Disruption of the Striated Muscle Glycogen Targeting Subunit PPP1R3A of Protein Phosphatase 1 Leads to Increased Weight Gain, Fat Deposition, and Development of Insulin Resistance

Recombinant phage containing genomic DNA of the G_M locus were isolated from a 129/Ola mouse library using a cDNA probe encoding the NH₂-terminal 691 amino acids of G_M. The G_M targeting vector was constructed by subcloning an 8.2-kb EcoRI/XbaI fragment with the XbaI cut end filled in using Klenow enzyme, into the EcoRI/XhoI (filled in) sites of a modified Bluescript pKS⁺ in which the XbaI site was destroyed. The construct was then digested with StyI to excise a 2.6-kb region spanning the entire exon 1 of G_M (encoding amino acids 1–262) and including 0.168 kb of the promoter region and 1.638 kb of the first intron. Subsequently, the StyI cut plasmid DNA was modified by the ligation of complementary oligonucleotides to create an XbaI site at this position. A loxP-flanked HSVtk/neo cassette containing a herpes simplex virus thymidine kinase gene (conferring gancyclovir sensitivity) and a neomycin phosphotransferase gene (conferring resistance to the antibiotic G418) under the control of a phosphoglycerate kinase promoter was inserted into the newly generated plasmid at the position of the XbaI site. The resultant G_M targeting vector comprised a 2.5-kb 5′ homology arm and a 3.2-kb 3′ homology arm separated by the HSVtk/neo cassette. Before electroporation, the targeting vector was linearized at the unique NotI site.

E14-TG2a-IV ES cells (10⁷; 129/Ola strain), cultured according to standard protocol (23), were electroporated with linearized targeting vector and selected after 48 h in G418. G418-resistant ES cell clones were obtained after 10 days and screened by Southern blotting of XbaI and PstI digests of clone DNAs using hybridization probes flanking and external to the vector homology arm sequences (see Fig. 1). Several G418-resistant clones were identified containing the targeted disruption, and one of these was chosen for an additional round of transfection to remove the HSVtk/neo cassette from the targeted allele by Cre-mediated recombination between its directly repeated flanking loxP sites. This involved electroporation with a Cre expression construct pBS185 (24) and application of gancyclovir selection after 6 days. Gancyclovir-resistant clones were isolated after 12 days, and excision of the cassette was confirmed by Southern blot analysis as described above and in Fig. 1. ES cells from two gancyclovir-resistant clones were then used to generate G_M^−/− mice, by standard protocol, using C57BL/6J blastocysts and pseudopregnant female mice as foster mothers (25). Chimeric male mice were test-crossed to C57BL/6 female mice and agouti offspring screened by Southern blotting for the presence of the disrupted G_M allele. Mice heterozygous for the disrupted G_M allele were back-crossed to C57BL/6J for several generations. Second- and third-generation back-cross heterozygotes were intercrossed to produce mice homozygous for the disrupted allele, heterozygous and wild type (WT), which were used in the analyses described herein. All mice were maintained in temperature- and humidity-controlled conditions with a 12-h light/12-h dark cycle and allowed food and water ad libitum. Mice were genotyped by Southern blotting of tail DNA digested with XbaI (Fig. 1B) or PstI as described for ES clone DNAs above.

