Human Skeletal Muscle Expresses a Glycogen-Targeting Subunit of PP1 That Is Identical to the Insulin-Sensitive Glycogen-Targeting Subunit G_L of Liver

Vastus lateralis biopsies were obtained from three healthy, male volunteers (21 ± 0 years; 72.3 ± 3.3 kg; 176 ± 1 cm; BMI 23.3 ± 1.35 kg/m²). The Tayside Ethics Committee approved the protocol, and the nature, purpose, and possible risks were explained to each subject before written consent was obtained. Subjects fasted overnight for 12 h. Under local anesthesia (1% lignocaine, no adrenaline), an incision of 10 mm length and depth was made in the skin and muscle fascia at ∼20 cm above the knee, using an aseptic technique. Vastus lateralis biopsies (∼100 mg) from the right quadriceps femoris muscle were obtained using Tilley-Henkel forceps. Samples were blotted dry and grossly dissected free of fat and connective tissue, frozen in liquid nitrogen and stored at −80°C before analysis. Soleus and gastrocnemius biopsies were taken from a patient in the fasted state under general anesthesia, with the patient’s consent, during lower-limb surgery and were immediately frozen in liquid nitrogen. Liver biopsy samples were from control subjects or patients who had been previously studied for glycogen storage diseases, and the tissue was stored at −80°C (37).

The National Center for Biotechnology Information (NCBI) expressed sequence tag and cDNA databases were searched for sequences relating to rat G_L using the tBLASTN algorithm. A human testis cDNA clone in plasmid pT7T3d-Pac (accession no. AA954470) and a mouse embryonic cDNA in plasmid pME18S-FL3 (accession no. AI115058) were provided by the IMAGE consortium. Sequence analysis of the plasmid inserts was performed in both directions on an Applied Biosystems 373A automated DNA sequencer using Taq dye terminator cycle sequencing by the University of Dundee DNA sequencing service managed by Dr. Nick Helps (www.dnaseq.co.uk). The human clone, which contained 200 bp that were highly similar to nucleotide (nt) 1–200 of rat G_L, was used in a hybridization screen of a human Bacterial Artificial Chromosome (BAC) library (performed by Genome Systems, St. Louis, MO). Three positive human genomic clones in pBELOBAC gave identical patterns after cleavage with EcoRI, HindIII, PstI, SacI, and AvaI; Southern blotting; and hybridization with the G_L coding region. Sequence analysis showed that the clones were identical from −800 to +852 of the G_L gene.

Human liver and testis Matchmaker cDNA libraries (Clontech, Palo Alto, CA) were used as templates to amplify nt 1–200 of G_L using the Hi-Fidelity PCR system (Roche Diagnostics, Lewes, Sussex, U.K.) according to the manufacturer’s instructions with the forward primer 1 (nt 1–23, 5′ ATGGCTGTGGACATCGAGTACAG 3′) and reverse primer 2 (complementary to nt 210–180, 5′ GTTGTCTGCGAA GGACACCCGC 3′). Marathon-Ready liver and skeletal muscle cDNA libraries (Clontech) were used as templates to amplify the full-length coding sequence for G_L by PCR using the forward primer 1 and primer 3 (complementary to nt 852–827, 5′ GTAGTAGGGCCCTAGCTTTTCATATC 3′). 5′ and 3′ rapid amplifications of cDNA ends [5′ and 3′ rapid amplification of cDNA ends (RACE), (38)] were performed using the same libraries. Gene-specific primers were primer 1 and primer 4 (complementary to nt 401–374, 5′-CCG ATT TCT AAA GTC TAA GTA ATC TGC-3′) for the 3′ and 5′ RACE respectively. PCR products were resolved on 1% (wt/vol) agarose gel, gel-purified, cloned into pCR2.1 TOPO vector (Invitrogen, Carlsbad, CA), and sequenced in both directions using M13 forward and reverse primers.

