TNF-α Antagonism with Etanercept Decreases Glucose and Increases the Proportion of High Molecular Weight Adiponectin in Obese Subjects with Features of the Metabolic Syndrome

Abstract

Context and Objective: Obesity is associated with activation of the TNF-α system, increased inflammatory markers, and insulin resistance. Although studies in rodents suggest that attenuation of TNF activity improves glucose homeostasis, the effect of prolonged inhibition of TNF-α with etanercept on inflammation and glucose homeostasis in a human model of obesity is not known.

Design and Participants: Forty obese subjects with features of metabolic syndrome were randomized to etanercept or placebo, 50 mg twice weekly for 3 months, followed by 50 mg once weekly for 3 months.

Outcome Measures: Subjects underwent oral glucose tolerance testing and measurement of serum inflammatory biomarkers and adipokines. Subcutaneous fat biopsy was performed in a subset for measurement of adipokine and TNF-α mRNA expression.

Results: Visceral adiposity was significantly associated with serum concentrations of TNF receptor 1 (TNFR1), TNFR2, and vascular cell adhesion molecule-1 and adipose tissue expression of TNF-α and SOCS-3 (all P < 0.05). Insulin resistance as assessed by homeostasis model assessment was significantly associated with TNFR1, C-reactive protein, IL-6, and soluble intracellular adhesion molecule-1 (sICAM-1) (all P < 0.05). Etanercept significantly improved fasting glucose (treatment effect vs. placebo over 6 months, −10.8 ± 4.4%, P = 0.02). Etanercept also increased the ratio of high molecular weight adiponectin to total adiponectin (+22.1 ± 9.2% vs. placebo, P = 0.02), and decreased levels of sICAM-1 (−11 ± 2% vs. placebo, P < 0.0001). In contrast, body composition, lipids, C-reactive protein, and IL-6 were unchanged after 6 months.

Conclusions: Prolonged therapy with etanercept improved fasting glucose, increased the ratio of high molecular weight to total adiponectin, and decreased sICAM-1 in obese subjects with abnormal glucose homeostasis and significant subclinical inflammation.

Etanercept treatment decreases fasting glucose and increases the ratio of HMW:total adiponectin in obesity, without any effect on body composition or lipids.

Subclinical inflammation is mediated in part by activation of the TNF-α system and may contribute to the development of insulin resistance in obesity. Circulating TNF-α levels and TNF-α expression in adipocytes are increased severalfold in obese compared with lean individuals, and TNF-α decreases with weight loss (1,2,3). Data regarding the effects of TNF-α modulation on glucose regulation in humans are limited. Earlier studies using a single dose of TNF-α antagonists showed no effect on glucose homeostasis (4,5). More recently, however, TNF-α antagonism in patients with rheumatoid arthritis or psoriasis has demonstrated modest improvements in glucose homeostasis (6,7,8). To further investigate the role of TNF-α on glucose and inflammatory parameters in obesity, we investigated the effects of immunoneutralization of TNF-α with etanercept for 6 months in this population with abnormal glucose homeostasis and subclinical inflammation. The current study, using a longer duration of treatment and higher initial dosing of etanercept, extends our observations from a previous short-term study (9). With longer-term dosing, we found that TNF-α antagonism with etanercept reduced fasting glucose, increased the ratio of high molecular weight (HMW) adiponectin to total adiponectin, and reduced soluble intracellular adhesion molecule-1 (sICAM).

Subjects and Methods

Forty men and women with obesity and metabolic syndrome (10) were recruited from December 2006 to March 2009. Written, informed consent was obtained from each subject. The study was approved by the Massachusetts General Hospital and Massachusetts Institute of Technology Institutional Review Boards. Inclusion criteria included age 18–60 yr, body mass index (BMI) higher than 30 kg/m², and metabolic syndrome, defined using modified World Health Organization criteria [either fasting insulin ≥ 10 μU/ml or fasting glucose 110–125 mg/dl and at least one of the following: systolic blood pressure ≥ 140 mmHg, diastolic blood pressure ≥ 90 mmHg, triglyceride > 150 mg/dl, or high-density lipoprotein < 35 mg/dl (males) or < 39 mg/dl (females) (10)]. Exclusion criteria included hemoglobin less than 11 g/dl, creatinine higher than 1.5 mg/dl, serious chronic or recurrent infectious disease, diabetes mellitus, inflammatory or autoimmune conditions, known cardiovascular disease, immunosuppressant use, statin use, history of malignancy or demyelinating disorder, pregnancy, and initiation of niacin, antihypertensives, or fibrates within 6 wk before baseline.

