Identification and Small Molecule Inhibition of an Activating Transcription Factor 4 (ATF4)-dependent Pathway to Age-related Skeletal Muscle Weakness and Atrophy

Background: Aging reduces skeletal muscle strength and mass. Results: The transcription factor ATF4 is required for age-related muscle weakness and atrophy, and the small molecules ursolic acid and tomatidine reduce ATF4 activity, weakness, and atrophy in aged skeletal muscle. Conclusion: ATF4 is an essential mediator of muscle aging. Significance: These results identify new strategies for reducing weakness and muscle loss during aging.

Skeletal muscle weakness and atrophy are among the most pervasive and disruptive effects of aging. In nearly all people, even elite athletes, a subtle loss of muscle strength begins between the ages of 30 and 40. Over the next two to three decades, strength continues to erode, whereas muscle mass typically declines to a lesser degree. As a result, reduced muscle quality (i.e. strength per unit muscle mass) is a hallmark of the aging process (1,2). By the age of 65, overt muscle loss (agerelated muscle atrophy or sarcopenia) is apparent in many individuals, and nearly all elderly persons report a gradual loss of strength and muscle over the course of their lives. The clinical consequences of age-related weakness and muscle loss are significant. Weakness limits activity, impairs quality of life, contributes to falls and fractures, and can create a vicious cycle of muscle disuse and further muscle loss and weakness. In its later stages, age-related muscle atrophy can lead to frailty, debilitation, and loss of independent living. All of these issues are becoming more prevalent as the elderly population increases. For example, in the United States, the number of persons over 65 years old is predicted to more than double between 2010 and 2040 (3).
Despite their broad impact, age-related muscle weakness and atrophy cannot be reliably prevented by physical therapy or current nutritional approaches, and a pharmacologic therapy does not exist. The development of effective interventions has been somewhat hindered by the fact that the molecular basis of age-related muscle weakness and atrophy is largely unknown.
The slow progression of age-related skeletal muscle atrophy represents a significant barrier to its experimental study and suggests that the condition may reflect subtle molecular changes that accumulate in skeletal muscle over many years. At the cellular level, age-related muscle atrophy shares some features with acute forms of muscle atrophy caused by fasting, muscle disuse, or systemic illness, which reduce muscle mass and strength over the course of days or weeks rather than years. For example, aging, fasting, muscle disuse, and systemic illness all cause a reduction in skeletal muscle fiber size and a loss of skeletal muscle protein. However, it is also clear that age-related muscle atrophy differs from acute muscle atrophy in some important ways. For example, at least some mediators of acute muscle atrophy (e.g. MAFbx/atrogin-1, MuRF1, and AMP kinase) also appear to protect muscle from effects of aging, and thus, chronic deficiencies of those proteins reduce muscle atrophy during acute stress conditions but accelerate the loss of muscle mass and/or quality during aging (4 -7). A specific protein target for reducing the loss of muscle quality, strength, and mass during aging has not yet been found.
In the current study, we investigated mechanisms of skeletal muscle weakness and atrophy during aging as well as potential interventions for these conditions. The potential interventions we investigated were two structurally dissimilar small molecules, ursolic acid and tomatidine. Ursolic acid is a naturally occurring pentacyclic triterpene acid present in several edible herbs and fruits, including apples (8). Tomatidine is a naturally occurring steroidal alkaloid derived from tomato plants and green tomatoes (9). We previously found that ursolic acid and tomatidine reduce acute skeletal muscle atrophy caused by fasting and muscle disuse in young adult mice (10 -12). In addition, we found that ursolic acid and tomatidine increase muscle strength and quality, and they stimulate muscle hypertrophy when they are administered to healthy young adult mice (10,11). The mechanisms of action of ursolic acid and tomatidine in skeletal muscle are not well understood; however, both compounds stimulate protein synthesis, protein accretion, and cellular hypertrophy in cultured skeletal myotubes, indicating a direct effect on skeletal muscle cells (10,11,13). The effects of ursolic acid and tomatidine on age-related muscle weakness and atrophy were not known. However, despite the existence of mechanistic differences between acute and age-related muscle atrophy, we hypothesized that acute and age-related muscle atrophy might share some common molecular mediators, and thus, age-related muscle weakness and/or atrophy might be reduced by ursolic acid and/or tomatidine.
Plasmids and Recombinant Adenoviruses-p-ATF4-FLAG encodes wild-type mouse activating transcription factor 4 (ATF4) 2 (NM_009716) with three copies of the FLAG epitope tag at the NH 2 terminus under the control of the CMV promoter. Recombinant adenoviruses expressing GFP alone, GFP plus FLAG-tagged ATF4, or GFP plus FLAG-tagged ATF4⌬bZIP were described previously (14). ATF4⌬bZIP (15) is a full-length ATF4 construct containing three copies of the FLAG epitope tag at the NH 2 terminus as well as 292 GLY-EAAA 298 substituted for 292 RYRQKKR 298 .
