Insulin and Exercise Decrease Glycogen Synthase Kinase-3 Activity by Different Mechanisms in Rat Skeletal Muscle*

Glycogen synthase activity is increased in response to insulin and exercise in skeletal muscle. Part of the mechanism by which insulin stimulates glycogen synthesis may involve phosphorylation and activation of Akt, serine phosphorylation and deactivation of glycogen synthase kinase-3 (GSK-3), leading to dephosphorylation and activation of glycogen synthase. To study Akt and GSK-3 regulation in muscle, time course experiments on the effects of insulin injection and treadmill running exercise were performed in hindlimb skeletal muscle from male rats. Both insulin and exercise increased glycogen synthase activity (%I-form) by 2–3-fold over basal. Insulin stimulation significantly increased Akt phosphorylation and activity, whereas exercise had no effect. The time course of the insulin-stimulated increase in Akt was closely matched by GSK-3α Ser21 phosphorylation and a 40–60% decrease in GSK-3α and GSK-3β activity. Exercise also deactivated GSK-3α and β activity by 40–60%. However, in contrast to the effects of insulin, there was no change in Ser21 phosphorylation in response to exercise. Tyrosine dephosphorylation of GSK-3, another putative mechanism for GSK-3 deactivation, did not occur with insulin or exercise. These data suggest the following: 1) GSK-3 is constitutively active and tyrosine phosphorylated under basal conditions in skeletal muscle, 2) both exercise and insulin are effective regulators of GSK-3 activity in vivo, 3) the insulin-induced deactivation of GSK-3 occurs in response to increased Akt activity and GSK-3 serine phosphorylation, and 4) there is an Akt-independent mechanism for deactivation of GSK-3 in skeletal muscle.

Insulin and contractile activity are the most biologically relevant regulators of glycogen metabolism in skeletal muscle. In the postprandial state, insulin promotes muscle glucose uptake and the storage of glucose as glycogen (1). In contracting skeletal muscles glycogenolysis is activated, whereas the period following muscle contraction is characterized by a marked increase in glycogen synthesis (2). The conversion of UDP-glucose to glycogen by glycogen synthase is the rate-limiting step in glycogen synthesis, and this enzyme is regulated by both allosteric and phosphorylation-dephosphorylation mechanisms (3). Glycogen synthase is serine phosphorylated on multiple sites, and insulin treatment results in the hierarchal dephosphorylation of several of these sites, leading to activation of the enzyme and increased glycogen synthesis (3). There is evidence for both protein phosphatase-1 activation and glycogen synthase kinase-3 (GSK-3) 1
Although there has been considerable progress in elucidating signals regulating insulin activation of glycogen synthase in various cell types, there is little understanding of the molecular mechanisms leading to stimulation of glycogen synthase with exercise in skeletal muscle. In the current investigation we compared the effects of insulin and exercise on the regulation of glycogen synthesis in skeletal muscle and determined whether the activation of glycogen synthase was accompanied by changes in the phosphorylation state and activity of Akt and GSK-3. We show that both insulin and exercise increase glycogen synthase activity and that the time course of synthase activation is closely mirrored by deactivation of GSK-3. Insulin-induced GSK-3 deactivation is associated with GSK-3␣ serine phosphorylation and increased Akt activity, whereas exercise deactivates GSK-3 in the absence of changes in serine or tyrosine phosphorylation of GSK-3 or increased Akt activity.

EXPERIMENTAL PROCEDURES
Materials-GSK-3␣ and Akt1 antibodies, Phospho-GS2 substrate peptide, and Akt/PKB-specific substrate peptide were purchased from Upstate Biotechnology Inc. (Lake Placid, NY). GSK-3␤ antibody was purchased from Transduction Laboratories (Lexington, KY). [␥-32 P]ATP was from NEN Life ScienceProducts, and protein A-and G-Sepharose beads were from Amersham Pharmacia Biotech. Reagents for protein assays and electrophoresis were purchased from Bio-Rad (Rockville Center, NY). Chemiluminescence reagents were from Amersham Pharmacia Biotech, and all other standard chemicals were obtained from Sigma.
