Reversal of Obesity- and Diet-Induced Insulin Resistance with Salicylates or Targeted Disruption of Ikkβ

See allHide authors and affiliations

Science  31 Aug 2001:
Vol. 293, Issue 5535, pp. 1673-1677
DOI: 10.1126/science.1061620

This article has a correction. Please see:


We show that high doses of salicylates reverse hyperglycemia, hyperinsulinemia, and dyslipidemia in obese rodents by sensitizing insulin signaling. Activation or overexpression of the IκB kinase β (IKKβ) attenuated insulin signaling in cultured cells, whereas IKKβ inhibition reversed insulin resistance. Thus, IKKβ, rather than the cyclooxygenases, appears to be the relevant molecular target. Heterozygous deletion (Ikkβ +/−) protected against the development of insulin resistance during high-fat feeding and in obese Lepob/ob mice. These findings implicate an inflammatory process in the pathogenesis of insulin resistance in obesity and type 2 diabetes mellitus and identify the IKKβ pathway as a target for insulin sensitization.

Insulin resistance refers to a decreased capacity of circulating insulin to regulate nutrient metabolism. Individuals with insulin resistance are predisposed to developing type 2 diabetes, and insulin resistance is an integral feature of its pathophysiology. Chronic secretion of large amounts of insulin to overcome tissue insensitivity can lead, in predisposed individuals, to pancreatic β cell failure and concomitant defects in glucose and lipid metabolism. The prevalence of insulin resistance is high and rising, but only rare genetic causes have been identified. The molecular cause of acquired insulin resistance, which is promoted by sedentary lifestyle, obesity, fatty diet, and increased age, and is reversed by exercise and weight loss, is similarly unknown.

High doses of salicylates [4 to 10 g per day (g/day)], including sodium salicylate and aspirin, have been used to treat inflammatory conditions such as rheumatic fever and rheumatoid arthritis. These high doses are thought to inhibit nuclear factor kappa B (NF-κB) (1) and its upstream activator the IκB kinase β (IKKβ) (2), as opposed to working through cyclooxygenases (COXs), the classical targets of nonsteroidal anti-inflammatory drugs (NSAIDs). High doses of salicylates also lower blood glucose concentrations (3–7), although their potential for treating diabetes has been all but forgotten by modern biomedical science. We have investigated potential mechanisms of these hypoglycemic effects to identify potential mediators of insulin resistance and molecular targets for intervention. We have found that reduced signaling through the IKKβ pathway, either by salicylate inhibition or decreased IKKβ expression, is accompanied by improved insulin sensitivity in vivo. Our findings further indicate that the IKKβ pathway may contribute to insulin resistance in type 2 diabetes and obesity by impinging on insulin signaling.

We determined the effect of high doses of salicylates on the severe insulin resistance seen in genetically obese rodents. Twelve-week-old male Zucker fa/fa rats and 8-week-old male ob/obmice were treated for 3 to 4 weeks with 120 mg/kg/day of aspirin or sodium salicylate, administered by continuous subcutaneous infusion. Fasting blood glucose values and glucose tolerance were improved in Zucker fa/fa rats (Fig. 1A). Concomitant reductions in insulin concentrations (Fig. 1B) indicated a marked improvement in insulin sensitivity. Glucose tolerance in leanfa/+ animals was normal, and blood glucose concentrations were similar after aspirin treatment (Fig. 1C). Nevertheless, lower insulin concentrations in the aspirin-treated group (Fig. 1D) demonstrate improved insulin sensitivity, despite milder insulin resistance. The ability of high-dose aspirin to increase insulin sensitivity was further established in insulin tolerance tests (Fig. 1E). Intraperitoneal injection of insulin (2.0 U per kilogram of body weight) had essentially no effect on blood glucose concentrations in untreated fa/fa rats. However, the same insulin dose caused a decrease in blood glucose when given to aspirin-treated animals.

