Phosphorylation-dependent Translocation of Glycogen Synthase to a Novel Structure during Glycogen Resynthesis*

Glycogen metabolism has been the subject of extensive research, but the mechanisms by which it is regulated are still not fully understood. It is well accepted that the rate-limiting enzymes in glycogenesis and glycogenolysis are glycogen synthase (GS) and glycogen phosphorylase (GPh), respectively. Both enzymes are regulated by reversible phosphorylation and by allosteric effectors. However, evidence in the literature indicates that changes in muscle GS and GPh intracellular distribution may constitute a new regulatory mechanism of glycogen metabolism. Already in the 1960s, it was proposed that glycogen was present in dynamic cellular organelles that were termed glycosomas but no such cellular entities have ever been demonstrated. The aim of this study was to characterize muscle GS and GPh intracellular distribution and to identify possible translocation processes of both enzymes. Using in situ stimulation of rabbit tibialis anterior muscle, we show GS and GPh intracellular redistribution at the beginning of glycogen resynthesis after contraction-induced glycogen depletion. We identify a new “player,” a new intracellular compartment involved in skeletal muscle glycogen metabolism. They are spherical structures that were not present in basal muscle, and we present evidence that indicate that they are products of actin cytoskeleton remodeling. Furthermore, for the first time, we show a phosphorylation-dependent intracellular distribution of GS. Here, we present evidence of a new regulatory mechanism of skeletal muscle glycogen metabolism based on glycogen enzyme intracellular compartmentalization.

Glycogen metabolism has been the subject of extensive research, but the mechanisms by which it is regulated are still not fully understood. It is well accepted that the ratelimiting enzymes in glycogenesis and glycogenolysis are glycogen synthase (GS) and glycogen phosphorylase (GPh), respectively. Both enzymes are regulated by reversible phosphorylation and by allosteric effectors. However, evidence in the literature indicates that changes in muscle GS and GPh intracellular distribution may constitute a new regulatory mechanism of glycogen metabolism. Already in the 1960s, it was proposed that glycogen was present in dynamic cellular organelles that were termed glycosomas but no such cellular entities have ever been demonstrated. The aim of this study was to characterize muscle GS and GPh intracellular distribution and to identify possible translocation processes of both enzymes. Using in situ stimulation of rabbit tibialis anterior muscle, we show GS and GPh intracellular redistribution at the beginning of glycogen resynthesis after contraction-induced glycogen depletion. We identify a new "player," a new intracellular compartment involved in skeletal muscle glycogen metabolism. They are spherical structures that were not present in basal muscle, and we present evidence that indicate that they are products of actin cytoskeleton remodeling. Furthermore, for the first time, we show a phosphorylation-dependent intracellular distribution of GS. Here, we present evidence of a new regulatory mechanism of skeletal muscle glycogen metabolism based on glycogen enzyme intracellular compartmentalization.
Although the primary factors involved in the development of type 2 diabetes mellitus are still unknown, it is clear that insulin resistance of skeletal muscle glucose metabolism plays a key role. A defect in insulin activation of muscle glycogen synthase (GS), 1 a key enzyme in the regulation of glycogen synthesis, is one of the most consistent findings in patients with type 2 diabetes (1,2). However, since the mechanisms involved in muscle GS activation are still not fully understood, the defective steps in type 2 diabetic subjects have remained unidentified. A better understanding of skeletal muscle glycogen metabolism is needed to understand the primary factors involved in the development of the disease and to identify targets for new therapeutic strategies.