Study design and methods

This was a randomized, placebo-controlled, double-blind, 6-month intervention. Etanercept 50 mg vs. identical placebo was given twice weekly for the first 3 months, followed by 50 mg once weekly vs. placebo for the final 3 months. After screening visit, eligible subjects returned after a 12-h fast for a baseline visit that included blood sampling, standard 2-h 75 g oral glucose tolerance test, single-slice abdominal computed tomography scan (11), echocardiogram, assessment of peripheral artery tonometry [Endo-PAT 2000 (12)], and optional abdominal sc fat biopsy. Subjects returned for safety evaluations at 2 wk and at months 1, 2, 3, 4, and 5 for complete blood count and serum inflammatory cytokines and adhesion markers. At the 3-month visit, subjects also underwent echocardiogram. At the end of the study, subjects returned for a 6-month visit identical to the baseline visit.

Glucose and lipids were measured using standard methodology. Insulin was measured by a paramagnetic-particle chemiluminescence immunoassay (Beckman Access Immunoassay; Beckman, Coulter, Inc., Brea, CA). High-sensitivity C-reactive protein (hsCRP) was measured using a kit from Genzyme diagnostics (Cambridge, MA). Two subjects (one placebo and one etanercept) with baseline hsCRP more than 4 sd above the mean (>30 mg/liter) were excluded from hsCRP analysis as outliers. The following were measured in plasma by ELISA: TNF-α (IBL); TNF receptor 1 (TNFR1) (R&D Systems, Minneapolis, MN), TNFR2 (R&D Systems), IL-6 (Biosource, Camarillo, CA), adiponectin (R&D Systems), HMW adiponectin (Alpco, Salem, NH), sICAM-1 (R&D Systems), vascular cell adhesion molecule-1 (VCAM-1) (R&D Systems), and resistin (R&D Systems).

Eighteen subjects agreed to elective fat biopsy. Subcutaneous abdominal fat was obtained using 4-mm punch biopsy. Samples were immediately frozen for future analysis. RNA was extracted by mechanical homogenization using TRIzol (Invitrogen, Carlsbad, CA). Cleared lysates were mixed with 1 vol of 70% ethanol and applied to RNeasy mini columns for RNA isolation according to manufacturer’s protocol (QIAGEN, Valencia, CA). RNA was treated with deoxyribonuclease I according to manufacturer’s protocol (Invitrogen). The RNA was reverse transcribed using Superscript First Strand (Invitrogen). cDNA was treated with ribonuclease H according to the manufacturer’s protocol (Invitrogen). cDNA was amplified using TaqMan gene expression assays for adiponectin, TNF-α, and SOCS-3. Variability of cDNA levels were normalized by subtracting the cycle threshold (C_t) value of 18S from the target C_t value. Relative levels were quantified by measuring the 2^ΔCt. Expression data that were not normally distributed was logarithmically transformed using log₁₀.

Statistical analysis

Baseline hsCRP, TNF-α, IL-6, adiponectin, HMW adiponectin, sICAM-1, VCAM-1, and resistin were not normally distributed and were log transformed. Continuous baseline variables were compared using Student’s t test, and categorical variables were compared using Pearson χ². To compare treatment effects (etanercept vs. placebo) over 6 months on oral glucose tolerance test parameters, lipid, and body composition, ANCOVA was used in modeling that included baseline values, age, and race. For hsCRP, exploratory modeling showed that change over the initial 3 months (the period of higher dosing) was greater than change over 6 months. Therefore, separate comparisons are reported at 3 and 6 months. To analyze variables measured at multiple time points, within-subject percent changes at each time point were calculated, and then repeated-measures ANCOVA controlling for age and race was performed. One subject randomized to etanercept was subsequently prescribed metformin and was excluded from analyses of glucose. SAS was used for repeated-measures analysis, and other analyses were performed using SAS JMP. Data are presented as mean ± sem unless otherwise specified.

Results

Of 40 subjects who underwent baseline assessment, five subjects (two placebo and three etanercept) withdrew for personal reasons before the 3-month visit. In addition, one subject developed a side effect of leg tingling after study drug injections and was asked to withdraw from the study. Thirty-four patients (85%) completed the study.