Purified BAC or cosmid DNA was labeled with the fluorochrome Spectrum Red using Vysis Nick Translation Kit, as described in the manufacturer;s protocol (Vysis, Downers Grove, IL). The labeled probe was used for hybridization of metaphase chromosome spreads from lymphocyte cultures, and the slides were counterstained with 4′, 6-diamidino-2-phenylindole dihydrochloride.

Chromosome 8 and 10 cosmid libraries, constructed at Los Alamos National Laboratory, were kindly provided by the U.K. Human Genome Mapping Project (HGMP) Resource Center in the form of high-density gridded filters. The libraries had an average insert size of 30–40 kb and contained ∼20,000 clones providing a four- to fivefold coverage of each chromosome. The filters were hybridized with the G_L coding region (852 bp). Four chromosome 8 clones (LA0835p13, LA0848n12, LA0817a8, and LA0822f15 from the LA08NC01 library, kindly provided by the HGMP, were sequenced.

Northern blots (Clontech) contained ∼2 μg poly (A)⁺ RNA from different tissues. Alternatively, total RNA was prepared using TRIzol Reagent (Life Technologies, Paisley, U.K.). Blots were hybridized according to Clontech’s instructions with the last wash in 15 mmol/l NaCl:1.5 mmol/l sodium citrate:0.1% SDS at 55°C.

Rodents were killed by suffocation in CO₂, and tissues were excised, freeze clamped, and stored at −80°C. Heart and skeletal muscle were freeze-fractured −80°C and homogenized at 4°C in six volumes 50 mmol/l Tris-HCl, pH 7.5, 150 mmol/l NaCl, 2 mmol/l EGTA, 5% glycerol, 0.1% Triton X-100, 0.1% (vol/vol) 2-mercaptoethanol, and “Complete” protease inhibitor (Roche Diagnostics, Lewes, U.K). Homogenates were centrifuged at 16,000g for 10 min, and the supernatants were decanted, snap-frozen in liquid nitrogen, and stored at −80°C. Liver was homogenized in three volumes of 2 mmol/l EDTA, pH 7.0, 2 mmol/l EGTA, 250 mmol/l sucrose, 0.1% (vol/vol) 2-mercaptoethanol, and “Complete” protease inhibitor and were processed similarly. Proteins in the 16,000g supernatants were separated by 4–12% SDS-PAGE, transferred to nitrocellulose, and probed with 0.2 μg/ml affinity purified antibodies raised in sheep to rat G_L (24), human PP1β peptide (amino acids 316–327), human G_M (amino acids 1–243), or a control antibody raised in mice to GAPDH (Research Diagnostics, Flanders, NJ). Antibody binding was determined using anti-sheep and anti-mouse IgG antibodies conjugated to horseradish peroxidase (Pierce, Rockford, IL), followed by enhanced chemiluminescence (Amersham International, Amersham, Bucks, U.K.).

Searches of the NCBI databases with the full-length sequence of rat hepatic glycogen-targeting subunit of PP1, G_L, did not uncover any human liver cDNA but did identify a cDNA derived from a human testis cDNA library that contained 200 bp encoding 66 amino acids that were highly similar to amino acids 1–66 of rat G_L. PCR with primers constructed to the ends of this cDNA fragment led to the amplification of identical 200-bp cDNA fragments from both human testis and liver cDNA libraries. Subsequent PCR using the liver library with the 5′ primer and a 3′ primer complementary to the extreme COOH-terminal coding region of rat G_L allowed the isolation of cDNA for the human G_L coding region. The 5′ and 3′ untranslated regions (UTRs) and confirmation of the extreme NH₂- and COOH-terminal coding sequences was obtained by PCR using RACE methodology. Three genomic clones from a BAC library contained an identical region encoding G_L, and the sequence obtained for the G_L gene from −3,000 to +6,000 was identical to that reported in the human genome database recently. The liver G_L cDNA was colinear with the gene sequence (from −1,880 to +2,100). In contrast, a testis cDNA (accession no. AA 824377) was missing sequences from −3,600 to −147, while a retinoblastoma cDNA (accession no. AK024067) was missing sequences from −8,913 to −21, presumably as the result of alternative splicing.

