Metformin extends the lifespan of iMSUD mice by rescuing physiologic dysfunction and modulating mechanisms of mTORC1 activation

Background- Dysregulation of branched chain amino acid (BCAA) catabolism caused by mutations in the metabolizing enzyme-branched chain alpha-keto acid dehydrogenase (BCKDH) leads to the fatal newborn disorder Maple Syrup Urine Disease (MSUD) and is linked to chronic diseases such as type 2 diabetes. Results- Here we show that a MSUD mouse model with severely reduced BCKDH activity exhibits glucose intolerance, altered mTOR signaling, decreased levels of TCA cycle intermediates, reduced distribution of Type I skeletal muscle bers and activated AMP-activated protein kinase (AMPK) signaling, all of which can be ameliorated by long term treatment with metformin, which enhances the survival of these mice. Treatment with metformin was correlated with increased serum-levels of growth/differentiation factor GDF15 (GFD15) and adiponectin. We also found that branched chain keto acid regulates leucine-mediated activation of mTORC1 by preventing the dissociation of sestrin2-gator complex in vitro. Conclusions- In summary, our study suggests metformin could be an effective therapeutic candidate for MSUD patients and maps the crosstalk between key pathways that maintain metabolic homeostasis in MSUD. The iMSUD and wild-type mice used in this study were obtained by breeding Dbt tm1Geh/+ ; Tg(tetO-DBT)A1Geh/J; Tg(Cebpb-tTA)5Bjd) mice to generate iMSUD or Dbt1 +/+ (wild type) mice homozygous for the LAP-tTA and TRE-E2 transgenes (6). All the mice have free access to water and food and all animal care and experimental procedures were approved by the Institutional Animal Care and Use Committee at the BUCK Institute for Research on Aging. For the lifespan study with metformin diet, the wild type and iMSUD mice were supplied with control (5LG6, TestDiet) or metformin (5LG6 w/0.1% metformin, TestDiet) after weaning. The mice were regrouped after genotyping and assigned group of treatments were continued till naturally death happened.

Dysfunction in BCAA metabolism and an increase in fasting concentrations of BCAA are also associated with insulin resistance and type 2 diabetes in both animal models and human patients (9,10). Despite this accumulating evidence, it is unclear whether enzymes that control the metabolism of BCAA play a direct role in controlling glucose metabolism and physiology. Insights from previous studies (11) show that over half the capacity to metabolize BCAAs resides in skeletal muscle, with signi cant activity also in adipose tissue, liver and brain. The metabolism of these BCAAs (leucine, isoleucine and valine) occurs through two common enzymes, BCAA amino transferase (BCAT1 and 2) and BCKDH. BCATs use mitochondrial alpha-ketoglutarate as a substrate for the amino-transferase reaction, producing ketoisovaleric (KIV), keto-methylvaleric (KMV) and keto-isocaproic (KIC) acids from Valine, Isoleucine and leucine, respectively. The next enzyme in the pathway, BCKDH, represents the rst irreversible ' ux generating' step. In the classic form of MSUD (6,7), the BCKDH enzyme exhibits only 2-5% of its normal activity. In addition to the classic form, there are intermediate (12) (15-25% BCKDH activity) and intermittent (asymptomatic until 10-16 months or later) forms.
It has been recognized that MSUD patients display skeletal-muscle abnormalities, including lesions in myo bers (13). The mechanistic basis for skeletal muscle abnormalities in MSUD is not understood, and targeting this dysfunction has the potential to have important therapeutic outcomes in MSUD patients.
Animal models are well suited to explore the pathophysiology of MSUD and permit pre-clinical testing of drug interventions. Here we used the intermediate MSUD (iMSUD) model generated by Homniacs et. al. (7) which develops pathology similar to MSUD patients and survives past weaning. In our previous study, using a combination of metabolic pro ling and small molecule screening, we found that an acute treatment with metformin, a widely used anti-diabetic drug, diminishes MSUD-associated ketoacidosis in patient-derived broblasts and in an iMSUD mouse model (14). Here, we studied the long-term bene cial effects of metformin in iMSUD mice. Our studies reveal that BCAA metabolism plays a key role in maintaining TCA metabolite levels, glucose homeostasis and skeletal muscle function. These aspects are impacted by metformin in iMSUD mice. Finally, our studies reveal a novel arm of mTOR regulation by the leucine metabolite KIC and KMV, whose accumulation can prevent the canonical leucine-mediated activation of mTOR.

