Glycogen- and PP1c-targeting Subunit GM Is Phosphorylated at Ser48 by Sarcoplasmic Reticulum-bound Ca2+-Calmodulin Protein Kinase in Rabbit Fast Twitch Skeletal Muscle*

Multifunctional Ca2+-calmodulin-dependent protein kinase (CaMKII) is a Ser/Thr protein kinase uniformly distributed within the sarcoplasmic reticulum (SR) of skeletal muscle. In fast twitch muscle, no specific substrates of CaMKII have yet been identified in nonjunctional SR. Previous electron microscopy data showed that glycogen particles containing glycogen synthase (GS) associate with SR at the I band level. Furthermore, recent evidence implicates CaMKII in regulation of glucose and glycogen metabolism. Here, we demonstrate that the glycogen- and protein phosphatase 1-targeting subunit, also known as GM, selectively localizes to the SR membranes of rabbit skeletal muscle and that GM and GS co-localize at the level of the I band. We further show that GM, GS, and PP1c assemble in a structural complex that selectively localizes to nonjunctional SR and that GM is phosphorylated by SR-bound CaMKII and dephosphorylated by PP1c. On the other hand, no evidence for a structural interaction between GM and CaMKII was obtained. Using His-tagged GM recombinant fragments and site-directed mutagenesis, we demonstrate that the target of CaMKII is Ser48. Taken together, these data suggest that SR-bound CaMKII participates in the regulation of GS activity through changes in the phosphorylation state of GM. Based on these findings, we propose that SR-bound CaMKII participates in the regulation of glycogen metabolism, under physiological conditions involving repetitive raises elevations of [Ca2+]i.

CaMKII target in G M . We thus suggest that Ca 2ϩ -calmodulin activation of SR-bound CaMKII might play a pivotal role in the regulation of glycogen metabolism during physical exercise through phosphorylation of G M .
Animals and Muscles-New Zealand male adult rabbits were lawfully acquired and properly housed, fed, and taken care of in the Animal Colony of the Department of Experimental Biomedical Sciences of the University of Padova in compliance with Italian Law (Decreto Legge, September 27, 1992, no. 116). The adductor magnus was used as a representative fast twitch muscle (20).
Generation and Purification of His 6 -G M Fusion Proteins-Total cellular RNA extraction from rabbit adductor and cDNA synthesis were performed as described (21). G M cDNA was amplified to generate two forms of G M : G M - (40 -338) and G M -(69 -338). We used an identical cloning strategy with the same reverse primer and a different forward primer. The primers were designed based on the rabbit G M sequence (7) (GenBank TM accession number M65109) and are: 40 Forward (5Ј-CTC TTC ATA TGG CAC ATC ACC ATC ACC ATC ACA TTG ATG ACG ACG ACA AGT CCC CTC AAC CGA GAC GAG-3Ј, containing the NdeI restriction enzyme and a His 6 encoding sequences); 96 Forward (5Ј-CGA CAT ATG CAC CAC CAC CAC CAC CAC GTC GAC AAC TTT GGA TTC AAT-3Ј, containing the NdeI restriction enzyme and a His 6 encoding sequences); and 338 Reverse (5Ј-CTT CCT GAA TTC CTA TTC AGA AGC AGC ACA TCT AG-3Ј, containing the EcoRI restriction enzyme sequence). The first set of primers amplifies the region that encodes for a 299-aa fragment protein (NH 2 -terminal residues from aa 40 to 338). The second, amplifies the region from aa 69 to 338. PCR conditions were as follow: 94°C for 3 min, 94°C for 1 min, 65°C for 1 min, and 72°C for 2.5 min for 5 cycles followed by 30 cycles of 94°C for 1 min, 57°C for 1 min, and 72°C for 2.5 min. The extension step was performed using the Pfu DNA polymerase by Promega. The PCR products, purified using a Qiagen kit, were digested with NdeI and EcoRI enzymes and cloned into the pET 21aϩ vector (Novagen). The amplified mutated vector was used for heat-pulsing transformation of Novablue cells. Vector extraction was performed using a QIAprep Miniprep kit (Qiagen) After confirmation of the sequences, the plasmid was heat pulse transferred into the Escherichia coli strain BL21 (DE). Cultures were grown at 37°C in a medium containing ampicillin. When cultures reached an A 550 of 0.5 units, the His-tagged fusion protein fragments were induced with 0.5 mM isopropyl ␤-D-thiogalactopyranoside for 3 h at 37°C. His-tagged protein fragments were purified from bacterial lysate by nickel-nitrilotriacetic acid affinity chromatography (Qiagen) according to the manufacturer's instructions. Purified protein was quantified by densitometry of Coomassie Blue-stained gels, with reference to a calibration curve of bovine serum albumin (0.5-10 g).
