S6 kinase inactivation impairs growth and translational target phosphorylation in muscle cells maintaining proper regulation of protein turnover

Virginie Mieulet, Mila Roceri, Catherine Espeillac, Athanassia Sotiropoulos, Mickael Ohanna, Viola Oorschot, Judith Klumperman, Marco Sandri, Mario Pende


A defect in protein turnover underlies multiple forms of cell atrophy. Since S6 kinase (S6K)-deficient cells are small and display a blunted response to nutrient and growth factor availability, we have hypothesized that mutant cell atrophy may be triggered by a change in global protein synthesis. By using mouse genetics and pharmacological inhibitors targeting the mammalian target of rapamycin (mTOR)/S6K pathway, here we evaluate the control of translational target phosphorylation and protein turnover by the mTOR/S6K pathway in skeletal muscle and liver tissues. The phosphorylation of ribosomal protein S6 (rpS6), eukaryotic initiation factor-4B (eIF4B), and eukaryotic elongation factor-2 (eEF2) is predominantly regulated by mTOR in muscle cells. Conversely, in liver, the MAPK and phosphatidylinositol 3-kinase pathways also play an important role, suggesting a tissue-specific control. S6K deletion in muscle mimics the effect of the mTOR inhibitor rapamycin on rpS6 and eIF4B phosphorylation without affecting eEF2 phosphorylation. To gain insight on the functional consequences of these modifications, methionine incorporation and polysomal distribution were assessed in muscle cells. Rates and rapamycin sensitivity of global translation initiation are not altered in S6K-deficient muscle cells. In addition, two major pathways of protein degradation, autophagy and expression of the muscle-specific atrophy-related E3 ubiquitin ligases, are not affected by S6K deletion. Our results do not support a role for global translational control in the growth defect due to S6K deletion, suggesting specific modes of growth control and translational target regulation downstream of mTOR.

  • signal transduction
  • atrophy
  • autophagy

the phosphorylation of transcription and translation factors is a common mechanism to ensure rapid and long-term changes in gene expression depending on the environmental cues. In the 40S ribosomal subunits, the phosphorylation of ribosomal protein S6 (rpS6) was demonstrated more than 20 years ago in eukaryotic cells after exposure to growth factors, mitogens, and nutrients (12, 48, 49). rpS6 is phosphorylated in sequential order on five serine residues (Ser-236, Ser-235, Ser-240, Ser-244, Ser-247) in the carboxy-terminal tail of the protein (24). The kinases catalyzing this reaction have been identified initially as two isoforms, p70 and p85 S6 kinases (S6Ks), encoded by a single gene, S6K1 (22). However, the genetic inactivation of S6K1 in mice does not impair rpS6 phosphorylation because of the existence in the mammalian genome of a homologous gene coding for S6K2 that also has S6K activity (46). The combined deletion of S6K1 and S6K2 (S6K1;S6K2−/−) strongly reduces, but does not abrogate, rpS6 phosphorylation, demonstrating that both kinases are the main regulators of this event in vivo (34).

In resting cells, S6K1 is tethered to a translation preinitiation complex comprising the eukaryotic initiation factor-3 (eIF3) (16). When anabolic signals are received, S6K1 is phosphorylated by the mammalian target of rapamycin (mTOR) and phosphoinositide-dependent protein kinase-1 (PDK1) protein kinases (23). These modifications switch on the catalytic activity of S6K1, which dissociates from the eIF3 subunits and in turn phosphorylates multiple translational targets (16). The S6K1 substrates include not only rpS6 but also eIF4B at Ser-422 and the eukaryotic elongation factor-2 kinase (eEF2K) at Ser-366, which are also involved in the translational machinery (35, 52). These substrates share a consensus sequence with two arginine residues at the −3 and −5 positions from the phosphoserine. Of note, the MAPK-dependent p90 ribosomal protein S6Ks (Rsk) recognize a similar substrate consensus sequence and can partly compensate for the loss of S6K activity (34, 44, 52). Therefore, the combination of the mTOR inhibitor rapamycin and MAPK inhibitors is required to abrogate phosphorylation of rpS6, eIF4B, and eEF2K, indicating a certain degree of redundancy in this regulation by the mTOR/S6K and MAPK/Rsk pathways.

Studies defining the functional consequences of these posttranslational modifications have largely relied on mTOR inhibition by rapamycin. Rapamycin has multiple effects on protein synthesis, as it decreases amino acid uptake, cap-dependent translation initiation, and ribosome biogenesis while increasing protein degradation by autophagy and ubiquitination (7, 53). The effect on translation initiation is especially evident for mRNAs carrying a long 5′-untranslated region (5′-UTR) or a 5′-terminal oligopyrimidine tract (5′-TOP) (1, 19). However, mTOR activity affects a variety of molecular targets in addition to S6K, such as the eIF4E-binding proteins (4EBPs) eIF4G and eIF2 (53). Thus it remains to be demonstrated whether S6Ks and their substrates are directly involved in any of mTOR actions on protein synthesis. Overexpression of a rapamycin-resistant mutant allele of S6K1 protects against the inhibitory effect of rapamycin on cap-dependent translation and 5′-TOP mRNA recruitment onto polysomes (18). Inconsistent with these findings, the combined deletion of S6K1 and S6K2 in mouse embryonic fibroblasts (MEFs) and embryonic stem (ES) cells affects neither global translation initiation, as judged by the polysomal profile, nor 5′-TOP mRNA translation (34). In addition, the substitution of the five phosphorylatable rpS6 serine residues to alanine does not alter 5′-TOP mRNA translation and surprisingly increases translation initiation rates in MEFs (38). The overexpression of a phosphomimetic eIF4B mutant stimulates or inhibits cap-dependent translation depending on the reporter mRNA and experimental conditions (16, 35). The role of the S6K pathway in long-lived protein degradation by autophagy is also not firmly understood. rpS6 dephosphorylation correlates with the induction of autophagy in rat livers during fasting (2, 45). However, Drosophila S6K loss-of-function mutants do not display increased autophagy in standard feeding conditions and are actually protected against starvation-induced autophagy (43). Clearly, further studies are needed to address the contribution of S6K to protein synthesis and degradation.

