Stress-induced Skeletal Muscle Gadd45a Expression Reprograms Myonuclei and Causes Muscle Atrophy*

Background: In skeletal muscle, diverse stresses induce the transcription factor ATF4, which promotes muscle atrophy by an unknown mechanism. Results: ATF4 causes muscle atrophy by inducing Gadd45a, which reprograms myonuclear gene expression to repress anti-atrophy mechanisms and induce pro-atrophy mechanisms. Conclusion: Gadd45a is a critical stress-induced mediator of muscle atrophy. Significance: The ATF4/Gadd45a pathway is a potential therapeutic target in atrophic muscle. Diverse stresses including starvation and muscle disuse cause skeletal muscle atrophy. However, the molecular mechanisms of muscle atrophy are complex and not well understood. Here, we demonstrate that growth arrest and DNA damage-inducible 45a protein (Gadd45a) is a critical mediator of muscle atrophy. We identified Gadd45a through an unbiased search for potential downstream mediators of the stress-inducible, pro-atrophy transcription factor ATF4. We show that Gadd45a is required for skeletal muscle atrophy induced by three distinct skeletal muscle stresses: fasting, muscle immobilization, and muscle denervation. Conversely, forced expression of Gadd45a in muscle or cultured myotubes induces atrophy in the absence of upstream stress. We show that muscle-specific ATF4 knock-out mice have a reduced capacity to induce Gadd45a mRNA in response to stress, and as a result, they undergo less atrophy in response to fasting or muscle immobilization. Interestingly, Gadd45a is a myonuclear protein that induces myonuclear remodeling and a comprehensive program for muscle atrophy. Gadd45a represses genes involved in anabolic signaling and energy production, and it induces pro-atrophy genes. As a result, Gadd45a reduces multiple barriers to muscle atrophy (including PGC-1α, Akt activity, and protein synthesis) and stimulates pro-atrophy mechanisms (including autophagy and caspase-mediated proteolysis). These results elucidate a critical stress-induced pathway that reprograms muscle gene expression to cause atrophy.

A variety of stresses, including starvation, muscle disuse, systemic illness, and aging cause skeletal muscle atrophy, which is often debilitating. However, despite its broad impact, muscle atrophy lacks an effective medical therapy and its pathogenesis remains incompletely understood. Like many complex diseases, muscle atrophy is associated with widespread positive and negative changes in gene expression (1)(2)(3)(4)(5)(6)(7)(8). Some gene expression changes in atrophying muscle are known to promote atrophy, including induction of genes that promote proteolysis (9 -17) and repression of the gene encoding PGC-1␣, a transcriptional coactivator that promotes mitochondrial biogenesis and energy production (18,19). However, most atrophy-associated gene expression changes are unstudied, and it remains unknown if these changes contribute to muscle atrophy, and if so, to what extent.
To better understand the mechanisms of muscle atrophy, we recently investigated the role of activating transcription factor 4 (ATF4) 2 (also called CREB2), a basic leucine zipper (bZIP) transcription factor that mediates a variety of cellular stress responses (20). Oligonucleotide microarrays showed that starvation, denervation, diabetes, cancer, and renal failure increase ATF4 mRNA in skeletal muscle (1,2). We found that ATF4 overexpression in mouse skeletal muscle is sufficient to induce muscle fiber atrophy (21). Conversely, an RNA interference construct targeting ATF4 mRNA reduces muscle fiber atrophy induced by fasting (21). These studies indicated an important role for ATF4 in fasting-induced muscle atrophy, and raised the possibility that ATF4 might also mediate other types of muscle atrophy, such as disuse atrophy, which most commonly occurs when muscles are immobilized by limb casting or bedrest.
The mechanism by which ATF4 promotes muscle atrophy is not yet known. ATF4 does not increase atrogin-1/MAFbx or MuRF1 mRNAs (21), the first well characterized atrophy-associated transcripts, which are partially required for muscle atro-phy (9,22). This indicates the existence of a previously unrecognized pathway that operates in parallel to, or downstream of, known atrophy pathways. As an initial step toward finding potential ATF4 target genes, we used exon expression arrays to identify five mouse skeletal muscle mRNAs that are induced by both ATF4 overexpression and fasting: Gadd45a, Cdkn1a, Peg3, Ankrd1, and Csrp3 (21). Of these, Gadd45a is particularly intriguing because other microarray studies also associated Gadd45a induction with skeletal muscle atrophy in mice, pigs, and humans (3)(4)(5)(6)(7)(8). However, the role of Gadd45a in skeletal muscle is not known. Indeed, many mRNAs are induced in atrophic muscle, and at least some (including Ankrd1, atrogin-1, and MuRF1) are not sufficient to cause muscle atrophy (10,12,23). Thus, it is not known if ATF4 causes atrophy by inducing Gadd45a.
In the current studies, we initially tested whether ATF4 might play a broader role in muscle atrophy by generating and studying muscle-specific ATF4 knock-out (ATF4 mKO) mice. When it became clear that ATF4 promotes both fasting-and immobilization-induced atrophy, we continued our search for the downstream mechanism. Using unbiased exon expression arrays, we sought to discover mRNAs that were reduced in the muscle of ATF4 mKO mice and also induced by ATF4 overexpression in both skeletal muscle and cultured myotubes. When we found that Gadd45a was the only mRNA that met all of these criteria, we tested the hypothesis that Gadd45a might be a critical mediator of skeletal muscle atrophy.