The rate-limiting enzymes in glycogenesis and glycogenolysis are considered to be GS and glycogen phosphorylase (GPh), respectively. Both enzymes are regulated by reversible phosphorylation and by allosteric effectors. In vivo, nine GS sites susceptible to phosphorylation have been identified, which are phosphorylated by different protein kinases. In rabbit skeletal muscle GS, two sites (2 and 2a) are located near the N terminus, whereas the remaining seven (sites 3a-c, 4, 5, 1a, and 1b) are located within 100 residues of the C terminus. Phosphorylations at different sites have different effects on the enzyme activity. The most important sites involved in GS intrinsic activity regulation are 2, 2a, 3a, and 3b. Dephosphorylation of these sites increases GS activity much more than dephosphorylation of the remaining sites, which have little or no effect on the enzyme activity (3). A number of protein kinases have been identified that can phosphorylate GS, including glycogen synthase kinase 3, casein kinases 1 and 2, protein kinase C, Ca 2ϩ / calmodulin-dependent protein kinase-2, cAMP-dependent protein kinase, phosphorylase kinase, and AMP-activated protein kinase (for review see Ref. 4). Sequences of hierarchal phosphorylation of the N-terminal and C-terminal sites have been described. Phosphorylation of site 2 creates a recognition motif for casein kinase 1 to phosphorylate site 2a (5), whereas phosphorylation of site 5 by casein kinase 2 is a prerequisite for glycogen synthase kinase 3 to sequentially phosphorylate sites 4, 3c, 3b, and 3a (6). Covalent regulation of GS by phosphorylation is complex and still not fully understood. The binding of 1 The abbreviations used are: GS, glycogen synthase; GPh, glycogen phosphorylase; Glc-6-P, glucose 6-phosphate; SR, sarcoplasmic reticulum; CLF, chronic low frequency; TA, tibialis anterior; S1, 1,000 ϫ g supernatant; P1, 1,000 ϫ g pellet; S2, 10,000 ϫ g supernatant; P1, 10,000 ϫ g pellet; S3, 100,000 ϫ g supernatant; P3, 100,000 ϫ g pellet; GS1a, GS phosphorylated at site 1a; GS1b, GS phosphorylated at site 1b; GS2ϩ2a, GS phosphorylated at sites 2 and 2a; GS3aϩ3b, GS phosphorylated at sites 3a and 3b; GN, glycogenin; TEM, transmission electron microscopy. the allosteric activator glucose 6-phosphate (Glc-6-P) reverses inactivation of GS by phosphorylation (7) and increases the susceptibility of the enzyme to dephosphorylation by GS phosphatases (8). GPh is regulated by reversible phosphorylation at Ser 14 by phosphorylase kinase, resulting in activation of the enzyme. GPh is also subject to allosteric regulation, mainly by AMP (9). Dephosphorylation of the multiple GS sites and GPh Ser 14 is mainly catalyzed by the glycogen-associated phosphatases of type 1 (10). The dephosphorylated forms of GS and the phosphorylated form of GPh are active in absence of the allosteric activators, Glc-6-P and AMP, respectively. Therefore, the ratio of the activities in the presence and absence of the allosteric activator is a useful measure of the phosphorylation state for both enzymes.
Already in the 1960s, glycogen was proposed to be present in dynamic cellular organelles that were termed glycosomas (11) but to date no such cellular entities have been clearly defined. Nevertheless, increasing evidence in the literature supports the concept of compartmentalized glycogen metabolism. In skeletal muscle, a fraction of glycogen particles have been found to be associated with the sarcoplasmic reticulum (SR) (12,13). Furthermore, two glycogen fractions have been described that differ in molecular weight and protein content. Mature glycogen particles were termed macroglycogen, whereas smaller glycogen particles, detected as an intermediate product of macroglycogen, were termed proglycogen (14). Structural differences between the free and the SR-associated glycogen particles have been reported (15), which could be explained by differences in the proportions of proglycogen and macroglycogen. However, whether glycogen particles have different metabolic functions or are metabolized in different ways depending on their composition or intracellular localization is still not clear.
Glycogen-protein and protein-protein interactions play an important role in the regulation of glycogen metabolism (for review see Refs 16,17). 95% protein content in glycogen particles comprises GS, GPh, debranching enzyme, and phosphorylase kinase (18). In skeletal muscle at rest, GS is mainly found associated with glycogen particles (19,20) and myofibrils (21). In three different cultured cell types, muscle GS translocates from the nucleus to the cytosol in response to glucose (22). Furthermore, in fully differentiated rat skeletal muscle, GS translocates from a glycogen-enriched to an actin-enriched fraction in response to glycogen depletion induced by muscle contraction (23). For an efficient start of glycogen resynthesis, GS needs to associate with glycogenin (24), the initiator protein of glycogen synthesis. Both of these proteins have been reported to be associated with the actin cytoskeleton (23,25). GPh, on the other hand, has been found associated to SR (26), interacting directly or indirectly with ryanodine receptors and acting as a negative regulator of SR Ca 2ϩ release (27). All of the above support the concept of a complex, compartmentalized glycogen metabolism with different cellular compartments and pools of glycogen particles, which are dynamic entities subjected to constant synthesis and degradation and to which different proteins bind and dissociate depending on the metabolic state. Changes in the intracellular distribution of GS and GPh may constitute a new regulatory mechanism that could target their activity to different cellular compartments. However, the identity of the intracellular compartments involved and the molecular processes by which translocation of these enzymes are regulated are still unknown.
If there was no compartmentalization of glycogen metabolism, variation in glycogen content should be fully explained by changes in the activation states of GS and GPh and variations in the concentration of the respective allosteric activators, Glc-6-P and AMP. Metabolic situations have been reported where this is not the case (28 -30). In a previous study (30), we compared the time courses of changes in glycogen content and the activity ratio of GS, an indicator of the phosphorylation state of the enzyme. During continuous low frequency (CLF) stimulation of rabbit tibialis anterior (TA) muscles, we found that, after 30 min, muscle glycogen levels had reached a minimum. By that time the activity ratio of GS was already high; however, no net resynthesis of glycogen was detected until after 3 h of stimulation. The aim of the present study was to characterize the intracellular distribution of skeletal muscle GS and GPh and to monitor possible translocation processes that could explain this lack of correlation between glycogen content and the activation state of the enzyme.

