Blocking ActRIIB signaling and restoring appetite reverses cachexia and improves survival in mice with lung cancer


 The cancer anorexia-cachexia syndrome (CACS) is a common, debilitating condition with limited therapeutic options. The defining feature of CACS is weight loss, which suggests a state of negative energy balance. It is unclear whether this net reduction in energy is due solely to anorexia or if a combination of anorexia and increased energy expenditure (EE) occurs. To address this question, we induced lung cancer in mice and measured changes in food intake, EE, and body composition. Mice with CACS developed reductions in food intake, spontaneous activity, and EE. There was severe atrophy and markers of metabolic dysfunction in the adipose and skeletal muscle tissues as compared to mice without CACS and pair-fed wild-type mice. We used anamorelin fumarate (Ana), a ghrelin receptor agonist, alone or in combination ActRIIB-Fc, a ligand trap for TGF-β/activin family members, to reverse anorexia and skeletal muscle atrophy, respectively. Ana effectively increased food intake and the combination of drugs increased lean mass, restored spontaneous activity, and improved overall survival. These beneficial effects were limited to female mice. Our findings suggest that multimodal, gender-specific, therapies are needed to reverse CACS.


Introduction 1
The cancer anorexia cachexia syndrome (CACS) is a systemic metabolic disorder 2 associated with increased mortality and poor quality of life 1 . CACS is defined by weight 3 loss that preferentially affects tissues that store nutrients like skeletal muscle and adipose 4 tissue. The loss of body weight suggests a state of negative energy balance, which could 5 be the result of reduced food intake (anorexia), increased energy expenditure (EE), or a 6 combination of both. While anorexia is frequently observed in CACS, it remains unclear 7 whether the net reduction in energy balance is due solely to anorexia or if it is paired with 8 an increase in EE. 9 Total energy expenditure (TEE) is controlled centrally with input from peripherally derived 10 hormones like leptin, thyroid hormones, and glucocorticoids [2][3][4] . TEE can be divided into 11 activity energy expenditure (AEE) and resting energy expenditure (REE) 5 . In this 12 simplified model, REE encompasses the basal metabolic rate and the contribution of 13 adaptive thermogenesis determined by changes in temperature and diet. It seems 14 obvious that animals with cancer would have increased REE given the presence of highly 15 metabolic tumor cells, however, the data supporting this assumption are mixed 6 . 16 Cold-induced thermogenesis is a major contributor to REE and TEE in mice, especially 17 when housed under standard conditions at 22°C. At this temperature, brown adipose 18 tissue (BAT) is actively generating heat via mitochondrial uncoupling and futile cycles 7,8 . 19 The same heat-producing pathways can be induced in white adipose tissue (WAT) in a 20 process called browning, which has been implicated in the pathogenesis of weight loss 1 during CACS 9, 10 2 We and others have identified and characterized CACS in a genetically engineered 3 mouse model of lung cancer driven by oncogenic activation of Kras and loss of 4 Stk11/Lkb1 (referred to as KL mice) 11,12 . In this model, the majority of mice develop 5 weight loss, anorexia, and wasting of skeletal muscle and WAT independent of IL-6 and 6 tumor burden. The penetrance of CACS phenotype is incomplete and this feature allows 7 us to compare genetically identical, tumor-bearing mice with and without weight loss 8 under controlled conditions. 9 In this study, we use the KL model to interrogate the changes in body composition, food 10 intake, peripheral organ metabolism, and EE that occur following the induction of lung 11 cancer. Surprisingly, we find that mice with CACS have severely low TEE driven by 12 reductions in food intake and spontaneous activity. We attempted to reverse anorexia in 13 mice with CACS using an anti-GDF-15 antibody, however there was no improvement in 14 food intake. Next, we used anamorelin fumarate (Ana), a ghrelin receptor agonist, alone 15 or in combination with a ligand trap for TGF-β/activin family members (ActRIIB-Fc) to 16 reverse anorexia and preserve skeletal muscle. While Ana increased food intake and fat 17 mass in both groups, only the combination of Ana and ActRIIB-Fc increased lean mass, 18 restored spontaneous activity, and improved overall survival. These beneficial effects 19 were limited to female mice. Our findings suggest that multimodal, gender-specific, 20 therapies are needed to reverse CACS. 21 with CACS ( Figure 2B). These results suggest an overall reduction in pyruvate oxidation 1 in fast-twitch muscles. 