The TAS1R2 sweet taste receptor regulates skeletal muscle mass and fitness

Muscle fitness and mass deteriorate under the conditions of obesity and aging for reasons yet to be fully elucidated. Herein, we describe a novel pathway linking peripheral nutrient sensing and skeletal muscle function through the sweet taste receptor TAS1R2 and the involvement of ERK2-PARP1-NAD signaling axis. Muscle-specific deletion of TAS1R2 (mKO) in mice produced elevated NAD levels due to suppressed PARP1 activity, improved mitochondrial function, increased muscle mass and strength, and prolonged running endurance. Deletion of TAS1R2 in obese or aged mice also ameliorated the decline in muscle mass and fitness arising from these conditions. Remarkably, partial loss-of-function of TAS1R2 (rs35874116) in older, obese humans recapitulated the healthier muscle phenotype displayed by mKO mice in response to exercise training. Our findings show that inhibition of the TAS1R2 signaling in skeletal muscle is a promising therapeutic approach to preserve muscle mass and function.

Skeletal muscle function and mass can adapt to environmental stimuli by integrating mechanical, 36 hormonal, neuronal, and metabolic pathways. This process involves the activation of sensory 37 pathways to meet energy demands and balance anabolic and catabolic mechanisms. However, 38 the deterioration of these adaptive responses due to aging or obesity can lead to the loss of 39 muscle mass and function. Therefore, it is important to identify and target the molecular networks 40 that preserve or restore muscle health in the development of these conditions. 41 Nicotinamide adenine dinucleotide (NAD) is an endogenous metabolite involved in muscle 42 adaptation. It participates in redox reactions and serves as a substrate for poly(ADP-ribose) 43 polymerases (PARPs) and NAD-dependent deacetylases (sirtuins; SIRT) 1 . Specifically, the 44 activation of the NAD-SIRT1 axis is crucial for maintaining metabolic health 2 . Depletion of cellular 45 NAD is associated to insulin resistance, diabetes 3 , skeletal muscle atrophy and dysfunction 4 . 46 Strategies that enhance NAD synthesis (e.g. NAD precursor supplementation 5,6 , exercise 7,8 , 47 caloric restriction [9][10][11] ) or block its consumption (e.g. inhibitors of PARPs 12-14 or CD38 15,16 ) can 48 improve mitochondrial function, metabolic flexibility, insulin sensitivity and increase muscle mass. 49 For instance, the direct inhibition of PARP1 increases NAD availability for SIRT1 activation, 50 resulting in improvements in muscle fitness 13,14 . However, the physiological pathways that link 51 energy metabolism to PARP1 activity for the coordination of NAD bioavailability have yet to be 52 established. 53 G protein-coupled receptors (GPCRs) are central mediators of cellular function that integrate 54 extracellular and intracellular signaling pathways. They are typically activated by hormones and 55 neurotransmitters, but a growing number of specialized GPCRs have been identified that respond 56 to nutrients or endogenous metabolites 17 . Some of these GPCRs are found on the tongue, 57 intestine, and pancreas where they can directly sense energy substrates such as fatty acids, 58 amino acids, or sugars. For example, sugars are sensed by sweet taste receptors (STR), which 59 are formed by the obligate heterodimerization of TAS1R2 and TAS1R3, to trigger the secretion of 60 ATP (taste buds) 18 , incretins (L-cells) 19,20 , or insulin (beta-cells) 21,22 . Therefore, sugars such as 61 glucose, which typically enter cells to be used for energy or for generating regulatory metabolites, 62 can also independently activate signal transduction pathways through cell-surface GPCRs. In 63 support of a role of STR in endocrine physiology, whole-body genetic ablation of TAS1R2 (bKO) 64 protected mice from metabolic derangements associated with chronic high-fat diet (HFD), such 65 as hyperinsulinemia and liver steatosis 23 . Unexpectedly, bKO mice had higher lean mass in 66 response to HFD 23 , suggesting that the role of STR may extend beyond endocrine physiology. 67 These findings prompted us to investigate whether STR signaling exerts muscle-autonomous 68 effects on the regulation of muscle mass and fitness. 69

