JNK and tumor necrosis factor-alpha mediate free fatty acid-induced insulin resistance in 3T3-L1 adipocytes.

Lipid infusion and high fat feeding are established causes of systemic and adipose tissue insulin resistance. In this study, we treated 3T3-L1 adipocytes with a mixture of free fatty acids (FFAs) to investigate the molecular mechanisms underlying fat-induced insulin resistance. FFA treatment impaired insulin receptor-mediated signal transduction and decreased insulin-stimulated GLUT4 translocation and glucose transport. FFAs activated the stress/inflammatory kinases c-Jun N-terminal kinase (JNK) and IKKbeta, and the suppressor of cytokine signaling protein 3, increased secretion of the inflammatory cytokine tumor necrosis factor (TNF)-alpha, and decreased secretion of adiponectin into the medium. RNA interference-mediated down-regulation of JNK blocked JNK activation and prevented most of the FFA-induced defects in insulin action. Blockade of TNF-alpha signaling with neutralizing antibodies to TNF-alpha or its receptors or with a dominant negative TNF-alpha peptide had a partial effect to inhibit FFA-induced cellular insulin resistance. We found that JNK activation by FFAs was not inhibited by blocking TNF-alpha signaling, whereas the FFA-induced increase in TNF-alpha secretion was inhibited by RNA interference-mediated JNK knockdown. Together, these results indicate that 1) JNK can be activated by FFAs through TNF-alpha-independent mechanisms, 2) activated JNK is a major contributor to FFA-induced cellular insulin resistance, and 3) TNF-alpha is an autocrine/paracrine downstream effector of activated JNK that can also mediate insulin resistance.

Insulin resistance is characterized by a reduced ability of insulin to regulate glucose homeostasis in target tissues and is commonly associated with obesity, often preceding the clinical diagnosis of type 2 diabetes. In insulin-resistant states, signal transduction via the insulin receptor (IR) 2 is impaired with decreased activation of downstream obligate molecular intermediates, such as IRS-1, Akt, and PKC (1), that are involved in stimulating translocation of GLUT4 proteins to the cell surface.
Many factors have been reported to induce insulin resistance in vitro and in vivo, including lipid-and fat-derived free fatty acids (FFAs) (2). Lipid infusion and high fat feeding promote insulin-resistant states in rodents (3) and humans (4,5), and elevated plasma FFA concentrations typically correlate with obesity and decreased target tissue insulin sensitivity in humans. Although the underlying mechanisms are still unclear, evidence suggests that oversupplying FFAs causes intracellular accumulation of FFA-derived metabolic products (6 -8). These can activate the serine/threonine kinases stress-activated protein kinase or JNK (stress-activated protein or c-Jun amino-terminal kinases) (9), IKK␤ (10,11), and PKC (6,11,12), all of which can phosphorylate IRS-1 on serine residues. Consequently, IRS-1 activation through tyrosine phosphorylation is impaired, leading to a reduction in IR-mediated signaling and subsequent insulin resistance. Lipid infusion and high fat feeding also impair PI 3-kinase, Akt, and PKC/ activation in muscle (13,14), and a defect in muscle PI 3-kinase and PKC/ activity is also observed in obese and diabetic subjects (15).
The proinflammatory cytokine TNF-␣ is another important contributor to the development of insulin resistance (16). TNF-␣ levels are elevated in adipose tissue of various rodent obesity models and in obese humans (16,17), whereas a genetic defect in TNF-␣ signaling significantly improves IR signaling capacity and insulin sensitivity in dietinduced and genetically obese mice (18). How TNF-␣ causes insulin resistance in adipose and skeletal muscle tissues is, however, not yet fully understood. Several molecular mechanisms have been proposed that also involve the activation of JNK (19,20), PKC (20), and IKK␤ (10,20,21), and the resulting serine phosphorylation of IRS-1 and inhibition of IR kinase activity. TNF-␣ can also down-regulate the secretion of adiponectin, an effect that could cause insulin resistance (22). Adiponectin is an abundant circulating cytokine secreted by adipocytes that sensitizes liver and skeletal muscle to insulin in rodents and humans. Adiponectin gene expression in both primary human preadipocytes and differentiated 3T3-L1 cells is suppressed by TNF-␣ treatment (23).