2 We performed RNA-seq using the gastrocnemius muscle from mice with and without 3 CACS in order to identify transcriptional changes that could be related to changes in 4 skeletal muscle metabolism. At a whole-transcriptome level, mice with and without CACS 5 clustered independently in an unbiased principal component analysis ( Figure S5A). A 6 pathway enrichment analysis using the differentially expressed genes between the two 7 groups identified multiple downregulated gene sets related to mitochondrial function 8 ( Figure S5B). In accordance with these changes in gene expression, we found that the 9 mitochondrial DNA content and the abundance of mitochondrial proteins were reduced 10 specifically in the EDL of mice with CACS ( Figures 2C and 2D). However, other measures 11 of mitochondrial structure and function showed no difference. For example, there were 12 no consistent alterations at the ultrastructural level ( Figure S5C), no change in whole 13 muscle citrate synthase activity ( Figure S5F), and no difference in the oxidative 14 phosphorylation or electron transfer capacity of permeabilized soleus and EDL fibers 15 ( Figures S5D and S5E). Thyroid hormones are known to alter skeletal muscle metabolism 16 during weight loss but there was no change in the abundance of T3 and T4 in muscle 17 extracts ( Figure S5G). 18 To test the function of skeletal muscle at the whole-body level, we performed a maximal 19 exercise capacity test using a motorized treadmill. The total distance traveled, time until 20 exhaustion, and work performed significantly correlated with weight loss demonstrating a 21 reduction in exercise capacity in mice with CACS ( Figure 2E). Blood lactate levels after 22 AEE between CR and CACS. There was a trend for CCA to be lower in CR mice as 1 compared to WT, suggestive of improved work efficiency 20,21 . 2 Similar to CACS, CR reduced the mass of gWAT, BAT, and glycolytic (but not oxidative) 3 skeletal muscles (Figures S8A and S8B). Histologically, the gWAT and skeletal muscle 4 displayed atrophy with small adipocytes and reduced fiber cross-sectional area, 5 respectively ( Figure S8C). Contrary to what was observed in mice with CACS, the BAT 6 of CR mice was depleted of lipid droplets. Moreover, the UCP1 tissue staining was more 7 intense in the BAT and only identified in rare patches in the gWAT of CR mice ( Figure  8 S8D). 9 The changes in skeletal muscle were also unique following CR. The PFK activity tended 10 to increase in both the Soleus and EDL ( Figure 3E). Moreover, the citrate synthase activity 11 was significantly higher in EDL muscles of CR mice ( Figure 3F) without changes in the 12 abundance of the electron transport chain proteins ( Figure S8E). These changes in 13 skeletal muscle metabolism were associated with improved distance traveled during an 14 exercise performance test ( Figure 3G), however, the work performed was similar given 15 their lower body weight ( Figure 3H). If we consider CR mice as a control for chronic 16 anorexia, these data reveal distinct changes in adipose tissue browning and skeletal 17 muscle metabolism that occur in CACS, a condition with similar food intake, body 18 composition, and TEE. 19

Targeting GDF15 and activin A does not improve CACS in KL mice. 20