71
TAS1R2 regulates skeletal muscle mass and fitness 72 STR signaling genes are expressed in different mouse skeletal muscles (Extended Data Fig.1a), 73 in mouse myofibers in vivo ( Fig.1a and Extended Data Fig.1b), and in mouse and human primary 74 myocyte cultures (Extended Data Fig.1c,d). The active expression of Tas1r2 in adult skeletal 75 myofibers was further confirmed using reporter mice ( Fig.1b and Extended Data Fig.1e) to 76 partially circumvent issues with TAS1R2 antibody reliability (Supplementary Fig.1a). 77 To study the role of TAS1R2 in skeletal muscle, we developed mice with muscle-specific deletion 78 of Tas1r2 (mKO) (Supplementary Fig.1b; see methods) followed by a mouse with transgenic 79 overexpression of the human TAS1R2 in skeletal muscle (mTg) (Supplementary Fig.1c,d; see 80 methods). mKO mice had similar body mass (Extended Data Fig.1f) but increased lean mass 81 compared to WT controls, while reconstitution of TAS1R2 expression in mKO mice (i.e., mTg) 82 restored the WT phenotype (Fig.1c). Consistent with the effects on lean mass, skeletal muscle 83 mass ( Fig.1d and Extended Data Fig.1g) and myofiber cross sectional area ( Fig.1e) was 84 increased in mKO mice and these effects were independent of total body mass (Extended Data 85 Fig.1h), total number of myofibers and nuclei number (Extended Data Fig.1i,j). Increased muscle 86 mass in mKO mice was accompanied by increased muscle strength (Fig.1f) and running 87 endurance (Fig.1g) compared to mTg and WT controls. Moreover, mKO mice showed increased 88 running efficiency (i.e., lower O2 consumption) during moderate intensity exercise (Fig.1h) 89 consistent with reduced glucose clearance in skeletal muscle (Extended Data Fig. 2a) and 90 reliance on glucose oxidation (Fig.1i), which suggests improved mitochondrial function. Intact 91 myofibers (Fig.1j) or primary myocyte cultures (Extended Data Fig.2b) from mKO mice had 92 higher maximum coupled respiration compared to WT controls. Moreover, succinate 93 dehydrogenase (SHD)-positive oxidative muscle fibers (Extended Data Fig.2c), respiratory chain 94 complex protein content (Fig.1k), and mitochondrial density (Fig.1l) were also increased in mKO 95 muscles. The enhanced mitochondrial and oxidative profile of mKO muscles was not due to 96 differences in fiber type distribution (Extended Data Fig.2d). However, principal component 97 analysis (PCA) showed that STR signaling gene expression strongly correlated with genes 98 expressed in Type I and IIa oxidative myofibers (Extended Data Fig.2e). Finally, we assessed 99 the performance of mKO mice after a moderate exercise training program. Four weeks of 100 exercise wheel training improved running endurance in both genotypes, but mKO mice had better 101 responses than WT (i.e., training x genotype interaction, p=0.010), independent of baseline 102 performance and training volume ( Fig.1m and Extended Data Fig.2f). Taken together our data 103 suggest that the deletion of TAS1R2 sugar sensor in skeletal muscle increases muscle mass and 104 fitness and improves adaptations to exercise training. 105 Deletion of Tas1r2 rescues muscle dysfunction associated with obesity and aging 106 Next, we investigated the effects of TAS1R2 on preserving muscle function in ob/ob mice, a 107 genetic model of obesity that manifests muscle atrophy 24 . Genetic deletion of TAS1R2 in ob/ob 108 mice (ob:bKO) displayed reduced fat mass and increased lean mass (Fig.2a) despite similar 109 gains in total body mass (Extended Data Fig.3a). Like in mKO mice, myofiber CSA was higher 110 in bKO and ob:bKO mice compared to lean (WT) and obese (ob:WT) controls ( Fig.2b Extended 111 Data Fig.3b). However, deletion of TAS1R2 did not affect fiber type distribution in muscles of 112 ob/ob mice (Extended Data Fig.3c). Consistent with the muscle-specific phenotype in mKO 113 mice, ob:bKO mice had increased grip strength independent of body mass (Fig.2c), increased 114 mitochondrial protein content (Fig.2d) and SHD levels (Extended Data Fig.3d), and prolonged 115 running endurance (Fig.2e). 116 Aging is also strongly associated with muscle loss and dysfunction 25 . Compared to aged (27-117 month-old) WT controls, aged bKO mice had increased muscle size and mass (Fig.2f,g and  118 Extended Data Fig.3e), independent from total body mass or body size (Extended Data Fig.3f).

119
The effects on muscle mass were accompanied by improvements in muscle strength (Fig. 2h). 120 Aged bKO mice also had increased mitochondrial density (Fig.2i), and improved exercise 121 endurance (Fig. 2j) that cannot be attributed to differences in cardiac function (Extended Data 122 Fig.3g). Taken together the findings suggest that, like in young lean mice, deletion of Tas1r2 in 123 obese or aged mice displays improvements in muscle mass and function. 124 Partial loss-of-function of TAS1R2 in humans recapitulates the muscle phenotype of mKO 125 mice in response to exercise training 126 Ablation of TAS1R2 in mice produced a robust exercise-trained phenotype and preserved muscle 127 function in obese or aged mice. To translate this finding to humans, we evaluated the impact of 128 TAS1R2-Ile191Val partial loss-of function variant 26,27 in older obese individuals who participated 129 in a 6-month trial of diet-induced weight loss with exercise training (WLEX), diet-induced weight 130 loss alone (WL), or an education control (CON) 28,29 . Participants within each group were 131 retrospectively genotyped and classified as Ile/Ile (i.e., TAS1R2 conventional function) or Val/_ 132 (i.e., TAS1R2 partial loss-of-function). We examined parameters that are directly comparable to 133 those assessed in mice: glucose control, body composition, mitochondrial and aerobic capacity, 134 and muscle strength before and after the interventions (Supplementary Fig.2a,b) At baseline (i.e., pre-intervention), there were no major genotype differences in the assessed 140 variables within or between intervention groups (Extended Data Table.2) and the TAS1R2-141 Ile191Val allele frequency and distribution was typical (Extended Data The magnitude of responses varies between participants and individual variables. Therefore, we 152 considered cumulative effects of all measured outcomes to rank the performance of participants 153 from both genotypes for each of the four physiological categories (Fig.3n,o). In the WLEX group, 154 Val/_ carriers scored significantly higher in all categories compared to their Ile/Ile counterparts 155 (Fig.3p). Muscle strength was the only category that did not improve for both genotypes (Fig.3p), 156 but overall, the WLEX group performed significantly better than the WL group (muscle strength  157 WLEX vs. WL, p=0.006; Fig.3p,q). This indicates that diet-induced weight loss may have slightly 158 compromised muscle performance, but concomitant exercise training partially offset this effect. 159 Most notably, when the combined performance from all four categories was considered, most 160 Val/_ participant in the WLEX group responded substantially better to the exercise intervention 161 than any of the Ile/Ile participant, revealing a large effect size difference (Cohen's d=-1.79; Fig.3p  162 and Source Data F3). Our results demonstrate prominent genotype-dependent adaptations to 163 exercise training which, like in mice, show that TAS1R2 partial loss-of-function in humans confers 164 beneficial effects associated with muscle mass and function. 165

