Jun N-terminal kinase mediates activation of skeletal muscle glycogen synthase by insulin in vivo.

Mitogen-activated protein kinases (MAPKs) represent a conserved family of Ser/Thr protein kinases with central roles in intracellular signaling. Activation of three prominent members of the MAPK family, i.e. extracellular response kinases (ERK), jun N-terminal kinase (JNK), and p38, was defined in vivo in order to establish their role, if any, in the cardinal response of skeletal muscle to insulin, the activation of glycogen synthesis. Insulin was found to activate ERK, JNK, and p38 in skeletal muscle. The time courses for activation of the three MAPKs by insulin, however, are distinctly different. Activation of JNK occurs most rapidly, within seconds. Activation of p38 by insulin follows that of JNK, within minutes. Activation of ERK occurs last, 4 min after administration of insulin. The temporal relationship between the activation of ERK, JNK, p38 and the downstream elements p90rsk and PP-1 in vivo suggest that JNK, but neither ERK nor p38 MAPKs, mediates insulin activation of glycogen synthase in vivo. Activation of JNK by anisomycin in vivo mimics activation of glycogen synthase by insulin. Challenge by anisomycin and insulin, in combination, are not additive, suggesting a common mode of glycogen synthase activation. The p90rsk isoform rapidly activated by insulin is identified as RSK3. In addition, RSK3 can be activated by JNK in vitro. Based upon these data a signal linkage map for activation of glycogen synthase in vivo in skeletal muscle can be constructed in which JNK mediates activation of glycogen synthase via RSK3.

Insulin, acting via its cognate tyrosine kinase receptor, provokes a wide array of biological responses in vivo (for review, see Refs. 1 and 2), including increased protein, lipid and glycogen synthesis, glucose uptake, and counterregulation of lipolysis. The biological effects of insulin are mediated via net dephosphorylation/phosphorylation on Ser/Thr residues of key regulatory enzymes (3). Examples of proteins undergoing dephosphorylation include glycogen synthase, pyruvate kinase, and pyruvate dehydrogenase. Proteins displaying increased serylthreonyl-phosphate content include acetyl-CoA carboxyl-ase, ATP-citrate lyase, and S6 ribosomal protein. Several Ser/ Thr kinase networks including p70 S6 kinase, MAPK, 1 and p90 rsk have been implicated in mediating the actions of insulin in vivo (2). Identifying the precise linkage between any one of these kinase cascades and the multitude of insulin effects remains a challenging problem.
Skeletal muscle is a primary site for insulin-stimulated glucose uptake, accounting for greater than 80% of insulin-stimulated glucose uptake in humans (4). In addition, glycogen synthesis in skeletal muscle accounts for a large fraction of insulin-stimulated glucose disposal. Insulin stimulates glycogen synthesis via a coordinated response promoting inhibition of GSK-3, a Ser/Thr kinase (5), and activation of PP-1, a Ser/ Thr phosphatase, (6), resulting in net dephosphorylation of glycogen synthase (5,7). Both PP-1 and GSK-3 are substrates for p90 rsk (8,9), an ϳ90-kDa S6 kinase isoform (10) displaying insulin-sensitive kinase activity in skeletal muscle (9). In addition, both p70 S6 kinase and c-Akt/PKB have been shown to inhibit GSK-3 kinase activity (8,11,12). A physiological role of c-Akt/PKB in insulin-stimulated glycogen synthesis remains an interesting, but untested hypothesis. Several studies have shown that inhibition of p70 S6 kinase activity with rapamycin has no effect on insulin-stimulated glycogen synthesis in L6 myotubes (12,13) or mouse skeletal muscle in vivo (14), suggesting no major role for p70 S6 kinase in the regulation of glycogen synthase by insulin. Based upon in vitro studies, ERK1 and ERK2 have been proposed to be the upstream regulators of p90 rsk (9,15). Thus, a linear signaling cascade from the activated insulin receptor to dephosphorylated, active glycogen synthase via the Ras-MAPK/ERK network has been proposed (16,17).