It is unclear what drives anorexia and the changes in peripheral organ metabolism in KL 1 mice with CACS. In other cachexia models, GDF-15 induces anorexia through activation 2 of a brainstem "sickness center" 22-26 . Also, Activin-A, a ligand of the ActRIIB receptor, 3 can induce skeletal muscle atrophy and modulate adipocyte browning 25,27,28 . We checked 4 the levels of both proteins and saw that they were specifically increased in the serum of 5 mice with CACS in comparison to WT mice, KL mice without CACS, and CR mice (  Therefore, we sought to inhibit the action of these proteins using monoclonal antibodies 8 (mAb) and test the effects on food intake, TEE, and survival. Before testing the mAbs in 9 tumor-bearing mice, we performed a pilot study to test their safety. WT mice were treated 10 with IgG control (20 mg/kg, QW, SQ), anti-GDF15 mAb (10 mg/kg, QW, SQ) alone, or a 11 combination of anti-GDF15 mAb and ActRIIB-Fc (20 mg/kg, QW, SQ), for 2 weeks. Food 12 intake, body weight, and lean mass were significantly increased with the combination 13 therapy ( Figures S9A, S9B, S9C), and fat mass was subtly reduced in the IgG and anti-14 GDF15 treated mice ( Figure S9D). No alterations were observed in EE, VO2, RER, and 15 activity among the treatment arms ( Figures S9G, S9H, S9I, and S9J) leading to the overall 16 conclusion that these mAbs were safe in mice of this genetic background. 17 We proceeded with a prospective RCT in a cohort of KL mice. Mice were induced with 18 AdCre and then monitored weekly for changes in body weight and food intake. Once the 19 mice reached 15% weight loss, they were randomized to one of the 3 intervention arms. 20 There was no difference in the weight loss at the start of treatment (Week 0) among the 21 arms, as expected ( Figure 4C). After the first week of treatment, the mice treated with the 1 combination of anti-GDF15 mAb and ActRIIB-Fc had significantly less weight loss than 2 mice treated with the control IgG ( Figure 4C). This trend persisted at week 2 of treatment 3 however did not reach statistical significance because many of the IgG-treated mice 4 reached euthanasia criteria (30% weight loss) after 1 week of treatment. Therefore, we 5 compared the changes in body weight among groups at the "End" of treatment, which 6 was defined as the date of euthanasia. In this assessment, the mice treated with the 7 combination therapy showed a significant attenuation in weight loss ( Figure 4D). This 8 response was mostly driven by two female mice where CACS was reversed. The 9 combination therapy also protected against the loss of glycolytic skeletal muscles but not 10 oxidative muscles or adipose tissue ( Figures 4E, 4F, S10A, S10B, and S10C). There were 11 no changes in overall survival, EE, and activity by either treatment (Figures 4G, 4H, and 12 4I), as well as in food intake, VO2, and RER (Figures S10D, S10E, and S10F). Anamorelin hydrochloride was recently approved by the pharmaceutical regulatory 16 authority of Japan for the treatment of patients with CACS 29 . Anamorelin is a non-peptide 17 ghrelin receptor agonist that has been reported to induce food intake and improve survival 18 in patients with CACS 30-32 . Therefore, we planned a prospective RCT in KL mice using 19 anamorelin fumarate (Ana) and ActRIIB-Fc. This combination therapy was determined to 20 be safe in a pilot study of WT mice (data not shown). Using a similar design as the GDF15 21 13 trial, we randomized KL mice into 3 intervention arms (Control, Ana, or Ana + ActRIIB-1 Fc). Mice treated with Ana + ActRIIB-Fc showed a significant increase in food intake and 2 weight loss after 2 weeks of treatment (Figures S11A and S11B). The combination 3 therapy also protected against the loss of fat and lean mass ( Figures S11C and S11D). 4 There were no changes in overall survival and EE by either treatment (Figures S11E, 5 S11F, and S11G); however, spontaneous activity was restored in the mice treated with 6 Ana + ActRIIB-Fc ( Figure S11H) and the gastrocnemius, a primarily glycolytic muscle, 7 was protected (Figures S11I and S11J). 8 During this trial, we noticed that female mice seemed to respond better than male mice. 9 Therefore, we stratified the cohort by gender and interrogated the therapeutic efficacy of 10 these interventions. In this analysis, it was clear that male mice did not benefit from either 11 intervention (Figures S12). However, there was a strong effect in female mice. Ana 12 significantly increased food intake, weight, and fat mass in the female mice ( Figures 5A-13 C). The addition of ActRIIB-Fc to Ana further improved weight due to an increase in lean 14 mass ( Figures 5B and 5D). Remarkably, CACS was fully reversed in two-thirds of the 15 female mice treated with the combination of Ana and ActRIIB-Fc without changes in lung 16 mass, a surrogate for tumor burden ( Figure 5E). The combination therapy restored 17 spontaneous activity and improved overall survival ( Figures 5E and 5F). 18

