Circulating Adiponectin Levels Are Reduced in Nonobese but Insulin-Resistant First-Degree Relatives of Type 2 Diabetic Patients

Forty-five FDR were recruited by advertisements in local newspapers and via questionnaires, and 40 CON were randomly selected among men in the county council register for Göteborg. Inclusion criteria were subjects with two first-degree relatives or one first-degree relative and two second-degree relatives with type 2 diabetes (grandparents, uncle, or aunt); male sex (to exclude variation in insulin sensitivity during the menstrual cycle); normal glucose tolerance; and no evidence of hypertension, endocrine disease, or obesity (BMI >30 kg/m²) (Table 1). The control group consisted of subjects who did not have a known family history of diabetes but fulfilled the remaining criteria. FDR and CON were similar with respect to age (43 ± 9 vs. 45 ± 7 years), cigarette smoking (3.7 ± 6.8 vs. 5.7 ± 10.9 pack-years), and use of smoke-free tobacco (6.4 ± 11.4 vs. 2.3 ± 8.0 g/day, NS), respectively. All participants gave informed consent and the study was approved by the Ethical Committee of Göteborg University.

Blood pressure was measured with a standard mercury sphygmomanometer on the right arm after the subjects had been resting in the supine position for at least 5 min. Mean values were determined from two independent measurements taken at 5-min intervals.

All subjects underwent a 75-g oral glucose tolerance (OGTT) test after having fasted overnight. Baseline samples were drawn from an antecubital vein and stored at −20° or −80°C until analysis.

Glucose was injected intravenously in an antecubital vein over a period of 60 s (0.3 g/kg body wt of 30% glucose) to measure the acute insulin response. Twenty minutes after the glucose injection, insulin (Actrapid; Novo Nordisk, Copenhagen, Denmark) was administered intravenously as a bolus of 0.03 units/kg. Blood samples were collected at 20 time points during the intravenous glucose tolerance test (IVGTT) at –5, –1, 2, 4, 6, 8, 10, 14, 19, 22, 30, 40, 50, 60, 90, 100, 120, 140, 160, and 180 min. Insulin sensitivity index (S_i) was calculated using the Bergman MINIMOD computer program (18).

Body fat was calculated from the bioelectrical impedance method (BIA-103; RJL Systems, Detroit, MI) (19).

Glucose was analyzed in venous blood using an automatic glucose analyzer (Yellow Springs Instruments, Yellow Springs, OH), and plasma insulin was analyzed with a standard radioimmunoassay having 40% cross-reactivity with proinsulin (Pharmacia, Uppsala, Sweden). Proinsulin was measured with the Mercodia Proinsulin ELISA (Mercodia AB, Uppsala, Sweden).

Lipid concentrations were determined with an automated Cobas Mira analyzer (Hoffman-LaRoche, Basel, Switzerland) as previously reported (20,21), and LDL cholesterol was calculated according to Friedewalds formula (LDL cholesterol = total cholesterol − HDL cholesterol − 0.45 × triglyceride level [mmol/l]). LDL peak particle diameter was measured with gradient gel electrophoresis (22).

C-reactive protein (CRP) was measured with an immunoenzymometric assay for quantitative determination of human CRP (CRP IEMA Test; Medix Biochemica, Kauniainen, Finland), while serum intracellular adhesion molecule (ICAM), vascular cell adhesion molecule (VCAM), and E-selectin were measured by the quantitative sandwich enzymatic immunoassay technique using Parameter kits (catalogue no. BBE 3, 1B, 2B; R&D Systems, Minneapolis, MN). The optical densities were read in a Perkin Elmer HTS 7000 Plus BioAssay fluorescent and absorbance microplate reader.

Fibrinogen and PAI-1 were analyzed using standard methods, and Apo E genotype was determined by a PCR technique. Adiponectin was analyzed by a quantitative immunoblotting technique (23).

For descriptive purposes, mean and SD were used. The Mann-Whitney U test was used for comparisons between the groups. To adjust for confounding variables when comparing the groups, logistic regression was used. All correlations were analyzed with Spearman’s nonparametric correlation coefficient. To adjust for confounding variables in the correlation analysis, Spearman’s partial nonparametric correlation coefficient was calculated. To select independent predictors, only variables with a univariate correlation (P < 0.1) were chosen. Then, a stepwise multiple regression analysis was used after transforming the dependent variable to normal distribution by calculating normal score using Blom’s method (24).

All tests were two-tailed and conducted at 5% significance level. SAS 8.2 was used for statistical calculations.

The basic characteristics of the 85 participants are shown in Table 1. The FDR had higher BMI, waist circumference, body fat, and blood pressure levels. However, Apo E phenotype distribution was similar for the two groups (data not shown).

All participants had normal glucose tolerance, but the glucose concentrations during the OGTT were higher in the FDR as compared with the CON (Table 1). Furthermore, plasma insulin at 2 h was increased, and the FDR were also insulin resistant as determined with the Minimal Model computer program.

Lipoprotein concentrations, inflammatory markers, PAI-1, cellular adhesion molecules, and adiponectin levels are presented in Table 1. Comparisons between the groups showed that PAI-1 and LDL cholesterol levels were significantly higher in the FDR. Moreover, the FDR also had significantly lower adiponectin levels while CRP was higher, but this difference only reached borderline significance (P < 0.1). The lower adiponectin levels in the relatives remained also when the small differences in BMI, body fat, and waist were adjusted for (data not shown). Thus, high propensity for type 2 diabetes is associated with lower adiponectin levels also when adjusted for body fat and of its localization.

