Total and High–Molecular Weight Adiponectin in Relation to Metabolic Variables at Baseline and in Response to an Exercise Treatment Program: Comparative evaluation of three assays

A total of 60 Caucasian men and women were studied in the context of a study on insulin resistance, with 20 subjects each having NGT, IGT, and type 2 diabetes on the basis of a 75-g oral glucose tolerance test (OGTT) according to American Diabetes Association criteria (17). Subjects were enrolled in 60 min of supervised physical training sessions 3 days per week. Each training session included 20 min of biking or running, 20 min of swimming, and 20 min of warming up/cooling down periods, as previously described (5). At baseline and after 4 weeks of training, blood samples were obtained in the fasting state and measurements of anthropometric parameters were performed. The study was approved by the ethics committee of the University of Leipzig. All subjects gave written informed consent.

Metabolic and anthropometric assessment was performed as previously described (5). Insulin sensitivity was assessed in all subjects at baseline and after 4 weeks of training using the euglycemic-hyperinsulinemic clamp method. Briefly, after an overnight fast intravenous catheters were inserted into antecubital veins in both arms. One was used for the infusion of insulin and glucose; the other was used for the frequent sampling. After a priming dose of 1.2 nmol/m² insulin, the infusion with insulin was started with a constant infusion rate of 0.28 nmol/m² body surface per min and continued for 120 min. After 3 min, the variable 20% glucose infusion rate was added. The glucose infusion rate was adjusted during the clamp to maintain the blood glucose at 5.0 mmol/l. Bedside blood glucose measurements were performed every 5 min.

Serum adiponectin levels were measured using radioimmunoassay (LINCO Research, St. Charles, MO) (ADIPO_L) with a sensitivity of 1 ng/ml and an intra-assay coefficient of variation (CV) of 6.6% and also using ELISA (Mediagnost, Reutlingen, Germany) (ADIPO_M), as previously described (18). In addition, serum levels of total adiponectin, as well as HMW adiponectin, were determined using a novel ELISA (ALPCO Diagnostics, Salem, NH) (ADIPO_A). The sensitivity of this assay was 0.04 ng/ml. The recovery rate was 99–103% for total adiponectin and 97–105% for HMW adiponectin. The effect of serial dilutions has been tested on human serum samples, and linearity and specificity of the assay has been documented (16). Total and HMW adiponectin values per subject time point were obtained together in the same assay, and the ratio of HMW to total adiponectin per subject time point was calculated by dividing the respective values. All respective samples before and after exercise were measured together in the same assay.

In the current study, 10 different serum samples were used as internal controls to estimate precision for both HMW and total adiponectin. Concentrations of adiponectin, ranging from 2.02 to 11.54 μg/ml, were consecutively measured four times. For total adiponectin, the intra-assay CV was 5.3% (range 2.8–8.4) and the average interassay CV was 7.4% (3.2–9.5). The same human serum samples also were analyzed after treatment with proteinase K (used to digest adiponectin multimers), and precision was estimated at different concentrations of HMW adiponectin (0.4–6.6 μg/ml). For HMW adiponectin, the intra-assay CV was 6.8% (range 5.4–8.4) and the interassay CV was 8.7% (7.2–11.5). Based on our serum samples, the intra-assay CV was 4.2% and the interassay CV was 5.6% for total adiponectin, while for HMW adiponectin the intra-assay CV was 6.4% and the interassay CV was 7.9%. The CV of the assay was also evaluated in heparin-treated plasma samples and was 9.1% for total adiponectin and 7.9% for HMW adiponectin based on 16 pairs of internal controls. To our knowledge, this is the first study to evaluate the precision of this assay for both serum and plasma based on a large number of samples (16).