Figure 1

In vivo effects of aspirin in Zucker fatty rats and ob/ob mice. (A to F) Twelve-week-old male Zucker fa/fa rats and (Gand H) 8-week-old ob/ob(Lepob/ob ) and ob/+ mice were given free access to food and water. Aspirin (120 mg/kg/day) was continuously infused for 3 to 4 weeks by Alzet pumps (pump 2ML2 was used in rats, pump 2002 was used in mice) implanted subcutaneously between the scapulae of the animals. In all panels, data are mean ± SEM values, diamonds and dashed lines represent control implantation of a pump with vehicle only, and solid circles and solid lines represent 3 or 4 weeks of treatment with aspirin. For glucose tolerance tests (GTT), glucose (2.0 g/kg) was administered by oral gavage (rats) or intraperitoneal injection (mice) after an overnight fast. (A and C) Blood glucose and (B and D) serum insulin concentrations were determined during oral glucose tolerance tests in (A and B) Zuckerfa/fa rats or (C and D) fa/+ rats; six animals were in each treatment group. In (A) and (C), *P < 0.05, **P < 0.01 for vehicle versus aspirin treatment, Student's t test. (B) P = 0.0004 and (D)P = 0.02, for vehicle versus aspirin treatment, integrated areas under the curves; Student's t test. (E) For insulin tolerance tests (ITT), insulin (2.0 U/kg) was injected intraperitoneally after an overnight fast (six Zuckerfa/fa rats, P = 0.0001, for vehicle versus aspirin treatment, integrated areas under the curves; Student'st test). (F) Cholesterol (Chol), triglyceride (TG), long-chain FFA, and ALT concentrations were measured in sera from fasting Zucker fa/fa rats (for FFA, P < 0.005 for three-group analysis of variance; *P = 0.01,t = 0 versus t = 1 week of aspirin treatment, **P < 0.002, t= 0 versus t = 3 weeks of aspirin treatment, Dunnett's adjusted pairwise comparison. For TG, *P <0.05,t = 0 versus t = 3 weeks of aspirin treatment; Student's t tests). Glucose tolerance tests are shown in (G) ob/ob and (H) ob/+ mice. (G) P = 0.004 and (H) P < 0.0001 for 10 mice treated with vehicle versus 10 mice treated with aspirin, integrated areas under the curves; Student's t tests. [Insets in (G) and (H)] Fasting serum insulin concentrations (ng/ml); **P < 0.01 for 10 mice treated with aspirin versus 10 controls treated with vehicle alone. Aspirin treatment did not significantly affect food intake or body weights in any of these studies. Animal care was in accordance with institutional and NIH guidelines.

Increased triglyceride concentrations in the blood of Zucker rats fell from 494 ± 68 mg/dl to 90 ± 58 mg/dl during 3 weeks of aspirin treatment (Fig. 1F). The concentrations of free fatty acid (FFA) dropped as well, from 3.1 ± 0.3 mM to 1.1 ± 0.2 mM. The decrease in the amount of circulating FFA occurred within 1 week of aspirin treatment, preceding reductions in the amounts of triglyceride and glucose in the blood. This is consistent with the hypothesis that increased FFA concentrations contribute to the pathogenesis of hyperglycemia and hypertriglyceridemia. Cholesterol concentrations were unaffected. Hepatotoxicity was not observed at the high aspirin and salicylate doses used, judging from the consistently normal circulating concentrations of the liver enzyme alanine aminotransferase (ALT), seen throughout our studies (Fig. 1F). Serum salicylate concentrations were 0.81 ± 0.25 mM.

Ob/ob mice (Lepob/ob ) are a more relevant model for type 2 diabetes, as the animals are diabetic in addition to being obese and severely insulin-resistant. Fasting blood glucose values and glucose tolerance were significantly improved by aspirin treatment (Fig. 1G). Heterozygous, leanLepob/+ mice exhibited postprandial hyperglycemia but were less insulin resistant. Glucose intolerance inLepob/+ mice was normalized with aspirin treatment (Fig. 1H). Insulin concentrations were reduced during aspirin therapy in both Lepob/ob andLepob/+ mice. Neither aspirin nor salicylate affected food intake or body weight in Zucker fa/fa rats orob/ob mice.

