Tissue-Specific Effects of Rosiglitazone and Exercise in the Treatment of Lipid-Induced Insulin Resistance

Male Sprague-Dawley rats (∼4 weeks of age) were given ad libitum access to a high-fat diet (58% kcal fat; Research Diets, New Brunswick, NJ) for 4 weeks to induce insulin resistance. During the subsequent 4-week experimental period, rats continued to eat a high-fat diet and were randomly allocated to one of the following groups (n = 24/group): high-fat control, high-fat exercise training, high-fat rosiglitazone treatment, or a combination of high-fat exercise training and high-fat rosiglitazone treatment. Exercise training consisted of treadmill running for 1 h/day, 5 days/week on a 15% incline, at a speed that was gradually increased during the 1st week of training to 32 m/min. High-fat rosiglitazone-treated rats received a diet containing 50 ppm rosiglitazone (GlaxoSmith Kline, Stevenage, U.K.), which they consumed ad libitum at an average dose of 2.08 ± 0.06 mg · kg⁻¹ · day⁻¹. A fifth group of rats (chow-fed controls; n = 24) was fed a normal chow diet (17% kcal fat; Research Diets) for the duration of the study (8 weeks) and acted as a control for high-fat controls. After the experimental period, animals were fasted for 8–12 h before undergoing hindlimb perfusion for the measurement of basal 3-O-methylglucose transport (n = 8/treatment group), insulin-stimulated 3-O-methylglucose transport (n = 8/treatment group), or [¹⁴C]palmitate uptake and oxidation (n = 8/treatment group). Exercise training animals undertook their last training bout 36–48 h before hindlimb perfusion.

Rats were anesthetized and surgically prepared for hindlimb perfusion as previously described (17). The basic perfusate medium contained 30% washed time-expired human erythrocytes (Ogden Medical Center, Ogden, UT) and was continuously gassed with a mixture of 95% O₂-5% CO₂ and warmed to 37°C.

Insulin-stimulated perfusions were performed in the presence of 500 μU/ml insulin, whereas basal perfusate contained no insulin. Glucose transport was measured over an 8-min period using a perfusate that contained 8 mmol/l nonmetabolizable glucose analog 3-O-methylglucose (32 μCi 3-[³H]methylglucose · mmol⁻¹ · l⁻¹; PerkinElmer Life Sciences, Boston, MA) and 2 mmol/l mannitol (60 μCi [1-¹⁴C]mannitol · mmol⁻¹ · l⁻¹; PerkinElmer Life Sciences) as an extracellular space marker. Rates of basal and insulin-stimulated skeletal muscle 3-O-methylglucose transport (17) and insulin receptor substrate 1 (IRS1)–associated phosphatidylinositol 3-kinase (PI 3-kinase) activity (18) were calculated as previously described.

Rats were fasted for 8–12 h before being anesthetized and having their right hindlimbs surgically prepared for perfusion. Before cannulation, unperfused tissue from the left hindlimb was excised and stored for later analysis of AMPK activity and phosphorylation. Skeletal muscle fatty acid metabolism was assessed as previously described (19) by perfusing hindlimbs (7.5 ml/min) with a recirculated perfusate containing 4% FA-free BSA, 500 μmol/l albumin-bound palmitate, and 5 μCi albumin-bound [1-¹⁴C]palmitate (PerkinElmer) for 60 min. Arterial and venous samples for the analysis of ¹⁴CO₂ were taken immediately after a 20-min equilibration period and at 60 min. The liberation and collection of ¹⁴CO₂ from perfusate samples was performed by injecting 2 ml anaerobically collected perfusate into a sealed flask containing an equal volume of 1 mol/l acetic acid. The released ¹⁴CO₂ was trapped by an insert containing a strip of filter paper saturated with 500 μl benzothonium hydroxide and quantified using liquid scintillation counting.

AMPK activity, pT172 phosphorylation, and total protein content were analyzed as previously described (20). Activities were calculated as picomoles of phosphate incorporated into the SAMS peptide per minute, per milligram of protein subjected to immunoprecipitation. Acetyl CoA carboxylase (ACC) was affinity purified and quantified using post-AMPK immunoprecipitation supernatants as previously described (20).