Sixteen subjects were randomized to etanercept and 24 to placebo. The groups were equally matched with respect to sex, but the etanercept group was younger than the placebo group (41 ± 2 vs. 47 ± 2 yr, P = 0.02). There were no significant differences in other baseline demographic or clinical parameters (Supplemental Table 1, published on The Endocrine Society’s Journals Online web site at http://jcem.endojournals.org). Baseline hsCRP concentrations were increased (mean, 6.3 ± 1.0 mg/liter). Baseline relationships between inflammatory biomarkers, body composition, and measures of glucose homeostasis, including adipose gene expression, are shown in Supplemental Table 2. Adipose TNF-α and SOCS-3 expression were significantly associated with visceral adiposity but not with markers of insulin resistance.

Fasting glucose decreased significantly in the etanercept group compared with the placebo group, with a treatment effect of −10.8 ± 4.4%, controlling for age and race, in the ANCOVA model (P = 0.02, Fig. 1A). Using unadjusted group means comparing change at 6 months, this amounted to a 13 mg/dl difference in fasting glucose between the groups, with glucose decreasing 8 mg/dl in etanercept and increasing 5 mg/dl in the placebo group. Two-hour glucose increased in both groups over the study, although there was a trend toward an effect of etanercept to attenuate this increase (treatment effect, −11 ± 6 mg/dl, P = 0.08). Etanercept did not significantly affect fasting insulin (treatment effect, 0.2 ± 1.2 μU/ml, P = 0.84), 2-h insulin (treatment effect, −6.7 ± 4.8 μU/ml, P = 0.17), or homeostasis model assessment for insulin resistance (treatment effect, 0.05 ± 0.32, P = 0.87).

Figure 1.

Effect of etanercept on glucose, hsCRP, and adiponectin: changes in etanercept (black bars) and placebo (white bars) groups. Error bars are se. A, Percent change in fasting glucose at 3 and 6 months; P = 0.02 for etanercept vs. placebo by repeated-measures ANCOVA. B, Change in log₁₀ hsCRP at 3 and 6 months. *, P = 0.03 for change in etanercept vs. placebo at 3 months controlling for age and race. Change between groups is not significantly different at 6 months. C, Aggregate percent change in total adiponectin, HMW adiponectin, and ratio of HMW to total adiponectin over 6 months. *, P < 0.05 for etanercept vs. placebo by repeated-measures ANCOVA.

At baseline, five subjects in the etanercept group and four in the placebo group had fasting glucose of 100 mg/dl or higher. Of these, three in the etanercept group vs. one in the placebo group had normalized fasting glucose by 6 months. Of subjects with normal fasting glucose at baseline, zero subjects in the etanercept group vs. two in the placebo group developed fasting glucose of 100 mg/dl or higher by 6 months.

As shown in Supplemental Table 3, etanercept did not affect body composition or lipid parameters. There were no significant changes in measurements of dietary intake, including daily caloric, protein, fat, and carbohydrate intake (data not shown). There was no change in endothelial function as measured by peripheral artery tonometry (reactive hyperemia index treatment effect, −0.07 ± 0.10, P = 0.46).

hsCRP decreased significantly in the etanercept group compared with placebo at 3 months (treatment effect, −0.13 ± 0.06 log₁₀ mg/liter, P = 0.03), but the decrease at 6 months was not statistically significant (treatment effect, −0.08 ± 0.06 log₁₀ mg/liter, P = 0.20, Fig. 1B). Table 1 shows percent change at 3 and 6 months for other inflammatory markers. As expected, due to its sequestration of TNF-α in serum and its cross-reactivity with the TNFR2 assay, etanercept increased measurements of TNF-α and TNFR2 (Table 1). Etanercept significantly increased the ratio of HMW to total adiponectin (treatment effect, +22 ± 9% vs. placebo, P = 0.02, Fig. 1C). sICAM-1 significantly decreased with etanercept (treatment effect, −11 ± 2% over 6 months, P < 0.0001, Table 1), whereas there was no change in VCAM-1.

Table 1.