TAS1R2 regulates NAD levels in skeletal muscle 166
Mice with genetic ablation of PARP 13,14 or TAS1R2 23 exhibit striking phenotypic similarities 167 ( Fig.1,2), which suggested that the functions of these 2 proteins might be intertwined. mKO 168 muscles had reduced protein poly(ADP)-ribosylation (PAR), a measure of PARP activity (Fig.4a), 169 while PAR levels in WT and mTG muscles were comparable (Extended Data Fig.5a). mKO 170 muscles also had increased NAD levels compared to WT controls and mTg muscles (Fig.4b), 171 consistent with reduced PARP activity 13,14,30 . The TAS1R2 genotype effect on NAD was 172 recapitulated in murine primary myocyte cultures and in muscles from ob:bKO and aged bKO 173 mice (Extended Data Fig.5b,c,d). LC/MS analysis also showed elevated NAD levels in mKO 174 muscles whereas the concentrations of NAD precursors 30 , NAM and NMN (Extended Data 175 Fig.5e), were unchanged. Accordingly, the protein levels of the rate liming enzyme of NAD 176 synthesis, nicotinamide phosphoribosyl transferase (NAMPT), were also comparable in WT and 177 mKO muscle (Fig.4a). Pharmacological inhibition of PARP1/2 with PJ34 increased muscle NAD 178 in WT mice 14 at levels similar to vehicle-treated mKO, but it failed to further increase NAD in mKO 179 muscles, which suggests that the lower PAR levels in mKO muscles were sufficient to maximize 180 NAD (Fig.4c,d). mKO muscles had reduced peroxisome proliferator activated receptor-γ 181 coactivator-1α (PGC1α) acetylation ( Fig.4e) which is consistent with the increased NAD and a 182 modestly elevated SIRT1 protein levels 13,14 (Fig.4f). The activation of NAD-SIRT1 axis was also 183 accompanied by a moderate activation of AMPK 10,31 (Fig.4f), but other related pathways remained 184 unchanged (Extended Data Fig.5f). Taken together, our data indicate that TAS1R2-mediated 185 regulation of NAD is coupled, at least in part, to PARP1 activity. 186 STR link ambient glucose sensing to the regulation of NAD in skeletal muscle 187 C2C12 muscle cells that were either shifted to low glucose 10,32 (Fig.4g, green trace) or treated 188 with PJ34 13 (Fig.4g, gray trace) for 2 days increased NAD levels compared to cells maintained in 189 high glucose alone (Fig.4g, black trace). Notably, when 3-O-methylglucose (3-OMG) -a 190 nonmetabolizable glucose analogue 33 that has equal affinity for TAS1R2 as glucose 34 -was 191 added to cells maintained in low glucose, NAD levels were reversed to those seen under the high 192 glucose condition (Fig.4g, blue trace). As with the chronic treatment above, C2C12 cells in low 193 glucose spiked with equal molarity glucose or 3-OMG for 6h, displayed a comparable drop in NAD 194 levels ( Fig.4h, black bars) that was associated with a corresponding increase in PARP activity 195 (Fig.4i). Consequently, inhibition of PARP with PJ34 reversed NAD levels in both treatments 196 ( Fig.4i, gray bars). Notably, the drop in NAD levels in response to 3-OMG addition was partially 197 prevented with gurmarin, a STR inhibitor (Fig.4j neurons 36-38 , so we activated ERK1/2 in muscle cells with PMA and found it also produced 204 proportional increases in PARP activity (Extended Data Fig.6a). Baseline ERK1/2 205 phosphorylation was similar between genotypes (Extended Data Fig.6b), so to test the STR-206 ERK-PARP1 axis in skeletal muscle we activated STR via intra-muscular injection of sucralose 207 or 3-OMG (Extended Data Fig.6c,d) which rapidly induced ERK1/2 phosphorylation in WT, but 208 not mKO muscles. Similarly, aspartame, a human-specific TAS1R2 agonist (Supplementary 209 Fig.1e), activated ERK1/2 in "humanized" mTg muscles, but not in WT muscles (Fig.5a). Since 210 PARP1 is nuclear, we determined the activation status and localization of ERK1/2 in our models. 211 ERK1, but predominately ERK2, was induced in both the cytoplasm and nucleus in response to 212 aspartame in mTg muscles (Fig.5a). ERK2 was also co-immunoprecipitated with PARP1 in 213 muscle nuclear isolates, revealing these two proteins interact (Extended Data Fig.6e). Using an 214 ERK2-specific phospho-PARP1 antibody (Extended Data Fig.6e), we found increased ERK2-215 mediated PARP1 phosphorylation 36 and PARP activity in mTg muscles treated with aspartame 216 (Fig.5b). In contrast, treatment with EGF, which also robustly activates the ERK pathway 39 , did 217 not increase PARP1 phosphorylation or its activity in skeletal muscle (Extended Data Fig.6f).