Red quadriceps muscle lysates (60 μg), liver lysates (60 μg), or adipose tissue lysates (30 μg) were solubilized in Laemmli buffer, separated by SDS-PAGE, and transferred to polyvinylidine fluoride membranes. The membranes were blocked and incubated overnight at 4°C with antibodies specific for phospho-Akt Ser473, Akt, Akt1, Akt2 (1:1,000; 4058, 9272, 2967, and 2962; Cell Signaling), AS160 (21) (a gift from David James, Garvan Institute, Sydney), PPARγ coactivator 1 (PGC-1) (AB3242; Chemicon International), PEPCK (10004943; Cayman Chemical), or GLUT4 (46701-704; Biogenesis). The immunoreactive proteins were detected with enhanced chemiluminescence (Amersham Biosciences) and quantified by densitometry. FAT/CD36 and FABPpm content were determined as described previously (22).

Skeletal muscle triacylglycerol (TAG), diacylglycerol (DAG), and ceramide contents were determined as outlined previously (12).

Serum glucose was determined using the MediSense2 Blood Glucose Testing system (MediSense Australia, Melbourne, Australia). Serum insulin and adiponectin levels were determined using enzyme immunoassay kits (Sensitive Rat Insulin RIA kit [Linco Research]; Mouse/Rat Adiponectin ELISA kit [Otsuka Pharmaceutical]). Serum free fatty acids (FFAs) were determined using an enzymatic colorimetric method (NEFA C; Wako Chemicals).

Results are presented as mean ± SE. Differences arising from dietary interventions were determined using an unpaired t test (chow-fed vs. high-fat controls), and P values were adjusted using a Bonferroni correction to account for multiple comparisons. Differences between high-fat treatment groups were determined using a one-way ANOVA with a Student-Newman-Keuls post hoc test (SigmaStat version 3.1.1.).

Body mass was increased by high-fat feeding and rosiglitazone treatment but was decreased by exercise training (P < 0.05, Table 1). Changes in body mass were mirrored by changes in epididymal fat pad mass (Table 1). Serum adiponectin concentration was elevated in high-fat rosiglitazone (P = 0.03 vs. high-fat controls) and decreased in high-fat exercise training (P = 0.04 vs. high-fat controls). Serum FFA concentration was decreased in high-fat exercise training and in high-fat rosiglitazone treatment and exercise training (P < 0.02 vs. high-fat control; Table 1).

Insulin-stimulated glucose transport in the red quadriceps muscle of high-fat controls (high-fat controls) was reduced by ∼37% (P = 0.01; Fig. 1A) compared with chow-fed controls (chow-fed controls) and was impaired a further ∼32% (P = 0.05) in high-fat rosiglitazone. Exercise training increased insulin-stimulated glucose transport in fat-fed and rosiglitazone-treated animals (P < 0.05). Insulin-stimulated IRS1-associated PI 3-kinase activity in high-fat controls was reduced by 39% compared with chow-fed controls (P = 0.05; Fig. 1B). Exercise training increased insulin-stimulated PI 3-kinase activity in fat-fed (P = 0.01) but not in rosiglitazone-treated animals (high-fat exercise training vs. high-fat rosiglitazone treatment and exercise training, P < 0.01). Basal muscle glycogen content (Fig. 1C) was increased in high-fat exercise training and high-fat rosiglitazone treatment and exercise training (P < 0.05 vs. high-fat controls) and decreased with high-fat rosiglitazone (P < 0.05 vs. high-fat controls).

Muscle TAG (Fig. 2A) and DAG (Fig. 2B) content increased 20 and 13%, respectively, in high-fat rosiglitazone (P < 0.05 vs. high-fat controls) but were normalized by exercise training (high-fat rosiglitazone vs. high-fat rosiglitazone treatment and exercise training; P < 0.05). Muscle ceramide content (Fig. 2C) increased ∼23% with high-fat feeding (chow-fed vs. high-fat controls, P = 0.02) but was unaffected by either exercise training or rosiglitazone treatment. Increased muscle lipid storage was associated with increased [¹⁴C]palmitate uptake in the perfused hindlimb (Fig. 2D). Palmitate uptake tended to be ∼40% higher in high-fat controls compared with chow-fed controls (P = 0.07) and increased a further 40% in high-fat rosiglitazone (P = 0.03). Rosiglitazone-induced increases in palmitate uptake were associated with a ∼50% increases in skeletal muscle FAT/CD36 protein content in high-fat rosiglitazone and in high-fat rosiglitazone treatment and exercise training (P < 0.05 vs. high-fat control; Fig. 2F). No changes in the protein content of FABPpm were observed (data not shown). Although [¹⁴C]palmitate uptake was not increased in high-fat exercise training (Fig. 2D), a twofold increase in ¹⁴CO₂ production was observed (P = 0.03 vs. high-fat control; Fig. 2E), suggesting that increased lipid oxidation may be a mechanism responsible for normalized lipid levels in high-fat rosiglitazone treatment and exercise training. Increased lipid oxidation with exercise training was associated with ∼30% increase in the expression of PGC-1 (Fig. 3D) in high-fat exercise training (P = 0.03 vs. high-fat controls) and in high-fat rosiglitazone treatment and exercise training (P = 0.07 vs. high-fat controls).