Recently the role of ERKs in insulin action in vivo has been subject to debate (17). Saltiel and coworkers (18) have shown that inhibition of MEK1 and MEK2, the upstream regulators of ERK, has no effect on several insulin-mediated responses, including glycogen synthesis and glucose uptake in adipocytes. Experiments using the same inhibitor to examine insulin-mediated inhibition of GSK-3 in L6 myotubes have produced similar results (12), suggesting alternative pathways to ERK signaling. One possible explanation for these findings may be the involvement of MAPKs other than ERK. Recently, several ad-* This work was supported in part by grants from the NIDDK, National Institutes of Health (to C. C. M.), American Cancer Society (to C. C. M.), and the NCI, National Institutes of Health (to R. J. D.). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 1 The abbreviations used are: MAPK, mitogen-activated protein kinase; ERK, extracellular response kinase; JNK, jun N-terminal kinase; MEK, MAPK/ERK kinase; RSK, ribosomal S6 kinase; GSK-3, glycogen synthase kinase-3; PKB, protein kinase B; PP-1 and -2A, protein phosphatase-1 and -2A, respectively; MBP, myelin basic protein; UDP-Glc, uridine 5Ј-diphosphoglucose; PMSF, phenylmethylsulfonyl fluoride; DTT, dithiothreitol; Glu-6-P, glucose-6-phosphate; PAGE, polyacrylamide gel electrophoresis; MES, 4-morpholineethanesulfonic acid; FPLC, fast protein liquid chromatography; GST, glutathione S-transferase; ATF2, activating transcription factor 2. ditional members of the MAPK family have been cloned, including jun N-terminal kinases (JNKs) -1-3 (19 -22) and p38 (23,24), the mammalian homolog of the protein kinase hog1 in Saccharomyces cerevisiae. In mammalian cell culture systems, both p38 and JNK have been shown to be activated in response to hyperosmotic stimuli (25,26), cytokines (22, 26 -28), protein synthesis inhibitors (29), and ultraviolet light (19) and are perhaps involved in apoptosis (30 -32) and the heat shock response (33).
Reports of insulin-stimulated MAPK activity (34) preceded cloning of JNK (19 -22) and p38 (23,24), necessitating reevaluation of the role of MAPK family members in insulin action. In the present study, we create a signal linkage map in vivo defining the temporal relationships between insulin action and mediation by three prominent members of the MAPK family, i.e. ERK, JNK, and p38.

EXPERIMENTAL PROCEDURES
Materials-BDF1 mice were obtained from Taconic Farms, Inc. (Germantown, NY). [␥-32 P]ATP (ϳ7000 Ci/mmol) was purchased from ICN Biomedicals Inc. Myelin basic protein was obtained from Sigma. Anti-ERK1 and -ERK2 monoclonal antibodies were purchased from Zymed Laboratories, Inc. (San Francisco, CA). Alkaline phosphatase-conjugated goat anti-rabbit and anti-mouse antibodies, protein kinase A inhibitor peptide, phosphorylase kinase, and glycogen phosphorylase b were purchased from Life Technologies, Inc. Source 15Q anion-exchange and Source 15S cation-exchange media, and glutathione-Sepharose were from Pharmacia Biotech Inc. The p90 rsk kinase S6 substrate peptide, anti-p90 rsk and RSK3 polyclonal antibodies, and Protein A/G Plus-agarose were purchased from Santa Cruz Biotechnology (Santa Cruz, CA). GSK-3␤ antiserum was a kind gift from Dr. E. Krebs (Univ. of Washington, Seattle, WA). The GSK-3␤ substrate, phospho-glycogen synthase peptide-2, was purchased from Upstate Biotechnology Inc. (Lake Placid, NY). Antibodies to c-jun and Ser-63 phospho-specific antibodies to c-jun were purchased from New England Biolabs Inc. (Beverly, MA). Production of antibodies specific for p38 and JNK1 were described previously (26). Monoclonal JNK antibodies were purchased from Pharmingen (San Diego, CA.). All other reagents were of the highest quality commercially available.
Insulin Administration in Vivo-BDF1 mice were fed ad libitum and maintained on normal light/dark cycles. The evening prior to experiments, mice were allowed access to drinking water only. Experiments were routinely conducted between 8:00 and 10:00 a.m. Mice (6 -8 weeks) were anesthetized and insulin or anisomycin was administered via intravenous injection as described previously (35). Control mice received injections with vehicle alone.