DISCUSSION 227
All cells possess glucose sensing mechanisms that link energy availability to cellular processes 228 governing homeostasis. These mechanisms rely on cell uptake of glucose and its transformation 229 to metabolic products that are detected by cellular sensors 40 . In this study, we showed that 230 TAS1R2, the sugar-sensing GPCR originally shown to mediate sweet taste perception on the 231 tongue 41 , is a plasma membrane glucose sensor in skeletal muscle that regulates muscle mass 232 and fitness in mice and humans. 233 TAS1R2 signaling in skeletal muscle activated PARP1 which consumed NAD and produced ADP-234 ribose polymers. On the other hand, attenuating the TAS1R2-PARP1 axis increased the NAD 235 level in skeletal muscles. PARP1 inhibition has been shown to raise NAD levels and cause SIRT1 236 activation and de-acetylation of PGC1α 13,14 . TAS1R2 loss-of-function phenocopied the PARP1 237 loss-of-function in mice which included improved muscle fitness (i.e., mitochondrial capacity, 238 muscle strength and running endurance) arising from activation of the NAD-SIRT1 axis and 239 AMPK 10,31 . The importance of this novel signaling cascade involving TAS1R2 in skeletal muscle 240 is supported by the ability of the TAS1R2 loss-of-function to mitigate the functional decline of 241 muscles associated with obesity or aging. Mechanistically, we showed that stimulation of STR in 242 skeletal muscle induced ERK2, which then activated PARP1 through specific phosphorylation 36-243 38 . The ERK2-PARP1 interaction, which was previously unknown in skeletal muscle, is not 244 indiscriminate. It is likely coupled to specific extracellular stimuli, including the activation of 245 TAS1R2 receptor. Our findings also revealed that NAD levels in vitro could be regulated by 246 glucose dependent on the activation of STR-PARP1 axis. Therefore, we demonstrate for the first 247 time that the well-known metabolic adaptions of NAD in response to ambient glucose 248 fluctuations 9,10 do not solely rely on glucose uptake and metabolism but can also involve nutrient 249 sensing GPCRs. In fact, the affinity values of glucose for TAS1R2 42 coincide with the physiological 250 blood glucose concentrations, suggesting possible physiological ramifications in response to 251 feeding states. 252 One of the main phenotypic attributes of the mKO mice, increased skeletal muscle mass, was not 253 displayed by the PARP1 loss-of-function mice. This finding suggested that TAS1R2 signaling 254 intersected with additional pathways that regulate anabolic processes. STR activation during 255 states of energy surplus (e.g., postprandial) might serve as a bioenergetic checkpoint to curtail 256 redundant muscle growth and/or futile energy cycles. Despite nutrient abundance, STR-deficient 257 muscles do not constrain the PARP1-NAD-SIRT1 axis and anabolism, thereby enabling further 258 positive adaptations in muscle fitness and mass, respectively. In summary, TAS1R2-dependent 259 pathways inform myocytes about the peripheral energy status to modulate, along with other 260 known mechanisms, the intracellular processes that govern muscle structure and function. 261 Genetic manipulations in mice are a powerful approach to decipher the function of genes and 262 their role in the development of disease. However, the resulting mouse phenotypes often do not 263 translate with fidelity to human pathophysiology due to fundamental mechanistic differences and 264 populational heterogeneity inherent to human biology. However, the bKO mice and the partial 265 loss-of-function TAS1R2-Ile191Val human variant (rs35874116) both displayed reduced glucose 266 excursions during an OGTT 20,26 . Hence, we used the TAS1R2-Ile191Val variant to translate the 267 mouse TAS1R2 findings to humans. We performed a retrospective analysis of a randomized 268 interventional study conducted in older obese individuals as this study specifically assessed 269 several measured variables associated with muscle mass, strength, mitochondrial capacity and 270 aerobic performance 28,29 . This approach was necessary to enable direct comparisons between 271 the human and rodent loss-of-function phenotypes. 272 The exercise intervention (WLEX) produced only modest to no improvements in the associated 273 outcomes in the whole study population, a fact attributed to the heterogeneity of responses to 274 exercise training among participants 29,43 . However, retrospective consideration of the Ile191Val 275 variant as a factor revealed characteristics that phenocopied the mKO mice. Specifically, 276 participants with TAS1R2 partial loss-of-function (Val/_) exhibited increased muscle mass and 277 improved mitochondrial capacity and aerobic performance in response to the exercise 278 intervention, unlike participants with "wild type" TAS1R2 function (Ile/Ile). The exercise training 279 protocol was primarily focused on improving aerobic performance, which likely explains why 280 muscle strength remained unchanged for both genotypes. Notably, the genotype differences in 281 response to the exercise intervention were comprehensive. An analysis of cumulative effects 282 revealed partitioning between genotypes, whereas Ile/Ile participants were low or no responders 283 to exercise, and Val/_ participants were high responders. The genetic and physiological factors 284 contributing to the heterogeneous response to exercise are under ongoing investigation 44 . 285 Therefore, the predictive value of this SNP should be further tested though specialized 286 interventional studies. Finally, Val/_ carriers also showed significant improvements in glucose 287 control. Specifically, the lower HbA1c in Val/_ carriers is consistent with previous observations 27 . 288 These outcomes were not seen in the Ile/Ile counterparts, even though both genotype groups had 289 similar baseline glycemic profiles and lost the same amount of body mass in response to the 290 intervention. It is not apparent which tissues expressing STR contributed to the effects on glucose 291 control, but inhibition of TAS1R2 signaling might have independent effects on glucose 292 homeostasis unrelated to skeletal muscle function. 293 Taken together, our pre-clinical and clinical data suggests that TAS1R2 signaling regulates 294 muscle structure and function. GPCRs are proven targets for pharmacological interventions 45 . 295 Hence, inhibition of TAS1R2 signaling could be a promising therapeutic approach for treating 296 muscle dysfunction associated with aging, obesity or other conditions. 297

ACKNOWLEDGMENTS 298
This work was supported by the National Institutes of Health (R01DK127444 to GAK; 299 R01AG021961 to BHG; U24DK097209 to University of Florida's Southeast Center for Integrated 300 Metabolomics; P30NS104177 and S10OD026842 to OSU Neuroscience Imaging Core/Brown, 301 PI), the National Institute of Food and Agriculture (NIFA-2018-67001-28246 to GAK), The 302 American Heart Association (AHA904048 to JS), and Advent Health (to GAK and REP). 303

DATA AVAILABILITY 304
The data generated and analyzed for the current study are available from the corresponding 305 author (GAK). The human data that support the findings of this study are not openly available 306 which is stored in a controlled access repository. We will share datasets within the restrictions of 307 institutional review board ethics approvals, upon reasonable request from PMC 308 (Paul.Coen@AdventHealth.com). 309           Table 3 Allele Frequency χ² P

Animals
All animal procedures were performed under the approval of The Ohio State University institutional animal care and use committee (IACUC). Mice were housed on a 12 h light/dark cycle with free access to water and standard diet. C57BL6/6J, routinely obtained every 5 th generation from the Jackson lab (Jax# 000664), were used to backcross all mouse models. Tas1r2 Cre knock in (Tas1r2 iCre ) was previously generated as described 1 . tdTomato mice (Td fl/fl ) were obtained from the Jackson lab (Jax#007914) and crossed with Tas1r2 iCre to generate a Tas1r2 reporter (Td Tas1r2 ). Myogenin-Cre mice (Myo Cre ) were a gift by Dr. Olson. RiboTag mice (Rpl22 fl/fl-HA ) were obtained from the Jackson Lab (Jax# 029977) and crossed with Myo Cre mice to generate a myocyte RiboTag reporter (Ribo Myo ). Tas1r2 floxed mice (Tas1r2 fl/fl ) were obtained from the KOMP Repository (#CSD25803) and crossed with MyoCre mice to generate muscle-specific knockouts (mKO). Human TAS1R2 transgenic mice (TG) were created by Ingenious targeting lab (NY) by placing a cDNA cassette containing a synthetic CAG promoter, a floxed NEO stop cassette, the consensus human TAS1R2 cDNA, an IRES-EGFP and a BGH polyA on the mouse ROSA26 locus. After backcrossing, the TG allele was extracted into the Tas1r2 fl/fl background and further crossed with the mKO line to generate muscle-specific transgenic mice (mTG). Specifically, we routinely bred Cre + mTG lines with Cre -Tas1r2 fl/fl dams to obtain mixed litters of wild type (WT), mKO and mTG experimental mice. Whole body Tas1r2 knockouts (bKO) were a gift from Dr. Zuker. Leptin deficient Lepob mice (ob) were obtained from the Jackson Lab (Jax #000632) and crossed with bKO mice to generate Tas1r2-deficient ob mice (ob:bKO). Mice were used for experimental purposes only after reaching sexual maturity at 12 weeks of age.

Mouse genotyping
Mice were ear-punched at weaning and punches were digested with 100µL of 50mM NaOH for 1h at 95°C. Samples were then cooled at room temperature, spun, and neutralized with 20µL of 1M Tris pH8 and 100µL of water. Clear homogenates were subsequently genotyped by PCR and electrophoresis using specific primers and conditions as stated in resource data.