S6K deletion in Drosophila fruitflies and mice decreases organismal size because of a specific defect in cell size, highlighting an essential role for S6K in cell growth control (27, 46). S6K1;S6K2−/− mice display an increased frequency of perinatal lethality and a 20% reduction of body weight (34). Of note, S6K1 deletion is sufficient to cause the small-size phenotype, suggesting that S6K2 activity does not participate in growth control. Although the weight of all organs is reduced, S6K1−/− adipose tissue, skeletal muscles, and pancreatic β-cells are predominantly affected (31, 33, 50). Importantly, S6K1 deletion in skeletal muscle cells mimics the inhibitory action of rapamycin and amino acid starvation on cell size while blunting the hypertrophic effects of insulin-like growth factor-1 (IGF1) and constitutively active Akt (MyrAkt) (31). Since recent studies have proposed that skeletal muscle mass results from a balance between protein synthesis and degradation (9), it is tempting to speculate that S6K1 may control cell growth through the regulation of one or more mTOR-dependent targets in protein turnover. To address this question, here we evaluate whether S6K1 and/or S6K2 deletion is sufficient to affect the phosphorylation of translational targets in muscle and liver cells. Since we uncover a muscle-specific control of translational target phosphorylation by S6K, we measure molecular markers of protein synthesis and degradation in S6K-deficient muscle cells and perform functional assays. Our results do not support the current model of cell growth control by S6K1 through global protein synthesis regulation and instead indicate distinct atrophy programs on inactivation of specific mTOR targets.


Animals and in vivo procedures.

Generation of S6K-deficient mice has been described previously (31, 34). For in vivo analysis of protein expression, 5-wk-old male mice were killed either after 48-h starvation or after 48-h starvation plus 1-h stimulation with oral administration of 1.4 g/kg leucine (Sigma) and intraperitoneal injection of 1 U/kg insulin (Sigma), or after 48-h starvation plus 1 h of normal chow refeeding. For in vivo analysis of atrophy-related gene expression, 5-wk-old male mice were killed either at the randomly fed state, after 48 h of starvation, or after 48 h of starvation plus 48 h of normal chow refeeding. Denervation-induced muscle atrophy of lower limb muscles was achieved by severing the right sciatic nerve in the midthigh region in anesthetized 2-mo-old male mice. Left sciatic nerves were exposed but not severed to serve as sham-operated controls. Mice were killed after 3, 7, or 12 days, and gastrocnemia muscles were weighed and frozen into liquid nitrogen. Animal procedures were approved by the French Ministry of Agriculture (authorization no. 75-42). Mice were killed by cervical dislocation.

Cell cultures.

Myoblast primary cultures were derived from gastrocnemia and tibialis anterior muscles of 4-wk-old male mice, as described previously (31). Cells were plated on 0.02% gelatin-coated dishes (type A from pig skin; Sigma) and grown in complete medium composed of DMEM-Ham's F12 (Gibco), 2% Ultroser G (Biosepra), 20% fetal calf serum (Gibco), penicillin, streptomycin, and l-glutamine. To differentiate muscle cells into myotubes, myoblasts were plated on matrigel-coated dishes (Becton Dickinson) in complete medium, and after 6-h adhesion, they were switched to DMEM-Ham's F12 + 2% horse serum. To measure the effect of IGF1 and/or rapamycin, myotubes at day 2 of differentiation were incubated with 250 ng/ml IGF1-R3 (Sigma) ± 20 nM rapamycin (Calbiochem) for an additional 2 days in DMEM-Ham's F12 + 2% horse serum. Primary hepatocytes from 12- to 14-wk-old male mice were isolated by liver perfusion as described previously (34). Hepatocytes were plated at 12 × 104 cells/cm2 in Primaria dishes (Falcon) in M-199 medium (Invitrogen) supplemented with 10% fetal bovine serum (FBS; Invitrogen), 0.1% BSA, and 100 nM dexamethasone. After 3 h of adhesion, cells were incubated in serum-free M-199 medium containing 0.1% BSA for 48 h and then in amino acid- and glucose-free DMEM + 0.02% BSA for 3 h. Hepatocytes were stimulated for 1 h in DMEM with growth factors corresponding to 10% FBS + 1 μM insulin (Sigma) + 4 nM EGF (Sigma) with or without 30-min pretreatment with 20 nM rapamycin (Calbiochem), 10 μM U0126 (Calbiochem), and 1 μM wortmannin (Calbiochem). All media contained penicillin (100 U/ml), streptomycin (100 μg/ml), and amphotericin B (Fungizone) (250 ng/ml). To quantify hepatocyte size, cells were fixed for 10 min with 4% paraformaldehyde and stained for 10 min with 10% Giemsa solution in PBS. Cell size was then analyzed on a Nikon E800 microscope equipped with a video camera and Lucia archive software.