The animals were studied after a 16-h overnight fast and killed by suffocation in CO₂, followed by cervical dislocation, unless otherwise stated. Tissues were freeze-clamped and stored at −80°C. After freeze fracturing, skeletal muscle was homogenized at 4°C in 6 vol of 50 mmol/l Tris-HCl (pH 7.5), 150 mmol/l NaCl, 2 mmol/l EGTA, 2 mmol/l EDTA, 5% glycerol, 0.1% Triton X-100, 0.1% (vol/vol) 2-mercaptoethanol, “Complete” protease inhibitor cocktail from Roche Diagnostics Ltd (Lewes, East Sussex), and okadaic acid (Calbiochem, Nottingham, U.K.) or microcystin (Life Technologies, Paisley, U.K.) where stated. The supernatant obtained after centrifugation at 16,000g for 10 min was snap-frozen in liquid nitrogen and stored at −80°C. Glucose levels were measured in blood taken rapidly from the tail veins of mice fasted for 12 h using a Glucotrend 2 meter (Roche Diagnostics). Glucose tolerance tests were performed on mice after a 16-h overnight fast. Mice were injected with 2 mg of d-glucose/g i.p., and blood glucose levels were determined immediately before and at 15, 30, 60, and 120 min after injection. Insulin tolerance was assessed by measuring blood glucose levels after mice had received an intraperitoneal injection of 0.75 mU/g insulin (Human Actrapid, 100 iU/ml; Novo Nordisk Pharmaceuticals Ltd, Crawley, Sussex, U.K.) after a 6-h fast. In vivo tissue glucose transport was determined after intraperitoneal injection of 2-deoxy-d-[1,2-³H]-glucose mixed with 20% dextrose (2 g/kg body wt; 10 μCi/mouse) according to Zisman et al. (26). For determining the effects of insulin on enzymes in skeletal muscle, mice that were anesthetized with 6 mg of Sagatal/100 g body wt (Rhone Merieux Ltd, Harlow, U.K.) were infused with 2 mU of insulin into the femoral vein over a period of 10 min. Animals were killed by cervical dislocation, and skeletal muscle was freeze-clamped and stored in liquid nitrogen. Responsiveness of the animals to insulin was confirmed by determination of blood glucose levels before and after insulin administration.

Antibodies to human glutathione-S-transferase‐G_M (1–243) were affinity-purified against maltose-binding protein fused to G_M. Antibodies to human PP1α (301-KNKGKYGQFSGLNPGG-316) and human PP1β (316-TPPRTANPPKKR-327) were affinity-purified against their respective peptides by Dr. Jane Leitch. The peptides were synthesized by Dr. G. Bloomberg (University of Bristol, U.K.), and antibodies were raised in sheep by Diagnostics Scotland (Penicuik, Midlothian, U.K.). Affinity-purified antibodies to glutathione-S-transferase‐human R5 protein and mouse R5 peptide (residues 36–49) have been described previously (20). Phospho-GSK-3 α/β (Ser 21/9) antibodies were purchased from New England Biolabs (Hitchin, Herts, U.K.). Immunoblotting was performed as described by Browne et al. (20).

PP1 activities were determined by release of [³²P]phosphate from phosphorylase a (10 μmol/l) or GS (3 μmol/l, phosphorylated with GSK-3β) in the presence of 4 nmol/l okadaic acid for 10 min at 30°C. For immunoadsorption of PP1-G_M and PP1-R5 with anti-G_M and anti-R5 peptide antibodies, respectively, lysates were prepared in the presence of 100 nmol/l okadaic acid. The immune pellets were washed five times in the presence of 4 nmol/l okadaic acid, and PP1 activity in the immune pellets were assayed as described above either before (“spontaneous” activity) or after (“total” activity) preincubation with 0.1 mg/ml “dissociating” peptide that causes the release of free PP1c from the glycogen-targeting subunit (20). One unit of activity is the amount of enzyme that catalyzes the release of 1 μmol of ³²P-phosphate per minute. For GS and phosphorylase assays, skeletal muscle was homogenized in 3 vol of 25 mmol/l Tris (pH 7.5), 50 mmol/l NaF, 2 mmol/l EDTA, 2 mmol/l EGTA, 1 mmol/l sodium orthovanadate, 1 μmol/l microcystin, 0.1% 2-mercaptoethanol, and “Complete” protease inhibitor cocktail (Roche) and centrifuged at 13,000g for 5 min. GS activity in the supernatant was determined by measuring incorporation of [¹⁴C]glucose from UDP-[¹⁴C] glucose into glycogen in the absence or presence of 7.2 mmol/l glucose-6-phosphate (G-6P) (27). Glycogen phosphorylase activity was determined by the incorporation of [¹⁴C]glucose-1-phosphate (Amersham Biosciences, Little Chalfont, Bucks, U.K.) into glycogen in the absence or presence of the allosteric activator AMP (2 mmol/l) (28). Glycogen was measured with anthrone reagent after extraction from skeletal muscle with 1 mol/l NaOH at 100°C for 60 min (29).

