The Effect of Thiazolidinediones on Plasma Adiponectin Levels in Normal, Obese, and Type 2 Diabetic Subjects

Ten subjects with type 2 diabetes and 17 nondiabetic control subjects (8 lean, BMI <27 kg/m² and 9 obese, BMI >27 kg/m²) participated in the study. The clinical characteristics of the subjects are given in Table 1. Studies in subjects with type 2 diabetes were performed after discontinuation of their antihyperglycemic medications for ∼2 weeks. All participants were in good general health and had normal kidney and liver function. Subjects were excluded if they had active cardiac, hepatic, or renal disease or if they had long-term complications from diabetes. None of the control subjects were taking any medication to alter glucose tolerance. Among the diabetic subjects, five were being treated with sulfonylureas, one with metformin, two with combination sulfonylurea and metformin, and two with diet alone. The diabetic subjects were asked to discontinue taking their medications at least 2 weeks before their baseline studies. Each subject underwent an 80 mU · m⁻² · min⁻¹ hyperinsulinemic-euglycemic glucose clamp at baseline, and this was repeated after 3 months of treatment with 600 mg/day troglitazone.

The study protocol was approved by the Human Subjects Committee of the University of California San Diego. Informed written consent was obtained from each subject.

Studies were performed in the morning after a 10-h overnight fast. At 0300, an 18-gauge intravenous catheter was inserted in an antecubital vein and a constant infusion of [3-³H]glucose (0.25 μci/min) (New England Nuclear, Boston, MA) was started. At 0700, another intravenous catheter was inserted in a retrograde fashion in a hand vein, with the hand placed in a hand warmer for sampling of arterialized blood. At ∼0800, four basal blood samples were drawn at 10-min intervals for measurement of plasma glucose concentration and specific activity, insulin, and nonesterified fatty acid (NEFA) concentrations. An intravenous infusion (Humulin; Eli Lilly, Indianapolis, IN) diluted in 0.15 mol/l saline containing 1% wt/vol human albumin was then begun at a rate of 80 mU · m⁻² · min⁻¹. Potassium and phosphate were given intravenously to compensate for the intracellular movement of these ions and to maintain normal blood levels. Blood samples were taken every 20–30 min for plasma specific activity, insulin, and NEFA. Additional blood samples were obtained every 5 min with a glucose analyzer (YSI 2700 analyzer; YSI, Yellow Springs, OH) for measurement of plasma glucose. A variable infusion of [3-³H]glucose-enriched 20% glucose was delivered to maintain a plasma glucose concentration of 5 mmol/l. During the last 40 min of the insulin infusion, blood samples were obtained at 10-min intervals for determination of plasma glucose concentration and specific activity, insulin, and NEFA.

For determination of [³H]glucose specific activity, 0.65 ml plasma was deproteinized with Ba(OH)₂/ZnSo₄. After centrifugation, the neutral supernatant was evaporated and the residue dissolved in 1 ml of water. After addition of 10 ml scintillation fluid (Ecoscint, Manville, NJ), ³H disintegrations per minute were determined in an ICN 36014 liquid scintillation counter (Titertek Instruments, Huntsville, AL) with the use of an external standard to correct for quenching. Quadruplicate aliquots of the labeled glucose infusate were added to nonradioactive plasma and processed in parallel with the plasma samples to allow calculation of the [3-³H]glucose infusion rate.

Plasma insulin was measured by a double antibody technique (26). The intra- and interassay coefficients of variation were 6.8 and 7.9%, respectively. Plasma NEFA was determined using an Acyl-CoA oxidase-based colorimetric kit (Wako NEFA-C; Wako, Richmond, VA) with intra- and interassay coefficients of variation of 2.4 and 3.3%.

Plasma adiponectin was measured by a radioimmunoassay kit (LINCO Research, St. Charles, MO). This kit is based on the double-antibody/polyethylene glycol technique using ¹²⁵I-labeled adiponectin and a multispecies adiponectin rabbit antiserum.

