO annotation from (B) were visualized as split donut charts around the nodes as follows: 1 – muscle contraction; 2 – chromatin assembly or disassembly; 3 – nucleic acid binding; 4 – proton transmembrane transport; 5 – cellular protein metabolic process; 6 – translation; 7 – mitochondrial respiratory chain complex assembly; 8 – electron transport; 9 – cellular protein-containing complex assembly; 10 – ATP metabolic process; 11 – ion transport; 12 – ATP metabolic process; 13 – ATP hydrolysis coupled ion transmembrane transport. Proteins without interaction partners within the network (singletons) are omitted from the visualization. 3.6. D2-HG promotes sex-dependent metabolic alterations We conducted a multi-omics analysis to assess the impact of oncometabolic stress at a systems level. First, we compared the proteomic and transcriptomic datasets regarding gene categories using a 2D annotation enrichment algorithm. This algorithm identifies both correlated and uncorrelated changes between two data dimensions. Our analysis revealed different strategies in males and females to compensate for the metabolic stress caused by the oncometabolite D2-HG ([214]Figure 6A and [215]Figure 6B). Male animals showed reduced proteins that are part of oxidative phosphorylation, whereas female animals upregulated the same set of proteins. Likewise, pathway enrichment analysis using metabolomics data revealed sex-dependent metabolic alterations ([216]Supplementary Fig. 7). Glycolytic intermediates and amino acids were downregulated in the skeletal muscle of male mice in response to D2-HG compared to female mice. Next, we identified protein-metabolite interactions (PMIs), metabolite–metabolite interactions (MMIs), and protein-protein interactions (PPIs) by solving a multiple linear regression (MLR) problems followed by Spearman's rank and Pearson correlation coefficient analysis. Annotations between proteins and metabolites were conducted using the Reactome pathway knowledgebase [[217]49] and CardioNet [[218]50]. In total, we identified 361 significant protein-metabolite interactions. We integrated 25 metabolites and 49 proteins into a network using Spearman's rank and Pearson correlation coefficient cut-off value of 0.75. The metabolic network is composed of 60 PMIs, 76 MMIs, and 229 PPIs ([219]Figure 6C). We identified four main clusters of interactions that encompassed the (1) pyrimidine/co-factor metabolism, (2) amino acid metabolism, (3) Krebs cycle (or tricarboxylic acid cycle), as well as (4) glycolysis and the pentose phosphate pathway. Several core proteins interacted with multiple metabolites, and PMIs preferentially occurred within a metabolic sub-network. Our analysis shows the combined regulation of metabolic function through up-or down-regulation of proteins drives adaptation in male mice. For example, the upregulation of betaine-homocysteine S-methyltransferase (Bhmt) and branched-chain keto acid dehydrogenase E1 subunit beta (Bckdhb) expression correlated with an increased level of aspartate and methionine. Correspondingly, we observed increased levels of AMP and glycolytic intermediates (e.g., glucose, fructose 6-phosphate) and upregulation of proteins involved in glycolysis and the pentose phosphate pathway. The findings suggest that prolonged treatment with D2-HG promotes the upregulation of glucose uptake and utilization in skeletal muscle. Our multi-omics analysis further demonstrated the close interaction between protein and metabolite profiles in response to oncometabolic stress and exposed metabolic vulnerabilities. Figure 6. [220]Figure 6 [221]Open in a new tab D2-HG promotes sex-dependent metabolic and proteomics remodeling in skeletal muscle. (A-B) Two-dimensional annotation enrichment analysis based on transcriptome (RNA-sequencing) and proteome expression in skeletal muscle tissue from male (A) and female (B) mice treated with D2-HG for 30 days. Significant metabolic pathways (KEGG and Reactome) and gene ontology terms are distributed along the proteome and transcriptome change direction. Types of annotation databases are color-coded, as depicted in the legend. (C) Pairwise Spearman correlation networks of multi-omics data based on MS-based proteomics and targeted MS-based metabolomics. Nodes represent metabolites and proteins, while edges represent metabolite–metabolite, protein–protein, and metabolite–protein interactions. Metabolites and proteins are depicted as squares or circles, respectively. Nodes were color-coded by P-value, and size represents the median fold change relative to untreated control animals. Statistical Analysis using a cut-off of 0.75 for Spearman's rank correlation coefficient. (D) Summary of the main discoveries. 