To examine the physiological role of G_M, we disrupted the murine G_M gene by homologous recombination, deleting all of exon 1, which encodes the translation initiation codon as well as the PP1 and glycogen-binding domains (Fig. 1A). The G_M^−/− mice created had no obvious developmental or morphological defects in the first 3 months of life and gave the expected ratio of male/female offspring for an autosomal gene. Analysis of skeletal muscle from the G_M^−/− mice showed no G_M protein on immunoblotting and decreased levels of PP1β, whereas the level of PP1α was unchanged (Fig. 1C), suggesting that the PP1β isoform interacts with G_M and that its expression is decreased in the absence of the targeting subunit. Assay of immunoadsorbed PP1-G_M using glycogen phosphorylase as substrate confirmed the absence of PP1-G_M activity (Fig. 1D). Consistent with these results, PP1 activity in skeletal muscle lysates of G_M^−/− and G_M^−/+ mice was decreased by 49 and 22%, respectively, compared with WT levels, indicating that G_M binds approximately half of the PP1 in skeletal muscle (Fig. 1E).

A major metabolic difference between G_M^−/− and WT mice was that the glycogen levels in skeletal muscle of G_M^−/− mice were drastically decreased to ∼10% of their WT littermates (Fig. 2A). This result is consistent with the tenet that PP1-G_M dephosphorylates and activates GS, leading to an increase in glycogen synthesis. Although fasting blood glucose levels of G_M^−/− and WT mice up to 11 months of age showed no statistically significant differences (Fig. 2B), 11-month-old mice exhibited glucose intolerance (Fig. 2C). After intraperitoneal administration of a bolus of glucose, blood glucose levels in G_M^−/− male mice rose to ∼40% higher levels than those observed in WT mice, and blood glucose was much slower to return to basal levels (Fig. 2C). In addition, mice older than 11 months developed insulin resistance (Fig. 2D). The weights of G_M^−/− and WT mice were similar at 3 months of age, but thereafter, both male and female G_M^−/− mice gained weight more rapidly, weighing ∼20% more than WT littermates when older than 1 year (Fig. 3A and C). Deposition of fat was readily apparent in G_M^−/− mice with increased fatty tissue around the heart and massive increases in abdominal fat (Fig. 3B and D). G_M^−/− mice older than 1 year were not only obese but also showed a small increase in body length compared with their WT littermates, suggesting that alteration of metabolism influences the growth of mice, which is known to continue throughout adult life.

In mice older than 12 months, basal glucose transport, estimated by glucose tolerance tests incorporating the tracer 2-deoxy-d-[1,2-³H]-glucose (2-DOG) (26), was decreased in skeletal muscle by 69% in G_M^−/− mice compared with WT (Fig. 4A), indicating that this tissue is insulin-resistant in G_M^−/− mice. In contrast, glucose transport into adipose tissue (Fig. 4B), glycogen levels, and GS activity in adipose tissue (Table 1) were similar in G_M^−/− and WT mice. Incorporation of 2-DOG into glycogen, through the direct pathway (30), was not significantly elevated in the adipose tissue (WT 0.030 ± 0.01, G_M^−/− 0.069 ± 0.01 μmol · min⁻¹ · mg⁻¹ glycogen, P < 0.2) or liver (WT 0.107 ± 0.02, G_M^−/− 0.101 ± 0.02 μmol · min⁻¹ · mg⁻¹ glycogen; P < 0.5) of G_M^−/− mice. However, low levels of 2-DOG incorporation and low glycogen content make precise ascertainment of these values difficult for adipose tissue. Overall, it seems likely to be largely the presence of more adipose tissue in older G_M^−/− mice (abdominal adipose tissue is increased approximately fourfold compared with WT) that accounts for the uptake of glucose that is not utilized by skeletal muscle.

To assess whether basal or insulin-stimulated GS activity was impaired in skeletal muscle by disruption of the G_M gene, we infused, intravenously, saline or doses of insulin to achieve levels only slightly higher than physiological concentrations and took samples of muscles from the hind limbs of G_M^−/− mice and their WT littermates. The basal GS activity was only 0.1 in the skeletal muscle of G_M^−/− mice compared with 0.3 for WT littermates (Fig. 5A). Consequently, although insulin increased the GS activity by 1.5- to 2-fold in G_M^−/− and WT mice, the activity in G_M^−/− mice did not reach even the basal level observed in WT mice, rising only to 0.2 after stimulation with insulin. Consistent with the effects of insulin on GS activity, there was significant phosphorylation of GSK-3α at Ser21 and GSK-3β at Ser9 (Fig. 5B). The low basal and insulin-stimulated GS activities are in accordance with the very low glycogen level observed in G_M^−/− mice, but, interestingly, differ from the usual situation in which the activity of GS is regulated by the glycogen content, a low glycogen level being associated with a high activity state of GS (31). Our results demonstrate that loss of G_M overrides the normal effects of glycogen on GS activity. Loss of G_M also leads to an activation of phosphorylase (Fig. 5C), indicating that the lower glycogen content of G_M^−/− skeletal muscle may arise from the combined effects of inactivation of GS and activation of phosphorylase. However, the total levels of both phosphorylase and GS are decreased by 40–50% in G_M^−/− mice compared with WT, probably because the enzymes are more easily degraded when their glycogen-binding sites are limited by the low glycogen levels.

