Adiponectin Expression From Human Adipose Tissue: Relation to Obesity, Insulin Resistance, and Tumor Necrosis Factor-α Expression

This study involved 62 weight-stable subjects ages 23–61 years. All subjects gave informed consent, and the research was approved by the institutional review board. Subjects initially underwent a 75-g oral glucose tolerance test, and subjects with diabetes (fasting glucose >126 mg/dl, 2-h glucose >200 mg/dl) were excluded. Of the 62 subjects, 15 had impaired glucose tolerance based on a 2-h glucose test of 140–200 mg/dl, and 4 of these subjects had impaired fasting glucose based on a fasting glucose of 110–126 mg/dl. Subjects then underwent a frequently sampled intravenous glucose tolerance test (FSIVGT) and an adipose tissue biopsy, which were performed on separate days.

Characteristics of the study subjects are shown in Table 1. Blood lipids, glucose, and HbA_1c were measured using standard clinical assays. Of the 62 subjects studied, 50 were women and 12 were African American. The subjects ranged from lean to very obese (BMI range 19–65 kg/m²). Some subjects demonstrated moderate dyslipidemia, but no subject demonstrated fasting triglycerides >400 mg/dl. Body composition was determined using bioelectric impedance (18).

The measurement of in vivo insulin sensitivity was performed in the fasting state using the classic tolbutamide-modified minimal model analysis of the FSIVGT (19,20), which has been validated against the euglycemic clamp (21,22). Four basal blood samples were obtained at time 0. Patients were then given an intravenous glucose bolus (11.4 g/m²) and, 20 min later, an injection of tolbutamide (125 mg/m²). Frequent blood sampling was then performed according to the standard protocol. Glucose was measured using the glucose oxidase method in a glucose analyzer and insulin was measured using radioimmunoassay (RIA; Endocrinology Laboratory of the Indiana University School of Medicine, Indianapolis, IN). The insulin sensitivity index (S_I) was calculated using the MINMOD program, along with the acute insulin response to glucose (acute insulin response to glucose [AIR_glu]) (20).

Abdominal subcutaneous adipose tissue (∼5 g) was removed from each patient by incision. Some of the tissue was immediately frozen in liquid N₂ for later RNA extraction, whereas the rest of the tissue was placed into cold Dulbecco’s modified Eagle’s medium (DMEM) for other assays. To measure the secretion of adiponectin and TNF-α, ∼500 mg of adipose tissue were minced and placed into serum-free DMEM (pH 7.4, 10 mmol/l HEPES) at 37° for varying times, as previously described (23). To compare cytokine secretion among different subjects, we measured cytokine levels in the medium after 2 h at 37°. All data were normalized to either adipose DNA content (24) or cell number to control for differences in fat cell size. Cell number was measured using the method of DiGirolamo et al. (25).

The measurement of adiponectin protein used a radioimmunoassay (Linco Research, St. Charles, MO). This assay demonstrates a 4.3% intra-assay variation and a 7.1% interassay variation. This assay method was used to measure adiponectin in fasting plasma as well as secretion by adipose tissue, as described above. Adiponectin mRNA levels were measured by Northern blotting using the cDNA to human adiponectin, and the same blots were reprobed with the cDNA to 18S RNA as a constitutive probe (Fig. 1B). To quantitate adiponectin mRNA expression, the blots were analyzed by densitometry, and the adiponectin/18S RNA ratio was given as arbitrary units.

TNF-α mRNA levels were measured as previously described using competitive RT-PCR (26). In brief, 0.4 μg of total RNA from adipose tissue were added to increasing quantities of a competing RNA construct containing an internal 49-nt deletion. After the RT and PCR reactions, the products were resolved on a 2% agarose gel and the ethidium bromide−stained gel was quantitated. Data are expressed as the “number of copies” per microgram of total RNA, where “number of copies” refers to the number of copies of cRNA added at the equivalence point between TNF-α mRNA product and cRNA. TNF-α and IL-6 protein were measured using enzyme-linked immunosorbent assays (R&D Systems, Minneapolis, MN), and leptin was measured using an RIA (Linco Research).

All data are expressed as means ± SE. Analysis of trends was performed using linear regression. When comparing two groups, a Student’s t test was used, and to analyze data among groups of three or more, a one-way ANOVA was performed and secondary analysis was performed with the Student’s t test with Bonferroni correction.