Animals-Fed male Sprague-Dawley rats (190 g) were divided into three treatment groups: basal, exercise, and insulin. Insulin-stimulated animals were studied 5, 10, or 30 min after intraperitoneal insulin injection (maximal, 20 units/rat), and exercised animals were studied immediately after 5, 10, 30, or 60 min of treadmill running at 25 m/min, 10% grade. Serum glucose concentrations were determined by hexokinase assay (26) and serum insulin concentrations by radioimmunoassay (27). Rats were killed by decapitation, and the gastrocnemius muscles from both legs were quickly dissected, divided into red and white fractions, frozen in liquid N 2 , and stored at Ϫ80°C until processed.
Glycogen and Glycogen Synthase Assays-Approximately 60 mg of muscle tissue was Polytron homogenized in 1 ml of buffer containing 50 mM Tris-HCl, 5 mM EDTA, 100 mM sodium fluoride, pH 7.8. Glycogen synthase activity in the absence or presence of 6.7 mM glucose 6-phosphate was determined as described previously (28) and is reported as the ratio of the glycogen synthase activity in the absence of glucose 6-phosphate to that in the presence of glucose 6-phosphate (%I form). To measure muscle glycogen concentrations, homogenates were hydrolyzed in 2 N HCl, neutralized, and assayed spectrophotometrically for glucose content as described previously using a hexokinase-dependent assay kit (26).
Statistical Analysis-All data are expressed as the means Ϯ S.E. Statistical analysis was performed using SAS General Linear model. Differences were considered significant when p Յ 0.05. Fig. 1 shows the time course for the activation of the %I-form of glycogen synthase in red gastrocnemius muscle in response to treadmill running exercise and maximal insulin injection. Only 5 min of exercise was necessary to produce a significant activation of glycogen synthase, but the highest levels of activity were not observed until 30 or 60 min of running exercise (3-fold above basal). Insulin was slightly less effective in increasing glycogen synthase activity (2-fold above basal), but maximal activation occurred more rapidly (10 min). Similar results were observed in white gastrocnemius muscle (data not shown). Muscle glycogen content was decreased by exercise at all time points, whereas glycogen levels were unaffected by insulin injection (data not shown).

Exercise and Insulin Effects on Glycogen Metabolism-
Exercise Does Not Act through Akt to Regulate Glycogen Synthesis-To determine whether the magnitude and time course of glycogen synthase activation with insulin and exercise corresponded to changes in Akt, we measured both Akt activity and serine phosphorylation in red and white gastrocnemius muscle. Similar to the activation of glycogen synthase, Akt activity ( Fig. 2A) and serine phosphorylation (Fig. 2, B and C) were significantly increased with insulin in the red gastrocnemius muscle. Maximal insulin-stimulated Akt activation occurred earlier than peak glycogen synthase activation (5 min versus 10 min), and Akt phosphorylation/activity had returned to base-line levels by 30 min after insulin injection. Insulin also increased Akt activity in white gastrocnemius with a similar time course of activation, although peak stimulation was somewhat less than in the red muscle (2.5-fold above basal; data not shown). In contrast to the consistent stimulation of Akt with insulin, exercise did not alter Akt activity or serine phosphorylation in the red (Fig. 2) or white skeletal muscle (data not shown). The temporal relationship between Akt and glycogen synthase activation is consistent with the hypothesis that part of the mechanism by which insulin stimulates glycogen synthesis is through stimulation of Akt. In contrast, exercise must utilize a different mechanism to increase glycogen synthase activity.
Both Insulin and Exercise Decrease GSK-3 Activity-Consistent with the stimulation of Akt, insulin rapidly decreased GSK-3␣ and ␤ activities in red gastrocnemius muscle (Fig. 3). With 5 min of insulin stimulation, both GSK-3␣ and ␤ were decreased by approximately 40% and remained significantly lower than basal after 30 min of insulin treatment. In contrast with the lack of effect of exercise on Akt activity and phosphorylation, exercise decreased GSK-3␣ and ␤ activities in the muscle (Fig. 3). Activities of both isoforms were significantly decreased within 10 min of exercise and remained depressed with 30 and 60 min of exercise (Fig. 3). This time course of GSK-3 deactivation (Fig. 3) followed closely with the time course of activation of glycogen synthase (Fig. 2). The pattern and magnitude of decrease in GSK-3␣ activity with insulin and exercise was similar in the red and white gastrocnemius muscle. However, decreases in GSK-3␤ activity in response to insulin and exercise were highly variable in white muscle (data not shown), suggesting that GSK-3␤ may be less sensitive to metabolic alterations in the more glycolytic fibers.