EXPERIMENTAL PROCEDURES
Mouse Protocols-ATF4 mKO mice were generated and genotyped as described in supplemental Fig. S1, A-H. ATF4 mKO mice were compared with ATF4(L/L);MCK-Cre(0/0) littermates, and all experiments used 9 -12-week-old males. C57BL/6 mice were also males, obtained from the National Cancer Institute at ages 6 -8 weeks, and used for experiments within 3 weeks of their arrival. Fasting, unilateral hindlimb denervation, and electroporation of mouse TA muscles were performed as described previously (21,24). Unilateral TA muscle immobilization was performed under isoflurane anesthesia using an Autosuture Royal 35W skinstapler (Tyco Healthcare, Point Claire, QC, Canada) as described previously (25,26). All mice were housed at 21°C in colony cages with 12-h light/12-h dark cycles. Mice were maintained on standard chow (Harlan Teklad formula 7013) and provided with water ad libitum. All animal procedures were approved by the Institutional Animal Care and Use Committee of the University of Iowa.
Immunohistochemistry, H&E Staining, and Light Microscopy of Mouse Muscle-For fiber type analysis, mouse tibialis anterior muscles (TAs) were harvested and fixed in 10% zinc formalin for 16 h, processed with RMC1530 parafin tissue processor, and then embedded in paraffin. A Leica RM2135 ultramicrotome was used to prepare 5-m sections, which were then deparaffinized and subjected to epitope retrieval with Antigen Unmasking Solution (Vector Labs H-3300) and a Pelco Biowave. Nonspecific peroxidase activity was quenched with 3% H 2 O 2 in methanol. Blocking and primary antibody incubation utilized the mouse on mouse (M.O.M.) kit (Vector Labs, BMK-2202) and either fast myosin heavy chain (Sigma, number M4276) or slow myosin heavy chain (Sigma, clone NOQ7.5.4.D, number M8421). Slides were then washed and incubated with Envision plus anti-mouse HRP (Dako K4001) antibody followed by visualization utilizing DAB (DAB peroxidase substrate Kit, 3,3Ј-diaminobenzidine kit, Vector Labs SK-4100). To localize Gadd45a, TAs were fixed in 4% paraformaldehyde for 16 h, placed in 30% sucrose (w/v) for 24 h, and then embedded in tissue freezing medium. A Microm HM 505E cryostat was then used to prepare 8-m sections, which were rinsed 3 times with PBS (pH 7.4) and then blocked with PBS containing 5% normal goat serum (NGS) for 1 h, followed by an overnight incubation with a 1:50 dilution of rabbit monoclonal anti-FLAG (Sigma, number F2555) in PBS containing 5% normal goat serum at 4°C. After incubation, muscle sections were washed 3 times with PBS and then incubated with Alexa 568conjugated secondary antibody (1:400) for 1 h at room temperature. Muscle sections were then washed 3 times with PBS and then covered with Vectashield mounting medium. For H&E staining, sections were prepared using the same technique used for fiber type analysis, and then stained with H&E according to standard protocols. To analyze transfected fibers, TA sections were prepared and imaged as described previously (21). All sections were examined and photographed on an Olympus IX-71 microscope equipped with a DP-70 camera. Image analysis was performed using ImageJ software. Muscle fiber diameters were measured using the lesser diameter technique as described previously (29).
Transmission Electron Microscopy (TEM) of Mouse Muscle-Mouse TA muscles were fixed in 2.5% glutaraldehyde in 0.1 M cacodylate buffer (pH 7.4) at 4°C overnight, rinsed 3 times with 0.1 M calcodylate buffer, and then postfixed and stained with 1% osmium tetroxide (OsO 4 ) and 1.5% potassium ferrocyanide (K 4 Fe(CN) 6 ) in 0.2 M cacodylate buffer for 1.5 h at room tem-perature. Skeletal muscle sections were then stained with 2.5% uranyl acetate for 30 min, dehydrated by a series of ethanol dilutions (50 -100%), infiltrated with graded mixtures of propylene oxides and Epoxy Resin 12, and then embedded in 100% Epoxy Resin 12. Ultra-thin sections (Ϸ85 nm) were cut using a Leica UC6 ultramicrotome and stained with 2% uranyl acetate and lead citrate. Sections were examined and photographed with a JEM-1230 transmission electron microscope equipped with a Gatan Ultra Scan 2k ϫ 2k CCD camera. Myonuclear diameter was measured with the lesser diameter method and ImageJ software.
C2C12 Myotube Culture and Infection-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 number  containing antibiotics (100 units/ml of penicillin, 100 g/ml of 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%. All assays except protein degradation (described below) were performed 48 h post-infection on day 9.
Myotube Protein Synthesis and Protein Degradation-[ 3 H]Leucine (120 Ci/mmol) and [ 3 H]tyrosine (40 Ci/mmol) were obtained from ARC. For analysis of protein synthesis, [ 3 H]leucine incorporation into cultured myotubes was determined as described previously (30). Protein degradation assays were performed according to a previously described protocol (16): myotubes were incubated with [ 3 H]tyrosine (4 Ci/ml) for 20 h (to label long-lived proteins) and then switched to chase medium (DMEM, antibiotics, and 2 mM unlabeled tyrosine) for 2 h. Myotubes were then rinsed with PBS, after which 1 ml of chase medium 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. Medium samples were collected 36 h post-infection and mixed with TCA (15% final concentration) for 1 h at 4°C. Precipitated proteins were washed twice with 10% TCA and twice with 95% EtOH, and then radioactivity was measured by liquid scintillation analysis. The acid-soluble radioactivity reflects the amount of proteins degraded and was expressed relative to the total cellular radioactivity present at the time of infection.