Total glucose appearance rate (R_a) and glucose disposal rate (R_d) were calculated from the [3-³H]glucose data using the non-steady-state equations of Steele (27). A distribution volume of 0.19 l/kg and a pool fraction of 0.5 were used in the calculation (28). The basal R_d for subjects with type 2 diabetes was corrected for urinary glucose excretion. Hepatic glucose output was calculated as the difference between total glucose R_a and the rate of exogenous glucose infusion.

Results are expressed as means ± SE, unless otherwise indicated. To compare the adiponectin levels among lean, obese, and diabetic subjects before and after TZD treatment, adiponectin plasma level data were modeled by repeated-measures ANOVA using TZD treatment as the repeated variable. The effect of insulin during the clamp and TZD treatment was analyzed by ANOVA using PROC GLM, with insulin as the repeated effect. PROC GLM and PROC MIXED (SAS, Cary, NC) were used for these analyses. A P value <0.05 was considered statistically significant.

Fasting plasma glucose was significantly lower after 12 weeks of treatment compared with baseline in type 2 diabetic subjects (9.1 ± 0.9 mmol/l vs. 11.1 ± 0.9 mmol/l, P < 0.005) but was unchanged in lean (4.8 ± 0.1 vs. 4.8 ± 0.1 mmol/l) and obese subjects (4.9 ± 0.1 vs. 5.0 ± 0.1 mmol/l). Consistent with previous experience, three of the diabetic subjects could be classified as nonresponders, defined as <10% improvement in fasting plasma glucose after TZD treatment (responders 12.1 ± 1.8 vs. 8.1 ± 1.3 mmol/l, P < 0.005; nonresponders 12.7 ± 1.3 vs. 12.2 ± 0.9 mmol/l, P = 0.28). HbA_1c was not significantly different after treatment in any of the groups (lean 5.4 ± 0.1 vs. 5.8 ± 0.2%, obese 5.4 ± 0.2 vs. 5.8 ± 0.2%, diabetic subjects 8.8 ± 0.5 vs. 9.2 ± 0.6%).

No significant change in body weight occurred after 12 weeks of TZD treatment in any group.

Fasting insulin levels were significantly higher in the obese and diabetic subjects than in the lean control subjects (P < 0.05 and P < 0.005, respectively) (Table 2). After TZD therapy, insulin levels decreased in both obese and diabetic subjects and remained unchanged in the lean subjects. Taken as a whole group, fasting insulin was significantly lower (P = 0.02) after treatment.

Plasma triglyceride concentrations were higher in the diabetic subjects than in the lean control subjects (P < 0.01), as were NEFAs (P < 0.025) (Table 2). There was no statistically significant difference in total or HDL cholesterol levels between groups. After treatment with TZDs, triglyceride levels improved in the diabetic and obese subjects (P < 0.02), with an increase in HDL cholesterol in the diabetic group, whereas levels remained unchanged in the lean subjects.

Mean plasma insulin levels during the last 30 min of the glucose clamps were similar before and after TZD treatment and did not differ among groups (baseline: lean 122 ± 10, obese 131 ± 6, diabetic subjects 129 ± 7 mU/l; post-TZD: lean 119 ± 9, obese 132 ± 9, diabetic subjects 118 ± 6 mU/l).

As seen in Fig. 1, R_d during the pretreatment euglycemic clamp studies was lower in the diabetic subjects (3.4 ± 0.5 mg · kg⁻¹ · min⁻¹) than in the lean (12.3 ± 0.4) or obese subjects (6.7 ± 0.7) (P < 0.001 for both). After 12 weeks of treatment, R_d was significantly improved in the diabetic and obese groups (P < 0.05) but remained unchanged in the lean subjects. Hepatic glucose output was completely suppressed during the glucose clamps in all groups and was not affected by TZD therapy.

Compared with lean control subjects (16.7 ± 2.7 μg/ml), baseline adiponectin levels were lower in the obese group (12.1 ± 1.8) and further decreased in obese diabetic subjects (9.0 ± 1.7), and the differences between the lean and diabetic subjects were highly significant (P < 0.01) (Fig. 2). After 1, 2, and 3 months of TZD treatment, adiponectin levels increased uniformly in all groups. As can be seen in Fig. 2, the drug effect to increase adiponectin levels was maximal at 1 month and persisted throughout the study. The effect was approximately twofold in all groups, and when all of the subjects were taken together, the average increase was 12.2 ± 2.3 vs. 25.7 ± 2.6 μg/ml (P < 0.0001). Remarkably, adiponectin levels increased in each of the 27 subjects included in the study, indicating the uniformity and consistency of this effect (Fig. 3). It is also of interest that TZD treatment was fully effective at increasing adiponectin levels in the lean control subjects, despite the fact that no other metabolic effects of TZDs were observed in these subjects.