For immunoblot analysis, cells were washed twice with cold PBS and then scraped from the culture dish into lysis buffer B (50 mM Tris, pH 8, 1% NP-40, 120 mM NaCl, 20 mM NaF, 1 mM benzamidine, 1 mM EDTA, 1 mM EGTA, 15 mM sodium pyrophosphate, 1 mM PMSF, 2 mM sodium orthovanadate, 2 μg/ml leupeptin, 5 μg/ml aprotinin). Homogenates were sonicated for 6 min and spun at 8,000 g for 10 min at 4°C to remove cell debris. Muscle and liver tissues were frozen in liquid nitrogen and powdered manually with a mortar and pestle. A volume of glass beads (Sigma G-8772) and a volume of lysis buffer (20 mM Tris, pH 8, 5% glycerol, 138 mM NaCl, 2.7 mM KCl, 1% NP-40, 20 mM NaF, 18 mM pepstatin A, 5 mM EDTA, 1 mM sodium orthovanadate, 20 mM leupeptin, 1 mM dithiothreitol) equivalent to the tissue volume were added. The homogenates were vortexed 10 s and put back on ice for 10 s; this was repeated four times. The samples were then agitated for 15 min at 4°C until complete lysis and spun at 8,000 g for 10 min at 4°C to remove fat and debris. Protein extracts were resolved by SDS-PAGE before transfer to nitrocellulose membranes (Protran) and incubation with the following primary antibodies: anti-phosphorylated Ser-235/Ser-236 rpS6, anti-phosphorylated Ser-422 eIF4B, anti-eIF4B, anti-phosphorylated Thr-56 eEF2, anti-phosphorylated Thr-389 p70S6K, anti-phosphorylated Thr-359/Ser-363 p90Rsk (Cell Signaling Technology), anti-LC3 (a kind gift from N. Mizushima Tokyo Medical and Dental University, Tokyo, Japan), and anti-p70S6K (C-18) (Santa Cruz Biochemicals).

Analysis of protein synthesis by [35S]methionine and [35S]cysteine incorporation.

Muscle cells (20,000 cells/cm2) were plated on matrigel-coated six-well plates in complete medium. After 4 days of differentiation in DMEM-Ham's F12 + 2% horse serum, cells were washed twice with cysteine- and methionine-free DMEM. Then, 1 ml of this DMEM was added together with 2% dialyzed horse serum and 40 μCi of [35S]methionine and [35S]cysteine (Promix, Amersham) for 2 h at 37°C. The amount of radiolabel was saturating, as the amount in the medium did not decrease significantly during the 2-h incubation. The protein synthesis rate was determined as described previously (6). Briefly, the cells were washed twice in ice-cold PBS, and radiolabel incorporation into protein was assessed by lysing the cells in 100 μl of 0.2% Triton and then removing three aliquots of 10 μl. Two aliquots were counted in a scintillation counter to evaluate the label in total cell lysate. The third aliquot was used for protein quantitation by the colorimetric DC Protein Assay (Bio-Rad); 100% ice-cold trichloroacetic acid (TCA) was added to the remaining 70 μl to a final concentration of 10% to precipitate the protein. After 10 min on ice, the solutions were spun at 12,000 g to pellet the precipitated proteins, and the amount of free radiolabel was assessed by removing two aliquots of the supernatant and counting them in a scintillation counter. The amount of radiolabel incorporated into protein was calculated by subtracting the value of the nonincorporated label from the value of the label in the total cell lysate.

Protein/DNA content.

This experiment was done in parallel with the analysis of protein synthesis by [35S]methionine and [35S]cysteine incorporation. Cells (20,000 cells/cm2) were plated on matrigel-coated six-well plates in complete medium. After 4 days of differentiation in DMEM-Ham's F12 + 2% horse serum, cells were washed twice in ice-cold PBS, scraped in 400 μl of PBS, and spun at 3,300 g for 5 min at 4°C. Lysis buffer (100 μl; 0.1% NP-40, 0.5 mM EDTA) was added to the cell pellet, and the samples were frozen into liquid nitrogen, thawed, and sonicated for 5 min. The DNA content was evaluated using a fluorescent DNA quantitation kit (Bio-Rad). Briefly, Hoechst-33258 dye was added to an aliquot. The dye bound to double-stranded DNA was measured with a fluorometer (excitation, 360 nm; emission, 460 nm). Salmon sperm DNA was used for concentration standard curves. The protein quantitation was evaluated on an aliquot of the same sample using BSA as standard.

Polysome fractionation.

Sucrose density gradient centrifugation was employed to separate the subpolysomal from the polysomal ribosome fractions. Cells (25,000 cells/cm2) were plated in matrigel-coated 15-cm plates in complete medium. After 4 days of differentiation in DMEM-Ham's F12 + 2% horse serum, myotubes were washed twice in hypotonic buffer (5 mM Tris·HCl, pH 7.5, 1.5 mM KCl, 2.5 mM MgCl2). The cells were scraped in 400 μl of extraction buffer [5 mM Tris·HCl, pH 7.5, 1.5 mM KCl, 2.5 mM MgCl2, 0.5% Triton X-100, 0.5% sodium deoxycholate, 120 U/ml RNase inhibitor (Promega), 3 mM dithiothreitol]. The lysates were spun at 3,000 g for 8 min at 4°C. An aliquot of the supernatant was removed to measure protein concentration. Heparin was added to the supernatants at a final concentration of 1 mg/ml. The extracts were rapidly frozen into liquid nitrogen and stored at −80°C. The same amount of protein for each sample (800 μg) was adjusted to the same final volume (600-μl maximum) and layered on a 0.5–1.5 M linear sucrose gradient (20 mM Tris·HCl, pH 7.5, 80 mM NaCl, 5 mM MgCl2, 1 mM dithiothreitol) and centrifuged in a SW41 rotor at 36,000 rpm for 2 h at 4°C. Following centrifugation, with the use of the density gradient fractionation system (Isco), the gradient was displaced upward through a flow cell recording absorbance at 260 nm.

Quantitative real-time PCR.

