We prospectively studied patients with RA who were started on anti-TNF agents due to high disease activity, defined by a disease activity score in 28 joints (DAS28) > 5.1 despite treatment with disease-modifying anti-rheumatic drugs (DMARDs) 23 . Patients fulfilled the American College of Rheumatology 1987 classification criteria for RA 24 . Patients aged < 18 years or with history of DM were excluded. Anti-TNF agents (infliximab, adalimumab or etanercept) were given at approved standard doses. The dosage of background DMARDs and glucocorticoids, as well as of anti-hypertensive and lipid-lowering treatments remained stable during the study period. Detailed medical and drug history was obtained, and the use of tobacco was recorded. Disease activity (DAS28) was assessed at baseline and after 12 weeks of treatment 25 . Coronary heart disease was defined as presence of any of the following: angina pectoris, myocardial infarction or coronary revascularization procedures (bypass grafting, percutaneous coronary intervention). As a control group, we consecutively enrolled seven patients with RA who were started on abatacept (CTLA4.Ig), another class of biologic therapy which targets T-cell costimulation rather than TNF, due to high disease activity (DAS28 > 5.1). All patients signed informed consent forms, and the study protocol was approved by the University Hospital of Heraklion Ethics Committee (13/12/99).

Blood was drawn after 12-hour overnight fasting. Serum concentrations of total cholesterol (TC), high-density lipoprotein cholesterol (HDL-C), and triglycerides (TG) were measured using an automated chemistry analyzer (Olympus AU-600, Tokyo, Japan). Low-density lipoprotein cholesterol (LDL-C) was calculated according to the Friedewald formula except for samples with serum TG > 400 mg/dl (4.5 mmol/L). Apolipoprotein (apo)-A, apo-B and lipoprotein a (Lpa) were analyzed by immunonephelometry (BN II analyzer, Eschborn, Germany). Plasma glucose concentration was determined by standard enzymatic colorimetric assay (Roche Diagnostics, Mannheim, Germany). C-reactive protein (CRP) was measured by immune nephelometry (BN II analyzer, Eschborn, Germany) and erythrocyte sedimentation rate (ESR) by an automated analyzer (Ves-matic 20, Florence, Italy). Serum insulin was determined by chemiluminescent microparticle immunoassay (ARCHITECT Insulin Reagent Kit, 8K41, Abbott Diagnostics, Lake Forest, Illinois, USA)

While the hyperinsulinemic euglycemic clamp technique is the gold standard for evaluating insulin sensitivity, the Homeostasis Model Assessment (HOMA) for insulin resistance and the Quantitative Insulin Sensitivity Check Index (QUICKI) are widely used as noninvasive surrogate markers of insulin resistance and sensitivity, respectively 26 27 . Pancreatic beta cell function was assessed by the HOMA for beta cell function (HOMA-B) 28 . We used the National Cholesterol Education Program (NCEP) Adult Treatment Panel (ATP) III criteria for metabolic syndrome 29 .

PBMCs were isolated by Ficoll-Histopaque (Sigma-Aldrich, St. Louis, MO, USA) density-gradient centrifugation of heparinized venous blood. PBMCs were lysed with RIPA buffer containing 50 mM Tris-HCl, pH 7.2, 150 mM sodium chloride, 1% NP-40, 12 mM sodium deoxycholate, 3 mM sodium dodecyl sulfate, 4 mM sodium azide, 0.57 mM phenylmethysulfonyl fluoride and complete protease inhibitor cocktail. Protein concentrations were determined by Bradford protein assay (Bio-Rad laboratories, Inc, Hercules, CA, USA).

Proteins were separated by 8% sodium dodecyl sulphate-polyacrylamide gel electrophoresis (SDS-PAGE) and transferred to nitrocellulose membrane (Whatman, Schwerte, Germany). Membranes were probed overnight with anti-phospho-IRS1-Ser³¹² (Upstate, Lake Placid, NY, USA) and anti-phospho-Akt-Ser⁴⁷³ (Cell Signalling, Frankfurt, Germany) monoclonal antibodies. Blots were also probed with anti-β-actin (Santa Cruz, CA, USA) to control for total protein content. The secondary antibodies horseradish peroxidase (HRP)-conjugated anti-mouse IgG and anti-rabbit IgG were used (Santa Cruz, CA, USA). Enhanced chemiluminescence (ECL) from Thermo Scientific (Vantaa, Finland) was used for detection, and densitometry analysis was performed with Image J program.

