Research ArticleMUSCLE BIOLOGY

mTORC1 Promotes Denervation-Induced Muscle Atrophy Through a Mechanism Involving the Activation of FoxO and E3 Ubiquitin Ligases

See allHide authors and affiliations

Science Signaling  25 Feb 2014:
Vol. 7, Issue 314, pp. ra18
DOI: 10.1126/scisignal.2004809


Skeletal muscle mass and function are regulated by motor innervation, and denervation results in muscle atrophy. The activity of mammalian target of rapamycin complex 1 (mTORC1) is substantially increased in denervated muscle, but its regulatory role in denervation-induced atrophy remains unclear. At early stages after denervation of skeletal muscle, a pathway involving class II histone deacetylases and the transcription factor myogenin mediates denervation-induced muscle atrophy. We found that at later stages after denervation of fast-twitch muscle, activation of mTORC1 contributed to atrophy and that denervation-induced atrophy was mitigated by inhibition of mTORC1 with rapamycin. Activation of mTORC1 through genetic deletion of its inhibitor TSC1 (tuberous sclerosis complex 1) sensitized mice to denervation-induced muscle atrophy and suppressed the kinase activity of Akt, leading to activation of FoxO transcription factors and increasing the expression of genes encoding E3 ubiquitin ligases atrogin [also known as MAFbx (muscle atrophy F-box protein)] and MuRF1 (muscle-specific ring finger 1). Rapamycin treatment of mice restored Akt activity, suggesting that the denervation-induced increase in mTORC1 activity was producing feedback inhibition of Akt. Genetic deletion of the three FoxO isoforms in skeletal muscle induced muscle hypertrophy and abolished the late-stage induction of E3 ubiquitin ligases after denervation, thereby preventing denervation-induced atrophy. These data revealed that mTORC1, which is generally considered to be an important component of anabolism, is central to muscle catabolism and atrophy after denervation. This mTORC1-FoxO axis represents a potential therapeutic target in neurogenic muscle atrophy.


Motor innervation is an important regulator of skeletal muscle mass and function. Denervation—loss of motor innervation—results in muscle atrophy. Patients with traumatic nerve injury and motor neuron diseases—for example, amyotrophic lateral sclerosis, spinal muscular atrophy, post-polio syndrome, and progressive bulbar palsy—suffer from highly morbid denervation-associated muscle wasting for which there is currently no effective therapy (1, 2).

Muscle mass is maintained by a balance between protein synthesis and degradation (38). Adenosine triphosphate–dependent protein degradation, mediated by the ubiquitin proteasome system (UPS), is increased in atrophying muscle through activation of the muscle-specific E3 ubiquitin ligases atrogin [also known as MAFbx (muscle atrophy F-box protein)] and MuRF1 (muscle-specific ring finger 1) (911). Use of a proteasome inhibitor (12) or genetic deletion of each of the E3 ubiquitin ligases reduces denervation muscle atrophy (9), indicating that UPS-mediated protein degradation is a major pathway underlying this process. However, the molecular mechanisms coupling nerve activity and this UPS-mediated protein degradation are just emerging.

The axis consisting of histone deacetylases (HDACs) 4 and 5 and the transcription factors Dach2 and myogenin has been reported to control both the suppression and induction of numerous denervation-regulated genes (1315) including those encoding atrogin and MuRF1 (16). Myogenin directly mediates the transcriptional activation of the genes encoding these E3 ubiquitin ligases, and deletion of HDAC4, HDAC5, or myogenin can reduce denervation-induced muscle atrophy (16, 17). Another upstream regulator of atrogin and MuRF1 is the FoxO family of transcription factors (FoxO isoforms 1, 3, and 4). Overexpression of FoxOs induces expression of the genes encoding atrogin and MuRF1 and results in muscle atrophy (18, 19). Interference with FoxO function with dominant-negative (DN) forms of FoxO preserves soleus muscle from immobilization-induced atrophy (20) and also spares muscle wasting in cancer cachexia and sepsis (21). However, perhaps due to the difficulty in creating mice that are completely null for FoxO because of the presence of three FoxO isoforms in skeletal muscle, the role of FoxOs in denervation-induced atrophy has not been fully explored genetically.

In contrast to the UPS-mediated protein degradation that plays a role in muscle atrophy, protein synthesis mediated by mTOR (mammalian target of rapamycin) facilitates cell growth and proliferation (2224). mTOR, an evolutionarily conserved serine/threonine protein kinase, forms two distinct complexes: mTOR complex 1 (mTORC1) and mTOR complex 2 (mTORC2). mTORC1, composed of mTOR, raptor, mLST8 (also known as GβL), and PRAS40, is sensitive to rapamycin inhibition (25). mTORC1 exerts its effect on translation by initiating phosphorylation of S6 kinase (p70S6K) and ultimately activating the downstream components involved in protein translation, such as 4EBP1, S6, and eIF4 (22, 26). mTORC2, containing rictor, mSIN1, and mLST8, is insensitive to rapamycin, and disruption of mTORC2 does not cause phenotypic changes in skeletal muscle (27).

mTORC1 activity can be regulated by its components, such as raptor (27, 28), as well as its upstream inhibitor TSC1 (tuberous sclerosis complex 1), the deletion of which increases the activity of mTORC1 (29). mTORC1 activity appears to be essential to muscle growth. Inhibition of mTORC1 activity in normal skeletal muscle through either genetic deletion of raptor (27) or mTOR (30) or overexpression of the mTORC1 inhibitor TSC1 (31) causes muscle atrophy. Genetic deletion of the mTORC1 downstream target S6K also causes muscle atrophy in normal muscle (32, 33). These data indicate that a basal amount of mTORC1-S6K activity is required for normal muscle growth during development.