Muscle tissues were frozen in liquid nitrogen and powdered with a mortar and pestle. Total RNA from muscle cells was prepared using Trizol reagent according to the manufacturer's protocol (Life Technologies, Paisley, UK). Single-strand cDNA was synthesized from 400 ng of total RNA with random hexamer primers and RT Superscript II (Invitrogen). Quantitative real-time PCR (QRT-PCR) was performed using a LightCycler Instrument (Roche Diagnostics) according to the manufacturer's instructions and the QuantiTect SYBR Green kit (Qiagen). Briefly, the final volume of 20 μl contained 8 ng of cDNA, 50 μM primer mix, and 10 μl of QuantiTect SYBR Green kit. The reactions were carried out in capillaries (Roche Instruments) for 40 cycles. We determined the relative amounts of the mRNAs studied by means of the second-derivative maximum method, with LightCycler analysis software version 3.5 and β-actin as the housekeeping gene for all studies. The murine primer sequences used were as follows: β-actin, forward 5′-CTGGCTCCTAGCACCAAGAT-3′ and reverse 5′-GGTGGACAGTGAGGCCAGGAT-3′; atrogin-1, forward 5′-GGCGGACGGCTGGAA-3′ and reverse 5′-CAGATTCTCCTTACTGTATACCTCCTTGT-3′; muscle ring finger 1 (MuRF1), forward 5′-ACAACCTGTGCCGCAAGTG-3′ and reverse 5′-AGGACAACCTCGTGCCTACAAG-3′. The results of QRT-PCR were given in arbitrary units and expressed as fold changes in mRNA levels relative to wild-type controls.

Electron microscopy.

For electron microscopy of extensor digitorum longus muscle tissues, we used conventional fixation-embedding procedures based on glutaraldehyde-osmium fixation and Epon embedding, as previously described (40). For electron microscopy of myotube cultures, cells were fixed in 2% glutaraldehyde in 0.1 M cacodylate buffer, pH 7.2, for 2 h at room temperature. After a rinsing with cacodylate buffer, cells were postfixed in a mixture of 1% OsO4 and 1% K4Ru(Cn)6 for 2 h at 4°C, stored overnight in 70% ethanol, and dehydrated through a series of increasing ethanol concentrations. After the embedding in Epon according to the standard protocol, ultrathin sections were obtained and counterstained with lead citrate and uranyl acetate. At least 25 nonoverlapping fields of the sections were photographed at a final magnification of ×10,000. Autophagosome, mitochondria, and cytosol volume densities were quantified by point-counting morphometry using a 1-cm grid.


It was previously demonstrated that the growth factor-activated mTOR/S6K and MAPK/Rsk pathways converge on the regulation of rpS6, eIF4B, and eEF2K (34, 44, 52). To evaluate whether S6K deletion was sufficient to affect these putative S6K translational targets, phosphorylation of rpS6 Ser-235/236, eIF4B Ser-422, and eEF2 Thr-56 was measured in mouse tissue in vivo 1 h after administration of insulin and leucine. The phosphorylation state of eEF2 Thr-56, the eEF2K site, provides a readout of eEF2K activity that is negatively controlled by the growth factor pathways (52). In this study, we limited our analysis to skeletal muscles and liver, two major targets of the insulin and nutrient signaling pathways. The combined deletion of S6K1 and S6K2 genes strongly attenuated the insulin- and leucine-induced eIF4B phosphorylation and completely abrogated rpS6 phosphorylation in gastrocnemius muscle (Fig. 1A). Surprisingly, these conditions were not sufficient to affect eEF2 phosphorylation, precluding the in vivo analysis of this event. These data indicate that S6Ks are the main kinases responsible for rpS6 and eIF4B phosphorylation in skeletal muscles after leucine and insulin administration.

Fig. 1.

Control of eukaryotic initiation factor-4B (eIF4B) phosphorylation is tissue specific. Western blot analysis of protein extracts from wild-type, S6K1−/−, S6K2−/−, and S6K1;S6K2−/− gastrocnemius muscle (A) and liver tissue (B) was carried out with anti-phosphorylated S6 (Ser-235/236), anti-phosphorylated eIF4B (Ser-422), anti-phosphorylated eukaryotic elongation factor-2 (eEF2; Thr-56), and total eIF4B, with total eIF4B as a loading control. Five-week-old male mice of the indicated genotype were killed after 48-h starvation (starved) or after 48-h starvation and 1-h stimulation with 1.4 g/kg leucine and 1 U/kg insulin (Leu + Ins). The phospho-eIF4B (P-eIF4B)-to-total eIF4B ratio of signal intensity is indicated, as assessed by ImageJ software. S6K, S6 kinase.

The deletion of S6K1 or S6K2 has differential effects on muscle growth that correlate with S6K1 but not S6K2 activity (31). Therefore, we addressed whether the two S6K family members differentially regulated these putative S6K translational targets. The effect of the single deletion of S6K1 or S6K2 on rpS6 and eIF4B was less pronounced than that of the combined deletion, indicating that both kinases participate in this regulation (Fig. 1A). The levels of eIF4B phosphorylation were comparable in S6K1−/− and S6K2−/− muscles, while rpS6 phosphorylation was more affected by S6K2 deletion. Thus the growth responses of the distinct S6K genotypes do not correlate with the phosphorylation status of rpS6 and eIF4B.

We next evaluated these posttranslational modifications in liver from the same mice after insulin and leucine administration. rpS6 phosphorylation was abolished in liver as well as in muscle lacking both S6Ks (Fig. 1B). Conversely, in liver from S6K1;S6K2−/− mice, the eIF4B phosphorylation was largely intact. Thus rpS6 is a preferential S6K substrate in both muscle and liver tissues, while eIF4B phosphorylation is predominantly carried by S6K1/S6K2 in skeletal muscles and by an S6K-independent pathway in liver.