Exercise Deactivates GSK-3 through a Ser 21 Phosphorylation-independent Mechanism-Phosphorylation of the GSK-3 isoforms on Ser 21 or Ser 9 results in a decrease in enzyme activity. To determine whether the insulin-and exercised-induced decreases in GSK-3 activity were associated with increased serine phosphorylation, we immunoblotted muscle lysates with an antibody that recognizes GSK-3␣ only when phosphorylated on Ser 21 . Fig. 4 shows that insulin increased GSK-3 Ser 21 phosphorylation in both red and white skeletal muscle in a time-dependent manner, strikingly similar to the deactivation of the enzyme and the activation and phosphorylation of Akt. On the other hand, the deactivation of GSK-3 activity with exercise was not accompanied by an increase in Ser 21 phosphorylation of GSK-3␣. These data provide compelling evidence that exercise regulates GSK-3 activity in skeletal muscle by a mechanism other than Ser 21 phosphorylation.
GSK-3␣ and ␤ may be constitutively active in resting cells, and this activated state may be maintained by phosphorylation on Tyr 279 and Tyr 216 , respectively. Because some studies have suggested that tyrosine dephosphorylation may regulate kinase activity in vivo (14,15,31), we determined whether insulin or exercise altered GSK-3 tyrosine phosphorylation in the muscle. Immunoblotting muscle lysates with a phosphospecific antibody that recognizes GSK-3␤ when phosphorylated on Tyr 216 failed to demonstrate any effect of insulin or exercise

FIG. 2. Time course of Akt1 activity and phosphorylation in skeletal muscle from rats following maximal insulin stimulation (open circles) or treadmill exercise (closed circles).
A, Akt1 activity expressed as percentages of basal activity, normalized to a control sample analyzed with each assay. Results are the means Ϯ S.E., n ϭ 4 -7/group. *, different from control basal (time 0), p Ͻ 0.05. B, Akt phosphospecific immunoblot. This representative phosphorimage shows an increase in Akt phosphorylation following maximal insulin stimulation, but no change following treadmill exercise. Aliquots of muscle proteins (90 g) were resolved by SDS-polyacrylamide gel electrophoresis and immunoblotted with anti-Akt-P Ser473 antibody. C, Akt1 phosphorylation expressed in arbitrary units, normalized to a reference standard used for comparison among blots. Results are the means Ϯ S.E., n ϭ 4 -7/group. *, different from control basal (time 0), p Ͻ 0.05.

FIG. 3. Time course of GSK-3␣ and GSK-3␤ activities in red skeletal muscle from rats following maximal insulin stimulation (open circles) or treadmill exercise (closed circles).
A, GSK-3␣ activity expressed as percentages of basal activity, normalized to a control sample analyzed with each assay. Results are the means Ϯ S.E., n ϭ 4 -7/group. *, different from control basal (time 0), p Ͻ 0.05. B, GSK-3␤ activity expressed as percentages of basal activity, normalized to a control sample analyzed with each assay. Results are the means Ϯ S.E., n ϭ 4 -7/group. *, different from control basal (time 0), p Ͻ 0.001. (Fig. 5). Furthermore, immunoprecipitation/immunoblotting with antibodies to phosphotyrosine, GSK-3␣, and GSK-3␤ showed significant phosphorylation in the basal state but no regulation with insulin or exercise (data not shown).

