The Effects of Rosiglitazone on Insulin Sensitivity, Lipolysis, and Hepatic and Skeletal Muscle Triglyceride Content in Patients With Type 2 Diabetes
We examined the effect of three months of rosiglitazone treatment (4 mg b.i.d.) on whole-body insulin sensitivity and in vivo peripheral adipocyte insulin sensitivity as assessed by glycerol release in microdialysis from subcutaneous fat during a two-step (20 and 120 mU · m−2 · min−1) hyperinsulinemic-euglycemic clamp in nine type 2 diabetic subjects. In addition, the effects of rosiglitazone on liver and muscle triglyceride content were assessed by 1H-nuclear magnetic resonance spectroscopy. Rosiglitazone treatment resulted in a 68% (P < 0.002) and a 20% (P < 0.016) improvement in insulin-stimulated glucose metabolism during the low- and high- dosage−insulin clamps, respectively, which was associated with ∼40% reductions in plasma fatty acid concentration (P < 0.05) and hepatic triglyceride content (P < 0.05). These changes were associated with a 39% increase in extramyocellular lipid content (P < 0.05) and a 52% increase in the sensitivity of peripheral adipocytes to the inhibitory effects of insulin on lipolysis (P = 0.04). In conclusion, these results support the hypothesis that thiazolidinediones enhance insulin sensitivity in patients with type 2 diabetes by promoting increased insulin sensitivity in peripheral adipocytes, which results in lower plasma fatty acid concentrations and a redistribution of intracellular lipid from insulin responsive organs into peripheral adipocytes.
Although thiazolidinediones (TZDs) are now widely used to treat type 2 diabetes, their mechanism of action remains largely unknown. It has been well established that TZDs enhance insulin sensitivity by acting as ligands for the transcription peroxisome proliferator−activated receptor-γ (PPAR-γ) (1,2). However, PPAR-γ is predominantly expressed in adipose tissue, whereas the improvement in insulin sensitivity occurs predominately in skeletal muscle (3–5) where PPAR-γ expression is relatively low (1,2). This paradox suggests that the TZDs may indeed modulate key communication signals between fat and muscle, such as leptin (6), adiponectin (7), tumor necrosis factor-α (TNF-α) (8), resistin (9), and fatty acids (10–13). In support of fatty acids being an important factor in this regard, recent studies have demonstrated a strong relationship between tissue lipid content and insulin resistance in both skeletal muscle (14–20) and liver (15,16,21).
In this study, we examined the hypothesis that TZDs improve insulin sensitivity in patients with type 2 diabetes by promoting the redistribution of triglyceride from liver and muscle to peripheral adipocytes. In that regard, we analyzed the effects of rosiglitazone, a TZD, in subjects with type 2 diabetes using several different complementary methodologies. 1H-nuclear magnetic resonance (NMR) spectroscopy was used to assess triglyceride content in liver and muscle in conjunction with the hyperinsulinemic-euglycemic clamp, and indirect calorimetry was used to assess insulin sensitivity. The effects of rosiglitazone on peripheral adipocyte lipolysis during the clamp were assessed by microdialysis measurements of glycerol release from subcutaneous fat.
RESEARCH DESIGN AND METHODS
The participants in this study were nine nonsmoking healthy subjects with type 2 diabetes. Their clinical and metabolic profiles are detailed in Table 1. Generally, this was a middle-aged, obese group of patients with well-controlled diabetes. At the time of enrollment, six subjects were taking metformin, a sulfonylurea, or both, and the remaining three subjects were managing their diabetes with diet alone. Informed written consent was obtained from all subjects after the aims and potential risks of the study were explained to them. The protocol was approved by the Yale University School of Medicine Human Investigation Committee.
After a medical screening, a wash-out period of 10–14 days was allowed for the discontinuation of any prior oral antidiabetic regimen. For the 3 days before the start of the trial, subjects consumed a weight-maintenance diet according to the guidelines of the American Diabetes Association (60% carbohydrate, 20% protein, 20% fat), evenly divided among three meals Subjects were then admitted to the Yale/New Haven Hospital General Clinical Research Center, where they underwent a series of metabolic investigations. After these baseline studies, subjects took rosiglitazone (4 mg b.i.d., p.o.) for 10–12 weeks. The subjects were then readmitted to the center and underwent the identical metabolic investigations conducted during the baseline evaluation. During the treatment period, subjects were evaluated every 2 weeks to ensure compliance and to assess them for adverse events, including liver dysfunction.