Metformin induces lifespan-extension and body weightmaintenance in iMSUD mice
Our previous study has identi ed that metformin inhibits BCAA-derived ketoacidosis and promotes metabolic homeostasis in MSUD patient-derived broblast cells (14). To determine whether metformin extends the survival and/or delays pathology in iMSUD mice, we used the strategy of delivery in the food as described previously (15). We found metformin did not measurably alter the body weight and median survival of iMSUD mice ( Figure S1A and B), which live approximately two month of age. We note, however, one observation about the metformin survival curve that merits discussion. One mouse treated with metformin survived to 120 days, which is longer than we have ever observed with these mice. This divergence at the end of the survival curve, while not statically signi cant, warrants consideration in design of different delivery strategy of metformin. In addition, iMSUD is an inborn error of metabolism in both human patient and mouse disease model (6)(7)(8)16). Considering that the chow-mediated administration of metformin could occur only after weaning (21 days of age) and the dosage of metformin can be variable based on food consumption, we next treated iMSUD mice with different doses of metformin by daily intra-peritoneal (IP) injection started at 5 days of age. Low-dose (125 mg/kg bw) administration of metformin increased the survival of iMSUD mice in both sexes ( Fig. 1A and B). The survival is dose-dependently increased at a higher dose (250 mg/ kg bw) of metformin with a median survival of 32 days (metformin-treated) vs 24 days (PBS-treated) in females and a median survival of 33 days (metformin-treated) vs 24 days (PBS-treated) males ( Fig. 1A and 1B, Log-rank (Mantel-Cox) test, P < 0.05). In our study, iMSUD mice weigh about half as much as wild type mice at weaning ( Figure S1B). We found only the higher dosage of metformin (250 mg/kg bw) caused a continuous increase in body weight of iMSUD mice over their lifespan ( Fig. 1C and D).
Metformin has been shown to mediate body weight and energy balance through increasing the circulating levels of the peptide hormone growth/differentiation factor 15 (GDF15) (17). GDF15 has also been shown to lower body weight through a brain-stem-restricted receptor (18). In order to investigate the effect of metformin on the increase in body weight of iMSUD mice, we measured the concentration of GDF15 in serum of metformin-treated iMSUD mice. We found the concentration of GDF15 was increased after metformin treatment in both wild type (WT) and iMSUD mice (Fig. 1E). We also measured the concentration of adiponectin -another adipokine regulates body weight. We found that the concentration of adiponectin in the serum was also increased upon metformin treatment (Fig. 1F). This suggests that the increased GDF15 concentration is dependent on the treatment of metformin in both wild type and iMSUD mice. However, the increase in total fat content was only observed in metformin-treated iMSUD mice (Fig. 1G). To further support of this nding, we measured the protein-expression level of fatty acid synthesizing enzyme fatty acid synthetase (FAS) in metformin-treated iMSUD mice. We found that FAS protein expression was increased by metformin in both skeletal muscle and subcutaneous fat (ingWAT) of iMSUD mice (Fig. 1H). These results suggest that metformin-induced body weight increase might depend on the increase in fatty acid synthesis pathway in both adipose tissue and skeletal muscle in iMSUD mice.

Metformin-induced decreased glucose intolerance in iMSUD mice
Previous studies suggest that elevated circulating BCAA levels have been correlated with severity of insulin resistance (19,20). Metformin has been known to improve the regulation of glucose tolerance (21)(22)(23). Therefore, we tested the physiological effects of BCKDH loss on the glucose homeostasis of iMSUD mice. We found there was signi cant glucose intolerance in iMSUD mice ( Fig. 2A and B). The increased area under the curve (AUC) demonstrates that iMSUD mice have dysregulated glucose disposal capacity. We also found that metformin restored glucose tolerance in iMSUD mice to levels similar to that of wild type mice ( Fig. 2A). This result of BCAA metabolism via BCKDH controls glucose homeostasis suggests a mechanistic role and not merely a correlative role for BCAA metabolism in controlling glucose homeostasis. To test this hypothesis, we then investigated whether the iMSUD-related dysfunction in glucose tolerance and its rescue by metformin were related to the levels of insulin and fasting glucose. We found that there was a decrease in insulin level in the serum of iMSUD mice (Fig. 2C), but this was not rescued by the treatment with metformin. There was a decrease in fasting glucose levels in iMSUD mice, which was also not signi cantly affected by treatment with metformin (Fig. 2D). These results suggest that metformin-induced decreased glucose intolerance in iMSUD mice is not involved with modulation of insulin and glucose level.
It is well known that metformin regulates the protein activation/expression of IGF-1 receptor, which plays an important role in regulating glucose uptake into skeletal muscle and liver (24). Therefore, we measured IGF-1 signaling in metformin-treated liver and skeletal muscle samples using western blotting. We found that the expression/activation of IGF-1R was increased in the liver of metformin-treated iMSUD mice, which is consistent with previous reports ( Figure S2A) (25). The expression/activation of IGF-1R was diminished in the skeletal muscle of iMSUD mice, and metformin mitigated this effect ( Figure S2A).