Comparisons of descriptive characteristics, expressed as means ± SD, were conducted using one-way ANOVA with Bonferroni-corrected post hoc tests and were repeated using nonparametric Kruskal-Wallis. Nonparametric Spearman correlation coefficients were calculated among baseline measures of study variables, HMW, and total adiponectin, as well as between changes in total and HMW adiponectin with changes in measures of insulin sensitivity. Comparisons of baseline and after training measures were made using both paired t tests and Wilcoxon’s rank-sum tests among all subjects and then stratified by sex and glucose tolerance group (NGT, IGT, or type 2 diabetes). A level of α = 0.05 was used to determine statistical significance. Statistical analyses were performed using SPSS version 8 (SPSS, Chicago, IL).

Participants with IGT and type 2 diabetes were significantly older and had higher BMI, waist-to-hip ratio, percentage body fat, fasting plasma glucose, 2-h OGTT glucose, fasting plasma insulin, fasting leptin, and total and LDL cholesterol compared with subjects with NGT (Table 1). ADIPO_L and ADIPO_M total adiponectin measures were strongly correlated to each other (r = 0.96) and inversely associated with baseline athropometric variables (BMI, waist-to-hip ratio, and percentage fat mass), as well as baseline parameters of insulin resistance (fasting plasma glucose, 2-h OGTT glucose, and fasting plasma insulin and positively associated with whole- blood glucose uptake) (Table 2). Moderately strong associations were also apparent with baseline lipid profile, with negative correlations of total adiponectin with total and LDL cholesterol and positive associations with HDL. The novel ELISA, ADIPO_A, produced values of total and HMW adiponectin that were highly correlated (r = 0.96) but did not show strong associations with baseline body composition, measures of insulin sensitivity, or lipid profile. HMW and total adiponectin as quantified by ADIPO_A showed weak associations with ADIPO_L and ADIPO_M (Table 2). Strength of associations between ADIPO_L and ADIPO_M total adiponectin with anthropometric measures, parameters of insulin sensitivity, and lipid profile ranged from ±0.32 to 0.61, whereas the correlations of these variables were much weaker with total and HMW adiponectin as measured by ADIPO_A (±0.03 to 0.27).

With respect to identifying insulin resistance, ADIPO_L and ADIPO_M better predicted these conditions than ADIPO_A total or HMW (Fig. 1). The area under the curve (AUC) for insulin resistance, as defined by whole-body glucose uptake <40 μmol · kg⁻¹ · min⁻¹ was 0.96 (95% CI 0.91–1.00; P < 0.0001) and 0.92 (0.83–1.00; P < 0.0001) for ADIPO_L and ADIPO_M, respectively, and each had significantly better predictive performance over ADIPO_A (P = 0.0001 for comparison with ADIPO_L and P = 0.002 with ADIPO_M), with an AUC of 0.65 (0.51–0.79; P = 0.10), and HMW adiponectin (P = 0.0001 for comparison with ADIPO_L and P = 0.001 with ADIPO_M), with an AUC of 0.63 (P = 0.20). Similar results were observed when defining insulin resistance by homeostasis model assessment >2.5 (ADIPO_L AUC = 0.96 [95% CI 0.91–1.02], ADIPO_M 0.92 [0.83–1.00], ADIPO_A 0.65 [0.511–0.79], and HMW 0.628 [0.48–0.77]) with predictive values consistent with those reported recently (12). Analysis using the ratio of HMW to total adiponectin did not materially change the results (data not shown). Similarly, AUC indicated greater ability to predict IGT of ADIPO_L (AUC = 0.95; P < 0.001) and ADIPO_M (AUC = 0.90; P < 0.001) than ADIPO_A for measuring total (AUC = 0.65; P = 0.05) and HMW adiponectin (AUC = 0.63; P = 0.10) .