Baseline values and percent change in serum inflammatory markers

	Baseline values (mean± sem)		3 months		6 months
	Etanercept	Placebo	Treatment effect vs. placebo (% change)	P value	Treatment effect vs. placebo (% change)	P value
TNF-α (pg/ml)	3.4 ± 1.8	0.9 ± 0.5	+55 ± 25	0.08	+51 ± 17	0.03
TNFR2 (pg/ml)	2366 ± 111	2499 ± 125	+103 ± 6	<0.0001	+98 ± 4	<0.0001
IL-6 (pg/ml)	10.5 ± 1.6a	7.2 ± 0.7	−4 ± 22	0.85	−1 ± 14	0.92
Total adiponectin (ng/ml)	3306 ± 265	4376 ± 601	−12 ± 5	0.01	−12 ± 4	0.004
HMW adiponectin (ng/ml)	761 ± 145	1418 ± 257	+3 ± 17	0.85	+13 ± 11	0.26
HMW/total adiponectin	0.31 ± 0.13	0.38 ± 0.10	+13 ± 14	0.35	+22 ± 9	0.02
sICAM-1 (ng/ml)	231 ± 26	211 ± 13	−12 ± 2	<0.0001	−11 ± 2	<0.0001
VCAM-1 (μm)	494 ± 28	562 ± 48	−1 ± 6	0.84	−2.7 ± 4	0.51
Resistin (ng/ml)	8.8 ± 1.0	7.4 ± 0.5	−21 ± 18	0.24	−20 ± 11	0.08

^aBaseline IL-6 significantly different between etanercept and placebo groups (P = 0.03). No other baseline variables were significantly different between etanercept and placebo groups. 

In sc adipose tissue, there was a strong association at baseline between TNF-α expression and SOCS-3 expression (r = 0.92; P < 0.0001) but no association between TNF-α expression and adiponectin expression (P = 0.48). Etanercept did not significantly affect expression of TNF-α mRNA (P = 0.31), adiponectin mRNA (P = 0.89), or SOCS-3 mRNA (P = 0.54).

Discussion

In this study, we demonstrate that treatment with the TNF-α antagonist etanercept decreases fasting glucose, increases the ratio of HMW to total adiponectin, and decreases sICAM in obese subjects. Our data stand in contrast to earlier trials reporting no effect of TNF-α antagonism on glucose homeostasis in individuals with diabetes or insulin resistance (4,5,9), but these studies were of short duration. The relatively higher dose and longer duration of treatment in this study may be needed to demonstrate improvements in glucose in obese patients. The improvements in fasting glucose and adiponectin after etanercept occurred in the absence of changes in body composition or lipid parameters, suggesting a direct effect on TNF-α on these parameters. Homeostasis model assessment for insulin resistance did not change, and we were unable to perform euglycemic clamps; thus, it remains unclear whether etanercept directly improved insulin sensitivity.

TNF-α may affect glucose homeostasis by altering levels of adiponectin. In vitro studies have shown that TNF-α decreases expression of adiponectin mRNA in adipocytes (13,14). In the current study, we did not observe effects of etanercept on adiponectin mRNA expression in sc adipose tissue. We did observe changes in relative serum concentrations of total and HMW adiponectin, however, suggesting a possible effect of etanercept on the posttranslational assembly of adiponectin into multimers. This finding has not been previously reported. To our knowledge, in vitro mechanisms by which TNF-α may affect posttranslational processing of adiponectin have not been described, and further study will be required to validate this finding. TNF-α may also affect insulin sensitivity through up-regulation of SOCS-3 expression (15). Ghanim et al. (16,17) have demonstrated that SOCS-3 mRNA expression is increased in the inflammatory milieu of obesity and is associated with increased BMI and insulin resistance as well as high-fat diet. We report a very strong correlation between TNF-α expression and SOCS-3 expression in sc adipose tissue at baseline as well as an association between SOCS-3 and visceral adipose tissue, but our data do not demonstrate significant associations between SOCS-3 and BMI or measures of glucose homeostasis. We do not show an effect of etanercept on adipose SOCS-3 expression.

In this study, we demonstrate a significant effect to reduce sICAM, suggesting an effect of TNF on endothelial function, but we did not observe a clinical effect using a noninvasive test of vascular function. Additional longer-term studies are needed to investigate the effects of etanercept on glucose homeostasis, inflammatory cytokines, and vascular adhesion molecules and the mechanisms by which TNF-α may affect posttranslational processing of adiponectin in vivo. Regarding the clinical implications of our study, etanercept has potential side effects, including increased susceptibility to infection, which would prevent its widespread use in obesity. However, in individuals who require antiinflammatory therapy for an autoimmune condition and have concomitant obesity and metabolic dysregulation, etanercept may have the beneficial side effect of improving glucose homeostasis, and this is a potentially important consideration for the large numbers of patients with rheumatoid arthritis and psoriasis, many of whom are obese and have increased risk of cardiovascular disease (18,19). Overall, this study is the longest and largest to date to show effects of TNF-α blockade on a clinically relevant glucose parameter and adipocytokine and inflammatory indices. Our results support the concept that increased TNF-α in obesity is linked to impaired glucose homeostasis. However, it remains to be determined whether TNF-α is a viable target for the treatment of diabetes and other complications of obesity.