Mouse Protocols-Animals were housed in colony cages in a specific pathogen-free animal facility at 21°C with 12-h light/ 12-h dark cycles and were maintained on standard chow (Harlan Teklad formula 7013) unless otherwise indicated. In all studies, investigators who obtained the results were blinded to the intervention (dietary or genetic). Ursolic acid or tomatidine was added to standard chow formula 7013 at concentrations of 0.27 (w/w) and 0.05% (w/w), respectively, by Harlan Teklad, as described previously (10,11). 22-month-old male C57BL/6 mice were obtained from the National Institute on Aging and randomly assigned to standard chow diets lacking or containing ursolic acid or tomatidine. Forelimb grip strength was determined using a triangular pull bar attached to a grip strength meter (Columbus Instruments) as described previously (10). Ex vivo studies of skeletal muscle force generation were performed as described previously (11). For skeletal muscle transfection experiments, male C57BL/6 mice were obtained from the National Cancer Institute (NCI) at ages 6 -8 weeks old and used for experiments within 2 weeks of their arrival. Transfection of mouse skeletal muscle with plasmid DNA was performed as described previously (15). In the SUnSET assay (16), mice were administered an intraperitoneal injection of puromycin (0.04 mol/g of body weight, dissolved in 100 l of sterile saline), and then 30 min later, mice were euthanized, and skeletal muscles were harvested and frozen in liquid N 2 for further analysis. Muscle-specific Atf4 knock-out (ATF4 mKO) mice were generated and genotyped as described previously (14).
Male ATF4 mKO mice (i.e. ATF4(L/L);MCK-Cre(Tg/0) mice) were studied at 22 months of age and compared with male ATF4(L/L);MCK-Cre(0/0) littermates. All animal procedures were approved by the Institutional Animal Care and Use Committee of the University of Iowa.
Histology-Skeletal muscle H&E stain and fiber type-specific stain were performed as described previously (11). All histological sections were examined and photographed using an Olympus BX-61 automated upright microscope equipped with CellSens digital imaging software. Image analysis was performed using ImageJ software. Skeletal muscle fiber size was analyzed by measuring the lesser diameter (minimal Feret diameter) of muscle fibers as recommended elsewhere (17).
Quantification of Ursolic Acid and Tomatidine in Mouse Plasma-Mouse plasma was collected as described previously (10), and plasma levels of ursolic acid and tomatidine were determined with validated liquid chromatography/mass spectrometry (LC/MS) assays. In the ursolic acid assay, betulinic acid was used as the internal standard. Stock solutions of ursolic acid and betulinic acid were generated by dissolving the compounds in 100% methanol at a concentration of 200 g/ml, and working solutions were prepared by serial dilution of the stock solutions in 100% methanol. Standard curve samples were generated by adding ursolic acid to 100 l of blank mouse plasma at concentrations ranging from 1 to 1000 ng/ml. Control samples contained 0, 10, 100, or 700 ng/ml ursolic acid in 100 l of blank mouse plasma. Betulinic acid (5 l of a 10 g/ml stock) was added to the standard curve samples, control samples, and experimental samples (100 l of plasma from mice treated with control diet, ursolic acid diet, or tomatidine diet). Ursolic acid and betulinic acid were then extracted from plasma samples as follows. 600 l acetonitrile was added to each sample, and the samples were vortexed for 1 min and allowed to sit for 5 min. 400 l of ethyl acetate was added to each sample, and the samples were vortexed, centrifuged for 3 min at 14,000 rpm, and then transferred to a Phree Phospholipid Removal plate (product number 8E-S133-TGB, Phenomenex, Torrance CA). A vacuum was applied, and the eluent was collected, transferred to a 13 ϫ 100-mm glass tube, and dried under nitrogen. The dried samples were reconstituted in 100 l of mobile phase (5 mM ammonium acetate in ultrapure water (15%) and methanol (85%)) and then transferred to HPLC injection vials. LC/MS was performed with a Waters Acquity UPLC system equipped with binary solvent manager, sample manager, and column manager operating under Empower software (Waters Corp., Milford, MA) and a Shimadzu 2010A LC/MS platform in atmospheric pressure chemical ionization negative mode operating under LC/MS Solutions software (version 3.41.324; Shimadzu, Columbia, MD). The analytical column was an Acquity UPLC high strength silica column (C 18 , 1.8 m, 2.1 ϫ 100 mm; Waters Corp.) with a Phenomenex C 18 ULTRA UHPLC SecurityGuard column. Separation conditions were as follows: sample temperature, 7°C; column temperature, 40°C; sample was injected in partial loop with overfill injection mode; injection volume, 15 l. The analysis was isocratic at 0.4 ml/min flow. Solvent A (15%) was 5 mM ammonium acetate; solvent B (85%) was methanol. The total run time for an LC/MS analysis was 7 min. The scan interval was 0.3 s with the following parameters: microscan, 0.1 atomic mass unit; atmospheric pressure chemical ionization temperature, 450°C; curved desolvation line temperature, 200°C; heat block temperature, 200°C; nitrogen flow rate, 2.5 liters/min; detector voltage, 1.6 kV. The monitored mass-to-charge ratio for both ursolic acid and betulinic acid was 455.45. The retention times for ursolic acid and betulinic acid were 3.85 and 3.25 min, respectively. The assay was linear over the concentration range of 1-1000 ng/ml, and the coefficient of variation was less than 15% at all control concentrations. In the tomatidine assay, solanidine was utilized as the internal standard. Stock solutions of tomatidine and solanidine were generated by dissolving the compounds in 100% methanol, and working solutions were prepared by serial dilution of the stock solutions in 100% methanol. Standard curve samples were generated by adding tomatidine to 100 l of blank mouse plasma at concentrations ranging from 4 to 1000 ng/ml. Control samples contained 0, 10, 100, or 700 ng/ml tomatidine in 100 l of blank mouse