Insulin clamp procedure.
At 8:00 a.m., a 3-h primed (corrected for ambient fasting plasma glucose level), continuous infusion of [6,6-2H]glucose (2 mg · m−2 body surface area · min−1) and [1,1,2,3,3-2H5]glycerol (∼0.02 mg · m−2 body surface area · min−1) were begun through an antecubital vein. During the third hour of infusion, a retrograde cannula was inserted into a vein in the contralateral hand, which was warmed for the sampling of arterialized venous blood. A small volume of 0.9% NaCl flowed through the sampling cannula to maintain patency. Blood samples were withdrawn at 10-min intervals during the final 40 min of the third hour for the measurement of plasma glucose and insulin levels and glucose isotope enrichment. After 3 h of isotope infusion, a two-step hyperinsulinemic-euglycemic clamp was initiated, as previously described (3). During the first step, a low dosage of insulin was infused at 20 mU · m−2 · min−1 for 120 min to assess insulin’s effects on peripheral adipocyte lipolysis. During the second step, a high dosage of insulin was infused at 120 mU · m−2 · min−1 for 240 min to assess insulin’s effects on peripheral glucose disposal and suppression of endogenous glucose production.
During the final 30 min of the basal infusion and each step of the insulin clamp, indirect calorimetry was performed using the ventilated hood technique (Deltatrack Metabolic Monitor, Sensorimedics, Anaheim, CA) to measure basal fasting rates of glucose and lipid oxidation, as previously described (22).
Adipose tissue microdialysis.
The microdialysis probe (CMA/60; CMA, Microdialysis, Solna, Sweden) and details of the microdialysis experiments have been previously described (23,24). Briefly, a dialysis tubing (20-mm shaft, 30-mm membrane, 20 kDa) was attached to the end of double-lumen polyurethane tubing. The perfusate entered through the outer lumen, flowed to the microdialysis membrane, and was subsequently collected (dialysate) throughout the outer lumen. Once steady state was reached, the composition of the dialysate reflected that of the interstitial fluid (25). After an overnight fast, and before the baseline infusions were begun for the insulin clamp, two dialysis probes were inserted percutaneously into the subcutaneous tissue of the abdomen, 4–6 cm lateral to the umbilicus. To estimate regional adipose tissue blood flow, a microdialysis ethanol technique was used, as previously described and validated (26,27). Briefly, an ethanol-based perfusate (50 mmol/l) was infused through each probe using a microinjection pump (CMA/106 for a perfusate rate of 0.3 μl/min) and a Harvard pump (2 μl/min). After an initial 30-min wash-out period, dialysate samples were collected every 30 min for the entire 9-h period (3-h baseline period and 6-h period for the two-step clamp).
1H-NMR spectroscopy of hepatic and intramuscular triglyceride content.
At 7:00 a.m on the morning after the insulin clamp, and after a 12-h fast, subjects were transported by wheelchair to the Yale Magnetic Resonance Center. Localized 1H-NMR spectra of the soleus muscle were acquired on a 2.1T Biospec system (Bruker Instruments, Billerica, MA), using a coil assembly consisting of two circular hydrogen-1 coil loops (13-cm diameter each), arranged spatially to generate a quadrature field. During the measurements, the subject remained supine, and the gastrocnemius-soleus complex of the right leg was positioned within the homogeneous volume of the magnet. Scout images were acquired to position the volume of interest. Volumes of interest (15 mm × 15 mm × 25 mm) were centered within the soleus muscle and positioned to avoid vascular structures and gross adipose tissue deposits. Localized shimming in the soleus was performed using FASTERMAP (28), with typical line widths of ∼10 Hz obtained. Localized proton spectra were then collected using a PRESS sequence with the following parameters: repetition time (tR) = 3 s, echo time (tE) = 21.1 ms, 8,192 data points over 5,000 Hz spectral width, and 128 scans. Signals in the time domain were multiplied by a Gaussian function before Fourier transformation and manual phase correction. 1H resonances were assigned to water and methyl/methylene of triglycerides from their chemical shift, in agreement with Schick et al. (29), and were line fitted using the Mac-Nuts-PPC software package (Acorn NMR, Livermore, CA). Intramyocellular lipid (triglyceride) content (IMLC) and extramyocellular lipid content (EMLC) were calculated from the peak areas of intra- and extramyocellular CH2, respectively, with respect to the water peak area, and corrected for T1 and T2 relaxation effects. IMLC and EMLC content were then expressed as a percentage of water content.