Source 15Q FPLC Chromatography-After administration of insulin in vivo, hind limb skeletal muscle was removed and immediately homogenized for 30 s in ice-cold buffer A supplemented with 0.5% Triton X-100, 5 g/ml aprotinin, 5 g/ml leupeptin, and 0.1 mM PMSF using a Willems Polytron tissue homogenizer adjusted to setting 6. After 10 min on ice, the homogenate was centrifuged at 14,000 ϫ g for 15 min, and the supernatant was transferred to a fresh tube for protein determinations (36). Five mg of soluble protein resuspended in 1 ml of buffer A was loaded onto a Source 15Q anion-exchange resin column (1.5 ml) equilibrated in buffer A. After washing the column with 3 column volumes of buffer A, the bound proteins were eluted with a 25-ml linear gradient of NaCl (0 -1 M) in buffer A. Twenty-five fractions (1 ml) were collected at a flow rate of 1 ml/min. ERK and p38 Protein Kinase Assays-Aliquots of column fractions (15 l) were mixed with 5 l of 4 ϫ kinase assay buffer containing either MBP (2 mg/ml) or GST-ATF2 (2 g) as substrates. The reactions were started with the addition of [␥-32 P]ATP (10 Ci/tube, 100 M final concentration). After 15 min at 30°C, the reactions were terminated with 5 l of 5 ϫ Laemmli solution. The samples were heated at 100°C for 3 min and briefly centrifuged. The denatured samples were subjected to SDS-PAGE on 14 (MBP) or 12% (GST-ATF2) acrylamide gels. Upon staining with Coomassie Brilliant Blue R250 and autoradiography, the bands corresponding to either MBP or GST-ATF2 were excised from the gel. Gel slices were solubilized and then the 32 P label was quantified by liquid scintillation counting. The autoradiographs were subjected to scanning densitometry, and the densities of products were quantified using a Bio-Rad GS-670 imaging densitometer fitted with Molecular Analyst software.
jun N-terminal Protein Kinase Assays-Solid-state protein kinase assays using GST-fusion proteins encoding for the N-terminal region of jun-(1-79) were performed as described previously (37). Briefly, an aliquot of soluble skeletal muscle protein (2.5 mg) in ice-cold buffer B supplemented with 5 g/ml leupeptin and 5 g/ml aprotinin was mixed with 20 g of GST-jun prebound to GSH-Sepharose. After 2 h at 4°C with continuous mixing, the samples were centrifuged at 14,000 ϫ g for 2 min and washed twice in buffer B and once with JNK assay buffer. The samples were then resuspended in 20 l of JNK assay buffer and placed on ice. The kinase reactions were started with the addition of [␥-32 P]ATP (10 Ci/tube, 100 M final concentration). After 15 min at 30°C, the reactions were terminated with 5 l of 5 ϫ Laemmli buffer, heated at 100°C for 3 min, and then subjected to SDS-PAGE on 12% acrylamide gels. Upon staining with Coomassie Brilliant Blue R250 and autoradiography, the bands corresponding to GST-jun were excised from the gel. The gel slices were solubilized and then the 32 P label was quantitated by liquid scintillation counting. In addition, autoradiographs were subjected to scanning densitometry and quantification as described above.
Immunoprecipitations-Mice were administered with insulin or anisomycin as described, and hind limb skeletal muscle was removed at the times indicated. The skeletal muscle biopsies were immediately homogenized in ice-cold buffer C supplemented with 5 g/ml aprotinin, 5 g/ml leupeptin, and 1.0 mM PMSF. After 10 min on ice, the lysate was centrifuged at 14,000 ϫ g for 10 min. An aliquot (800 g of protein) of the clarified lysate was mixed with the indicated antibodies and resuspended to 0.5 ml in the above buffer. After 2 h at 4°C with constant rotation, 40 l of Protein A/G Plus-agarose was added, and the incubation was continued for an additional 1.5 h. The immunoprecipitates were pelleted by centrifugation at 14,000 ϫ g for 5 min. The pellets were washed 3 times with buffer C supplemented with 50 mM NaCl. Immunoprecipitates were then washed twice with 1 ϫ kinase assay buffer and resuspended in a final volume of 40 l of the same buffer. A 20-l aliquot was subjected to SDS-PAGE on 7.5% acrylamide gels and immunoblot analysis.
Immunoblot Analysis-Portions of column fractions 1-12 were precipitated with trichloroacetic acid and prepared for SDS-PAGE on 12% acrylamide gels, as previously described (38). The proteins resolved by SDS-PAGE were then transferred to nitrocellulose and probed with the indicated antibodies, as previously described (39). Immunoreactive proteins were made visible using secondary antibodies coupled to alkaline phosphatase and colorimetric development.
Immunocomplex Protein Kinase Assays-The immunoprecipitate was resuspended in 1 ϫ kinase assay buffer, and a 20-l aliquot was mixed with or without substrate (S6 peptide for p90 rsk , 500 M final concentration or glycogen synthase peptide-2 for GSK-3␤, 200 M final concentration). The reactions were initiated with the addition of [␥-32 P]ATP (10 Ci/tube, 100 M final concentration). After 15 min at 30°C, the reactions were terminated with 10 l of trichloroacetic acid (25%). The reactions were spotted onto P-81 phosphocellulose paper discs and washed twice in 75 mM H 3 PO 4 , 5 min for each wash. The discs were washed once with acetone and air dried, and the 32 P incorporated into protein was measured by liquid scintillation counting. Kinase activity termed "specific" was determined by subtracting the radioactivity detected in the absence of from that detected in the presence of substrate.
