Interleukin-6 Induces Cellular Insulin Resistance in Hepatocytes

Porcine insulin and recombinant human IL-6 were purchased from Sigma (St. Louis, MO) and R&D Systems (Minneapolis, MN), respectively. Cell culture reagents such as Dulbecco’s modified Eagle’s medium (DMEM) and fetal bovine serum (FBS) were purchased from Life Technologies (Burlington, ON, Canada). Anti-IR antibody (α-IR-1) was a gift from Dr. S. Jacobs (Wellcome Research Laboratories, Research Triangle Park, NC). The phospho-Akt (Ser-473) antibody was purchased from Cell Signaling Technology (Beverly, MA). Antiphosphotyrosine antibody (4G10), polyclonal rabbit anti-p85, and anti-IR substrate (IRS)-1 were from Upstate Biotechnology (Lake Placid, NY). [C¹⁴]glucose was purchased from NEN (Boston, MA). Male C57/BL6 mice were from Charles River Laboratories (Wilmington, MA).

HepG2 human hepatocarcinoma cells were obtained from American Type Tissue Culture Collection (Rockville, MD). Cells were maintained at 37°C in a humidified 5% CO₂ environment in DMEM/high glucose supplemented with 10% FBS.

HepG2 cells were grown to confluence in 10-cm dishes before overnight treatment in serum-free DMEM/high glucose containing 1% BSA. Cells were then treated with recombinant human IL-6 at 20 ng/ml for 0–8 h. At selected time points, cells were stimulated for 3 min with insulin at the indicated concentrations. After two washes with cold PBS, cells were lysed in lysis buffer A containing 50 mmol/l Tris, pH 7.4, 140 mmol/l NaCl, 1% Triton X-100, 50 mmol/l NaF, 10 mmol/l tetrasodium pyrophosphate, 25 mmol/l benzamidine, protease inhibitor mixture (Calbiochem), 2.5 μmol/l pervanadate, 2 mmol/l phenylmethylsulfonyl fluoride, and 10% glycerol. Lysates were passed 10 times through an 18-gauge needle, centrifuged at 10,000g for 10 min, and then adjusted to equal amounts of protein as determined by the Bradford method (19). IRs were immunoprecipitated using an anti-human IR antibody (α-IR-1) bound to protein G-Sepharose. Immune complexes were washed four times with wash buffer (1% Triton X-100, 100 mmol/l Tris-HCl, pH 7.4, and 150 mmol/l NaCl) and separated by SDS-PAGE on 5–20% gradient gels (20). In experiments in which IRS-1 was immunoprecipitated, the same protocol was followed, except protein A Sepharose was used. After separation by SDS-PAGE, the samples were subjected to Western blot analysis using appropriate antibodies at 4°C overnight. To determine the amount of phosphorylated Akt, lysates from HepG2 cells, treated as previously described, were separated by SDS-PAGE and subjected to Western blot analysis using anti-phospho Akt antibody. Blots were visualized by enhanced chemiluminescence (Amersham Pharmacia). Quantitation was performed using a ChemiImager (Alpha Innotech, San Leandro, CA).

Mouse hepatocytes were prepared for primary culture by nonrecirculating collagenase perfusion, as adapted from Klaunig et al. (21). In brief, the liver was initially perfused with Ca²⁺-free Hank’s buffer and then dissociated with collagenase type IV (1 mg/ml; Sigma) in Hank’s buffer plus 5 mmol/l CaCl₂. Isolated hepatocytes were resuspended in Hank’s buffer containing 1.2 mmol/l CaCl₂ and 0.6 mmol/l MgSO₄. Cells were then filtered through 90-micron nylon mesh, counted, and tested for viability using trypan blue exclusion.