We isolated tissues from treated animals to analyze various signaling proteins (8–10). Insulin receptor (IR) tyrosine phosphorylation, one of the earliest responses to insulin binding, was barely detectable in the liver and muscle of insulin-resistant Zucker rats (Fig. 2, A and B). Increased stimulation occurred in corresponding tissues from aspirin- and salicylate-treated animals, suggesting an increase in insulin responsiveness. Signaling from the IR to insulin receptor substrates (IRSs), phosphatidylinositol 3-kinase, and the PDK1 protein kinase is required for the maintenance of metabolic homeostasis. Phosphorylation of the protein kinase AKT, a subsequent step in this cascade, correlates with IR activation in tissues from Zucker rats (Fig. 2, A and B). The blunted insulin-stimulated phosphorylation of AKT in the liver and muscle of untreated Zucker rats was increased after aspirin or salicylate treatment, providing a biochemical correlate for increased in vivo insulin sensitivity. The electrophoretic mobility of IRS-1 from rat livers increased with aspirin and salicylate treatment [Web fig. 1 (11)], suggesting a decrease in serine-threonine (Ser-Thr) phosphorylation (this is a known inhibitor of insulin signaling). Basal IKK activity was elevated in tissues from Zuckerfa/fa rats relative to those from lean fa/+controls (12).

Figure 2

Signaling proteins in tissues from aspirin- and salicylate-treated Zucker fa/fa fatty rats. Animals treated for 3 to 4 weeks with aspirin or sodium salicylate received intravenous injections of insulin or saline 7 min before being killed. Tissues were harvested and frozen immediately in liquid nitrogen. Liver and plantaris muscle samples were homogenized and signaling proteins were identified by Western blotting (13). (A) Liver or (B) muscle homogenates were immunoprecipitated with antiIR and proteins were identified by blotting with anti-pY and anti-IR. Alternatively, tissue homogenates were separated by SDS-PAGE and proteins were identified by blotting with antibodies specific to phospho-AKT (pAKT) and antibodies to AKT. IB, immunoblot; IP, immunoprecipitate.

Studies with cultured cells were used to investigate potential mechanisms relating salicylate treatment to the in vivo reversal of insulin resistance. Treatment of 3T3-L1 adipocytes with tumor necrosis factor–α (TNF-α) induced “insulin resistance,” as judged by decreases in insulin-stimulated tyrosine phosphorylation of the IR β-subunit and IRS-1, to 42 ± 11% and 37 ± 9%, respectively, of that in insulin-stimulated cells not treated with TNF-α (Fig. 3, A and B) (13). This was reversed by prior treatment of cells with high doses of aspirin (5 mM). The amounts of IR and IRS-1 proteins were similar in all cells. TNF-α activates the IKK complex (14). Phosphatase inhibitors such as okadaic acid and calyculin A also activate IKKβ (15,16), but without activating upstream elements in the TNF-α signaling cascade, and these inhibitors also induce insulin resistance in isolated tissues and cultured cells (17,18). Calyculin A reduced insulin-stimulated tyrosine phosphorylation of the IR and IRS-1, to 29 ± 12% and 16 ± 2%, respectively, of that in untreated cells, and this was prevented by incubating the cells with aspirin (Fig. 3, A and B). Similar results were obtained after okadaic acid treatment (19). The reduced electrophoretic mobility of IRS-1 due to calyculin A treatment was reversed with aspirin (Fig. 3C), further suggesting that aspirin's ability to reverse insulin resistance might result from reduced levels of Ser-Thr phosphorylation of components in the insulin action cascade.