Sweetener consumption
An automated phenotyping homecage system (TSE PhenoMaster) was used to perform behavioral and metabolic tests in single-caged mice. The specificity of the TG construct was tested in bTG and WT mice by recording daily liquid intake of water or 5mM aspartame and 600ppm sucralose in 3 subsequent, alternate days.

Body composition
Body composition was measured in duplicate using an EchoMRI instrument, calibrated daily with 40.5g of canola oil, and percent composition was calculated after subtraction of free water. Body length and tibia length were measured at harvest using pre-calibrated Traceable digital calipers (Fisher Scientific).

Grip tests
Forelimb strength was measured with a 1027SM-D66 grip strength meter (Columbus Instruments). Mice forelimbs were positioned evenly on a slanted grid and mice were pulled horizontally from the tail for 5 times, spaced at least 30 seconds to avoid exhaustion. Test were performed in duplicate in different days and values were averaged.

Exercise training
Mice were housed individually for 30 days in standard cages containing an in-cage running wheel (STARR Life Sciences) to provide voluntary access to physical exercise. Wheel revolutions through glass reed switches and recorded every 30min with VitalView data acquisition software (STARR Life Sciences). Revolutions were transformed to distance in cm through an estimated internal diameter of 11.0998 cm. The distance run per day was averaged to estimate the training volume.

Exercise calorimetry
Gas exchange during treadmill exercise was performed as previously described 2 . Briefly, mice were acclimated to an enclosed, motorized treadmill connected to O2 and CO2 sensors (Columbus Instruments, Columbus, OH) by performing an acclimation run (10 m/min for 10 min). Two days later, mice were placed back in the enclosed treadmill and were allowed to acclimate for 30 min. The exercise test then began with a 2 min run at 10 m/min at a 0% incline which was then adjusted to 20 m/min at a 10% incline for 60 min. Mice were encouraged to run using an electric shock grid at the back end of the treadmill (1.5 mA, 200 msec pulses, 4 Hz). Oxygen consumption (VO2) and carbon dioxide production (VCO2) in ml/(kg•min) were continuously measured with an airflow rate of 1 m/min. RER was calculated as VCO2/VO2.

Peripheral glucose uptake during exercise
Tissue glucose clearance was determined as previously described 2 . Indwelling catheters were surgically implanted in the carotid artery and jugular vein for blood sampling and infusions, respectively. Following a 5-day recovery period, mice were acclimated to an open, motorized treadmill (Columbus Instruments, Columbus, OH) by performing a 10 min run at 10 m/min. Two days later, 5h-fasted mice were placed in the treadmill and underwent a 30 min exercise bout at 16 m/min, 0% incline. Mice were encouraged to run using an electric shock grid at the back end of the treadmill (

Mouse in vivo treatments
Administration of 10mg/kg PJ34 was performed IP twice a day for 7 days. For intramuscular sweetener injections, 5h fasted mice were anesthetized with pentobarbital and 10µL of sucralose (100mM), 3OMG (30mM) or aspartame (5mM) were injected with a 20µL Hamilton syringe in the vastus medial using the contralateral leg as saline-injected control. Muscles were dissected exactly 10min after each injection and snap-frozen in liquid nitrogen.

Transthoracic echocardiography.
Cardiac function and heart dimensions were determined by 2-dimensional echocardiography using a Visual Sonics Vevo 3100 Ultrasound (Visual Sonics) on mice under anesthesia (initially 2.5% isoflurane mixed with 1 liter per minute 95%O2/5%CO2 and maintained with 1-1.5% isoflurane). A single observer blinded to mouse genotypes performed echocardiography and data analysis. M-mode tracings were used to measure left ventricular dimensions at end diastole and end systole. Ejection fraction (EF%) was calculated by: EF(%) = (EDV -ESV)/EDV, where ESV represents end systolic volume and EDV represents end diastolic volume.

CT scanning
In vivo muscle CSA was assessed under 2% isoflurane anesthesia using a SkyScan 1276 computed tomography scan with a 40.9µm resolution. Images were subsequently converted to NIFTI format using FIJI (ImageJ 1.53t) and skin, bones and muscles were reconstructed in ITK SNAP 3.8.0 with before calculating the posterior muscle CSA at 1/3 of the tibia intercondylar notch.

Harvest and tissue processing
All tissue harvests were performed in vivo under 165 mg/kg pentobarbital anesthesia using heat pads to sustain body temperature. Muscles were harvested tendon to tendon starting with the left leg and immediately snap-frozen in liquid nitrogen. Samples for RNA extraction were stored at -80 °C until use. Samples for protein determination and metabolomics were lyophilized overnight at -50 °C and 0.5mbar and then pulverized in 2ml Sarstedt tubes containing 2 RNase-free plastic beads in a Precellys homogenizer at 0°C x 7200RPM x 20sec x 15sec pause x 3 cycles.

Primary myocyte isolation and culture
Mouse hindlimb muscles were dissected in PBS, sliced into 1mm pieces in HBSS and digested for 45min at 37°C using 400U/ml collagenase and 0.08 U/ml dispase II. Tissues were then triturated by aspiration with a 1ml pipette, incubated for 45min more and the reaction was stopped by addition of pre-warmed pre-plating medium (PPM, 90% low glucose DMEM, 10% FBS, 1x glutamax, 1% penn-strep). Cells were strained, spun at 300g, re-suspended in PPM and incubated into a 100mm tissue dish for 3h. Myoblasts were then released from the culture dish by gentle agitation leaving the adherent fibroblasts behind and the medium was spun at 300g.

C2C12 culture and studies
C2C12 (ATCC CRL-1772) were cultured in high glucose DMEM with 10% fetal bovine serum and 1% Penn/Strep using 75cm 2 flasks. Cells were passaged in 24-well plates for NAD or 6-well plates for signaling studies at a density of 52632 cells/cm 2 and media was changed to high glucose DMEM with 2% horse serum and 1% Penn/Strep upon confluence to start the differentiation process. NAD experiments were performed by switching to low glucose DMEM 2% horse serum and 1% Penn/Strep at day 4 of differentiation and supplementing it with combinations of glucose (25mM), 3-o-methylglucose (25mM) or PJ-34 (1uM) during the following 2 days. After 2 days in low glucose, selected experiments evaluated the acute effects of glucose (25mM), 3-omethylglucose (25mM), PJ-34 (1uM) or gurmarin (30ug/ml) by spiking stock volumes in the wells and harvesting 6h after. Signaling experiments were performed at day 6 of differentiation after 10min of equilibration in EBSS and subsequent incubation of 2x stocks in EBSS to achieve 10mM sucralose, 10µM PD98059, 10ng/ml EGF or 0.1mM PMA.