Metformin rescues muscle dysfunction in iMSUD mice
It has been recognized that MSUD patients display skeletal-muscle abnormalities, including lesions in myo bers (13). Our previous study shows reduced mean ber cross-sectional area in quadriceps and gastrocnemius muscles of iMSUD mice (14). The result of lifespan extension by metformin treatment suggests the skeletal muscles of iMSUD mice might be the target of metformin treatment. Force measurement revealed that iMSUD mice have lower grip strength when compared to wild type mice (Fig. 3A), demonstrating a muscle functional de cit in iMSUD mice. We next analyzed the effect of BCKDH dysfunction on skeletal muscle ber type distribution. Gastrocnemius and soleus skeletal muscle sections were stained for muscle ber-type I (slow twitch oxidative), IIA (fast twitch oxidative) and IIB (fast twitch glycolytic) using previously characterized antibodies (26) (Fig. 3B). Both muscle groups showed similar trends, but the most signi cant effects were observed in soleus, where iMSUD mice have decreased levels of the slow twitch oxidative ber-type I and increased levels of fast twitch glycolytic ber-type IIB (Fig. 3C). No changes were observed in type IIA ber content. Type I bers are associated with higher mitochondrial metabolism, and their loss has been shown to be correlated with metabolic diseases such as Type II diabetes (27,28). Treatment of iMSUD mice with metformin restored the levels of type I bers in iMSUD mice nearly to that of wild type mice (Fig. 3C). Metformin partially restored the grip strength of iMSUD mice (Fig. 3A), showing that the rescue in ber type composition was correlated with improvement in skeletal muscle function.

Metformin reverses elevated AMPK signaling in iMSUD mice
It's well known that metformin activates AMPK, including hepatocytes (29) and muscle cells (22,(30)(31)(32). We found that iMSUD mice had increased levels of phosphorylated AMPK (Thr172) and the canonical AMPK target, ACC (Ser79) in skeletal muscle or liver tissues (Fig. 3D), which suggests that loss of BCKDH activates the AMPK energy-stress response pathway. The treatment of metformin had increased phosphorylation of both AMPK and ACC in wild type mice. Surprisingly, metformin has an opposite effect in iMSUD mice and relieves the hyper-activation of the AMPK energy stress pathway (Fig. 3D).

