Hypoadiponectinemia and Proinflammatory State: Two Sides of the Same Coin?

RESEARCH DESIGN AND METHODS

The population-based Cooperative Health Research in the Region of Augsburg Survey 4 (KORA S4) study population has been described extensively (25–29). In the age range 55–74 years, 120 individuals with newly diagnosed type 2 diabetes (and thus without antidiabetic treatment, which could interfere with adiponectin and immune marker levels) and 242 subjects with impaired glucose tolerance (IGT) were available for analysis. Control subjects with normal glucose tolerance (NGT; n = 244) were randomly selected after frequency matching for age and sex.

The measurement of anthropometric parameters, metabolic and immunological variables (from fasting blood samples), blood pressure, and the assessment of alcohol intake, smoking, and physical activity has been described previously (25–29). Hypertension was defined as use of antihypertensive treatment or systolic blood pressure ≥140 mmHg or diastolic blood pressure ≥90 mmHg. In the NGT group, 42 men and 31 women were on antihypertensive treatment. In the IGT/type 2 diabetic group, 89 men and 85 women received antihypertensive treatment. Serum adiponectin concentrations were determined using the human adiponectin radioimmunoassay from Linco Research (St. Charles, MO). Mean intra- and interassay variations in control sera were 5.5 and 9.2%, respectively.

Statistical analyses

Adiponectin concentrations were compared between men and women using Wilcoxon’s test and between the different glucose tolerance status groups using the Kruskal-Wallis test (three groups), and in case of significance, Wilcoxon tests for each pairwise group comparison. The associations of anthropometric, metabolic, and immunological variables with adiponectin were estimated after adjustment for age and sex or after adjustment for age, sex, IGT, type 2 diabetes, BMI, hypertension, total cholesterol, HDL cholesterol, and uric acid using linear regression models with log values of adiponectin as dependent variables and the aforementioned variables and one of the immunological parameters as independent variables. Immune markers entered the linear regression models as dummy variables based on quartiles of normoglycemic subjects. P < 0.05 was considered statistically significant. Calculations were carried out using the SAS statistical package version 8.2 TS2M0.

RESULTS

Clinical, anthropometric, metabolic, and immunological characteristics of the study sample are summarized in Tables 1 and 2. Serum adiponectin concentrations were highly dependent on sex and on glucose tolerance status. Adiponectin levels (median [25th–75th percentile]) were significantly higher in women than in men (10.6 [7.6–13.6] vs. 6.6 μg/ml [4.8–9.0]; P < 0.0001). Adiponectin concentrations in women with NGT, IGT, or type 2 diabetes were 12.2 (9.5–15.0), 9.7 (7.2–12.5), and 9.0 μg/ml (6.9–11.7), respectively. In men, adiponectin levels in subjects with NGT, IGT, or type 2 diabetes were 7.8 (5.4–9.7), 6.1 (4.6–8.5), and 5.9 μg/ml (4.3–7.8), respectively. In both women and men, adiponectin levels of normoglycemic subjects were significantly higher than in subjects with IGT or with type 2 diabetes, whereas adiponectin concentrations between subjects with IGT and type 2 diabetes did not differ statistically significantly.

First, the associations of adiponectin with leukocyte count, acute-phase proteins, cytokines, cytokine receptors, and chemokines were evaluated in subjects with NGT or with disturbed glucose metabolism (IGT/type 2 diabetes) after adjustment for age and sex (Table 1). In the NGT group only, we found a positive correlation between adiponectin and tumor necrosis factor-α (TNF-α; P = 0.023). However, there was no evidence for a significant correlation between adiponectin and any of the other immunological markers in both NGT and IGT/type 2 diabetic groups.

In contrast, there were strong positive correlations between adiponectin and age or HDL cholesterol and strong negative correlations with fasting insulin, homeostasis model assessment of insulin resistance, HbA_1c (A1C), and triglycerides in both groups and additionally with fasting glucose, uric acid, and serum albumin in the IGT/type 2 diabetic group (Table 2). Adiponectin levels were also significantly associated with indexes of obesity and body fat distribution (e.g., inverse associations with BMI, waist circumference, and waist-to-hip ratio [WHR]) (data not shown), but these associations did not remain significant after adjustment for glucose tolerance status, sex, and age (Table 2).

Furthermore, we looked for associations between adiponectin and immunological variables in the total study sample using multivariable linear regression analyses (Table 3). After adjustment for sex, IGT, and type 2 diabetes as major modulators of adiponectin levels, only eotaxin was significantly associated with adiponectin (P = 0.015), unlike any of the other immunological markers (P = 0.123 for TNF-α). In a second model, adjusting for sex, IGT, and type 2 diabetes, as well as for other variables that were associated with either adiponectin or immune marker concentrations (age, BMI, hypertension, total cholesterol, HDL cholesterol, and uric acid), the positive correlation between adiponectin and eotaxin persisted (P = 0.004), whereas all other associations remained nonsignificant (Table 3).

CONCLUSIONS

In the present study, serum adiponectin levels in general were not associated with a wide panel of immunological markers that were chosen to represent different aspects of immunity. In particular, there was no statistically significant association between adiponectin and proinflammatory immune mediators such as C-reactive protein (CRP), interleukin (IL)-6, and IL-8 in both univariable and multivariable analyses, although an inverse association could be assumed if adiponectin were a potent anti-inflammatory protein. However, there was some evidence for positive correlations of TNF-α and eotaxin with adiponectin.