However, the functional role of mTORC1 in adult denervated muscle remains unclear. It is an important question whether increased protein synthesis through activation of mTORC1 might minimize denervation muscle atrophy, which is a pressing clinical problem. However, mTORC1 activity is, in fact, induced in denervated, atrophying muscles [(34, 35) and this study]. Given our understanding of mTORC1 function in regulating muscle mass in the normal (innervated) situation, the finding of increased mTORC1 activity in atrophying, denervated muscle indicated to us that perhaps this increase represents an adaptive response to denervation that is not sufficient to prevent muscle atrophy. We also, however, considered the alternate hypothesis that mTORC1 serves a different role in the denervation scenario than its role in the maintenance of normal skeletal muscle mass.

We thus set out to explore the functional role of mTORC1 in denervation muscle atrophy both by inducing mTORC1 activity through genetic deletion of the mTORC1 inhibitor TSC1 and by inhibiting mTORC1 activity with rapamycin treatment. Our results indicate that inhibition of mTORC1 activity ameliorates denervation-induced muscle atrophy in fast muscles, and that this effect is mediated through FoxO. After denervation, rapamycin-sensitive mTORC1 activates only a few translation initiation factors, and thus, mTORC1 plays a lesser role in protein synthesis than in degradation after muscle denervation. We further demonstrate that denervation-activated mTORC1 inhibits Akt activity, thereby activating FoxOs and E3 ubiquitin ligases and accelerating denervation muscle atrophy. In addition to this previously unknown role for mTORC1, we show that complete deletion of FoxOs in skeletal muscle induces resistance to denervation muscle atrophy and causes hypertrophy in normal, innervated muscle.


mTORC1-S6K and Akt-FoxO signaling pathways are reciprocally regulated in denervated muscle

We first examined mTORC1 activity 3 days (designated as early) and 15 days (designated as late) after denervation in normal mouse hindlimb. Three days after denervation, Western blot analysis indicated that the phosphorylation of some of the components in the mTORC1 pathway, such as mTOR, S6K, and S6, was moderately increased, indicating a mild activation of mTORC1 pathway, whereas the phosphorylation and activity of Akt and FoxO1 remained unchanged (Fig. 1A, left panels). In muscle, 15 days after denervation, the phosphorylation of mTORC1 and of two of its downstream effectors (S6K and S6) was markedly increased. At this later time point, the phosphorylation of Akt and FoxO1 was reduced compared to that in innervated muscle (Fig. 1A, right panels). The total protein abundance of the mTORC1 components mTOR and raptor was also increased after denervation, suggesting that activation of mTORC1 by denervation could result from both the induction of phosphorylation of mTOR and the increased abundance of mTORC1 components (Fig. 1A). Quantitation confirmed a reciprocal regulatory pattern between the activity of the mTORC1 pathway, represented by the ratio of phosphorylated S6K to total S6K, and the activity of Akt-FoxO cascade, represented by the ratio of phosphorylated Akt to total Akt and phosphorylated FoxO1 to FoxO1 (fig. S1). Furthermore, immunostaining on cryosections of innervated and denervated muscles indicated that mTORC1 activity, as assessed by phosphorylated S6, was increased in skeletal muscle fibers after denervation (Fig. 1B). Additionally, we observed that FoxO1 protein accumulated in nuclei of muscle fibers 15 days after denervation, consistent with a reduction in Akt activity that leads to decreased phosphorylation of FoxO1 and its relocation into myonuclei (Fig. 1C). Together, these data show that denervation activates the mTORC1-S6K pathway, suppresses Akt activity, and activates FoxO.

Fig. 1 Time-dependent regulation of mTORC1-S6K and Akt-FoxO pathways in response to skeletal muscle denervation.

(A) Protein extracts from gastrocnemius muscles 3 and 15 days after denervation were subjected to Western blot analysis. Actin was used as a loading control. Inn, innervated; Den, denervated (n = 2 randomly selected mice per treatment shown; see Materials and Methods). (B) Confocal images of cryosections of innervated and denervated (15 days) tibialis anterior (TA) muscles that were costained with anti-pS6 (red) and the membrane marker wheat germ agglutinin (WGA, green). Scale bars, 50 μm. Images are representative of three mice per treatment. (C) Confocal images of cryosections of innervated and denervated (15 days) tibialis anterior muscle stained with anti-FoxO1 (red), WGA (green), and 4′,6-diamidino-2-phenylindole (DAPI) (blue). Scale bars, 15 μm. Images are representative of three mice per treatment.

Constitutively active mTORC1 sensitizes muscle to, rather than preventing, denervation-induced atrophy

Because mTORC1 activation is thought to increase protein synthesis in most situations, we initially hypothesized that the induction of mTORC1 activity in denervated muscle may be a compensatory mechanism directed toward preventing muscle atrophy. To test this concept, we established a mouse model of increased mTORC1 activity by deleting the mTORC1 inhibitor TSC1 in a muscle-specific manner. Floxed TSC1 mice (TSC1L/L) were bred with MCK-Cre mice (fig. S2), and the resulting TSC1 knockout (KO) mice showed increased mTORC1 activity in skeletal muscle fibers (Fig. 2, A and B). In young adults (6 to 8 weeks of age), the TSC1 KO mice appeared phenotypically normal, without noticeable differences in muscle morphology or mass compared to that of wild-type mice (Fig. 2, A and C). However, TSC1 KO mice showed muscle atrophy at 3 days after denervation, suggesting that increased mTORC1 activity sensitized skeletal muscle to denervation-induced atrophy (Fig. 2D). The expression of atrophy-associated genes, such as those encoding atrogin and MuRF1, was induced to a greater extent by denervation in both TSC1 KO mice and wild-type mice (Fig. 2E). The expression of these atrophic genes was also increased in innervated TSC1 KO muscle compared to their wild-type counterparts, suggesting that mTORC1 activation may play a role in activating the E3 ligase–mediated protein degradation pathway.

