Reductions in Plasma Cytokine Levels With Weight Loss Improve Insulin Sensitivity in Overweight and Obese Postmenopausal Women

The women in this study were postmenopausal (had not menstruated for at least 1 year and had plasma follicle-stimulating hormone levels >30 mIU/ml) and were 50–70 years old, healthy, overweight, and obese (BMI >25 kg/m², range 25–41). Only women who were weight stable (<2.0-kg weight change in previous year) and sedentary (<20 min of aerobic exercise twice a week) were recruited to participate. Women were screened by medical history questionnaire, physical examination, fasting blood profile, and a graded exercise treadmill test in an attempt to exclude those with cardiovascular disease. The women underwent a 2-h 75-g oral glucose tolerance test during screening to exclude women with diabetes (23). All women were nonsmokers and showed no evidence of cancer, liver, renal, or hematological disease, or other medical disorders. The Institutional Review Board of the University of Maryland approved all methods and procedures. Each participant provided written informed consent to participate in the study.

Sixty-two women met the study inclusion/exclusion criteria and were enrolled. Fifty-eight of these women (46 Caucasians and 12 African Americans) had complete baseline data, including completion of a 3-h glucose clamp. Nineteen of these women were on hormone replacement therapy for at least 3 years before enrollment, which did not change for the duration of the study. Thirty-seven of the women studied at baseline completed 6 months of dietary- or exercise-induced weight loss (30 Caucasians and 7 African Americans, 12 of which were on hormone replacement therapy).

For 6 months, all women attended weekly weight loss classes led by a registered dietitian for instruction in the principles of a hypocaloric diet that followed the American Heart Association Step I (24) guidelines with restriction of caloric intake by 250–350 kcal/day, as previously described (13). Compliance to the weight-loss classes was ∼80%. Twenty-five women also exercised three times a week on nonconsecutive days using treadmills and cycle ergometers at >60% Vo_2max for 45 min (n = 16) or using eight exercises on pneumatic variable-resistance machines (Keiser K-300; Keiser Sports Health, Fresno, CA) and dumbbells (n = 9) (13,25).

Vo_2max was measured using a continuous treadmill test protocol that has been previously described (25). Height and weight were measured to calculate BMI as weight (kg)/height (m²). Fat mass, lean tissue mass, and bone mineral content (BMC) were determined by dual-energy X-ray absorptiometry (Model DPX-L; Lunar Radiation, Madison, WI) using the 1.3z DPX-L extended analysis program. Fat-free mass (FFM) is reported as lean tissue plus BMC. A single 5-mm computed tomography scan was taken at the L₄-L₅ region using a General Electric Hi-Light Scanner to determine relative proportions of visceral adipose tissue (VAT) area and subcutaneous adipose tissue (SAT) area (12).

All subjects were weight stabilized (<1 kg) for at least 2 weeks before metabolic testing before and after the interventions and were provided with a eucaloric diet for 2 days before the clamp by a registered dietitian to control nutrient intake, as previously described (26). All testing was performed in the morning after a 12-h overnight fast. Glucose clamps were performed 36–48 h after any exercise. Peripheral tissue sensitivity to exogenous insulin was measured using the hyperinsulinemic-euglycemic clamp technique (27). Arterialized blood was obtained from a dorsal heated hand vein (28). Basal glucose, insulin, and cytokine levels were measured in the fasted state. A 10 min, priming and continuous infusion of insulin (240 pmol · m⁻² · min⁻¹, Humulin; Eli Lilly, Indianapolis, IN) and a 20% glucose solution infusion was performed for 180 min. During the clamp, blood samples were obtained every 5 and 10 min for the determination of plasma glucose and insulin levels.

The plasma glucose and insulin levels during all clamps averaged 5.19 ± 0.05 mmol/l and 478 ± 7 pmol/l (n = 96, i.e., 58 pre- and 37 postintervention). This was 97.6 ± 0.1% of the desired goal with a coefficient of variation (CV) of 4.9 ± 0.2% in all clamps (n = 96).

Blood samples were collected in heparinized syringes and placed in prechilled test tubes containing 1.5 mg EDTA/ml blood. The blood samples were centrifuged at 4°C, and plasma was stored at −70°C until analysis. Plasma glucose was measured with the glucose oxidase method (Beckman Instruments, Fullerton, CA). Insulin was determined by radioimmunoassay (Linco, St. Louis, MO). All cytokines were measured by Quantikine enzyme-linked immunosorbent assay kits (high sensitivity for IL-6 and TNF-α) (R&D Systems, Minneapolis, MN). The inter- and intra-assay CVs were <6% for IL-6 and the soluble receptor assays and <12% for TNF-α. C-reactive protein was measured using an automated immunoanalyzer (Immulite; Diagnostics Products, Los Angeles, CA) with inter- and intra-assay CVs of 7.5 and 4.4%, respectively. Samples for glucose, insulin, and cytokines were measured in duplicate, and the average of the two values were used in the statistical analyses. The duplicate samples for the cytokines required a CV <15%, or the sample was reanalyzed.