Histological Analysis of Myotubes-All myotube imaging was performed on an Olympus IX-71 microscope equipped with a DP-70 camera and epifluorescence filters. Image analysis was performed using ImageJ software. To analyze myotube size, we averaged three width measurements per GFP-positive myofiber. To localize Gadd45a, myotubes were washed two times with ice-cold PBS, fixed in 4% paraformaldehyde for 10 min, and permeabilized with PBS (pH 7.4) containing 0.5% Triton X-100 for 15 min. Permeabilized myotubes were blocked with PBS containing 1% bovine serum albumin (BSA) and 5% normal goat serum for 1 h, followed by an overnight incubation with 1:50 dilution of rabbit monoclonal anti-FLAG (Sigma, number F2555) in PBS containing 1% BSA at 4°C. After incubation, the myotubes were washed 3 times with PBS and then incubated with Alexa 568-conjugated secondary antibody (1:400) for 1 h at room temperature. Myotubes were then washed 3 times with PBS and then covered with Vectashield mounting medium. For trypan blue staining, myotubes were rinsed 3 times with PBS, stained with 0.2% trypan blue (in PBS) for 5 min at room temperature, and then rinsed 2 times with PBS. As a positive control for cell death, myotubes were treated with 80% ethanol for 20 min prior to staining with 0.2% trypan blue.
Exon Arrays and Quantitative Real-time RT-PCR (qPCR) in Muscle and Myotubes-Extraction of skeletal muscle RNA and RNA hybridizations to Mouse Exon 1.0 ST arrays (Affymetrix) were performed as described previously (21). Myotube RNA was extracted using TRIzol solution (Invitrogen), and then purified using the RNeasy kit and RNase-free DNase Set (Qiagen). qPCR analysis of mouse ATF4, Gadd45a, Bax, Casp3, Ar, Thra, Ghr, Slc2a4, Hk2, Suclg1, Cox11, Ppargc1a, Tfam, Nos1, Trim63, Fbxo32, Ctsl, Bnip3, Map1lc3a, and Vegfa mRNA was performed using TaqMan Gene Expression Assays (Applied Biosystems). For qPCR studies, first strand cDNA was synthesized in a 20-l reaction that contained 2 g of RNA, random hexamer primers, and components of the High Capacity cDNA reverse transcription kit (Applied Biosystems). qPCR was carried out using a 7500 Fast Real-time PCR System (Applied Biosystems). All qPCR were performed in triplicate and the cycle threshold (C t ) values were averaged to give the final results. To analyze the data, we used the ⌬C t method, with the level of 36B4 mRNA serving as the invariant control.
Analysis of Caspase Activity-Mouse TA muscles were snap frozen in liquid N 2 , and homogenized in 1 ml of cold lysis buffer solution containing 50 mM Tris (pH 7.4), 150 mM NaCl, 0.1% (v/v) Triton X-100, 1.5 mM MgCl 2 , 1 mM sodium EDTA, 1 mM sodium EGTA, and cOmplete Mini protease inhibitor mixture (Roche Applied Science) using a polytron (Tissue Master 240, Omni International) for 1 min on setting number 8. C2C12 myotube homogenates were prepared by scrapping PBSwashed myotubes into cold lysis buffer solution (above) that contained cOmplete Mini protease inhibitor mixture (Roche Applied Science) then lysed with 10 passes through a 22-gauge needle. Muscle and myotube homogenates were centrifuged at 4°C and 10,000 ϫ g for 10 min, and caspase activity assays were set-up in white-walled 96-well plates; each assay contained 20 g of protein from the sample supernatant mixed with an equal volume of caspase reagent (Promega, Madison, WI). Reactions were incubated on a rocker for 30 min at room temperature, and then luminescence was measured on a SpectraMax L luminescence microplate reader (Molecular Devices, Sunnyvale, CA). All reactions were performed in triplicate and values were averaged to give the final results.
Analysis of Mitochondrial DNA-Mouse skeletal muscle DNA was extracted using the QIAamp DNA mini kit (Qiagen). Mitochondrial and nuclear DNA was quantified by qPCR; reactions contained, in a final volume of 20 l, 10 ng of muscle DNA, 660 nM forward and reverse primers, and 10 l of 2ϫ Power SYBR Green Master Mix (ABI). Ndufv1 and mtDNA primer sequences were previously described (31,32). qPCR was carried out using a 7500 Fast Real-time PCR System (Applied Biosystems). All qPCR were performed in triplicate and the C t values were averaged to give the final results. To analyze the data, we used the ⌬C t method, with the level of 36B4 mRNA serving as the invariant control.
Statistical Analysis-Unless otherwise noted in the figure legends, we used paired t tests to compare within subject samples and unpaired t tests for all other comparisons.

Loss of ATF4 Delays Skeletal Muscle Atrophy Induced by
Fasting or Immobilization-To generate ATF4 mKO mice, we flanked the coding region of the mouse ATF4 gene (exons 2 and 3) with LoxP restriction sites, and then excised the floxed ATF4(L) allele by crossing ATF4(L/L) mice to transgenic mice carrying Cre recombinase under control of the muscle creatine kinase (MCK) promoter (supplemental Fig. S1, A-H) (33). As expected, the MCK-Cre transgene specifically excised the ATF4(L) allele in skeletal muscle and heart, reducing ATF4 mRNA in skeletal muscle by Ͼ95% ( Fig. 2F and supplemental Fig. S1H). Residual ATF4 mRNA may be from satellite cells and non-muscle cells, which do not express MCK-Cre (34). ATF4 mKO were born at the expected Mendelian frequency and lacked any overt phenotype up to 9 months of age (the longest period of observation). Relative to littermate control mice lacking MCK-Cre, ATF4 mKO mice possessed normal total body, skeletal muscle, heart, and liver weights ( Fig. 1A and supplemental Fig. S1J). Histological examination of ATF4 mKO skel-etal muscle revealed normal percentages and sizes of type I and type II muscle fibers, and no signs of degeneration, regeneration, or inflammation ( Fig. 1, C and E, and supplemental Fig.  S1I). Thus, skeletal muscle ATF4 expression was not required for skeletal muscle development, and its absence did not induce muscle hypertrophy.