4. Discussion Our study revealed that the oncometabolite D2-HG promotes autophagy activation in skeletal muscle and broad proteomic and metabolomic remodeling dependent on sex. The response to oncometabolic stress includes early (e.g., energy substrate metabolism, altered redox state, and autophagy activation) and late events (e.g., structural protein remodeling, protein quality control, and chromatin remodeling) ([222]Figure 6D). The combination of in vitro and in vivo multi-omics studies provided a comprehensive insight into the link between metabolic alteration and protein response pathways. We demonstrated that D2-HG impairs mitochondrial function in myotubes, which increases NAD^+ levels and activation of LC3-II, a key regulator of autophagy, through deacetylation by the nuclear deacetylase Sirt1. Oncometabolic stress causes muscle atrophy in mice, reduced grip strength, and increased expression of autophagy markers. Finally, a multi-omics analysis of transcriptomic, proteomics, and metabolic data revealed a systems-wide sex-dependent remodeling in skeletal muscle in response to oncometabolic stress. Several pathways regulate the activation of proteolytic systems in mammalian cells. Several lines of evidence indicate that D2-HG mediates autophagy and skeletal muscle atrophy. As no single autophagy marker exists, we used a series of experiments to answer whether D2-HG activates autophagy using cultured rat (L6), mouse (Sol8), and human myotubes. Gene expression shows a time-dependent activation of autophagy after treatment with D2-HG. Previous studies showed that LC3-II formation precedes an increased p62 expression during the activation of autophagy in nutrient-starved cells [[223][51], [224][52], [225][53]]. Our data indicate that within 2 h after D2-HG treatment, Beclin1 expression increases rapidly, followed by a decreased p62 expression after 8 h. Correspondingly, we observed an association between increased LC3 and Murf1 expression within 12 h. These findings were corroborated in live-cell imaging using GFP-tagged LC3. Murf1 is an E3 ubiquitin ligase mediating the ubiquitination and subsequent proteasomal degradation of structural proteins (e.g., cardiac troponin I/TNNI3) and other sarcomere-associated proteins. Its role in muscle atrophy and hypertrophy is to regulate an anti-hypertrophic protein kinase C-mediated signaling pathway, which results in increased muscle protein degradation [[226]54]. Our data support the conclusion that D2-HG rapidly activates autophagy and protein degradation upon cellular exposure, even in a nutrient-rich environment independent of skeletal muscle cell origin. Mitochondrial function is critical to maintain oxidative metabolism in skeletal muscle and contributes to stress adaptation [[227][55], [228][56], [229][57], [230][58]]. This study demonstrates that oncometabolic stress impairs mitochondrial ATP provision and NADH utilization. These findings agree with our previous studies in isolated working rat hearts, showing that D2-HG inhibits α-KGDH [[231]14]. Disruption of mitochondrial function impacts cytosolic NADH through stimulation of reductive carboxylation via glutamate-derived α-KG and regeneration of cytosolic NADH to support glycolysis. Cytosolic reductive carboxylation of glutamate supports glycolytic flux for ATP provision in conditions where mitochondrial function is insufficient to recycle cytosolic NADH [[232]59]. Our targeted metabolomics and computational network analysis concur with these findings by demonstrating increased glutamate levels and an accumulation of glycolytic and PPP intermediates. Several studies have shown an association between α-KG levels and autophagy. However, our findings indicate that the impaired utilization of NADH and increased NAD^+ may be the primary driver of autophagy activation in myotubes. Using mutant LC3, we confirmed that autophagy activation in response to D2-HG is mediated through the deacetylation of LC3 by the nuclear deacetylase Sirt1. These findings are consistent with several recent reports that nuclear deacetylation of LC3 is driving autophagy during cell starvation [[233]44,[234][60], [235][61], [236][62], [237][63]]. Acetylation has emerged as an essential post-translational modification