When adjusting for age, there were significant negative correlations between adiponectin and proinsulin (r −0.64, P < 0.001), PAI-1 activity (r −0.56, P < 0.001), fasting insulin (r −0.55, P < 0.001), and the acute insulin response (r −0.40, P < 0.05) in the FDR (Table 2). In addition, E-selectin showed a negative correlation of borderline significance (r −0.27, P < 0.1). We also found significant positive correlations between adiponectin and HDL cholesterol (r 0.48, P < 0.01) and S_i (r 0.41, P < 0.01). In the CON group, there were no significant correlations between adiponectin and proinsulin or HDL cholesterol, but otherwise essentially similar correlations appeared as compared with those found in the FDR (data not shown).

In stepwise multiple regression analyses with adiponectin as a dependent variable in the FDR, insulin, proinsulin, HDL cholesterol, and PAI-1 levels remained significant throughout partial correlation models adjusted for age, waist, and S_i and were entered as independent variables. The analyses excluded all variables, with the exception of HDL cholesterol (parameter estimate –0.05, SE 0.01, F value 12.6, P = 0.002) and fasting plasma proinsulin (parameter estimate 1.53, SE 0.43, F value 10.7, P = 0.002), that remained significant explanatory variables of adiponectin, with a total explanatory power of 51%.

In the CON, a similar analysis revealed that fasting insulin (parameter estimate –0.12, SE 0.04, F value 7.7, P = 0.009) explained 48% of the variability in the adiponectin concentration.

The two salient findings of the present study are 1) circulating adiponectin levels are significantly lower in healthy individuals with high propensity for type 2 diabetes, adjusted for BMI, body fat, or waist circumference; and 2) adiponectin levels, when adjusted for age, are related to several key risk factors for CVD in the FDR group. These factors include both metabolic risk factors related to insulin sensitivity (like hyperinsulinemia, proinsulin, HDL cholesterol, and PAI-1) and cellular adhesion molecules (E-selectin). Thus, these data support experimental studies that adiponectin can improve insulin sensitivity and action in obese rodent models, as well as several observations suggesting that it exerts antiatherogenic effects (12,25–30).

In contrast, a recent study of FDR and CON found similar adiponectin levels in the two groups and no relationship between adiponectin and different measures of insulin sensitivity in the FDR group. However, the authors found significantly reduced levels of adiponectin mRNA in subcutaneous adipose tissue from FDR as compared with CON (31). We did not examine adiponectin mRNA in this study, but recent findings in our laboratory are consistent with lower adiponectin mRNA levels in nonobese insulin-resistant subjects (data not shown). This is a novel finding, although it is well established that plasma adiponectin levels are decreased in other insulin-resistant states, such as obesity and type 2 diabetes (9,13,15,32).

Interestingly, small insulin-sensitive adipocytes appear to secrete more adiponectin, since obese monkeys with hypercellular but small adipocytes have higher plasma adiponectin levels than obese monkeys with fat cell hypertrophy (33). By analogy, activation of PPAR-γ, a transcription factor of key importance for adipocyte differentiation, also leads to increased adiponectin levels (34). Yu et al. (35) also recently reported that troglitazone, a synthetic PPAR-γ ligand, increased plasma adiponectin levels in lean, obese, and type 2 diabetic subjects after 3 months. Moreover, adiponectin correlated with HDL cholesterol, which is in agreement with the present data as well as other recent studies (14,16,17). Taken together, these observations support the concept that the cellular expression of adiponectin is under the control of PPAR-γ and that it is more closely related to insulin sensitivity than obesity.

In this study, we also found that adiponectin was inversely correlated with proinsulin. This was true after adjustment for age, waist circumference, and insulin sensitivity and has, to our knowledge, not been reported previously. The reason for this is unclear, but may be a consequence of the increased insulin levels and, initially, insulin resistance. However, other possibilities, including effects on the intracellular processing of insulin, cannot be excluded. Hyperproinsulinemia has been shown to be a risk factor for cardiovascular disease (36), and this may also be related to lower adiponectin levels. Similarly, we found in the present study a strong correlation between circulating PAI-1 levels and adiponectin. PAI-1 is also a risk factor for CVD (37), and its secretion from the adipose tissue is increased by cytokines like TNF-α (38). Adiponectin has been shown to reduce TNF-α secretion as well as the TNF-α–induced expression of adhesion molecules in endothelial cells (12).

Taken together, our data lend further support to the hypothesis that adiponectin can exert antiatherogenic effects in humans.

In conclusion, adiponectin levels were significantly reduced in nonobese but insulin-resistant FDR with a high propensity for type 2 diabetes. In addition, an association was found between adiponectin, proinsulin, HDL cholesterol, and other facets of the insulin resistance syndrome. Thus, adiponectin may be an important mediator of the relationship between insulin resistance and atherosclerosis, and thus could be an important target for future diabetes therapy.

This work was supported by the Inga-Britta and Arne Lundberg Foundation, the Swedish Diabetes Association, Novo Nordisk Foundation, the Regional Health Care Authority of West Sweden, Swedish Medical Research Council (grant no. K2002-72X-03506-31C) and the European Union (project QLG1-CT-1999-00674).

We thank Eva Alfvegren, Erika Löfstedt, and Caroline Moberg for technical assistance and for their help with recruiting the subjects. Statistical advice was provided by Gunnar Ekeroth.