Fig. 2 Activated mTORC1 did not prevent denervation-induced muscle atrophy but rather activated FoxO and increased expression of ubiquitin E3 ligase genes.

(A) Activation of the mTORC1 pathway in gastrocnemius muscle of TSC1 KO mice as determined by phosphorylation of S6K protein (n = 3 mice per genotype and treatment). (B) Cross sections of tibialis anterior muscle stained by WGA or for phosphorylated S6. Scale bars, 50 μm. Images are representative of three mice per genotype. (C) Muscle weight of wild-type (WT) and TSC1 KO mice at 8 weeks of age. Gastroc, gastrocnemius (n = 4 mice per genotype). (D) Whole muscles (tibialis anterior and gastrocnemius) were weighed after 3 and 15 days of denervation and normalized to their contralateral, innervated counterparts (n = 4 mice per genotype, *P < 0.05). (E) mRNA expression of E3 ubiquitin ligases in innervated and denervated gastrocnemius muscles from WT and TSC1 KO mice. Results are normalized to γ-actin expression (n = 4 mice per genotype, *P < 0.05). (F) Total protein extracts from gastrocnemius muscles were immunoblotted with anti-ubiquitin (Ub) antibody. Actin was used as a loading control. Gray density was plotted as a continuous function on the right panel, and mean density of detectable ubiquitinated proteins is shown in the lower panel (n = 3 mice per genotype, *P < 0.05). a.u., arbitrary units. (G) Protein extracts from gastrocnemius muscles were immunoblotted with the indicated antibodies. n = 3 mice per genotype. (H) mRNA expression of HDAC4 and myogenin (Mgn) was normalized to that of γ-actin (n = 3 mice per genotype). (I) Cryosections of tibialis anterior muscle from WT and TSC1 KO mice were stained with anti-FoxO1 (red), DAPI (blue), and WGA (green), and the images were merged to show the myonuclear localization of FoxO1. Scale bars, 15 μm. Images are representative of three mice per genotype.

Constitutively active mTORC1 is sufficient to activate FoxO and the ubiquitin-proteasome system by suppressing Akt in skeletal muscle

Consistent with the induction of E3 ubiquitin ligases (Fig. 2E), the extent of protein ubiquitination was also significantly induced in the TSC1 KO skeletal muscle, ~150% of the total protein ubiquitination in wild-type skeletal muscle (Fig. 2F). TSC1 KO muscle showed increased phosphorylation of S6 but reduced phosphorylation of Akt and FoxO1 (Fig. 2G), similar to the phosphorylation patterns in denervated muscle (Fig. 1A). Like FoxO, the HDAC4-myogenin pathway can increase expression of genes encoding E3 ubiquitin ligases (16, 17), but we did not find any changes in the expression of the genes encoding myogenin or HDAC4 in TSC1 KO muscle (Fig. 2H). By immunostaining, we found that FoxO1 protein accumulated in TSC1 KO muscle and colocalized with nuclei, suggesting that FoxO1 is translocated into myonuclei with mTORC1 activation (Fig. 2I). Therefore, increased mTORC1 activity stimulates a denervation-like signaling cascade that accelerates muscle atrophy in response to denervation, although it does not induce muscle atrophy in normal innervated muscle at a young age.

Lack of FoxOs induces muscle fiber hypertrophy and resistance to denervation-induced atrophy

Because FoxO1 is activated in denervated muscle as well as in muscle with increased mTORC1 activity, and because FoxOs transcriptionally activate the genes encoding the E3 ubiquitin ligases atrogin and MuRF1, FoxOs would appear to be likely downstream mediators of mTORC1-induced atrophy in denervated muscle. We generated FoxO KO mice to determine whether FoxOs are required for denervation-induced muscle atrophy. Because FoxO1, FoxO3, and FoxO4 are present in skeletal muscle and may have overlapping functions, we simultaneously deleted all three genes in a muscle-specific manner by breeding floxed FoxO1/3/4 mice with MCK-Cre mice (fig. S3A). We confirmed the successful deletion of all three FoxO isoforms (fig. S3, B and C). Adult (8 weeks) homozygous FoxO1/3/4 triple KO mice (FoxO TKO) appeared phenotypically normal but had significantly higher body weight compared to wild-type mice or the single or double FoxO KO mice (Fig. 3A). The weights of the tibialis anterior, gastrocnemius, and soleus muscles were also all significantly increased in the FoxO TKO mice (Fig. 3B). The cross-sectional area of tibialis anterior muscle fibers in FoxO TKO mice was increased to 180% of wild-type mice (Fig. 3C). This increased muscle fiber size may result from reduced protein degradation because protein ubiquitination was significantly reduced in FoxO TKO muscles (Fig. 3D).

Fig. 3 Triple deletion of FoxO1, FoxO3, and FoxO4 resulted in muscle hypertrophy and reduced protein degradation and mitigated denervation muscle atrophy.