The ADIPO_L and ADIPO_M techniques showed a more highly significant effect of exercise training on total adiponectin in all subjects than that seen on total or HMW adiponectin using ADIPO_A (Table 3). No significant change in total adiponectin was observed among subjects with NGT using ADIPO_L or ADIPO_M, but substantial increases in adiponectin were detected in participants with IGT or type 2 diabetes. In contrast, ADIPO_A showed a similar magnitude of effect of training in subjects with NGT, IGT, and type 2 diabetes, and the change in total adiponectin was less pronounced and did not achieve statistical significance in the IGT and type 2 diabetic groups. Stratifying by sex showed a similar pattern of more highly significant results using ADIPO_L and ADIPO_M methods than ADIPO_A, with a slightly stronger effect of training on adiponectin in men than women (data not shown). Additionally, changes with exercise correlated more strongly with changes in total adiponectin as measured by ADIPO_L and ADIPO_M than by total or HMW for free fatty acids (r = 0.38, P < 0.01 and r = 0.30, P = 0.02 vs. r = 0.03, P = 0.85 and r = 0.03, P = 0.81, for ADIPO_L, ADIPO_M, ADIPO_A, and HMW, respectively), total cholesterol (r = −0.34, P < 0.01 and r = −0.28, P = 0.03 vs. r = 0.17, P = 0.20 and r = 0.18, P = 0.16, respectively), triglycerides (r = −0.34, P < 0.01 and r = −0.20, P = 0.13 vs. r = 0.11, P = 0.41 and r = 0.09, P = 0.48, respectively), and insulin (r = −0.20, P = 0.12 and r = −0.18, P = 0.18 vs. r = 0.06, P = 0.67 and r = 0.03, P = 0.82, respectively). Similarly, the magnitudes of correlations of change in 2-h OGTT glucose and whole-blood glucose uptake were greater with change in ADIPO_L and ADIPO_M than ADIPO_A or HMW adiponectin; however, all associations failed to reach statistical significance (not shown). Change in HDL with training was not significantly associated with changes in total adiponectin or HMW as measured by any assays. Finally, given our SD of 0.3, we had 80% power to detect a difference in AUC of 0.15 between different adiponectin assays.

We confirm herein that total adiponectin as measured by ADIPO_L and ADIPO_M is significantly correlated with insulin sensitivity and metabolic variables. Additionally, total adiponectin as measured by ADIPO_L and ADIPO_M is clearly superior to either HMW or total adiponectin as measured by ADIPO_A at identifying the presence of insulin resistance and predicting levels of metabolic variables. Adiponectin levels increase with exercise and significantly correlate with improvements in lipid profile when total adiponectin is measured by ADIPO_L and ADIPO_M but not with total adiponectin or HMW adiponectin measured by ADIPO_A. Finally, we did not detect any clinically significant difference between HMW adiponectin and total adiponectin, when both are measured by ADIPO_A, at predicting insulin sensitivity and metabolic variables. The associations between metabolic variables and HMW adiponectin, when measured by techniques other than the assays used herein, remains to be seen.

Previous studies (10,13,14,19) investigating the differential associations of multimeric forms of adiponectin have used either Western blot technique or, subsequently, a recently described ELISA technique to quantify the proportion of total adiponectin circulating in the HMW form (12). A recent article (12) reporting that the HMW adiponectin–to–total adiponectin ratio is slightly superior to total adiponectin (AUC 0.713 vs. 0.615) at predicting the presence of insulin resistance and the metabolic syndrome used the same ELISA technique used herein. We replicated herein the predictive value of these measurements, as expressed by almost identical receiving-operating characteristic curves in this and the previous study (12). While we report that HMW and total adiponectin, as measured by the recently reported ALPCO ELISA (16), were not different in predicting the presence of insulin resistance or IGT, we also report that the other two total adiponectin assays evaluated herein are significantly better in this respect. This study was adequately powered to detect a 0.15 difference in AUC, which is clinically important.