Oxygen consumption in isolated fibers and primary cultures
Measurement of mitochondrial function (oxygen consumption) in muscle fibers and cells was performed as described 6 . Oxygen consumption in saponin-permealized fibers was performed using Oxygraph 2K (Oroboros, Austria). Maximal respiration supported by electron flux was measured with the addition of ADP (500 mM) after stabilization on saturating pyruvate-malate (2M). Steady-state O2 flux for was determined and normalized to fiber dry weight using Datlab 6 software (Oroboros). Primary cells were seeded at 6000cells/well into XF96 culture plates, differentiated, and assayed in XF assay medium containing 4.5g/L glucose and 0.11g/L pyruvate. Basic respiration was assayed by sequentially injecting 0.8ug/ml oligomycin, 1µM FCCP and 2µM antimycin.

RNA isolation, gene expression and analysis
Frozen tissue samples were homogenized with 1mL of TRIzol (Ambion), 3

Ribotag RNA isolation and analysis
Muscle samples were harvested, washed in PBS with 100 μg/ml cycloheximide, and frozen in liquid nitrogen. Frozen samples were pulverized and homogenized in 1mL of cold polysome buffer (20 mM Tris pH 7.4, 10 mM MgCl, 200 mM KCl, 2 mM DTT, 1% Triton X-100, 100 μg/ml CHX) using 2mL Dounce homogenizer. Homogenates were transfered to a 1.5ml RNase-free tube and centrifuged at 17,000xg, 4°C, for 10 minutes. The supernatant was brought to a fresh 1.5mL tube and 100µL of sample was saved as input. The remaining supernatant was incubated with 4μL of anti-hemagglutinin antibody (catalog number) for 3-6 hours at 4°C with constant rotation. 50μL of AG Magnetic Beads (catalog number) were washed 5 times with polysome buffer, re-suspended into 50μL of polysome buffer, and added to the samples. Following overnight rotation, the samples were placed on a magnetic rack and washed 5x with 500μL of high salt buffer (20 mM Tris pH 7.4, 10 mM MgCl, 300 mM KCl, 2 mM DTT, 1% Triton X-100) before elution with 300μL of TRIzol.
The supernatant was mixed with an equal volume of ethanol and RNA was immediately extracted using a purification kit (Zymo Micro). Reverse transcription and real time PCR were performed thereafter with our standard protocol. Gene enrichment was calculated as the ration between elution and input expression. Associations with cell markers were evaluated through principal component analysis of eluate gene expressions after normalization with 18s cDNA.

Muscle fiber CSA, fiber number and nuclei number
Muscles were dissected, fixed with 1% PFA in PBS for 1h at RT, and cryo-protected in 30% sucrose overnight. Muscles were then frozen in OCT using nitrogen-cold methyl-butane, kept on dry ice and cut in 12µm sections. Slides were then brought to RT, washed with PBS 0.05% tween, washed again with PBS, and incubated overnight in a humid chamber at 4°C with primary antibodies. Slides were then brought to RT, washed with PBS 0.05% tween, washed again with PBS, and incubated for 1h in a humid chamber at RT with secondary antibodies. Slides were then washed with PBS 0.05% tween, incubated with DAPI, washed with PBS, mounted with IF mounting medium (Southern Biotech Fluoromount G 0100-01) and placed in a cooled to RT foilcovered dessicating chamber with EMD-Millipore-Sigma DX0017 beads. Images were captured with Cytation 1 using Gen5 software (BioTek). Images were then stitched and analyzed with a laminin primary mask to define the fibers. Fiber type and associated nuclei were determined by the image channels.

Immunohistochemistry and tissue fluorescence
For SDH, tissues were frozen in OCT, then sectioned and stained within 5 days of freezing. Slides were incubated in SDH solution (48mM succinate, 1mM 1-methoxyphenzine methosulphate, 1.5mM Nitroblue tetrazolium in PBS) for 10 minutes and then dipped into ddH2O to stop the reaction. Slides were washed in 1X PBS and blocked for 45 minutes. Laminin was applied as primary antibody and used thereafter as primary mask. mKO slides were imaged using Cytation 1 and Gen5 software (BioTek). OB mice slides were imaged in the SBP core using Aperio Scanscope (Leica) software. For TdTomato tissue fluorescence, muscles were cryo-protected in 30% sucrose overnight, then frozen in OCT, sectioned, mounted (Prolong Gold Antifade Mountant), and imaged using Zeiss LSM 900 confocal with a 581nm filter and Zen 3.0 Zeiss Confocal software.

Electron Microscopy
Transmission electron microscopy imaging of muscle samples was performed by the OSU imaging core. Soleus samples were specially chosen for its fast fixation, thus preserving mitochondrial ultrastructure. Samples were dissected and fixed in 2.5% glutaraldehyde in 0.1 M phosphate buffer. Samples were postfixed with 1% osmium tetroxide and then en bloc stained with 1% aqueous uranyl acetate, dehydrated in a graded series of ethanol, and embedded in Eponate 12 epoxy resin (Ted Pella Inc., Redding, CA). Ultrathin sections were cut with a Leica EM UC7 ultramicrotome (Leica microsystems Inc., Deerfield, IL) and collected on copper grids. Images were acquired with an FEI Technai G2 Spirit transmission electron microscope (Thermo Fisher Scientific, Waltham, MA) operating at 80kV, and a Macrofire (Optronics, Inc., Chelmsford, MA) digital camera and AMT image capture software. For random sampling during imaging, at least 3 myofibers per subject were identified and then each was imaged in at least 4 random locations. The process was repeated at 11500x and 34000x.