plasma. Solanidine (5 l of a 10 g/ml stock) was added to the standard curve samples, control samples, and experimental samples (100 l of plasma from mice treated with control diet, ursolic acid diet, or tomatidine diet). Tomatidine and solanidine were then extracted from plasma samples as follows. Samples were applied to Isolute SLEϩ cartridges (6 ml; Biotage, AB, Uppsala, Sweden) in a Cerex solid phase extraction processor (Varian, Palo Alto, CA) and allowed to adsorb onto the packing for 10 min. Samples were eluted with 5 ml of hexanes:dichloromethane (2:1 volume ratio) under gravity flow for ϳ15 min. Eluates were collected in 13 ϫ 100-mm glass tubes and evaporated under flowing nitrogen at 25°C. Dried samples were reconstituted in 200 l of water containing 40% methanol and 0.05% acetic acid and then transferred to HPLC injection vials. LC/MS was performed with a Shimadzu LC/MS-2010A mass spectrometer operated using electrospray interface in positive ion mode controlled using LC/MS Solutions software (version 3.41.324). The block and curved desolvation line temperature were both set at 250°C. Data were collected in the selected ion monitoring mode at 416.45 (tomatidine) and 398.45 atomic mass units (solanidine). The analytical column was a Phenomenex Kinetex C 18 (100 ϫ 2.1 mm, 2.6 m) preceded by a Phenomenex C 18 ULTRA UHPLC SecurityGuard column. Separation conditions were as follows: sample temperature, 24°C (ϩ3°C); column temperature, 40°C; sample injection volume, 10 l. A gradient of 0.25 ml/min was used for the analysis. Solvent A was water with 0.05% acetic acid and 5 mM ammonium acetate, and solvent B was methanol (75%) mixed with acetonitrile (25%). The gradient started at 40% B and increased to 90% B at 5 min, was held for 3 min, and then was changed back to baseline at 9 min. The total run time for LC/MS analysis was 12 min. The retention times for tomatidine and solanidine were 5.4 and 4.6 min, respectively. The assay was linear from 4 to 1000 ng/ml, and the coefficient of variation for all control samples was less than 15%.

Mechanism and Small Molecule Inhibition of Sarcopenia
mRNA Expression Analyses-For microarray analyses, skeletal muscle RNA was extracted with TRIzol (Invitrogen) and purified with an RNeasy kit and RNase-free DNase set (Qiagen). RNA was processed and hybridized to Mouse Ref-8 version 2.0 BeadChip arrays (Illumina) by the Southern California Genotyping Consortium (University of California, Los Angeles) as described previously (18). Following hybridization, arrays were washed, blocked, stained, and dried (Little Dipper processor). Arrays were scanned with an iScan reader, and data were extracted and analyzed with BeadStudio software (Illumina). Quantitative real time RT-PCR (qPCR) was performed using TaqMan Gene Expression Assays (Applied Biosystems) and methods described previously (14).
Myotube Experiments-Mouse C2C12 myoblasts were obtained from ATCC (CRL-1772) and maintained at 37°C and 5% CO 2 in Dulbecco's modified Eagle's medium (DMEM) (ATCC catalog number  containing antibiotics (100 units/ml penicillin and 100 g/ml streptomycin sulfate) and 10% (v/v) fetal bovine serum (FBS). Myoblasts were set up for experiments on day 0 in 6-well plates at a density of 2.5 ϫ 10 5 cells/well. On day 2, differentiation was induced by replacing 10% FBS with 2% horse serum. On day 7, cells were rinsed once with PBS, and then 1 ml of DMEM containing adenovirus (multiplicity of infection, 250) was added to each well. Two hours later, 1 ml of DMEM containing 1% horse serum plus antibiotics was added to each well. On day 8, cells were rinsed twice with PBS, and then 2 ml of DMEM containing 2% horse serum and antibiotics was added to each well. Infection efficiency was Ͼ90%. On day 9 (48 h postinfection), myotubes were harvested for analysis of Eif4ebp1 and 4E-BP1 expression or used for analysis of global protein synthesis. For analysis of protein synthesis, [ 3 H]leucine incorporation was determined as described previously (14).
Statistical Analysis-Results of experiments are shown as mean Ϯ S.E. Statistical significance was determined using t tests, one-way ANOVA with Dunnett's multiple comparison test, or false discovery rate (FDR) analysis as noted in the figure legends.

Ursolic Acid and Tomatidine Reduce Age-related Skeletal
Muscle Weakness and Atrophy-Mice have been reported to develop age-related muscle atrophy by the age of 22 months (2,5,19). To confirm this finding and establish an experimental system for our studies, we compared 22-month-old mice with 6-month-old mice, which are considered fully grown young adults. The cohorts of 6-and 22-month-old mice were matched for total body weight (35.4 Ϯ 1.0 and 35.7 Ϯ 1.8 g, respectively; p ϭ 0.86). However, the combined weight of the largest muscle groups in the hind limb and forelimb (quadriceps femoris and triceps brachii, respectively) was significantly reduced in the 22-month-old mice (by 18 Ϯ 3%; Fig. 1A), indicating age-related muscle loss. We performed histological analyses of the quadriceps, a functionally important muscle group that is known to be severely affected by age-related muscle atrophy in humans (20). In both 6-and 22-month old mice, the quadriceps was primarily composed of type IIb fibers with the remainder composed of type IIx fibers (82 Ϯ 2% type IIb fibers and 18 Ϯ 2% type IIx fibers in 6-month-old mice versus 84 Ϯ 2% type IIb fibers and 16 Ϯ 2% type IIx fibers in 22-month-old mice). The relative amounts of IIb and IIx fibers were not affected by age (p ϭ 0.49), and the size of IIx fibers was not affected by age (Fig.  1C). Importantly, however, the size of IIb fibers was significantly reduced in 22-month-old mice (Fig. 1, B-D). Furthermore, 22-month-old mice exhibited a significant reduction in grip strength (decreased by 12 Ϯ 4%; Fig. 1E) and a significant reduction in specific force (i.e. muscle quality; decreased by 36 Ϯ 11%; Fig. 1F). These age-related reductions in muscle mass, type IIb fiber size, strength, and muscle quality are indicative of age-related muscle atrophy as it has been described in both mice and humans (1,2,20).