In the present study, we treated 3T3-L1 adipocytes with a mixture of FFAs to study the molecular mechanisms underlying hyperlipidemiainduced insulin resistance in adipose tissue. We demonstrated that FFAs induce insulin resistance by activating JNK and by increasing the secretion of TNF-␣, which in an autocrine/paracrine fashion acts on cells to impair adipocyte function. Blocking JNK activation rescued the cellular and molecular defects induced by FFAs, whereas blocking TNF-␣ signaling only rescued some of these defects. Furthermore, blocking TNF-␣ signaling did not prevent activation of JNK by FFAs, indicating that TNF-␣ is a key but not the sole downstream effector of JNK-mediated, FFA-induced insulin resistance in 3T3-L1 adipocytes.

EXPERIMENTAL PROCEDURES
Materials-Antibodies against adiponectin and phosphotyrosine were purchased from Affinity Bioreagents (Golden, CO) and Transduction Laboratories (Lexington, KY), respectively. Antibodies to Akt1/2, IR␤, and NF-B, horseradish peroxidase-linked secondary antibodies, neutralizing antibodies to TNF-␣ and TNF receptors RI and RII, and control unconjugated Armenian hamster IgG were from Santa Cruz Biotechnology, Inc. (Santa Cruz, CA); SOCS-1 and SOCS-3 antibodies were from Abcam (Cambridge, MA); and all other antibodies were from Cell Signaling (Beverly, MA).
Dulbecco's modified Eagle's medium and fetal bovine serum were purchased from Invitrogen; radioisotope was from ICN (Costa Mesa, CA). Recombinant p110 CAAX and control recombinant adenoviruses were prepared as previously described (28).
Cell Culture and Treatment-Mouse 3T3-L1 preadipocytes (American Type Culture Collection, Manassas, VA) were cultured and differentiated into adipocytes as previously described (29). Unless indicated otherwise, adipocytes were used 10 -12 days after differentiation. After overnight incubation in serum-free Dulbecco's modified Eagle's medium supplemented with 0.1% FFA-free bovine albumin (FFA-free BSA; ICN Biomedicals, Aurora, OH), 3T3-L1 adipocytes were treated with various concentrations of an FFA mixture composed of lauric, myristic, linoleic, oleic, and arachidonic acids (Sigma) or ethanol/PBS vehicle in the presence of 0.1% FFA-free BSA, for 1-6 h at 37°C. FFAfree BSA was also used at 2.0% to simulate physiological serum albumin levels. These correspond to 20:1 and 3:1 ratios of FFA/BSA, respectively, given 300 M FFA in 0.1% BSA and 1 mM FFA in 2.0% BSA, respectively.
Western Blotting and Enzyme-linked Immunosorbent Assay (ELISA) Analysis-Serum-starved 3T3-L1 adipocytes were incubated with or without FFA or insulin stimulation (bovine; Sigma) as indicated in each experiment. After harvesting conditioned medium for analysis, cells were lysed in cold buffer containing 50 mM HEPES, pH 7.4, 150 mM NaCl, 200 mM NaF, 20 mM sodium pyrophosphate, 10% glycerol, 1% Triton X-100, 4 mM sodium orthovanadate, 2 mM phenylmethylsulfonyl fluoride, and 1 mM EDTA. For Western blot analysis, whole cell lysates (20 g) or conditioned media (30 l) were resolved by SDS-PAGE and transferred onto polyvinylidene difluoride membranes (Immobilon-P; Millipore Corp., Bedford, MA). Following blocking, membranes were probed with specific antibodies and subsequently incubated with horseradish peroxidase-linked secondary antibodies for chemiluminescent detection (Pierce). Blots were stripped in Restore TM Western blot stripping buffer (Pierce) and reprobed as indicated. Conditioned medium (50 l) was also assayed for mouse TNF-␣ using an ELISA kit (BIO-SOURCE, Camarillo, CA) following the manufacturer's protocol.
