Skeletal Muscle Adaptation to Exercise Training : AMP-Activated Protein Kinase Mediates Muscle Fiber Type Shift

Muscle-specific transgenic mice expressing the inactivating D157A mutation in their α2 subunit of AMPK (α2iTG) were generated on an FVB background as described previously (23). Basal and contraction-stimulated AMPKα2 activity of α2iTG mice is almost abolished, whereas basal AMPKα1 activity is normal, and contraction-stimulated AMPKα1 activity is ∼50% decreased compared with that of their wild-type littermates (23,24). The α2iTG mice used for this study were heterozygotes for the inactivating mutation. Muscle-specific R70Qγ1 transgenic mice (γ1TG) expressing the activating R70Q mutation in their γ1 subunit of AMPK were generated on an FVB background as described previously (25). Basal total AMPK activity in γ1TG mice is approximately threefold higher than that in wild-type littermates (25). The γ1TG mice used for this study were heterozygotes for the activating mutation. Physiological and behavioral characterization using the Comprehensive Lab Animal Monitoring System in both transgenic animal models showed that the respective transgene did not significantly alter locomotive activity, oxygen consumption, carbon dioxide generation, heat generation, or food and water consumption (23; M.F.H., L.A.W., L.J.G., unpublished data). Mice were housed at a constant temperature (20–22°C) with a 12 h:12 h light-dark cycle and received LabDiet rodent chow (Purina Mills, St. Louis, MO) and water ad libitum.

Seven-week-old female α2iTG and γ1TG mice and corresponding wild-type littermates were randomly assigned to housing in individual cages with or without running wheels (Nalgene, Rochester, NY) for a total of 6 weeks. Completed wheel revolutions and time spent running were continuously monitored. On day 42, all mice were removed from their cages 3–4 h before being killed by cervical dislocation. Preliminary experiments from our laboratory have shown that wheel cage running has the most pronounced training effects (as determined by increases in citrate synthase activity) in triceps brachii. We therefore chose to use this muscle for our studies. After the mice were killed, triceps muscles from both forelegs were immediately removed. One triceps of each animal was snap frozen in liquid nitrogen and stored at −80°C. The other triceps was embedded in Tissue-Tek O.C.T. compound (Sakura Finetek, Torrance, CA), quick frozen in isopentane cooled by liquid nitrogen, and stored at −80°C until cryosections were prepared for histology. All experiments were in accordance with guidelines of the Institutional Animal Care and Use Committee of the Joslin Diabetes Center and the National Institutes of Health.

At weeks 0, 2, 4, and 6, blood glucose, body weight, and food consumption were recorded. Blood glucose measurements were performed between 3 and 4 p.m. by obtaining blood samples from the tail of fully conscious mice, using a OneTouch Ultra portable glucometer (Lifescan, Mipitas, CA). Body weight and food consumption were determined using a digital scale (model CS200; Ohaus, Pine Brook, NJ).

Frozen muscles were pulverized and homogenized as described previously (26), and protein concentrations were determined via the Bradford assay. Skeletal muscle proteins (40 μg) were resolved by SDS-PAGE for Western blot analysis (27). Antibody-bound proteins were visualized on film using chemiluminescence detection reagents (Perkin-Elmer Life Sciences, Boston, MA). Protein bands were scanned with an ImageScanner (Amersham Biosciences) and quantitated by densitometry (FluorChem 2.0; Alpha Innotech, San Leandro, CA). Commercially available primary antibodies were PGC-1α (Calbiochem, San Diego, CA), hexokinase II (Santa Cruz Biotechnology, Santa Cruz, CA), and GLUT4 (Chemicon International, Temecula, CA). The PGC-1β antibody was a gift from Dr. Bruce Spiegelman and was produced as described (28).