Because RNAi-mediated knockdown of ATF4 reduces atrophy of TA muscle fibers during fasting (21), we anticipated that ATF4 mKO mice would be resistant to fasting-induced muscle atrophy. After 24 or 48 h of fasting, ATF4 mKO TA muscles and muscle fibers were significantly larger than those of control mice (Fig. 1, A-C, and supplemental Fig. S1J). Other skeletal  A-C, ATF4 mKO mice and littermate controls were allowed ad libitium access to food or fasted by removing food but not water for 24 or 48 h. **, p Ͻ 0.01. *, p Ͻ 0.05. A, TA muscle weights. Each data point represents the mean Ϯ S.E. from Ն9 mice. B, TA muscle fiber size in the absence and presence of a 48-h fast. Left, mean fiber diameters Ϯ S.E. Right, fiber size distributions. Data are from Ն250 muscle fibers per TA, from Ն3 mice per condition. C, representative H&E stains from B. D-F, ATF4 mKO mice and littermate controls were subjected to unilateral TA immobilization for 0, 3, or 7 days. D, TA muscle weights. In each mouse, weight of the immobile (atrophied) TA was normalized to the weight of the mobile (non-atrophied) TA, which was set at 1. Each data point represents the mean Ϯ S.E. from Ն10 mice. *, p Ͻ 0.01. E, representative H&E stains from D. F, TA muscle fiber size. In each mouse, Ն350 muscle fibers were measured in each TA, and the mean fiber diameter in the immobile TA was normalized to the mean fiber diameter in the mobile TA, which was set at 1. Each data point represents the mean Ϯ S.E. from Ն5 mice. *, p Ͻ 0.05. muscles, including fast-and slow-twitch muscles (biceps and soleus, respectively), also exhibited reduced atrophy (supplemental Fig. S1J). Interestingly, loss of ATF4 appeared to have a greater protective effect after 24 h of fasting than after 48 h of fasting. This was apparent in the TAs (Fig. 1, A-C) and in the quadriceps and triceps (supplemental Fig. S1J). These data suggested that loss of ATF4 reduced fasting-induced atrophy by delaying its progression.
We asked whether loss of ATF4 might delay muscle atrophy induced by a different stress: muscle immobilization. To test this, we immobilized one TA with a surgical staple (25,26), leaving the contralateral, mobile TA as an internal control. Relative to littermate control TAs, ATF4 mKO TAs underwent less muscle and muscle fiber atrophy during the first 3 days of immobilization ( Fig. 1, D-F). However, after 7 days of immobilization, the amount of muscle and muscle fiber atrophy was equivalent between the two genotypes ( Fig. 1, D-F). Thus, loss of ATF4 delayed immobilization-induced muscle atrophy. Collectively, these results indicate that ATF4 is partially required for early, essential events in immobilization-and fasting-induced skeletal muscle atrophy.

Identification of Gadd45a as a Transcript That Is Reduced in ATF4 mKO Muscle and Increased by ATF4 Overexpression in
Both Mouse Muscle and Cultured C2C12 Myotubes-Overexpressing ATF4, but not a transcriptionally inactive ATF4 construct (ATF4⌬bZIP), induces skeletal muscle fiber atrophy in mice (21). To develop a complementary in vitro system, we infected C2C12 skeletal myotubes with adenovirus co-expressing ATF4 and GFP (Ad-ATF4). Control myotubes were infected with adenoviruses expressing only GFP (Ad-GFP) or GFP plus ATF4⌬bZIP, a full-length, transcriptionally inactive ATF4 construct that does not induce atrophy in mouse skeletal muscle fibers (21). Immunoblot analysis confirmed that Ad-ATF4 and Ad-ATF4⌬bZIP generated ATF4 and ATF4⌬bZIP, respectively ( Fig. 2A). Similar to its effect in mouse muscle (21), ATF4 overexpression induced myotube atrophy (Fig. 2, B and C).
To determine how ATF4 promotes muscle atrophy, we used genomewide exon expression arrays to search for mRNAs that satisfied three criteria: 1) induced by Ad-ATF4 in myotubes; 2) reduced by loss of ATF4 in mouse muscle; and 3) induced by ATF4 overexpression in mouse muscle. To determine effects of Ad-ATF4, we compared myotubes that were infected with Ad-ATF4 or Ad-ATF4⌬bZIP. To determine effects of ATF4 mKO, we compared TA muscles from fasted ATF4 mKO mice and ATF4(L/L) littermate controls. To determine effects of ATF4 overexpression in mouse muscle, we compared C57BL/6 TA muscles that were transfected with plasmid encoding mouse ATF4 or empty vector, as described previously (21). Using p Յ 0.01 as the threshold for statistical significance, only one mRNA, Gadd45a, satisfied all three criteria ( Fig. 2D and supplemental Tables S1 and S2). qPCR analysis confirmed that Ad-ATF4 increased Gadd45a mRNA in C2C12 myotubes (Fig.  2E), and that Gadd45a mRNA was significantly reduced in ATF4 mKO muscle (Fig. 2F). Previous qPCR studies confirmed that ATF4 overexpression increases Gadd45a mRNA in C57BL/6 muscle (21). In contrast, atrogin-1 or MuRF1 mRNAs are not increased by ATF4 overexpression (21), and they were not reduced in ATF4 mKO muscle (Fig. 2F). These data led us to hypothesize that ATF4 might cause muscle atrophy by inducing Gadd45a.