Glycogen Synthase Assay-BDF1 mice, which had been starved overnight for 18 h, were administered with insulin or anisomycin in vivo. At the times indicated, a portion of the gastrocnemius muscle was removed and homogenized in ice-cold buffer D supplemented with 5 g/ml aprotinin, 5 g/ml leupeptin, and 0.1 mM PMSF. The homogenate was centrifuged at 16,000 ϫ g for 15 min at 4°C. A 200-l aliquot of the clarified extract was centrifuged (1500 ϫ g for 5 min) through a 2-ml column of Sephadex G-50 (superfine) that had been preequilibrated in buffer F to remove low molecular weight substances that might interfere with the assay. Glycogen synthase activity was determined by measuring the incorporation of [ 14 C]UDP-Glc into glycogen in the absence or presence of 10 mM Glu-6-P as described (40). The activity ratio of glycogen synthase (Ϫ/ϩ10 mM Glu-6-P) was calculated.
Protein Phosphatase-1 Assay-After an overnight fast, mice were treated with insulin or anisomycin as described above. Hind limb skeletal muscle was removed at the times indicated and immediately homogenized in 1 ml of ice-cold buffer D supplemented with 5 g/ml aprotinin, 5 g/ml leupeptin, and 0.1 mM PMSF. The homogenate was centrifuged at 16,000 ϫ g for 10 min at 4°C. Protein phosphatase-1 activity was measured at 30°C in the presence of okadaic acid (3 nM) as described (41), using 32 P-labeled glycogen phosphorylase a as a substrate. Endogenous PP-2A activity is completely inhibited using 3 nM okadaic acid, whereas PP-1 activity is largely unaffected (42). PP-1 activity was calculated by measuring the release of 32 P-labeled inorganic phosphate from the 32 P-labeled substrate.
Source 15S Chromatography-Flow-through fractions (3 mg of protein) from the Source 15Q column were diluted to 10 ml with equilibration buffer E and applied to a 1.0-ml Source 15S cation-exchange resin column. After washing the column, bound proteins were eluted with a 25-ml gradient of NaCl (0 -1 M) in buffer E). One-ml fractions were collected at a flow rate of 1 ml/min. Aliquots (15 l) of each fraction were mixed with 4 ϫ kinase assay buffer and S6 peptide (500 M final concentration). The reactions were initiated with the addition of [␥-32 P]ATP (10 Ci/tube, 100 M final concentration). After 15 min at 30°C, the reactions were terminated with 10 l of trichloroacetic acid (25%). The reactions were spotted onto P-81 phosphocellulose paper discs and washed twice in 75 mM H 3 PO 4 , 5 min for each wash. The discs were washed once with acetone and air dried, and the incorporated 32 P was measured by liquid scintillation counting.
RSK3 Kinase Assays-JNK was immunoprecipitated from skeletal muscle homogenates (2.5 mg of soluble protein) prepared from mice receiving either vehicle or insulin for 0.5 min in vivo, as described above. The immunoprecipitate was resuspended in 150 ml of ice-cold 1 ϫ kinase assay buffer. Aliquots (15 l) of fraction 7 from Source 15S column chromatographs of skeletal muscle from mice treated with vehicle alone (i.e. latent RSK3) were assayed for S6 kinase activity in the absence or presence of JNK (5 l). The JNK immunoprecipitates were assayed for S6 kinase activity in the absence of any added fraction 7 as a control. Reaction conditions were identical to those described above.
RSK3 Immunodepletion Experiments-Aliquots (100 l) of latent RSK3 that had been isolated by Source 15S column chromatography of skeletal muscle homogenates prepared from mice treated without insulin were resuspended to 500 l with ice-cold buffer C supplemented with 5 g/ml aprotinin, 5 g/ml leupeptin, and 1.0 mM PMSF in duplicate. RSK3 antibody (1:100 dilution) or preimmune serum was added to the appropriate tubes and incubated at 4°C for 2 h with constant rotation. Protein A/G Plus-agarose was added to both tubes, and the incubation was carried out for an additional 1.5 h. The immunocomplexes were pelleted by centrifugation, and the supernatant removed and stored at Ϫ70°C. The pelleted immunoprecipitate was washed twice with ice-cold buffer C and then resuspended in 1 ϫ kinase assay buffer. Aliquots (15 l) of the immunoprecipitate and of the supernatant were assayed for S6 kinase activity in the absence or presence of immunoprecipitated JNK, as described above.