A 10-mg portion of pulverized triceps from each animal was homogenized on ice in a lysis buffer (pH 6.5) containing 300 mmol/l KCl, 100 mmol/l KH₂PO₄, 50 mmol/l K₂HPO₄, and 1 mmol/l EDTA and then centrifuged at 12,000g for 1 min. Total protein content was assessed via the Bradford assay. Electrophoretic separation of MHC isoforms was performed as described previously (29) with slight modifications. Briefly, 1 μg of skeletal muscle proteins was resolved using an SDS-PAGE gel (0.75 mm) consisting of a stacking gel with 4% acrylamide-bis (50:1) and 30% glycerol and a separating gel with 8% acrylamide-bis (50:1) and 30% glycerol. The upper running buffer consisted of 300 mmol/l Tris (base), 450 mmol/l glycine, 0.3% SDS, and 10 mmol/l mercaptoethanol. The lower running buffer consisted of 50 mmol/l Tris (base), 75 mmol/l glycine, and 0.05% SDS. The gels were run on an electrophoresis Minigel system (Bio-Rad Laboratories, Hercules, CA) at 4°C at a constant current of 6 mA/gel for 25 h. After the electrophoresis, the gels were fixed for 20 min in 7% acetic acid and 50% methanol, stained for 1 h with GelCode Blue Stain Reagent (Pierce Biotechnology, Rockford, IL), and washed with double-distilled H₂O overnight. Gels were scanned with the ImageScanner, and the relative proportion of each MHC isoform was quantitated by densitometry. Bands were identified according to their migration characteristics as described previously (29). In triceps muscle, electrophoresis resulted in two distinct bands. The upper band corresponded to combined MHC IIa and IIx (referred to as MHC IIa/x) and the lower band to MHC IIb. In soleus muscle, an additional lower band was identified as MHC I, and, as expected, this band was not present in triceps, which is a mainly glycolytic muscle as determined by ATPase staining.

Mid-belly cross-sections of triceps muscle were cut at 8 μm in a cryostat (−20°C). After drying for 5 min at room temperature, the sections were incubated for 30 min at 37°C in the incubation solution (pH 7.6) containing 6.5 mmol/l sodium phosphate monobasic, 43.5 mmol/l sodium phosphate biphasic, 0.6 mmol/l nitroblue tetrazolium (N6876; Sigma), and 50 mmol/l sodium succinate. The sections were then rinsed three times for 30 s each in physiological saline and for 5 min in 15% ethanol and coverslipped with Faramount aqueous mounting medium (DakoCytomation) (30).

Citrate synthase activity was determined in muscle lysates spectrophotometrically at room temperature (412 nm) to detect the transfer of sulfhydryl groups to 5,5′-dithiobis-2-nitrobenzoate using saturating concentrations of substrates and cofactors: 80 mmol/l Tris-HCl, 0.1 mmol/l 5,5′-dithiobis-2-nitrobenzoate, 0.4 mmol/l acetyl-CoA, and 0.6 mmol/l oxaloacetate (31). Citrate synthase activity was calculated as micromoles per minute per gram of protein.

Data are means ± SE. All data were compared using two-way ANOVA and Tukey's post hoc analysis. Pearson's correlation coefficients were used to describe the linear association between variables. The differences between groups were considered significant when P < 0.05.

Body weight and blood glucose concentrations were not different between the study groups throughout the training period (Tables 1 and 2). All training groups showed a tendency for an increase in food intake compared with the sedentary control groups, which was statistically significant in the trained α2iTG and γ1TG mice. Running distances and running speeds did not differ between transgenic mice and their wild-type littermates and were maintained throughout the training period (Tables 1 and 2 and Figs. 1A and 2A).

We hypothesized that one of the responses of skeletal muscle to chronic intermittent (exercise training) or chronic sustained (γ1TG mice) AMPK activation may be the mediation of muscle fiber type transformation. To study fiber type changes, we performed electrophoretic separation of MHC isoforms in triceps muscle, a mainly glycolytic muscle with very few type I fibers, as determined by ATPase staining. Electrophoretic separation of MHC isoforms resulted in two distinct bands that were identified according to their migration characteristics as previously described (29). The lower band corresponded to MHC IIb, and the upper band corresponded to a combination of MHC IIa and IIx (Fig. 1B). Further separation of MHC IIa and IIx was not achieved, a difficulty that has been reported earlier (32). To signify that the highest migrating band may contain both MHC IIa and IIx isoforms, we refer to this band as IIa/x.

