Increased Insulin Sensitivity in Gsα Knockout Mice*

The stimulatory guanine nucleotide-binding protein (Gs) is required for hormone-stimulated cAMP generation. Gnas, the gene encoding the Gs α-subunit, is imprinted, and targeted disruption of this gene in mice leads to distinct phenotypes in heterozygotes depending on whether the maternal (m−/+) or paternal (+/p−) allele is mutated. Notably, m−/+ mice become obese, whereas +/p− mice are thinner than normal. In this study we show that despite these opposite changes in energy metabolism, both m−/+ and +/p− mice have greater sensitivity to insulin, with low to normal fasting glucose levels, low fasting insulin levels, improved glucose tolerance, and exaggerated hypoglycemic response to administered insulin. The combination of increased insulin sensitivity with obesity in m−/+ mice is unusual, because obesity is typically associated with insulin resistance. In skeletal muscles isolated from both m−/+ and +/p− mice, the basal rate of 2-deoxyglucose uptake was normal, whereas the rate of 2-deoxyglucose uptake in response to maximal insulin stimulation was significantly increased. The similar changes in muscle sensitivity to insulin in m−/+ and +/p− mice may reflect the fact that muscle Gsα expression is reduced by ∼50% in both groups of mice. GLUT4 expression is unaffected in muscles from +/p− mice. Increased responsiveness to insulin is therefore the result of altered insulin signaling and/or GLUT4 translocation. This is the first direct demonstration in a genetically altered in vivo model that Gs-coupled pathways negatively regulate insulin signaling.

Heterotrimeric guanine nucleotide-binding proteins (G proteins) 1 transduce signals from seven transmembrane receptors to intracellular effectors. Each G protein is composed of distinct ␣, ␤, and ␥ subunits (1). The ␣-subunit binds guanine nucleotide and interacts with specific effectors, such as adenylyl cyclase, phospholipase C, and ion channels. The ␣-subunit for G s (G s ␣) is ubiquitously expressed and transmits the stimulatory signal from hormone-bound receptors to adenylyl cyclase and is therefore required for hormone-stimulated cAMP generation. Both catecholamines (whose receptors activate G s ) and cAMP have been implicated as negative regulators of insulin signaling and insulin-stimulated glucose transport in various cell types, including adipocytes and muscle cells (2)(3)(4)(5)(6), although this has not been found universally (7)(8)(9). In one study admin-istration of cholera toxin (which constitutively activates G s ␣) to rats led to decreased insulin sensitivity and glucose uptake in skeletal muscle (10).
Heterozygous inactivating mutations of the gene encoding G s ␣ (GNAS1 at 20q13. Ref. 11) cause Albright hereditary osteodystrophy, an autosomal dominant disorder characterized by obesity, short stature, and skeletal defects (12). Paternal transmission of GNAS1 mutations produces offspring with Albright hereditary osteodystrophy alone (pseudopseudohypoparathyroidism), whereas maternal transmission produces offspring who also have multihormone resistance (termed pseudohypoparathyroidism type Ia) (13), suggesting that the G s ␣ gene is imprinted. Genomic imprinting is an epigenetic phenomenon characterized by parental allele-specific differences in gene expression (14 -16). In mice G s ␣ is expressed primarily from the maternal allele in some tissues, such as renal proximal tubules and adipose tissue, but is biallelically expressed in most other tissues (17,18). Although this could explain the presence of hormone resistance in pseudohypoparathyroidism type Ia and its absence in pseudopseudohypoparathyroidism, the imprinting of G s ␣ in humans has yet to be directly demonstrated (19).
GNAS1 (and the murine homolog Gnas) are now known to produce two additional gene products by use of alternative promoters and first exons, which splice onto a common second exon. XL␣s, a novel isoform of G s ␣, is expressed exclusively from the paternal allele, whereas NESP55, a chromograninlike neurosecretory protein, is expressed exclusively from the maternal allele (20 -22). Both are expressed primarily in neuroendocrine tissues, and their biological function has not been elucidated (23)(24)(25)(26)(27). XL␣s has been shown recently to be able to activate adenylyl cyclase in vitro but does not appear to be activated by G s -coupled receptors (28,29).