Global dysfunction in oxidative carbon metabolism caused by BCKDH knockdown
To gain a broader picture of physiological dysfunction in iMSUD mice, we performed metabolic pro ling of tissue and serum. We pro led relative levels of BCAA, leucine derived KIC and TCA metabolites from liver, skeletal muscle and serum (Fig. 4A, Figure S3 and S4). As expected, we found an accumulation of leucine/isoleucine and keto-isocaproic acid in all tissues and serum. Treatment with metformin was able to diminish the accumulation of leucine/isoleucine and KIC signi cantly in liver, skeletal muscle tissue and serum. Interestingly, iMSUD mice had signi cantly lower overall metabolism as judged by a decrease in TCA cycle metabolites and lactate in tissue and serum. The nal break-down product of BCAA (including leucine) is the mitochondrial anaplerotic substrate acetoacetate. Acetoacetate is readily converted to beta-hydroxybutyrate (βHB) by βHB dehydrogenase. The levels of βHB were signi cantly decreased in skeletal muscle of iMSUD mice (Fig. 4A, Figure S4). Metformin has been shown to promote mitochondrial metabolism via regulation of redox state (33)(34)(35)(36)(37), activation of AMPK (38)(39)(40)(41)(42) and increased the production of lactate (43)(44)(45)(46). Here, metformin was unable to restore intra-tissue lactate or TCA metabolite levels in iMSUD mice unlike in wild type mice, suggesting that BCKDH function is required for metformin action (Fig. 4A). To test whether the role of BCKDH in TCA and glucose metabolism observed in vivo was cell autonomous, we studied the effects of knocking down BCKDH using cultured cell system (Fig. 4B). Upon BCKDH knock down using shRNA in mouse C2C12 myoblasts, we observed an extracellular accumulation of KIC and KMV as expected. Intriguingly, there was a dramatic decrease in TCA and glycolytic metabolites in C2C12 shBCKDH cells, con rming that BCKDH plays a critical role in regulating TCA and glycolytic homeostasis. Finally, treatment with 1 mM metformin was unable to promote glycolytic and TCA metabolism in shBCKDH unlike in shNT (scrambled non-targeting shRNA control) C2C12 cells. This also con rms that metformin requires functional BCKDH for its metabolic action.
iMSUD mice have diminished mTOR signaling, which is modulated by metformin Global changes in mitochondrial and energy metabolism in iMSUD mice suggest that the mTOR signaling pathway, a key nutrient sensor and controller cellular physiology (3) may be altered. To test our hypothesis, we examined the activation of mTORC1 and mTORC2 in the skeletal muscle tissue of iMSUD mice. Our results indicate that both mTORC1 (P-S6) and mTORC2 (P-Akt) signaling is decreased in iMSUD mice (Fig. 5A). Treatment with metformin increased the activation in both mTORC1 and mTORC2 signal pathways. It is well known that the BCAA leucine is a direct activator of mTORC1 signal pathway (47)(48)(49). One of the hallmarks of iMSUD mice is the accumulation level of leucine, which rises up to 0.5 mM and KIC up to 1 mM in serum (14). It was interesting that metformin-mediated induction of mTOR signaling was correlated with diminished KIC levels (Fig. 5A), possibly suggesting that KIC may play a role in the inhibition of mTOR signaling.
Leucine is known to activate the mTORC1 signaling complex through stabilization of the GATOR1 complex by direct binding to the sestrin-2 protein (50). Activation of mTORC1 signaling complex causes the re-localization of the mTORC1 complex to the lysosomal membrane (51). To investigate the effect of accumulated KIC or KMV on mTORC1 complex activation, we tested whether KIC and/or KMV could directly modulate mTORC1 activity in HEK293T cells. Both mTORC1 and LAMP1 proteins (a lysosomal marker) were immuno-stained, and cells were imaged to assess the co-localization of mTOR and the lysosomal compartment as a readout of mTORC1 activation. Interestingly, treatment with 1 mM KIC and/or 1 mM KMV inhibited leucine-induced mTORC1 translocation (Fig. 5B).
To further examine the downstream regulating mechanisms of mTORC1 activation by accumulated KIC or KMV, we tested the effect of KIC and KMV on leucine-induced phosphorylation of the canonical mTOR substrate p70S6K (Fig. 5C), nding that treatment with KIC or KMV interfered with p70S6K phosphorylation. In HEK293T cells, the activation of mTORC1 can be induced by both leucine and arginine (3,47,51). We then tested whether this effect was speci c to leucine-induced mTORC1 activation by alternatively activating mTORC1 phosphorylation using arginine, which disrupt the CASTOR1-GATOR2 protein complex by binding to CASTOR1 and then activate mTORC1 in cells (52). Interestingly, KIC and KMV co-treatment had no effect on arginine-mediated activation of mTORC1 (Fig. 5C).
It has been shown that leucine induces the disruption of sestrin2-GATOR2 interaction to regulate mTORC1 activation (47). To investigate whether KIC could affect the leucine-induced disruption of sestrin2-GATOR2 interaction, HEK293T cells overexpressing FLAG-sestrin2 were treated with leucine and KIC or KMV. We immunoprecipitated FLAG-sestrin2 and probed for pulldown of two GATOR2 complex components, WDR59 and MIOS. We found that leucine treatment disrupted the interaction of sestrin2 with both MIOS and WDR59 as expected. Interestingly, both KIC and KMV promoted the interaction of sestrin2 with MIOS and WDR59, even upon leucine co-treatment (Fig. 5D). These results suggest that pathogenic accumulation of KIC or KMV in iMSUD mice could block leucine-induced activation of mTORC1 by preventing the dissociation of the sestrin2-GATOR2 complex (Fig. 5E).