To ascertain whether PP1-G_M was being stimulated by insulin in WT mice, we assayed its activity in the absence and presence of a peptide that dissociates PP1 from G_M after immunoadsorption of PP1-G_M from the muscle extracts (20). These assays showed that in WT mice, PP1-G_M activity was stimulated slightly by insulin and that the total PP1 activity associated with G_M was increased (Fig. 6A). As expected, PP1-G_M activity was absent from G_M^−/− mice, indicating that the assays were specific for this particular glycogen-targeted form of PP1. Immunoblotting with an anti-PP1β antibody raised against a COOH-terminal peptide revealed an increase in PP1β immunostaining after insulin stimulation (Fig. 6B).

Although the level of insulin-stimulated GS activity was low in G_M^−/− mice, insulin stimulated basal GS activity 1.5- to 2-fold in G_M^−/− mice (Fig. 5A), raising the question of which phosphatase was dephosphorylating GS under these conditions. We therefore tested whether PP1-R5/PTG, which is expressed at appreciable levels in skeletal muscle (18), responds acutely to insulin in WT and G_M^−/− mice. After infusion with saline, assays carried out after immunoadsorption of PP1-R5/PTG revealed a decrease in the basal activity in the skeletal muscle of G_M^−/− mice, probably as a consequence of the low glycogen levels (Fig. 6C). It is interesting that although PP1-R5/PTG activity showed little or no stimulation in WT mice after infusion with insulin, the activity of PP1-R5/PTG doubled in G_M^−/− mice upon insulin stimulation and now approached the level of activity found in WT mice. In addition, as found for the G_M complex, the immunostaining of the PP1β present in the PP1-R5/PTG immunopellets increased (Fig. 6D). The low levels of PP1-R6 activity in murine skeletal muscle precluded ascertainment of whether this glycogen-targeted form of PP1 was acutely regulated by insulin in WT or G_M^−/− mice.

The very low glycogen levels in skeletal muscle of G_M^−/− mice result from a decreased −/+G-6P GS activity ratio and an increased −/+AMP phosphorylase activity ratio. They are in accordance with G_M being the most abundant glycogen-targeting subunit of PP1 in rodent skeletal muscle, dephosphorylating both GS (with activation) and phosphorylase (causing inactivation). The phenotype demonstrates that PP1-G_M is an essential enzyme for the maintenance of the normal skeletal muscle glycogen levels. In addition, our data show that G_M specifically interacts with PP1β and not other isoforms of PP1 (Fig. 6B), consistent with previous studies (32,33).

The increased weight gain of G_M^−/− mice from 3 months of age and massive increases in adipose tissue present at 12 months of age are consistent with the concept that when blood glucose cannot be converted into glycogen in skeletal muscle, much of this glucose is taken up by adipocytes and converted into fat deposits, which increase gradually with the age and lead to obesity. Glucose intolerance, which is clearly evident in G_M^−/− mice at 11 months of age, and the subsequent development of insulin resistance may arise as a consequence of the increased fatty acids present in G_M^−/− mice of this age (34,35). Because glucose intolerance and insulin resistance usually precede overt diabetes, our results suggest that disruption of G_M may predispose to insulin resistance and diabetes. In accordance with this data, association of PPP1R3/G_M mutations with insulin resistance have been noted in some human populations (36–38) but not in others (39–41). For example, a mutation-lowering expression of PPP1R3 (G_M) was associated with diabetes in the Pima Indians (37) and a Japanese population (38). In addition, interaction of a truncated G_M with a peroxisome proliferator‐activated receptor γ mutant was found to give rise to a rare severe insulin resistance phenotype (42).