Sequencing of a mouse embryonic cDNA clone identified the complete open reading frame of a putative G_L protein of 284 amino acids. Comparison of the amino acid sequence of rat G_L with the deduced amino acid sequence of human and mouse G_L reveals high levels of identity between the proteins from the three species and shows complete conservation of the PP1 and phosphorylase a binding regions. Human G_L is 89 and 88.5% identical to rat and mouse G_L respectively, and mouse G_L is 94% identical to rat G_L (Fig. 1).

Initial mapping of the human G_L gene using a human 120- to 150-kb BAC clone as a probe gave two locations at 8p23.1 and 10q25. Screening of chromosome 8 and 10 cosmid libraries with a G_L probe identified potential positive clones on the chromosome 8 filter but no positives on the chromosome 10 filter. Sequence analysis of four chromosome 8 clones hybridizing with a G_L probe showed that they all contained a region encoding G_L identical to that in the BAC genomic clones. Mapping using a smaller cosmid clone (30–50 kb) as a probe gave only the location 8p23 (Fig. 2), consistent with the map location cited in the HGMP database.

Restriction mapping, Southern blotting, and sequencing of three BAC genomic clones established that there was only one G_L gene within each clone, indicating that G_L is encoded by a single gene in the human genome.

Analysis of the tissue distribution of human G_L mRNA shows that G_L cDNA hybridizes to a single RNA species of 6.7 kb (Fig. 3A). However, surprisingly, highest levels are observed to be in skeletal muscle and heart and not in liver. In addition, G_L was present at low levels in placenta, lung, and kidney.

As this was an unexpected result, the rat mRNA expression was reanalyzed. The highest levels of 5-kb G_L mRNA were present in liver as reported previously (12), but a longer exposure time revealed that rat G_L mRNA shows appreciable expression in the heart and is evident at lower levels in other tissues, such as lung, kidney, and spleen (Fig. 3B). Mouse G_L mRNA (4.4 kb) shows a similar tissue distribution to rat G_L mRNA, with highest levels being present in liver and appreciable levels present in heart (Fig. 3C). Although there were no detectable levels of G_L mRNA in rat and mouse skeletal muscle on the Clontech blot (Fig. 3B and C), subsequent analyses of total RNA from skeletal muscle of mouse (Fig. 3D) and rat (data not shown) detected low levels of G_L mRNA.

Analyses of G_L at the protein level focused upon the insulin-sensitive tissues heart, liver, and skeletal muscle, as these tissues showed the highest levels of G_L expression at the mRNA level. Immunoblotting of rat and mouse tissue extracts with an anti-rat G_L antibody confirmed that liver exhibits the highest levels of G_L (Fig. 4A and B). Bands of the same size (33 kDa) were also present in rat kidney, heart, testis, spleen, and lung, and very low levels were detectable in skeletal muscle, indicating that G_L is not exclusively liver specific, as had been thought to be the case previously.

Analysis of human liver and skeletal muscle biopsy samples for the presence of G_L by immunoblotting with the anti-rat G_L antibody, revealed a protein migrating at the same size as rat liver G_L in both human liver and skeletal muscle extracts (Fig. 5). Although the mRNA analysis suggested that higher levels of human G_L mRNA are present in skeletal muscle than in liver, immunoblotting showed human G_L protein to be present at appreciable and fairly similar levels in these tissues.