NAD cyclic assay
NAD was determined by a cycling assay based on alcohol dehydrogenase after the acidic destruction of NADH 7 . 500µL of ice-cold 0.6M perchloric acid was added to pelleted cells or to 2-3mg of lyophilized tissue placed in 500µL skirted tubes with 5mg of glass beads. Samples were homogenized at 0°C in a Precellys homogenizer with 3 cycles of shaking at 7200RPM x 20sec and 15sec pause, repeated 4 times. The homogenates were centrifuged at 12000g during 20min and 4°C, and 300µL of acidic extract was saved in a new ice-cold 1.5ml tube and the acidic pellet was saved for protein determination. Samples were then diluted 1:10 (cells) or 1:100 (tissues) with 100mM PBS right before starting the assay. The assay was performed in triplicate in black opaque 96 well plates containing 5µL of sample and standards and 95µL of freshly-made cycling assay (100mM PBS, 0.1% BSA, 10mM NAM, 2% ethanol, 10µM FMN, 20µM resazurin, 100µg/µL ADH, 10ug/ml diaphorase) and 530/590 fluorescence was recorded for 30min. NAD from muscle samples was normalized by mg of dry tissue. For cell determinations, the total protein per well was used to normalize NAD. The acidic cell pellet was centrifuged again at 12000g during 20min and 4°C and any remaining supernatant was discarded. The pellet was then neutralized with 50µL 1M NaOH, and homogenated at 7200RPM x 20sec three times or until solubilization. The basic homogenate was then spun and neutralized with 250µL 100mM PBS before measuring protein with a BCA kit.

Muscle metabolomics
Targeted, quantitative metabolomics of pyridines was performed from lyophilized gastrocnemius muscle as previously described 8 . Briefly, oxidized nucleotides were extracted in 0.5M perchloric acid, stabilized with 1M ammonium formate and separated by liquid chromatography on a Hypercarb column (Thermo) with a mobile phase of 10 mM ammonium acetate pH 9.5 and a gradient of acetonitrile. Nucleotides were quantified by single reaction monitoring using a Quantiva triple quadrupole (Thermo) operating in positive ion mode and an electrospray ionization capillary voltage of 3500 V.

Protein extraction
4-6mg of lyophilized powdered tissue was weighed in 2ml Sarstedt tube and protein was extracted with 200µL of RIPA buffer supplemented with protease and phosphatase inhibitors (Roche). Samples were placed 10min on ice and then homogeneized with 3 cycles of shaking at 7200RPM x 20sec and 15sec pause in a Precellys. After spinning at 12000g during 30min, supernatants were placed in new tubes and protein was measured immediately with a BCA kit. Samples were then brought to 2mg/ml with supplemented RIPA buffer and 6x reducing buffer and stored at -20°C if not used immediately.

Immunoblotting
Samples were heated for 10 minutes at either 65 °C (samples probed for parylation) or 85 °C and 10-30µg of protein per lane was loaded into standard 7-10% reducing acrylamide gels for electrophoresis. Protein transfer onto PVDF (for mitochondrial proteins) or nitrocellulose membrane was performed in Towbin buffer at 100 V for 90 minutes inside a cold 4°C room. After the transfer, blots were Ponceau-stained, cut into pieces to match the molecular weight of a target and a housekeeping protein, destained, and blocked with 5% dehydrated milk in TBST buffer.
Blots were then probed overnight at 4°C with corresponding primary antibodies and continuous rocking. After 3 washes in TBST, blots were incubated for 1h with secondary antibodies, washed again and visualized with SuperSignal reagent in a Bio-Rad ChemiDoc imaging system. Original images are shown at "Resource Data images" file.

Nuclear isolation
Lyophilized, pulverized muscle powder (12-15 mg) was resuspended in 300 µL PBS and homogenized via dounce homogenizer. Homogenate was centrifuged at 1900 x g, 4 °C for 15 minutes, and supernatant was isolated as cytosolic fraction. The pellet was resuspended in 150 µL CER I buffer and incubated 10 min on ice. Buffer CER II (8.25 µL) was added to each sample, incubated 1 min on ice, then centrifuged at 16,000 x g, 4 °C for 5 min. Supernatant was collected and added to the cytosolic fraction. Pellet was resuspended in 75 µL NER and incubated on ice for 40 min, with vortexing every 10 min. Samples were centrifuged at 16,000 x g, 4 °C for 10 minutes. Supernatant was collected as nuclear fraction. Sample protein concentration was determined via BCA assay.

Immunoprecipitation
Protein samples (450 µg of each) were diluted to 1 µg/µL in RIPA buffer with protease and phosphatase inhibitors. To pre-clear samples, 50 µL Protein A/G-conjugated agarose beads were added to each sample and incubated for 60 min. at 4 C while rotating. Beads and supernatant were separated magnetically, and 50 µL of supernatant was isolated as "input" sample.
Remaining supernatant was incubated with 4 µL PARP antibody overnight at 4 °C with rotation. Protein A/G-conjugated agarose beads (50 µL) were added to each sample and incubated at 4 °C for 2 hours. Beads and supernatant were separated magnetically, and supernatant was discarded. Beads were washed 5X in 500 µL cold RIPA with inhibitors. SDS 6X reducing buffer was added (15 µL) and samples were incubated at 65 °C for 8 minutes to elute protein, after which 15 µL RIPA with inhibitors was added and protein samples were magnetically separated from beads.

Study design and participant characteristics
The study was a randomized controlled trial with a parallel group design between conducted between 2012 and 2017 at the University of Pittsburgh and the AdventHealth Translational Research Institute (AH TRI). Eighty-four physically inactive men and women aged between 60-80 years of age and with obesity were randomized into one of three 6-month interventions with a 1:1:1 allocation ratio; Health education control (CON;); calorie restriction-induced weight loss (WL), and weight loss with exercise (WLEX). Written informed consent was obtained from all participants prior to participation. Randomization was performed electronically using a random allocation sequence designated by the study statistician. A permuted-blocks approach using blocks of random sizes of 4 and/or 8 was used, with groups stratified by gender. The study coordinator was responsible for participant enrollment and group assignment. Outcome assessors were blinded to group assignment. Sixty-one participants received allocated intervention. The study protocol was approved by both University of Pittsburgh Research Ethics Board and Institutional Review Board of AdventHealth. The study was also registered on ClinicalTrials.gov (NCT02230839).
Participants were enrolled in the study if they met the following screening criteria: BMI ≥ 30 kg/m2; stable weight over the last 6 months; physically inactive (≤1 continuous exercise session/week); non-smoking; resting systolic blood pressure (SBP) < 150 mmHg and diastolic blood pressure (DBP) < 95 mmHg. Exclusion criteria included: clinically significant CVD including history of myocardial infarction within the past year; peripheral vascular disease; hepatic, renal, muscular/neuromuscular, or active hematologic/oncologic disease; presence of bruits in the lower extremities, history of pulmonary emboli; peripheral neuropathy; anemia; substance abuse.