Discussion 19
In this study, we performed a comprehensive analysis of the changes in food intake, 20 peripheral organ metabolism, and TEE that occur in mice following induction of lung 21 cancer. We show that KL mice with CACS have anorexia and suppressed TEE. The 1 reduction in TEE is more than what is predicted by changes in lean mass alone. This 2 physiologic adaptation to weight loss has been observed in other mouse models of CACS 3 and humans undergoing CR 33,34 . In this setting, the TEE reduction is due, in part, to 4 reduced skeletal muscle glycolytic metabolism and improved work efficiency that can be 5 prevented with low-dose leptin treatment 35,36 . Similarly, KL mice with CACS have atrophy 6 of glycolytic muscle fibers 11 , reduced muscle PFK activity, and low levels of leptin; these 7 changes may improve muscle efficiency and contribute to the decrease in TEE during 8

CACS. 9
Low levels of leptin are also known to contribute to the dramatic increase in spontaneous 10 activity that we observed in mice following CR. This phenomenon has been previous 11 described as "food-seeking behavior" or "semi-starvation-induced hyperactivity" and it 12 can be suppressed by replacing leptin or increasing housing temperature 37,38 . We find 13 that mice with CACS do not display hyperactivity despite the presence of anorexia, 14 hypoleptinemia, and low housing temperature. In fact, spontaneous activity is reduced in 15 KL mice with CACS. Interestingly, we find that increasing the housing temperature 16 normalizes activity in mice with CACS. These data suggest that the neurohormonal 17 pathways regulating thermogenesis may suppress spontaneous activity in mice with 18

CACS. 19
Our data highlights the dramatic changes that occur to adipose tissue during CACS. KL 20 mice develop increased rates of lipolysis, browning, and atrophy of the WAT adipocytes. 21 Data from other mouse models and human studies of CACS show that WAT lipolysis is 22 an essential feature of CACS 22,39,40 , however, the role of browning is more controversial. 1 In certain models, browning exacerbates the negative energy state 9,41 , but this finding is 2 not consistent with clinical studies in subjects with lung cancer nor our data from the KL 3 mice 42-49 . We speculate that the browning observed in KL mice with CACS occurs in 4 response to an increased demand for thermogenesis following the loss of the abdominal 5 "insulation" provided by skeletal muscle and adipose tissues. Furthermore, we identified 6 histologic and biochemical evidence of BAT dysfunction in mice with CACS. The BAT 7 lipid droplets were found to be enlarged and associated with reduced UCP1 mRNA 8 expression and protein abundance consistent with whitening of the BAT. This phenotype 9 has been observed in mouse models of diet-induced obesity where it can be reversed 10 with fenofibrate, a PPARα-agonist 50,51 . Interestingly, we previously showed that 11 fenofibrate can prevent CACS in KL mice so the role of BAT in this syndrome needs 12 further study 11,50,51 . 13 Our results identify the distinct alterations in skeletal muscle metabolism that occur during 14 weight loss from CACS in comparison to weight loss from CR. In both conditions, we see 15 similar reductions in body weight, skeletal muscle mass, and TEE; however, the reduction 16 in exercise tolerance and markers of oxidative metabolism only occur in CACS. It is 17 unclear if the lack of spontaneous activity and limited exercise tolerance is due to malaise 18 (i.e. CNS-mediated) or a primary deficit in skeletal muscle. In support of the latter, Kamei 19 et al. have shown that overexpression of Forkhead box protein O1 (FoxO1) in skeletal 20 muscle is enough to suppress spontaneous activity, and we have previously shown that 21 the muscles from KL mice with CACS have increased expression of this protein 52 . We 22 also found evidence for a reduction in the gene expression and protein abundance of 1 several proteins involved in the electron transport chain in the EDL of mice