Site-directed Mutagenesis-Alanine substitution in G M -(40 -338) was generated by introducing a mutation with the use of the QuikChange site-directed mutagenesis kit (Stratagene) according to the manufacturer's specifications. The mutagenic primers were designed complementary to opposite strands of the vector and containing two bases mismatched (bold letters) in correspondence of serine 48 to obtain alanine. They were as follow: sense primer 5Ј-CCG AGC AGA CGA GGT GCG GAA TCT TCT GAA GAG GTC-3Ј; and antisense primer 5Ј-GAC CTC TTC AGA AGA TTC CGC ACC TCG TCT GCT CGG-3Ј. PCR conditions were as follows: 1 cycle of 95°C for 30 s and 16 cycles of 95°C for 30 s, 65°C for 1 min, and 72°C for 7 min. The amplified mutated vector was used for heat-pulsing transformation of Novablue cells. Vector extraction was performed using a QIAprep Miniprep kit (Qiagen). Mutation was confirmed by DNA sequencing of the region containing the mutation. The cloning strategy of the mutated G M -(40 -338)-S48A was identical of that described under "Purification of His-G M Fusion Proteins. Mutated His tag G M -(40 -338)-S48A was purified from bacterial lysate by nickel-nitrilotriacetic acid affinity chromatography (Qiagen) according to the manufacturer's instructions.
Preparative Procedures-To avoid ␤-adrenergic stimulation and to maximize association of glycogen to SR, rabbits were anesthetized with Zoletil 100 (10 mg/kg per vena) and injected intravenously with propranolol (3 mg/kg), as described by Walker et al. (14). After 10 min, the animals were killed by intravenous injection of 1.5 ml of Tanax (Intervet International B.V., Boxmee, The Netherlands) in accordance with procedure approved by the Animal Committee of the University of Padova. Bilateral adductor muscles (a representative fast twitch muscle) were homogenized in 0.3 M sucrose (20% weight/volume), 5 mM imidazole, pH 7.0, 1 mM EGTA, 1 g/ml leupeptin, and 100 M phenylmethylsulfonyl fluoride. Muscle microsomes enriched in content of nonjunctional SR were isolated by the method of Chu et al. (22) by centrifuging muscles homogenates at 7700 ϫ g for 10 min. The postmyofibrillar supernatant was recentrifuged for 20 min at 15,000 ϫ g. The final post-mitochondrial supernatant was centrifuged at 150,000 ϫ g for 90 min to obtain soluble cytoplasm and a microsomal fraction. SR membranes were further purified by isopycnic sucrose density centrifugation (23) to yield four distinct fractions, labeled F1-F4 from top to bottom of gradients. The Ca 2ϩ pump membrane and the junctional face membrane (JFM) were dissociated by incubating F4 membranes with 0.25% detergent Chaps at low Ca 2ϩ and centrifuging, as described (24). Ca 2ϩ pump membrane corresponded to the solubilized material. Crude JFM was further purified by the method of Sacchetto et al. (25). The membrane fractions were resuspended in 0.3 M sucrose, 5 mM imidazole, pH 7.4, 100 M phenylmethylsulfonyl fluoride, 1 g/ml leupeptin, divided in aliquots, and stored at Ϫ80°C, until used. Protein concentration was determined by the method of Lowry et al. (26), using bovine serum albumin as a standard.
Measurement of Glycogen-Glycogen content of membrane fractions was measured under the general conditions described by Gomez-Lechon et al. (27) and Lees et al. (28). Membranes resuspended (0.5-1 mg/ml) in 0.2 M sodium-acetate buffer, pH 4.8, were incubated at 40°C with 1750 milliunits/100 l of amyloglucosidase. At the end of 2 h of incubation, distilled water was added to the assay mixture up to final volume of 1 ml and neutralized with 5.5 l of 1 N NaOH. Glucose liberated after enzymatic hydrolysis of glycogen was determined by a colorimetric method, using a glucose assay kit (Sigma). A blank of the reaction was performed by incubating SR membranes without amyloglucosidase. A calibration curve was constructed with known amounts of glycogen.