affecting every autophagic cascade step. During periods of nutrient deficiency, cells initiate autophagy to replenish substrates for macromolecular synthesis. Our study advances this concept by proving that oncometabolic reprogramming activates autophagy through a Sirt1-LC3 cascade even in a nutrient-rich environment. Metabolic stress evokes several cellular signaling pathways, including activation of AMPK and inactivation of mTOR. We did not observe an increased activation of AMPK in our in vitro experiments. The silencing of Sirt1 and p300 overexpression was sufficient to attenuate autophagy. Our data suggest that LC3-dependent mechanisms primarily drive autophagy activation in response to D2-HG. Our studies indicate that reductive α-KG metabolism drives the protein acetylation changes within the autophagic cascade. Mechanistic challenges arise from linking metabolic changes to corresponding protein post-translational modifications and whether they are direct or indirect effects induced by metabolites. The argument supporting our conclusions that autophagy is mediated directly through metabolic changes in the presence of D2-HG is the timing and concentration-dependent results. As discussed above, metabolic alterations and acetylation of LC3 are sensitive to cellular α-KG levels and mitochondrial function. When D2-HG is elevated to 0.125 mM in L6 myotubes, mitochondrial respiratory capacity is impaired, and LC3 gene expression and lipidation increase within 12 h. Overexpression of p300 reverses the LC3 lipidation to control levels. Together, these findings prove that the observed effects are directly mediated through metabolic changes. Systems biology approaches aim to capture molecular interactions at every level, ranging from gene transcription to proteins, metabolites, and flux [[238]50,[239]64,[240]65]. These data-driven approaches integrate large-scale datasets from RNA-sequencing, proteomics, and metabolomics and complement hypothesis-driven experimental studies. Our study builds on these inductive strategies to identify which gene-protein-metabolic interactions drive adaptation in skeletal muscle during oncometabolic stress. Indeed, we identified a sex-specific metabolic, proteomic, and transcriptomic pattern that provides insight into the long-term consequences of D2-HG-induced remodeling. Specifically, we found that proteins involved in autophagy or proteasomal degradation machinery are increasingly expressed while structural proteins were decreased. Our data indicate that D2-HG treatment increases the expression of proteins regulating chromatin organization, cytoskeleton organization, and cell signaling in both sexes. However, metabolic alterations induced by D2-HG affect the expression of proteins in a sex-dependent manner. Sex differences in muscle wasting and autophagy are increasingly recognized [[241][66], [242][67], [243][68], [244][69]]. Estrogens and androgens are essential regulators of muscle mass and function [[245]70]. Recent studies demonstrated that when autophagy is activated, acetylation of p300 is attenuated by Sirt1 or Sirt2 to modulate the degree of autophagic flux [[246]71,[247]72]. Other studies showed that female mice have less basal ubiquitin-proteasome activity and increased autophagy activity than male mice [[248]73,[249]74]. Our data support these previous findings and demonstrate that autophagic activation optimizes metabolic adaptation in female mice. This metabolic response is characterized by an increased expression of proteins associated with oxidative phosphorylation and maintenance of metabolite levels. Our data provide evidence for sex-dependent etiologies for cancer cachexia development and the need to optimize therapies for skeletal muscle pathologies based on biological sex. However, the role of sex in regulating autophagy and muscle wasting requires further studies. We demonstrate autophagic activation by D2-HG using skeletal muscle myotubes from humans, mice, and rats. However, our study is limited by characterizing skeletal muscle alterations only in mouse models. Recent case reports in patients with high blood D2-HG levels indicate that skeletal muscle remodeling is a clinical concern that requires further investigation [[250]75]. Multi-omics analysis is inherently limited by annotating transcriptomics, proteomics, and metabolomics datasets using curated databases (e.g., HMDB, PubChem, KEGG). Lipid metabolism is underrepresented in pathway-based databases, which