G_L was examined for variation in the level of expression in different types of muscle. In mouse (Fig. 4B) and rabbit (data not shown), G_L levels were very low in the gastrocnemius, psoas, and soleus muscles and no variation was observed. The expression of G_L is appreciable in three different human muscles, with the level in soleus being slightly higher than those in the vastus lateralis and the gastrocnemius (Fig. 5B). Assessment of the G_L-to-G_M ratio was determined by specific immunoprecipitation assays in the presence of a peptide that dissociates the glycogen-targeting subunits from PP1c (25). In all three human muscles, the total PP1c activity bound to G_L was found to be ∼40% of that bound to G_M (Fig. 5B and data not shown).

Although the sizes of the human G_L mRNA and protein were identical in liver and skeletal muscle, there remained a possibility that their sequences differed due to alternative splicing of the mRNA in the two tissues. G_L cDNA was therefore amplified by PCR from a human skeletal muscle cDNA library and compared with that amplified by PCR from the liver cDNA library. Amplification of the coding region gave fragments of identical size and sequence in skeletal muscle and liver (Fig. 6).

RACE methodology shows that the 5′ UTRs from both skeletal muscle and liver cDNA libraries commenced at −1,800 with a sequence identical to the genomic sequence. Amplification of the 3′ UTRs gave fragments that terminated at +2,100 with the liver cDNA library and terminated at +2,100 and +5,380 with the skeletal muscle cDNA library, the sequence of all fragments being identical to the genomic G_L sequence. The larger skeletal muscle cDNA amplified (−1,880 to +5,380) is comparable in size to the mRNA (6.7 kb) detected in both liver and muscle by blotting (Fig. 3A).

We show here that the glycogen-targeting subunit, G_L, of PP1 is very highly conserved, with >88% identity in amino acid sequence between mammalian species. In comparison, G_M shows 73% identity between rabbits and humans (10) and R5/PTG shows 84% identity between mouse and humans (14,15). The regulatory phosphorylase a binding site at the extreme COOH-terminus of G_L is invariant between species, suggesting that allosteric regulation of G_L-PP1c activity by phosphorylase a is crucially important in mammals. Although G_L was believed to be specific to liver (3,11,12), analyses of human, rat, and mouse tissues show that the G_L mRNA is present at significant levels in heart and at low levels in several rat tissues, including kidney and lung. More surprising is that G_L mRNA is more abundant in human skeletal muscle than in liver. Immunological detection of G_L protein at high levels in liver and low levels in heart and skeletal muscle of rat and mouse correlates with the mRNA results for these species. Human G_L protein was detected immunologically in biopsy samples at fairly similar levels in liver and skeletal muscle. The disparity between the human mRNA and protein analyses did not appear to arise from different levels of expression of G_L in different types of muscle, since the level in soleus, comprising slow fatigue-resistant fibers, was only slightly higher than those in the gastrocnemius and vastus lateralis muscles that are composed of both fast and slow fibers in similar proportions. In situ hybridization and sequence analysis of genomic clones indicates that there is only a single human gene encoding G_L at 8p23.1. There was no evidence for alternative splicing of the liver and skeletal muscle mRNA since the cDNA sequences from these tissues were identical and only a single mRNA species was detected by Northern blotting of several tissues in each species. A possible explanation is that the translation of that mRNA may be controlled differentially in liver and skeletal muscle by tissue-specific factors. Although three further human genes (PPP1R3E, PPP1R3F, and PPP1R3G) encoding related glycogen-targeting subunits of PP1 have recently been reported (39), the encoded proteins show <30% identity to G_L. PPP1R3F and PPP1R3G encode proteins larger than G_L (79 and 38 kDa, respectively). The sizes of the PPP1R3E mRNA transcripts (5.9 and 7.2 kb) encoding a 30.6-kDa protein are different from that for G_L (data not shown), and PPP1R3E 3′ UTR contains no significant sequence similarities to G_L 3′ UTR.