Localized 1H-NMR spectra of the liver were obtained using a 12 × 14 cm butterfly 1H observation coil placed rigidly over the lateral aspect of the abdomen. Placement of the volume of interest (15 mm × 15 mm × 15 mm) within the liver was ensured by imaging the liver with a multi-slice gradient echo sequence. Before each measurement, the water signal was optimized using a shimming procedure, where lipid signal (predominantly from subcutaneous fat) was suppressed using an inversion recovery method optimized to null the subcutaneous fat signal. Localized proton spectra were collected using a modified PRESS sequence with the following parameters: tR = 3 s, tE = 24.1 ms, 8,192 data points over 5,000 Hz spectral width, and 64 scans. A saturation slice centered on the chest wall was applied to prevent contamination from subcutaneous adipose tissue. The spatial position of the saturation slice was determined for each subject from the scout image. A Lorentzian filter of 5 Hz was applied before Fourier transformation and manual phase correction. Hepatic triglyceride content was calculated from the area of intrahepatic CH2 resonance relative to the area of the water resonance, using the integration routine of Paravision software (Bruker), and then expressed as a percentage of water content.
Body composition analysis.
Dual-energy X-ray absorptiometry (DEXA) scanning (Hologic QDR-4500 W; Hologic, Bedford, MA) was performed with the subject in the supine position. Fat and lean body mass in each arm and leg, the trunk, and the head were calculated assuming 17% brain fat and lean body consisting of 73% water (30).
Plasma glucose concentrations were measured using the glucose oxidase method with a Beckman Glucose Analyzer II (Beckman Instruments, Fullerton, CA). Plasma immunoreactive insulin concentrations were measured with a double-antibody radioimmunoassay (Diagnostic System Labs, Webster, TX). Plasma fatty acid concentrations were determined using a microfluorimetric method (31). Gas chromatography−mass spectrometry (GCMS) analysis of enrichments of [6,6-2H]glucose and [1,1,2,3,3-2H5]glycerol in plasma and infusates were performed using the penta-acetate derivative of glucose and the triacetate derivative of glycerol, respectively (32). GCMS analysis was performed with a Hewlett-Packard 5971A Mass Selective Detector (Wilmington, DE) operating in the electron ionization mode. Ions with m/z 200, 201, and 202 were monitored for moles percent enrichment in glucose, and ions with m/z 145, 146, and 148 were monitored for moles percent enrichment in glycerol.
Basal and insulin-stimulated glucose and lipid oxidation rates were measured by the ventilated hood technique using a Deltatrack Metabolic Monitor (Sensorimedics, Anaheim, CA), as previously described (22).
Adipose tissue microdialysis.
Microdialysate glycerol was measured using an enzyme-linked colorimetric determination of 0.5-μl samples by a CMA 600 microdialysis analyzer (CMA 600 Microdialysis; N. Chelmsford, MA), which has an intra-assay coefficient of variation of 2.0% and an interassay coefficient of variation of 2.0%. Ethanol concentrations in the perfusate were determined by using the automated YSI analyzer (YSI, Yellow Springs, CA).
The rates of glucose infusion were calculated in 20-min blocks between 80 and 120 min during the low-dose insulin clamp and between 300 and 360 min of the high-dose insulin clamp. The data were corrected for urinary glucose and glucose space, averaged for each step of the clamp study, and are expressed as μmol glucose/m2-min. Clamped glucose disposal rate (GDRc) was calculated using the following formula: where GPc represents endogenous glucose production during the clamp and GIR represents the mean glucose infusion rate (mg/m2-min) during the clamp procedure (300 to 360 min).
Basal glucose production (GPb) was calculated using the following formula: where f = basal [6,6-2H]glucose infusion rate (mg/min), BSA = body surface area (m2), enrichmentinf = [6,6-2H]glucose infusate enrichment (%), and enrichmentplasma = steady-state basal plasma [6,6-2H]glucose enrichment (%). (The term “enrichment” refers to the ratio fo glucose isotope to naturally occurring (12C) glucose, expressed as a percentage.)
