Metformin induces muscle atrophy by transcriptional regulation of myostatin via HDAC6 and FoxO3a

Abstract

Background: Skeletal muscle atrophy is a severe condition that involves loss of muscle mass and quality. Drug intake can also cause muscle atrophy. Biguanide metformin is the first-line and most widely prescribed anti-diabetic drug for patients with type 2 diabetes. The molecular mechanism of metformin in muscle is unclear.

Methods: Myostatin expression was investigated at the protein and transcript levels after metformin administration. To investigate the pathways associated with myostatin signalling, we used real-time polymerase chain reaction, immunoblotting, luciferase assay, chromatin immunoprecipitation assay, co-immunoprecipitation, immunofluorescence, primary culture, and confocal microscopy. Serum analysis, physical performance, and immunohistochemistry were performed using our in vivo model.

Results: Metformin induced the expression of myostatin, a key molecule that regulates muscle volume and triggers the phosphorylation of AMPK. AMPK alpha2 knockdown in the background of metformin treatment reduced the myostatin expression of C2C12 myotubes (-49.86 ± 12.03%, P < 0.01) and resulted in increased myotube diameter compared with metformin (+46.62 ± 0.88%, P < 0.001). Metformin induced the interaction between AMPK and FoxO3a, a key transcription factor of myostatin. Metformin also altered the histone deacetylase activity in muscle cells (>3.12-fold ± 0.13, P < 0.001). The interaction between HDAC6 and FoxO3a induced after metformin treatment. Confocal microscopy revealed that metformin increased the nuclear localization of FoxO3a (>3.3-fold, P < 0.001). Chromatin immunoprecipitation revealed that metformin induced the binding of FoxO3a to the myostatin promoter. The transcript-level expression of myostatin was higher in the gastrocnemius (GC) muscles of metformin-treated wild-type (WT) (+68.9 ± 10.01%, P < 0.001) and db/db mice (+55.84 ± 6.62%, P < 0.001) than that in the GC of controls (n = 4 per group). Average fibre cross-sectional area data also showed that the metformin-treated C57BL/6J (WT) (-31.74 ± 0.75%, P < 0.001) and C57BLKS/J-db/db (-18.11 ± 0.94%, P < 0.001) mice had decreased fibre size of GC compared to the controls. The serum myoglobin level was significantly decreased in metformin-treated WT mice (-66.6 ± 9.03%, P < 0.01).

Conclusions: Our results demonstrate that metformin treatment impairs muscle function through the regulation of myostatin in skeletal muscle cells via AMPK-FoxO3a-HDAC6 axis. The muscle-wasting effect of metformin is more evident in WT than in db/db mice, indicating that more complicated mechanisms may be involved in metformin-mediated muscular dysfunction.

Keywords: AMPK; FoxO3a; HDAC6; Metformin; Muscle atrophy; Myostatin.

Figure 1

Metformin up‐regulates myostatin expression. (A) Comparison of the relative mRNA expression of myostatin using real‐time PCR (qRT‐PCR). C2C12 myotubes were stimulated with metformin (2 mM) for 12 h. (B) C2C12 myotubes were stimulated with metformin (2 mM) for the indicated times. PCR was normalized using GAPDH expression. Bars represent the mean ± SEM. (C) The expression of myostatin was examined by western blotting (WB). C2C12 myotubes were stimulated for 36 h with indicated concentrations of metformin (0.01–2 mM). Cell lysates were analysed using anti‐myostatin and anti‐β‐actin antibodies. (D) C2C12 myotubes were incubated with metformin (2 mM) for the indicated times. Cell lysates were analysed using anti‐myostatin and anti‐β‐actin antibodies. (E) The expression of myostatin was evaluated by WB. C2C12 myotubes were stimulated with the biguanides, metformin (2 mM), buformin (50 $\mathrm{\mu M}$), and phenformin (50  $\mathrm{\mu M}$) for 24 h. Cell lysates were analysed using anti‐myostatin and anti‐β‐actin antibodies. (F) Myotube morphology was examined via haematoxylin and eosin (H&E) staining after metformin treatment (0–2 mM). (G) Analysis of the diameter of cultured myotubes. Myotube diameter was calculated using ImageJ. Bars represent the mean ± SEM. (H, I) Primary TA myotubes were stimulated with metformin (2 mM) for 24 h. (J) C2C12 myotubes were pre‐treated with Myostatin siRNA (100 nM) and incubated with metformin (2 mM). Myotube morphology was examined via haematoxylin and eosin (H&E) staining. (K) Analysis of the diameter of cultured myotubes. Myotube diameter was calculated using ImageJ. (L, M) Comparison of the relative mRNA expression of MuRF1 and MAFbx32 using real‐time PCR (qRT‐PCR). C2C12 myotubes were stimulated with metformin (2 mM) for the indicated times. PCR was normalized using GAPDH expression. (N) Comparison of the relative mRNA expression of myostatin using real‐time PCR (qRT‐PCR). Human skeletal myoblasts were stimulated with metformin (0.1 and 2 mM) for 12 h. Gene expression was normalized using GAPDH. (O, P) Creatine kinase (CK) and lactate dehydrogenase (LDH) activity of human skeletal muscle cells. CK and LDH activity was measured in cytoplasmic extracts and normalized using total protein. Results are expressed as the mean ± SEM. Scale bar, $100\mathrm{\mu m}$. *P < 0.05, **P < 0.01, ***P < 0.001 compared to the control.