The glucose clamp studies involve a 5-h infusion of insulin at 80 mU · m⁻² · min⁻¹, which provided the opportunity to study the effects of insulin administration on adiponectin levels. As shown in Fig. 4, there was an ∼20% decrease in adiponectin levels after insulin administration before TZD treatment in the lean (16.6 ± 2.7 vs. 13.3 ± 2.0 μg/ml, P = 0.027), obese (11.9 ± 2.3 vs. 9.9 ± 1.8 μg/ml, P = 0.004), and diabetic subjects (9.0 ± 1.7 vs. 7.9 ± 1.6 μg/ml, P = 0.019). Overall, the mean values were 12.2 ± 2.3 compared with 10.2 ± 2.3 μg/ml, before and after insulin administration, respectively. Repeated-measures ANOVA revealed that the effect of insulin administration on adiponectin levels was highly significant (P < 0.01). Interestingly, the insulin-induced decrease in adiponectin levels during the clamp was not observed after the 3-month TZD treatment period (P = 0.48).

To determine whether similar results were seen in normal rats, ACRP30 levels were measured in 3-month-old male rats before and after 2 weeks of TZD treatment, and as summarized in Fig. 5, a 2.2-fold increase was observed. Glucose clamps were also performed in these animals (Fig. 5B). During the glucose clamp studies, R_d increased from basal values of 11.2 mg · kg⁻¹ · min⁻¹ to insulin-stimulated values of 49.1 mg · kg⁻¹ · min⁻¹, and this was associated with an ∼50% reduction in circulating ACRP30 levels (Fig. 5B).

In the pretreatment state, adiponectin levels were positively correlated with R_d (r = 0.46, P < 0.01) and HDL cholesterol levels (r = 0.59, P < 0.001) and negatively correlated with fasting insulin (r = −0.39, P < 0.05) and plasma triglyceride (r = −0.61, P < 0.001). Similar statistically significant correlations were observed after TZD treatment. When the change in fasting adiponectin levels was plotted against the change in insulin sensitivity before and after TZD treatment, no statistically significant relationship was observed.

Adiponectin is a member of the adipocytokine family of fat cell-secreted proteins. It is exclusively expressed in adipose tissue and circulates at relatively high (microgram per milliliter) concentrations. It shares sequence homology with human type VIII and X collagen and complement factor C1q and TNF-α (8,9,29,30), and the rodent form of this protein is called ACRP30. Although its exact function is not known, adiponectin/ACRP30 and/or the globular head portion of this protein may promote weight loss in mice (31), stimulate free fatty acid oxidation in muscle, enhance insulin sensitivity (32), and possess anti-inflammatory (33) and antiatherogenic properties (34,35).

It has been reported that plasma levels of adiponectin are lower in obesity, type 2 diabetes, and coronary artery disease (21–23) and that adipose tissue ACRP30/adiponectin mRNA levels are decreased in obese ob/ob mice and obese humans (9). Circulating adiponectin levels are also inversely correlated to triglyceride and fasting insulin concentrations as well as the magnitude of insulin resistance (21,23,24). Caloric restriction in humans and mice resulted in higher levels of adiponectin (31), with a similar increase reported after weight loss induced by gastric surgery (36). Conversely, high-fat feeding leads to decreased ACRP30 secretion in rodents (24).