To further strengthen our observations, we set up primary cultures of skeletal muscle and liver tissues from wild-type and S6K1;S6K2−/− mice and combined the genetic deletions with pharmacological inhibitors of the growth factor pathways. Both cell types were stimulated with nutrients, serum, and growth factor peptides to reach a maximal response. In wild-type muscle cells, rapamycin treatment inhibited rpS6 and eIF4B phosphorylation, confirming that these translational targets are under control of the mTOR pathway in muscle (Fig. 2A). In addition, rapamycin increased eEF2 phosphorylation, revealing in cultured cells a negative control of eEF2K activity by the mTOR pathway that was not evident in vivo. In S6K1;S6K2−/− myotubes, basal and growth factor-stimulated rpS6 phosphorylation was abrogated. Although eIF4B phosphorylation was attenuated compared with wild-type cells, it was not abolished, and it remained sensitive to growth factor stimulation (Fig. 2, A and B). Of note, rapamycin had no effect on basal and growth factor-induced eIF4B phosphorylation in S6K1;S6K2−/− myotubes, indicating the involvement of an mTOR-independent pathway that partially compensates for the loss of S6K activity. In contrast, the sensitivity of eEF2 phosphorylation to rapamycin did not differ in wild-type and S6K1;S6K2−/− myotubes. Taken together, these data demonstrate that S6Ks regulate rpS6 and eIF4B phosphorylation in muscle cells, whereas eEF2K activity is under the control of an mTOR-dependent but S6K-independent mechanism. S6K deletion is sufficient to cause a dramatic loss of rpS6 phosphorylation, while eIF4B phosphorylation is partially rescued by an mTOR-independent pathway.

Fig. 2.

Effect of S6K deletion on the phosphorylation of ribosomal protein S6 (rpS6), eIF4B, and eEF2 in skeletal muscle cells and hepatocytes. Western blot analysis of protein extracts from wild-type and S6K1;S6K2−/− myotubes (A) was carried out with the indicated antibodies. Myotubes from the indicated genotypes were allowed to differentiate for 2 days in 2% horse serum and were then incubated for an additional 2 days in the presence of either vehicle (none) or 250 ng/ml IGF1-R3 (IGF) or 20 nM rapamycin (Rap) or both (IGF + Rap). The phospho-eIF4B-to-total eIF4B ratio of signal intensity is indicated, as assessed by ImageJ software. B: histogram shows the average ± SE of the phospho-eIF4B-to-total eIF4B ratio from 3 distinct experiments. Myotubes were treated with vehicle or 250 ng/ml IGF1-R3 in the presence or absence of 20 nM rapamycin, as described in A, and immunoblot signals were quantified using ImageJ software. C: Western blot analysis of protein extracts from wild-type and S6K1;S6K2−/− hepatocytes was carried out with the indicated antibodies. Hepatocytes from the indicated genotypes were grown in amino acid-, glucose-, and serum-free DMEM + 0.2% BSA for 3 h (none) and then stimulated for 1 h in DMEM supplemented with growth factors (GF; 10% FBS + 1 μM insulin + 4 nM EGF) in the presence or absence of 20 nM rapamycin ± 10 μM U0126 (U0) ± 1 μM wortmannin (Wort). Rsk, ribosomal S6K.

In hepatocytes, the phosphorylation of rpS6 was largely but not completely dependent on mTOR and S6K activities (Fig. 2B). As previously reported (34), the residual rpS6 phosphorylation in rapamycin-treated and S6K-deficient cells was blocked by the MAPK inhibitor U0126, suggesting that Rsk participates in rpS6 phosphorylation but to a lesser extent than S6K. In contrast to muscle cells, in wild-type liver cells, rapamycin did not affect the control of eEF2 and eIF4B phosphorylation by growth factors, suggesting that in liver, the mTOR pathway is not predominant over the other pathways regulating these targets (Fig. 2B; Ref. 44). Consistent with these findings, eIF4B and eEF2 regulation was not impaired in S6K-deficient hepatocytes. Altogether, these findings confirm the in vivo data, demonstrating that rpS6 is a preferential S6K substrate compared with eIF4B. Moreover, the contributions of mTOR-dependent and -independent pathways on these phosphorylation events vary among distinct tissues.

Adult hepatocytes in culture comprised a population of mononucleated and binucleated cells. S6K deletion significantly decreased the size of binucleated cells (3,476 ± 192-μm2 surface of wild-type vs. 2,684 ± 286-μm2 surface of S6K1;S6K2−/− hepatocytes; n = 3 independent primary cultures, P < 0.05), whereas mononucleated cells were less affected (2,311 ± 152-μm2 surface of wild-type vs. 2,088 ± 272-μm2 surface of S6K1;S6K2−/− hepatocytes; n = 3 independent primary cultures). Taken together, our findings indicate that the small size of S6K-deficient hepatocytes is not accompanied by a defect in eIF4B and eEF2 phosphorylation.

To investigate the involvement of other growth factor-activated pathways in the phosphorylation of eIF4B and eEF2, hepatocytes were treated with the MAPK inhibitor U0126, with the phosphatidylinositol 3-kinase (PI3K) inhibitor wortmannin, and with the mTOR inhibitor rapamycin alone or in combination. MAPK inhibition was sufficient to block eEF2K activity, as judged by eEF2 phosphorylation, while rapamycin and wortmannin had no effect (Fig. 2B). Conversely, the combination of PI3K and MAPK inhibitors was required to completely abrogate eIF4B phosphorylation in wild-type hepatocytes, suggesting that S6K, Rsk, and a PI3K-dependent kinase participate in eIF4B phosphorylation in liver. Since the constitutively active allele of Akt leads to increased eIF4B phosphorylation in a rapamycin-insensitive manner (data not shown), and the Akt kinases recognize a similar substrate consensus sequence as S6K and Rsk, we propose that Akt family members may be the PI3K-dependent kinases that regulate eIF4B.

To gain further insight into the differential regulation of rpS6 and eIF4B in muscle and liver, we compared S6K and Rsk activities by using phosphospecific antibodies that recognize S6K1/S6K2 or Rsk1/Rsk3 when they are phosphorylated, respectively, at Thr or Ser residues in the hydrophobic region that are critical for activity. Interestingly, Rsk phosphorylation was higher in hepatocytes than in skeletal muscle cells, while S6K1 phosphorylation was roughly equivalent in these two cell types (Fig. 3). Conversely, S6K expression was higher in muscle cells than in liver. Although other mechanisms may also be involved, the differential expression and activity levels of S6K and Rsk in these two tissues may contribute to the predominance of one pathway over the other in the control of the translational targets.