Primary mouse hepatocytes were seeded onto collagen-coated 6-well tissue culture plates (Corning, Corning, NY) at 0.5 × 10⁶ cells/well in serum-free Waymouth’s media supplemented with 100 units/ml penicillin/streptomycin and 10 nmol/l dexamethasone. A total of three wells were seeded for each condition. The cells were incubated overnight at 37°C in a humidified atmosphere containing 5% CO₂. After 12 h, cells were pretreated for 90 min with IL-6 (20 ng/ml) in serum-free low-glucose DMEM. Cells were then treated with 1.5 μCi [¹⁴C]d-glucose (300 mCi/mmol/l) and a final concentration of either 0, 1, or 100 nmol/l insulin. After 3 h, the reaction was terminated by washing the cells three times with ice-cold PBS. Cells were solubilized with 0.3 ml 10N KOH. Cold glycogen carrier (4 mg) was added to the lysates, and samples were boiled for 30 min. Glycogen was precipitated with two volumes ethanol overnight on ice. Precipitated glycogen was centrifuged at 10,000g for 10 min. Pellets were washed once with 66% ethanol, resuspended in 0.5 ml water, and counted by scintillation counting. All procedures involving laboratory animals were approved by the University of Rochester Committee on Animal Resources.

Data were analyzed using a Student’s t test (Stataquest) to compare results between time points. P < 0.05 was considered statistically significant.

To examine the effect of IL-6 on IR signal transduction, HepG2 cells were serum starved overnight and then exposed to IL-6 (20 ng/ml) for 0–8 h. This concentration of IL-6 has been shown by Paysant et.al. (22) to produce a near-maximum effect (fibrinogen production) in HepG2 cells. At the indicated time points after initiation of IL-6 exposure, the HepG2 cells were subjected to a 3-min treatment with 0.5 or 100 nmol/l insulin. The effect of IL-6 alone was also assessed. Immunoprecipitates of the IR (Fig. 1) were subjected to Western blot analysis using an antiphosphotyrosine antibody to determine IR autophosphorylation. As expected, insulin increased autophosphorylation of the IR at both the physiologic and supraphysiologic concentrations, but IL-6 had no effect on IR autophosphorylation in the absence or presence of insulin at the time points examined. Similarly, IL-6 had no effect on tyrosine phosphorylation of IRS-1 in the absence of insulin or in response to 100 nmol/l insulin (Fig. 2). However, IL-6 did inhibit tyrosine phosphorylation of IRS-1, which was initiated by the physiologic insulin concentration (0.5 nmol/l). Tyrosine phosphorylation of IRS-1 decreased to 43 ± 10% of control levels at 1 h after IL-6 treatment. A partial recovery was observed at 2 h, followed by a return to suppression of tyrosine phosphorylation (55 ± 20% of control levels) at 4 h. Such a transient loss of the IL-6 response in HepG2 cells has been reported by Wang et.al. (23) when examining c-jun protein expression in response to IL-6 at 10 ng/ml.

Association of the p85 subunit of phosphatidylinositol (PI)-3 kinase with IRS-1 is a rapid and essential response to insulin-dependent tyrosine phosphorylation of IRS-1. Both 0.5 and 100 nmol/l insulin increased this association, as detected by Western blot analysis of IRS-1 immunoprecipitates (Fig. 3). Once again, IL-6 suppressed this insulin-dependent effect only in cells exposed to the physiological insulin concentration. A small suppression was seen at 120 min with 100 nmol/l insulin but was not statistically significant. As with IRS-1 phosphorylation, the inhibition of p85 association was maximum at 60 min, with a decrease to 58 ± 31% of controls. After a loss of the IL-6-dependent inhibition of IRS-1/p85 association at 2 h, association again was inhibited at 4 and 8 h. IL-6 alone was also observed to increase association of the p85 subunit with IRS-1, although the magnitude of the effect was substantially less than that of the physiologic insulin concentration.

Finally, lysates from insulin- and IL-6-treated cells were assayed for the presence of activated Akt using Western blot analysis (Fig. 4). Insulin at both 0.5 and 100 nmol/l activated Akt, as assessed by detection of Akt phosphorylation at ser473. In HepG2 cells exposed to 0.5 nmol/l insulin, IL-6 again produced an inhibition, with a statistically significant suppression of Akt activation to 52 ± 12% of controls at 1 h and comparable inhibition at 4 and 8 h. In agreement with data in Figs. 2 and 3, a transient diminution of this inhibition was observed at the 2-h time point. In contrast to the other parameters of IR signaling that were unaffected by IL-6 when activated by 100 nmol/l insulin, activation of Akt in response to 100 nmol/l insulin was suppressed by IL-6. This inhibition was more gradual, beginning at 60 min but progressing through 8 h. Finally, IL-6 alone had no effect on the activation state of Akt, despite its modest increase in IRS-1/p85 association. In some experiments, basal Akt activity was detectable (Fig. 8), but this was not affected by IL-6.