(A) Tibialis anterior, soleus (Sol), gastrocnemius muscle (Gas.), and whole body were weighed in WT mice, FoxO TKO mice, and breeding intermediates. The results for WT mice are indicated by the horizontal, dotted line at 100% (n = 4 mice per genotype, *P < 0.05). (B and C) Muscle fiber size in WT and FoxO TKO muscle. (B) Distribution of the cross-sectional area (CSA) of at least 600 tibialis anterior muscle fibers per specimen. Graph is a representative of n = 4 mice per genotype. (C) Mean cross-sectional area (n = 4 mice per genotype, *P < 0.05). (D) Western blot analysis for protein ubiquitination; Ponceau S–stained total protein was used to show equal loading and to normalize gray density (representative image is shown of n = 3 mice per genotype, *P < 0.05). (E) Time-dependent changes in gastrocnemius muscle weight after denervation in WT and FoxO TKO mice (n = 4 mice per genotype and treatment, *P < 0.05). The weight of the denervated (D) muscle was normalized to its contralateral innervated (I) muscle, and results were expressed as percentages (D/I). (F) Muscle weight loss in tibialis anterior, soleus, and gastrocnemius muscle 15 days after denervation in WT and FoxO TKO mice. The values for contralateral, innervated muscles were set at 100% and indicated by the horizontal line (n = 4 mice per genotype and treatment, *P < 0.05). (G) Cryosections of tibialis anterior muscles were stained with WGA. Representative images are shown. (H) Quantitation of cross-sectional area of 800 to 1000 fibers in each tibialis anterior muscle (n = 4 mice per genotype and treatment, *P < 0.05). (I) Protein extracts from muscles 15 days after denervation and the contralateral innervated controls were immunoblotted with anti-ubiquitin. Ponceau-S staining of total protein is shown as a loading control. n = 2 mice per genotype and treatment. (J) Mean fold changes in the amount of ubiquitinated proteins (total ubiquitinated proteins normalized to total protein stained with Ponceau S) (n = 2 mice per genotype and treatment). (K and L) mRNA expression of regulatory factors, normalized to that of γ-actin, in gastrocnemius muscles 3 days (K) or 15 days (L) after denervation. Fold changes were calculated by normalizing mRNA in denervated muscles to that in the contralateral controls (n = 4 mice per genotype and treatment, *P < 0.05).

To delineate the role of FoxOs in denervation-induced muscle atrophy, we examined denervation-dependent changes in the FoxO TKO mice. In both wild-type and FoxO TKO mice, denervation resulted in a time-dependent loss of muscle weight, with no change at 3 days but a significant reduction at 15 days after injury. However, denervation-induced atrophy was attenuated 33% in FoxO TKO mice compared to wild-type controls (Fig. 3E). Muscle mass was preserved to about the same degree in the tibialis anterior, gastrocnemius, and soleus muscles (Fig. 3F). Reductions in fiber cross-sectional area were similar to those in muscle weight (Fig. 3, G and H). Although the cross-sectional area of TKO muscle was significantly reduced after denervation, the cross-sectional area in that group was similar to that in the innervated wild-type muscle because normally innervated FoxO TKO muscle was hypertrophic.

Denervation-induced protein ubiquitination was significantly reduced in FoxO TKO mice both in total and when normalized to total protein (Fig. 3, I and J). Denervation-induced expression of the genes encoding the E3 ubiquitin ligases atrogin and MuRF1 in FoxO TKO mice was similar to that in wild-type mice 3 days after denervation and was accompanied by a robust induction of the mRNA encoding myogenin at that time point (Fig. 3K). However, the denervation-induced increase in mRNA abundance of atrogin and MuRF1 at 15 days after denervation did not occur in FoxO TKO mice (Fig. 3L). At this time point, the expression of the mRNAs encoding myogenin and HDAC4 was still significantly increased compared to pre-denervation amounts, but about 50% of that in wild-type denervated muscle. This suggests that FoxOs transcriptionally activate genes encoding E3 ubiquitin ligases only at later times after denervation, and that they could affect the HDAC-myogenin pathway to some degree, directly or indirectly.

Inhibition of mTORC1 with rapamycin ameliorates denervation-induced muscle atrophy

Because mTORC1 activation stimulates pathways similar to those activated by denervation, we suspected that denervation-induced mTORC1 activity may directly contribute to muscle atrophy by suppressing Akt and activating FoxO. To test this idea, we treated mice with denervated hindlimbs with rapamycin to inhibit mTORC1 activity. Rapamycin treatment at the doses used did not influence body weight or innervated muscle weight (table S1) but significantly prevented both the reduction in cross-sectional area (Fig. 4A) and the muscle weight loss (Fig. 4B) that occurred after denervation. The denervation-induced expression of the genes encoding the E3 ubiquitin ligases atrogin and MuRF1 (Fig. 4C) and protein polyubiquitination (Fig. 4D) was also significantly reduced by rapamycin treatment. However, denervation-induced expression of the gene encoding myogenin was not affected by rapamycin (Fig. 4E). This further supports the notion that myogenin was not involved in mTORC1-dependent regulation of muscle mass at the late stage of denervation. In contrast, rapamycin treatment rescued the denervation-suppressed phosphorylation of Akt (at Thr308) and FoxO1 (at Ser256) as well as inhibited the denervation-induced phosphorylation of S6 (Fig. 4F), suggesting that mTORC1 activity serves as an upstream regulator of Akt and FoxO in denervated muscle. These observations suggest that the inhibitory effect of rapamycin on denervation-induced atrophy is mediated by the normalization of the mTORC1-Akt-FoxO axis.

Fig. 4 Inhibition of mTORC1 activity by rapamycin mitigates denervation muscle atrophy via regulating FoxO.

(A and B) Inhibition of mTORC1 activity with rapamycin (Rapa) treatment partially rescues fiber size (A) and weight (B) in muscle 15 days after denervation. (A) Cross-sectional area of muscle fibers stained with WGA. The innervated, contralateral control was set at 100% and indicated by the horizontal, dotted line. At least 600 fibers per muscle were used (n = 6 mice, *P < 0.05). Representative images are shown. Scale bars, 50 μm. (B) Denervated muscle weight (tibialis anterior and gastrocnemius) was normalized to the contralateral innervated muscles, which was set at 100% and indicated by the horizontal, dotted line (n = 6 mice per genotype and treatment, *P < 0.05). (C) mRNA expression of atrogin and MuRF1 was normalized to γ-actin. Fold changes were calculated as the ratio of mRNA in denervated muscles to that in their contralateral controls (n = 6 mice per genotype and treatment, *P < 0.05). (D) Total protein extracts from gastrocnemius muscles were immunoblotted with antibodies against ubiquitin and actin. Actin was used as a loading control. n = 2 mice per genotype and treatment. (E) Myogenin mRNA abundance was measured. n = 6 mice per genotype and treatment, *P < 0.05. (F) Western blot analysis was performed on protein extracts from gastrocnemius muscles to measure the phosphorylation and total abundance of the indicated proteins. Equal loading was shown by Ponceau-S staining and actin. Phosphorylation of a protein was normalized to the total abundance of the protein, and fold changes are shown (right). n = 2 mice per genotype and treatment. (G) FoxO is required for the protective effect of rapamycin. Control or FoxO-DN constructs and a GFP-expressing reporter gene were electroporated into denervated tibialis anterior muscles. Rapamycin was administered, and 15 days after denervation, the cross-sectional areas of the GFP+ fibers (>80) in tibialis anterior muscles were measured and normalized to the contralateral innervated control. Representative images are shown. Scale bars, 50 μm. n = 3 mice per genotype and treatment. *P < 0.05. ns, no significant difference.