with CACS; 2 however, there was no change in the oxygen flux of the permeabilized EDL when 3 measured ex vivo. Given that the EDL contains significant numbers of non-atrophied type 4 IIA fibers 11 , we speculate that the activity of these highly oxidative fibers is masking any 5 change in ex vivo oxygen flux. 6 In an attempt to treat anorexia and the loss of skeletal muscle mass during CACS, we 7 treated KL mice with an anti-GDF15 mAb alone or in combination with an ActRIIB-Fc 8 mAb, which induces hypertrophy of glycolytic muscle fibers 53 . Both GDF15 and Activin A 9 are elevated in mouse models and humans with CACS 54-57 , and the systemic inhibition 10 of either signaling pathway can prevent CACS in other mouse models 22,24,58-60 . However, 11 we did not observe significant improvements in food intake or TEE with either treatment. 12 This result may be due to the overall low abundance of GDF15 in the KL mice as 13 compared to other models where anti-GDF15 therapy has shown benefit 22,24,60 . While 14 the addition of ActRIIB-Fc to anti-GDF15 did not improve food intake, it did significantly 15 delay the progression of CACS and preserve skeletal muscle mass. More importantly, we 16 find that the combination of Ana with ActRIIB-Fc significantly improved body composition, 17 activity, and overall survival. The degree of survival improvement we observed is on par 18 with the effects of chemotherapy and immunotherapy in this model 61,62 . 19 The beneficial effects of Ana and ActRIIB-Fc were limited to female mice. We have a 20 limited understanding of the basic mechanisms underlying sex differences in CACS 63 . 21 There are known differences in body composition, EE, and peripheral organ metabolism 1 between men and women 64 . Gender discrepancies have also been observed in animal 2 models of disuse atrophy and cardiac cachexia [65][66][67] . Additional studies are required to 3 identify the key pathophysiologic differences that drive the differential therapeutic 4 response between male and female KL mice. 5 Our data show that CACS in female mice with lung cancer can be overcome by 6 stimulating appetite and blocking catabolic activin signaling. There are late-stage clinical 7 compounds available for the immediate translation of our findings. Anamorelin was 8 recently approved by the pharmaceutical regulatory authority of Japan for the treatment 9 of patients with CACS, and bimagrumab, a fully human monoclonal antibody that prevents 10 ligand binding to ActRIIB, is safe and increases lean mass in adults with sarcopenia 29,68 . 11 Despite the improvement in lean mass, these patients did not show any benefits on 12 physical function and overall survival. Our data suggests that a combination of 13 multimodal, gender-specific, therapies are needed for effective reversal CACS. Translational Science Center (U54 GM104940) and the Scott Rodeo for use of the rodent 8 treadmill. The GDF15 antibody, Anamorelin, and ActRIIB-Fc antibody were provided by 9 Pfizer, Inc. Catecholamines assays were performed by the VUMC Hormone Assay and 10 Analytical Services Core which is supported by NIH grants DK059637 and DK020593.   meter and blood from the tail vein before CO2 asphyxiation. Following euthanasia, 12 whole blood was collected via cardiac puncture and placed into pre-treated tubes for 13 serum/plasma isolation. Next, the liver, gonadal adipose, kidney, and skeletal muscles 14 (gastrocnemius, quadriceps, tibialis anterior, EDL, and Soleus) were dissected, 15 weighed, and flash-frozen in liquid nitrogen. All tissues were subsequently stored at −80 16 °C until further processing. to indirect calorimetry for a period of 3 consecutive days under a 12h light-dark cycle. 