For the hyperinsulinemic-euglycemic clamps, the mean concentration of glucose and insulin was calculated for each sample time point. The trapezoidal rule was used to calculate the integrated response over 30-min intervals from 30 to 180 min for each subject. The integrated response was divided by its time interval to compute mean concentrations. Glucose utilization (M) for 30-min intervals was calculated from the amount of glucose infused after correction for glucose equivalent space (glucose space correction). Insulin sensitivity was expressed as M/I, which represents the amount of glucose metabolized per unit of plasma insulin (I), and was calculated by dividing the glucose utilized by the insulin concentration during the last 60 min of the clamp for each subject.

The CRP, IL-6, and TNF data were not normally distributed, therefore the logarithm was used for parametric statistical analyses. Associations at baseline were tested using linear regression analyses. Differences between pre- and postintervention measures of all variables were determined using a paired t test. Univariate regression analysis was used to determine predictors of glucose utilization and insulin sensitivity. Stepwise regression analysis was used to determine whether changes in markers of inflammation were significant predictors of changes in glucose utilization. Statistical significance was set at P < 0.05 for all tests. All data were analyzed by SPSS statistical software (SPSS, Chicago, IL). All values are expressed as means ± SEM.

The baseline characteristics of the women in this study are shown in Table 1. Although all women were overweight or obese, sedentary, and of a similar age, there was wide variation in glucose utilization, insulin sensitivity, and inflammatory markers.

Pearson correlation coefficients between all of the inflammatory markers and Vo_2max, BMI, fat percentage, VAT, M, and M/I at baseline are shown in Table 2. CRP and IL-6 concentrations correlated negatively with Vo_2max, and CRP, sTNFR-1, and IL-6 concentrations correlated positively with BMI. Only IL-6 concentrations were directly related to percentage of fat, whereas both sTNFR-1 and sTNFR-2 concentrations correlated positively with VAT. Glucose utilization and insulin sensitivity were lower in women with higher CRP concentrations (Fig. 1) and tended to be lower in women with higher IL-6 levels.

We next determined whether M and M/I were independently related to CRP at baseline in these women. Since univariate analyses showed that race (M: r = −0.41 and M/I: r = −0.43; P < 0.01), BMI (M: r = −0.31 and M/I: r = −0.32; P < 0.05) and Vo_2max (M: r = 0.26 and M/I: r = 0.29; P < 0.05) were the only other variables associated with M and M/I, we used linear regression analyses to co-vary the relationship between M and M/I and CRP for these variables. The results showed that CRP concentrations remained significantly associated with M (partial r = −0.30; P = 0.03) and M/I (partial r = −0.32; P = 0.02), indicating that CRP is a predictor of both glucose utilization and insulin sensitivity independent of total and regional adiposity, race, and aerobic fitness.

Body weight, percentage body fat, fat mass, FFM, VAT, and SAT decreased (P < 0.05) (Table 3), whereas Vo_2max (in liters per minute) did not change after weight loss. Fasting plasma glucose and insulin levels decreased by 6 and 10%, respectively (P < 0.001). Glucose utilization increased by 10% (in μmol · kg⁻¹ · min⁻¹) and 16% (in μmol · kg_FFM⁻¹· min⁻¹) (P < 0.05), and insulin sensitivity increased 23% (in μmol · kg⁻¹· min⁻¹) and 16% (in μmol · kg_FFM⁻¹· min⁻¹, P < 0.05). CRP, IL-6, and sTNFR-1 concentrations decreased (P < 0.05, Table 3), but concentrations of TNF-α, sTNFR-2, and IL-6sR did not change. Changes in M and M/I were inversely related to changes in fat mass (r = −0.38 and r = −0.41, P < 0.05) and VAT (r = −0.49 and r = −0.57, P < 0.01), whereas only the change in M/I was inversely related to the loss of SAT (r = −0.48, P < 0.01).

We next examined whether the improvements in glucose metabolism were related to declines in inflammatory markers. Changes in M and M/I were inversely related to changes in CRP, sTNFR-1 and IL-6 (Table 4 and Fig. 2A and B). Changes in M and M/I were not related to changes in IL6sR, TNF-α, or sTNFR-2. The changes in total fat mass, VAT, CRP, IL-6, and sTNFR-1 were put in a stepwise regression model to determine the strongest independent predictor of changes in M and M/I with weight loss. Changes in VAT entered the model first (r = −0.49, P < 0.01), followed by changes in sTNFR-1 (cumulative r = −0.64, P < 0.01) to predict changes in M. Changes in VAT and IL-6 were both independent predictors of changes in M/I (r = −0.57 and cumulative r = −0.68, P < 0.01).