Assay of Glycogen Synthase Activity-GS activity was assayed at 37°C, by the method of Thomas et al. (29), measuring the incorporation of [ 3 H]glucose from UDP[ 3 H]glucose into glycogen. Membranes (final protein concentration, 0.4 mg/ml) were incubated in duplicate in a standard assay mixture (final volume, 50 l), containing 50 mM Tris-HCl, pH 7.8, 10 mM EDTA, 50 mM NaF, 15 mg/ml glycogen, 2 M okadaic acid, 10 mM UDP[ 3 H]glucose (specific activity, 0.15 Ci/mol), and 10 mM glucose 6-phosphate. At the end of 2 min of incubation, the assay mixture was spotted on Whatman C glass filters. The filters were washed twice for 30 min with ice-cold 70% ethanol (10 ml/filter) and dried. [ 3 H]Glucose incorporated into glycogen was measured by liquid scintillation. Background radioactivity was measured in the absence of the enzyme. Zero time controls were prepared by mixing membranes and test mixture on ice and immediately spotting the assay mixture on filters.
Phosphorylation Assays-CaMKII phosphorylation assay: Phosphorylation of SR proteins and of G M -(40 -338) by endogenous CaMKII was carried out, as described previously (21), using a standard medium containing 100 M free Ca 2ϩ and 1 M CaM, unless otherwise stated. Incubation was at 30°C for 30 min. Phosphorylation by exogenous CaMKII of His-tagged G M recombinant fragments was carried out using a commercially available CaMKII (New England Biolabs). Prior to substrate phosphorylation, CaMKII was activated by incubation with ATP/Mg 2ϩ in the presence of CaCl 2 and calmodulin as indicated by the manufacturer. Activated CaMKII enzyme was then added to the assay mixture (250 units/50 l) containing substrate proteins. Composition of phosphorylation assay medium was identical to that used for endogenous phosphorylation. Incubation was at 30°C for 30 min.
PKA Phosphorylation Assay-Phosphorylation by exogenous PKA of SR membranes and of His-tagged G M fusion proteins was carried out at 30°C as described by Damiani et al. (30), except that the incubation time was 30 min. Incubation medium contained 5 M cAMP (final concentration) and 2.6 M PKA (5-10 units/50 l). All of the phosphorylation reactions were started by the addition of 50 M [␥-33 P]ATP (specific radioactivity, 0.10 Ci/mmol) and were terminated by adding SDS-solubilizing buffer to samples. 33 P-Labeled protein were detected by autoradiography (Hyperfilm Amersham) or by a Bio-Rad model GS-250 Molecular Imager. Bound radioactivity was quantified by using a ␤-particle sensitive screen and a model GS-250 Molecular Imager (Bio-Rad). One unit of exogenous CaMKII and PKA is defined as the amount of enzyme that will transfer 1 pmol of phosphate to specific substrate in 1 min at 30°C.
Protein Dephosphorylation-Isolated SR membranes were incubated at 30°C with 50 M [␥-33 P]ATP in assay medium for phosphorylation by endogenous CaMKII. After 30 min of incubation, CaMKII was inhibited by 2 M staurosporin. As reported in detail in the legends to Fig. 5, protein dephosphorylation by endogenous PP1c was carried out in the same medium for 30 min at 30°C.
Phosphoamino acid Analysis-G M , 33 P-phosphorylated by exogenous CaMKII as described above, was blotted to nitrocellulose membrane, localized by autoradiography, and digested by trypsin. The tryptic peptides were subjected to acidic hydrolysis (6 M HCl for 4 h at 110°C), and the radiolabeled phosphoamino acids were separated by high voltage paper electrophoresis at pH 1.9, as described by Perich et al. (31).
Preparation of Antisera against G M Fusion Proteins-Chicken polyclonal antibodies against the purified His-tagged G M -(40 -338) fusion protein were raised, as previously reported by Damiani et al. (32). Guinea pig polyclonal antibodies were raised against a 11.2-kDa peptide corresponding to aa 754 -852 of rabbit G M (7). The immunogen was expressed and purified by on-column purification through an IMAC column and was then used to raise the antiserum (Harlan Sera-Lab Ltd., Bicester Oxon, UK).