AMPKα1 activity was unchanged by high-fat feeding but increased ∼30% in high-fat exercise training (P = 0.01 vs. high-fat controls) and ∼37% in high-fat rosiglitazone treatment and exercise training (P = 0.006 vs. high-fat controls; Fig. 3A). AMPKα2 activity was increased ∼36% by high-fat feeding (P = 0.05, high-fat vs. chow-fed controls) but was not further altered by exercise training or rosiglitazone (Fig. 3B). Changes in both AMPKα1 and -α2 activities were associated with similar changes in the total protein levels of these isoforms (P < 0.05, Fig. 3C). Phosphorylation of AMPKα1 at T172 was similar between high-fat controls and high-fat rosiglitazone but tended to increase in high-fat exercise training and in high-fat rosiglitazone treatment and exercise training (P = 0.1 vs. high-fat controls). High-fat–induced increases in activity and protein levels in the AMPKα2 isoform were associated with increased phosphorylation of the enzyme at T172 (chow-fed vs. high-fat controls, P = 0.01). We also examined the protein levels and phosphorylation status of the AMPK substrate, ACC. Although no changes in the total protein level of ACC were observed (Fig. 3C) there was a significant increase in the level of phosphorylation of the enzyme at S218 with high-fat feeding (chow-fed vs. high-fat controls, P = 0.04).

Compared with chow-fed controls, total Akt protein was reduced ∼25% by high-fat feeding (high-fat control; P = 0.02) and was restored by exercise training (high-fat controls vs. high-fat exercise training; P = 0.007; Fig. 4A). The increase in total Akt was entirely attributable to an increase in the Akt1 isoform (Fig. 4C), which was elevated ∼50% by exercise training (high-fat controls vs. high-fat exercise training; P = 0.04). Rosiglitazone did not change Akt1 protein content and appeared to blunt the exercise training–induced increase in Akt1 (high-fat exercise training vs. high-fat rosiglitazone treatment and exercise training; P = 0.04). No changes in the Akt2 isoform (Fig. 4D) were observed after either treatment. Phosphorylation of total Akt at the Ser473 residue (Fig. 4B) was also altered by exercise training, with a 50% decrease in high-fat exercise training (P = 0.01 vs. high-fat controls). AS160 protein was elevated ∼15% in high-fat (chow-fed vs. high-fat controls; P = 0.004) and was normalized in high-fat exercise training (P = 0.04) and in high-fat rosiglitazone treatment and exercise training (P = 0.03; Fig. 4E). GLUT4 protein content was decreased ∼20% in high-fat controls (P = 0.007) but was increased ∼50 and ∼40% in high-fat exercise training and in high-fat rosiglitazone treatment and exercise training, respectively (P < 0.001; Fig. 4F).

GLUT4 content in adipose tissue was reduced by ∼20% in high-fat controls compared with chow-fed controls (P = 0.01) and tended to be impaired a further ∼20% by exercise training (high-fat exercise training, P = 0.07; Fig. 5A). Rosiglitazone increased adipose tissue GLUT4 content in both fat-fed (high-fat controls vs. high-fat rosiglitazone, P = 0.04) and exercise-trained (high-fat exercise training vs. high-fat rosiglitazone treatment and exercise training, P = 0.007) animals. Adipose tissue PGC-1 protein content was decreased by exercise training (high-fat controls vs. high-fat exercise training, P < 0.05; Fig. 5B) and increased by rosiglitazone treatment (P = 0.05). PEPCK content in adipose tissue was increased by high-fat feeding (chow-fed vs. high-fat controls, P < 0.05; Fig. 5C) and was elevated further by rosiglitazone (P < 0.01). AMPKα1 activity and protein content were elevated in high-fat compared with lean controls (P < 0.05; Fig. 5D and E). Rosiglitazone treatment tended to decrease AMPKα1 activity (P = 0.1) and protein content (P < 0.03) while increasing the expression of ACC protein in adipose tissue (P < 0.05; Fig. 5F).