Figure 2

Metformin increases myostatin expression via AMPK. (A) The level of p‐AMPK was evaluated by western blotting (WB). C2C12 myotubes were stimulated for 36 h with indicated concentrations of metformin (0.01–2 mM). Cell lysates were analysed using anti‐phospho‐AMPK (Thr¹⁷²) and anti‐AMPK antibodies. (B) C2C12 myotubes were incubated with metformin (2 mM) for the indicated times. Cell lysates were analysed using anti‐phospho‐AMPK (Thr¹⁷²), and anti‐AMPK antibodies. (C) The expression of myostatin was evaluated by WB. C2C12 myotubes were pre‐treated with the AMPK inhibitor, Compound C (30 $\mathrm{\mu M}$), and incubated with metformin (2 mM). Cell lysates were analysed using anti‐myostatin and anti‐β‐actin antibodies. (D) Comparison of the relative mRNA expression of myostatin using real‐time PCR (qRT‐PCR). C2C12 myotubes were pre‐treated with AMPK α2 siRNA (100 nM) and incubated with metformin for 24. PCR was normalized using GAPDH. Bars represent the mean ± SEM. (E) The expression of myostatin was evaluated by WB analysis. C2C12 myotubes were pre‐treated with AMPK α2 siRNA (100 nM) and incubated with metformin (2 mM) for 24 h. Cell lysates were analysed using anti‐myostatin, anti‐AMPK α2 and anti‐β‐actin antibodies. (F) C2C12 myotubes were pre‐treated with AMPK α2 siRNA (100 nM) and incubated with metformin (2 mM). Myotube morphology was examined via haematoxylin and eosin (H&E) staining. (G) Analysis of the diameter of cultured myotubes. Myotube diameter was calculated with ImageJ. Bars represent the mean ± SEM. (H) Primary TA myotubes were pre‐treated with AMPK α2 siRNA and incubated with metformin (2 mM). Myotube morphology was examined via H&E staining. (I) Analysis of the diameter of cultured primary myotubes. Primary myotube diameter was calculated with ImageJ. Bars represent the mean ± SEM. (J) The expression of AMPK α was evaluated by WB analysis. CRISPR/AMPK α (2 µg) was transfected for 24 h. Cell lysates were analysed using anti‐AMPK α, and anti‐β‐actin antibodies. (K) The expression of myostatin was evaluated by WB analysis. C2C12 myotubes were pre‐treated with CRISPR/AMPK α and incubated with metformin (2 mM) for 24 h. Cell lysates were analysed using anti‐myostatin and anti‐β‐actin antibodies. (L) C2C12 myotubes were pre‐treated with CRISPR/AMPK α (2 µg) and incubated with metformin (2 mM). Myotube morphology was examined via haematoxylin and eosin (H&E) staining. (M) Analysis of the diameter of cultured myotubes. Myotube diameter was calculated with ImageJ. Results are expressed as the mean ± SEM. Scale bar, $100\mu \mathrm{m}$. **P < 0.01, ***P < 0.001 compared to the control.