Inflammation likely plays a key role in the development of type 2 diabetes. Thus, interventions that reduce inflammation and improve insulin resistance would have important clinical implications. Our results show that in postmenopausal women, increases in glucose utilization and insulin sensitivity with weight loss are related to reductions in circulating cytokines, specifically sTNFR-1 and IL-6, and this relationship is independent of decreases in total body and visceral fat. Moreover, this occurs with a relatively modest amount of weight loss.

Concentrations of CRP, TNF-α, and IL-6, considered sensitive markers of subclinical systemic inflammation, are elevated in obesity and type 2 diabetes (1,7). Our study in postmenopausal women focuses on the relationship between markers of inflammation and insulin resistance. At baseline, CRP is the only inflammatory marker associated with insulin-stimulated glucose uptake and insulin sensitivity. This is consistent with previous reports in postmenopausal women (9) and in Pima Indians (10), as well as findings in the Insulin Resistance Atherosclerosis Study (IRAS), where CRP was associated with insulin sensitivity determined by the frequently sampled intravenous glucose tolerance test (29). Thus, it is clear that insulin resistance is linked to CRP in diverse populations.

In animal models, TNF-α induces insulin resistance by decreasing insulin receptor and IR substrate phosphorylation, while producing other modifications in the insulin-signaling cascade (31). In humans, insulin resistance is associated with increased expression of skeletal muscle TNF-α (30). Yet we and others (9) did not find a relationship between plasma TNF-α and glucose utilization, suggesting that local rather than systemic TNF-α may be more important to insulin resistance.

Neither glucose utilization nor insulin sensitivity was related to circulating concentrations of sTNFR-1 or -2 in these women before intervention. This is in contrast to a report (8) in younger obese women where glucose utilization was associated with sTNFR-2 levels. In addition, sTNFR2 decreased with a 12-week exercise training regimen, and this decline was related to an increase in insulin sensitivity (8). However, in our study, the decrease in sTNFR-1, and not sTNFR-2, concentration was an independent predictor of the improvement in glucose utilization with weight loss. Although the biological functions of the TNF receptors are not completely understood, TNFR-1 is involved in proinflammatory and cytotoxic responses, whereas TNFR-2 may merely vary the response produced by TNFR-1 (32). In addition to TNF-α, other cytokines can bind to these receptors, which are elevated in inflammatory states, even in the absence of change in TNF-α concentrations (33). Accordingly, our finding that changes in inflammation are associated with improvements in insulin-stimulated glucose uptake raises the possibility that weight loss induced by diet and exercise training may provide the pathway to concurrently reduce inflammation and insulin resistance.

IL-6 is another cytokine produced in adipose tissue and released into the systemic circulation. Although Vozarova et al. (10) found that glucose utilization was negatively related to plasma IL-6 concentrations in Pima Indians, we did not see a significant relationship between these parameters in obese postmenopausal women before treatment. This is similar to other reports in postmenopausal women (9) and obese subjects with and without diabetes (34) when glucose utilization is determined at low physiologic insulin concentrations. However, when IL-6 is measured in adipose tissue and glucose utilization is determined at maximal insulin concentrations, IL-6 concentrations are inversely correlated with maximal insulin responsiveness in vivo and in vitro (34). This suggests that measurement of local IL-6 concentrations may be more sensitive than that measured in the systemic circulation. Yet, we did find that change in IL-6 concentration is an independent predictor of the change in insulin sensitivity with weight loss, suggesting that systemic IL-6 may be an important mechanistic mediator of improvements in glucose homeostasis.

We conclude that improvements in glucose metabolism with modest weight loss in overweight and obese postmenopausal women are associated with decreases in cytokine concentrations, suggesting that a reduction in inflammation is a potential mechanism that mediates changes in insulin resistance. Future studies should be directed at the examination of either skeletal muscle or adipose tissue cytokine production and whether modulation of local inflammatory markers with weight loss impacts changes in systemic insulin resistance.

This study was supported by funds from National Institutes of Health Grants K01-AG00747 and R01-AG19310 (to A.S.R.) and R29-A614066 and R01-AG/DK20583 (to B.J.N.) and the Department of Veterans Affairs, Geriatrics Research, Education, and Clinical Center (GRECC) in Baltimore, Maryland.

Our appreciation is extended to the women who participated in this study. We are grateful to the nurses, laboratory technicians, exercise physiologists, and Drs. Dora Berman and Andrew P. Goldberg of the Division of Gerontology and GRECC for their assistance in this project.