RESULTS AND DISCUSSION
Activation of ERK, p38, and JNK by Insulin in Vivo-Insulin was administered intravenously to BDF1 mice. Using MBP as the substrate, two peaks of insulin-stimulated kinase activity were detected following Source 15Q anion-exchange column chromatography of homogenates prepared from skeletal muscle treated with insulin for 0.5-4 min (Fig. 1A). Immunobloting demonstrates that the MBP kinase activity present in fractions 4 and 5 of peak I co-elutes with ERK1 and ERK2 (Fig. 1B). Immunoblotting with antibodies to p38 revealed immunoreactivity in fractions 7 and 8 (Fig. 1A, peak II) (Fig. 1C), suggesting, for the first time, that insulin may activate this member of the MAPK family. Using a GST fusion protein containing the N-terminal 109 amino acids of ATF2 as a p38-specific substrate (26), we examined the activation of p38 by insulin. JNK, which also phosphorylates ATF2 (43), was not detected in fractions 7 and 8 (not shown). The time course for activation of p38 using GST-ATF2 as substrate (Fig. 1D) agrees well with the time course obtained with the MPB substrate (peak II, Fig. 1A). The presence of insulin-sensitive MBP-and ATF2-directed kinase activity and the positive staining with anti-p38 antibodies in fractions 7 and 8 demonstrate p38 activation as a novel pathway for insulin signaling in skeletal muscle in vivo.
Time courses for the activation of ERK and p38 by insulin display similar kinetics (Fig. 1, E and F). Insulin promotes a robust, 8-fold activation of ERK activity that was first evident at 4 min after insulin administration (Fig. 1E). Whereas activation of ERK by insulin was not detectable at 2 min, activation of p38 was 1.3-fold over basal by 2 min (Fig. 1F). By 4 min, following administration of insulin, p38 activation was nearly 4-fold (Fig. 1F).
Using rat adipocytes, Kim and Kahn (44) reported the Ser/ Thr phosphorylation of c-jun at 1 min after stimulation by insulin. We explored the activation of JNK in vivo using the solid-state kinase assay (37) and homogenates of skeletal muscle. A GST fusion protein harboring the first 79 amino acids of jun was employed as a substrate. Insulin administration in vivo stimulated JNK activation (Fig. 2). JNK activity was essentially nil in the absence of insulin stimulation (T ϭ 0 min). At the shortest sampling point measurable (30 s), challenge with insulin provoked an ϳ4-fold increase in JNK activity (Fig.  2, A and B). However, the stimulation of JNK by insulin in skeletal muscle in vivo was found to be a transient response. The peak response was obtained at the earliest sampling, and JNK activity returned to nearly basal levels within 4 min of insulin administration.
Several growth factors including insulin have been shown to increase AP-1 activity in vivo (45)(46)(47). The increased activity can be attributed to, in part, phosphorylation of c-jun on Ser residues 63 and 73 in the N-terminal transactivation domain (48,49). Using antibodies specific for the Ser-63-phosphorylated form of c-jun, we analyzed c-jun phosphorylation in skeletal muscle from mice administered with insulin in vivo. Immunostaining of c-jun immunoprecipitates with antibodies specific for the phosphorylated Ser-63 residue demonstrates increased phosphorylation in response to insulin (Fig. 2C). Levels of c-jun, in contrast, remain constant during this period. Increased phosphorylation of the Ser-63 residue of c-jun is evident as early as 1 min and is maximal within 2 min of administration of insulin. These results agree well with the time course of JNK activation (Fig. 2B) and establish c-jun in skeletal muscle as one of the earliest MAPK targets for insulin action in vivo.
Activation of Glycogen Synthase by Insulin in Vivo-Activation of glycogen synthase by insulin requires net dephosphorylation of glycogen synthase, thought to involve an MAPK-dependent step. The MAPKs involved would activate p90 rsk , provoking activation of PP-1 (responsible for dephosphorylation of glycogen synthase), while inhibiting GSK-3 (responsible for phosphorylation of glycogen synthase). The temporal relationships between activation of these three downstream, insulin-sensitive elements and the time courses for insulin activation of each of the MAPKs were analyzed. Skeletal muscle displayed a sharp increase in glycogen synthase activity in response to insulin (Fig. 3A). Increased glycogen synthase activity is detected as early as 2 min after administration of insulin and is essentially the same at 4 min (Fig. 3A). Activation of p90 rsk , in contrast, is detectable within 30 s, much earlier than reported previously (9,50), and is maximal by 2 min (Fig. 3B). Activation of skeletal muscle PP-1, following insulin administration, is nearly as rapid as activation of p90 rsk (Fig. 3C). Maximal activation of PP-1 by insulin (ϳ1.75-fold increase over basal) is not observed until 2 min after insulin injection. By 4 min, PP-1 activity displays a decline from the peak level, but remains significantly elevated (1.5-fold). Inhibition of GSK-3 activity is first detected at 1 min following insulin injection, decreasing the GSK-3 activity by 40% (Fig.  3D). By 2 min after insulin administration, inhibition of GSK-3 is maximal. The time courses displayed for the three MAPKs and the profile for insulin activation of downstream elements responsible for activation of glycogen synthase suggest a prominent role for JNK, but not for ERK or for p38.