EXPERIMENTAL PROCEDURES
Animal Procedures-All of the experiments were approved by the Animal Experiments Inspectorate of the Danish Ministry of Justice. Female New Zealand rabbits were anesthetized, electrodes were implanted, and chronic low frequency (10 Hz) stimulation of TA muscle was performed as previously described (30). In all of the experiments, the left leg was stimulated, whereas the right leg was used as control.
Muscle Homogenate Fractionation Study-TA muscles were carefully exposed, rapidly frozen in liquid nitrogen, powdered, and homogenized in 10 volumes of ice-cold 50 mM HEPES (pH 7.8), 100 mM NaF, 10 mM EDTA, 10 g/ml aprotinin, 10 g/ml leupeptin, 1 mM phenylmethylsulfonyl fluoride, 1 mM benzamidine, and 30 mM ␤-mercaptoethanol buffer. Homogenates were centrifuged at 1,000 ϫ g for 10 min at 4°C, obtaining supernatant (S1) and pellet (P1). S1 was centrifuged at 10,000 ϫ g for 15 min at 4°C, obtaining supernatant (S2) and pellet (P2). S2 was centrifuged at 100,000 ϫ g for 1 h at 4°C, obtaining supernatant (S3) and pellet (P3). Glycogen content, GS and GPh assays, and immunoblotting were performed as previously described (30). For Western blot analyses, protein concentration in S1 was determined and 30 g of protein from the S1 homogenate fraction were loaded. The same volume was loaded for the rest of the fractions. Particulate fractions were resuspended with a volume of homogenate buffer equal to the corresponding recuperated supernatant. GS and GPh were immunodetected with antibodies against GS (Chemicon) and GPh (generously provided by Dr. A. A. DePaoli-Roach, Indiana University). Quantification was performed using Quantity One software (Bio-Rad).
Immunofluorescence and Transmission Electron Microscopy in Single Muscle Fibers-Muscles were fixed by perfusion, and single fibers were teased and processed for fluorescence microscopy and transmission electron microscopy (TEM). Immunocytochemistry was performed as previously described (31). GS and GPh were immunodetected with the same primary antibodies used for Western blot. Production of the antibodies against GS phosphorylated at sites 2ϩ2a (Ser 7 and Ser 10 ) and sites 3aϩ3b (Ser 640 and Ser 644 ) has been previously described (32). Antibodies recognizing GS phosphorylated at site 1a (Ser 697 ) and site 1b (Ser 710 ) were raised against the peptides CEWPRRApSCTSSTG (residues 692-703 of human GS) and CSGSKRNpSVDTATS (residues 704 -716 of human GS), respectively. Peptides were synthesized with an additional cysteine at the N terminus coupled to keyhole limpet hemocyanin and used to immunize sheep as described previously (33). Factin was stained with Alexa-568 phalloidin (Molecular Probes, Eugene, OR), and antibodies against fast myosin (Sigma-Aldrich), ␤-actin (Abcam), ␥-actin (Abcam), ␣-actinin (Abcam), vinculin (Abcam), gelsolin (Abcam), and tropomyosin smooth (Chemicon) were used for immunostaining of the corresponding proteins. Secondary antibodies were conjugated to either Alexa-568 or Alexa-488 (Molecular Probes).
Image Acquisition and Manipulation-Confocal images were collected on a TCS SP2 microscope (Leica) through a Plan-Apo ϫ63/1.32 oil objective at 20°C as previously described (31). Confocal Z-stack images were collected from the surface of the muscle fibers 0.25-m apart on the z-plane. Images were analyzed using Metamorph software (Universal Imaging Corp.).
Muscle Fixation for TEM-For optimal structural preservation, vascular perfusion was performed with two different glutaraldehyde-based fixatives in succession, as previously described (34). Following perfusion-fixation, TA muscles were isolated and fixed in the secondary fixative overnight and subsequently postfixed in 1% OsO 4 in 0.12 M sodium cacodylate buffer (pH 7.4) for 2 h. The specimens were dehydrated in a graded series of ethanol, transferred to propylene oxide, and embedded in Epon according to standard procedures. Ultrathin sections were cut with a Reichert-Jung Ultracut E microtome and collected on one-hole copper grids with Formvar supporting membranes. Staining with uranyl acetate and lead citrate was performed, and sections were examined in a Philips CM 100 transmission electron microscope operated at an accelerating voltage of 80 kV. Images were collected using a Megaview 2 camera and processed with the Analysis software package.