Discussion
The ndings in this study support the idea that metformin may be an effective therapeutic agent for MSUD. Metformin enhances survival in both sexes and promotes weight maintenance (Fig. 1A-D). In this context, metformin increases the fat content of iMSUD mice and concomitantly increases the expression of both the fatty acid synthase (FAS) enzyme and the adipokine adiponectin (Fig. 1E). Recently it has been shown that the levels of GDF15 are increased upon metformin treatment, and that this hormone modulates the effects of metformin on weight loss upon high-fat diet condition (53). It is interesting that treatment of both iMSUD and wild type mice show an increase in levels of GDF15 in the serum, but this instead results in gain in weight and fat content in iMSUD mice. In our study, metformin induces the body weight loss in wild type mice ( Figure S1B) which is consistent with previous nding. These results suggest that the effect of GDF15 might depend on the genetic content and metabolic background of mice. The adipokine adiponectin has been shown to improve glucose metabolism, lipid metabolism and insulin sensitivity in numerous tissues (54)(55)(56) and therefore it is interesting that its levels are increased with metformin treatment. The resting insulin levels do not change upon metformin treatment (Fig. 2C), therefore it is possible that the effects of adiponectin might be via improvement of insulin signaling and glucose homeostasis. iMSUD mice have decreased glucose tolerance ( Fig. 2A and B), lower insulin levels (Fig. 2C) and decreased fasting glucose levels (Fig. 2D). Therefore, loss of BCKDH activity associated with iMSUD causes altered glucose homeostasis. It is known that perturbations in BCAA levels in the serum are closely associated with loss of glucose homeostasis and risk of Type II Diabetes (57-60). The iMSUD mice show that a functional BCKDH is required to maintain glucose homeostasis. Treatment with metformin increased IGF-R phosphorylation associated with improved insulin signaling ( Figure S2) and restored glucose clearance capacity (Fig. 2B), without increasing levels of insulin in the serum (Fig. 2C) or improving fasting glucose (Fig. 2D). This suggests that metformin is not able to restore all aspects of glucose homeostasis, but possibly improves insulin signaling and glucose disposal in iMSUD mice.
Study shows that skeletal muscle tissue is a major site of glucose disposal upon metformin treatment (22,61). Our previously result shows that in iMSUD mice there is a loss of skeletal muscle mass (14). More speci cally, we show here that it is the loss of the mitochondrial-rich Type I bers ( Fig. 3B and C) in iMSUD mice, which is partially restored by metformin treatment. This also corresponds to the increased grip strength in mice treated with metformin, which shows a functional improvement of this tissue.
The iMSUD related loss of BCKDH causes decreased TCA components both in vitro and in vivo ( Fig. 4A and B), suggesting that BCAA are a signi cant source of TCA intermediates in all tissues tested, but most signi cantly in liver. This decrease in TCA intermediates is correlated with the diminished energy supply, as judged by the activation of the AMPK pathway (Fig. 3D). It has been proposed that metformin activates AMPK (38,40,42) and in agreement with this, wild type mice have increased phosphorylation of both AMPK and ACC upon metformin treatment. Surprisingly, metformin has the opposite effect in iMSUD mice (Fig. 4D). These data suggest that the previously characterized activation of AMPK by metformin is mostly observed in the wild type mice (22,(30)(31)(32), and that the idea of metformin as an AMPK activator does not extend to all metabolic contexts, especially when the activity of BCKDH is reduced. Future studies are needed to assess the effects of metformin in other altered metabolism contexts.
The activity of the mTOR pathway is lower or unchanged in iMSUD in skeletal muscle tissue of iMSUD mice. This is striking since these mice have signi cantly high levels of leucine the canonical activator of mTOR (47). We present evidence that suggests the inhibitory effects of the metabolites KIC and KMV on mTOR signaling using HEK293T cells imaging of mTOR translocation to lysosomes (Fig. 5B) and western blotting on p70S6K phosphorylation (Fig. 5C). Our results suggest that iMSUD related accumulation of metabolites KIC and KMV might inhibit the leucine mediated activation of mTOR signaling. Our work also suggests that one of the ways by which these organic acids might affect mTOR is by preventing the leucine mediated dissociation of the sestrin2-GATOR2 association ( Fig. 5D and E). Understanding the connectivity between BCAA catabolism and mTOR signaling is critical because both pathways are important for proper metabolism and are altered both by aging and in the context of metabolic disease.

Conclusion
Overall, we show that the iMSUD associated loss of BCKDH controls TCA and glucose homeostasis, and that its dysfunction results in the accumulation of KIC and KMV, which can inhibit the leucine-induced activation of mTOR. Therefore, mTOR positively senses leucine levels, but is also sensitive to the status of BCAA metabolism via KIC or KMV. The mechanism of action of metformin depends on the metabolic background, and it acts as an AMPK activator in iMSUD instead of its canonical role as an AMPK inhibitor.