We also observed that the phosphorylation of IRS1 (insulin receptor substrate 1) was increased in denervated muscle and was inhibited by rapamycin treatment (Fig. 4F). This is consistent with the observation that activation of the mTORC1 pathway leads to phosphorylation of IRS1 (in mouse, Ser302 and Ser307), which leads to the inhibition of the phosphatidylinositol 3-kinase (PI3K)–Akt pathway (3639). Thus, phosphorylated IRS1 may mediate the crosstalk between the mTORC1-S6K pathway and the Akt-FoxO pathway and is linked to the reduction of phosphorylated Akt (Thr308). In addition, the phosphorylation of Ser473 in Akt (phosphorylated Akt-Ser473 over total Akt) was also reduced after denervation, but rapamycin treatment only showed a mild rescue effect on this reduction, compared to that of Thr308 (phosphorylated Akt-Thr308 over total Akt) (Fig. 4F). Although mTORC2 phosphorylates Akt at Ser473, the abundance of the mTORC2 components mLST8 and rictor was either increased or unchanged by denervation (Fig. 4F). These results suggest that mTORC2 is not responsible for the reduced phosphorylation of Ser473 in Akt in denervated muscle. Together, denervation represses the phosphorylation of Akt at Thr308 and Ser473, and rapamycin treatment appears to rescue the phosphorylation of Akt mainly at Thr308 in denervated muscle.

Inhibition of atrophy by rapamycin in denervated muscle requires FoxO

Although rapamycin treatment restored the phosphorylation of Akt and FoxO and reduced post-denervation-induced atrophy, it was important to determine whether FoxO was required. Our FoxO TKO model was not ideal to investigate this issue, because pre-denervation muscle hypertrophy in FoxO TKO mice could confound potential protective effects of rapamycin. Thus, we generated a DN form of FoxO that lacked the transactivation domains of FoxO1 but retained the DNA binding domain (DBD) and a nuclear localization signal (NLS) (fig. S4A). In cultured cells, we found that DN FoxO reversed the FoxO-induced activity of a FoxO reporter gene (fig. S4B). In addition, DN FoxO significantly reduced atrogin promoter activity in cultured myotubes (fig. S4C). We then electroporated the denervated tibialis anterior muscles with control or FoxO-DN plasmids and a green fluorescent protein (GFP) reporter plasmid in the presence or absence of rapamycin. Measurement of the cross-sectional area of the GFP-positive fibers demonstrated that the DN form of FoxO protected against denervation-induced muscle atrophy (Fig. 4G). Thus, blockage of FoxO after muscle denervation was effective in reducing muscle atrophy, indicating that pre-denervation induction of muscle hypertrophy was not required. Furthermore, although rapamycin treatment prevented the reduction in cross-sectional area after denervation in fibers electroporated with the control plasmid, rapamycin did not have an additional effect on fibers electroporated with the DN form of FoxO. This result suggests that the protective effect of rapamycin on denervated muscle is mediated through regulation of FoxO (Fig. 4G).

Rapamycin-sensitive mTORC1 plays a lesser role in regulating protein synthesis in denervated muscle

We showed that activated mTORC1 induced protein degradation in denervated muscle, but it is unlikely that this is the sole function of activated mTORC1 in denervated muscle. mTORC1 induces protein synthesis by promoting the protein phosphorylation of regulators of eukaryotic translation initiation, including members of the 4EBP and eIF4 families. In contrast to the unchanged total protein abundance of 4EBP1, phosphorylation of several sites on 4EBP1 was increased, including Thr37, Thr46, and Ser65 (Fig. 5A), suggesting that denervation could activate protein synthesis by triggering the phosphorylation of 4EBP1. Phosphorylated 4EBP1 dissociates from eIF4E and activates translation initiation (40). However, none of these phosphorylation events were sensitive to rapamycin treatment. Therefore, rapamycin-sensitive mTORC1 did not regulate the phosphorylation of 4EBP1 in denervated muscle. Similarly, phosphorylation of eIF4E and total protein abundance of eIF4H were increased in a rapamycin-insensitive manner in denervated muscle. In contrast, the increased phosphorylation of eIF4B at Ser422 by denervation was reversed by rapamycin treatment (Fig. 5A). Denervation also resulted in the suppression of phosphorylation of eIF4G, which was also partially recovered by rapamycin treatment (Fig. 5A). Phosphorylation of eIF4B at Ser422 leads to enhanced interaction with eIF3, promoting protein translation (41), and phosphorylation of eIF4G inhibits cap-dependent protein translation (42). Therefore, decreased phosphorylation of eIF4G and increased phosphorylation of eIF4B would be predicted to favor protein synthesis in denervated muscle. The normalization of these denervation-induced changes by rapamycin would compromise protein synthesis. However, our finding that other members of the eIF4 family, as well as 4EBP1, show increased abundance after denervation, in a rapamycin-insensitive manner, suggests that rapamycin has only a limited effect in compromising protein synthesis in denervated muscle. This small atrophy-promoting effect of rapamycin appears to be overwhelmed by the atrophy-preventing effects delineated above.