Clamped glucose production (GPc) during the second step (120 mU/m2-min) of the hyperinsulinemic-euglycemic clamp was calculated using the following formula: where GIR = mean glucose infusion rate (mg/m2-min) during the clamp procedure (300 to 360 min), enrichmentinf = exogenous [6,6-2H]glucose infusate enrichment (%), and enrichmentplasma = steady-state clamped plasma [6,6-2H]glucose enrichment (%).
The nonprotein respiratory quotients for 100% oxidation of fat and for oxidation of carbohydrates were 0.707 and 1.00, respectively (22). Nonoxidative glucose metabolism was calculated by subtracting the amount of glucose oxidized from the total amount of glucose infused.
Glycerol turnover (Ra[Glycerol]) at baseline and during the two steps of the insulin clamp was calculated using the following formula: where f = basal [2H5]glycerol infusion rate (mg/min), BSA = body surface area (m2), enrichmentinf = [2H5]glycerol infusate enrichment (%), and enrichmentplasma = steady-state basal plasma [2H5]glycerol enrichment (%).
The percent glycerol suppression from adipocytes was calculated using the following formula: where Glyb represents the serum glycerol concentration at baseline and Gly120′ represents that at 120 min of the clamp.
Adipose blood flow.
The change in local adipose blood flow during the hyperinsulinemic-euglycemic clamp was estimated using ethanol (ETOH) recovery rates, which are expressed as a percentage using the following formula: where ETOHin represents the ethanol concentration measured in the perfusate, whereas ETOHout represents the ethanol concentration in the dialysate.
All statistical analyses were performed using Statview software (Abacus Concepts, Berkely, CA). All data are presented as means ± SEM. Pre- to posttherapy values are compared using a paired t test, with significance reached at P < 0.05.
Table 2 shows the changes in pertinent fasting laboratory studies before and after therapy. After 3 months of rosiglitazone treatment, mean fasting plasma glucose concentrations decreased by 12.2 ± 3.7% (P = 0.02), mean insulin concentrations decreased by 23.7 ± 6.9% (P = 0.03), and mean fasting plasma fatty acid concentrations decreased by 38.8 ± 4.6% (P = 0.003). Although there were no significant changes in serum hepatic transaminase levels, both alkaline phosphatase (−15.7 ± 3.8%; P = 0.007) and γ-glutamyl-transpeptidase (−38.8 ± 4.6%; P = 0.001) were significantly reduced by rosiglitazone therapy. No appreciable change in HbA1c was detected, given that the baseline value reflected the degree of glycemic control achieved with the subjects’ prior antidiabetic regimen, whereas the disproportionately higher fasting plasma glucose concentration likely reflected an acute deterioration in glycemic control during the washout period.
Insulin levels were similar during both steps of the insulin clamp before and after treatment (Fig. 1). During the low-dosage clamps, the mean GIR required to maintain euglycemia increased by 68%, from 56.2 ± 4.7 to 94.2 ± 5.6 mg · m−2 · min−1 before and after rosiglitazone, respectively (P < 0.0001). During the high-dosage clamps, the mean GIR required to maintain euglycemia increased by 20%, from 296.0 ± 21.5 to 356.0 ± 24.7 mg · m−2 · min−1 before and after rosiglitazone, respectively (P = 0.002). The corresponding increases in the mean GDR were 119% during the low-dosage clamps (34.8 ± 9.2 vs. 76.2 ± 9.9 mg · m−2 · min−1; P = 0.002) and 18% during the high-dosage clamps (340.2 ± 57.0 vs. 402.2 ± 37.1 mg · m−2 · min−1; P = 0.10). Rosiglitazone had no effect on basal glucose production (101.8 vs. 93.2 mg · m−2 · min−1; P = 0.12) or on the ability of insulin to suppress glucose production during the second step of the hyperinsulinemic-euglycemic clamp that was maximally suppressed in both studies.
Glucose and lipid oxidation.
Basal and insulin-stimulated rates of glucose and lipid oxidation were unchanged by rosiglitazone therapy (Table 2).
Whole-body glycerol turnover.