Figure 3

Metformin regulates myostatin, via FoxO3a. (A) Comparison of the relative FoxO3a mRNA expression using real‐time PCR (qRT‐PCR). C2C12 myotubes were stimulated with metformin (2 mM) for 12 h. (B) C2C12 myotubes were stimulated with metformin (2 mM) for the indicated times. PCR was normalized using GAPDH. The bars represent the mean ± SEM. (C) The expression of FoxO3a was evaluated by western blotting (WB). C2C12 myotubes were stimulated for 36 h with indicated concentrations of metformin (0.01–2 mM). Cell lysates were analysed using anti‐FoxO3a and anti‐β‐actin antibodies. (D) The expression of FoxO3a was evaluated by WB. C2C12 myotubes were stimulated with metformin (2 mM) for the indicated times. Cell lysates were analysed using anti‐FoxO3a and anti‐β‐actin antibodies. (E) A co‐immunoprecipitation assay used lysate from metformin‐ (2 mM, 36 h) treated C2C12 myotubes. Immunoprecipitations was performed with the anti‐FoxO3a antibody. Cell lysates were analysed using anti‐IgG, anti‐FoxO3a, anti‐p‐AMPK (Thr¹⁷²), anti‐AMPK, and anti‐β‐actin antibodies. IgG was used as the negative control. (F) Comparison of the relative mRNA expression of myostatin using real‐time PCR (qRT‐PCR). C2C12 myotubes were pre‐treated with FoxO3a siRNA (100 nM) and incubated with metformin for 24 h. PCR was normalized using GAPDH. (G) C2C12 myotubes were pre‐treated with FoxO3a siRNA (100 nM) and incubated with metformin (2 mM). Myotube morphology was examined via haematoxylin and eosin (H&E) staining. (H) Analysis of the diameter of cultured myotubes. Myotube diameter was calculated with ImageJ. Results are expressed as the mean ± SEM. Scale bar, $100\mu \mathrm{m}$. **P < 0.01, ***P < 0.001 compared to the control.

Figure 4

HDAC6 is involved in the up‐regulation of myostatin. (A) Comparison of the relative HDAC6 mRNA expression using real‐time PCR (qRT‐PCR). C2C12 myotubes were stimulated with metformin (2 mM) for 12 h. (B) C2C12 myotubes were stimulated with metformin (2 mM) for the indicated times. PCR was normalized using GAPDH expression. Bars represent the mean ± SEM. (C) HDAC6 expression was evaluated by western blotting (WB). C2C12 myotubes were stimulated with metformin (2 mM) for 24 h. Cell lysates were immunoblotted using anti‐HDAC6 and anti‐β‐actin antibodies. (D) A co‐immunoprecipitation assay used lysates from metformin‐treated (2 mM, 36 h) C2C12 myotubes. Immunoprecipitation was performed with the anti‐FoxO3a antibody. Cell lysates were immunoblotted using anti‐IgG, anti‐FoxO3a, anti HDAC6, and anti‐β‐actin antibodies. IgG was used as a negative control. (E) Comparison of the relative HDAC6 mRNA expression using real‐time PCR (qRT‐PCR). C2C12 myotubes were pre‐treated with AMPK α2 siRNA (100 nM) and incubated with metformin for 24 h. PCR was normalized using GAPDH expression. Bars represent mean ± SEM. (F) The expression of HDAC6 was evaluated by western blotting (WB). C2C12 myotubes were pre‐treated with AMPK α2 siRNA (100 nM), incubated with metformin (2 mM) for 24 h, and analysed by WB analysis. Cell lysates were analysed using anti‐HDAC6 and anti‐β‐actin antibodies. (G) C2C12 myotubes were pre‐treated with CRISPR/AMPK α sgRNA (2 µg), incubated with metformin (2 mM) for 24 h, and analysed by WB analysis. Cell lysates were immunoblotted using anti‐HDAC6 and anti‐β‐actin antibodies. (H) Comparison of the relative mRNA expression of myostatin using real‐time PCR (qRT‐PCR). C2C12 myotubes were pre‐treated with HDAC6 siRNA (100 nM) and incubated with metformin for 24 h. PCR was normalized using GAPDH expression. Bars represent the mean ± SEM. (I) C2C12 myotubes were pre‐treated with HDAC6 siRNA, incubated with metformin (2 mM) for 24 h, and analysed by WB analysis. Cell lysates were immunoblotted using anti‐myostatin, anti‐HDAC6, and anti‐β‐actin antibodies. (J) C2C12 myotubes were pre‐treated with HDAC6 siRNA (100 nM) and incubated with metformin (2 mM). Myotube morphology was examined via haematoxylin and eosin (H&E) staining. (K) Analysis of the diameter of cultured myotubes. Myotube diameter was calculated with ImageJ. Results are expressed as the mean ± SEM. Scale bar, $100\mu \mathrm{m}$. Results are expressed as the mean ± SEM. **P < 0.01, ***P < 0.001 compared to the control.