Adiponectin was measured in plasma, in the medium secreted from adipose tissue, and adiponectin mRNA levels were measured by Northern blotting. To examine the relation between adiponectin and obesity, adiponectin expression was measured in subjects covering a wide range of BMI and percent body fat. When all subjects were considered together, there was a significant inverse association between BMI and plasma adiponectin (r = −0.39, P < 0.002, n = 62) (Fig. 1A). In addition, adiponectin mRNA was assessed by Northern blotting, which yielded a single band for adiponectin (Fig. 1B). There was a strong correlation between adiponectin mRNA levels and plasma adiponectin levels (r = 0.80, P < 0.001, n = 24; data not shown), and likewise a significant inverse correlation between adiponectin mRNA and BMI (r = −0.55, P < 0.01, n = 24; data not shown).

Among subjects with similar BMIs, there was considerable variation in plasma adiponectin level. As can be seen in Fig. 1A, some of this variation could be attributed to lower adiponectin levels in men. To more appropriately analyze this association and account for gender differences, adiponectin expression was analyzed in relation to percent body fat. This resulted in a stronger association, along with significant gender differences. With increasing body fat, there was an overall significant decrease in plasma adiponectin when all subjects were considered together (r = −0.27, P < 0.02, n = 62). However, as shown in Fig. 2, this association was stronger when women and men were considered separately (women: r = −0.50, P < 0.001, n = 50; men: r = −0.60, P < 0.05, n = 12). At similar degrees of body fat, women demonstrated considerably higher plasma adiponectin levels than men, although this difference was less apparent with obesity and high body fat levels. To more directly compare men and women, we examined plasma adiponectin levels in normal weight and mildly overweight men and women (BMI <30 kg/m²). As shown in Table 2, these men and women had the same BMI and the women had higher body fat, as expected. According to the trend described in Fig. 1A, involving lower plasma adiponectin levels with increasing adiposity, women would be expected to have lower plasma adiponectin levels because of their higher body fat. Instead, as shown in Table 2, women had significantly higher plasma adiponectin levels (8.6 ± 1.1 and 14.2 ± 1.6 μg/ml in men and women, respectively; P < 0.02) and also higher adiponectin mRNA levels.

Insulin sensitivity was expressed as S_I using the MINMOD analysis of the FSIVGT, as described in research design and methods. Figure 3A shows the relation between adiponectin and S_I. There was a significant positive relation between S_I and plasma adiponectin (r = 0.67, P < 0.0001, n = 51) and a similar relation between S_I and adiponectin mRNA levels (r = 0.53, P < 0.01, n = 24; data not shown). This relation between adiponectin expression and S_I was unaffected by gender (data not shown). To determine whether adiponectin affected insulin secretion, the AIR_glu was assessed in these subjects. No relations between insulin secretion, using the AIR_glu, and adiponectin was observed (data not shown). Previous studies have reported the product of S_I × AIR_glu, called the disposition index (DI), as a heritable trait that indicates the ability of the β-cell to compensate for insulin resistance (19,27). We observed no significant relation between DI and adiponectin expression (data not shown).

As described by numerous previous studies, there was significant relations between obesity and insulin sensitivity. Among the subjects in this study, there was a significant correlation between BMI and S_I (r = −0.60, P < 0.001) and a similar correlation with percent body fat. We wanted to determine whether there was a significant relation between plasma adiponectin levels and insulin resistance that was independent of obesity. To assess this, we examined adiponectin expression in subjects who had similar degrees of adiposity, but were discordant for S_I. Therefore, paired subjects were identified who were matched for BMI (± 3 kg/m²), age (± 10 years), and gender, but who were discordant for S_I. Because S_I is a continuous variable that varies across the population, we used the median value of our population, which is 2.6 units, to separate insulin-sensitive from insulin-resistant subjects. Using these criteria, we were able to match 11 subjects with S_I <2.6 with 11 subjects with S_I >2.6 (in each subject pair, the difference between S_I values was >2.0 units). The result of this matching is shown in Fig. 3B. There were no significant differences between the insulin-resistant and insulin-sensitive group with regard to BMI, age, percent body fat, or plasma leptin, but there were significant differences in S_I, based on the selection. Although these subjects were well matched with regard to adiposity, the insulin-sensitive subjects demonstrated a two-fold higher level of serum adiponectin (5.6 ± 0.6 and 11.2 ± 1.1 μg/ml in insulin-resistant and insulin-sensitive subjects, respectively; P < 0.0005).