Figure 3

Aspirin effects on insulin signaling in 3T3-L1 adipocytes. 3T3-L1 adipocytes were serum-starved for 16 hours and treated or not treated with 5 mM aspirin for 2 hours and either 6.0 nM mTNF-α (20 min) or the phosphatase inhibitor calyculin A (at 2.0 nM for 30 min). After a 5-min stimulation with 10 nM insulin, the cells were chilled and solubilized and proteins were immunoprecipitated with (A) anti-IR or (B and C) anti-IRS1. Proteins separated by SDS-PAGE were identified by Western blotting with anti-pY, anti-IR, or anti-IRS1. Control blots (not shown) indicated that protein amounts of IR and IRS1 did not differ significantly between treatments. Ins, insulin. (A and B) Phosphorylation levels were quantified by densitometry and expressed relative to insulin-stimulated controls; the numbers of individual experiments are shown in parentheses (*P < 0.01 for minus aspirin, plus insulin, minus versus plus mTNF-α, or minus versus plus calyculin A; **P < 0.05 for minus versus plus aspirin of mTNF-α or calyculin A–treated pairs; Student's ttest).

Fao hepatoma cells are an insulin-responsive model for liver as opposed to fat. TNF-α treatment decreased tyrosine phosphorylation of the insulin receptor substrate IRS-2 (Fig. 4A) (20). Decreased IRS-2 tyrosine phosphorylation was reversed by aspirin or sodium salicylate. Aspirin and sodium salicylate are equipotent inhibitors of IKKβ (2), whereas aspirin is ∼100-fold more potent toward the cyclooxygenases (21). Our findings suggest that IKKβ, and not COX1 nor COX2, might alter insulin signaling. Additional NSAIDs were used to further evaluate potential molecular mediators. Ibuprofen and naproxen, which inhibit both COX1 and COX2, did not reverse TNF-α–induced insulin resistance (Fig. 4C). The selective COX2 inhibitor NS-398 similarly had no effect. Studies were conducted with doses of the drugs known to have biological effects (2, 21). These pharmacological profiles further point to IKK as the target of these effects and demonstrate that COX1 and COX2, the classical targets for NSAIDs, do not mediate the antidiabetic effects of aspirin and salicylate. This issue was addressed further using mice with reduced COX expression; homozygous and heterozygous deletions of either COX1 or COX2 had no effect on carbohydrate or lipid metabolism in insulin-resistant mice (22).

Figure 4

(A) NSAID effects in Fao hepatoma cells. Fao cells were serum-starved for 16 hours followed by 2-hour incubations at 37°C with 5 mM aspirin, 10 mM sodium salicylate, 25 μM ibuprofen, 25 μM sodium naproxen, or 25 μM NS-398 (a selective COX2 inhibitor). Cells were then stimulated sequentially with 6.0 nM mTNF-α for 20 min and 10 nM insulin for 5 min. Cells were chilled and solubilized and proteins were immunoprecipitated with anti-IR and detected by Western blotting using anti-pY. Phosphorylation was quantified by densitometry and is expressed relative to insulin-stimulated controls; the numbers of individual experiments are shown in parentheses. (B andC) Induction of insulin resistance with IκB kinases and reversal with dominant inhibitors. (B) Myc-tagged NIK or Flag-tagged IKKβ or (C) kinase-deficient IKKβ were expressed in HEK 293 cells. The pRK7 (cytomegalovirus) vectors encoding the kinases were obtained from M. Rothe (Tularik); transfections of 50 to 60% confluent cells (1.5 μg of DNA per well; six well dishes) using FuGene 6 (Boehringer-Mannheim) were as recommended. Experiments were initiated 36 hours after transfection and 16 hours after removal of serum from the culture medium. (B) Cells were stimulated for 5 min with 10 nM insulin unless otherwise indicated. (C) Cells were treated for 40 min with 6 nM mTNF-α, followed by 5-min stimulations with 10 nM insulin. In (B) and (C), protein expression levels ranged from 10 to 20 times that of the endogenous proteins.