DISCUSSION
The current results demonstrate that physical exercise decreases GSK-3 activity in skeletal muscle and that the magnitude of GSK-3 deactivation is similar to that observed with maximal insulin stimulation in vivo. In addition to the novel finding that exercise decreases GSK-3 activity in muscle, our data also suggest that there are distinct mechanisms leading to GSK-3 deactivation in response to exercise and insulin. For the insulin-induced deactivation of GSK-3, the findings are consistent with an Akt-dependent phosphorylation of GSK-3 on Ser 21 . Previous studies have provided direct evidence that Akt can phosphorylate GSK-3 on Ser 9 or Ser 21 (24,31). Although we could not make this type of assessment in the intact animal, it is compelling that the time course of Akt activation and GSK-3 phosphorylation with in vivo insulin stimulation was so similar ( Figs. 2A and 4B). The magnitude and time course of GSK-3␣ serine phosphorylation also strikingly resembles the deactivation of GSK-3␣ (Figs. 3A and 4B). Furthermore, the return of Akt activity to near basal activity occurred more rapidly than the dephosphorylation of GSK-3. Thus, the pattern of regulation of muscle Akt and GSK-3 in the intact animal supports the hypothesis that insulin results in the sequential phosphorylation and activation of Akt and GSK-3 in skeletal muscle.
In contrast to the effects of insulin to cause serine phosphorylation of GSK-3, exercise did not result in Ser 21 phosphorylation of GSK-3␣. The inability of exercise to elicit GSK-3 serine phosphorylation is consistent with experiments demonstrating that Akt activity (current study and Refs. 30, 32, and 33) and PI 3-kinase activity (34) are not increased in response to treadmill running exercise or muscle contractions induced by electrical stimulation. There are other kinases that have been suggested to serine phosphorylate GSK-3 (35,36), and we have shown that two of these molecules, the p90 ribosomal S6 kinase 2 (RSK2) and the c-Jun NH 2 -terminal kinase (37), are highly activated by the same exercise protocol used in the current study. This suggests that high levels of RSK2 and c-Jun NH 2terminal kinase activity in skeletal muscle do not necessarily lead to increased Ser 21 phosphorylation of GSK-3 in vivo.
There are other examples in the literature of an Akt-independent mechanism for deactivation of GSK-3. In mouse fibroblasts, Wingless inactivation of GSK-3 is wortmannin-insensitive (38). Evidence that GSK-3␤ Ser 9 phosphorylation is not required for a decrease in enzyme activity also comes from a study showing that mutation of Ser 9 to alanine does not diminish lithium effects on GSK-3-mediated tau phosphorylation (7). Interestingly, a recent study has demonstrated that lithium increases glycogen synthase activity by a wortmannin-independent mechanism in rat skeletal muscle (39). Thus, exercise and lithium may work by a similar, serine phosphorylationindependent mechanism for deactivation of GSK-3.
In the activated state, GSK-3␣ and ␤ are tyrosine phosphorylated (Tyr 279 and Tyr 216 , respectively) (13)(14)(15), making it plausible that dephosphorylation of these residues could facilitate enzyme deactivation. Although one study has demonstrated tyrosine autophosphorylation of GSK-3 (14), a clear mechanism for tyrosine dephosphorylation in vivo has not been established, and data are conflicting regarding tyrosine dephosphorylation of GSK-3 as a result of insulin stimulation (15,31). Because we did not observe Ser 21 phosphorylation of GSK-3␣ with exercise, we hypothesized that the enzyme may be regulated by tyrosine dephosphorylation. However, we found no evidence for exercise regulation of GSK-3 tyrosine phosphorylation in the current study and found stable levels of tyrosine phosphorylation under all conditions. This raises the possibility that exercise deactivates GSK-3 by phosphorylation at an alternative site. Indeed, some isoforms of protein kinase C can phosphorylate and inactivate GSK-3, and the exact sites of phosphorylation have not been determined (15,38,40). We could also speculate that glycogen or glycogen-bound molecules function as allosteric regulators of GSK-3 activity in skeletal muscle. If glycogen itself could act as an allosteric activator of GSK-3, this would limit glycogen synthesis under conditions where glycogen is abundant. Following exercise, decreased glycogen content within the muscle fibers could reduce the allosteric activation of GSK-3, leading to increased glycogen synthesis.