Fig. 5 Rapamycin-sensitive mTORC1 plays a lesser role in regulating protein synthesis in denervated muscle.

(A) Protein lysates from innervated and denervated gastrocnemius muscles from mice that received or did not receive rapamycin treatment were immunoblotted for various translation initiation factors of the 4EBP and eIF4 families (n = 2 randomly selected mice per treatment shown). (B) Model for the temporal regulation of signaling cascades in neurogenic muscle atrophy. Denervation rapidly activates HDAC4, HDAC5, Dach2, and myogenin, which appears to trigger the early-phase induction of E3 ubiquitin ligases. Denervation later induces activation of FoxO through mTORC1-dependent inhibition of Akt, which appears to be responsible for the induction of E3 ubiquitin ligases at the later phase. Protein degradation induced by these early and late cascades after denervation is apparently more robust than protein synthesis, the net result being muscle atrophy.

Phosphatidic acid and phospholipase D are unlikely to contribute to the activation of mTORC1 in denervated muscle

The activation of mTORC1 in denervated muscle appears to be Akt-independent, and the underlying mechanism remains to be elucidated. Our data indicate that both increased abundance of mTORC1 components and activation of upstream signaling molecules of mTORC1 may be involved. It has been reported that the mTORC1 activators phospholipase D (PLD) and phosphatidic acid (PA) are up-regulated in skeletal muscle by mechanical stimuli (43) and that they regulate muscle fiber size (44). We thus examined the abundance and activity of PA and PLD, and found that there was no substantial increase in abundance or activity after denervation (figs. S5 and S6). Therefore, the mTORC1 activators PLD and PA are unlikely to contribute to the Akt-independent mTORC1 activation in denervated muscle.

Rapamycin treatment does not prevent the atrophy of denervated soleus muscle

In contrast to fast muscles, loss of slow soleus muscle in denervation was not rescued by rapamycin treatment (fig. S7A). Denervation increased the phosphorylation of S6 to a much lesser extent in soleus than in gastrocnemius. Rapamycin treatment did not substantially alter the phosphorylation or total protein amounts of Akt and FoxO in denervated soleus muscle (fig. S7B). Therefore, it appears that mTORC1 functions differently in the soleus than in other muscles. Indeed, in TSC1 KO mice, the soleus muscle suffers less atrophy after denervation (45) and undergoes age-dependent hypertrophy, in contrast to the atrophic changes in fast TSC1 KO muscles in these situations (46). The distinct regulation of mTORC1 activity in soleus muscle is worthy of further investigation.


Motor neuron disease or injury may result in denervation muscle atrophy, leading to disability and even mortality. Understanding the molecular regulatory mechanisms underlying this process may lead to novel therapeutic interventions. Here, we used a mouse denervation model to demonstrate that mTORC1, which promotes cell growth and anabolic metabolism in most circumstances, surprisingly plays a central role in effecting muscle atrophy in denervation. Denervation-induced mTORC1 activated E3 ubiquitin ligases by altering the Akt-FoxO cascade. Inhibition of either mTORC1 or FoxO mitigated denervation-induced muscle atrophy. Our findings imply that rapamycin and similar agents have therapeutic potential in muscle-wasting disorders.

Increased protein degradation is mediated by Akt-independent activation of mTORC1 in denervated muscle

mTORC1 promotes protein translation by controlling the activity of initiation factors (40, 47). mTORC1 activity is required for muscle growth (27, 30, 31) and thus is generally linked to gains in muscle mass. However, mTORC1 activity decreases with postnatal motor innervation (48). In adult mice, we and others have observed that rapamycin treatment does not influence body and muscle weight (4951), indicating that mTORC1 activity in adult muscle is not as important as it is during muscle development. We showed that a reinduction of mTORC1 activity after denervation promotes muscle atrophy. Denervation-induced mTORC1 is Akt-independent, in contrast to Akt-dependent activation of mTORC1 by insulin-like growth factor or constitutively active Akt. Akt-dependent activation of mTORC1 prevents denervation muscle atrophy (5255) because activated Akt simultaneously inhibits FoxO-E3 ligases while activating mTORC1. Thus, mTORC1 promotes muscle atrophy only when Akt is inactive, such as during denervation. This dual function of mTORC1 in skeletal muscle is similar to that of myogenin, which is also required for muscle development but triggers denervation-induced muscle atrophy (16, 17, 5658).

The HDAC4-myogenin and mTORC1-FoxO pathways coordinate denervation-induced muscle atrophy

The molecular network consisting of HDAC4, Dach2, and myogenin regulates gene expression in denervated muscle (14, 15) and promotes denervation-induced atrophy by inducing the expression of genes encoding E3 ubiquitin ligases (16, 17). Myogenin binds to an E box in the promoter sequence of these E3 ligases (16), adjacent to the FoxO DNA binding element (18). Both of these cis elements are evolutionarily conserved (table S2). Because the amount of myogenin binding to the E box decreases between 3 and 7 days after denervation (16), and mTORC1-FoxO is activated 3 days after denervation, it seems that there might be a temporal relay between the two signaling pathways that mediate denervation-induced muscle atrophy—HDAC4-myogenin is activated early, and mTORC1-FoxOs is activated late in this process (Fig. 5B). Consistently, deletion of FoxOs had no effect on the induction of E3 ubiquitin ligases until the later post-denervation time point. If and how these pathways—HDAC4-myogenin and mTORC1-FoxO—regulate and interact with one another remains to be elucidated. In addition, both activation of mTORC1 (Fig. 2) and overexpression of myogenin (16) in normal innervated muscle can induce E3 ubiquitin ligases, but neither of these manipulations alone results in muscle atrophy. Therefore, coordinated activation of both of these signaling cascades might be necessary for the development of muscle atrophy.