IKK complexes contain the heterodimeric kinases IKKα and IKKβ and the scaffolding protein IKKγ (14). Either the IKKβ catalytic subunit or NIK, an upstream activator, was expressed in HEK 293 cells. Insulin stimulated the activation of IR (Fig. 4B), IRS-2, and AKT (19); activation was attenuated by expression of IKKβ or NIK (Fig. 4B) [Web fig. 2 (11)], and attenuated activation was reversed by treatment with aspirin (19). IκBα levels were reduced by IKKβ or NIK expression [Web fig. 2 (11)]. Because IKKα and IKKβ form heterodimers, overexpression of kinase-deficient IKKα or IKKβ inhibits endogenous components of the complex and signaling to NF-κB (23). We therefore asked whether the expression of dominant inhibitory IKKβ(K44A) blocked the induction of insulin resistance. TNF-α treatment reduced insulin-stimulated IR activation to 29 ± 2% of untreated controls, and expression of IKKβ(K44A) reversed the TNF-α–inhibited effects (Fig. 4C) [Web fig. 2 (11)]. Active IKKβ thus promotes insulin resistance in cultured cells, and the inactive, dominant inhibitory kinase blocks TNF-α induced insulin resistance.

Studies with mice having targeted disruption of the Ikkβlocus further tested potential roles of IKKβ in the development and reversal of insulin resistance. Ikkβ−/− mice die in utero because of enhanced liver apoptosis, whereas heterozygous Ikkβ+/− mice were reportedly normal (24, 25). Fasting glucose and insulin concentrations were consistently lower inIkkβ+/− compared toIkkβ+/+ littermates (Fig. 5, A and B). The potential protective effect of reduced Ikkβ gene dose was tested in crosses between Ikkβ+/− andLepob/ob mice. Fasting blood glucose concentrations were reduced inIkkβ+/−Lepob/ob mice compared with those in Ikkβ+/+Lepob/ob littermates (Fig. 5C). Glucose tolerance in theIkkβ+/−Lepob/ob mice was improved compared to Ikkβ+/+Lepob/ob littermates (Fig. 5D), although insulin concentrations were indistinguishable (12). These findings are consistent with improved insulin sensitivity inIkkβ+/−Lepob/ob mice compared to that in Ikkβ+/+Lepob/ob littermates. Also consistent with improved metabolic control, plasma FFA concentrations were lower in theIkkβ+/−Lepob/ob mice (1.86 ± 0.12 mM) than theirIkkβ+/+Lepob/ob littermates (2.24 ± 0.09 mM; P = 0.03).