Intracellular Redistribution of Muscle GS and GPh at the
Beginning of Glycogen Resynthesis-The previously introduced lack of net glycogen synthesis after 30 min and 1 h of stimulation, when GS activity ratio was already high (30), could have been due to a counteracting high GPh activity ratio. To investigate whether that was the case, we measured GPh activity ratio at rest and after 30 min, 1, and 3 h of CLF stimulation in rabbit TA muscles. GPh activity ratio decreased from 0.67 Ϯ 0.03 in resting muscle to 0.03 Ϯ 0.01 and 0.03 Ϯ 0.00 at 30 min and 1 h, respectively (p Ͻ 0.01, n ϭ 6 -9). By 3 h of stimulation, the GPh activity ratio had increased to 0.50 Ϯ 0.09, which was not significantly different from basal levels. Thus, the lack of net glycogen synthesis was not due to a high rate of glycogenolysis because the activity ratio of GPh at 30 min and 1 h was very low. The question arose of whether GS required been associated with a particular cellular compartment to initiate glycogen resynthesis. To address this question, we performed subcellular fractionation by sequential centrifugation of basal and 3-h-stimulated muscle homogenates. We obtained a S1 and P1, S2 and P2, and S3 and pellet P3. To determine whether glycogen particles associated with the different fractions were differentially affected by stimulation, glycogen content was quantified in the different homogenate fractions from basal and stimulated muscles. The results show that glycogen levels are significantly decreased to a similar extent in all of the fractions of stimulated muscles compared with the respective basal fractions (Fig. 1A).
To detect possible changes in the intracellular distribution of GS and GPh, total activities were measured and Western blots were performed on the different fractions. In response to stimulation, GS and GPh total activities did not change significantly in the different homogenate fractions, with the exception of the S3 fraction where a significant increase in the total activity of both enzymes was detected (Fig. 1B). Consistent with these results, the protein levels of both enzymes were significantly increased in the S3 fraction and significantly decreased in the P3 fraction of stimulated muscle homogenates (Fig. 1C). Thus, the above results show that, in response to 3 h of stimulation, muscle GS and GPh move from cellular structures that pellet at 100,000 ϫ g (P3), probably glycogen particles, to cytosolic structures, which remain in the soluble fraction (S3). Furthermore, to determine whether there were any differences in the activation state of GS and GPh depending on their intracellular distribution, the activity ratio of both enzymes was measured on the different homogenate fractions. After stimulation, the activity ratio of GS significantly increased and the activity ratio of GPh significantly decreased in all of the fractions (Fig. 1D). No significant differences were detected in either GS or GPh activity ratios between the different fractions from either basal or stimulated muscles, indicating that the activity ratio or activation state of both enzymes is independent of their distribution between muscle homogenate fractions.
Phosphorylation-dependent GS Translocation to a Spherical Structure, a Novel Glycogen Metabolism Intracellular Compartment-To visualize the cellular structures with which GS associates in basal and stimulated muscles, we performed a morphological study by confocal immunofluorescence microscopy using antibodies against total GS and antibodies designed and produced to recognize different specific phosphorylated forms of GS (Fig. 2). In Fig. 3A, representative images of muscle GS intracellular distribution in basal and 3-h-stimulated single muscle fibers are shown. In resting muscle, GS is mainly found in the perinuclear region with a polarized arrangement and at the myofibrillar cross-striations with a periodicity corresponding to that of sarcomeres. A total of 414 single muscle fibers from four stimulated muscles were analyzed. In 36% fibers, GS showed a clear association with spherical structures with maximum dimensions of 0.5-1 m, which FIG. 1. GS and GPh redistribution between homogenate fractions in response to stimulation. Basal and 3-h CLF-stimulated TA muscle homogenates were fractionated by differential centrifugation obtaining: S1 and P1, S2 and P2, and S3 and P3. A, glycogen content was measured and is expressed as mol of glucose/g wet tissue. Results show a decreased glycogen content in all of the stimulated muscle homogenate fractions (the decreased percentage of glycogen content is indicated below the bars). B, GS and GPh total activities in S3 and P3 fractions from basal and stimulated muscles are expressed as milliunits/mg protein and units/mg protein, respectively. C, representative Western blots of GS and GPh and quantification of protein content in the S3 and P3 fractions are presented. Note that, in response to stimulation, there was a significant increase in both GS and GPh total activities and protein levels in the S3 homogenate fraction. At the protein level, a significant decrease in both enzymes was detected in the P3 homogenate fraction. D, activity ratios of GS and GPh are indicative of the activation state of both enzymes. GS and GPh activity ratios were increased and decreased, respectively, in all of the stimulated muscle homogenate fractions. Results are expressed as the means Ϯ S.E. (n ϭ 5) of basal (black filled bars) and stimulated (gray filled bars) muscles. Statistical significance was assessed by one-way ANOVA (analysis of variance) and is indicated as follows: *, p Ͻ 0.05; **, p Ͻ 0.01; and ***, p Ͻ 0.005.