Mouse strain, housing, diets and drug treatments
The iMSUD and wild-type mice used in this study were obtained by breeding Dbt tm1Geh/+ ; Tg(tetO-DBT)A1Geh/J; Tg(Cebpb-tTA)5Bjd) mice to generate iMSUD or Dbt1 +/+ (wild type) mice homozygous for the LAP-tTA and TRE-E2 transgenes (6). All the mice have free access to water and food and all animal care and experimental procedures were approved by the Institutional Animal Care and Use Committee at the BUCK Institute for Research on Aging. For the lifespan study with metformin diet, the wild type and iMSUD mice were supplied with control (5LG6, TestDiet) or metformin (5LG6 w/0.1% metformin, TestDiet) after weaning. The mice were regrouped after genotyping and assigned group of treatments were continued till naturally death happened.
For the intraperitoneal injection of metformin of the lifespan study in iMSUD mice, PBS-dissolved metformin solution was given to mice at 125 or 250 mg/kg of mouse body weight via intraperitoneal injection daily started from 5-days-old age. After determination of genotype after weaning, both wild type and iMSUD mice were continually given injected till natural deathbed occurred.

Mouse physiological measurements
Body fat composition was determined by in vivo magnetic resonance tomography imaging (EchoMRI) at 3 weeks-old age. The muscle strength was determined by a grip strength behavior task. The grasp strength of each mouse of fore limbs was measured on a pull-grid assembly connected to a grip strength meter (Bioseb). The mouse was drawn along a straight-line parallel to the grid until the grip was broken, providing the peak force in grams. This was repeated at least 4 times within 5 mins intervals between measurements.
Glucose tolerance tests were performed at 4 weeks-old mice. Overnight fasted (16 hrs) mice with free access to drinking water were received 2 g/kg body weight of glucose intraperitoneally, and blood glucose levels were measured before and 30, 60, 90, and 120 min after glucose injection.

Immuno uorescence staining
For the muscle ber type determination, skeletal muscle samples of mice were incubated in 0.5% paraformaldehyde for 3-6 hrs and then incubated with 30% sucrose/PBS at 4°C overnight. The skeletal muscle samples are embedded using OCT (Tissue-Tek) and then cut into 10 um thick cross-sections with a cryostat. Immuno uorescence analysis of MHC expression was performed with primary antibodies against MHC1 (BA-F8, DSHB), MHCIIa (SC-71, DSHB) and MHCIIb (BF-F3, DSHB). Slides were visualized using confocal microscope (Nikon) and nal panaroma images were composed using Image composer editor software (Microsoft research).

Cell line
Human HEK293T cells (CRL-3216) and mouse C2C12 cells (CRL-1722) were purchased from ATCC (ATCC.org). Cells were grown at 37'C and 5% carbon-di-oxide. Cells were grown in standard tissue culture treated dishes purchased from Corning. Inc. Cells were grown in standard DMEM media (Corning) with 5mM glucose and 10% FBS (Corning Catalog # 35-010-CV). Cells were seeded at approximately 50% con uency or at 5000 cells per 96-well plate and 50,000 cells per 6-well plate. Media was changed every 72 hours. Cells were split at 1:5 ratio upon reaching 80% con uency. Differentiation of C2C12 cells was initiated by 10% FBS serum withdrawal and changed into 2% hourse serum in growth media. Penicillin-Streptomycin was added during passaging.

Immunoprecipitation assay
After HEK293T cells were treated with indicated stimulation for 10 min, cells were rinsed one time with ice-cold PBS and immediately lysed with Triton lysis buffer (1% Triton, 10 mM β-glycerol phosphate, 10 mM pyrophosphate, 40 mM HEPES pH 7.4, 2.5 mM MgCl2 and 1 tablet of EDTA-free protease inhibitor (Roche). The cell lysates were cleared by centrifugation at 13,000 rpm at 4°C in a microcentrifuge for 10 minutes. Protein concentration was determined by DC assay and 2 mg protein/sample was used for anti-FLAG-immunoprecipitations. The FLAG-M2 a nity gel was washed 3 times with lysis buffer. 30 µl of a 50/50 slurry of the a nity gel was then added to clari ed cell lysates and incubated with rotation for 2 hours at 4°C. The anti-FLAG-M2 beads were washed two time with lysis buffer. Immunoprecipitated proteins were denatured by the addition of 50 µl of sample buffer and boiling for 5 minutes as described, resolved by 8% SDS-PAGE, and analyzed by immunoblotting.