Adipose tissue expresses a number of cytokines, and previous studies have implicated adipose tissue TNF-α, IL-6, and leptin in obesity-related insulin resistance (23,28). To determine whether adiponectin expression was related to the expression of other cytokines, plasma IL-6, leptin, TNF, and TNF expression were measured in these subjects. There was no significant relation between adiponectin expression and plasma leptin (r = −0.12, NS; data not shown). In previous studies, plasma IL-6 correlated strongly with S_I (23). Even though IL-6 and adiponectin were both strongly associated with S_I, there was only a nonsignificant inverse trend between plasma IL-6 and plasma adiponectin (r = −0.24, P = 0.10, n = 48). However, there was a significant relation between adiponectin and TNF-α expression in adipose tissue. As shown in Fig. 4, subjects with high plasma adiponectin levels demonstrated low levels of TNF mRNA expression (r = −0.47, P < 0.005, n = 36). In a similar fashion, adiponectin mRNA levels were significantly associated with TNF mRNA levels (r = −0.48, P < 0.05, n = 20; data not shown). Previous studies demonstrated that TNF-α secretion from adipose tissue was strongly associated with obesity-related insulin resistance (23), suggesting that TNF-α may function in a paracrine fashion in adipose tissue. To examine the relation between TNF secretion and adiponectin, we examined TNF secretion in tertiles of adiponectin mRNA. As shown in Fig. 5, the low, middle, and high tertiles of adiponectin mRNA levels were associated with a significant inverse relation with adipose TNF-α secretion. Gender did not influence the relation between adiponectin and TNF-α. Therefore, the expression of adiponectin by adipose tissue was significantly associated with TNF-α expression, including TNF-α secretion, but not with the expression or plasma levels of IL-6 or leptin.

Adiponectin is a 29-kDa protein, expressed exclusively in adipocytes, that was initially called adipocyte complement−related protein of 30 kDa (ACRP30) because of its homology to complement factor C1q (29,30). Since the initial description of adiponectin, numerous animal and human studies have examined the relation of adiponectin to obesity, diabetes, and other associated syndromes. Plasma levels of adiponectin are low in obese, compared with lean, subjects; lower in diabetic, compared with nondiabetic, subjects; and lower in patients with coronary artery disease (17,31,32). In addition, plasma adiponectin levels increase following weight loss (33).

Several studies have demonstrated an association between plasma adiponectin and markers of insulin resistance. Among very obese subjects, a weak association between plasma adiponectin and insulin sensitivity has been shown using the steady-state plasma glucose method (33). In another study, plasma insulin was weakly associated with plasma adiponectin, but this relation was lost when the investigators controlled for BMI (17). Using the euglycemic clamp in a population of Pima Indians and Caucasians, Weyer et al. (34) demonstrated a strong inverse association between plasma adiponectin and insulin resistance, an association that was independent of obesity. In addition, patients and rodents treated with thiazolidinediones, which improve insulin sensitivity, demonstrated higher plasma adiponectin levels (35–37). Finally, a longitudinal study in rhesus monkeys demonstrated a significant association between insulin sensitivity and plasma adiponectin, along with a progressive decline in adiponectin with the development of obesity and insulin resistance (38).

At present, a receptor and precise target tissue for adiponectin action has not been described. However, several studies in rodents have provided evidence that adiponectin is important in tissue insulin action. In a mouse model of lipoatrophy, adiponectin levels were very low, and the administration of adiponectin in combination with leptin restored insulin sensitivity (13). That study demonstrated decreased lipid content in liver and muscle and improved lipid oxidation in muscle with adiponectin administration. The infusion of an adiponectin cleavage product containing the globular head of the protein resulted in a decrease in plasma nonesterified fatty acid, along with an increase in muscle lipid oxidation (14). Other studies have demonstrated an improvement in hepatic insulin sensitivity (15), and the infusion of mice with adiponectin during a euglycemic clamp demonstrated that the hypoglycemic effect of this cytokine was attributable to a suppression of hepatic glucose production (39). Together, these studies in rodents suggest that adiponectin may play an important role in preventing the development of lipotoxicity in insulin-sensitive tissues that commonly occurs in the metabolic syndrome (40).