The fact that G_L appears to be present at much higher levels in human skeletal muscle than in mouse, rat, and rabbit skeletal muscle is unusual but not without precedent. Other proteins have been shown to exhibit species-specific differences in expression. For example, inhibitor-1, which is expressed in a wide variety of tissues, is found in rabbit, pig, sheep, and guinea pig liver, but is absent from mouse and rat liver, despite being present in other tissues in these rodents (40). The abundance of G_L in human skeletal muscle compared with that in rodents and rabbits indicates that the latter animals may not be appropriate models from which to gain an understanding of hormonal regulation of human muscle glycogen synthase phosphatase.

The high level of G_L protein in human skeletal muscle is also of considerable importance for understanding the mechanisms by which insulin regulates glycogen synthesis in this tissue. Increasing evidence indicates that insulin acts via the PI-3 kinase/PKB pathway to inhibit GSK3 and elicit an activation of glycogen synthase. However, the mechanisms by which insulin may mediate activation of glycogen-targeted PP1, thereby causing dephosphorylation and activation of glycogen synthase, are unclear. The most abundant glycogen-targeted form of PP1 in skeletal muscle is PP1G_M, but recent studies indicate that this glycogen-targeted form of PP1 does not appear to be regulated by insulin. Although Ser48 and Ser67 of rabbit G_M become phosphorylated in response to adrenaline, these residues are not phosphorylated significantly in response to insulin by MAPKAP kinase 1 or any other insulin-stimulated protein kinases (34). Overexpression of G_M/PPP1R3 in L6 myotubes resulted in increased basal and insulin-stimulated glycogen synthesis but did not increase the approximately twofold stimulation of glycogen synthesis by insulin, suggesting that G_M-PP1c may not be the major insulin-stimulated protein phosphatase in skeletal muscle (41). In G_M null mice, insulin stimulation of glycogen synthase in skeletal muscle does not differ from that in wild-type mice (36). In addition, in these G_M “knockout” mice, an insulin-stimulated glycogen synthase phosphatase, which was not thought to be the more widely distributed R5 and R6 complexes with PP1, was still detected. The studies we report here raise the possibility that G_L-PP1c may contribute to this insulin-stimulated phosphatase activity.

Previous studies have shown that the hepatic levels of G_L are reduced in streptozotocin-induced diabetes and restored by insulin treatment (24,25), but the low levels of G_L in rodent muscle make similar investigations difficult in this tissue. Modulation of G_L-PP1c interaction with the allosteric inhibitor phosphorylase a is more likely to underlie the short-term action of insulin, although the pathway by which insulin may decrease the levels of phosphorylase a in skeletal muscle is less clear. Unlike the studies with liver, where changes in cAMP levels in response to insulin are well documented, it is thought that in most situations changes in cAMP levels may not underlie the ability of insulin to activate glycogen synthase in skeletal muscle (1). Nevertheless, pretreatment of isolated rat diaphragm with insulin lowered the amount of cAMP produced in response to adrenaline (42,43). Another report shows that while insulin had no effect on basal cAMP release from perfused rat hind limb, it inhibited by 50% the elevated cAMP release produced in response to adrenaline (44). In adrenalectomized animals, isoproterenol, a β-adrenergic agonist, produced a twofold rise in inhibitor-1 phosphorylation, an effect that was completely prevented by the addition of insulin, and the antagonism of isoproterenol by insulin was shown to correlate with a decrease in the muscle content of cAMP (45). Overall, it appeared likely that for insulin to modulate cAMP levels, adrenaline concentrations should be >10 nmol/l in normal physiological range. These effects of insulin may be mediated by its ability to activate cAMP phosphodiesterase in skeletal muscle (43,46,47). It will be important to demonstrate whether insulin acting via this or another as yet unknown pathway can enhance the activity of the PP1c-G_L complex in human skeletal muscle.

The work was supported by funds from the U.K. Medical Research Council (to P.T.W.C). S.M. was the recipient of a Cooperative Awards in Science and Engineering postgraduate studentship from the Biotechnology and Biological Research Sciences Council, U.K., and Novo Nordisk, Bagsvaerd, Denmark.

We thank Chris Armstrong, Norman Pratt, and Mike Rennie for helpful advice.