Figure 5

FoxO3a regulates myostatin by binding to its promotor region via subcellular localization. (A) Cells were transfected with the p‐myostatin‐Luc firefly luciferase reporter plasmid (1 µg) and β‐galactosidase control reporter plasmid (100 ng). The next day, metformin was administered for 12 h. The cell extracts were analysed using a Promega luciferase assay kit. The graphs display the mean ratios of the firefly reporter and β‐galactosidase control. Bars represent the mean ± SEM. (B) Framework of the mouse myostatin gene. The arrows indicate the location of putative FoxO3a binding site in the myostatin promotor region. (C) ChIP experiments were performed using digested chromatin from C2C12 myotubes. Purified DNAs were analysed by standard PCR methods using ChIP myostatin primers. (D) Subcellular localization of FoxO3. The C2C12 myotubes were fractionated into nuclear and cytoplasmic fractions after metformin treatment for the indicated times. The fractions were immunoblotted to detect anti‐FoxO3a, cytosolic protein anti‐GAPDH, nuclear proteins, anti‐lamin B antibodies. (E) C2C12 myoblasts treated with metformin were subjected to immunofluorescence (IF) analysis with anti‐MHC (green), anti‐FoxO3a (red), and confocal microscopy. (F) Results of immunofluorescence (IF) nuclear intensity quantification, expressed as the mean ± SEM. Scale bar, 20 $\mu \mathrm{m}$. *P < 0.05, ***P < 0.001 compared to the control.

Figure 6

Reduced muscle atrophy effects of metformin in db/db mice. (A) The expression of p‐AKT (ser⁴⁷³) was determined via western blotting (WB) analysis. C2C12 myotubes were pre‐treated with insulin (100 nM) and TNF‐α (2.5 nM) for 24 h, and then stimulated with insulin (100 nM) for 15 min. Cell lysates were analysed using anti‐phospho‐AKT (ser⁴⁷³), and anti‐AKT antibodies. (B) Comparison of the relative mRNA expression of myostatin using real‐time PCR (qRT‐PCR). C2C12 myotubes were pre‐treated with insulin (100 nM) and TNF‐α (2.5 nM) for 24 h, and then incubated with metformin for 24 h. PCR was normalized using GAPDH. The bars represent the mean ± SEM. (C) Blood glucose level (mg/dL) in metformin‐treated db/db mice, measured after overnight fasting. Bars represent the mean ± SEM. (D) Muscle fibre cross‐sectional area in the GC muscles of db/db mice. Data show the fibre size distribution. (E) Average fibre cross‐sectional area (CSA) of GC muscles. Bars represent the mean ± SEM. (F‐I) Real‐time PCR (qRT‐PCR) comparison of the relative mRNA expression of myostatin in GC, TA, QC, and EDL muscles obtained from control and metformin‐ (250 mg/kg) treated db/db mice. Only the GC samples showed significantly increased myostatin expression. PCR was normalized using GAPDH. Bars represent the mean ± SEM. (J) Grip strength test; fore and hindlimb (four paws) grip force measurements. Tests were performed weekly and averaged across three trials. Data are expressed as the mean ± SEM. n = 10/group. (K) Enzyme‐linked immunosorbent assay (ELISA). Data indicate the serum myoglobin level (ng/mL). The serum myoglobin levels of the db/db mice showed no significant change from the controls. Results are expressed as the mean ± SEM. **P < 0.01, ***P < 0.001 compared to the control.

Figure 7

Metformin negatively affects muscle function in wild‐type mice. (A) Data indicate the muscle fibre cross‐sectional area of the GC muscle wild‐type mice. Data show the fibre size distribution. (B) Average fibre cross‐sectional area (CSA) of GC muscles. Bars represent the mean ± SEM. (C–F) Real‐time PCR (qRT‐PCR) comparison of the relative mRNA expression of myostatin in the GC, TA, QC and EDL muscles of control and metformin‐treated wild‐type mice. Only the GC muscle showed significantly increased myostatin expression. PCR was normalized using GAPDH. The bars represent the mean ± SEM. (G) Grip strength test; fore‐/hindlimb (four paws) grip force measurements. Tests were performed weekly and averaged across three trials. The metformin‐treated wild‐type mice showed decreased muscle grip strength. Data are expressed as the mean ± SEM. n = 10/group. (H) Enzyme‐linked immunosorbent assay (ELISA). Data indicate the serum myoglobin level (ng/mL). The serum myoglobin levels of the wild‐type mice decreased significantly compared with the controls. Results are expressed as the mean ± SEM. **P < 0.01, ***P < 0.001 compared to the control.