To test the hypothesis that ursolic acid and tomatidine might reduce age-related skeletal muscle weakness and atrophy, we provided weight-matched cohorts of 22-month-old mice ad libitum access to diets lacking or containing either 0.27% ursolic acid or 0.05% tomatidine for 2 months ( Fig. 2A). We previously found that these doses of ursolic acid and tomatidine stimulate skeletal muscle hypertrophy and increase strength and specific force in young adult mice (10,11,21). The diets containing ursolic acid and tomatidine were well tolerated by 22-month-old mice. Altogether in these studies, we observed 32 mice on each diet (control, ursolic acid, or tomatidine), and by the end of the 2-month treatment period, four control mice had died of natural causes, one ursolic acid-treated mouse died, and no tomatidine-treated mice died. As expected, the diet containing ursolic acid generated measurable plasma levels of ursolic acid, and the diet containing tomatidine generated measurable plasma levels of tomatidine ( Fig. 2A).
We found that ursolic acid and tomatidine significantly increased skeletal muscle weight by 9 Ϯ 2 and 10 Ϯ 2%, respectively (Fig. 2B). In addition, both compounds significantly increased the size of type IIb muscle fibers in the quadriceps without increasing the size of type IIx fibers ( Fig.  2, C-E). Ursolic acid and tomatidine did not alter the relative amount of IIb and IIx fibers in the quadriceps (there were 97 Ϯ 2, 96 Ϯ 2, and 96 Ϯ 1% IIb fibers in control, ursolic acid-treated, and tomatidine-treated quadriceps, respectively; all p Ͼ 0.05). In addition to reducing atrophy of type IIb fibers, ursolic acid and tomatidine significantly increased grip strength (by 12 Ϯ 3 and 10 Ϯ 2%, respectively; Fig. 2F) and significantly increased specific force (by 30 Ϯ 8 and 33 Ϯ 11%, respectively; Fig. 2G). In contrast to their effects on skeletal muscle, ursolic acid and tomatidine did not significantly alter total body weight or the weights of heart, liver, or fat pads (Fig. 3, A-F). Taken together, these data indicate that ursolic acid and tomatidine reduced age-related muscle weakness and atrophy.
In Aged Skeletal Muscle, Ursolic Acid and Tomatidine Generate Similar mRNA Expression Signatures, Which Are Composed of Hundreds of Small Positive and Negative Changes in mRNA Levels-Age-related muscle atrophy is a complex and slowly progressive process that remains poorly understood at the molecular level. Because ursolic acid and tomatidine reduced age-related muscle weakness and atrophy, we reasoned that they might provide insight into the molecular pathogenesis of age-related muscle loss. To test this hypothesis, we performed additional analyses of quadriceps muscles of 22-month-old mice that had consumed diets lacking or containing ursolic acid or tomatidine for 2 months (as in Figs. 2 and 3).
In young adult mice, ursolic acid and tomatidine promote muscle hypertrophy by increasing activity of the protein kinase mTORC1 (10,11,21). We therefore tested the possibility that one or both of the compounds might increase phosphorylation (activity) of S6 kinase, a key downstream target of mTORC1, in aged skeletal muscle. However, we found that ursolic acid and tomatidine did not increase the steady-state level of S6 kinase phosphorylation in aged muscle (Fig. 4A). Similarly, we did not detect any change in phosphorylation (activity) of Akt (Fig. 4B), an anabolic protein kinase that mediates the effect of ursolic acid on mTORC1 (10,21). Although we cannot rule out the possibility that ursolic acid and/or tomatidine may have stimulated Akt and/or mTORC1 at some point during the 2-month treatment period, it was clear that Akt/mTORC1 signaling was not significantly increased at the end of the treatment period. The absence of sustained mTORC1 activation in ursolic acidand tomatidine-treated muscles is perhaps not surprising because sustained activation of mTORC1 does not inhibit agerelated muscle atrophy but rather appears to promote it (22). Furthermore, a small molecule inhibitor of mTORC1, rapamycin, does not accentuate or reduce age-related muscle atrophy (23). These results led us to investigate other potential mechanisms.