were found throughout the muscle fibers (Fig. 3A). To elucidate whether the observed redistribution of GS in response to 3-h CLF stimulation was phosphorylation-dependent, we studied the distribution of several phosphorylated forms of the enzyme in basal and stimulated muscle using phosphospecific antibodies (Fig. 3B). When phosphorylated at site 1a (GS1a), GS was found in the perinuclear region and at the cross-striations in resting muscle fibers, whereas in the stimulated fibers, it was mainly found to be associated with the spherical structures. GS phosphorylated at sites 3a and 3b (GS3aϩ3b) showed a similar distribution pattern associated with perinuclear structures and the cross-striations in resting muscle and with the spherical structures after stimulation. On the other hand, when phosphorylated at site 1b (GS1b), GS was found only at the crossstriations in both basal and stimulated muscle. Finally, when phosphorylated at sites 2 and 2a (GS2ϩ2a), GS presented a dotted distribution pattern not associated with perinuclear structures, cross-striations, or the spherical structures in both basal and stimulated muscle. Taken together, our results show that the fraction of total GS, which is associated with the spherical structures, can be phosphorylated at sites 1a and 3aϩ3b but is dephosphorylated at sites 1b and 2ϩ2a. These results demonstrate for the first time an intracellular distribution of GS that is dependent on its phosphorylation state.
To study changes in the phosphorylation state of GS in response to stimulation, the percentage of the signal obtained with the antibody against total GS that overlapped with the signal obtained using each of the phosphospecific antibodies was quantified and taken as an indication of the percentage of GS phosphorylated at that site. The fraction of total GS overlapping with GS1a decreased from 28.8 Ϯ 2.5 to 13.1 Ϯ 1.8% (p Ͻ 0.05) after 3 h of stimulation. Phosphorylation at site 1b also decreased from 34.8 Ϯ 2.1 to 14.7 Ϯ 4.1% (p Ͻ 0.01) after stimulation. The percentage of total GS overlapping with GS2ϩ2a decreased from 12.6 Ϯ 0.8 to 5.8 Ϯ 1.0% after stimulation (p Ͻ 0.05). No significant difference was detected in the percentage of total GS overlapping with GS3aϩ3b upon stimulation (25.9 Ϯ 8.6% and 30.3 Ϯ 15.7% in rested and stimulated muscle fibers, respectively). Thus, stimulation for 3 h decreased phosphorylation of GS at sites 1a, 1b, and 2ϩ2a without apparently changing phosphorylation at site 3aϩ3b.
Glycogen Phosphorylase Translocates to the Same Spherical Structures in Response to Stimulation-To determine whether the spherical structures were also a target for the translocation of GPh shown by the above presented biochemical data, coimmunostaining of GS and GPh was performed in single muscle fibers from basal and stimulated muscles (Fig. 4). In basal muscles, GPh presented a dotted distribution found mainly in the perinuclear regions and at the cross-striations. After stimulation, a fraction of GPh translocated to the spherical structures at which GS was also associated, whereas the remainder gave an unchanged distribution. Quantification of GS and GPh co-localization showed no significant difference in the percentage of overlap in signals (83.2 Ϯ 7.6% and 56.1 Ϯ 19.2% in basal and stimulated muscles, respectively).
A New Dynamic Intracellular Compartment Involved in Muscle Glycogen Metabolism-To identify the spherical structures to which GS and GPh translocate in response to stimulation, basal and stimulated muscles were fixed for transmission electron microscopy and Nanogold immunostaining for GS was performed on single muscle fibers. We show representative images of the spherical structures (Fig. 5A, arrows). Their average size was 0.25 ϫ 0.04 m with some measuring as much as 0.5 ϫ 1 m. These structures were mainly located in the subsarcolemmal region and in the I bands of the sarcomeres (Fig. 5, A and E). Most of the observed structures were found near transverse tubules and surrounded by SR elements (Fig.  5, A and C). These spherical structures were not membranelimited, and they showed a triangular or hexagonal lattice pattern, indicating a repetitive crystalline structure (Fig. 5B). No such structures were observed in resting muscles. When immunogold staining against GS was performed, we observed that GS was clearly associated with these structures (Fig. 5, D  and E), which is consistent with the finding that glycogen particles are also associated with the spherical structures (Fig.  5B, arrow).