The mTORC1-mediated activation of FoxO appears to be through the S6K-IRS-Akt feedback route (3739), which reportedly causes insulin resistance. We demonstrated that this feedback pathway results in muscle atrophy in an animal model of denervation. Future work may involve testing whether other feedback routes, such as Grb10 (59, 60), also mediate the mTORC1-initiated suppression of Akt in denervated muscle.

Last, although FoxOs regulate E3 ubiquitin ligases, and blockage of FoxO can relieve atrophy in targeted muscle fibers during disuse and sepsis (20, 21), it has been unclear whether and when FoxOs mediate denervation-induced muscle atrophy. Here, we used both genetic deletion of FoxO genes and temporary inhibition of FoxO with a dominant-negative form to uncover the functional roles of FoxOs in normal and denervated muscles. Due to the potential functional overlap of the isoforms, as well as the lethality of FoxO1 deletion in all tissues (61), we created muscle-specific deletion of FoxO1, FoxO3, and FoxO4 together. The data from these mice show that FoxO is critical to protein turnover in both innervated and denervated muscle. Blocking FoxO activity is thus a plausible approach to induce muscle hypertrophy and prevent neurogenic muscle atrophy. Although FoxOs can activate E3 ubiquitin ligases and cause muscle atrophy, we demonstrate that FoxOs do not play a role in the early induction of E3 ubiquitin ligases after denervation. Rather, they are only responsible for the late induction of E3 ubiquitin ligases after denervation.

Inhibition of mTORC1 or FoxO activity may have therapeutic benefit in neuromuscular disorders

Our study uncovers an mTORC1-FoxO-E3 ubiquitin ligase cascade that partially mediates denervation muscle atrophy. In adult muscle, a moderate dose of rapamycin helped maintain skeletal muscle mass after denervation without affecting normal muscle. This suggests that rapamycin therapy may be beneficial in motor neuron diseases and nerve injury. Our data also provide evidence supporting the concept that blockage of FoxOs represents another promising approach to relieve the muscle atrophy associated with neuromuscular disorders.



The animal care and experimental procedures followed the protocol approved by University of Michigan Committee on Use and Care of Animals and VA Palo Alto Healthcare System Institutional Animal Care and Use Committee. NSA (CF-1) mice were purchased from Harlan. The generation of TSC1 lox/lox mice with exons 17 and 18 of Tsc1 flanked by loxP sites by homologous recombination has been described (62, 63). We generated muscle-specific TSC1 KO mice through breeding TSC1 lox/lox mice (C57BL/6J) with the hemizygote MCK-Cre mouse line (FVB, Cre+/−; gift from D. Alessi, University of Dundee), in which Cre is under the control of muscle-specific creatine kinase. Briefly, the TSC lox/lox mice were bred with MCK Cre/+ to obtain a TSC lox/+, Cre/+ line, and then the latter were crossed with TSC1 lox/lox mice to generate a TSC1 lox/lox, Cre/+ line. The F3 generation was acquired by breeding TSC1 lox/lox with TSC lox/lox, Cre/+ line. The offspring were genotyped with Cre primers. The Cre+ mice were TSC1 KO mice, whereas the Cre− mice were used as wild-type control of the same genetic background. Genotyping was performed as described previously (29).

The generation of FoxO1/3/4 lox/lox mice (FVB) with exon 2 (FoxO1, on chromosome 3), exon 3 (FoxO3, on chromosome 10), and exon 1 (FoxO4, on chromosome X) flanked by loxP sites by homologous recombination was generated in Harvard Medical School and reported before (37). FoxO1/3/4 lox/lox female mice were bred with MCK-Cre (FVB, The Jackson Laboratory) male mice to generate F1 offspring with a genotype of FoxO1/3/4 lox/−, Cre+/− (~50%). Male mice of FoxO1/3/4 lox/−, Cre+/− were then bred with female FoxO1/3/4 lox/lox to generate F2 offspring with a genotype of FoxO1−/−, 3−/−, 4−/− (~12.5%), as well as other intermediate offspring with genotypes FoxO1+/−, 3−/−, 4−/− (~12.5%), FoxO1−/−, FoxO3+/−, FoxO4−/− (~12.5%), FoxO1+/−, FoxO3+/−, FoxO4−/− (~12.5%), and 50% of Cre− with various combinations of floxed FoxO1/3/4. These Cre+, FoxO1/3/4 floxed (FoxO1−/−, 3−/−, 4−/−) mice were further bred with FoxO1/3/4 floxed mice to generate Cre+ (~50%, FoxO TKO) and Cre− (~50%, control) FoxO1/3/4 floxed mice. The Cre− mice were used as wild-type control of the same genetic background. Genotyping primers were described before (64). The primers are listed in table S3.

Sciatic nerve injury and in vivo rapamycin treatment of mice

Sciatic nerve injury was performed in an animal operating room with aseptic techniques as described previously (14). Briefly, adult mice (~8 weeks) were anesthetized with ketamine (100 mg/kg) and xylazine (10 mg/kg). After removing hair, the local skin was disinfected with betadine and 75% ethanol. A small incision was made at mid-thigh level, and about 0.5 cm of the nerve trunk was removed while the animal was under deep anesthesia. Animals were under intensive care until sample collection.