Immunoprecipitation-Immunoprecipitation experiments were carried out as described (21). 250 g of F4 protein, after solubilization, was incubated for 2 h in cold room with the guinea pig anti-G M antibody to (1:1000 dilution). Protein A-Sepharose (40 l) was added to the mixture, incubated for 1 h, and sedimented using an Eppendorf centrifuge. After washing, the pellets were solubilized with SDS solubilization buffer.
Gel Electrophoresis and Immunoblotting-SDS-PAGE (33) and immunoblotting were carried out, as described (21). Slab gels were stained with Coomassie Blue and then with Stains All. Apparent M r values were calculated from a graph of relative mobilities versus log M r of standard proteins. The blots were probed with: (i) mouse monoclonal antibodies to ␣-actinin (Sigma), RyR1 (sarcoplasmic reticulum Ca 2ϩ release channel, skeletal isoform; BioMol, Plymouth Meeting, PA), GP 53 (a SR glycoprotein of 53 kDa) (ABR, Golden, CO), GS (Chemicon International, Temecula, CA), and PP1c (Santa Cruz Biotechnology, CA); (ii) guinea pig polyclonal antibodies to G M ; or (iii) chicken polyclonal anti-bodies to G M . Antibody binding was detected by immunoenzymic staining (21). Densitometry of blotted proteins, after immunostaining, was carried out using a Bio-Rad model GS-670 imaging densitometer. GS content was determined by densitometry of proteins immunostained with mouse monoclonal antibody GS, with reference to a calibration curve of purified GS . G M content was determined by densitometry of blots immunostained with chicken polyclonal antibodies to G M , with reference to a calibration curve obtained with purified G M -(40 -338) .
Immunofluorescence Microscopy-Muscle cryosections were fixed with paraformaldehyde and immunostained by incubating with primary antibodies to G M and GS (dilution 1:100), followed by incubation with the appropriate secondary antibody conjugated with TRITC or fluorescein isothiocyanate (Dako), as described by Sacchetto et al. (34). The images were captured by a B/W chilled CCD camera (Hamamatsu, Japan), transmitted to an interactive image analysis system equipped with image memory (High Fish Beta, version 2.0) and image processing software (Image Processing, version 3.4, Casti Imaging, Venice, Italy). The images were printed on a CP-D1E printer (Mitsubishi, Japan).

Localization of G M to SR-When longitudinal cryosections
from rabbit adductor muscle were probed with a guinea pig anti-G M antibody, a cross-striated pattern was observed (Fig.  1A). By comparison with phase contrast microscopy, the fluorescent striations were found to correspond to the I band of the sarcomere. Fig. 1B shows that G M co-localized with GS, which also is resident in the I band (35). The overlapping between the two proteins was virtually complete. It was verified that nonspecific antibody or secondary alone did not give this fluorescence pattern (not shown).
The subcellular localization of G M was further investigated by differential centrifugation of muscle homogenates. The immunoblot analysis shown in Fig. 1C clearly shows that most G M sedimented mostly in the microsomal fraction, which was devoid of contamination by myofibrillar proteins and highly enriched in SR membranes, as demonstrated by the localization of the SR-specific protein markers, RyR1 and GP 53 . It is noteworthy that G M and GS co-fractionated throughout the entire preparation. A fraction enriched in nonjunctional SR membranes was obtained from muscle microsomes by isopycnic-sucrose density centrifugation (22). This procedure, at variance with that of Saito et al. (23), which yields a much higher proportion of membranes deriving from the terminal cisternae of junctional SR, yields mainly vesicles derived from nonjunctional SR (22). The nonjunctional SR origin of this fraction was confirmed by [ 3 H]ryanodine binding measurements, carried out at optimal conditions of free Ca 2ϩ (100 M) and ionic strength (1 M KCl). The specific SR illustrated in Fig. 2A gave a B max value of 1.1 pmol/mg protein, which is almost 20-fold lower than that previously reported for purified terminal cisternae (19.0 pmol/mg protein) (24). The same fraction was consistently found to be highly enriched in glycogen (927 Ϯ 207 g/mg protein, average Ϯ S.E. of three determinations), as well as in GS activity (0.82 Ϯ 0.04) and in GS protein (179 Ϯ 20 g/mg protein). Of interest, fraction F4 was also characterized by a high density of G M and PP1c (Fig. 2B). Based on this evidence, henceforth all experiments were carried out using the F4 fraction.