Gadd45a Is Required for Muscle Fiber Atrophy Induced by Immobilization, Fasting, and Denervation-To begin to investigate the function of Gadd45a, we asked whether it is required for immobilization-induced atrophy. To test this, we transfected bilateral TAs of C57BL/6 mice with plasmids encoding  artificial miRNAs targeting Gadd45a (miR-Gadd45a). TAs of control mice were transfected with plasmid expressing a nontargeting control miRNA (miR-Control). All plasmids co-expressed EmGFP as a transfection marker. Plasmid transfection was achieved via electroporation, which transfects terminally differentiated muscle fibers, but not satellite or connective tissue cells (35). Three days after transfection, we performed unilateral TA immobilization, and 1 week later, we harvested and compared bilateral TAs. In control (mobile) muscles, miR-Gadd45a did not alter muscle fiber size (Fig. 3, A and B); thus, reduction of Gadd45a, like loss of ATF4, did not induce fiber hypertrophy. However, in immobilized muscles, miR-Gadd45a prevented the induction of Gadd45a mRNA (Fig. 3A), and reduced muscle fiber atrophy (Fig. 3, A and B). Similar results were obtained with a second miR-Gadd45a construct that targeted a different region of the Gadd45a transcript (supplemental Fig. S2A). These data indicate that Gadd45a is required for immobilization-induced atrophy.
To determine whether Gadd45a might play a broader role in muscle atrophy, we examined the effects of miR-Gadd45a during fasting and muscle denervation. To investigate fasting, we transfected miR-Gadd45a into one TA, and miR-Control into the contralateral TA. We then subjected mice to a 24-h fast, and found that miR-Gadd45a significantly impaired fasting-induced muscle fiber atrophy (Fig. 3C). Similar results were obtained with a second miR-Gadd45a construct (supplemental Fig. S2B).
Like immobilization and fasting, muscle denervation strongly induces atrophy and Gadd45a mRNA (36). To test the role of Gadd45a in denervated muscle, we transfected miR-Control or miR-Gadd45a bilaterally, then transected one sciatic nerve to induce atrophy, leaving the contralateral leg as an intrasubject control. One week later, we compared innervated and denervated muscles. Under control conditions, denervation reduced muscle fiber size by 22 Ϯ 3% (Fig. 3D). However, in the presence of miR-Gadd45a, denervation reduced muscle fiber size by only 12 Ϯ 2% (Fig. 3D), indicating a 45% reduction in denervation-mediated atrophy. Interestingly, in all three atrophy models that we examined (immobilization, fasting, and denervation), miR-Gadd45a protected type II but not type I muscle fibers from atrophy (supplemental Fig. S2, C-F). The percentages of type I and type II fibers were unchanged (supplemental Fig. S2, C-F).
To determine whether Gadd45a is required for atrophy induced by ATF4 overexpression, we co-transfected plasmid encoding ATF4 with miR-Control or miR-Gadd45a. We  . Gadd45a is required for skeletal muscle fiber atrophy induced by immobilization, fasting, and denervation. A and B, Gadd45a is required for immobilization-induced muscle fiber atrophy. On day 0, bilateral C57BL/6 TA muscles were transfected with either 20 g of p-miR-Control or 20 g of p-miR-Gadd45a, as indicated. All plasmids carried EmGFP as a transfection marker. On day 3, right hindlimbs were immobilized. On day 10, bilateral TA muscles were harvested for analysis. A, upper panel, mRNA levels were determined by qPCR and normalized to levels in mobile, p-mir-Control-transfected TA; data are mean Ϯ S.E. from 3 muscles per condition. Lower left, mean fiber diameters Ϯ S.E. from 5 TAs per condition. Lower right, fiber size distributions. Statistical differences were determined using a linear mixed model with a random effect for mouse (58). Different letters are statistically different (p Յ 0.05). B, representative images. C, Gadd45a is required for fasting-induced muscle fiber atrophy. On day 0, C57BL/6 TA muscles were transfected with either 20 g of p-miR-Control (left leg) or 20 g of p-miR-Gadd45a (right leg). On day 9, mice were fasted for 24 h and then harvested for analysis. Left, mean fiber diameters Ϯ S.E. from Ն4 TAs per condition. *, p Ͻ 0.01. Right, fiber size distributions. D, Gadd45a is required for denervation-induced muscle atrophy. On day 0, C57BL/6 TA muscles were transfected bilaterally with either 20 g of p-miR-Control or 20 g of p-miR-Gadd45a. On day 3, the left sciatic nerve was transected. On day 10, bilateral TA muscles were harvested. Left, mean fiber diameters Ϯ S.E. from Ն5 TAs per condition. Statistical differences were determined using a linear mixed model with a random effect for mouse; different letters are statistically different (p Յ 0.05). Right, fiber size distributions. E, Gadd45a is required for ATF4-mediated muscle atrophy. C57BL/6 TA muscles were transfected with 10 g of p-ATF4-FLAG ϩ either 20 g of p-miR-Control (left TA) or 20 g of p-miR-Gadd45a (right TA), then harvested 10 days later. Left, mean fiber diameters Ϯ S.E. from 5 TAs per condition. *, p ϭ 0.03. Right, fiber size distributions.

JOURNAL OF BIOLOGICAL CHEMISTRY 27295
found that miR-Gadd45a increased fiber size, indicating reduced ATF4-mediated atrophy (Fig. 3E). Taken together, these data indicate that Gadd45a is required for atrophy induced by immobilization, fasting, denervation, and ATF4 overexpression.
To determine whether Gadd45a overexpression might induce muscle fiber atrophy in vivo, we transfected plasmid encoding Gadd45a into C57BL/6 TA muscle. The contralateral TA muscle was transfected with empty plasmid vector (pcDNA3). To identify transfected muscle fibers, bilateral TA muscles were co-transfected with plasmid encoding eGFP (pCMV-eGFP). Immunoblot analysis confirmed Gadd45a overexpression specifically in the TA muscle that was transfected with Gadd45a plasmid (Fig. 4D). Relative to control transfected fibers, muscle fibers transfected with Gadd45a were significantly smaller (Fig. 4E). To test whether Gadd45a-mediated atrophy requires ATF4, we transfected Gadd45a plasmid into ATF4 mKO TA muscles. Gadd45a induced atrophy under both fed conditions (Fig. 4, F and G) and fasted conditions (supplemental Fig. S3A), indicating that ATF4 is not required for Gadd45a-mediated atrophy. Gadd45a overexpression did not alter the percentages of type I or type II muscle fibers, and it promoted atrophy of type II but not type I fibers (supplemental Fig. S3B). Thus, increased Gadd45a expression causes atrophy both in vitro and in vivo.