We originally identified ursolic acid and tomatidine as small molecules whose collective effects on gene expression in human cell lines are roughly opposite to the changes in skeletal muscle gene expression that occur during muscle atrophy (10,11). This suggested the hypothesis that ursolic acid and tomatidine might reverse at least some changes in skeletal muscle gene expression that promote age-related muscle atrophy. To begin to test this idea, we used genome-wide mRNA expression arrays to assess mRNA expression in quadriceps muscles of  22-month-old mice that had consumed diets lacking or containing ursolic acid or tomatidine for 2 months. Using p Յ 0.025 by two-tailed unpaired t test as an arbitrary cutoff for statistical significance, we found that ursolic acid affected the signals from 1289 mRNA probes (about 5% of the Ͼ25,000 probes measured on the arrays; supplemental Table 1), and tomatidine affected signals from 1031 mRNA probes (supplemental Table 2). The magnitude of these changes in mRNA levels was small with few

Control
Tomatidine Ursolic Acid changes greater than 2-fold. Remarkably, there was a strong correlation between the effects of ursolic acid and tomatidine: in most cases, if one compound was judged to have a statistically significant effect on a particular mRNA, then the other compound affected that mRNA in the same general direction (up or down) even if the effect was not judged to be statistically significant (Fig. 5, A and B). When the data were subjected to an FDR analysis, mRNA probe signals that met a cutoff of FDR Յ 0.1 (supplemental Table 3) constituted a subset of the mRNA probe signals that met the less stringent cutoff of p Յ 0.025, and there remained a strong correlation between the effects of ursolic acid and tomatidine (R 2 ϭ 0.92; p Ͻ 0.0001). Thus, the antiatrophic effects of ursolic acid and tomatidine in aged skeletal muscle were associated with numerous small positive and negative changes in skeletal muscle mRNA expression, and there was a surprisingly high concordance between the effects of the two compounds.
Gene Set Enrichment Analysis Identifies ATF4 as a Transcription Factor Whose Activity Is Reduced by Ursolic Acid and Tomatidine in Aged Skeletal Muscle-Because ursolic acid and tomatidine had similar effects on skeletal muscle mRNA levels, we used gene set enrichment analysis (24) to search for basic cellular processes that might be stimulated or inhibited by ursolic acid and tomatidine in aged skeletal muscle. We found that both compounds induced seven gene sets that promote either muscle contraction or mitochondrial bioenergetics (Fig. 6A). Consistent with the gene set enrichment analysis results, mRNA expression arrays and qPCR analysis showed that ursolic acid and tomatidine increased mRNAs encoding proteins involved in muscle growth and/or mitochondrial bioenergetics (e.g. transferrin receptor and mitochondrial creatine kinase, encoded by Tfrc and Ckmt2, respectively) as well as mRNAs encoding the contraction-induced transcripts IL-15 and Fndc5 (Figs. 5, A and B, and 6C).
We also used gene set enrichment analysis to search for gene sets that were repressed by ursolic acid and tomatidine. Altogether, we found that both compounds repressed 25 gene sets that represent a diverse range of cellular functions, including "activation of genes by ATF4" (Fig. 6B). ATF4 is a bZIP transcription factor subunit that mediates a variety of cellular stress responses (25,26). The potential role of ATF4 in age-related muscle atrophy had not been investigated.   Fig. 2 that were treated for 2 months with control diet, ursolic acid diet, or tomatidine diet. Quadriceps mRNA was analyzed with Mouse Ref-8 version 2.0 BeadChip arrays, which contain Ͼ25,000 probes to Ͼ19,000 mRNA transcripts. n ϭ 5-6 arrays per condition. Effects of ursolic acid and tomatidine on individual probe signals were determined by normalizing log 2 signals from ursolic acid-treated mice and tomatidine-treated mice, respectively, to log 2 signals from control mice. Statistical significance was arbitrarily defined as p Յ 0.025 by a two-tailed unpaired t test. A, scatter plot showing the effects of ursolic acid and tomatidine on the 1289 mRNA probes whose signals were significantly altered by ursolic acid. Some mRNAs of interest are highlighted. B, scatter plot showing the effects of tomatidine and ursolic acid on the 1031 mRNA probes whose signals were significantly altered by tomatidine. Some mRNAs of interest are highlighted.
All of the gene sets influenced by ursolic acid and tomatidine (Fig. 6, A and B) represent potentially important processes in age-related muscle atrophy. However, we chose to focus on the activation of genes by ATF4 set for three reasons. First, we wished to identify a cellular process that actively drives agerelated muscle atrophy, and we reasoned that such a process would most likely be found within the gene sets that were repressed by ursolic acid and tomatidine. Second, ATF4 is known to be required for some, but not all, forms of acute muscle atrophy (14, 15, 18, 27). Although endogenous ATF4 protein cannot be reliably detected in skeletal muscle (presumably due to its low abundance, very short half-life, and the absence of high quality antibodies), targeted knock-out of the Atf4 gene in skeletal muscle fibers partially prevents acute muscle atrophy during fasting and limb immobilization (14,15,18). Conversely, forced expression of ATF4 is sufficient to induce atrophy of skeletal muscle fibers in vivo and skeletal myotubes in vitro (14,15).