Differential Distribution of G M between Ca 2ϩ Pump Membrane and JFM-The treatment of SR membranes with low concentrations of the detergent Chaps can be used (24) to separate the Ca 2ϩ pump membrane, i.e. the highly specialized membrane l of nonjunctional SR consisting of SERCA1 protein (22), from the membrane domains of the SR that are enriched in Ca 2ϩ release channels, RyR1. The RyR1-enriched SR region is named JFM (24), because of its structural association with transverse tubules. We took advantage of this preparative procedure to investigate the partition of G M associated to F4 membranes between Ca 2ϩ pump membrane and the JFM. In agreement with the prevalent nonjunctional SR origin of membranes forming the F4 fraction, more than 80% of total SR protein was solubilized. Fig. 2 shows that well defined junctional SR markers, such as RyR1 (Fig. 2C) and FKBP-12 (Fig.  2D), were present only in the JFM (Fig. 2, C and D, lanes 3) (24). On the other hand, the Ca 2ϩ pump membrane was completely devoid of these proteins, while being enriched in SERCA1 and GP 53 content (Fig. 2, C and D, lanes 2). Western blot analysis with antibodies to G M , GS, and PP1c showed that all of these proteins localized exclusively to the Ca 2ϩ pump membrane (Fig. 2D).
Co-immunoprecipitation of G M , GS, and PP1c-F4 membranes were solubilized and immunoprecipitated with a guinea pig polyclonal antibody to G M protein. Fig. 3A shows that a prominent G M band was specifically immunoprecipitated by this antibody. G M protein neither was bound to protein A in the absence of antibody nor was immunoprecipitated by a nonrelated anti-␣-actinin antibody (Fig. 3A). Fig. 3B shows that G M co-immunoprecipitated not only with PP1c, as expected (36), but also with GS. The specificity of this result was validated as in Fig. 3A (data not shown).
Phosphorylation-Dephosphorylation of G M by SR-bound CaMKII and PP1c-G M is phosphorylated by PKA at Ser 48 and Ser 67 (Refs. 4 and 5; see also Ref. 7). Based on the consensus sequence (37), another potential PKA target site of G M is Thr 215 . The analysis of these potential phosphorylation sites reveals that the amino acid sequence (RXX(S/T)) also corresponds to a consensus sequence for CaMKII (37,38). Because rabbit SR membranes contain CaMKII (see Ref. 40 for a review), we tested whether that G M is phosphorylated by SRbound CaMKII.
To address this point, F4 membranes were incubated with [␥- 33  medium. As we have already reported (39,24,21), the incorporation of radiolabeled phosphate into SR proteins is poor, if it occurs at all, in the absence of exogenously added Ca 2ϩ -calmodulin. On the contrary, in the presence of these effectors several proteins were phosphorylated, including a 60-kDa protein corresponding to autophosphorylated CaMKII (21,40). Additional SR proteins phosphorylated by CaMKII were 95-kDa triadin (39) and 36-kDa glyceraldheyde-3-phosphate dehydrogenase (41) (Fig. 4A, lane 1). As already reported (39), a prominent 33 P-labeled protein was also detected at about 85 kDa. This phosphoprotein proved to be GS, based on Western blot analysis with anti-GS antibody. 2 The addition of 100 M KN-93, a specific inhibitor of CaMKII that competes for calmodulin binding to the enzyme (42), almost completely abolished the phosphorylation of SR proteins (Fig. 4A, lane 2) (see also Ref. 21). Taken together, these results strongly support the notion that the Ca 2ϩ -calmodulin-dependent phosphorylation of SR proteins is mediated by endogenous CaMKII. Fig. 4A shows that 33 P-labeled protein bands were also detectable in the M r range of 160,000 -170,000, including a protein that displayed the same mobility of G M as judged by Western blot analysis (Fig. 4A, lane 3). The low extent of the protein phosphorylation is consistent with the low content of G M in F4 membranes (about 0.5 g/mg of SR protein; see "Experimental Procedures"). Because another SR protein, called HRC, with the apparent M r of 170,000 has been described to be phosphorylated by endogenous CaMKII (39,40) to confirm the phosphorylation of G M , 33 P-labeled SR proteins were solubilized and immunoprecipitated with the guinea pig anti-G M antibody. Fig. 4B shows that the 33 P-labeled protein of about 160 kDa was indeed G M . As a positive control, we performed parallel experiments in which F4 membranes were phosphorylated with exogenous PKA. Also in this case, the anti-G M antibody immunoprecipitated a 160-kDa 33 P-labeled protein that was immunostained with anti-G M antibody (Fig. 4C).