In wild-type mice, exercise training significantly increased the relative proportion of type IIa/x fibers (Fig. 1B). Exercise training also induced an increase of type IIa/x fibers in α2iTG mice. However, the proportion of type IIa/x fibers in trained α2iTG mice was significantly lower than that in their trained wild-type littermates. Some of the variability in fiber type adaptations in trained wild-type mice can be explained by the differences in total running distance. Figure 1C illustrates the fact that running distance correlated strongly with the relative proportion of type IIa/x fibers in wild-type mice. In α2iTG mice, this correlation was absent.

In a second approach to investigate the role of AMPK activity in fiber type transformation, we performed muscle fiber type analysis in triceps of sedentary and trained γ1TG mice, a genetic model of chronic AMPK activation (25) (Fig. 2B). Training significantly increased type IIa/x fibers in wild-type mice. Sedentary γ1TG mice had a ∼2.6-fold higher proportion of type IIa/x fibers than their sedentary wild-type littermates, and exercise training did not further increase type IIa/x fibers in the transgenic animals. Figure 2C illustrates total running distances and corresponding type IIa/x fibers in wild-type and γ1TG mice. There was no significant correlation between running distance and type IIa/x fibers in these training groups.

Exercise training induces a variety of metabolic adaptations that lead to improvements in fatty acid oxidation (33) and glucose metabolism (34). To study whether AMPK mediates exercise training–induced increases in oxidative capacity, we assessed mitochondrial markers. Citrate synthase activity was ∼14% lower (P < 0.05, main effect) in triceps of α2iTG mice compared with their wild-type littermates (Fig. 3A). Exercise training, however, induced a similar increase in citrate synthase activity in α2iTG and wild-type mice. Succinate dehydrogenase staining of triceps mid-belly cross-sections supports these results (Fig. 3B). The stain was more intense in muscle sections from all trained animals, suggesting that ablation of AMPKα2 activity has little effect on the increase in mitochondria in response to exercise training.

In sedentary γ1TG mice, citrate synthase activity was ∼13% higher than that in their wild-type littermates (Fig. 4A). Similarly, succinate dehydrogenase staining suggested an increase of mitochondria in sedentary γ1TG mice compared with their wild-type littermates as shown in representative micrographs (Fig. 4B). We did not observe any further increase in mitochondrial markers attributable to exercise training in γ1TG mice.

The transcriptional coactivator PGC-1α has previously been linked to fiber type II to I transformation and mitochondrial biogenesis (13,17). The PGC-1α homolog, PGC-1β, has recently been proposed to drive muscle fiber type IIx formation (35). In our study, PGC-1β protein was not altered in either line of transgenic mice compared with wild-type animals or after exercise training (data not shown). In contrast, protein expression of PGC-1α was influenced by genotype as well as training. Ablation of AMPKα2 activity in α2iTG mice was associated with a 32% decrease in PGC-1α protein expression (P < 0.05, main effect) (Fig. 5A). Exercise training, however, increased PGC-1α in α2iTG mice to an extent similar to that in wild-type mice (54% vs. 55%). Sedentary γ1TG mice expressed elevated levels of PGC-1α protein compared with their corresponding wild-type littermates (Fig. 5B), and there was a tendency for training to increase PGC-1α in γ1TG mice (28%, P = 0.06).

Muscle glucose uptake is controlled by glucose transport across the cell membrane, mainly by GLUT4, as well as intracellular phosphorylation by hexokinase II (36,37). Both GLUT4 and hexokinase II have previously been reported to be upregulated by exercise training (38–40). In the current study, training-induced increases in GLUT4 protein expression did not reach statistical significance (main effect: P = 0.06 [Fig. 6A] and 0.12 [Fig. 6B]), and no differences were observed between genotypes. Hexokinase II protein expression was not different in sedentary α2iTG and wild-type mice (Fig. 6C). However, whereas exercise training induced a ∼65% increase in hexokinase II in wild-type mice, there was no significant change in α2iTG mice. γ1TG mice showed an overall higher expression of hexokinase II than their wild-type littermates with no significant additional increase after exercise training (Fig. 6D).

