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=x−μ<
/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