To investigate whether G M could be dephosphorylated by endogenous PP1c, F4 proteins were first 33 P-phosphorylated in the presence of Ca 2ϩ and calmodulin (Fig. 5, lane 1). After 30 min of incubation, endogenous CaMKII was inhibited by adding 2 M staurosporin to the assay medium. In this particular case, we ought to use staurosporin instead of KN-93, because, once autophosphorylated, CaMKII remains active even in the absence of Ca 2ϩ -calmodulin (autonomous form). The membranes were then solubilized and immunoprecipitated with anti-G M antibody. Fig. 5 clearly demonstrates that radiolabeled G M was completely dephosphorylated by endogenous PP1c (lanes 1 and 2). Indeed, the level of G M phosphorylation remained constant when the PP1c inhibitor okadaic acid was also present in the incubation medium (Fig. 5, lane 3).
Identification of CaMKII Phosphorylation Sites of G M -All of the phosphorylation sites thus far identified in G M are clustered within 28 residues (Ser 40 to Ser 67 ) located at the NH 2 terminus of the molecule. Therefore, we investigated the possibility that CaMKII might phosphorylate G M within this phosphoregulatory region. A His-tagged recombinant protein corresponding to aa 40 -338 of G M (G M -(40 -338)) was expressed in E. coli and purified according to the manufacturer's instructions. Fig. 6A shows that, when added to SR fraction, G M -(40 -338) was phosphorylated by endogenous CaMKII (lane 5) and that this phosphorylation was almost completely abolished by KN-93 (lane 6). Fig. 6B shows that the G M -(40 -338) fusion protein was also phosphorylated by exogenous, commercially available CaMKII. No phosphorylation of G M -(40 -338) fusion protein occurred in the absence of activated CaMKII (not shown).
Based on the known consensus sequence for CaMKII (37,38), there are six potential Ser or Thr phosphoacceptor sites in G M -(40 -338) fusion protein: Ser 48  cial exogenous CaMKII showed that only serine residue(s) were affected by the kinase (results not shown).
Further experiments were carried out to identify the target residues of CaMKII. To this purpose, an additional His-tagged G M fusion protein lacking aa 40 -68 (G M -(69 -338)) was generated. Fig. 7A shows that the phosphorylability of G M -(69 -338) by both CaMKII and PKA was completely lost upon deletion of the sequence containing Ser 48 and Ser 67 (lanes 6, 8). These results suggest that the serine residue(s) phosphorylated by CaMKII are located in the 40 -68-amino acid sequence. It is noteworthy that PKA phosphorylation of G M -(40 -338), which reached a stoichiometry of 1 mol/mol protein, induced a substantial shift of the Coomassie Blue band toward higher molecular mass (Fig. 7A, lane 1), whereas CaMKII phosphorylation did not affect the protein mobility (lane 3).
To discriminate between the two potential CaMKII phosphorylation sites, a recombinant G M -(40 -338) protein was generated in which Ser 48 was mutated to alanine (G M -(40 -338) S48-A ). As expected, when G M -(40 -338) S48-A was incubated with PKA, a substantial decrease of 33 P incorporation into the protein was observed, as compared with wild-type G M -(40 -338) (Fig. 7B,  lanes 3 and 4). The inability of CaMKII to phosphorylate the point-mutated G M -(40 -338) S48A (Fig. 7B, lanes 5 and 6) demonstrated that the only target residue affected by CaMKII is Ser 48 .
Phosphorylation of Ser 48 by PKA does not dissociate PP1c from G M . Therefore, we compared the effect of CaMKII and PKA phosphorylation on the interaction of PP1c with SR membranes. To maximize the phosphorylation by CaMKII, SR membranes were incubated in a medium containing not only Ca 2ϩ and calmodulin but also exogenous CaMKII. Under these experimental conditions and as determined by densitometry of radiolabeled proteins, we observed that the extent of phosphorylation by CaMKII of G M protein was comparable with that obtained after PKA phosphorylation of SR proteins (Fig. 8A). As a further indication of the increased efficiency of CaMKII phosphorylation under these experimental conditions, we found that the amount of radioactivity incorporated into GS was more than 2-fold higher (22.050 arbitrary counts) than that observed for PKA-dependent phosphorylation (9859 arbitrary counts). When the association of PP1c to G M was moni-tored by Western blot analysis, a substantial amount of PP1c was released from SR membranes into the supernatant, following phosphorylation by PKA (Fig. 8, lanes 3 and 4). On the contrary, CaMKII-mediated phosphorylation of G M only negligibly affected the association of PP1c to SR (lanes 5 and 6). It was verified by Western blot analysis that association of GS to G M was unaffected under either condition (not shown). DISCUSSION In skeletal muscle CaMKII is targeted to SR by a specific anchoring protein, ␣CaMKII association protein (43). Even though both the ␣CaMKII association protein and CaMKII are uniformly distributed within junctional and nonjunctional SR (40), the function of nonjunctional SR CaMKII is still unknown in fast twitch muscle.