Fig. 3.

Differential expression and activity levels of S6K and Rsk in muscle and liver tissue. Western blot analysis of protein extracts from wild-type and S6K1;S6K2−/− muscle and liver tissues or myotubes and hepatocytes was carried out with anti-phosphorylated p90Rsk (Thr-359/Ser-363), anti-p70S6K, anti-phosphorylated p70S6K (Thr-389), and anti-tubulin, the latter as loading control. Muscle and liver tissues from the indicated genotypes were extracted after 48-h starvation and 1-h stimulation with 1.4 g/kg leucine and 1 U/kg insulin (Leu + Ins). Myotubes from the indicated genotypes were allowed to differentiate for 2 days in 2% horse serum and then were incubated for an additional 2 days in the presence of 250 ng/ml IGF1-R3 (IGF); hepatocytes from the indicated genotypes were grown in amino acid-, glucose-, and serum-free DMEM + 0.2% BSA for 3 h and then stimulated for 1 h in DMEM supplemented with growth factors (GF; 10% FBS + 1 μM insulin + 4 nM EGF).

Since S6K deletion affected rpS6 and eIF4B phosphorylation in skeletal muscle, we set up to assess protein synthesis in this cell type. First, we assayed the incorporation of radiolabeled methionine into proteins in cultured wild-type and S6K1;S6K2−/− muscle cells at day 4 of differentiation, when a difference in rpS6 and eIF4B phosphorylation was clearly observed between the two genotypes. At this stage, muscle cells were differentiated into multinucleated myotubes. The total amount of protein per myonuclear number was decreased in S6K-deficient myotubes (Fig. 4A). Similarly, the methionine incorporation per myonuclear number was decreased in S6K1;S6K2−/− cells (Fig. 4B). However, when the methionine incorporation was normalized to the total protein content, no difference was observed between wild-type and mutant cells (Fig. 4C). Since S6K1;S6K2−/− myotubes displayed an ∼25% reduction of myotube diameter with no defect in myonuclear number per cell (31), these data suggest that the reduced incorporation of radiolabeled methionine in S6K-deficient cells is not the primary cause of cell atrophy but rather a consequence of their reduced overall protein content and cell size. To more precisely assess translation initiation, we fractionated polysomes, monosomes, and free ribosomal subunits over a sucrose gradient, as the amount of ribosomes engaged in polysomes vs. monosomes reflects the global translational rates. As observed in fibroblasts, hepatocytes, and embryonic stem cells, S6K1;S6K2−/− muscle cells displayed a similar polysomal profile in resting conditions compared with wild types, and rapamycin produced a decrease in the polysomal fraction in both genotypes (data not shown and Ref. 34). Thus our analysis does not reveal a major defect of global translation initiation as a consequence of S6K inactivation, though we cannot rule out a regulation of specific mRNA classes. Of note, S6K deletion mimics the rapamycin effect on muscle cell size but not on global translation initiation.

Fig. 4.

Rate of protein synthesis in S6K-deficient cells. Myotubes from the indicated genotypes were allowed to differentiate for 4 days in 2% horse serum. A: total amount of protein per myonuclear number. The assay was performed by measuring protein and DNA concentrations and calculating the ratio of protein to DNA. Each sample was run in triplicate, and histograms are means ± SE of 3 independent experiments. B: radiolabeled amino acid incorporation per myonuclear number. The assay of protein synthesis rates was performed by measuring the amount of [35S]methionine (35S met) and [35S]cysteine incorporation into cellular proteins over 2 h and normalized to total DNA content. C: radiolabeled amino acid incorporation per total protein content. The assay was performed as described in B, although the data of incorporated radiolabeled amino acids were normalized to total protein. Each experiment was done in triplicate, and histograms are means ± SE of 3 independent experiments.

Since S6K-deficient muscles are atrophic, and nutrient deprivation does not cause further reduction in skeletal muscle mass (31), we reasoned that an atrophic program leading to protein degradation either may be constitutively turned on in S6K1;S6K2−/− muscle cells or its induction may be blunted. Starvation initiates two distinct mechanisms, the proteasome-mediated degradation of ubiquitinated proteins and the autophagosome-mediated digestion of cellular components. Both processes release amino acids that are used by muscles and other tissues as mitochondrial energy substrates. The onset of these atrophy programs may be followed by appropriate molecular markers. First, we assessed the induction of autophagy by detecting the posttranslational modifications of the LC3 protein, which is cleaved and conjugated to phospholipids during the early steps of autophagy (21). These events increase the electrophoretic mobility of LC3 and can be revealed by immunoblot analysis with anti-LC3 antibodies. As shown in Fig. 5A, starvation led to a decrease in the slower-migrating full-length LC3 (LC3 I) and the appearance of a faster-migrating band corresponding to cleaved LC3 (LC3 II). This marker of autophagy was also detected in starved S6K1;S6K2−/− muscles, suggesting that autophagy does not require S6K activity.

Fig. 5.

Autophagy is not affected by S6K inactivation. A: Western blot analysis of protein extracts from wild-type and S6K1;S6K2−/− gastrocnemius muscle was carried out with anti-LC3 and anti-eIF4B antibodies, the latter as a loading control. Five-week-old male mice were either starved for 48 h (starved) or starved for 48 h and then refed with normal chow for 1 h (refed). B: detection of autophagosome by electron microscopy analysis of myotube cultures. Left: early (arrowhead) and late (arrows) autophagosomes and mitochondria (asterisks) in a representative electron microscopy image of 48-h rapamycin-treated wild-type myotubes. Middle and right: the histograms are the ratio of autophagosome volume density vs. total cytosolic volume and mitochondrial volume density vs. total cytosolic volume. At least 25 nonoverlapping fields were analyzed in 1 culture.