The Reported Spherical Structure Appears to Be a Product of Actin Cytoskeleton Reorganization-In the late 1960s, structures similar to these observed here were described in myopathic and non-myopathic skeletal muscle fibers and were termed crystalloid bodies (35,36). Up until now, the composition and cellular function of these structures appear to have been unknown. Nevertheless, previous ultrastructural and histochemical studies have shown that the crystalloid bodies have two distinct components with particles of glycogen attached to filaments of unknown nature. They are formed by parallel filaments of 6 -10 nm, beaded periodically by electron-dense particles of 10 -18 nm in a lattice, hexagonal, or parallel-ripple pattern (37). The contractile apparatus of skeletal muscle is composed of repeating units, i.e. sarcomeres, which are formed of ordered arrays of actin-containing thin filaments and myosin-containing thick filaments. In some of the present structures, we observed filaments from the sarcomeres merging into their electro-dense pattern (Fig. 6, A-C). To investigate whether the filaments of unknown nature forming the spherical structures corresponded to any of the sarcomeric contractile filaments, actin thin filaments (F-actin) were stained using fluorescent phalloidin and immunostaining of myosin thick filaments was performed using a monoclonal antibody against the fast myosin isoform (Fig. 6D). No staining of the spherical structures was detected, excluding any involvement of the contractile filaments in their formation. Nevertheless, it is known that four actin isoforms were expressed in a tissue-specific manner, i.e. ␣-skeletal actin, ␣-cardiac actin, ␣-smooth muscle actin, and ␥-smooth muscle actin, predominant in adult skeletal and cardiac-striated muscles and in vascular and enteric smooth muscles, respectively (38). Two other actin isoforms were ubiquitously expressed, i.e. cytoplasmic ␤-actin and ␥-actin. Cytoskeleton isoactins form less stable polymers than skel- FIG. 2. GS-phosphospecific antibodies. Muscle GS can be phosphorylated at least at nine sites by different protein kinases. A schematic model indicating the different phosphorylation sites of rabbit skeletal muscle GS is shown. Two sites (sites 2 and 2a) are located near the N terminus, whereas the remaining seven (sites 3a-c, 4, 5, 1a, and 1b) are located within 100 residues of the C terminus. Phosphospecific antibodies against GS phosphorylated at sites 2ϩ2a, 3aϩ3b, 1a, and 1b were raised immunizing sheep with the indicated peptides, which correspond to the indicated residues of the protein.
etal muscle actin; therefore, actin microfilaments are highly dynamic, whereas the actin myofibrillar system is more stable (39). To investigate whether any of the cytoplasmic actin isoforms formed part of the spherical structures, immunostaining against ␤-actin and ␥-actin was performed (Fig. 6D). The re-sults show that ␤-actin was specifically found in the structures, whereas ␥-actin showed no specific association with them. Furthermore, regulation of actin organization and dynamics is a highly complex process that involves a number of actin-binding proteins, which are regulated by a complex network of signaling cascades. Actin-binding proteins are classified according to their function as cross-linking, capping, branching, severing, and sequestering proteins. ␣-Actinin and vinculin are crosslinking actin-binding proteins (40), and cross-linking of actin filaments could play an important role in the formation of the spherical structures. We investigated whether ␣-actinin and/or vinculin were the actin-binding proteins cross-linking the filaments in the spherical structures. Co-immunostaining against GS and ␣-actinin showed co-localization of both proteins associated to the spherical structures (Fig. 6D). However, no vinculin was detected associated to the reported spherical structures (Fig. 6D). Furthermore, rearrangement of actin microfilaments must involve the activity of at least an actinsevering protein. Here we investigated the possible involvement of gelsolin in the formation of the spherical structures. Gelsolin is a pH-and Ca 2ϩ -dependent actin-severing protein, which primary function is to sever actin microfilaments and cap their barbed ends. Co-immunostaining against GS and gelsolin showed no association of gelsolin to the spherical structures (Fig. 6D). Moreover, tropomyosins are actin regulatory proteins that tightly regulate microfilament organization (41) and protect actin filaments against severing by gelsolin (42). Smooth muscle tropomyosin has been shown to compete with gelsolin and bind to the capped ends of actin filaments after been severed by gelsolin (43). To investigate whether smooth muscle tropomyosin was associated to the spherical structures, co-immunostaining with GS was performed (Fig. 6D). The re-

FIG. 3. Muscle GS phosphorylationdependent intracellular redistribution.
A, total GS intracellular distribution was analyzed by fluorescence immunostaining of single muscle fibers from basal and 3-h CLF-stimulated TA muscle (Bars, 10 m). White squares represent the higher magnification regions taken to visualize the distribution of total GS (red) and the different phosphospecific GS forms (GS-P). B, distribution of GS phosphorylated at sites 1a (GS1a), 1b (GS1b), 2 and 2a (GS2ϩ2a), and 3a-b (GS-3aϩ3b) is shown (green). On the right, a merged image is presented and co-localization between both signals can be seen in yellow (Bars, 2.5 m). Note that when associated to the spherical structures, GS is not phosphorylated at sites 1b and 2ϩ2a. Secondary antibodies conjugated to Alexa-568 and Alexa-488 were used to immunodetect total GS and the different phosphospecific forms of GS, respectively. The presented images are representative of four independent experiments. From each immunostaining, an average of 20 fibers was analyzed.