Previous electron microscopy studies in resting skeletal muscle (44) showed that glycogen particles associate to SR at the level of the I band, where the highest density of nonjunctional SR tubules is observed. This association is preserved under conditions of sucrose density centrifugation (45), likely because of the presence of the glycogen-and PP1c-targeting subunit G M . It is also well established that glycogen granules contain enzymes involved in their metabolism, such as GS and glycogen phosphorylase (46). Recent studies implicated CaMKII in regulation of glucose metabolism in skeletal muscle (41, 47). This involvement seems reasonable, because during physical exercise cytosolic [Ca 2ϩ ] increases, thereby activating Ca 2ϩcalmodulin-dependent enzymes. As a matter of fact, activity of CaMKII is increased by exercise in human skeletal muscle (48).
Based on immunofluorescence and subfractionation studies, here we demonstrated that G M , one of the key regulatory components of glycogen metabolism, selectively localizes to SR. Furthermore we show that G M is a component of the highly specialized Ca 2ϩ pump membrane forming nonjunctional tubules of SR.
Our subcellular localization studies do not support the previous report by Walker et al. (14) suggesting the presence of two pools of G M in skeletal muscle, one cytosolic and one bound to the SR. Our experiments demonstrate unequivocally that G M previously detected in the low speed, post-mitochondrial supernatant was due to the contamination by unsedimented SR vesicles. Based on this evidence, we conclude that in skeletal muscle only one form of G M is present that localizes to the nonjunctional SR.
Our work also shows that a structural complex containing G M , GS, and PP1c exists in native SR membranes. Using COS7 cells or C 2 C 12 myotubes transiently expressing G M, Liu and Brautigan (19) reported that transfected G M immunoprecipitated with endogenous GS. However, our data provide the first direct experimental evidence for the presence of this complex in native nonjunctional SR membranes isolated from fast twitch skeletal muscle.
The main goal of the present study was to investigate whether nonjunctional SR-bound CaMKII might phosphorylate key proteins of glycogen metabolism. We have shown that endogenous CaMKII phosphorylates G M and that phospho-G M is dephosphorylated by endogenous PP1c. By using G M recombinant fragments and site-directed mutagenesis, we demonstrate that CaMKII phosphorylated Ser 48 in the NH 2 terminus of G M . This finding is consistent with previous observations (10,15) reporting high levels of Ser 48 phosphorylation in vivo, as a consequence of phosphorylation by muscle contractionactivated protein kinase(s) (14). This serine residue is phosphorylated by PKA in vivo (14) and in vitro (4), as well as, possibly, by an insulin-dependent protein kinase (14,19,49). When G M is phosphorylated at Ser 48 , PP1c activity toward GS and phosphorylase kinase is much higher than that of unphosphoryl- ated G M (19,49). This, in turn, should lead to reactivation of GS and hence to the resynthesis of glycogen (14).
Our results support the concept that G M , like GS, is a multisite phosphorylated protein and that the domain of the protein critical for regulation by reversible phosphorylation is the region near the amino terminus, termed the phosphoregulatory domain (49). In this study, we considered the possibility that G M might directly interact with CaMKII, via the transmembrane domain of G M , on the analogy of what was observed for the slow twitch muscle integral SR membrane protein, phospholamban (13). However, co-immunoprecipitation experiments failed to provide evidence supporting a direct protein-protein interaction between CaMKII and the heteromeric G M , GS, and PP1c protein complex (not shown).