The above data suggested that IL-6-dependent inhibition of IR signaling was dependent on the insulin concentration. To more thoroughly characterize the effect of IL-6 on dose-dependent insulin signaling, HepG2 cells were pretreated with IL-6 (20 ng/ml) and then stimulated with various insulin concentrations. A 90-min time point was chosen as a compromise between the time to maximum IL-6-dependent inhibition of Akt with physiologic insulin (60 min) and the apparently longer time course with supraphysiologic concentrations (≥2 h). IL-6 had no effect on dose-dependent IR autophosphorylation under these conditions (data not shown), confirming data in Fig. 1. However, tyrosine phosphorylation of IRS-1 exhibited decreased sensitivity to insulin as a result of IL-6 exposure—it was undetectable until 10 nmol/l insulin (Fig. 5). This is somewhat different from the results in Fig. 2, which show that IL-6 did not totally suppress IRS-1 phosphorylation induced by 0.5 nmol/l insulin. These results suggest that the true nadir in IRS-1 phosphorylation is at 90 min (Fig. 5), not 60 min (Fig. 2). Control cells exhibited tyrosine phosphorylation of IRS-1 in response to 0.3 nmol/l insulin and threefold maximum phosphorylation at 10 nmol/l. Although basal phosphorylation of IRS-1 was sometimes detected, it did not appear to be altered by the presence of IL-6.

No effect of IL-6 on tyrosine phosphorylation of the IR or IRS-1 in the absence of insulin was detected. However, IL-6 did affect basal p85 association with IRS-1 (Fig. 6). This was also observed in Fig. 3. The data in Fig. 6 show that cells exposed to IL-6 without insulin exhibit a 1.7-fold increase in IRS-1/p85 complexes when compared with controls. The association of p85 with IRS-1 also increased in response to insulin in both IL-6-treated and control cells. The amount of IRS-1/p85 complexes was comparable until 100 nmol/l insulin. At this concentration, control cells showed an additional response to insulin, whereas IL-6-treated cells did not. When IRS-1/p85 complex formation in IL-6-treated and control cells was normalized to the individual insulin-free basal levels (Fig. 6 inset), insulin-dependent p85 association was clearly inhibited by at least 50% at all concentrations. The apparent increased sensitivity to insulin of p85/IRS-1 association versus IRS-1 tyrosine phosphorylation in the presence of IL-6 (Fig. 6 versus Fig. 5) is most likely due to the more sensitive Western blot method for detection of p85 at these low insulin concentrations.

Finally, insulin increased phosphorylation of Akt at 0.3 nmol/l in untreated cells and reached a maximum of sixfold at 10 nmol/l. Cells pretreated with IL-6 exhibited a complete loss of Akt activation in response to all insulin concentrations (Fig. 7). These results again suggest that a maximum effect of IL-6 is attained at 90 min. This is the optimum time point for all insulin concentrations under the conditions used in this study. As with IRS-1 tyrosine phosphorylation, basal activation of Akt was sometimes observed but not affected by IL-6 treatment. These data suggest that IL-6 induces a decrease in both the sensitivity and maximum effect of insulin on p85 association with IRS-1 and activation of Akt in HepG2 cells.