We also report herein that both total and HMW adiponectin increase with exercise but that correlations with metabolic parameters are indistinguishable when total adiponectin and HMW adiponectin are considered (as measured using a commercially available ELISA from the same source). Our findings are in agreement with those reported by Bobbert et al. (14), who found that both HMW and total adiponectin increased to a similar degree with exercise and weight loss in 17 obese men and women, but there was no correlation between adiponectin (HMW or total) and insulin sensitivity either before or after weight loss in this study. HDL was the only metabolic variable correlated with total or HMW adiponectin in the previous study (14). The small study group, which included subjects with NGT, IGT, and type 2 diabetes and/or the interassay variability as well as the relative insensitivity of the Western methodology compared with radioimmunoassay or ELISA may have accounted for these prior findings (14). We have used a more sensitive and precise methodology, i.e., ELISA, to quantitate HMW adiponectin, and our sample size was 3.5 times larger. Importantly, both total adiponectin and HMW adiponectin showed similar correlation with insulin sensitivity and metabolic variables, although HMW adiponectin (measured using the current commercially available ELISA) was not significantly associated with any study variable in the current study. The magnitude of the correlation coefficients was similar to those previously reported using a larger sample size (16), however, supporting the validity of our measurements.

Similar to changes in total adiponectin, the process regulating the production of adiponectin oligomers is incompletely understood in humans. Posttranslational modification of lysine residues within the collagenous domain of the molecule appears to be involved and adiponectin glycosylation is reduced in individuals with diabetes, explaining the low HMW adiponectin observed in this population (11). Several studies where HMW adiponectin is measured by Western blot have reported a selective reduction in HMW adiponectin in subjects with type 2 diabetes (10,11,20) and a preferential elevation following treatment with the insulin sensitizer rosiglitazone (19). Two human mutations that impair the formation of HMW adiponectin have been described in individuals with type 2 diabetes, suggesting, albeit not proving, a possible role for HMW adiponectin in the pathogenesis of diabetes (10).

In the initial description of the ELISA system for selective measurement of adiponectin multimers, Ebinuma et al. (16) compared Western blotting to their selective ELISA. They found a strong correlation between multimer ELISA and densitometry results for HMW adiponectin using samples from 16 healthy volunteers, but the assay has not been tested in individuals with IGT and diabetes, in whom low HMW adiponectin levels have been described. It is also important to note that the precision of the total and selective ELISAs in the previous study are based on measurement of two human samples only (16). We expand herein by providing more extensive evaluation, and although we duplicate associations withtotal adiponectin, we report that the specific assay used may not be as sensitive as other commercially available total adiponectin assays, possibly due to differences in the antibody used.

Strengths of our study include the larger study size compared with the previous study (14) and the detailed measurement of body composition, lipids, and parameters of insulin sensitivity before and after a 4-week exercise intervention, allowing for assessment of associations both at baseline and of changes in study variables with exercise. While it is possible that laboratory measurement error may have played a role in our finding of lack of superiority of HMW adiponectin relative to total adiponectin in predicting insulin resistance, the consistency of the magnitudes of correlations from our study with the previous studies using ADIPO_L (13), ADIPO_M (21), and ADIPO_A (12) indicate that it is more likely that our results are due to the relatively poor performance of the ADIPO_A and HMW adiponectin assay. Further studies, using additional methods to measure HMW adiponectin, are needed to draw firm conclusions regarding the biological activity of HMW adiponectin.

In summary, using a recently developed, commercially available ELISA technique, we failed to demonstrate a stronger association of insulin sensitivity and metabolic variables with HMW adiponectin versus total adiponectin measured by existing assays in individuals with NGT or IGT or type 2 diabetes. This data does not support the superiority of HMW adiponectin, as measured by the novel ELISA, over total adiponectin in assessing insulin sensitivity and changes with physical training. Moreover, we document significant differences in the ability of total adiponectin to predict the presence of insulin resistance when measured using different commercially available assays.

This study was supported by a grant from the Deutsche Forschungsgemeinschaft (DFG), Clinical Research Group “Atherobesity” KFO 152 (project BL 833/1-1) (to M.B.). C.S.M. is supported by a Bessel Award of the Humboldt Foundation and National Institute of Diabetes and Digestive and Kidney Diseases R01-58785.