The physiological significance of the CaMKII-mediated phosphorylation of G M Ser 48 is complex. Because CaMKII is activated during exercise, a role for CaMKII in inhibition of GS should be conceivable. However, it is also possible that phosphorylation of Ser 48 of G M by CaMKII plays a role during the recovery period. On this respect, it is interesting to remember the unique regulatory features of CaMKII, which, once autophosphorylated, becomes independent on regulation by Ca 2ϩ and calmodulin. Therefore the signal conveyed by this autonomous form of CaMKII might well outlast that of adrenergic  2, 4, and 6) were phosphorylated by PKA (lanes 3 and 4) or activated CaMKII (lanes 5 and 6), as described for A. Lanes 1 and 2, Coomassie Blue staining; lanes 3-6, autoradiography. stimulation. CaMKII-dependent phosphorylation of Ser 48 might provide a potential mechanism for reactivating glycogen resynthesis, by increasing the rate of dephosphorylation of GS during the recovery period after exercise, when adrenergic stimulus is terminated.
Ser 48 phosphorylation might also provide a mechanism for regulation of autophosphorylated Ca 2ϩ -calmodulin-independent CaMKII. In fact, we observed that, in addition to G M and GS, PP1c also dephosphorylates CaMKII (not shown). As a matter of fact, this finding identifies PP1c as the major CaMKII phosphatase thus far identified in skeletal muscle SR. This observation is even more striking, because PP1c activity seems to be specific for glycogen-associated substrates (49), such as GS, glycogen phosphorylase, and G M , as also shown here. Therefore, activation of PP1c secondary to CaMKII-dependent phosphorylation of G M Ser 48 might result in dephosphorylation of autonomous CaMKII and hence termination of CaMKII signal.
It is well known that direct phosphorylation of GS by several protein kinases, such as PKA, CK1, or CK2, poorly inhibits GS activity (50,51). On this respect, the most relevant kinase is GSK3 (50,51). Like PKA, CK1, and CK2, SR-bound CaMKII also phosphorylates GS. 3 In preliminary experiments we found that the inhibition of GS activity by this phosphorylation was minimal (not shown). It is also known that GS activity can be regulated by the synergic action of two protein kinases. This is the case of phosphorylase kinase and CK1 (52), PKA and CK1 (53), and CK2 and GSK3 (50,51). Therefore, the possibility exists that CaMKII and PKA may act synergistically in the regulation of GS activity in a complex fashion also depending on the intensity and duration of exercise (54). It is well known that levels of circulating catecholamines are directly related to the intensity of the exercise.
Finally, the finding that CaMKII phosphorylates G M at site 1 suggests a synergy with GSK3 because it is well known that GSK3 can phosphorylate G M at Ser 40 and Ser 44 only following the primary phosphorylation of Ser 48 . Thus far, this was demonstrated only for PKA (15). Our results provide that CaMKII might substitute for PKA in this priming role for GSK3. It should be remembered that the functional role of GSK3 phosphorylation of G M is still undefined.
In summary, our data indicate the existence of a functional relationship between SR-bound CaMKII and the heteromeric protein complex formed by G M , GS, and PP1c. This interaction takes place in the Ca 2ϩ pump membrane of nonjunctional SR. The finding that Ser 48 of G M is phosphorylated by CaMKII supports the hypothesis that SR-bound CaMKII is involved in the modulation of glycogen metabolism in skeletal muscle during exercise. FIG. 8. Phosphorylation of G M by CaMKII does not dissociate PP1c from SR membranes. A, F4 membranes were incubated for 30 min at 30°C with 50 M [␥-33 P]ATP either with exogenous PKA or in assay medium containing 100 M Ca 2ϩ , 1.0 M calmodulin, and activated exogenous CaMKII. After 5-10% SDS-PAGE, 33 P-labeled proteins were detected by autoradiography of dried gels, and the radioactive areas corresponding to radiolabeled G M were determined by densitometry (see "Experimental Procedures"). B, F4 membranes were incubated for 30 min at 30°C with 50 M unlabeled ATP as in A, in assay medium without addition (lanes 1 and 2), or containing 100 M Ca 2ϩ , 1.0 M calmodulin and activated exogenous CaMKII (lanes 3 and 4) or PKA (lanes 5 and 6). At the end of the incubation, the membrane suspensions were centrifuged with an air-driven ultracentrifuge at 22 p.s.i. for 30 min to separate pellets and supernatants. The proteins were analyzed by 5-10% SDS-PAGE, and the blots were probed with the anti-PP1c antibody. Lanes 1, 3, and 5, pellets; lanes 2, 4, and 6, supernatants.