This study examined the expression of adiponectin from adipose tissue in relation to insulin sensitivity in nondiabetic human subjects. Although plasma adiponectin levels were generally higher in lean compared with obese subjects, the relation was not very strong when expressed in terms of BMI. Some of the discrepancy between obesity and plasma adiponectin, however, could be accounted for by differences between men and women and by expressing obesity in terms of percent body fat. As shown in Fig. 2, the differences in plasma adiponectin between men and women were considerable, especially among subjects who were lean or mildly overweight. Among subjects with a BMI <30 kg/m², women predictably have a higher percent body fat. If plasma adiponectin levels were determined predominantly by fat mass, as with leptin, then one would expect women to have a lower plasma adiponectin, based on the overall inverse relation between adiponectin and obesity. However, women with a BMI <30 kg/m² actually express about twofold more adiponectin than men of the same BMI. It is interesting that this striking male/female difference was observed in one previous study (31), but was not observed in a study involving 144 subjects, of which 121 were Pima Indians (34). The reason for the higher adiponectin expression in women is unclear, but perhaps the higher adiponectin expression, along with differences in body composition, help explain the ability of women to maintain a normal insulin sensitivity despite higher body fat.

The relation between adiponectin and insulin resistance was stronger than the relation with obesity, and no gender differences were observed. In addition, the association between adiponectin and insulin sensitivity was independent of obesity, as demonstrated by the matching of subjects who were concordant for age and BMI but discordant for S_I. In contrast, leptin was not different in these groups. In a previous study, we demonstrated that TNF secretion and plasma IL-6 were also associated with insulin resistance, independent of obesity (23). Thus, these data suggest that adipose tissue expresses a number of cytokines that collectively are strongly associated with peripheral insulin sensitivity.

Although adiponectin levels were significantly associated with S_I, we found no evidence of an association between adiponectin levels and insulin secretion. This is of interest because of potential links between islet lipid content and insulin secretion. Troglitazone and other peroxisome proliferator−activated receptor-γ agonists increase adiponectin levels (35–37), lower islet lipid content, and improve β-cell function in Zucker diabetic rats (41). Because adiponectin has been linked to reduced lipid accumulation in tissues (13), it is possible that adiponectin also reduces islet lipid content, leading to increased insulin secretion or an improved β-cell response to insulin resistance. Although these studies did not fully test this concept, we could find no evidence for improved β-cell compensation in subjects with high adiponectin levels.

Several lines of evidence have suggested that TNF-α overproduction by adipose tissue may be involved in obesity-related insulin resistance. TNF is expressed at higher levels in the adipocytes of obese rodents and humans (26,42,43), and TNF knockout mice do not become insulin resistant with diet-induced obesity (44). In addition, the infusion of a soluble TNF-binding protein into insulin-resistant fa/fa rats improved insulin sensitivity and improved the defect in insulin receptor autophosphorylation (42,45). A recent study described a relation between adiponectin and TNF-α in adiponectin knockout mice (16). Although adiponectin knockout mice demonstrated normal blood glucose and insulin in the basal state, they became highly insulin resistant when fed a high-fat diet. TNF expression was higher in the adiponectin knockout mice, and the administration of adiponectin in these mice resulted in an improvement in insulin resistance along with a decrease in TNF expression. Thus, these studies in mice suggest that TNF-α and adiponectin may be antagonists of each other or that one cytokine may control the expression of the other.

This study examined the expression of adiponectin and TNF-α in human adipose tissue and found a significant association between adiponectin and TNF-α expression. In previous studies, we found that the strongest association between TNF-α and insulin resistance was with the adipose-secreted component of TNF, and not with plasma TNF-α, suggesting a paracrine function of TNF-α in adipose tissue (23). In these studies, we found that there was a strong relation between adiponectin expression and TNF secretion from adipose tissue. In addition, TNF secretion from human adipose tissue is significantly associated with IL-6 secretion; plasma IL-6 is also associated with insulin resistance, independent of obesity (23). Taken together, these data suggest an emerging paradigm of adiponectin and TNF-α as important control points in adipose tissue that may lead to the expression of other cytokines (e.g., IL-6) and be linked to the development of tissue lipid accumulation and the insulin-resistance syndrome.

This study was supported by a merit review grant from the Veterans Administration, General Clinical Research Center Grant M01-RR-14288, and National Institutes of Health Grant DK-39176.

We thank the many people who participated in these studies, as well as Denise Hargrove and Leslie Langum for assistance with subject recruitment and the nurses of the General Clinical Research Center at the University of Arkansas for Medical Sciences and Central Arkansas Veterans Healthcare System. We also thank Dr. Richard Bergman for supplying the MINMOD program and Sarah Dunn for excellent secretarial assistance.