Gadd45a Enters Myonuclei and Induces Myonuclear Remodeling-To begin to determine how Gadd45a promotes atrophy, we used immunohistochemistry to localize Gadd45a in myotubes and muscle fibers. Consistent with previous findings in non-muscle cells (37), Gadd45a was predominantly nuclear in myotubes (Fig. 5A) and muscle fibers (Fig. 5B), suggesting that Gadd45a promotes muscle atrophy by altering a process within myonuclei. To further investigate this possibility, we used TEM to examine nuclear morphology in muscle fibers that had undergone Gadd45a-mediated atrophy. Our positive control was muscle denervation, which increases Gadd45a mRNA to a similar level as Gadd45a overexpression (Table 1 and supplemental Fig. S4A). One week of muscle denervation induced the classical ultrastructural changes previously described by Korényi-Both: "the nuclei lose their cigarlike shape and become swollen, rounded and plump, with prominent nucleoli" (38) (Fig. 5C). Interestingly, 1 week of Gadd45a overexpression induced similar changes in nuclear morphology (Fig. 5D). These data suggest that Gadd45a plays an important role in the myonuclear remodeling that occurs during muscle atrophy.
Gadd45a Generates 40% of the mRNA Expression Changes That Occur during Muscle Denervation-In other cell types, nuclear remodeling is associated with altered gene expression (39). Thus, the finding that Gadd45a altered myonuclear structure suggested that it might contribute to gene expression changes that occur during muscle atrophy. To test this, we used exon expression arrays to compare effects of denervation and Gadd45a on levels of Ͼ16,000 mRNAs. Gadd45a was overexpressed in ATF4 mKO muscle to eliminate any potential contribution from ATF4. Using p Յ 0.01 as the threshold for statistical significance, we found that denervation significantly altered levels of 1674 mRNAs, decreasing 965 and increasing 709. Of the 965 mRNAs decreased by denervation, 40% were significantly decreased by Gadd45a, 3% were increased and 57% were unaffected (Fig. 5E). Of the 709 mRNAs increased by denervation, 40% were significantly increased by Gadd45a, 2% were decreased, and 58% were unaffected (Fig. 5E). Altogether, we identified Ͼ600 mRNAs whose levels were similarly altered by denervation and Gadd45a overexpression (supplemental Table  S3). Thus, increased Gadd45a expression generates many, but not all, of the positive and negative mRNA expression changes in denervated muscle.  Gadd45a Represses Anti-atrophy Genes and Induces Pro-atrophy Genes-Gadd45a altered levels of many mRNAs whose roles in muscle atrophy are unknown (supplemental Table S3). However, some patterns could be discerned. For example, both denervation and Gadd45a repressed many interconnected mRNAs involved in anabolic signaling, protein synthesis, glucose uptake, glycolysis, oxygen delivery, mitochondrial biogenesis, the citric acid cycle and oxidative phosphorylation ( Table  1). We used qPCR to validate 11 representative changes, including repression of mRNAs encoding PGC-1␣, the growth hormone receptor, androgen receptor, GLUT4, hexokinase-II, VEGF-A, nNOS, and thyroid hormone receptor-␣ (supplemental Fig. S4A). Gene set enrichment analysis also indicated that Gadd45a repressed growth and energy-yielding pathways; 10 gene sets were significantly depleted by both denervation and Gadd45a, including growth hormone signaling, insulin signaling, glycolysis, and the citrate cycle (supplemental Fig. S4B). Because anabolic signaling and PGC-1␣ prevent skeletal muscle atrophy (18,19,24,35,40,41), these data suggested that FIGURE 5. Gadd45a is a myonuclear protein that alters myonuclear structure and reprograms skeletal muscle gene expression. A and B, immunohistochemical detection of FLAG-tagged Gadd45a in myotube nuclei (A) and skeletal muscle fiber nuclei (B). In A, myotubes were infected with Ad-Gadd45a for 48 h before immunohistochemistry. In B, mouse muscle fibers were transfected with 2 g of p-eGFP ϩ 20 g of p-Gadd45a-FLAG and then harvested 10 days later for immunohistochemistry. C and D, Gadd45a alters myonuclear morphology in a manner similar to muscle denervation. C, left-sided sciatic nerves of C57BL/6 mice were transected, and bilateral TA muscles were harvested 7 days later for TEM analysis. Top, representative images. Bottom, effect of denervation on the lesser diameter of myonuclei. Data are mean Ϯ S.E. from Ͼ50 myonuclei per condition. *, p Ͻ 0.01. D, C57BL/6 TA muscles were transfected with 20 g of pcDNA3 (left TA) or 20 g of p-Gadd45a-FLAG (right TA), then harvested 7 days later for TEM analysis. Top, representative images. Bottom, effect of Gadd45a on the lesser diameter of myonuclei. Data are mean Ϯ S.E. from Ͼ30 myonuclei per condition. *, p Ͻ 0.01. E, effects of denervation and Gadd45a on skeletal muscle mRNA levels.