20 During this period, we measured food and water intake, spontaneous activity, and 21 volume of oxygen and carbon dioxide consumed. This data allows us to estimate total 22 energy expenditure (TEE) and the respiratory exchange ratio (RER). We recorded the 23 KL mice in two phases: 1-Acclimation, first 24 hours of measurement; 2-Fed, 24 hours 24 after the end of phase 1, mice were fed ad libitum. A penalized spline regression model 25 19 was used to estimate the resting energy expenditure (REE), activity energy 26 expenditure (AEE), and caloric cost of activity (CCA). To calculate CCA, we fit a simple 27 linear regression model between activity rate and total energy rate for each mouse. The 28 slope of each line is the CCA for each respective mouse. To calculate AEE, the 29 mouse's activity rate was multiplied by its CCA. REE was calculated by subtracting AEE 30 from TEE. REE and AEE were then smoothed using a second-order polynomial 31 smoothing spline. This method allows for the calculation of AEE while taking into 32 account time-varying REE. 33 34 Body Composition Analysis. Mice were weighed, and body composition (fat mass, 1 free fat mass, and water mass) was measured using an EchoMRI-100H 2n1 with a 2 horizontal probe configuration (EchoMRI, Houston, TX). 3 4 Exercise capacity test. Mice were acclimated (30 min at 8m/min) to a motorized 5 treadmill one week before the maximal exercise capacity test. Shocking grids with 6 frequency set at: 75 per minute and intensity at: 45% (3.4mA) were located at end of the 7 treadmill to force the mice to run at their maximum. On the day of the test, the protocol 8 was initiated with 3 min acclimation without any speed. Start speed was set to 8m/min 9 followed by incremental adjustments of 2.5m/min every 3 minutes until fatigue was 10 reached. Fatigue was defined as the mouse being stationary on the shocking grid for 20 11 seconds with no attempts to climb off the treadmill. Maximum speed, time and laps were 12 then recorded and used to calculate, total time, total distance, and work. Lactate was 13 quantified before and after the exercise protocol using a point of care device (Nova 14 Biomedical). prospective, randomized, controlled, intervention trials using KL tumor-bearing mice 23 (RCT 1 and RCT2). RCT1 was performed using mAb against GDF15 alone or in 24 combination with ActRIIB-Fc. RCT2 was performed using Ana alone or in combination 25 with ActRIIB-Fc. Analysis of data from our previous cohorts determined that 7 mice per 26 group were required to detect a 20% change in mean weight loss (α=0.05, β=0.9) in 27 tumor-bearing mice. In order to account for early mortality and the proportion of mice 28 without CACS, a total of 39 mice (22 males, 17 females) were induced. Of those, 30 29 mice (18 males, 12 females) reached 10-15% weight loss and underwent 30 randomization. The randomization was performed in blocks of 6 and stratified by gender 31 72 . The pre-specified primary outcome was the percent weight loss at 2 weeks following 32 the start of treatment. Secondary outcomes included weight loss at the time of 33 euthanasia, overall survival, body composition, food intake, spontaneous activity, 34 skeletal muscle mass, and white adipose tissue mass. 35 Serum and Tissue Metabolites. Blood was centrifuged (10,000 × g for 10 min at 4 °C), 1 and the serum or plasma was stored at −20 °C. Serum β-hydroxybutyrate, TG (Stanbio 2 Laboratory), and NEFA (Wako Life Sciences) were determined using commercially 3 available kits. Serum insulin, corticosterone (APLCO Diagnostics), Leptin (Milipore, 4 cat.# EZML-82K), Activin-A (DAC00B, R&D Systems), GDF-15 (MGD150, R&D 5 Systems) levels were quantified by ELISA. Plasma epinephrine and norepinephrine 6 were measured by HPLC via chromatography data station 73,74 by the Vanderbilt 7 Hormone and Analytical Services Core (sensitivity of 0.5 ng/ml for the mouse samples). 8 Serum T3 and T4 were determined by radioimmunoassay in a double antibody 9 technique. T4 was measured by using I125-labeled T4 (MP Biomedicals Cat# 10 06B257231) and 1st antibody developed in rabbit (Sigma Cat# T2652). T3 was 11 measured by using I125-labeled T3 (MP Biomedicals Cat# 06B254282) and 1st 12 antibody developed in rabbit Sigma Cat# T2777). Tissue metabolites (including T3 and 13 T4) were extracted from gastrocnemius (whole muscle) using 80% methanol 75 . 14 Targeted LC/MS analyses were performed on a Q Exactive Orbitrap mass spectrometer 15 (Thermo Scientific) coupled to a Vanquish UPLC system (Thermo Scientific) as 16 previously described 11 . Metabolites were identified on the basis of exact mass within 5 17 ppm and standard retention times. Relative metabolite quantitation was performed 18 based on the peak area for each metabolite. All data analyses were done using scripts 19 written in-house by the WCM Proteomics and Metabolomics Core Facility. 