Liver TAG was increased 36% by high-fat feeding (P = 0.05), but was reduced in all other treatment groups (P < 0.001 vs. high-fat controls; Fig. 6A). Exercise training resulted in a small (∼13%) but significant increase in liver FAT/CD36 expression (P = 0.03; Fig. 6B), whereas rosiglitazone treatment decreased the expression of this fatty acid transporter (P < 0.02 vs. high-fat exercise training). Liver PGC-1 expression was increased (P = 0.01 vs. chow-fed controls; Fig. 6C) and PEPCK expression was decreased (P = 0.04 vs. chow-fed controls; Fig. 6D) by high-fat feeding, with little effect of exercise training or rosiglitazone.

It is well-established that chronic rosiglitazone treatment improves whole-body insulin sensitivity (23). Less clear, however, is whether this enhanced insulin action is directly attributable to the effects of rosiglitazone on skeletal muscle or other insulin-sensitive tissues. In the present study, we observed that exercise training, but not rosiglitazone treatment, reversed skeletal muscle insulin resistance associated with a high-fat diet. Contrary to our hypothesis, rosiglitazone treatment decreased insulin-stimulated glucose transport (P = 0.05, Fig. 1A) in the perfused hindlimb compared with high-fat controls.

Reduced skeletal muscle insulin sensitivity with rosiglitazone treatment seems a paradoxical finding given the ability of the drug to improve whole-body insulin sensitivity (23). Other investigations have demonstrated enhanced skeletal muscle and whole-body insulin sensitivity after short-term (4 days [7] or 2 weeks [24]) rosiglitazone treatment in high-fat–fed rats. We are not the first, however, to report a failure of TZDs to improve insulin resistance induced by high-fat feeding in rodents (25,26). Our observation of increased muscle FAT/CD36 expression and lipid storage provides a potential mechanism for an absence of rosiglitazone-induced muscle insulin sensitization. It is possible that the longer duration of high-fat feeding (8 weeks) and rosiglitazone treatment (4 weeks) used in the present study provided sufficient time to allow for the accumulation of muscle lipids, resulting from increased CD36 expression. A time-course experiment would be necessary to determine whether muscle insulin sensitization observed after shorter term (<2 weeks) TZD treatment is abolished as treatment duration increases in fat-fed animals.

The “lipid steal hypothesis” of TZD action postulates that by increasing fatty acid storage in adipose tissue, TZDs enhance insulin sensitivity by reducing the toxic accumulation of lipids in other insulin-sensitive tissues such as skeletal muscle and liver (5). However, there is mounting evidence that rosiglitazone acts to improve insulin sensitivity in the absence of decreased muscle lipid storage in humans (9,27) and animal models of insulin resistance (7,10,12,28), suggesting that lipid steal from muscle may not be necessary for rosiglitazone-induced insulin sensitization. Although some investigations have demonstrated that lipid storage is decreased in muscle after rosiglitazone treatment (3,4,24), rosiglitazone-induced reductions of liver lipid accumulation have been more consistently observed (5,7,9,29). In the present study, 4-week rosiglitazone treatment resulted in a 20% increase in muscle TAG content (Fig. 3B). As noted previously, we attribute this increase in lipid storage to an increase in lipid uptake (Fig. 1B) resulting from a rosiglitazone-induced upregulation of FAT/CD36 (Fig. 4A), a finding that is consistent with that of Wilmsen et al. (30), who found that 4-day rosiglitazone treatment of skeletal muscle cells taken from type 2 diabetic subjects resulted in a 2.5-fold increase in this fatty acid transporter. In the combined treatment group (high-fat rosiglitazone treatment and exercise training), exercise training prevented rosiglitazone-induced increases in TAG and DAG. Reduced lipid storage with exercise training was associated with increased palmitate oxidation and elevated PGC-1 expression in high-fat exercise training and in high-fat rosiglitazone treatment and exercise training (Fig. 3D). Thus, an increased capacity to oxidize fatty acid provides a potential mechanism by which exercise prevents the increase in muscle lipid storage observed in animals treated with rosiglitazone alone.