Third, a variety of impairments in skeletal muscle protein synthesis have been implicated as important contributing factors in age-related muscle atrophy (2, 28, 29), and we hypothesized that ATF4-mediated gene expression might have a capac-  ity to reduce protein synthesis in skeletal muscle. This hypothesis was based on the finding that ursolic acid and tomatidine repressed several mRNAs that encode proteins that can potentially influence global protein synthesis, including Eif4ebp1, Eif2s2, and Eif3c (which encode proteins that regulate translation initiation); Slc6a9 and Slc38a2 (which encode amino acid transporters); and Lars, Nars, Mars, and Sars (which encode aminoacyl-tRNA synthetases) (Fig. 5, A and B). All of these transcripts arise from established ATF4 target genes (30), and Eif4ebp1 in particular encodes a well established inhibitor of global protein synthesis, 4E-BP1. qPCR analysis confirmed that ursolic acid and tomatidine reduced Eif4ebp1 mRNA in aged skeletal muscle (Fig. 6C), and importantly, the reduction in Eif4ebp1 mRNA was accompanied by a reduction in 4E-BP1 protein (Fig. 6D). Ursolic acid and tomatidine did not alter the level of mTORC1-mediated 4E-BP1 phosphorylation (inactivation), consistent with the finding that the compounds did not alter mTORC1-mediated S6 kinase phosphorylation. These data suggested that ursolic acid and tomatidine reduced the level of active (unphosphorylated) 4E-BP1 and perhaps other inhibitors of protein synthesis. Together, these considerations led us to focus on ATF4 as a potential negative regulator of skeletal muscle protein synthesis and a potentially important driver of age-related skeletal muscle atrophy. ATF4 Reduces Skeletal Muscle Protein Synthesis-To begin to investigate the potential effect of ATF4 on skeletal muscle pro-tein synthesis, we overexpressed ATF4 in skeletal muscle fibers of young adult (2-month-old) mice (Fig. 7A). As expected, ATF4 increased Eif4ebp1 mRNA (Fig. 7B) and 4E-BP1 protein (Fig. 7A). To determine the effect of ATF4 on muscle protein synthesis, we used the SUnSET method, which assesses in vivo incorporation of low dose puromycin into elongating polypeptide chains (16). We found that ATF4 significantly reduced the incorporation of puromycin into total skeletal muscle protein (Fig. 7C), indicating a reduction in global protein synthesis. Thus, increased ATF4 expression is sufficient to reduce protein synthesis in skeletal muscle. As a complementary system for investigating the effect of increased ATF4 expression, we used a recombinant adenovirus to overexpress ATF4 in cultured C2C12 myotubes, an in vitro model of skeletal muscle (Fig. 7D). In this system, we compared wild-type ATF4 with a transcriptionally inactive ATF4 construct that contains seven point mutations in the bZIP domain (ATF4⌬bZIP; Fig. 7D). Similar to its effects in skeletal muscle in vivo, ATF4 increased Eif4ebp1 mRNA and 4E-BP1 protein in cultured skeletal myotubes (Fig. 7D). In contrast, the transcriptionally inactive ATF4⌬bZIP construct did not increase either Eif4ebp1 mRNA or 4E-BP1 protein (Fig. 7D). We then assessed global protein synthesis with a traditional metabolic labeling method and found that ATF4 significantly decreased protein synthesis relative to ATF4⌬bZIP (Fig. 7E). These data suggest that ATF4-dependent gene expression leads to a reduction in skeletal muscle protein synthesis.

Targeted Reduction in Skeletal Muscle ATF4 Expression Increases Protein Synthesis in Aged Skeletal Muscle and Reduces
Age-related Muscle Weakness and Atrophy-To determine whether a reduction in ATF4 expression might increase protein synthesis in aged skeletal muscle, we studied 22-month-old ATF4 mKO mice. These mice carry two copies of a floxed Atf4 allele as well as the muscle creatine kinase (MCK)-Cre transgene, which excises the floxed Atf4 alleles in skeletal muscle fibers and heart (14,18,31,32). We compared the ATF4 mKO mice with littermate control mice, which also carry two copies of the floxed Atf4 allele but lack the MCK-Cre transgene. Importantly, ATF4 mKO mice undergo normal muscle development and are phenotypically normal under basal conditions as young adults (14,18). Fig. 8A shows the 20 mRNA transcripts whose levels were most reduced in 22-month old ATF4 mKO quadriceps relative to quadriceps from 22-month old littermate controls (the complete list is shown in supplemental Tables 4 and 5). As expected, the largest reduction was in Atf4 mRNA (Fig. 8A), and this was accompanied by modest reductions in multiple mRNAs that encode regulators of protein synthesis, including Eif4ebp1 (Fig. 8A). The reduction in Eif4ebp1 mRNA was associated with a reduction in 4E-BP1 protein (Fig. 8B). Aged ATF4 mKO muscles did not contain a reduced level of Gadd45a mRNA, which mediates acute atrophic effects of ATF4 during fasting and limb immobilization (14) but is also known to be controlled by an ATF4-independent mechanism (27). We used the SUnSET method to assess quadriceps protein synthesis in 22-month-old control and ATF4 mKO mice. Relative to control muscles, ATF4 mKO muscles had a significantly higher level of puromycin incorporation into total muscle protein (Fig. 8C). Thus, a targeted reduction in skeletal muscle ATF4 expression increases global protein synthesis in aged skeletal muscle. Effects of ATF4 mKO on individual probe signals were determined by normalizing log 2 signals from ATF4 mKO mice to log 2 signals from control mice. Statistical significance was arbitrarily defined as p Յ 0.025. The data show the 20 mRNAs whose levels were most reduced in ATF4 mKO muscle relative to control muscle; some mRNAs have two log 2 changes because they were represented by two probes on the array. Asterisks indicate probes that were also considered significantly reduced by FDR analysis (i.e. FDR Յ 0.1). B, quadriceps protein extracts were subjected to immunoblot analysis with anti-4E-BP1 and anti-phospho-4E-BP1. Left, representative immunoblots and Ponceau S stains. Right, quantification of total 4E-BP1. Each data point represents one mouse, and horizontal bars denote means Ϯ S.E. C, skeletal muscle protein synthesis was assessed by administering intraperitoneal injections of puromycin, harvesting the quadriceps muscles 30 min later, and then subjecting quadriceps protein extracts to immunoblot analysis with an anti-puromycin antibody. Left, representative immunoblot and Ponceau S stain. Right, quantification of incorporated puromycin. Each data point represents one mouse, and horizontal bars denote means Ϯ S.E. In B and C, p values were determined with t tests. FilGAP, filamin A-binding GTPase-activating protein.