Role of p70 S6 Kinase in Activation of Glycogen Synthase by Insulin-Rapamycin has been shown to inhibit p70 S6 kinase activity (51). Inhibition of p70 S6 kinase activity with rapamycin in vivo was achieved with 25 g of rapamycin (intramuscular injection). Although effectively blocking p70 S6 kinase activation, rapamycin does not alter either basal glycogen synthase activity or the ability of insulin to activate glycogen synthase. Basal glycogen synthase activity in the absence and presence of rapamycin is unchanged, 0.11 Ϯ 0.02 and 0.12 Ϯ 0.0 (n ϭ 3), respectively. Glycogen synthase activity ratios from mice treated with insulin in the absence and presence of rapa- mycin are similar also, 0.35 Ϯ 0.03 and 0.38 Ϯ 0.05 (n ϭ 3), respectively. The inability of rapamycin to alter insulin activation of glycogen synthase demonstrates that p70 S6 kinase does not contribute to insulin signaling to glycogen synthase in skeletal muscle in vivo.
Activation of JNK by Anisomycin Mimics Insulin Action-Although the temporal relationship between activation of JNK and of glycogen synthase by insulin is consistent with a role for JNK, the hypothesis was tested further using compounds shown to selectively activate specific members of the MAPK family. Protein synthesis inhibitors anisomycin (52,53), and to a lesser extent cycloheximide (54), promote selective activation of JNK, but not ERK. The activation of JNK occurs at concentrations of anisomycin less than that necessary to block protein synthesis (55), although varying with cell type (56). Intravenous administration of anisomycin increases glycogen synthase activity in a dose-and time-dependent manner (Fig. 4, A and  B). Administration of a sub-maximal dose of anisomycin (2 g/g of body weight) in combination with insulin fails to provoke an additive increase in glycogen synthase activity, suggesting that the two stimuli may share a common mode of glycogen synthase activation (Fig. 4A). Anisomycin was found to be insulinomimetic with respect to the downstream elements, p90 rsk , PP-1, and GSK-3 (Fig. 5, A-C). Anisomycin administration activates p90 rsk and PP-1 while inhibiting GSK-3 activity.
We tested directly whether activation of MAPKs by anisomycin in the skeletal muscle is selective for JNK. Anisomycin promotes a robust activation of JNK, expressed at 4 min after intravenous administration (Fig. 6, A and B). JNK activity is increased 4-fold over basal by anisomycin. ERK and p38 activities, in contrast, are not altered in mice treated with anisomycin (Fig. 6C), confirming the selective nature of this JNK activator in vivo in skeletal muscle. Thus, anisomycin activates glycogen synthase (Fig. 4B) in the absence of increases in the activities of either ERK1 and ERK2 or p38 (Figs. 1A and 5C). These data support a central role for JNK as a mediator of insulin action with respect to skeletal muscle glycogen synthase. Insulinomimetic effects of anisomycin were reported earlier in L6 muscle cells in which recruitment of GLUT4 glucose transporters to the cell surface and enhanced glucose uptake were observed following challenge with anisomycin (57). Differences in the time courses for the action of insulin versus anisomycin likely reflect differences in their solubilities and sites of action. The current work extends these early studies performed in vitro by providing evidence that anisomycin activates skeletal muscle glycogen synthase in vivo via the JNK signaling pathway.
Activation of RSK3 in Skeletal Muscle by Insulin in Vivo and by JNK in Vitro-Two isoforms of p90 rsk , RSK1 and RSK2, have been identified (58 -61). Both isoforms are expressed in skeletal muscle and are activated in response to several growth factors including insulin (58 -60). Activation can be achieved with ERK in vitro, identifying ERK as an upstream activator of RSK1 and RSK2 (9,15). Several lines of evidence suggested subsequently that RSK2 was the p90 rsk isoform mediating activation of glycogen synthase by insulin in skeletal muscle, thus creating the model for a linear cascade involving ERK and RSK2 in mediating activation of glycogen synthase by insulin (16,17). This model has been challenged by recent studies of the effects of MEK1 and MEK2 inhibitors on glycogen synthase activation, demonstrating that neither ERK nor RSK1 or RSK2 have major roles in activation of glycogen synthase by insulin (12,18).