Based on the above, it is apparent that ACRP30/adiponectin levels tend to be higher in states of insulin sensitivity and decreased in states of insulin resistance. Coupled with the findings that administration of adiponectin can lead to enhanced insulin action in vitro and in vivo in mice, it is possible that this peptide may play a physiologic role in maintaining normal insulin sensitivity and glucose homeostasis. To explore these ideas, we assessed the effect of TZD treatment in modulating the circulating adiponectin levels in lean, obese, and obese diabetic subjects. Previous reports have shown that TZD treatment can lead to an increase in adiponectin levels in subjects with normal and impaired glucose tolerance and patients with type 2 diabetes (37–40). Our results confirm this finding and show that TZD treatment led to a uniform increase in adiponectin levels in each subject, and this effect was observed by 1 month of treatment and remained throughout the 3-month treatment period. When taken as a group, the increase in adiponectin levels was about twofold across all subjects. Since we did not obtain measurements before 1 month, our studies do not indicate the time of onset of the TZD effect on adiponectin levels. However, a recent study by Combs et al. (40) demonstrated an increase in adiponectin levels in normal subjects after only 14 days of treatment with rosiglitazone. This is supported by our studies in rats showing a similar increase after 2 weeks of TZD treatment.

The uniformity of the effect of TZDs on adiponectin levels was striking. Thus, each individual subject displayed an increase, including the normal subjects, in whom no other metabolic effect of TZD treatment had been observed. Furthermore, it is well known that not all type 2 diabetic subjects respond to TZD therapy; in our studies, we found that three of the diabetic subjects did not show metabolic improvement with TZD treatment, consistent with previous reports (41–43). Interestingly, in each of these TZD “nonresponders” (arbitrarily defined as subjects with <10% decrease in fasting plasma glucose after treatment), an increase in adiponectin levels was observed. The mean increase was 10.5–22.3 μg/ml in the nonresponders compared with 8.5–19.2 μg/ml in the responders, indicating that the effect of TZD treatment to increase adiponectin levels was comparable, regardless of whether the subjects showed a glycemic or insulin-sensitizing effect of the drug. Although adiponectin administration has been shown to improve insulin sensitivity and lower glucose levels in mice and to be inversely correlated with the magnitude of insulin resistance, the fact that adiponectin levels increase in normal rats, lean subjects, and nonresponders indicates that adiponectin is unlikely to be the sole cause of TZD-induced insulin sensitivity.

Since insulin resistance is associated with hyperinsulinemia, the relationship between adiponectin levels and insulin sensitivity also implies an inverse relationship between adiponectin and insulin levels, and this was confirmed in the current studies. This raises the possibility that insulin might have an independent effect to modulate adiponectin production from adipocytes. During the glucose clamp studies in both humans and rats, we found that insulin infusion led to a decrease in adiponectin (or ACRP30) concentrations, consistent with the interpretation that insulin can exert an effect on adipocytes to decrease production and/or secretion of this adipocytokine. Thus, it is possible that the chronic hyperinsulinemia associated with insulin-resistant states leads to downregulation of ACRP30/adiponectin concentrations. This raises the idea that the effect of TZDs on adiponectin/ACRP30 levels are at least partly mediated by TZD-induced changes in insulin levels, which are secondary to the effect of TZDs to enhance insulin sensitivity. Although TZD-induced changes in insulin levels may partly explain the changes in adiponectin/ACRP30 levels observed in the current studies, they are unlikely to represent the complete effect of these agents, since in vitro studies have shown that TZDs can directly increase adipocyte ACRP30 mRNA.

It is most likely that the effect of insulin to lower circulating adiponectin/ACRP30 levels involves inhibition of adipocyte secretion. Since blood concentrations fall from 20 to 50% (humans versus rats, respectively) after a few hours of insulin administration, these studies suggest that the circulating half-life of this adipocytokine is in the range of several hours. This rate of turnover would be consistent with a role for this protein as a regulator of metabolic processes. Clearly, it would be of important interest to determine the more detailed pharmacokinetics of this peptide as well as its regulatory controls.

Since our results demonstrate a robust effect of TZD administration to increase circulating adiponectin levels, independent of the effects of these agents to produce insulin sensitization or any other metabolic effect, measurement of adiponectin levels in clinical studies may serve as a convenient biomarker for drug administration. It will be important to examine the dose-response effect of TZD administration on adiponectin levels as well as the early time course of this effect.

This work was supported by a research grant form the National Institutes of Health (DK 33651), the Veterans Administration San Diego Healthcare System Research Service, and the Whittier Institute for Diabetes.

The authors thank Jacob Rabenstein for his excellent technical assistance.