We generated mice with an insertional disruption of Gnas exon 2 (GsKO), an exon common to all known Gnas transcripts (17). Homozygotes (Ϫ/Ϫ) die during early postimplantation development. Heterozygotes with disruption of the paternal (ϩ/pϪ) or maternal (mϪ/ϩ) allele have distinct phenotypic manifestations (17), leading to early death in the majority of heterozygotes. mϪ/ϩ and ϩ/pϪ mice that survive past weaning have diametrically opposite abnormalities in fat mass, resting metabolic rate, and activity levels (18). Adult ϩ/pϪ mice are lean, hypermetabolic, and hyperactive, whereas mϪ/ϩ mice are obese and have lower resting metabolic rate and activity levels. These effects may be due to increased and decreased sympathetic activity in ϩ/pϪ and mϪ/ϩ mice, respectively. In this report we demonstrate that both lean ϩ/pϪ mice and obese mϪ/ϩ mice are more sensitive to insulin than normal mice, with increased insulin-stimulated glucose uptake in skeletal muscle. This is the first direct demonstration in a genetically altered in vivo model that G s -coupled pathways negatively regulate insulin signaling.

EXPERIMENTAL PROCEDURES
Mice-Mice with insertion of a neomycin resistance cassette into exon 2 of Gnas were previously created by targeted mutagenesis (17). Female mutants were mated to wild type CD-1 males (Charles River) to generate mϪ/ϩ mice, and male mutants were mated to wild type CD-1 females to generate ϩ/pϪ mice. Wild type littermates of mϪ/ϩ and ϩ/pϪ mice are designated mϩ/ϩ and ϩ/pϩ mice, respectively. Animals were maintained on a 12-h light/dark cycle (6 a.m./6 p.m.) and a standard pellet diet (NIH-07, 5% fat by weight).
Glucose and Insulin Tolerance Tests-For the measurement of fasting serum glucose and insulin levels, animals were sacrificed after an overnight fast (16 h), and serum was collected. Serum insulin was measured by radioimmunoassay (Linco Research Immunoassay, St. Charles, MO). Serum glucose was measured using the glucose HK kit (Sigma). For glucose tolerance tests, mice were fasted overnight and anesthetized with an intraperitoneal injection of pentobarbital (35 g/g of body weight) and ketamine (70 g/g of body weight). Glucose (2.5 mg/g of body weight in 0.9% NaCl solution) was administered intraperitoneally, and blood was collected from the retroorbital veins in heparinized capillary tubes at time 0 (prior to injection) and at 20, 40, 60, and 120 min after injection. Glucose was measured using the glucose HK kit (Sigma). For insulin tolerance tests, fasted mice were anesthetized with either pentobarbital and ketamine or pentobarbital alone. Human regular insulin (Humulin, 0.75 mIU/g of body weight in 0.9% NaCl solution) was administered intraperitoneally, blood was collected at 0, 15, 30, 60, 90, and 120 min, and glucose was measured using either the glucose HK kit or the Lifescan Surestep glucometer (Johnson & Johnson). Anesthesia had no significant effect on base-line glucose levels or response to insulin.
Measurement of 2-Deoxyglucose Uptake in Isolated Skeletal Muscle-In vitro glucose uptake into isolated extensor digitorum longus muscle was measured using [ 3 H]2-deoxyglucose, with [ 14 C]mannitol to correct for extracellular fluid (30). Briefly, muscles were isolated, allowed to recover for 3 h in Krebs-Heinseleit buffer containing glucose (6 mM), mannitol (18 mM), and pyruvate (2 mM) in a 95% O 2 , 5% CO 2 atmosphere, and then incubated in the presence or absence of 10 milliunits/ml insulin (Humulin, Eli Lilly) for 20 min, followed by addition of 100 M (2.4 Ci) [1,2-3 H]2-deoxyglucose and 0.7 Ci of [1-14 C]mannitol. After 20 min at 30°C, the tissue 3 H and 14 C were measured, and the tissue 3 H uptake into fibers was determined after adjustment for 14 C uptake into extracellular spaces.