The data presented here reveal in vivo activation of a p90 rsk mediated by JNK, but not by either ERK or p38. Additionally, RSK2 is not activated by JNK (p54 MAPK) in vitro (54). Recently, another p90 rsk , RSK3, has been identified that, like RSK1 and RSK2, is expressed in skeletal muscle and can be activated by several growth factors including insulin when overexpressed in a variety of cell lines (61)(62)(63). RSK3 phosphorylates the regulatory subunit of glycogen-bound PP-1, a physiological regulator of glycogen synthase in vivo (63). In contrast to RSK1 and RSK2, RSK3 is not activated by ERK in vitro (62). Regulation of RSK3 activity by other MAPKs has not been reported. Based upon our data and these other results, we explored whether or not RSK3 activity of skeletal muscle is increased after challenge with insulin in vivo. Second, if RSK3 were activated by insulin, we explored whether the temporal activation of RSK3 is consistent with the time course established earlier for activation of glycogen synthase in skeletal muscle. Finally, we sought to explore if RSK3 is activated by JNK in vitro.
For these studies, we chose to isolate RSK3 by a series of chromatographic steps. Skeletal muscle homogenates were prepared from mice receiving insulin or vehicle for up to 4 min in vivo, as before. Detergent extracts of skeletal muscle (0.5% Triton X-100, final concentration) were applied first to a Source 15Q FPLC column, followed by chromatography of the flowthrough fractions on a Source 15S FPLC column. Eluent fractions from the Source Q column were assayed for S6 peptidedirected kinase activity and subjected to immunoblot analysis with RSK3 antibodies. Although two major peaks of S6 peptide kinase activity elute from the Source Q column, no RSK3 immunoreactivity was observed in either (not shown). Assaying the flow-through fractions from the Source Q column reveals significant insulin-stimulated S6 kinase activity that does not bind to the Source Q column, suggesting the presence of additional S6 kinases (Fig. 7A). Maximal S6 kinase activity is observed in the flow-through fraction of muscle homogenates prepared from skeletal muscle of mice 1 and 2 min after administration of insulin in vivo. This temporal increase in S6 kinase activity agrees well with insulin-stimulated p90 rsk activity (Fig. 3B).
To characterize the S6 kinase activity further, the flow-through fractions prepared from skeletal muscle of mice that were treated with or without insulin for 1 min (maximal S6 kinase activity) were fractionated by FPLC, on a Source 15S cation-exchange resin. Proteins bound to the Source 15S matrix were eluted with a 0 -1 M linear gradient of NaCl, and the eluent fractions were assayed for S6 peptide-directed kinase activity (Fig. 7B). A single peak of insulin-stimulated S6 kinase activity eluting in fractions 6 -8, is obtained after fractionation on the Source 15S column, (Fig. 7B). Immunoblotting of fractions 5-9 with RSK3 antibodies demonstrates prominent positive immunoreactivity in fraction 7, lesser staining in fractions 6 and 8. The co-elution of RSK3, as determined by Western blotting, with the peak of insulin-stimulated S6 kinase activity in fraction 7 demonstrates that RSK3 is rapidly activated (within 1 min) in skeletal muscle treated with insulin in vivo.
To determine if RSK3 is indeed the p90 rsk activated by JNK in response to insulin treatment, we performed mixing experiments using the latent RSK3 and JNK isolated from mice administered with insulin in vivo. The latent RSK3 was isolated (fraction 7) by Source 15S chromatography of skeletal muscle from control mice treated with vehicle alone. The JNK was immunoprecipitated from skeletal muscle of mice administered with insulin in vivo for 0.5 min. Maximal insulin-stimulated JNK activity is observed at 0.5 min (Fig. 2). Interestingly, when immunoprecipitates of JNK from mice treated with or without insulin for 0.5 min are assayed for S6 kinase activity as a control, a 1.8-fold increase in S6 kinase activity is found in association with the JNK immunoprecipitate from insulintreated mice (Fig. 7C). Although speculation, the data suggest that JNK is associated in a complex with a p90 rsk isoform, as has been reported for ERK in combination with either murine (64 -65) or avian p90 rsk (66). Mixing the JNK immunoprecipitate from the insulin-treated mice with aliquots of latent RSK3 resulted in a marked increase (Ͼ4.0-fold) in S6 kinase activity, further demonstrating that RSK3 is activated by JNK in vitro. The magnitude of the increase in RSK3 activity stimulated by JNK in vitro is similar to that observed after insulin treatment in vivo.