To determine effects of denervation, left-sided sciatic nerves of C57BL/6 mice were transected, and bilateral TA muscles were harvested 7 days later. Bilateral TA muscle mRNA levels were then measured with exon expression arrays, and levels in denervated muscles were normalized to levels in contralateral innervated muscles. To determine effects of Gadd45a, ATF4 mKO TA muscles were transfected and harvested as in D. Bilateral TA muscle mRNA levels were then measured with exon expression arrays, and levels in Gadd45a-transfected muscles were normalized to levels in contralateral control muscles. n ϭ 4 arrays per condition. Statistical significance was defined as p Յ 0.01 by paired t test. E, denervation significantly altered levels of 1674 mRNAs (out of Ͼ16,000 mRNAs measured). The pie chart shows effects of denervation and Gadd45a on these mRNAs. Gadd45a reprograms myonuclei in a manner that removes barriers to muscle atrophy. In addition to reducing mRNAs that maintain muscle, Gadd45a induced genes that promote atrophy. For example, both denervation and Gadd45a increased several mRNAs involved in lysosome-and caspase-mediated proteolysis (Table  1), which are essential for muscle atrophy (15)(16)(17). Interestingly, both denervation and Gadd45a increased Runx1 mRNA and six known Runx1 targets (supplemental Fig. S4C); these mRNAs limit muscle damage in denervated, atrophying muscle (42). We confirmed some representative changes using qPCR (supplemental Fig. S4A). Gene set enrichment analysis also supported the notion that Gadd45a increased pro-atrophy mRNAs; 13 gene sets were significantly enriched by both denervation and Gadd45a (supplemental Fig. S4B). These included several stress-signaling pathways that promote muscle atrophy (e.g. NF-B, p53, TLR, and TNFR1 pathways (13,43,44)), suggesting that Gadd45a mediates or resembles these pro-atrophy pathways.
Because atrogin-1 and MuRF1 also mediate a component of proteolysis during muscle atrophy (9 -12, 14), we examined the effect of Gadd45a on atrogin-1 and MuRF1 mRNAs. Although denervation increased atrogin-1 or MuRF1 mRNAs (supplemental Fig. S4A), Gadd45a did not increase these mRNAs at early or late time points in skeletal muscle (supplemental Fig.  S4D) or myotubes (supplemental Fig. S5B). Thus, atrogin-1 and MuRF1 are among the 58% of denervation-induced mRNAs that are not under the control of Gadd45a. Taken together, these data indicate that Gadd45a orchestrates a multitude of nuclear changes that are predicted to promote skeletal muscle atrophy.
In contrast to its effect on anabolic mRNAs, Gadd45a increased mRNAs involved in autophagy (including Map1lc3a, which encodes LC3) and the caspase pathway (including Casp3, which encodes caspase-3) ( Table 1 and supplemental Fig. S4A). This suggested that Gadd45a might also increase proteolysis, In each muscle, PGC-1␣, LC3-II, and caspase-3 signals were normalized to the actin signal, and in each mouse, levels in the presence of Gadd45a were normalized to levels in the absence of Gadd45a. Data are mean Ϯ S.E. from 4 mice. *, p Ͻ 0.05. B, Gadd45a reduces mitochondrial DNA. C57BL/6 TA muscles were transfected and harvested as in A for qPCR analysis of mitochondrial DNA (mtDNA), which was normalized to the amount of nuclear DNA (nDNA) in the same muscle. Data are mean Ϯ S.E. from 7 mice. *, p Ͻ 0.02. C, Gadd45a reduces Akt and GSK-3␤ phosphorylation. C2C12 myotubes were infected with control virus (Ad-ATF4⌬bZIP) or Ad-Gadd45a, and then harvested 48 h later for SDS-PAGE and immunoblot analysis with the indicated antibodies. Left, representative immunoblots. Right, quantification. Phospho-Akt and phospho-GSK-3␤ signals were normalized to the actin signal from the same sample. Levels in Ad-Gadd45a-infected myotubes were then normalized to levels in control myotubes. Data are mean Ϯ S.E. from 4 experiments. *, p Ͻ 0.05. D, Gadd45a reduces protein synthesis. C2C12 myotubes were infected with control virus (Ad-ATF4⌬bZIP) or Ad-Gadd45a. Protein synthesis was assessed 48 h later by measuring [ 3 H]leucine incorporation. Levels in Ad-Gadd45a-infected myotubes were then normalized to levels in control myotubes. Data are mean Ϯ S.E. from 5 experiments. *, p Ͻ 0.01. E, Gadd45a increases proteolysis. C2C12 myotubes were incubated with [ 3 H]tyrosine for 20 h, washed with chase medium for 2 h, and then infected with control virus (Ad-ATF4⌬bZIP) or Ad-Gadd45a in fresh chase medium. Protein degradation was assessed 36 h later by measuring [ 3 H]tyrosine release. Levels in Ad-Gadd45a-infected myotubes were then normalized to levels in control myotubes. Data are mean Ϯ S.E.; n ϭ 8. *, p Ͻ 0.05. F, Gadd45a induces autophagosome formation. C57BL/6 TA muscles were transfected as in A, and harvested 7 days later for TEM analysis. G, Gadd45a increases caspase-mediated proteolysis. C57BL/6 TA muscles were transfected and harvested as in A, and then caspase-mediated proteolysis was measured. In each mouse, the level in the presence of Gadd45a was normalized to the level in the absence of Gadd45a. Data are mean Ϯ S.E. from 7 mice. *, p Ͻ 0.01. which we confirmed in myotubes (Fig. 6E). To further examine the potential role of autophagy, we studied atrophic muscle fibers overexpressing Gadd45a, and found that they contained increased total and lipidated LC3 (Fig. 6A), as well as autophagosomes (Fig. 6F). Similarly, Gadd45a increased two key autophagy mRNAs (Bnip3 and Ctsl), as well as Bnip3 protein, in myotubes (supplemental Fig. S5, B and C). To further examine the potential role of the caspase pathway, we studied atrophic muscle fibers overexpressing Gadd45a and found that they contained increased caspase-3 protein (Fig. 6A). As a result, caspase-mediated proteolysis was also increased (Fig. 6G). Similarly, Gadd45a increased caspase-mediated proteolysis in cultured myotubes (supplemental Fig. S5D), without causing myotube death (supplemental Fig. S5E).