Results are expressed as the mean and 95% confidence interval (95% CI) or as proportions. Parameters showing abnormal distribution were logarithmically transformed. In a post-hoc analysis, we used the median HOMA-IR value (3.18) to define high insulin resistance, and we compared the recorded variables in RA patients with high versus low insulin resistance using the independent samples t-test (continuous variables) or the chi-squared test (categorical variables). The Pearson's test was used for bivariate correlation analysis. Logistic regression (backward selection) was further performed to calculate odds ratios (ORs) for high versus low HOMA-IR according to demographic and lifestyle (gender, age, tobacco use), metabolic (obesity, hypertension, use of statins, metabolic syndrome), and clinical parameters, including the DAS28, health assessment questionnaire (HAQ), CRP, and RA treatment modality (glucocorticoids, methotrexate). Changes in recorded parameters between baseline and the end of the 12-week study period were analyzed by paired t-test or the Wilcoxon signed ranks test as appropriate. To address whether the effect of anti-TNF treatment on insulin resistance is confounded by changes in disease activity, we performed repeated measures analysis of variance (ANOVA) entering HOMA-IR values at baseline and at 12 weeks after treatment as dependent variables (within-subject variables), the insulin resistance group (high versus low baseline HOMA-IR) as grouping factor (between-subject variable), and the change in DAS28 score as a covariate. P values (two-tailed) < 0.05 were considered statistically significant. SPSS software version 18.0 (SPSS, Inc) was used for the statistical calculations.

We prospectively studied 61 patients with RA (43 women) of mean age 60 years and mean disease duration 7.2 years (Table 1). All patients had active disease with mean DAS28 score 5.8 and they were started on one of the following anti-TNF agents: infliximab (n = 49), adalimumab (n = 11), or etanercept (n = 1). Fifty six patients (92%) were on single (n = 36) or combination (n = 20) therapy with DMARDs, including methotrexate (n = 42), leflunomide (n = 15), and hydroxychloroquine (n = 17). Thirteen patients were on glucocorticoids. Patients treated with abatacept had comparable demographic and clinical characteristics and metabolic profiles to patients treated with anti-TNFagents (data not shown).

We used the median HOMA-IR at baseline to assign patients to the low (HOMA-IR ≤ 3.18, n = 31) and high (HOMA-IR > 3.18, n = 30) insulin resistance group. Patients in the high insulin resistance group had higher mean disease activity (P = 0.011), fasting serum insulin (P < 0.001), plasma glucose (P < 0.001), triglycerides (P = 0.015), and systolic blood pressure levels (P = 0.024) than patients with low insulin resistance (Table 1). Metabolic syndrome was more frequent in RA patients with high insulin resistance (63% versus 19% in those with low insulin resistance, P = 0.001). In accordance with previous reports 30 , we found significantly increased pancreatic beta cell function, assessed by the HOMA-B index, in RA patients with high insulin resistance.

By using a cut-off of HOMA-IR > 2.0 31 , 74% (n = 45) of RA patients were classified as having insulin resistance. Patients with HOMA-IR > 2.0 at baseline had higher serum insulin (P < 0.001), HOMA-B (P < 0.001), and triglycerides (P = 0.014), and lower QUICKI (P < 0.001) than those with HOMA-IR ≤ 2.0. RA patients with insulin resistance had a trend for higher DAS28 (P = 0.105) and significantly higher HAQ (P = 0.037). Fasting plasma glucose levels were not significantly different between the two groups, although 61% in patients with insulin resistance versus 6% in those without insulin resistance had metabolic syndrome (P = 0.001) (data not shown).

HOMA-IR correlated with DAS28 (r = 0.31, P = 0.016) and fasting triglycerides (r = 0.38, P = 0.003), and HOMA-B correlated with body mass index (BMI) (r = 0.28, P = 0.035) (data not shown). In multivariate regression analysis, presence of the metabolic syndrome and DAS28 score > 6.86 (75^th percentile) were independent predictors for high insulin resistance with ORs of 10.8 (95% CI 2.8 to 42.6, P = 0.001) and 6.4 (95% CI 1.2 to 32.9, P = 0.027), respectively. BMI > 30 kg/m² was associated with an OR of 3.8 (95% CI 1.1 to 12.4, P = 0.030) for increased pancreatic beta cell function.

Patients with high baseline insulin resistance demonstrated significant improvement in DAS28 (P = 0.013), HAQ (P = 0.012), and CRP (P = 0.007) after 12 weeks of treatment with anti-TNF agents (Table 2). A similar trend was observed in patients with lower baseline insulin resistance, yet only reduction in HAQ was statistically significant (P = 0.018). Anti-TNF therapy resulted in significant improvement in HOMA-IR (P < 0.001), QUICKI (P < 0.001), and HOMA-B (P = 0.001) only in RA patients with high baseline insulin resistance (Table 2). In these patients, significant decreases in fasting plasma glucose (P = 0.016), serum insulin (P < 0.001), and serum triglycerides (P = 0.039), and increase in serum HDL-C (P = 0.022) levels were observed.

We repeated the analysis in RA patients with baseline HOMA-IR ≤ 2.0 (n = 16) or > 2.0 (n = 45). Anti-TNF treatment reduced DAS28 (P = 0.007), HAQ (P = 0.002), and CRP (P = 0.002) in patients with insulin resistance but not in those with HOMA-IR ≤ 2.0 (data not shown). TNF blockade resulted in significant reduction in serum insulin (P < 0.001) and improvements in HOMA-IR (P = 0.001), HOMA-B (P = 0.001), and QUICKI (P = 0.001) in patients with HOMA-IR > 2.0. In contrast, there were no significant changes in serum triglycerides (P = 0.058), HDL-C (P = 0.101), homocysteine (P = 0.312) and blood glucose (P = 0.166). These findings suggest that the metabolic effects of anti-TNF treatment may be more pronounced in RA patients with very high insulin resistance levels compared to those with low-level or no insulin resistance.