FIG. 4. GS and GPh translocate to the spherical structures in response to stimulation. GS and GPh intracellular distribution was analyzed in basal and 3-h CLF-stimulated single muscle fibers by fluorescence co-immunostaining. Total GS (red) and GPh (green) intracellular distribution is shown in basal (upper panel) and stimulated (lower panel) single muscle fibers. On the right, a merged image is presented and co-localization between both signals can be seen in yellow (Bars, 2.5 m). Note that GPh translocates to the same spherical structures than GS in response to stimulation. Secondary antibodies conjugated to Alexa-568 and Alexa-488 were used to immunodetect GS and GPh, respectively. The presented images are representative of four independent experiments. From each, an average of 20 fibers was analyzed.
sults show the presence of smooth muscle tropomyosin in the reported cellular spherical structures. In summary, the abovepresented data reveal the presence of ␤-actin, ␣-actinin, and smooth muscle tropomyosin in the spherical structures.

DISCUSSION
The aim of the present study was to characterize the intracellular distribution of skeletal muscle GS and GPh and to identify possible translocation processes that could explain the lack of correlation between glycogen content and the activation state of both enzymes, as previously reported (30). During CLF stimulation of rabbit TA muscle, muscle glycogen levels reached a minimum after 30 min of stimulation. By that time, the activity ratio of GS was already high but no net resynthesis of glycogen was detected until after 3 h of stimulation. In the present study, we showed that the lack of glycogen resynthesis was not due to a high GPh activity ratio. Furthermore, we showed a clear intracellular redistribution of both GS and GPh after 3 h of CLF stimulation (Figs. 3 and 4), which coincides with the beginning of muscle glycogen resynthesis. During CLF stimulation, a negative correlation has been found between the reduction of isometric force and metabolic recovery with the first signs of recovery detectable after 1 h of stimulation (44). The force reduction coincided with an abrupt decline in the electromyographic amplitude, suggesting refractoriness of a large population of muscle fibers (45). Our results showed redistribution of GS in a percentage of stimulated muscle fibers, which could correspond to those muscle fibers that are in a refractory state, and thereby in early stages of metabolic recovery and glycogen resynthesis. These observations, taken together with our previous findings (30), indicate that association of GS with the spherical structures may be necessary to initiate glycogen resynthesis, introducing a potential new regulatory mechanism of glycogenesis. This is supported by recent studies where a similar intracellular reorganization of GS in response to insulin and glucose has been reported in adipocytes and cultured hepatocytes (46,47). In those studies, as well as in the current study, redistribution of GS was detected at the beginning of glycogen synthesis after metabolic stress associated with marked glycogen depletion. Furthermore, it has previously been shown that muscle GS needs to associate with glycogenin (GN) for efficient initiation of glycogen synthesis (24) and that both proteins can be found associated with the actin cytoskeleton (23,25,50). In rat skeletal muscle, a release of GS and GN into the muscle cytosol was shown after severe glycogen depletion, and only after several hours, a 50% reas-sociation between these proteins was established in vitro, leading the authors to suggest that muscle cytosol may contain factor(s) that regulates glycogen biogenesis by modulating the association of GN and GS. Could it be that both proteins need to be associated with the spherical structures reported here to reinitiate glycogen synthesis? Considering that ␤-actin is present in the spherical structures (Fig. 6D), it is tempting to speculate that GS and GN may need to associate with them in order to commence glycogen synthesis. The time needed for the formation of the spherical structures could explain the delay of at least 30 min in glycogen resynthesis initiation previously observed (30) despite the findings that GS was already highly activated and GPh was highly inactivated.
Reversible phosphorylation of GS is well known to regulate its intrinsic activity. However, a phosphorylation-dependent intracellular distribution of GS has never been demonstrated before. Our results show that GS associated with the spherical structures is dephosphorylated at sites 1b and 2ϩ2a (Fig. 3B), suggesting that also the intracellular distribution of GS is regulated by reversible phosphorylation. The intrinsic enzymatic activity of GS is mainly regulated through the phosphorylation of sites 2ϩ2a and 3a-c (3), whereas phosphorylation at sites 5, 1a, and 1b have little or no effect on GS activity (48,49). Thus, the fraction of GS associated with the spherical structures is highly active because it is dephosphorylated at sites 2 and 2a. With the presented data in mind (Fig. 3B), it is tempting to speculate that phosphorylation at sites 1b and 2ϩ2a could regulate muscle GS intracellular localization.