An alternative mechanism by which rosiglitazone may improve whole-body insulin sensitivity is by enhancing skeletal muscle glucose uptake via activation of AMPK in this tissue (13). In this regard, we have previously demonstrated that chronic rosiglitazone restores AMPK activation in the skeletal muscle of fa/fa Zucker rats, a model that displays reduced AMPK activation as a result of deficient leptin signaling (20). In contrast, high-fat feeding in the present study resulted in a ∼36% increase in AMPKα2 activity (Fig. 3B), which was not further augmented by rosiglitazone treatment. Thus, when compared with models lacking intact leptin signaling (i.e., ob/ob mouse, fa/fa rat) (20,31,32), the absence of AMPK deficiency in the present model of diet-induced insulin resistance provides a likely mechanism for the absence of a rosiglitazone-induced improvements in AMPK activation and muscle glucose transport.

Our in vivo finding of increased AMPKα2 activity with high-fat feeding is in agreement with previous work demonstrating that acutely increasing fatty acid availability stimulates AMPK activity, ACC phosphorylation, and palmitate oxidation in L6 myotubes (33). In contrast, AMPKα1 activity was not affected by diet but was increased by exercise training, which was associated with improved insulin-stimulated glucose transport and glycogen storage. These observations are consistent with other reports of upregulation of AMPKα1 after exercise training in humans with type 2 diabetes (34) and animal models of insulin resistance (35). Our results suggest distinct roles for the AMPK-α subunit isoforms in skeletal muscle, with AMPKα2 activity being responsive to increased fatty acid availability and AMPKα1 activity being linked to exercise training–induced changes in glycogen storage and insulin sensitivity.

In addition to increasing AMPKα1 activity, exercise training may also improve glucose uptake via enhancement of the classic insulin signaling pathway. In the present study, exercise training increased IRS1-associated PI 3-kinase activity and normalized impairments to total protein levels in the Akt/AS160/GLUT4 signaling pathway caused by high-fat feeding. To our knowledge, the present investigation is the first to demonstrate changes in AS160 expression due to chronic high-fat feeding and exercise training. Larance et al. (21) established that decreasing AS160 expression by short hairpin RNA leads to increased plasma membrane GLUT4 content. However, it is unclear what the physiological effects of altered AS160 expression are in vivo. We also observed an increase in the expression of the Akt1 isoform (Fig. 4C) after exercise training. Akt1 has recently been implicated in the regulation of fatty acid uptake in skeletal muscle using isoform-specific siRNA-based gene silencing (36); however, more investigation into the specific functions of Akt1 and AS160 in exercise training–induced adaptations is needed.

Although exercise did enhance PI 3-kinase activation, we observed a seemingly paradoxical decrease in Akt phosphorylation at Ser473 in exercise-trained animals (Fig. 4B). However, discrepancies between Akt activation and upstream insulin signaling have been reported in several different models of insulin resistance (37–39). In previous investigations, exercise training resulted in increased (40) or unchanged (41) insulin-stimulated Akt Ser473 phosphorylation, indicating that Akt activation is not necessary for exercise-induced insulin sensitization. Recent evidence suggests that AMPK activation results in the dephosphorylation of Akt at Ser473 (42). Decreased Akt Ser473 phosphorylation in the present investigation was associated with ∼30% increases in AMPKα1 activity with both exercise training and combined therapy (Fig. 3A), providing a potential mechanism for exercise-induced reductions in Akt activation.

In contrast to exercise, we saw no change in IRS1-associated PI 3-kinase activity (Fig. 1B) or the total protein levels of Akt, AS160, or GLUT4 (Fig. 4) after 4 weeks of rosiglitazone treatment. These findings are in agreement with those of Karlsson et al. (43) who observed no enhancement of the IRS1/PI 3-kinase/Akt/AS160 signaling pathway after 26 weeks of rosiglitazone treatment (8 mg/day) in the skeletal muscle of patients with newly diagnosed type 2 diabetes. However, despite no apparent improvement in muscle insulin signaling, insulin-stimulated muscle glucose uptake was increased in these patients during a hyperinsulinemic-euglycemic clamp (43). In contrast, others have observed improvements in insulin signaling in patients with type 2 diabetes after rosiglitazone treatment (44,45). Thus, it is not clear whether rosiglitazone-induced insulin sensitization in humans is attributable to improved insulin signaling in muscle. Our results suggest that impairments to muscle insulin signal transduction that result from high-fat feeding are reversed by exercise training, but not rosiglitazone treatment. This contention is supported by recent clinical evidence demonstrating that lipid-induced insulin resistance in obese individuals is not prevented by rosiglitazone treatment (46).