We performed histological analyses of the quadriceps muscles of 22-month-old control and ATF4 mKO mice and found that ATF4 mKO muscles had significantly larger type IIb fibers (Fig. 9, B-D). The relative amounts of IIb and IIx fibers were unchanged (87 Ϯ 3 and 92 Ϯ 2% type IIb fibers and 13 Ϯ 3 and 8 Ϯ 2% type IIx fibers in control and ATF4 mKO mice, respectively; p ϭ 0.13). In addition to increasing muscle mass and IIb muscle fiber size, muscle-specific Atf4 gene deletion significantly increased in vivo grip strength (by 11 Ϯ 2%; Fig. 9E) and ex vivo specific force (by 24 Ϯ 6%; Fig. 9F). Taken together, these data indicate that a targeted reduction of skeletal muscle ATF4 activity reduces age-related muscle weakness and atrophy, similar to ursolic acid and tomatidine.

ATF4 Is Required for a Small Subset of the Gene Expression Changes That Are Generated by Ursolic Acid and Tomatidine in
Aged Skeletal Muscle-Because Atf4 gene deletion, ursolic acid, and tomatidine had similar morphological and functional effects in aged skeletal muscle, we asked whether Atf4 gene deletion might generate the same changes in skeletal muscle gene expression that were generated by ursolic acid and tomatidine. Fig. 10, A and B, show the effects of Atf4 gene deletion on mRNAs whose levels were significantly altered by ursolic acid or tomatidine in aged skeletal muscle. As expected, Atf4 gene deletion reduced levels of multiple ATF4-dependent mRNAs (e.g. Eif4ebp1, Lars, Nars, Mars, Sars, Slc6a9, Slc38a2, Eif2s2, and Eif3c), similar to ursolic acid or tomatidine (Fig. 10, A and B, and supplemental Tables 6 and 7). However, in contrast to ursolic acid and tomatidine, Atf4 gene deletion did not increase mRNAs involved in muscle contraction and mitochondrial bioenergetics, and the overall correlation between the effects of Atf4 gene deletion and the effects of the compounds was poor (Fig. 10, A and B). Thus, ATF4 mediates a small subset of the effects that ursolic acid and tomatidine exert on skeletal muscle gene expression, and inhibiting this small subset of ATF4-dependent effects appears to be sufficient to reduce age-related skeletal muscle weakness and atrophy.

Discussion
One of the goals of the current study was to identify a small molecule that reduces age-related skeletal muscle weakness and/or atrophy. To this end, we tested two small molecules,  ursolic acid and tomatidine, in elderly mice. Both ursolic acid and tomatidine were known to reduce acute muscle atrophy in young adult mice during fasting and muscle disuse, but their effects in aging were unknown. The current results indicate that 2 months of dietary supplementation with either ursolic acid or tomatidine significantly reduces age-related deficits in skeletal muscle mass, strength, and quality. These results identify two novel small molecule inhibitors of age-related skeletal muscle weakness and atrophy. To better understand mechanisms of age-related muscle atrophy and weakness, we performed an unbiased analysis of the effects of ursolic acid and tomatidine on gene expression in aged skeletal muscle. We found that ursolic acid and tomatidine generate hundreds of small positive and negative changes in mRNA levels in aged skeletal muscle. Moreover and despite their structural differences, ursolic acid and tomatidine generate remarkably similar mRNA expression signatures, which, based on pathway analysis, appear to impact a broad range of cellular processes. Although these data do not speak to the direct or acute effects of ursolic acid and tomatidine in aged skeletal muscle, they do suggest that the end result of a 2-month treatment with either compound is a subtle reprogramming of skeletal muscle gene expression. Furthermore, these subtle changes in gene expression are accompanied by a reduction in age-related muscle weakness and atrophy and may reflect a moderation of at least some of the molecular effects of skeletal muscle aging.
All of the genes that are induced or repressed by ursolic acid and tomatidine in aged skeletal muscle could potentially contribute to age-related muscle weakness and atrophy. In the current study, we focused our attention on a small subset of genes that are repressed by ursolic acid and tomatidine. This particular subset of ursolic acid-and tomatidine-sensitive genes is known to be directly activated by ATF4, and our data indicate that a targeted reduction in skeletal muscle ATF4 expression not only represses this subset of genes but also reduces agerelated muscle weakness and atrophy. To our knowledge, ATF4 is the first example of a skeletal muscle protein in mammals that is required for the decline of skeletal muscle strength, quality, and mass during aging. The discovery of ATF4 as a key mediator of age-related muscle weakness and atrophy provides an important foundation for future investigations into the molecular basis of skeletal muscle aging.