While HepG2 cells are a useful model to investigate aspects of hepatic insulin action, primary hepatocytes remain a more physiologic model. The objective in the following experiments was to confirm that IL-6 inhibits IR signal transduction as well as inhibits glycogen synthesis, a metabolic end point of insulin action. Primary hepatocytes were isolated from mice and plated on collagen-coated dishes overnight in serum-free media. Paysant et al. (22) have shown that IL-6 at 20 ng/ml elicits a near-maximum response in HepG2 cells. To examine IL-6 sensitivity in primary hepatocytes, the concentration-dependent effect of IL-6 on STAT-3 tyrosine phosphorylation was examined. As shown in Fig. 8A, STAT-3 phosphorylation increased in response to 20 ng/ml IL-6 and showed further response at 60 ng/ml. These data indicated that the sensitivity to IL-6 was similar in HepG2 cells and primary hepatocytes. Therefore, 20 ng/ml IL-6 was used in the following experiments with primary hepatocytes to examine the effect of IL-6 on IR signaling.

Akt activation in response to a series of insulin concentrations was markedly inhibited by IL-6 (Fig. 8B). This inhibition was not as profound as the inhibition observed in HepG2 cells, but clearly demonstrated that IL-6 can inhibit IR signaling in primary hepatocytes. As shown in Fig. 8C, IL-6 inhibited activation of glycogen synthesis in response to physiologic insulin concentrations. In control cells, 1 nmol/l insulin increased glycogen synthesis by ∼50%. IL-6 inhibited this increase in glycogen synthesis by nearly 75%. The level of glycogen synthesis at 1 nmol/l insulin with an IL-6 treatment was no greater than IL-6 alone. A supraphysiological insulin concentration (100 nmol/l) overcame the IL-6-dependent inhibition of glycogen synthesis. The discordance between the ability of IL-6 to inhibit Akt activation in response to 100 nmol/l insulin, and the inability of IL-6 to inhibit glycogen synthesis at this insulin concentration has several possible explanations. First, the level of Akt activity that is reached with 100 nmol/l insulin, even in the presence of IL-6, may be sufficient to activate glycogen synthesis. Second, the glycogen synthesis assay requires exposure to IL-6 for 90 min followed by insulin for 3 h. Thus, the acute activation state of Akt after 3 min of insulin treatment may not reflect time-averaged glycogen synthesis over 3 h. Third, supraphysiologic insulin concentrations may be activating glycogen synthase through an IL-6-insensitive pathway such as the insulin-responsive serine/threonine phosphatase PP1. Finally, phosphoAkt (Ser473) may not reflect the potential of insulin to activate glycogen synthesis. Nonetheless, these data with primary hepatocytes support our results in HepG2 cells and provide evidence that IL-6 alters a major metabolic end point of insulin action in the liver.

Proinflammatory cytokines are thought to contribute to insulin resistance in both humans and animals (1–6). However, the molecular mechanism of cytokine-induced insulin resistance is not well characterized. The present study clearly demonstrates that IL-6 affects both the proximal and distal events in hepatic IR signal transduction. IL-6-induced inhibition was not an immediate event, but was observed after an exposure of 1–2 h, perhaps suggesting that protein synthesis is required. Work is ongoing to address the mechanism of IL-6-mediated inhibition. Insulin signal transduction was not inhibited at the level of IR autophosphorylation. However, insulin-dependent tyrosine phosphorylation of IRS-1, association of the p85 subunit of PI 3-kinase with IRS-1, and activation of Akt were all comparably inhibited, both temporally and in magnitude, by IL-6 exposure in HepG2 cells at physiological concentrations of insulin. This suggests a site of action of IL-6 at the level of IR/IRS-1 interaction. IL-6 was also capable of causing inhibition of insulin action in primary hepatocytes. Pretreatment of primary hepatocytes with IL-6 markedly inhibited both insulin-induced Akt activation and glycogen synthesis. This suggests that hepatocytes, in general, are physiologic targets for the inhibitory effect of IL-6 on insulin signaling.