We investigated the role of AMPK in skeletal muscle adaptations to exercise training. For this purpose, we used transgenic mouse models with either attenuated (α2iTG mice) or increased (γ1TG mice) AMPK activity and studied long-term effects of endurance training. Our main finding was that AMPKα2 is an important mediator of training-induced muscle fiber type changes as shown by decreased fiber type transformation in trained α2iTG mice. The fiber type profile of γ1TG mice with increases in type IIa/x muscle fibers further emphasizes the significance of AMPK activity in the determination of muscle fiber types. Previous studies showed that AMPK activity in γ1TG mice is not further increased by acute bouts of exercise (25). Hence, exercise training does not result in repeated periods of (additional) AMPK activation in these mice, which may explain the absence of a training-induced increase in type IIa/x muscle fibers in γ1TG mice.

Although inactive AMPKα2 in α2iTG mice was associated with a significant reduction in training-induced fiber type transformation, the fiber type IIb to IIa/x shift was not completely inhibited. These data suggest that AMPKα2 is not solely responsible for this skeletal muscle adaptation. Preserved AMPKα1 activity may account for the residual response. However, data from a variety of studies suggest that AMPK activity in skeletal muscle is almost exclusively a function of the α2 isoform (20,41). Hence, it is more likely that alternative fiber type–regulating pathways remain responsive in α2iTG mice and account for the residual fiber type adaptation. In particular, the Ca²⁺/calmodulin-dependent protein phosphatase calcineurin has been proposed as an important regulator of fiber type II to I transformation (42).

It seems possible that AMPKα2 and calcineurin signaling are parallel pathways in muscle fiber type adaptations, possibly being differentially regulated by certain stimuli, such as innervation patterns or exercise intensity. Our transgenic mouse model with chronically active AMPK (γ1TG) has a marked increase in type IIa/x fibers but no increase in type I fibers in triceps muscle. This response closely imitates the physiological response to endurance exercise training, which typically leads to muscle fiber type IIb to IIa transformations (7,8). Conversely, expression of constitutively active calcineurin mainly induces slow-twitch type I fibers (42). This kind of fiber type twitch has been seen after powerful stimuli such as very rigorous exercise training programs in humans (8) and electrical stimulation or cross-innervation in animal studies (43). Thus, calcineurin may be the predominant mediator of fiber type II to I transformations, whereas AMPKα2 appears particularly important for mediating type IIb to IIa/x transformations as seen in response to moderate intensity exercise training. It will be intriguing to further investigate Ca²⁺-dependent regulators of fiber type changes in combination with the AMPK pathway and their distinct responses to various stimuli.

AMPK activation has previously been reported to induce mitochondrial biogenesis (14,16). We did not directly assess mitochondrial content in the present study. However, the reduction in citrate synthase activity in α2iTG mice and its increase in γ1TG mice support the role of AMPK for baseline mitochondrial content. Thus, we were intrigued to find that training-induced increases in mitochondrial markers (citrate synthase activity and succinate dehydrogenase) occurred normally in α2iTG mice. These data suggest that AMPKα2 activity is not necessary for training-induced increases in mitochondria. Alternative pathways, such as the aforementioned calcineurin signaling pathway (44), may play a critical role in mediating exercise training–induced increases in mitochondria. However, recent reports have questioned the involvement of calcineurin in these training adaptations (45,46). Clearly, more studies are needed to comprehensively understand the mechanisms leading to exercise training-induced mitochondrial biogenesis.