To determine if RSK3 is the only S6 kinase activated by JNK in this fraction, we attempted to immunodeplete the latent RSK3 present in eluent (fraction 7) of the Source 15S chromatography. The extracts from skeletal muscle of mice treated without insulin separated on the Source 15S column were depleted using antibodies to RSK3. In the absence of added RSK3 antibody, the JNK-stimulated S6 kinase activity is found exclusively in the supernatant. When extracts are treated with the antibodies for RSK3, the JNK-dependent activity is found only in the immunoprecipitate. No S6 kinase activity is detected in the supernatant of the RSK3-immunodepleted fraction, demonstrating that RSK3 is responsible for the JNKstimulated S6 peptide phosphorylation. We establish that RSK3 is activated by insulin in vivo and that JNK immunoprecipitated from insulin-treated mice activates latent RSK3 in a manner that mimics insulin treatment in vivo. In combination with our description of the temporal relationship between MAPKs and activation of glycogen synthase by insulin, our data provide compelling evidence for a linkage map in skeletal muscle in vivo that places JNK and RSK3 in a linear cascade promoting activation of glycogen synthase by insulin (Fig. 8).
Recently Cross et al. (12) have shown that c-Akt/PKB can mediate insulin-dependent inhibition of GSK-3 in L6 muscle cells in an ERK/RSK2-and p70 S6 kinase-independent but wortmannin-sensitive manner, suggesting that PKB may mediate stimulation of glycogen synthase by insulin. Our data cannot rule out a possible role of PKB, operating in parallel with RSK3 to promote inhibition of GSK-3 in skeletal muscle. What further role, if any, PKB plays in insulin-stimulated glycogen synthesis in vivo requires further analysis.
We demonstrate that insulin activates multiple members of the MAPK family in vivo, including the most recently identified members JNK and p38. The distinct temporal activation of these enzymes in skeletal muscle highlights unique roles of JNK activation in insulin action. In addition, we demonstrate a previously unidentified regulation of RSK3 by insulin in skeletal muscle. The data from our in vitro studies demonstrate that RSK3 is activated by JNK, suggesting that RSK3 is a FIG. 7. RSK3 is activated by insulin in vivo and by JNK in vitro. A, measurements of S6 kinase activity in Source 15Q column flow-through fractions. Mice were treated with insulin in vivo for the times indicated. Homogenates of skeletal muscle were prepared and chromatographed by FPLC on a Source 15Q column, and the flow-through fractions were assayed for kinase activity using the S6 peptide substrate, as described under "Experimental Procedures." B, Isolation of RSK3 by Source 15S chromatography. Skeletal muscle homogenates were prepared from mice administered with insulin for the times indicated. Detergent extracts of skeletal muscle were resolved first by Source 15Q chromatography, and the flow-through fractions were prepared from mice treated with or without insulin for 1 min and then were applied to a Source 15S column. Proteins bound to the Source 15S matrix were eluted with a 0 -1 M linear gradient of NaCl (dotted line). The eluent fractions were assayed for kinase activity using S6 peptide as a substrate. Inset, eluent fractions 5-9 were subjected to SDS-PAGE on 7.5% acrylamide gels, transferred to nitrocellulose, and then immunoblotted with anti-RSK3 antibodies. C, RSK3 is activated by JNK in vitro. Latent RSK3, present in fraction 7 of Source 15S column chromatographs of skeletal muscle of mice treated without insulin, was mixed with JNK immunoprecipitated from mice administered with insulin in vivo for 0.5 min and then assayed for S6 peptide-directed kinase activity. D, immunodepletion of RSK3 abolishes p90 rsk activity isolated by chromatography on the Source 15S matrix. Aliquots of latent RSK3 were incubated in the absence or presence of RSK3 antibodies (1:100 dilution). After mixing for 2 h at 4°C, the immunocomplexes bound to Protein A/G Plus-agarose were pelleted by centrifugation. The resultant pelleted immunoprecipitate and supernatant were then assayed for S6 peptide-directed kinase activity in the absence or presence of JNK that had been immunoprecipitated from mice administered with insulin for 0. downstream effector of JNK in vivo. Based upon the current understanding of the role of Ser/Thr phosphorylation in the regulation of c-jun activity (45)(46)(47), of the control of glycogen synthesis, and of the present data, we propose a signal linkage map for insulin action in skeletal muscle (Fig. 8). The map requires coordinate control of specific kinases for activation of both c-jun and of glycogen synthase by insulin. Furthermore, the map provides for the temporal separation of the activation of downstream signaling elements (67,68). In addition to stimulating AP-1 transcriptional activity, JNK is offered as the MAPK-mediating activation of skeletal muscle glycogen synthesis by insulin.