While the effect of IL-6 on all the measured parameters of IR signaling was closely correlated in the presence of physiologic insulin concentrations, a lack of correlation was found between insulin-dependent p85/IRS-1 complex formation and Akt activation at supraphysiologic concentrations of insulin. The only aspect of IR signaling that was inhibited by IL-6 under these conditions was activation of Akt. Additionally, in HepG2 cells, IL-6 caused an association of the p85 regulatory subunit with IRS-1 in the absence of insulin, but this association was not sufficient to cause activation of Akt. A review of recent literature provides several examples of discordance between IRS-1/p85 association and Akt activation (24–27). Takano et al. (24) observed that growth hormone inhibited insulin-induced Akt activation without affecting insulin-dependent p85 association with IRS-1. Their data suggest that altered localization of p85/IRS-1 complexes was associated with an inability to activate Akt. Perrault et al. (27) observed that cytokine-induced iNOS expression decreased insulin sensitivity in insulin-sensitive tissues. In adipocytes with increased iNOS expression, insulin induced p85 association with IRS-1 and activitation of the p110 catalytic subunit without activating Akt (27). Therefore, even when PI 3-kinase is associated with IRS-1, it does not always activate downstream signaling components. IRS-1 has multiple phosphotyrosine binding sites where p85 can associate. It is possible that all p85/IRS-1 complexes are not equal in their ability to activate Akt. It is also possible that the observed dissociation between the presence of IRS-1/p85 complexes and activation of Akt reflects IL-6-dependent changes at the level of PDK1 or -2 activity, which is proposed to link PI 3-kinase to Akt activation. Thus, our results in Figs. 3 and 6 agree with several other reports that p85/IRS-1 association is not sufficient for Akt activation. In the case of IL-6, a second explanation is possible. While one site of inhibition by IL-6 may be at the level of IR/IRS-1 interaction, there may be a second site of action of IL-6 that inhibits Akt activation distal to IRS-1/PI 3-kinase assocation. It may be this second site of IL-6-dependent Akt inhibition that cannot be overcome by acute exposure to supraphysiological insulin concentrations (Figs. 4 and 7).

Each of the proinflammatory cytokines, TNF-α, IL-6, and IL-1, has recently been implicated in the development of insulin resistance and, possibly, type 2 diabetes (1–6). Many studies have investigated the role of TNF-α-induced insulin resistance in cell lines, animal models, and humans. Hotamisligil et al. (28) provided evidence that serine/threonine phosphorylation of IRS-1 is increased by TNF-α. They also provide evidence that serine/threonine-phosphorylated IRS-1 is an inhibitor of IR signaling (28). Multiple studies using muscle and adipose models support this hypothesis (7,29–35). The validity of this hypothesis in the liver is less certain. In both cell line and animal studies, the effect of TNF-α on hepatic IR signal transduction is frequently small or absent (7,29,36,37). While TNF-α may well have an effect on the liver, an alternate mechanism may be required.

Circulating IL-6 has been shown to be elevated two- to threefold in insulin-resistant states (5,10). The degree of correlation between these levels and severity of insulin resistance is actually higher than that of TNF-α (5). Interestingly, recent evidence suggests that some of the observed effects of TNF-α may be mediated by its ability to induce IL-6 and IL-6 receptor expression in tissues such as the liver and muscle (38,39). Our current report together with a growing body of literature argue that IL-6 plays a role in regulating insulin sensitivity. A recent study with IL-6 knockout mice, however, demonstrates the potential complexities of IL-6 responses in vivo. Wallenius et al. (40) observed that IL-6^−/− mice developed maturity-onset obesity and decreased glucose tolerance compared with lean wild-type controls. Intracerebroventicular but not intraperitoneal IL-6 replacement improved energy expenditure and obesity in these mice. The authors conclude that IL-6 in this model is acting centrally to influence obesity. How does the decreased glucose tolerance in obese IL-6-deficient mice not contradict our hypothesis that IL-6 plays a role in obesity-mediated insulin resistance? It is highly likely that increased adiposity in this IL-6-deficient mouse model releases several factors that contribute to insulin resistance (TNF-α, free fatty acids, etc). Only IL-6 was absent from the array of adipose tissue-derived factors. Thus, the insulin resistance in these animals, compared with lean controls, was not surprising. A comparison of glucose homeostasis and hepatic insulin sensitivity in high-fat diet-induced obese IL-6^−/− mice versus comparably obese controls would have directly addressed the effect of obesity-derived IL-6 on insulin sensitivity in insulin target tissues. This study does demonstrate, however, that IL-6 has complex systemic effects. By using HepG2 cells and primary hepatocytes, the experiments described in the current study isolated the effects on IL-6 on hepatic insulin sensitivity. While this study demonstrates that IL-6 inhibits IR signaling, the molecular mechanism of IL-6-dependent inhibition remains undefined.