To directly follow autophagosome formation, electron microscopy was performed in cultured myotube cells. The volume densities of early and late autophagosomes as well as mitochondria were quantified by morphometric analysis of differentiated myotubes in culture and normalized to the total cytosolic volume (Fig. 5C). The early autophagosomes were recognized as double-membrane vesicles, while late autophagosomes contained electron-dense lysosomal degradation products. As expected, a long-term rapamycin treatment increased autophagosome formation in wild-type myotubes. However, S6K deletion had no effect on autophagy. Of note, S6K1;S6K2−/− muscle cells displayed an increased mitochondrial mass, as previously reported for skeletal muscles (50). These data suggest that the small size of S6K1;S6K2−/− muscle fibers is not a consequence of increased autophagy.

Since the expression of atrogin-1 and MuRF1, two E3 ubiquitin ligases that target specific proteins to the proteasome, is rapidly upregulated in all forms of muscle atrophy and is required for muscle wasting (3, 41, 47), we assayed their mRNA levels by QRT-PCR after starvation of wild-type and S6K-deficient mice. As expected, in wild-type gastrocnemius and tibialis anterior muscles, a 2-day starvation period induced a dramatic upregulation of atrogin-1 and MuRF1 mRNAs that was suppressed by 2 days of refeeding (Fig. 6). Strikingly, S6K-deficient muscles exhibited the same pattern of atrogin-1 and MuRF1 mRNA expression compared with wild types. Thus S6K1;S6K2−/− muscles are smaller, and their mass is not decreased by starvation, despite a normal regulation of atrogin-1 and MuRF1.

Fig. 6.

Atrophic S6K-deficient muscles exhibit a normal regulation of atrogin-1 and MuRF1. Quantitative RT-PCR on gastrocnemia (GC; left) and tibialis anterior (TA; right) muscles from wild-type and S6K1;S6K2−/− mice was carried out with atrogin-1 primer (top) or MuRF1 primer (bottom). Five-week-old male mice were randomly fed (control), starved for 48 h (starved), or starved for 48 h and then refed with normal chow for 48 h (refed). Each sample was run in triplicate, and histograms are means ± SE of 5 animals/condition.

To evaluate whether S6K-deficient muscles displayed differential responses depending on the atrophic conditions, we measured the effect of denervation on muscle mass and atrophy-related E3 ubiquitin ligases. Sciatic nerve section led to MuRF1 expression in gastrocnemius muscles of both wild-type and S6K1;S6K2−/− genotypes (Fig. 7A). While the mass of S6K-deficient gastrocnemius muscles was largely resistant to starvation (31), denervation reduced by ∼20 and 30% the weight of S6K1;S6K2−/− gastrocnemius muscles, respectively, at days 7 and 12 after surgery (Fig. 7B). Thus S6K deletion does not lead to a minimal muscle cell size, as S6K1;S6K2−/− fibers are shrunken after denervation, similar to wild-type controls. In conclusion, the S6K inactivation confers a small-size phenotype that is selectively resistant to nutrient deprivation and does not stem from deregulation of two main pathways involved in starvation-induced muscle wasting, autophagy and atrophy-related E3 ubiquitin ligases.

Fig. 7.

Size of S6K-deficient muscles is reduced by denervation. A: quantitative RT-PCR was carried out with MuRF1 primers, and gene expression in denervated muscles was compared with that in contralateral muscle. The right hindlimb muscles from 2-mo-old male mice were denervated by severing the right sciatic nerve in the midthigh. The left hindlimb from each animal served as its own control. At days 3, 7, and 12 postdenervation, the right and left gastrocnemia muscles were weighed, and RNA samples were prepared. Each sample was run in triplicate, and histograms are means ± SE of 3 animals for each condition. B: muscle weight during denervation. Results are expressed as percentage of contralateral muscle weight and are averages ± SE of at least 3 mice of the indicated genotype at each time point.


In all eukaryotic cells, TOR integrates nutrient signals and coordinates cell growth, proliferation, and metabolism (53). The central role of mTOR is underlined by the striking phenotype of mTOR-null mouse embryos, which do not develop after the blastocyst stage (8, 28). mTOR resides in two large complexes (mTORCs) associated with either raptor protein, mTORC1, or rictor protein, mTORC2 (53). The functions of mTORC1 have been characterized by using the antibiotic rapamycin, a specific inhibitor of mTOR/raptor that has cytostatic actions and reduces cell size. It is well documented that mTORC1 interacts with translational machinery. mTOR/raptor can associate with eIF3 in a nutrient- and growth factor-dependent manner and regulate the phosphorylation of 4EBP, eIF4G, and S6K (15, 16). Rapamycin affects protein turnover at multiple steps, from ribosome biogenesis (19, 25) to translation initiation and elongation (1, 5) to protein degradation by increasing autophagy (45) and the expression of proteasome subunits and ubiquitin ligases (10, 39). However, it is unclear 1) whether these processes are coordinated, 2) the molecular targets of mTOR regulating these events, and 3) how growth is dependent on this control.