limits pathway enrichment analysis. Lastly, this study does not quantify the posttranslational regulation of proteins by metabolic alterations. Future studies using MS-based proteomics and ATAC-seq could be helpful to address this question. Our findings shed new light on the mechanisms underlying the strong relationship between cancer and skeletal muscle loss by showing how oncometabolic stress activates autophagy. Our study points to a critical interplay between acetylation and deacetylation of proteins to regulate metabolic adaptation and proteome remodeling. It remains unclear how metabolic changes may impact epigenetic remodeling and how substrate replenishment can prevent the observed transcriptional and post-transcriptional modifications. Our data clarify the importance of autophagy in cellular stress adaptation and provide insights into metabolic vulnerabilities driving skeletal muscle remodeling. Understanding the molecular mechanisms that enable muscle cells to adapt and compensate for oncometabolic stress will be necessary for therapeutic strategies targeting metabolic pathways in cachexia. Study approval All animal experiments were conducted according to the Institutional Animal Care and Use Committee and operated by the guidelines issued by The University of Texas Health Science Center at Houston. Data and materials availability The RNA-sequencing data have been deposited to the NCBI database (dataset identifier [251]GSE159772). The MS proteomics data have been deposited to the ProteomeXchange Consortium via the PRIDE [[252]76] partner repository (dataset identifier PXD022137). The metabolomics data and network files have been deposited to Mendeley [[253]77]. CRediT authorship contribution statement Yaqi Gao: Data curation, Formal analysis, Visualization, Writing – original draft, Writing – review & editing, Investigation. Kyoungmin Kim: Data curation, Formal analysis, Investigation, Writing – original draft, Writing – review & editing. Heidi Vitrac: Conceptualization, Formal analysis, Investigation, Methodology, Visualization, Writing – original draft, Writing – review & editing. Rebecca L. Salazar: Formal analysis, Investigation, Visualization, Writing – original draft, Writing – review & editing. Benjamin D. Gould: Formal analysis, Investigation, Writing – original draft, Writing – review & editing. Daniel Soedkamp: Formal analysis, Investigation, Writing – original draft, Writing – review & editing. Weston Spivia: Formal analysis, Methodology, Writing – original draft, Writing – review & editing. Koen Raedschelders: Investigation, Writing – original draft, Writing – review & editing. An Q. Dinh: Formal analysis, Investigation, Writing – original draft, Writing – review & editing. Anna G. Guzman: Investigation, Methodology, Writing – original draft, Writing – review & editing. Lin Tan: Data curation, Formal analysis, Investigation, Methodology, Writing – original draft, Writing – review & editing. Stavros Azinas: Formal analysis, Investigation, Writing – original draft, Writing – review & editing. David J.R. Taylor: Formal analysis, Investigation, Writing – original draft, Writing – review & editing. Walter Schiffer: Formal analysis, Investigation, Writing – original draft, Writing – review & editing. Daniel McNavish: Formal analysis, Investigation, Writing – original draft, Writing – review & editing. Helen B. Burks: Formal analysis, Investigation, Writing – original draft, Writing – review & editing. Roberta A. Gottlieb: Formal analysis, Resources, Writing – original draft, Writing – review & editing. Philip L. Lorenzi: Data curation, Formal analysis, Investigation, Resources, Writing – original draft, Writing – review & editing, Methodology. Blake M. Hanson: Formal analysis, Investigation, Methodology, Resources, Writing – original draft, Writing – review & editing. Jennifer E. Van Eyk: Data curation, Funding acquisition, Investigation, Methodology, Resources, Writing – original draft, Writing – review & editing. Heinrich Taegtmeyer: Conceptualization, Formal analysis, Funding acquisition, Writing – original draft, Writing – review & editing. Anja Karlstaedt: Conceptualization, Data curation, Formal analysis, Funding acquisition, Investigation, Methodology, Project administration, Resources, Software, Supervision, Validation, Visualization, Writing – original draft, Writing – review & editing. Declaration of competing interest The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper. Acknowledgments