Measurement of Total GLUT4 Expression in Skeletal
Muscle-Anterior tibialis muscles were dissected and immediately frozen in liquid nitrogen. Muscles were weighed, homogenized on ice with a Teflon pestle at 1:10 (w/v) in HEPES-EDTA-sucrose buffer (20 mM HEPES, 1 mM EDTA, 250 mM sucrose, pH 7.4, containing protease inhibitors (Complete capsules, Roche Molecular Biochemicals)), and solubilized in 2.3 M urea, 1.43% sodium dodecyl sulfate, 14.3 mM Tris, pH 6.8, and 100 mM dithiothreitol (all final concentrations). Samples were subjected to 10% SDS-polyacrylamide gel electrophoresis (Novex). After confirming equal loading of lanes by staining the gel with Ponceau S, proteins were transferred to nitrocellulose membranes by electroblotting. Filters were blocked, incubated with rabbit polyclonal antibody raised to the carboxyl terminus of GLUT4 (1:500 dilution), and then incubated with 0.25 Ci/ml [ 125 I]protein A (PerkinElmer Life Sciences) as previously described (31). Bands were quantified using a phosphorimaging device (Fuji BAS1000).
Statistical Analysis-All values are reported as the mean Ϯ S.E. Statistical significance was determined using the paired t test or multifactor ANOVA, with differences considered significant at p Ͻ 0.05 (2-tailed).

RESULTS AND DISCUSSION
To determine whether insulin sensitivity is altered in adult ϩ/pϪ and/or mϪ/ϩ mice, we measured glucose and insulin in fasted adult mice and glucose after intraperitoneal administration of glucose or insulin. Both ϩ/pϪ and mϪ/ϩ mice had significantly lower fasting serum insulin levels than their wild type littermates (Table I). For ϩ/pϪ mice five of six animals had fasting insulin levels below the detection limit of the assay (0.05 ng/ml). Simultaneous glucose levels were less than or equal to those of wild type littermates (Table I).
Glucose tolerance tests were next performed in 10-week-old male mice. Glucose levels were lower in both ϩ/pϪ and mϪ/ϩ mice compared with wild type littermates at all time points after injection of glucose (2.5 mg/g of body weight), with the genotype having a significant effect on the overall handling of glucose load (Fig. 1A). An overall decrease in the glucose curve was observed in both groups of mutant mice, although the return toward base-line glucose levels was more pronounced in ϩ/pϪ mice. This difference in glucose tolerance between ϩ/pϪ and mϪ/ϩ mice might be predicted, given that ϩ/pϪ mice are lean, whereas mϪ/ϩ mice are obese. We next examined the glucose response to administered insulin (0.75 mIU/g of body weight, intraperitoneal) (Fig. 1B). In this study base-line glucose was similar in all groups. In both ϩ/pϪ and mϪ/ϩ mice the hypoglycemic response to injected insulin was significantly greater than in wild type littermates.
Therefore, both ϩ/pϪ and mϪ/ϩ mice have increased insulin sensitivity, based on the findings of lower fasting insulin levels with normal or low fasting glucose, improved glucose tolerance, and a greater than normal acute hypoglycemic response to insulin. Increased insulin sensitivity might be expected in the lean ϩ/pϪ mice, because insulin sensitivity is generally inversely related to adiposity. However, in mϪ/ϩ mice increased insulin sensitivity is associated with obesity. The dissociation of obesity and insulin resistance has been described previously in a few other animal models (e.g. mice with a knockout of the adipocyte-specific aP2 gene; Ref. 32). The fact that mϪ/ϩ are obese whereas ϩ/pϪ are lean may explain why glucose tolerance and insulin sensitivity appear to be somewhat greater in ϩ/pϪ mice as compared with mϪ/ϩ mice (Fig. 1).
Most of the acute hypoglycemic response to insulin is normally due to increased glucose uptake in skeletal muscle, associated with translocation of the insulin-responsive glucose transporter GLUT4 from internal vesicles to the plasma membrane (33). Skeletal muscle (extensor digitorum longus) was removed from ϩ/pϪ and mϪ/ϩ mice and their wild type littermates, and the rates of 2-deoxyglucose uptake were measured in the presence of either no insulin (Fig. 2, Basal) or a maximal concentration of insulin (10 milliunits/ml; Fig. 2, Insulin). In the absence of insulin, rates of 2-deoxyglucose uptake were similar in muscles from ϩ/pϪ, mϪ/ϩ, and wild type mice. In contrast, in the presence of insulin the rates of 2-deoxyglucose uptake were significantly higher in muscles from both ϩ/pϪ and mϪ/ϩ mice when compared with those from wild type mice. The maximal responses to insulin (increase over basal) were 55 and 48% greater in muscles from ϩ/pϪ and mϪ/ϩ mice, respectively, than the response in muscles from wild type mice. These large differences in insulin-stimulated glucose up- b Five of six determinations were less than the detection limit of the assay and were assigned a value of 0.05 for determination of statistics.