Rapamycin (LC Laboratories) was initially dissolved in 100% ethanol, stored at −20°C, and further diluted in an aqueous solution of 5.2% Tween 80 and 5.2% polyethylene glycol 400 (final ethanol concentration, 2%) immediately before use (29). Denervated NSA mice (male, ~40 g) were injected immediately with rapamycin intraperitoneally at 1.5 mg/kg, or 6 mg/kg body weight, every other day. Control mice were injected with vehicle with the same amount of ethanol solvent with rapamycin group (6 mg/kg). Muscle samples were collected 15 days later. Muscle wet weight was recorded immediately. For immunostaining, muscles were frozen in optimum cutting temperature compound in dry ice-cold isopentane. For RNA and protein extraction, muscles were quickly frozen in liquid nitrogen. Student’s t test was used to evaluate the statistical significance. P < 0.05 was considered as statistically significant.

Each of these experiments was performed such that there were three to six separate experimental mice (for each genotype, dose, and time point) and three to six separate control mice per experiment. The blots in many of the figures show results from two randomly selected mice per experimental condition. When these two randomly selected specimens showed clear results, samples were not run on gels, nor additional specimens were analyzed.

Western blotting and real-time polymerase chain reaction

Protein abundance and protein phosphorylation were detected by Western blotting analysis following standard procedures. Antibodies against TSC1, FoxO1, phosphorylated FoxO1, S6, phosphorylated S6, Akt, phosphorylated Akt, and ubiquitin were purchased from Cell Signaling, and anti-actin was from Santa Cruz Biotechnology.

Gene expression was detected by real-time polymerase chain reaction (PCR). RNA (1 μg) was reverse-transcribed with oligo(dT) primer and SuperScript II (Invitrogen), and 1/20 of the complementary DNA (cDNA) mixture served as template for PCRs. Real-time PCR was performed on an ABI 7900HT with SYBR Green Master Mix (Thermo Fisher Scientific). The primers are listed in table S3.

Muscle immunostaining and cross-sectional area measurement

Freshly frozen muscle samples were sectioned on cryostat with 14-μm thickness. Standard procedure was used to perform the immunostaining against FoxO1. DAPI and WGA–Alexa Fluor 488 were purchased from Invitrogen. Slides were mounted in ProLong Gold Antifade Reagent (Invitrogen) and imaged by confocal fluorescence microscopy (LSM710, Zeiss). Muscle cross-sectional areas were measured on WGA-stained muscle cross sections by a technician blinded to experimental group. Three randomly selected regions from each section (~200 to 300 cells per image) and three consecutive sections were processed. Cross-sectional area was measured with Fiji, an enhanced version of ImageJ ( Student’s t test was used to evaluate statistical significance. P < 0.05 was considered as statistically significant.

Gene electrotransfer to skeletal muscle

Gene electrotransfer was performed as previously described (14). Briefly, 20 μg of control plasmid pcDNA3.1 or pcDNA3.1-FoxO-DN, together with a 2- μg pCS2-GFP plasmid, was electroporated by injection into tibialis anterior muscles with a Hamilton syringe, followed by eight electrical pulses with a duration of 60 ms with 100-ms interval with an ECM 830 electroporator (BTX). The electrical field intensity was 150 V/cm. The FoxO-DN construct was generated by cloning a cDNA fragment flanking the DBD and NLS into Topo-pcDNA3.1 vector. PCR primers used for the cDNA amplification are as follows: forward, gccgccgctgggccgctcgcggg; reverse, tcaggcagctcggcttcggctcttagcaaa.

PLD activity and PA concentration assays

PLD and PA assay kits were purchased from Cayman Chemical. The assay was performed according to the supplier’s instructions. Protein input was 12 μg per well. Samples were loaded onto a 96-well plate, and the fluorescence was read by a SpectraMax M2 (Molecular Devices), with excitation wavelength of 530 to 540 nm and emission wavelength of 585 to 595 nm.


Fig. S1. The temporal pattern of the phosphorylation of S6K, Akt, and FoxO1 after muscle denervation.

Fig. S2. Establishment of a muscle-specific mouse KO of the TSC1 gene.

Fig. S3. Establishment of mice with a muscle-specific TKO of FoxO1, FoxO3, and FoxO4.

Fig. S4. Establishment of a FoxO-DN construct to block FoxO activity.

Fig. S5. PLD enzymatic activity in innervated and denervated muscle.

Fig. S6. PA concentration in innervated and denervated muscle.

Fig. S7. The regulatory profile and functional role of mTORC1 in soleus muscle after denervation.

Table S1. The muscle (innervated) weight and body weight were not suppressed by rapamycin treatment.

Table S2. The evolutionarily conserved regulatory elements, forkhead and E box, in the promoter of the gene encoding the E3 ubiquitin ligase atrogin.

Table S3. The list of primers used in the current study.


Acknowledgments: We thank A. Olson (Stanford Neuroscience Microscopy Service, supported by NIH NS069375) for help with muscle cross-sectional area measurement, T. A. Rando (Stanford University) for helpful advice, A. Brunet (Stanford University) for the 6DBE construct, S. E. Alway (University of West Virginia) for advice on cross-sectional area measurement, and R. A. Miller and S. V. Brooks (University of Michigan) for personal communications. Funding: This study is supported by a Veterans Affairs merit review grant (J.B.S.), a Stanford SPARK Program and Spectrum Clinical and Translational Science Award grant (J.B.S. and H.T.), NIH grant DK083491 (K.I.), and NIH grants (K.-L.G. and D.G.). Author contributions: H.T., M.L., K.I., E.W., Andy Khuong, Amanda Khuong, S.S., and M.G. performed the experiments and collected the data. H.T., M.L., K.I., J.P., R.A.D., D.G., K.-L.G., and J.B.S. analyzed and discussed the data. J.P. and R.A.D. generated the FoxO1/3/4 floxed mice. H.T. and J.B.S. wrote the paper. Competing interests: H.T. and J.B.S. have filed a provisional patent entitled “mTOR and FOXO inhibition to prevent denervation atrophy of skeletal muscle” through the Stanford University Office of Technology Licensing. The other authors declare that they have no competing interests.
View Abstract

Stay Connected to Science Signaling

Navigate This Article