Viral preparation and transduction
HEK293T cells were grown in 6-well tissue culture treated dishes to 70-80 % con uence. Tubes with 125 ml 2.5 mM CaCl 2 and 375 uL DEPC water were prepared per viral transduction replicate. 500 µl of CaCl2 solution was aliquoted per viral tube. For each virus, 6.5 µg of psPAX2 (helper plasmid) and 3.5 µg of VSVG plasmid to were added to the master mix tube. 10 µg of shRNA plasmid was added with 500 µl of 2X BBS (pH=6.95) to each viral tube, and vortexed for 5 minutes. The tube was incubated for 20 minutes at room temperature. The viral solution was added drop-wise to HEK cells. After 48 hours, the viral supernatant was collected and centrifuged at 1200 rpm for 3 minutes. The Supernatant was ltered through a 0.44 micron lter and frozen in 1 ml aliquots at -80°C.
Cells were transduced at 50-60% con uence. Two µl of polybrene (1:1000) was added near the bottom of a 15 ml tube. 1 ml of growth media media was then added per 1 ml of viral particle containing supernatant. After 24 hours, viral media was removed and cells were removed with PBS and fresh media was added. After 72 hours a puromycin selection was performed. For C2C12 cells, 2 µg/ml puromycin solution was used for selection over 6 days.

Tissue metabolite extraction
Liquid-Liquid extraction (LLE) was performed based on previously reported protocols (52,62) with some modi cations. Approximately 25mg of cryohomogenized tissue was sonicated on ice with 0.5ml 50% methanol with 2µg/ml 13C1-leucine added as an internal standard. A volume of 1.5ml of CHCl 3 with 20pg/ml heptadecanoic acid was then added to each sample and vortexed for 1 min. Each sample was centrifuged at 10,000 x g for 30 min at 4°C to precipitate total protein and separate the lipid and aqueous phases. Following centrifugation, 250µl of the aqueous phase and 1.0ml of the lipid phase were recovered, concentrated by speedvac and N 2 , respectively, and stored at -80°C. The recovered protein fraction was rinsed (3x) with 500 µl of methanol, vortexed and centrifuged at 10,000 x g for 10min. Protein was resuspended in 60 µl of 8M urea + 50mM TEAB pH 7.5 and sonicated in a water bath for 10min. Protein was quanti ed for normalization.
C2C12 intracellular metabolite extraction Extracellular (spent) media was removed from C2C12 cells. C2C12 cells were then rinsed 3 times with phosphate-buffered saline (PBS) and quenched using 1ml 50% methanol. A volume of 2ml of CHCl 3 with 20pg/ml heptadecanoic acid was added to each sample and mixed for 1min at 4°C. Samples were centrifuged at 4,000 x g for 30min at 4°C, to fully precipitate total protein and separate aqueous and lipid phases. Following centrifugation, 1.5ml of aqueous phase were recovered, concentrated by N 2 , and stored at -80°C. The recovered protein fraction was rinsed (3x) with 500 µl of methanol, vortexed and centrifuged at 10,000 x g for 10min. Protein was resuspended in 60 µl of 8M urea + 50mM TEAB pH 7.5 and sonicated in a water bath for 10min. Protein was quanti ed for normalization. C2C12 extracellular aqueous metabolite extraction Extracellular (spent) media (200µl) was extracted using 200µl of 100% methanol with 2µg/ml 13 C1leucine added as internal standard and 600µl of CHCl 3 with 20pg/ml heptadecanoic acid and mixed for 1min at 4°C. Samples were centrifuged at 4,000 x g for 30min at 4°C, to fully precipitate total protein and separate aqueous and lipid phases. Following centrifugation, 100µl of aqueous phase was recovered and stored at -80°C.
Solvent A consisted of 20mM ammonium acetate pH 9.5 with 5% acetonitrile, solvent B was acetonitrile.
The starting gradient conditions were 95% B at a ow rate of 0.3ml/min. The following gradient program was used: 0 to 20min, 95-10% B, 25-30min 10% B, and 30.1-35min 95% B. Samples were kept at +4°C, and the injection volume was 10µL.
Mass spectrometric analysis was conducted using negative ion electrospray ionization in the multiple reaction monitoring mode (MRM) or multiple ion mode (MI) on an API 4000 QTRAP (Ab Sciex, Foster City, CA, USA) mass spectrometer tted with a TurboVTM ion source. The ionization parameters were set as follows: curtain gas (CAD); 20psi; collision gas: medium; ion spray voltage (IS): -4500V; Temperature (TEM): 550°C; Ion source Gas 1 (GS1); 60psi; and Ion source Gas 2 (GS2): 50psi. The compound-speci c parameters were established using the appropriate standards. AB SCiex's Analyst®v1.6.1 was used for all forms of data acquisition and method development.
Quanti cation and statistical analysis AB SCiex's Analyst®v1.6.1 and Varian Saturn software V6.4 was used for in-depth analysis of the LC-MS and GC-MS data, respectively. Software was used for calculating the peak areas for metabolites from muscle and cellular extracts. All other statistical analyses were conducted using GraphPad Prism 8 (GraphPad, La Jolla, CA, USA). The survival curves were completed using Kaplan-Meier curve. We used a log-rank (Mantel-Cox) test to perform the statistical analyses of the survival curves. All data were presented as mean ± s.e.m. Comparisons between groups were performed using unpaired, two-tailed Student's t-tests. P≤0.05 was considered signi cant and p≤0.005 considered highly signi cant.
Quanti cations of western blotting were performed using Image J (NIH). Statistical analysis for lifespan studies were performed using Prism 7 v7.01 (GraphPad) software. Data of lifespan study was analyzed utilizing parametric statistics including Student's t test or one-way ANOVA. Data are expressed as mean s.e.m, and statistical signi cance was accepted when p <0.05. Data are expressed as mean s.e.m, and statistical signi cance was accepted when p <0.05.