Figure 5

Glucose tolerance and insulin sensitivity inIkkβ+/– mice. (A and B)Ikkβ+/– mice were backcrossed four generations on a C57BL/6J background. Fasting blood glucose and fasting insulin concentrations were determined inIkkβ+/– and Ikkβ+/+ male littermates. Values represent the mean ± SEM for wild-type (WT) (diamonds, n = 9) andIkk+/– (circles, n = 6) mice. (A)P <0.001 and (B) P = 0.02 for integrated areas under the curves, Student's t tests for WT versusIkk+/– mice. (C and D)Ikkβ+/– offspring were crossed withLepob/+ (C57BL/6J) mice (Jackson Laboratory, Bar Harbor, ME). F1 maleIkk+/– Lepob/+ offspring were crossed with Ikk+/+Lepob/+ females, and F2 littermates were studied. (C) Fasting blood glucose concentrations in male littermates. Values represent the mean ± SEM for Ikk+/+Lepob/ob (diamonds,n = 7) andIkk+/–Lepob/ob (circles,n = 7); P = 0.003 for integrated areas under the curves; Student's t tests,Ikk+/–Lepob/ob versusIkk+/+ Lepob/ob . (D) Glucose tolerance tests were conducted with 9- to 12-week-old maleIkk+/−Lepob/ob andIkk+/+Lepob/ob littermates. Glucose (2.0 g/kg) was administered by intraperitoneal injection after an overnight fast. Values represent the mean ± SEM forIkk+/+Lepob/ob (diamonds,n = 12 males) and Ikk+/–Lepob/ob (circles, n = 11 males;P = 0.006 for integrated areas under the curves; Student's t tests,Ikk+/–Lepob/ob versusIkk+/+Lepob/ob ). (E toH) Beginning at 4 weeks of age, maleIkkβ+/– and Ikkβ+/+ littermates were maintained on high-fat diets (Research Diets D12331; 58% of calories from coconut oil). Fasting (E) glucose and (F) insulin concentrations were measured from 8- to 23-week-old male littermates (diamonds, n = 7 Ikk+/+ and circles, n = 8 Ikk+/− mice. (E)P <0.05 and (F) P = 0.01 for integrated areas under the curves, Student's t tests for WT versusIkk+/– mice. (G and H) Glucose tolerance tests were conducted with 18-week-old male littermates. Values represent the mean ± SEM for Ikk+/+ (diamonds,n = 10) and Ikk+/− (circles,n = 10). (G) *P <0.005; **P<0.01, Student's t tests for WT versusIkk+/– mice. (H) P = 0.03 for integrated areas under the curves, Student's t tests for WT versus Ikk+/− mice.

To further assess the effect of a reduction in Ikkβ gene dose, mice were fed a diet high in fat. Fasting glucose and insulin concentrations were consistently lower inIkkβ+/− compared toIkkβ+/+ littermates (Fig. 5, E and F). Reduced insulin concentrations in the Ikkβ+/− mice were maintained throughout glucose tolerance testing in 18-week-old littermates (Fig. 5H). Dietary intake and weights were indistinguishable between all pairs ofIkkβ+/− and Ikkβ+/+ littermates. These data demonstrate that a reduction inIkkβ gene dose reduces fasting glucose and insulin concentrations and protects against the development of insulin resistance in predisposed rodents.

Our findings demonstrate that increased IKK activity promotes insulin resistance, in obese rodents (12) when the kinase is overexpressed, or when IKK is activated by known stimulators. Conversely, reductions either in IKK activity or in the expression of its IKKβ subunit significantly improved insulin sensitivity. Even a 50% reduction in gene dosage improved in vivo glucose and lipid metabolism, which may explain why weak inhibitors of IKKβ, such as aspirin and sodium salicylate, have significant effects on glucose and lipid homeostasis. Although not recognized previously, there is an overlap between stimuli that activate IKK and conditions that promote insulin resistance, including proinflammatory cytokines such as TNF-α, hyperglycemia, phorbol esters and protein kinase C (PKC) enzymes, Ser-Thr phosphatase inhibitors, and bacterial lipopolysaccharide. These are either in vivo mediators of insulin resistance or experimental mimics in cultured cells. Our findings are consistent with potential links between chronic subacute inflammation and insulin resistance (26, 27), whether this is mediated by TNF-α produced in fat (28–31) or through TNF-α–independent mechanisms. As a potentially important example of the latter, in rodent muscle, FFA infusion activates PKC-θ (32), a known activator of IKK (33), and FFA-induced insulin resistance is suppressed by aspirin treatment and inIkkβ+/− mice (34). IKK activation through any mechanism initiates NF-κB–mediated transcription, which in certain cells would enhance the production of TNF-α. This positive feedback loop could perpetuate a vicious cycle of low-level inflammatory signaling, leading to insulin resistance. Our findings predict that IKK inhibition breaks this cycle. Too few tools are currently available to treat patients with insulin resistance and type 2 diabetes; IKKβ may provide a valuable target for the discovery of new drugs to treat these conditions.

  • * These authors contributed equally to this work.

  • To whom correspondence should be addressed. E-mail: Steven.Shoelson{at}


View Abstract

Stay Connected to Science

Navigate This Article