Our data showing a close correlation between the exerciseand insulin-induced decrease in GSK-3 activity and increase in FIG. 4

. Time course of GSK-3␣ serine phosphorylation in red and white skeletal muscle from rats following maximal insulin stimulation (open circles) or treadmill exercise (closed circles).
A and C, GSK-3␣ phosphospecific immunoblot. These representative phosphoimages show an increase in GSK-3␣ Ser21 phosphorylation following maximal insulin stimulation but no change following treadmill exercise in red (A) and white (C) skeletal muscle. Aliquots of muscle proteins (90 g) were resolved by SDS-polyacrylamide gel electrophoresis and immunoblotted with anti-GSK-3␣-P Ser21 antibody. B, GSK-3␣ Ser21 phosphorylation in red skeletal muscle expressed in arbitrary units, normalized to a reference standard used for comparison among blots. Results are the means Ϯ S.E., n ϭ 4 -7/group. *, different from control basal (time 0), p Ͻ 0.05.
FIG. 5. GSK-3␤ Tyr216 phosphorylation assessed by phosphospecific immunoblotting in red gastrocnemius skeletal muscle from rats in the basal state, following maximal insulin stimulation, and treadmill exercise. Aliquots of muscle proteins (90 g) were resolved by SDS-polyacrylamide gel electrophoresis and immunoblotted with anti-GSK-3␤ Tyr216 antibody. This representative phosphoimage shows no change in GSK-3␤ Tyr216 phosphorylation with either treatment. glycogen synthase activity are consistent with the hypothesis that GSK-3 functions to regulate glycogen synthesis in skeletal muscle. However, recent studies in adipocytes have suggested that activation of phosphatases is the primary factor regulating glycogen synthase activity (41,42). In adipocytes isoproterenolinduced deactivation of GSK-3 is not accompanied by glycogen synthase activation (41). In 3T3L1 fibroblasts glycogen synthase activity may be regulated by GSK-3 deactivation, but when these cells differentiate into adipocytes, control of synthase with insulin stimulation appears to shift to protein phosphatase-1 activation (42). On the other hand, in cultured human myoblasts (43,44) and myotubes (43), rat epididymal fat cells (45), and 3T3L1 adipocytes (46), the PI 3-kinase inhibitors wortmannin and LY294002 block the effect of insulin to inactivate GSK-3 and increase glycogen synthase activity. Furthermore, overexpression of Akt in L6 myotubes increases glycogen synthase activity via inhibition of GSK-3 activity (47). These types of inhibitor and overexpression studies have not yet been done in adult skeletal muscle, although it is well established that insulin causes both activation of protein phosphatase-1 and deactivation of GSK-3 in this tissue (3). Based on all the available data, it is likely that glycogen synthase regulation in skeletal muscle is regulated by both phosphatase and kinase activities.
Early studies of GSK-3 demonstrated the ability of the enzyme to phosphorylate and deactivate glycogen synthase in vitro, but it is now clear that there is a wide variety of cellular substrates for GSK-3. Particularly relevant to our finding that exercise deactivates GSK-3 are data showing that GSK-3 can phosphorylate transcription factors that have AP-1 activity (8 -10, 48) and regulate the eukaryotic initiation factor eIF2B (12). Exercise causes a selective increase in gene expression and increases rates of protein synthesis in skeletal muscle (49,50), but little is known about the molecular signaling mechanisms that convert skeletal muscle contractile activity into biochemical and gene regulatory responses. The effects of exercise to deactivate GSK-3 may function as an initial signaling event that results in changes in muscle protein metabolism and gene transcription. Interestingly, we have recently shown that the mitogen-activated protein kinase (51) and c-Jun NH 2 -terminal kinase signaling cascades (52) are also increased by exercise and muscle contraction in skeletal muscle. Thus, there is likely to be convergence of multiple signals at the nucleus in response to contractile activity, with integration of these signals critical in controlling transcription and ultimately determining muscle phenotype. We are beginning to develop an understanding of molecular signaling mechanisms in contracting skeletal muscle, and the current observation that exercise decreases GSK-3 activity should be an important step toward elucidating the molecular mechanisms that regulate glycogen synthesis, protein synthesis, and gene transcription in working muscle.