We next examined whether changes in insulin resistance/sensitivity and pancreatic beta cell function correlated with changes in RA disease activity. Significant improvements in the HOMA-IR, QUICKI, and HOMA-B indices were found in RA patients with high baseline insulin resistance irrespective of their clinical response according to the European League Against Rheumatism (EULAR) criteria (data not shown). In addition, repeated measures ANOVA revealed a significant effect of anti-TNF treatment (F test = 11.89, P = 0.001) and of the interaction of anti-TNF treatment with baseline insulin resistance group (F test = 24.68, P < 0.001) on HOMA-IR. Conversely, the interaction of anti-TNF treatment with the change in DAS28 had no significant impact on HOMA-IR (F test = 0.06, P = 0.810), suggesting that the effect of anti-TNF treatment on insulin resistance is not confounded by improvement in disease activity.

TNF may compromise insulin signaling by induction of serine 312 (Ser³¹²) phosphorylation of IRS-1 7 32 . Based on our findings of improved insulin resistance in patients with active RA following treatment with anti-TNF agents, we examined whether this effect is associated with altered IRS-1 phosphorylation. We used PBMCs, which provide a useful tool to determine alterations in the insulin signaling cascade in vivo as they include all of the signaling proteins present in insulin target tissues 33 34 . We purified total protein extracts from PBMCs of seven randomly selected RA patients at weeks 0 and 12 of anti-TNF therapy, and we quantified IRS-1 Ser³¹² phosphorylation by western blot. At study entry, all patients had active disease with median DAS28 7.4 (range 5.6 to 9.0), and high insulin resistance with median HOMA-IR 7.6 (range 3.1 to 10.3). At week 12, all but one patient demonstrated reduction in IRS-1 phosphorylation (Figure 1A). Phosphorylated IRS-1 normalized for β-actin protein expression was significantly reduced from 0.96 (arbitrary units) at week 0 to 0.57 at week 12 (Wilcoxon signed ranks test, P = 0.043) (Figure 1B). Both DAS28 and HOMA-IR improved over the study period (Figures 1C, D). We found no correlation between longitudinal changes in HOMA-IR and IRS-1 Ser³¹² phosphorylation levels.

IRS-1 mediates its metabolic effects through the PI-3K pathway which results in phosphorylation of AKT and other downstream molecules 32 . TNF-induced insulin resistance involves inhibition of AKT through a ceramide-activated protein-phosphatase (PP) 2A 35 . Individuals with insulin resistance display reduced insulin stimulation of the PI-3K pathway 5 , and insulin-stimulated AKT kinase activity is reduced in patients with T2DM 36 37 . Therefore, we measured phosphorylated AKT levels in protein extracts from PBMCs of RA patients at weeks 0 and 12 of anti-TNF therapy. Phosphorylated AKT levels, normalized for β-actin, increased in all seven RA patients (Figure 1E) from a median 0.24 at week 0 to 0.94 at week 12 (P = 0.001) (Figure 1F). We found no correlation between longitudinal changes in HOMA-IR or IRS-1 Ser³¹² phosphorylation and AKT phosphorylation levels.

To assess the specificity of the effect of TNF blockade on insulin resistance, we prospectively studied seven RA patients (median age 63 years, median disease duration 13 years) with high disease activity (median DAS28 5.92) treated with abatacept (CTLA4.Ig), a biologic agent that targets T-cell co-stimulation. All patients were receiving DMARDs (five on methotrexate, two on leflunomide), and three patients were receiving low-dose glucocorticoids. At baseline, three patients had low insulin resistance (median HOMA-IR 1.16, range 0.87 to 1.91), and four patients had high insulin resistance (median HOMA-IR 5.34, range 3.45 to 8.85).

At week 12 of abatacept treatment, disease activity was reduced in all seven RA patients (Wilcoxon signed ranks test, P = 0.046) (Figure 2A). Insulin resistance followed a variable trend, and HOMA-IR increased in three patients (two with low baseline insulin resistance), remained stable in one patient, and decreased in three patients (two with high baseline insulin resistance) (Figure 2B). Similarly, HOMA-B was not significantly altered during the study period (data not shown).

To examine the effect of CTLA4.Ig treatment on the insulin signaling cascade, we measured the levels of Ser³¹²-phosphorylated IRS-1 and phosphorylated AKT (p-AKT) in protein extracts from PBMCs (Figure 2C). We found no significant change in p-Ser³²¹ IRS-1 or in p-AKT levels in RA patients with either low or high baseline insulin resistance (Figure 2D, E). We found no correlation between changes in HOMA-IR and IRS-1 or AKT phosphorylation levels.