take in skeletal muscle largely account for the greater hypoglycemic effect of insulin in ϩ/pϪ and mϪ/ϩ mice. One potential mechanism for the observed increase in insulin-stimulated glucose uptake in skeletal muscle is increased expression of GLUT4 in this tissue. cAMP has been shown to negatively regulate the GLUT4 promoter (34,35), and therefore decreased expression of G s ␣ might be expected to lead to GLUT4 overexpression due to decreased intracellular cAMP concentrations. However, immunoblotting of whole homoge-nates of skeletal muscle (anterior tibialis) with a GLUT4-specific antibody showed that GLUT4 expression was unaffected in skeletal muscles of ϩ/pϪ mice (99 Ϯ 15% of the level in paired wild type littermates; n ϭ 4 pairs; see Fig. 3).
Decreased levels of G s ␣ expression and/or intracellular cAMP in skeletal muscle of mϪ/ϩ and ϩ/pϪ mice probably lead to increased insulin-stimulated glucose uptake either by increasing the responsiveness of the proximal insulin-signaling pathway to insulin or by increasing GLUT4 translocation or transporter activity. Several studies show that catecholamines or cAMP analogues can lead to decreased insulin binding or decreased insulin receptor kinase activity in response to insulin in cultured cells (3, 36 -47), although some studies failed to show an effect of these agents on insulin receptor function (48 -50). One potential mechanism for these effects is through serine/threonine phosphorylation of the insulin receptor by cAMP-dependent protein kinase (protein kinase A) (40,46,51). Serine/threonine phosphorylation of downstream components, such as IRS-1 and -2, may decrease the ability of these molecules to be activated by the insulin receptor (52). We are presently embarking on studies to directly examine the ability of insulin to activate the insulin receptor and other proximal insulin-signaling components in tissues from Gnas knockout mice. There is also evidence that catecholamines or cAMP may have a direct effect on GLUT4 transporter, perhaps through decreasing the accessibility of GLUT 4 to the cell surface (4,53,54). One recent study suggests that the rate of fusion of GLUT4-containing vesicles to the plasma membrane in the presence of insulin is reduced by catecholamines (4).
Although in many respects ϩ/pϪ and mϪ/ϩ have distinct phenotypes, both groups of mice seem to have similar changes in insulin sensitivity, despite the fact that one group is leaner than normal and the other is obese. We think this is most likely due to the fact that G s ␣ is not imprinted in skeletal muscle, and FIG. 1. Glucose and insulin tolerance tests in ؉/p؊ and m؊/؉ mice and their paired wild type littermates. A, after an overnight fast, glucose (2.5 mg/g intraperitoneal) was administered to 10-week-old male mice, and plasma glucose was measured at base line and 20, 40, 60, and 120 min after injection. After glucose administration both ϩ/pϪ and mϪ/ϩ mice had significantly decreased plasma glucose levels compared with paired control littermates (p Ͻ 0.05 by multifactor ANOVA; n ϭ 6 in each group). B, human regular insulin (0.75 mIU/g intraperitoneal) was administered to fasted 10-week-old male mice, and plasma glucose was measured at base line and 15, 30, 60, 90, and 120 min after injection. The results are expressed as the percent of base-line glucose at each time point, with the number of animals per group analyzed at each time point shown in parentheses. Both ϩ/pϪ and mϪ/ϩ mice had a greater reduction in plasma glucose levels than their control littermates in response to insulin (p Ͻ 0.05 by multifactor ANOVA). Base-line glucose was not significantly different between mutant and wild type mice (ϩ/pϩ, 165 Ϯ 8 mg/dl versus ϩ/pϪ, 147 Ϯ 9 mg/dl; mϩ/ϩ, 176 Ϯ 5 mg/dl versus mϪ/ϩ, 187 Ϯ 10 mg/dl; n ϭ 10 in each group). In both panels results for ϩ/pϪ mice (E) and their paired wild type (ϩ/pϩ) littermates (f) are shown on the left, and results for mϪ/ϩ mice (E) and their paired mϩ/ϩ littermates (f) are shown on the right.

FIG. 2.