20 21 RNA Sequencing and Analysis. Total RNA was extracted from gastrocnemius (whole 22 muscle) and gonadal WAT (whole depot) using TRIzol (Thermo Fisher), followed up by 23 a clean-up step using RNeasy kit (Qiagen). solution on ice. The muscle bundles were then mechanically separated under a 5 dissection microscope, placed into fresh BIOPS containing saponin (5 mg/mL), and 6 gently agitated at 4°C for 20 min. The fibers were then transferred to a mitochondrial 7 respiration medium (MiR05; 110 mM sucrose, 60 mM K +-lactobionate, 0.5 mM EGTA, 8 3 mM MgCl2, 20 mM taurine, 10 mM KH2PO4, 20 mM HEPES adjusted to pH 7.1 with 9 KOH at 37 °C; and 1 g/l de-fatted BSA), blotted on filter paper, and weighed. 2-5 mg of 10 permeabilized fiber bundles were transferred into the oxygraph chamber containing 2 11 mL of MiR05 until background respiration was stable. OXPHOS and ET capacity were 12 measured using the following concentrations of substrates, uncouplers, and inhibitors: buffer then cut into small strips, maintaining fiber orientation and post-fixed with 1% 23 OsO4-1.5%K-ferricyanide buffer for 60 min. After an additional wash with sodium 24 cacodylate 0.1 M buffer, the muscle was stained with 1.5% aqueous Uranyl acetate for 25 60 min. Samples were dehydrated in a graded ethanol series, followed by acetonitrile 26 for 15 min at room temperature. Samples were embedded in Embed 812 resin (Electron 27 Microscopy Sciences, Hatfield, PA). Samples were cut at 55-60 nm (silver-gold) using a 28 Diatome diamond knife (Diatome, USA, Hatfield, PA) on a Leica Ultracut S 29 ultramicrotome. Sections were contrasted with lead citrate and viewed on a JSM 1400 30 electron microscope (JEOL, USA, Inc., Peabody, MA) operated at 100 kV. Digital 31 images were recorded using a Veleta 2K x2K camera (EMSIS, GmbH). 32 33 Lipolysis Assay. The entire gWAT fat depot was isolated, weighed, and cut. 30-50 mg 34 of tissue was incubated in 600 ul per well of filtered lipolysis medium (DMEM, 2% BSA) 35 with or without Isoproterenol (1µM). At various time points during incubation at 37°C, the 36 medium was collected and glycerol concentration was measured using a free glycerol 37 determination kit (Sigma-Aldrich) according to the manufacturer's instructions. Sample 1 absorbance was measured at 540 nm, using the Epoch™ 2 Microplate 2 Spectrophotometer (BioTek), and glycerol content was normalized to the initial tissue 3 weight. 4 Muscle enzyme activity. Citrate Synthase (CS) and 6-phosphofructokinase (6-PFK) 5 enzyme activity were quantified from EDL and soleus protein lysate using commercially 6 available colorimetric assay kits (CS0720 Sigma-Aldrich Citrate Synthase; and 6-7 Phosphofructokinase Activity Assay -ab155898) according to the manufacturer's 8 instructions and quantified using an Epoch™ 2 Microplate Spectrophotometer (BioTek). 9 Quantification and Statistical Analysis 10 Statistical Analyses. Data are expressed as mean ± standard error of the mean 11 (SEM). Statistical significance for normally distributed data was determined using 12 Student's t-tests for comparisons of 2 groups or analysis of variance (ANOVA) followed 13 by Fisher LSD post-hoc tests for comparisons of 3 or more groups. For metabolic cage 14 analyses, ANOVA with repeated measures and Fisher LSD post-hoc tests were used.

15
Significance was set at P<0.05. Statistical analyses were performed with Prism 7 16 (GraphPad Software) unless otherwise indicated. Quantification of Western blots was 17 performed using ImageJ 1.53a. 18

Study Approval. 19
All animal care and treatments were carried out in compliance with Weill Cornell 20 Medical College Institutional Animal Care and Use Committee guidelines. 21 Data and Code Availability. All data and code to understand and assess the 22 conclusion of this research are available in the main text, supplementary materials, or 23 GEO Database (accession number TBD). 24

Lead Contact 26
Requests for resources and reagents should be directed to and will be fulfilled by the 27 Lead Contact, Marcus D. Goncalves (mdg9010@med.cornell.edu). 28

Materials Availability 29
All reagents are available from the Lead Contact under a material transfer agreement 30 with Weill Cornell Medicine. 31