This study identifies a new "player," a new intracellular compartment involved in skeletal muscle glycogen metabolism, which is a spherical cellular structure formed in skeletal muscle within 3 h of CLF stimulation. Are these cellular structures the hypothetical glycosoma? Since the 1960s, the idea of a cellular organelle involved in glycogen metabolism regulation has been discussed but such entity has never been demonstrated before. Our results show that glycogen particles, GS and GPh, can all be found associated with the reported spherical structures and that GS association to these structures is dependent on its phosphorylation state. Thus, it seems clear that these spherical structures play an important role in glycogen metabolism. Nevertheless, not much is known regarding the nature of these spherical structures; however, here we report the presence of ␤-actin, ␣-actinin, and smooth muscle tropomyosin in them. The similarity in size between filaments of the spherical structures and actin microfilaments (Fig. 6,   FIG. 5. TEM images of the spherical structures. A, a representative TEM image of the reported spherical structures (arrows) is shown. Note that these structures are mainly located at the I band of the sarcomeres surrounded by sarcoplasmic reticulum and close to the transverse tubules (Bar, 1 m). B, magnified detail of the spherical structure shown in C (Bar, 500 nm). Note that glycogen particles (B, arrow) are associated with these structures. D and E, immunogold labeling of GS was followed by silver enhancement and shows association of the enzyme with the spherical structures (Bars, 500 nm). The reported spherical structures can be located as well underneath the sarcolemma (E).
A-C) and the presence of tropomyosin and actinin in them is consistent with the possibility that they are formed of ␤-actin, despite the lack of phalloidin staining. Tropomyosin and actinin are known to occupy alternating domains along the actin filaments, and this could prevent the binding of phalloidin to F-actin (51). Tropomyosin modulates actin-myosin interaction and stabilizes actin filament structures. In skeletal muscle, in the absence of Ca 2ϩ , tropomyosin regulates contractility by sterically blocking actin myosin-binding sites. During contraction, Ca 2ϩ binds to the troponin complex, which causes a shift in the position of tropomyosin to the open state, leaving the actin myosin-binding sites exposed and allowing myosin to bind to actin (for review see Ref. 52). Evidence indicates that there is co-expression of different tropomyosin isoforms in skeletal muscle and that ␣and ␤-tropomyosins are known to be the most abundant; however, four minor isoforms have been reported (53). Our results show low levels of smooth muscle tropomyosin protein expression in rabbit TA muscle (Fig. 6D). In smooth muscle and non-muscle cells, troponin is absent and the precise role and structural dynamics of tropomyosin on actin are poorly understood. Nevertheless, functional differences between skeletal muscle and smooth muscle tropomyosins have been demonstrated (54). Furthermore, the pattern of the electro-dense lattice of these structures shows cross-links between the actin-like filaments. Considering the presence of ␣-actinin in the reported spherical structures (Fig. 6D), it seems reasonable to suggest that the actin-linking component in the spherical structures is partly or wholly composed of ␣-actinin. ␣-Actinin is a dimer composed of two anti-parallel monomers (55) with the actin-binding domains at opposite ends of the protein. In addition to its mechanical role, ␣-actinin interacts with proteins involved in a variety of signaling and metabolic pathways (for review see Ref. 56), i.e. glycogen phosphorylase (57) and fructose-1,6-bisphosphatase-aldolase (58).
FIG. 6. The spherical structures seem to be product of actin cytoskeleton reorganization. A, in some large spherical structures, actin filaments seemed to fuse into their electro-dense lattice pattern (Bar, 500 nm). B and C, magnification of the regions where contractile filaments seem to merge into the structures are shown. D, immunostaining against GS (red) was combined with phalloidin-Alexa 488 staining of F-actin and co-immunostaining of fast myosin, ␤-actin, ␥-actin, ␣-actinin, vinculin, gelsolin, and smooth tropomyosin (Tm-smooth) (green) (Bar, 2.5 m). Representative images are shown. On the right, merged images are presented and co-localization between both signals can be seen in yellow. The exact composition of the spherical structures needs to be elucidated, but the presented results show that ␤-actin, ␣-actinin, and smooth tropomyosin co-localize with GS associated with the spherical structures.
The functional significance of these interactions is unclear. However, the binding of metabolic enzymes to cytoskeletal proteins is a common mechanism of enzyme regulation (59) and it has been suggested that the tethering of these enzymes by ␣-actinin could contribute to the local availability of metabolites and energy generation (56). Our results add to the above evidence that ␣-actinin plays an important role in the communication between the sarcomeric machinery and the regulation of skeletal muscle metabolism. With all of the above in mind, we believe that the spherical structures are products of actin cytoskeleton remodeling in microdomains. This idea would be supported by previous studies in cultured smooth muscle cells where a similar relocalization of contractile proteins including tropomyosin, actin, and ␣-actinin has been described previously (51). Further studies are needed to elucidate the exact composition and the time course for the formation of these structures, as well as the potential translocation of other proteins related to glycogen metabolism. We are presently addressing these issues.