The transcriptional coactivator PGC-1α is known as a principal factor in muscle fiber plasticity by inducing mitochondrial biogenesis and fast-to-slow fiber type transformation (13). PGC-1α has been described to be downstream of AMPK signaling (12). In the present study, consistent with the proposed AMPK/PGC-1α signaling pathway, α2iTG mice showed decreased levels of PGC-1α protein, whereas γ1TG mice had increased PGC-1α expression. Interestingly, exercise training increased PGC-1α protein even when AMPKα2 activity was inhibited (α2iTG mice). Hence, in addition to AMPK, there must be other pathways that regulate PGC-1α protein expression during exercise training. Ca²⁺/calcineurin signaling as well as the p38/mitogen-activated protein kinase pathway are possible candidates that have been reported upstream of PGC-1α (13,47,48). Thus, PGC-1α may represent a point of convergence of various signaling pathways in skeletal muscle plasticity. The PGC-1α homolog, PGC-1β, has recently been shown to mediate the formation of type IIx muscle fibers (35). In contrast to PGC-1α, however, PGC-1β protein expression was unchanged in either line of transgenic mice and after exercise training. Hence, the present study provides no evidence for a role of PGC-1β in training-induced muscle fiber type transformations.

In states of impaired glucose metabolism, endurance training leads to improvements in glucose tolerance (49). Two major regulators of glucose uptake into the muscle cell are the glucose transporter GLUT4 and hexokinase II. Whereas training-induced increases in GLUT4 expression have been described in rats and humans (50), only a few studies have examined GLUT4 in trained mice. In our study, GLUT4 protein expression resulted in a nonsignificant increase in wild-type triceps muscle by 24 and 28%. These mild changes are comparable to previously published training effects (17–26% increases) in gastrocnemius muscle of wild-type and AMPKα2 knockout mice (16). The minimal changes in GLUT4 expression in trained mouse muscle are consistent with previous reports that adaptations of skeletal muscle to exercise training are more subtle in mice than in rats or humans, which may be due to the higher basal metabolism in mice (51).

Hexokinase II phosphorylates glucose upon entering the muscle cell and may be the major rate-limiting factor for glucose uptake in the insulin- or exercise-stimulated state (37). AMPK and exercise training have been suggested to play a role in the regulation of hexokinase (16,40). In our study, hexokinase II protein expression was substantially increased in association with exercise training and chronically activated AMPK. Interestingly, in triceps muscle of α2iTG mice, exercise training did not upregulate hexokinase II. This observation conflicts with data from Jorgensen et al. (16) showing normal increases in hexokinase II protein expression in trained gastrocnemius muscle of AMPKα2 whole-body knockout mice. The reason for this discrepancy is not entirely clear, as both AMPKα2 knockout and α2iTG mice lack AMPKα2 activity. Although the different mouse lines might yield conflicting results, it is more likely that there are muscle-specific differences between triceps and gastrocnemius in responsiveness to AMPK activity. The exercise-induced increase of hexokinase II suggests increased capacities for glucose phosphorylation as an adaptation to endurance training. Our results imply that this effect requires AMPKα2 activity, thus extending the importance of this enzyme for exercise training adaptations in skeletal muscle.

There is strong epidemiological evidence that regular physical activity improves insulin resistance (1). Furthermore, individuals with diabetes appear to have a distinct skeletal muscle phenotype with a prevalence of type IIx fibers and impaired oxidative capacity (18,52). These observations suggest that skeletal muscle adaptations to exercise training may play an important role in the prevention of diabetes. The underlying molecular mechanisms, however, are only partly understood. We demonstrate here that AMPK plays a critical role in training adaptations of skeletal muscle. Although our results suggest that the increase in oxidative capacity is mediated through AMPK-independent pathways, we show the significance of AMPK signaling for transforming skeletal muscle fiber type IIb to IIa/x as well as for increasing hexokinase II protein expression. In the long run, a better understanding of the molecular mechanisms that link exercise training to improvement or prevention of type 2 diabetes should reveal new targets for treatment and provide fundamental knowledge of the complex physiological processes in this disease.

This work was supported by the National Institutes of Health (Grants R01DK068626 and R01AR45670 to L.J.G., Grant DK35712 to L.A.W., Training Grant T32DK07260-29 to J.B., and Diabetes Endocrinology Research Grant DK36836) and by a fellowship within the Postdoc Program of the German Academic Exchange Service DAAD (to K.S.C.R.).