In contrast to its beneficial effects on glucose and lipid metabolism in skeletal muscle, exercise training appears to have an inhibitory role on these metabolic pathways in adipose tissue. Exercise training reduced adipose tissue GLUT4 (Fig. 5A), PGC-1 (Fig. 5B), and ACC (Fig. 5F) protein content. A combination of these adaptations would likely reduce glucose uptake and fat storage in adipose tissue, presumably functioning to spare glucose and lipid substrates for use in active skeletal muscle. Notably, the negative effects of exercise on key enzymes involved in glucose uptake and lipid storage in adipose were reversed by rosiglitazone treatment. In adipose tissue, rosiglitazone increased GLUT4 (Fig. 5A) and PGC-1 protein content (Fig. 5B), both of which are associated with insulin sensitivity in adipocytes (47). Rosiglitazone also increased the expression of key proteins involved in glyceroneogenesis, lipogenesis, and lipid storage in adipocytes, such as PEPCK (Fig. 5C) (48) and ACC (Fig. 5F). We also observed a rosiglitazone-induced decrease in adipose tissue AMPKα1 expression and a tendency (P = 0.1) for reduced AMPK activation. Given that AMPK activation (5-aminoimidazole-4-carboxamide ribonucleoside) with AICAR has been recently shown to inhibit glucose uptake and lipogenesis in adipocytes (49), our results are consistent with increased adipose tissue glucose uptake and lipid storage. In support of this contention, Virtanen et al. (50) demonstrated that enhanced glucose uptake into adipose tissue makes a significant contribution to rosiglitazone-induced insulin sensitization in patients with newly diagnosed type 2 diabetes. Our observations of increased adipose tissue mass combined with elevated adipose GLUT4 content and reduced AMPK activation provide a potential mechanism for increased adipose tissue glucose uptake after rosiglitazone treatment. Thus, it is possible that the increased adipose tissue mass commonly observed after TZD treatment (23) may provide a “glucose sink” that contributes to the ability of rosiglitazone to improve peripheral glucose disposal.

Our observation that lipid accumulation was reduced in the liver of rosiglitazone-treated animals (Fig. 6A) raises the possibility that this organ also acts as a potential target for the insulin-sensitizing actions of the drug. In support of this contention, Gastaldelli et al. (51) have recently reported that rosiglitazone treatment (8 mg/day for 12 weeks) decreased gluconeogenesis in patients with type 2 diabetes. Thus, it is probable that exercise-induced improvements in skeletal muscle insulin sensitivity, combined with the potential for rosiglitazone to improve adipose tissue glucose uptake, lipid storage, and hepatic insulin action, would produce additive improvements to whole-body insulin sensitivity as previously reported (16).

In conclusion, we provide novel data to demonstrate that exercise training, but not rosiglitazone treatment reverses skeletal muscle insulin resistance induced by lipid oversupply. Conversely, rosiglitazone, but not exercise training, increases the expression of proteins that regulate adipose tissue glucose transport and fatty acid storage. Although both interventions can improve whole-body insulin sensitivity, our results show that they produce divergent effects on protein expression and triglyceride storage in different tissues. Accordingly, exercise training and rosiglitazone may act as complementary therapies for the treatment of insulin resistance.

A.B. has received support from the Canadian Institutes of Health Research and is the Canada Research Chair in Metabolism and Health. M.A.F. is a National Health and Medical Research Council Principal Research Fellow. B.E.K. is an Australian Research Council Federation Fellow. B.B.Y. has received National Institutes of Health Grants GM-48680, GM-08395, and DK-57625. J.A.H. has received Australian Research Council Grant DP0663862.

We thank GlaxoSmithKline (Stevenage, United Kingdom) for the gift of rosiglitazone and Prof. David E. James for the gift of AS160 antibody.