Suzuki et al. (15) recently reported a G_M/R_GL knockout model in which the mice exhibited very low skeletal muscle glycogen levels. However, in contrast to our G_M^−/− mice, those in the study of Suzuki et al. did not show weight gain, fat deposition, glucose intolerance, or insulin resistance when examined up to 12 months of age. The routes for disposal of glucose that is not utilized for glycogen synthesis in the skeletal muscles of the Suzuki et al. G_M knockout mice are unclear. Both G_M knockout models have been produced in very similar genetic backgrounds of 129 crossed to C57BL/6J, although the 129 substrains differ (129/Ola compared with 129/Sv). The precise nature of the genetic modification that abrogates the PP1 and glycogen-binding domains of G_M in the two knockout models is slightly different; the one described herein deletes a larger region of both promoter and intron 1 sequence. Moreover, in this model, the selection markers have been removed by Cre/loxP recombination to avoid any potential complicating effects as a result of their interference with the expression of neighboring genes, whereas the selection marker is retained in the model of Suzuki et al. Nevertheless, the difference between the two G_M^−/− models is of considerable interest because it may provide an avenue to uncover the factors that lead to obesity, glucose intolerance, and insulin resistance when blood glucose conversion into muscle glycogen is inadequate for the postprandial removal of blood glucose.

In the G_M^−/− mice reported here, both the basal −/+G-6P GS activity ratio and −/+G-6P GS activity ratio measured after stimulation by near physiological doses of insulin were observed to be low. Disruption of the G_M-targeting subunit of PP1 therefore causes a major decrease in the GS activity ratio that cannot be overcome by insulin treatment under the conditions that we have studied. The data indicate that PP1-G_M may contribute to action of insulin on GS activity in skeletal muscle in these conditions. Although Suzuki et al. found that upon insulin stimulation GS activity in the skeletal muscle of G_M^−/− mice rose well above the basal levels in WT mice, they used higher doses of insulin (2 units/kg i.v.) to obtain this stimulation.

To examine the activities of different glycogen-targeted forms of PP1 separately, we developed a coupled glycogen-targeting subunit immunoadsorption and PP1 assay system (20). In contrast, previous assays have assessed the effect of hormones on total PP1 activity or a mixture of different glycogen-targeted forms of PP1 (15,43–46). Our PP1 assay is also performed in the presence of a peptide that dissociates the glycogen-targeting subunit from PP1 (measuring of total PP1 activity bound to the targeting subunit) or in the absence of the peptide (measuring the spontaneous activity of the glycogen-targeted PP1 complex). The activity of the free catalytic subunit of PP1 is unaffected by the dissociating peptide. Using this assay, we observed an increase in the total PP1 activity bound to G_M after insulin treatment in the skeletal muscle of WT mice. An increase in G_M-associated PP1β immunostaining after insulin stimulation was also observed and may reflect an real increase in the level of PP1β or alternatively the COOH-terminus of PP1β may undergo a posttranslational modification that allows it to be recognized more efficiently by the antibody. Because this region contains two threonine residues, it is possible that the activity of PP1β bound to G_M may be regulated by insulin via reversible phosphorylation.

In contrast to PP1-G_M, PP1-R5/PTG activity was not appreciably stimulated by insulin in the skeletal muscle of WT mice, but both the spontaneous and total PP1 activity associated with R5/PTG was stimulated by insulin in G_M^−/− mice with a parallel increase in PP1-β immunostaining. Thus, the absence of PP1-G_M leads to a compensatory effect of insulin on PP1-R5/PTG. In summary, our results are consistent with the following model. In WT mice, insulin-activates GS not only by inhibiting GSK-3 but also by activating PP1-G_M. The activation of PP1-G_M is caused by an increase in the activity or amount of PP1β bound to the glycogen-targeting subunit. In G_M^−/− mice, this modification of PP1β elevates the activity of the PP1-R5/PTG complex, but the activity of PP1-R5/PTG in the presence or absence of insulin is insufficient to restore GS activity to the levels observed in WT mice.

The work was supported by the Medical Research Council, U.K. and Diabetes U.K. The CGR Gene Targeting Laboratory was supported by the Biotechnology and Biological Sciences Research Council, U.K. M.D. is a recipient of a postgraduate studentship from the Royal Society and holds the Pat and Muriel McPherson Studentship of the Medical Research Council Protein Phosphorylation Unit. C.G.A. was supported by a postdoctoral research assistantship from Diabetes U.K.

We thank Gareth Browne and Ann Burchell for helpful advice and Derek Black for technical assistance.