Whole-body glycerol turnover was similar before and after treatment at baseline (7.6 ± 1.2 vs. 7.6 ± 1.0 mg · m−2 · min−1, respectively; P = 0.99), during the low-dosage clamp (3.2 ± 0.7 [before] vs. 2.8 ± 0.5 mg · m−2 · min−1 [after], respectively; P = 0.34), and during the high-dosage clamp (2.6 ± 0.4 [before] vs. 3.0 ± 0.5 mg · m−2 · min−1 [after], respectively; P = 0.34).
Adipocyte response to insulin.
Rosiglitazone therapy led to a 52% increase in the percent suppression of glycerol release from adipocytes by insulin during the low-dosage insulin clamp (28.5 ± 6.3 vs. 43.3 ± 7.4%; P = 0.04). There was no significant difference in baseline microdialysis glycerol concentration or percent suppression during the high-dosage insulin clamp (Fig. 2). Relative changes in adipocyte blood flow were estimated by measuring the ratio of ETOHout to ETOHin, representing the concentration of ethanol in the dialysate versus that in the perfusate during the microdialysis procedure, as previously described and validated (26,27). A lower ratio indicates decreased ETOH recovery and therefore an increase in local blood flow. There was no effect of rosiglitazone on the percent change in ETOH recovery during the low-dosage clamp (4.8 ± 2.3 [before] vs. 5.0 ± 4.7 [after]; P = 0.96).
Hepatic lipid content.
The 3 months of rosiglitazone treatment caused a 39% decrease in hepatic triglyceride content (9.3 ± 2.5 vs. 5.7 ± 1.5%; P < 0.05) (Fig. 3).
Muscle lipid content.
The IMLC did not change significantly after 3 months of rosiglitazone therapy (2.5 ± 0.4 vs. 2.7 ± 0.6%; P = 0.50). In contrast, the EMLC increased by 39%, from 1.9 ± 0.50 to 2.7 ± 0.5% (P = 0.03) after 3 months of rosiglitazone treatment (Fig. 3).
There were no detectable changes in body weight or body composition as assessed by DEXA between baseline recordings and at the conclusion of 3 months of rosiglitazone therapy: body weight 92.3 ± 3.0 vs. 93.9 ± 5.7 kg, P = 0.34; lean body mass 61.3 ± 3.4 vs. 59.9 ± 2.5 kg, P = 0.98; and percent body fat 29.9 ± 3.6 vs. 33.1 ± 4.5%; P = 0.78.
The 3 months of rosiglitazone treatment resulted in a significant improvement in insulin-stimulated glucose disposal that could mostly be attributed to an increase in nonoxidative glucose metabolism. These results are similar to the results from previous hyperinsulinemic-euglycemic clamp studies examining troglitazone treatment in patients with type 2 diabetes (3–5). This improvement in insulin responsiveness was associated with a marked reduction in hepatic triglyceride content and an increase in extramyocellular triglyceride content. Furthermore, we found that rosiglitazone treatment resulted in a significant improvement in peripheral adipocyte insulin responsiveness, as reflected by an increase in insulin suppression of peripheral adipocyte lipolysis and decreased plasma fatty acid concentrations. Overall, these results are consistent with the hypothesis that TZDs improve insulin sensitivity in patients with type 2 diabetes by activating PPAR-γ in peripheral adipocytes and promoting adipocyte differentiation (33,34). This results in an increase in smaller and more insulin-responsive adipocytes, which in turn leads to lower circulating plasma fatty acid concentrations and a redistribution of lipid (and intracellular fatty acid metabolites) from liver and muscle to peripheral adipocytes. Consistent with this theory are the recent findings that rosiglitazone’s antidiabetic effect is severely curtailed in an insulin-resistant transgenic mouse model of lipoatrophy (35).
The marked reduction that we observed in hepatic triglyceride content was associated with decreases in serum alkaline phosphatase and γ-glutamyl-transpeptidase, both markers of hepatic infiltration that are frequently elevated in steatohepatitis. These data suggest that TZD therapy may be useful in patients with nonalcoholic steatohepatitis, which has been associated with insulin resistance, obesity, and diabetes. Specific studies of patients with more severe fatty infiltration may further clarify this issue.
Surprisingly, we found that rosiglitazone caused no significant decrease in intramyocellular triglyceride content, despite a significant improvement in muscle insulin sensitivity. In contrast, Ye et al. (36) have recently reported that pioglitazone decreases triglyceride and fatty acyl CoA content in skeletal muscle of Zucker rats. Although species differences might explain these discrepant results, it is more likely that intramuscular triglyceride content is just a marker for another intramyocellular fatty acid metabolite, such as fatty acyl CoA, that is actually responsible for events leading to decreased insulin receptor substrate-1−associated phosphoinositol 3-kinase activity (34).