Basal and insulin-stimulated 2-deoxyglucose uptake in skeletal muscle isolated from ؉/p؊ and m؊/؉ mice and their wild type littermates. The rates of 2-deoxyglucose uptake by extensor digitorum longus muscles isolated from fasted 3-month-old male mϪ/ϩ and ϩ/pϪ mice (open bars) and their wild type littermates (filled bars) were measured as described under "Experimental Procedures." For each animal one muscle was studied in the absence of insulin (Basal, left), and the muscle from the opposite side was studied in the presence of 10 milliunits/ml insulin (Insulin, right). Results are shown as the mean Ϯ S.E. for each group. Asterisks indicate that in the presence of insulin 2-deoxyglucose uptake was significantly greater in muscles from mϪ/ϩ and ϩ/pϪ mice than from wild type mice (p Ͻ 0.05 by ANOVA; n ϭ 10 for wild type mice, 6 for mϪ/ϩ mice, and 5 for ϩ/pϪ mice). therefore its expression in skeletal muscle is decreased to a similar extent in ϩ/pϪ and mϪ/ϩ mice (18). Gnas also produces two other imprinted gene products, NESP55, a chromogranin-like protein that is expressed only from the maternal allele, and XL␣s, an alternative isoform of G s ␣ with a long amino-terminal extension that is only expressed from the paternal allele (20,21). It would seem unlikely that loss of these products leads to the observed increase in insulin sensitivity, because XL␣s expression is lost only in ϩ/pϪ mice, whereas NESP55 appears to be expressed in both mϪ/ϩ and ϩ/pϪ mice. 2 Moreover, we could not detect expression of either of these gene products in skeletal muscle of normal mice by reverse transcription-polymerase chain reaction (data not shown).
Hormones that counterregulate insulin, including catecholamines and glucagon, activate G s -coupled pathways. Therefore G s ␣ deficiency might be expected to produce an apparent increase in insulin sensitivity due to resistance to counterregulatory hormones. However, in patients with pseudohypoparathyroidism type Ia the glycemic response to maximal glucagon stimulation is normal (55,56). In our experiments the time course of the response to administered insulin in ϩ/pϪ and mϪ/ϩ mice is consistent with a direct effect on skeletal muscle glucose utilization independent of the effects of counterregulatory hormones. This was confirmed in our studies that directly examined insulin-stimulated glucose uptake in skeletal muscle in the absence of counterregulatory hormones.
Although our studies suggest that G s ␣ may negatively regulate insulin-stimulated glucose uptake, evidence suggests that two other heterotrimeric G proteins, namely G q/11 and G i2 , have the opposite effect on the metabolic response to insulin. Two studies in 3T3-L1 adipocytes demonstrated that G q/11 ␣ is required for insulin-stimulated GLUT4 translocation (57,58). One study suggested that G q/11 ␣ might work upstream of phosphoinositide 3Ј-OH kinase and be tyrosine phosphorylated by the insulin receptor (57), whereas the other study suggested that G q/11 ␣ works downstream of phosphoinositide 3Ј-OH kinase (58). Mice in which G i2 ␣ expression was decreased in liver and adipose tissue using an antisense RNA were insulin-resistant (59), whereas those expressing activated G i2 ␣ had increased insulin sensitivity (60 -62). G i2 ␣ appears to have biological effects opposite those of G s ␣ in several systems (63,64). Further studies are required to determine whether G s ␣ inhibits insulin signaling directly or through the actions of cAMP and/or cAMP-dependent protein kinase and what steps in the insulin receptor-GLUT4 pathway are affected by decreased G s ␣ expression.
FIG. 3. GLUT4 expression in skeletal muscle from ؉/p؊ mice and their normal littermates. Shown is an immunoblot of total homogenates (100 g of protein/lane) of skeletal muscle (anterior tibialis) removed from 3-month-old male ϩ/pϪ mice and paired wild type (ϩ/pϩ) littermates that was probed with a GLUT4-specific antibody. Individual sibling pairs are designated with brackets. Positions of molecular mass markers (in kDa) are designated on the left. The last lane (C) contains low density microsomes (20 g of protein) derived from rat adipocytes treated with insulin. Total GLUT4 expression in ϩ/pϪ mice was similar to that in paired ϩ/pϩ littermate controls (99 Ϯ 15% of ϩ/pϩ; n ϭ 4 pairs).