As a transcription factor, ATF4 could potentially influence many cellular processes in aged skeletal muscle. In the current study, we investigated the effects of ATF4 on one cellular process, skeletal muscle protein synthesis, which is known to be impaired in association with age-related muscle weakness and atrophy. Our data indicate that a targeted reduction in ATF4 expression is sufficient to increase protein synthesis in aged skeletal muscle, and conversely, forced expression of ATF4 is sufficient to decrease protein synthesis in young adult skeletal muscle. These results, coupled with the effects of ATF4 on muscle strength, quality, and mass, suggest that ATF4 may promote age-related muscle weakness and atrophy at least in part by reducing skeletal muscle protein synthesis. The mechanism by which ATF4 reduces protein synthesis is likely complex because ATF4 regulates multiple genes that could influence protein synthesis at multiple levels, including amino acid transport, aminoacyl-tRNA availability, and mRNA translation. One potentially relevant ATF4 target gene is Eif4ebp1, which encodes 4E-BP1, a well established repressor of cap-dependent mRNA translation.
ATF4 also promotes acute muscle atrophy during fasting (14,15), and many of the mRNAs that are positively regulated by ATF4 in aged skeletal muscle are also positively regulated by ATF4 in young adult skeletal muscle during fasting. Examples include Eif4ebp1, Eif2s2, Lars, Nars, Mars, Iars, Aars, and Gars (14,15). However, there also appear to be some important differences in the transcripts that are regulated by ATF4 during fasting and aging. The most notable example is Gadd45a, which is induced by ATF4 during fasting and contributes to muscle atrophy during fasting (14,15). Although aging is known to increase skeletal muscle Gadd45a expression (33)(34)(35)(36)(37), the current data indicate that ATF4 is not required for Gadd45a expression in aged skeletal muscle. This finding suggests that lowering Gadd45a mRNA is not absolutely required to reduce age-related muscle weakness and atrophy. It also sug- gests that Gadd45a expression in aged skeletal muscle may be mediated by HDAC4, which is known to induce Gadd45a mRNA in an ATF4-independent manner (27).
By identifying ATF4 as an indirect or direct target of ursolic acid and tomatidine, the current study identifies two small molecules that blunt ATF4-dependent effects in aged skeletal muscle. Understanding how ursolic acid and tomatidine reduce skeletal muscle ATF4 activity is an important and challenging area for future investigation. The current data do not indicate whether ursolic acid and tomatidine reduce ATF4 activity by the same or different mechanisms, and there are many potential mechanisms given the multitude of pathways, often interconnected, that regulate ATF4 activity. Importantly, a reduction in ATF4 activity cannot explain all of the effects of ursolic acid and tomatidine in skeletal muscle. For example, this mechanism cannot explain how ursolic acid and tomatidine stimulate muscle hypertrophy because a targeted reduction in skeletal muscle ATF4 expression does not induce muscle hypertrophy (14,15). In addition, ATF4 accounts for only a small portion of the mRNAs that are regulated by ursolic acid and tomatidine in aged skeletal muscle, and it seems improbable that all of the ATF4-independent effects of ursolic acid and tomatidine are functionally irrelevant. Rather, we speculate that age-related muscle atrophy proceeds via a highly complex molecular signaling network with several critical and vulnerable nodes. One of these nodes is represented by ATF4, but others likely exist and may be affected by ursolic acid and tomatidine in ways that we do not yet understand.
The progrowth effects of ursolic acid and tomatidine in young adult skeletal muscle (i.e. hypertrophy and recovery from atrophy) are associated with activation of mTORC1, a well established mediator of muscle growth (10,11). 3 In addition, the mTORC1 inhibitor rapamycin prevents ursolic acid-and tomatidine-mediated hypertrophy in cultured myotubes (11). 3 Interestingly, however, the current data indicate that ursolic acid and tomatidine reduce age-related muscle atrophy and weakness without eliciting a sustained increase in mTORC1 activity. Because of our study design, we cannot rule out the possibility that ursolic acid and tomatidine may transiently stimulate mTORC1 at an early treatment time point in aged skeletal muscle. However, given the potentially deleterious effect of sustained mTORC1 activity in skeletal muscle aging (22) and the finding that ursolic acid and tomatidine improve muscle strength, quality, and mass in aged muscle, it is perhaps not surprising that ursolic acid and tomatidine do not generate a sustained increase in mTORC1 activity in aged muscle. Taken together, the current data suggest that, whereas ursolic acid and tomatidine promote muscle growth via an mTORC1-dependent mechanism, they reduce age-related muscle weakness and atrophy via a mechanism that is at least somewhat less dependent on mTORC1 and more dependent on inhibition of atrophy mediators, such as ATF4.
In light of the current results, ursolic acid and tomatidine represent potential agents and/or lead compounds for medical treatment of age-related muscle weakness and atrophy. In addition, because ursolic acid and tomatidine naturally occur in food, they could potentially comprise or contribute to nutritional products aimed at preserving strength and muscle mass during aging. If ursolic acid-and tomatidine-based approaches are found to be safe and effective in humans, they could possibly be used alone, together, or in combination with physical therapy and other nutritional and pharmaceutical approaches. Similar to other complex chronic conditions such as type 2 diabetes, dyslipidemia, and hypertension, age-related muscle weakness and atrophy may ultimately require a repertoire of modalities that can be used alone or in combination depending on the clinical circumstances. In summary, the current study provides new insight into the molecular pathogenesis of agerelated skeletal muscle weakness and atrophy and elucidates new potential approaches for preventing and treating the loss of strength and muscle mass during aging.