Thus, as predicted by its effects on mRNA expression, Gadd45a reduced two proteins that inhibit muscle atrophy (PGC-1␣ and activated Akt), reduced mitochondria, increased three proteins that promote muscle atrophy (activated GSK-3␤, lipidated LC3, and caspase-3), inhibited a critical anabolic process (protein synthesis), and induced two key proteolytic systems (autophagy and caspase-mediated proteolysis). These data support the notion that Gadd45a causes muscle atrophy by reprogramming skeletal muscle gene expression.
In healthy muscle, ATF4 and Gadd45a levels are relatively low. However, acute stresses such as fasting and muscle disuse stimulate ATF4 expression (2), which contributes to the induction of Gadd45a expression. Gadd45a translocates to the nucleus, where it alters myonuclear morphology and induces widespread changes in skeletal muscle mRNA expression. mRNAs involved in anabolic signaling, protein synthesis, glucose uptake, glycolysis, oxygen delivery, mitochondrial biogenesis, citric acid cycle, and oxidative phosphorylation are repressed. Conversely, mRNAs involved in autophagy and caspase-mediated proteolysis are induced. By reprogramming skeletal muscle gene expression, Gadd45a stimulates multiple interconnected atrophy mechanisms in the cytosol. On one hand, Gadd45a reduces barriers to atrophy, including PGC-1␣ expression, Akt activity, protein synthesis, and mitochondria. On the other hand, Gadd45a increases mediators of atrophy, including activated GSK-3␤, activated caspase-3, and autophagy. Thus, Gadd45a coordinates a comprehensive program for skeletal muscle atrophy.
Because ATF4 and Gadd45a are not highly expressed under basal conditions, interventions that specifically target these proteins do not cause muscle fiber hypertrophy. However, the pathway emerges during stress, and thus reducing ATF4 or Gadd45a diminishes stress-induced muscle atrophy. Moreover, forced expression of ATF4 or Gadd45a induces atrophy in the absence of upstream stress. These data indicate a critical role in muscle atrophy, and suggest ATF4 and Gadd45a as potential therapeutic targets.
The ATF4/Gadd45a pathway is part of a larger signaling network with other important components. Loss of ATF4 only partially reduced Gadd45a expression, and it delayed but did not prevent muscle atrophy. This indicates that ATF4 plays an important early role in muscle atrophy, but other atrophy mechanisms compensate for the loss of ATF4 during prolonged stress. It also indicates the existence of other regulators upstream of Gadd45a. Potential candidates include FoxO transcription factors and p53, which are known to induce Gadd45a transcription in other settings (49,50). In addition, the microarray studies that pointed to Gadd45a as a key ATF4 target gene do not rule out the possibility that other important ATF4 target genes might also exist. Finally, Gadd45a generated Ͼ600 mRNA expression changes that occur during muscle denervation; however, this accounts for only 40% of the total changes in denervated muscles. This indicates the existence of other regulators that act in parallel to Gadd45a. Important mRNAs that are not controlled by Gadd45a include atrogin-1 and MuRF1. Because Akt activity and PGC-1␣ repress atrogin-1 and MuRF1 transcription (10,11,18), and because Gadd45a decreased Akt activity and PGC-1␣, it was surprising that Gadd45a did not increase atrogin-1 and MuRF1 mRNAs. This may reflect a requirement for other transcription factors that induce atrogin-1 and MuRF1, such as FoxO transcription factors (10,11), NF-B (13,43), and myogenin (12). Alternatively, atrogin-1 and MuRF1 transcription could lie upstream of Gadd45a. Furthermore, the current data do not rule out a role for atrogin-1 and MuRF1 proteins in Gadd45a-mediated atrophy. It is also important to note that Gadd45a regulates hundreds of mRNAs whose roles in muscle atrophy are not yet known.
By elucidating the critical importance of Gadd45a, this study generates some interesting questions for future investigation. In non-muscle cells, Gadd45a resides in ribosome-sized particles composed of both protein and RNA (51), and it promotes DNA demethylation by uniting thymine DNA glycosylase and activation-induced deaminase in a multimeric complex (52). In addition, Gadd45a and its paralog Gadd45b can function as transcriptional coactivators (53,54). This suggests Gadd45a might cooperate with other factors to generate changes in nuclear structure and skeletal muscle gene expression. In addition, it is interesting that Gadd45a induces the classical ultrastructural effects of early muscle denervation: rounded and swollen nuclei with prominent nucleoli. We speculate that myonuclear remodeling might be coupled to the reprogramming of genes. A similar model has been suggested in cancer cells (39).
Importantly, Gadd45a did not induce myonuclear pyknosis even though it increased caspase activity. Similarly, Gadd45a overexpression increased caspase activity in cultured myotubes without causing cell death. These findings are consistent with previous studies of denervated muscle: caspase-3 activity is required during the first 2 weeks of denervation-induced muscle atrophy (17), however, apoptotic loss of myonuclei does not occur in this time frame (55). It is also interesting that Gadd45a reduced mitochondria, but did not decrease the amount or size of type I fibers, which are particularly rich in mitochondria. RNAi targeting Gadd45a reduced atrophy in type II but not type I fibers. Selective effects on type II fibers may reflect previous findings that type II fibers are more prone to atrophy during fasting and denervation (56,57). Alternatively, it is possible that we missed an effect on type I fibers because of the time point we studied, or the muscle we studied (the TA, which contains many more type II fibers than type I fibers).
In summary, the current study elucidates a stress-induced pathway with a critical role in the signaling network that drives skeletal muscle atrophy. The centerpiece of this pathway is Gadd45a, a nuclear protein that stimulates myonuclear remodeling and widespread changes in skeletal muscle gene expression. By reprogramming muscle gene expression, Gadd45a reduces multiple barriers to atrophy and stimulates multiple atrophy mechanisms. This study also opens several new avenues for future investigation, including studies to determine whether this pathway might be a target for human therapeutics.