IL-6 is known to induce a family of negative regulators of cytokine signal transduction, the suppressors of cytokine signaling (SOCS) (41). Our laboratory and others have shown that SOCS-1, -3, and -6 can directly interact with the IR or IRS-1 and inhibit insulin signal transduction when overexpressed in hepatocytes and other cell lines (42–45). Preliminary investigations in our laboratory have indicated that IL-6-dependent induction of SOCS-3 mRNA and protein in the HepG2 cells corresponds with the time course reported here for IL-6-induced insulin resistance (J.J.S., R.A.M., unpublished observations). Our working hypothesis is that IL-6-dependent insulin resistance is mediated, at least in part, through induction of SOCS protein expression. This model not only can provide a potential mechanism to explain IL-6-dependent inhibition of the IR, but also the transient decrease in IL-6-dependent inhibition of IR signaling at the 2-h time point. IL-6 induces expression of SOCS proteins (specifically SOCS-3 in HepG2 cells), which increase in concentration at ∼1 h. We believe that SOCS proteins are direct mediators of IL-6-dependent inhibition of IR (42). The rise in SOCS concentration correlates with maximum IR inhibition by IL-6 (1 h). SOCS proteins, as their name implies, also suppress IL-6 receptor signaling by antagonizing Janus kinase activation by the receptor. This negative-feedback loop is well characterized (46). Thus, as SOCS proteins suppress IR and IL-6 receptors, they suppress induction of additional SOCS protein. SOCS expression decreases after 1 h and, consequently, we would argue that IL-6-dependent suppression of IR is decreased and IR signaling rebounds at 2 h. We have observed that SOCS3 mRNA peaks at 1 h (approximately eightfold), decreases toward basal levels at 2 h (less than twofold elevated), and increases again to two- to threefold by 4 h, consistent with a return of IL-6-dependent inhibition of the IR signaling. SOCS3 protein levels similarly peak at 1 h. Although not proven, we believe that IL-6-dependent SOCS expression provides a molecular explanation for IL-6-dependent inhibition of IR signaling and the temporary diminution of the IL-6 effect at 2 h.

Shoelson and coworkers (47,48) have provided evidence that salicylates can alleviate obesity-dependent and TNF-α-induced insulin resistance. Based on evidence that salicylates inhibit NF-κB inhibitor (IκB) kinase, a serine/threonine kinase in the nuclear factor (NF)-κB regulatory pathway, these investigators examined obesity-dependent insulin resistance in IκB knockout mice (48). They observed significantly decreased insulin resistance in the knockout mice model and proposed that IκB kinase is essential in mediating the serine/threonine phosphorylation of IRS-1 induced by TNF-α. While these studies are an important advance, an alternate interpretation of the results is possible. The TNF-α receptor activates IκB kinase, which releases the NF-κB transcription factor from inhibition. TNF-α-dependent induction of IL-6 is mediated through the NF-κB pathway. An interesting possibility is that the ability of salicylates to alleviate insulin resistance is, at least in part, mediated through its inhibition of TNF-α-dependent IL-6 induction. The inhibition of this induction may in turn lower expression of the SOCS proteins.

The data presented in this study indicate that chronic exposure to IL-6 inhibits IR signal transduction in both the HepG2 hepatocarcinoma cell line and primary mouse hepatocytes. The IL-6 effect is characterized by an inhibition of physiologic insulin-dependent IRS-1 tyrosine phosphorylation and association with the p85 subunit of PI 3-kinase, as well as a near complete inhibition of insulin-induced Akt activation in the HepG2 cell line. In the primary hepatocyte, IL-6 also inhibits Akt activation in response to insulin and inhibits a key physiological response to insulin, glycogen synthesis. While these data suggest that one site of action of IL-6 is at the level of IR/IRS-1 interaction, the precise mechanism for IL-6-dependent insulin resistance in hepatocytes has not been defined.

This work was supported by U.S. Public Health Service Grant R01-38138 (to R.A.M.).