TNF-α was associated with adiponectin levels in the age- and sex-adjusted model in individuals with NGT but not in the IGT/type 2 diabetic group or in the multivariable models. Most studies on the relationship between both proteins showed mutually antagonistic effects: adiponectin suppresses TNF-α expression or TNF-α–induced gene expression (9), while TNF-α reduces the expression of adiponectin (30). In addition, adipose tissue mRNA levels and plasma concentrations of TNF-α and adiponectin have been reported to be negatively correlated in nondiabetic subjects (14). Given the fact that some studies also found a somewhat surprising upregulation of TNF-α by adiponectin in their in vitro systems (31,32), the biological relevance of our finding remains to be elucidated.

Regarding the relation of eotaxin with adiponectin, there were positive but nonsignificant trends in the first analysis, which stratified the sample into subjects with NGT and IGT/type 2 diabetes. Both the sex- and glucose tolerance status–adjusted model and the fully adjusted model (also including age, BMI, hypertension, total cholesterol, HDL cholesterol, and uric acid) demonstrated a positive association between eotaxin and adiponection in the total study population. Eotaxin represents T helper 2 activity and is mostly expressed during allergic reactions (33). The authors are not aware of studies demonstrating a link between eotaxin and adiponectin expression. Eotaxin is also released from adipose tissue, but unlike adiponectin, stromal-vascular cells appear to be the major source of this chemokine (34). Since several recent studies did not find significant associations between eotaxin and type 2 diabetes (28) or cardiovascular disease (35), the meaning of our finding is difficult to assess.

To control the biological relevance of the adiponectin dataset, we performed additional analyses in the same study population and found, as expected, that adiponectin levels were significantly inversely correlated with variables that define the metabolic syndrome and strongly positively correlated with HDL cholesterol, another component of the metabolic syndrome. In the unadjusted analysis, we also found the previously described modulation of adiponectin levels by factors such as age, sex, glucose tolerance status, BMI, waist circumference, and WHR. The associations with indexes of obesity were considerably attenuated by adjusting for age and sex and by adjusting for glucose tolerance status. This finding may be attributable to our study sample, which exhibited a relatively small BMI or WHR range, since most of the elderly study participants were overweight or obese.

The overall lack of interdependence of adiponectin and inflammation was surprising since both hypoadiponectinemia and elevated leukocyte counts, as well as increased levels of CRP, serum amyloid A, fibrinogen, IL-6, sIL-6R, TNF-α (data for subgroup in 27), IL-8, RANTES (regulated on activation, normal T-cell expressed and secreted) (28), and macrophage-migration inhibitory factor (29) were associated with IGT and/or type 2 diabetes in the KORA S4 participants. Previous studies consistently demonstrated an inverse correlation between adiponectin and CRP (albeit weaker than the correlation between adiponectin and metabolic markers), which was not significant in our study population. However, there was some controversy about the association of adiponectin with IL-6, TNF-α, and fibrinogen (15–17,19–22,36), and other studies also reported no evidence for an association with sTNFR60, sTNFR80, serum amyloid A, and IL-8 (15,21–24). Some of these discrepancies may be explained by the widely different characteristics of the study populations regarding age, sex, glucose tolerance status, or degree of obesity; therefore, results from univariable analyses cannot be compared directly. Most analyses published thus far did not adjust for potential confounders, so data from multivariable regression, as performed in our study, are not currently available. There is some evidence that metabolic factors could have a major impact on the association of adiponectin with immune mediators since two studies reported that the inverse association between adiponectin and IL-6 (16) and CRP (18) disappeared after adjusting for obesity-related variables or for sex and BMI only, respectively.

In vitro approaches to characterize adiponectin functions suggested that adiponectin exerts anti-inflammatory properties in the interaction with leukocytes, endothelial cells, and adipocytes (8,9,10,11,12). These studies used a recombinant low–molecular weight form of adiponectin, which was derived from Escherichia coli and, thus, lacked posttranslational modifications and multimerization sites, whereas endogenous human adiponectin consists largely of middle–and high–molecular weight forms (37). This difference in adiponectin structure may explain the discrepancy of our in vivo data with previous in vitro studies. Given the frequent anatomical vicinity of adipocytes, lymph nodes, and blood vessels, it is conceivable that these in vitro data on the regulation of cytokines and chemokines by adiponectin represent paracrine effects that are not reflected by systemic levels in the circulation. Clinical studies to investigate the anti-inflammatory, antidiabetic, and antiatherogenic potency of the different molecular-weight forms of endogenous adiponectin would be highly desirable.

The cross-sectional design of the current study allows for the analysis of associations between systemic concentrations of adiponectin and immunological variables but not for causal investigation of the impact of adiponectin levels on disease risk or how an increase or decrease of adiponectin directly affects the markers of interest and vice versa. In addition, the selection of the study participants for this case-control study led to an overrepresentation of subjects with IGT and type 2 diabetes.

However, there are several strengths of the study that are relevant for the investigation of possible interdependence of adiponectin and immune markers. First, we used a well-characterized study population with extensive phenotype data covering many different aspects of systemic immunity and metabolism and allowing for multivariable analyses. Second, the oral glucose tolerance tests performed during the survey identified individuals with IGT and type 2 diabetes; therefore, the glucose tolerance status, as a potential confounder of relationships between adiponectin and immune markers, could be taken into account.

In conclusion, although adiponection levels exhibited strong negative correlations with several factors defining metabolic syndrome and a strong positive association with HDL cholesterol, there was no evidence for associations of adiponectin with a wide range of immunological parameters except eotaxin. We therefore conclude that hypoadiponectinemia and low-grade inflammation are independent and distinct factors and not just “two sides of the same coin.”