In contrast, rosiglitazone treatment caused a significant increase in EMCL, a finding that is consistent with rosiglitazone’s effects in promoting adipocyte differentiation and suggests that extramyocellular adipocytes behave similarly to peripheral adipocytes in response to TZDs. These findings are consistent with the results of Kelley et al. (37), who found that 3 months of troglitazone therapy in subjects with type 2 diabetes improved fasting plasma glucose concentrations while increasing total body fat mass. Body composition analysis by computed tomography revealed that this increase was exclusively attributable to an increase in peripheral adiposity, while visceral fat stores decreased. Similar results have been achieved by other groups using troglitazone (38–41) and pioglitazone (42) in patients with type 2 diabetes and lipodystrophy (43). This redistribution of body fat may result from the predilection for certain PPAR-γ agonists to induce pre-adipocyte differentiation in subcutaneous rather than visceral fat depots (45).
It is also possible that rosiglitazone improves insulin sensitivity in type 2 diabetes by altering the concentrations of certain adipocyte-derived hormones, such as leptin (6), TNF-α, adiponectin (7), or resistin (9). With regard to leptin, we found no significant effects of rosiglitazone on fasting or insulin-stimulated leptin concentrations. Previous in vitro and in vivo studies on the effects of TZDs on leptin production by adipocytes have yielded conflicting results (45–49). Nolan et al. (45) showed that co-incubation of adipocytes with troglitazone abolished the usual twofold increase in insulin-stimulated leptin production in vitro. In contrast rosiglitazone had no effect on plasma leptin concentrations in obese Zucker rats (46). In obese, nondiabetic human subjects, troglitazone caused no change in leptin levels, but has been shown to decrease leptin concentrations in diabetic subjects (47,48). Based on our results, an alteration in circulating leptin concentrations does not appear to play a major role in mediating the insulin-sensitizing effects of rosiglitazone. The effects of rosiglitazone on other circulating fat-derived factors, such as resistin, TNF-α, and adiponectin, which may modify insulin sensitivity in type 2 diabetic subjects, remain to be determined.
In summary, we found that 3 months of rosiglitazone treatment resulted in an improvement in an insulin responsiveness in type 2 diabetic subjects that was associated with a marked reduction in hepatic triglyceride content and an increase in EMLC. These changes were associated with an enhancement of insulin’s ability to suppress peripheral adipocyte lipolysis. Overall, these results support the hypothesis that TZDs enhance insulin sensitivity in patients with type 2 diabetes by promoting increased insulin sensitivity in peripheral adipocytes, which results in a redistribution of intracellular lipid from insulin responsive organs into adipocytes.
This work was supported by U.S. Public Health Service grants R01-DK-49230, P30-DK-45735, and M01-RR-00125; a U.S. Public Health Service K-23 award (K.F.P.); and a grant from Glaxo Smith-Kline (K.F.P.).
The authors thank the staff of the Yale University–New Haven Hospital GCRC for expert assistance with these studies.
Address correspondence and reprint requests to Kitt F. Petersen, M.D., Department of Internal Medicine, Yale University School of Medicine, 333 Cedar St., Fitkin 1, Box 208020, New Haven, CT 06520-8020. E-mail:.
Received for publication 3 October 2001 and accepted in revised form 31 October 2001.
S.E.I. has received honoraria and is on the speakers’ bureau for both Glaxco/SmithKline and Takeda Pharmaceuticals America and has also received research support from Takeda. G.I.S. has served as a research consultant for Glaxo/SmithKline Beecham.
DEXA, dual-energy X-ray absorptiometry; EMLC, extramyocellular lipid content; GDR, glucose disposal rate; ETOH, ethanol; ETOHin, ethanol concentration measured in the perfusate; ETOHout, ethanol concentration in the dialysate; GCMS, gas chromatography−mass spectrometry; GIR, glucose infusion rate; GP, endogenous glucose production; IMLC, intramyocellular lipid content; NMR, nuclear magnetic resonance; PPAR-γ, peroxisome proliferator−activated receptor-γ; TNF-α, tumor necrosis factor-α; TZD, thiazolidinedione.