Genetic inactivation of S6Ks represents an attractive model to address these questions for the following reasons. Mice carrying S6K1, S6K2, and the combined gene deletions are viable, and S6K1−/− mice, but not S6K2−/−, present a growth defect (34). Moreover, S6K1 deletion in skeletal muscles is sufficient to mimic the inhibitory effect of rapamycin on cell growth but not proliferation (31). Interestingly, the regulation of translational targets is tissue specific. We show that S6K-deficient muscles present an impaired rpS6 and eIF4B phosphorylation, whereas in liver, the MAPK/Rsk and PI3K/Akt pathways can compensate for the loss of S6K and partly or completely rescue the phosphorylation of rpS6 and eIF4B (Figs. 1 and 2). The predominance of the mTOR pathway in skeletal muscle is also demonstrated by the control of EF2 phosphorylation, which is rapamycin sensitive in muscle cells but not in hepatocytes (Figs. 1 and 2). However, we also show that EF2 kinase activity is not affected by S6K signaling, in contrast to a previous proposal (52). The tissue-specific difference of translational target phosphorylation may stem from a change in expression and/or activity of distinct signal transduction elements (Fig. 3). This regulation may contribute to the different physiological responses of liver and muscle to nutrient availability, as the mTOR pathway depends on nutrient levels, whereas MAPK and Akt activities are not modulated by these stimuli.

Here, we conclude that the atrophy of S6K-deficient muscles is not causally linked to a defect in rpS6 and eIF4B phosphorylation. In muscle, S6K2 activity contributes to the phosphorylation of these factors without affecting the muscle growth responses (Fig. 2 and Ref. 31), suggesting that muscle growth is controlled by an S6K1-specific mechanism independent of rpS6 and eIF4B phosphorylation. Consistent with these findings, S6K-deficient hepatocytes are smaller than wild types, and eIF4B phosphorylation is not affected (Fig. 1). Moreover, mice carrying a nonphosphorylatable rpS6 allele display a milder phenotype compared with S6K1-deficient mice (38). The rpS6 mutant mice in the adult stage do not present a reduction of total body size, and the defect of cell growth is limited to pancreatic β-cells. It is possible that rpS6 phosphorylation is required for the growth of specific tissues that do not include skeletal muscle. Alternatively, the defect of pancreatic β-cell growth in rpS6 mutant mice is a consequence of metabolic adaptations, as these mice present increased insulin sensitivity. S6Ks have additional substrates that are not implicated in the translational machinery, such as S6K1 Aly/Ref-like target (36), mTOR (17), insulin receptor substrate 1 (14), and Bcl-2/Bcl-XL-antagonist, causing cell death (13). Of note, some of these proteins may be S6K1-specific targets (36). Future studies could address their direct implication in muscle cell growth.

Recent studies have suggested that a change in the balance between protein synthesis and degradation determines the rate of muscle growth (9). During cachexia, nutrient deprivation and denervation, the E3 ubiquitin ligases atrogin-1 and MuRF1 are regulated by Akt/Foxo signaling, and this response precedes and contributes to rapid atrophy (3, 41, 47). In addition, the degradation of long-lived proteins by the formation of autophagic vacuoles accompanies several myopathy diseases (29, 30, 51). mTOR plays a major role in muscle growth, as rapamycin is a potent inhibitor of muscle hypertrophy and affects autophagy as well as ubiquitin ligase expression (4, 7, 32, 37, 39). Moreover, the rapamycin-insensitive mTORC2 regulates Akt (42) and should therefore affect Foxo-mediated atrogin-1 and MuRF1 gene expression. The muscle atrophy model triggered by inactivation of the mTOR substrate S6K1 in mice is characterized by thin size and complete resistance to nutrient deprivation (31). Surprisingly, here we show that S6Ks do not mediate the regulation of global translation, expression of atrophy-related genes, and autophagy by nutrient/mTOR signaling (Figs. 47), although S6K1 is permissive for cell growth depending on nutrient availability.

Our results indicate that muscle atrophy on S6K deletion is distinct from atrophy triggered by inactivation of other mTOR effectors. It is possible that S6K1 controls the expression of a specific subset of proteins that function as a size checkpoint in muscle. The existence of such factors has been proposed, after genetic screening for size mutants in yeast and Drosophila cells (11, 20). In both organisms, the majority of genes affecting cell size encode nucleolar proteins, suggesting that the rate of ribosome biogenesis may set cell size. Although S6K1-deficient cells do not present defects in 5′-TOP mRNA translation coding for ribosomal proteins (34), we cannot exclude the possibility that rDNA transcription, rRNA processing, or other activities required for ribosome biogenesis may be affected. Alternatively, S6K1 inactivation may decrease cellular components other than proteins. We have demonstrated recently that S6K1;S6K2−/− muscles display a sharp reduction in lipid content (V. Aguilar, A. Sotiropoulos, and M. Pende, unpublished data), possibly due to increased fatty acid β-oxidation (50). Future studies should address whether the amount of intracellular lipids participates in the cell volume control. Of note, a switch in energy substrate utilization is known to affect muscle size and sensitivity to nutrient availability, as oxidative fibers tend to be smaller and resistant to starvation compared with glycolytic fibers (26). As shown in Fig. 5 and as previously reported (50), S6K-deficient muscles display increased mitochondrial mass. Since S6K1;S6K2−/− muscle size is resistant to nutrient deprivation while maintaining sensitivity to denervation (Fig. 7 and Ref. 31), S6K may have the specific function in the mTOR pathway of regulating a metabolic program adapting cell size to nutrient availability. Definition of the distinct types of atrophy resulting from inactivation of mTOR effectors has the potential to target specific functions in human disease.


This work was supported by a grant to M. Pende from the INSERM Avenir Program (R01131KS), from Fondation de la Recherche Medicale (DEQ20061107956), and from Fondation Schlumberger pour l'Education et la Recherche and to A. Sotiropoulos and M. Pende from the Association Française contre les Myopathies (9971). V. Mieulet is a recipient of a fellowship from the Association Française contre les Myopathies.


We thank the Novartis Foundation and the George Thomas laboratory for the use of S6K mutant mice. We are grateful to the members of INSERM-U810 for support and to G. Posthuma (University Medical Center Utrecht) for advice on the morphometric analyses.


  • The costs of publication of this article were defrayed in part by the payment of page charges. The article must therefore be hereby marked “advertisement” in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.


View Abstract