Health education control (CON) intervention
Participants randomized to the CON group received bi-weekly general health education seminars on medication and type 2 diabetes management. However, they were not given specific exercise or dietary education.

Calorie restriction-induced weight loss (WL) intervention
The goal of the WL intervention was to lose 10% of baseline body weight. To achieve this goal, we used the Harris-Benedict equation with a correction for physical activity factor, to calculate a reduction of 500-1000 kcal/day based on baseline body weight. This reduction in caloric intake was prescribed along with a low-fat (< 30% of kilocalories from fat) diet. To encourage compliance, participants met individually with the Registered Dietitian and/or designated staff weekly to record body weight, review daily food logs, and receive updated dietary guidelines. To eliminate the confounding effects of acute caloric restriction on insulin sensitivity, participant weights were kept stable during the last two weeks of the intervention.

Weight loss and exercise (WLEX) intervention
The WLEX group completed a progressive 6-month exercise training program. The program consisted of exercise on 4-5 days per week, with 45 minutes per session (180 minutes per week) and consisting of mostly walking (outside and on an indoor treadmill) with the option of utilizing stationary cycling, elliptical and rowing machines. Endurance exercise was performed at 50-80% HRreserve. Exercise performed indoors at the training facility was supervised by a trained research assistant, while exercise performed outdoors was not supervised. The WLEX group was also prescribed two, non-consecutive resistance 30 min session exercise sessions per week starting at week 8. The exercises focused on major muscle groups using resistance exercise machines. Nine exercises (2-3 sets of 10-12 repetitions) were performed that alternated between upper and lower limbs, and trunk. The resistance exercises were performed at the highest weight the participant could achieve for the given number of reps (10-12) with proper form. When the participant reached 3x12 reps, the weight was increased, and the number of reps were decreased. Participants also met with the Registered Dietitian and received identical dietary instruction as the WL group.

Blood analyses
Glucose and HbA1C were measured by a fasting blood draw and analyzed in the clinical chemistry laboratory at AH TRI using standard assays. Blood was also drawn, centrifuged, and buffy coat was collected and frozen for DNA isolation and genotyping.

Cardiorespiratory fitness, muscle strength, and mitochondrial capacity
A cardiopulmonary graded exercise test was conducted by an exercise physiologist on the cycle ergometer using open circuit indirect calorimetry to measure maximal oxygen uptake (VO2peak). Following a standardized warm-up, participants exercised at a moderate intensity and resistance increased gradually until volitional fatigue.
Muscle strength and power were assessed at baseline and 6 months using a pneumatic-driven dynamometer (Biodex 4, Biodex Medical Systems, Inc., NY, USA). Following a one-minute warmup of free pedaling on a cycle ergometer, participants were seated on the Biodex machine with the lateral condyle of the knee lined up with the axis of rotation of the machine arm. Participants performed three tests on each leg at each resistance of 60, 120 and 180 degrees per second with a ~2-minute rest between each adjustment and a 5-minute rest between legs. Muscle average power and torque were calculated as the average of 60-, 120-and 180-degree tests.
In vivo muscle mitochondrial function was assessed at rest in the vastus lateralis using phosphorus ( 31 P) magnetic resonance spectroscopy on the 3-T magnet as previously described 9 .

Body composition
Weight and height were measured preand post-intervention and BMI was calculated. Waist circumference was measured using the Gulick II tape measure directly on the skin. Fat mass (FM; kg) and fat-free mass (FFM; kg) were determined by dual-energy X-ray absorptiometry (DXA) using a GE Lunar (GE Healthcare, UK). Skeletal muscle index was calculated as appendicular lean mass normalized by the square height.

Statistics
Inferential statistics were performed by null hypothesis significance testing at a significance level of 0.05 and assuming normal distributions. Two-tailed tests were systematically employed throughout to infer a difference against the null. One-tailed tests were performed in a small number of preclinical experiments that were specifically designed to test directional outcomes after substantial precedents were accumulated.
Preclinical data was analyzed with GraphPad Prism 9. Planned comparisons between two groups were analyzed by t-tests, while non-orthogonal planned comparisons were performed through Šidák-corrected values after ANOVA. Planned comparisons and post-hoc tests within 2factor designs were also Šidák corrected after ANOVA, except two western blot experiments run in multiple gels where t-tests were employed. Linear regressions were performed to either test for a significant association between variables or to compare adjusted means between two groups. For the latter, no significant difference between slopes was observed in any case.
Clinical data was analyzed with jamovi 2.2.5. Genotyping results were tested for Hardy-Weinberg equilibrium and allele distribution against a standard population using Chi square tests. For the retrospective evaluation of the genotype effects, we considered a factorial design (intervention group and genotype) which was corrected for sex, age and diabetic status to limit any effect of confounding factors (GAMLj linear regression module). Baseline values were tested for genotype differences (Welch) and random allocation (ANOVA and Tukey-corrected post-hocs).
Delta values were used to analyze the intervention effect as recommended for retrospective studies 10 . The main intervention effect (independent of genotype) on each measured outcome was tested through ANOVA and Tukey-corrected post-hoc. Genotype differences within the WLEX group were also tested within the regression module as a single independent contrast. The possibility that the two genotype subpopulations would respond differently to health education control (CON) was discarded by Welch tests (Extended Data Table 4). The intervention effects for each genotype subpopulation were then analyzed as independent Welch-corrected tests against the pooled control group to overcome an imbalance of the retrospective factorial design (i.e., CON: Ile n=10; Val n=5).
The primary goal of the retrospective analysis was to evaluate whether Ile/Ile and Val/_ participants had different cumulative responses to the exercise intervention. We evaluated this comprehensive response through the analysis of composite indexes. Z-scores were first calculated in Microsoft Excel 2016 by centering and scaling all delta adjusted values to the pooled control group. The main intervention direction of each variable (+1 or -1) was then employed to transform the z-score values before averaging them in composite response indexes. Genotype differences to the intervention were then analyzed by a Welch-corrected contrast, and its size effect was calculated as a Cohen's d. Genotype differences against the CON group were also analyzed by independent Welch tests to evaluate if the intervention was effective for each genotype.