Data and Software Availability
The LC-MS data will be available at the NIH Common Fund's Metabolomics Data Repository and B) Body weight recording of metformin diet treated wild type and iMSUD mice. Wild type mice with control diet (WT_control diet, n = 30); wild type mice with metformin diet (WT_metformin diet, n = 45); iMSUD mice with control diet (iMSUD_control diet, n = 21); iMSUD mice with metformin diet (iMSUD_metformin diet, n = 28).

Supplementary Figure 2. Metformin-induced increased activation of IGF-1R signals in iMSUD Mice.
A) Western blots of phospho-IGF-1R and IGF-1R in liver and skeletal muscle of wild type and iMSUD mice treated with PBS or metformin. Adjoining bar graph shows quantitation. B) Metformin has minor effect on glucose tolerance in wild type mice. Glucose tolerance assay were performed on at 50 days-old wild type animals with administration of glucose IP (1.2 g/ kg BW). Wild type with PBS treated for 50day (WT_PBS_50 days, n = 2); wild type mice with metformin (250 mg/kg bw) treated for 50 days (WT_MET_50 days, n = 21). AUC, area under curve. Data are mean ± s.e.m. P value by t-test.  Effect of metformin on glucose homeostasis, the level of insulin and fasting glucose in iMSUD mice. A&B) Metformin improves glucose tolerance in iMSUD mice. Glucose tolerance assessed were perform on at 27 days-old animals with administration of glucose IP (1.2 g/ kg BW). Wild type mice with PBS treatment (WT_PBS, n = 26); wild type mice with metformin treatment (WT_MET, n = 22); iMSUD mice with PBS treatment (iMSUD_PBS, n = 7); iMSUD mice with metformin treatment (iMSUD_MET, n = 8). AUC, area under curve. Data are mean ± s.e.m. P value by one-way ANOVA. C) Serum insulin in metformintreated wilt type and iMSUD mice. Wild type mice with PBS treatment (WT_PBS, n = 9); wild type mice with metformin treatment (WT_MET, n = 8); iMSUD mice with PBS treatment (iMSUD_PBS, n = 6); iMSUD mice with metformin treatment (iMSUD_MET, n = 11). Data are mean ± s.e.m. P value by one-way ANOVA. Partial restoration of TCA metabolites by metformin. A) Heat map of glycolytic and TCA metabolites. The relative level of glycolytic and TCA metabolites from liver, skeletal muscle (gastrastrocnemius) and serum of animals were determined by LC-Mass assay. The bar graphs show relative levels of leucine/isoleucine and ketoisocaproic acid (KIC) in different group (n = 3 or 4/group). The data was shown as mean± s.e.m.

B) Heat map shows relative levels of metabolites from C2C12 cells (extracellular and intracellular portion)
with shNT, shBCKDH with or without metformin (MET) treatment.

Figure 5
Inhibition of leucine mediated activation of mTOR in iMSUD A) Restoration of mTORC1 (p-S6) and mTORC2 (p-Akt) signaling by metformin treatment in iMSUD mice in skeletal muscle

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download.