GLUT4 Translocation Assay-This is a modified version of the method previously published (30). Briefly, on day 5 postdifferentiation, 1 ϫ 10 8 3T3-L1 adipocytes were electroporated with a plasmid containing the rat GLUT4 cDNA tagged with an HA epitope (in the first exofacial loop) and with GFP (at the carboxyl terminus; a generous gift from Dr. T. E. McGraw) using the Gene Pulser XCell (Bio-Rad) and plated onto coverslip bottom dishes (MaTek, Ashland, MA). Two days postelectroporation, the adipocytes were treated with or without 500 M FFA for 3 h in Dulbecco's modified Eagle's medium containing 0.1 or 2.0% BSA. The cells were then stimulated with 170 nM insulin for the indicated durations and fixed in 3.7% paraformaldehyde. HA-GLUT4-GFP was stained with mouse monoclonal anti-HA.11 (Covance, Princeton, NJ) followed by Cy3-conjugated secondary antibody (Jackson Immunolabs, West Grove, PA). Fluorescence quantification was performed on a Nikon TE300 inverted microscope using a Nikon ϫ40 numerical aperture 1.3 oil objective, a TILL Photonics II monochromator, and a 12-bit Hamamatsu Orca CCD camera using Simple PCI software (Hamamatsu, Bridgewater, NJ). At least 20 images were taken per time point, in which transfected cells were selected based on GFP expression.
Calculations-In transfected cells, for each time point, the intensities of the GFP and Cy3 signals were quantified, and background GFP and Cy3 fluorescent emissions were subtracted. The Cy3 fluorescence intensity for each HA-GLUT4-GFP-expressing cell (reflecting cell surface HA-GLUT4-GFP) was divided by the GFP fluorescence intensity (reflecting total cellular HA-GLUT4-GFP level) to determine the fraction of tagged GLUT4 transporter at the membrane (arbitrary units). Translocation following insulin stimulation was expressed as -fold translocation over basal.
2-Deoxyglucose (2-DOG) Uptake-The assay for glucose uptake was described previously (29) with some modifications. Following overnight serum starvation, adipocytes were glucose-starved for 1 h in Hepes-salt buffer containing 0.1 or 2% FFA-free BSA. FFA was then added to the cells at the indicated concentrations for the indicated times. 10 min before the end of the FFA treatment, the cells were stimulated with insulin (1.7 or 17 nM) at 37°C. Tracer glucose was then added for 10 min. After 20 min of insulin stimulation, glucose uptake was assayed in triplicate or quadruplicate wells for each condition using In experiments where cells were allowed to recover from the FFA treatments, the FFA-containing medium was removed, and the cells were incubated in fresh medium for 12 h prior to insulin stimulation and 2-DOG uptake assay. In experiments involving p110 CAAX overexpression, 3T3-L1 adipocytes were transduced at a multiplicity of infection of 10 plaque-forming units/cell for 16 h at 37°C with stocks of either p110 CAAX or control empty vector recombinant adenoviruses. At the end of the infection period, medium was replaced, and 48 h later, the cells were incubated in the starvation medium required for the 2-DOG uptake assay (see above).
JNK Protein Knockdown-On day 5 postdifferentiation, 1 ϫ 10 7 3T3-L1 adipocytes were transfected by electroporation with 2.5 nmol of high pressure liquid chromatography-purified siRNA oligonucleotides to mouse JNK2 or scrambled siRNA or equal volumes of siRNA resuspension buffer (Qiagen, Valencia, CA) and plated into 12-and 24-well tissue culture plates. 24, 48, and 72 h postelectroporation, cells were lysed and subjected to SDS-PAGE and Western blot analysis to evaluate intracellular JNK protein levels. In parallel, transfected 3T3-L1 adipocytes were serum-and glucose-starved, treated with FFA (500 M, 3 h; 1 mM, 1 h) with or without insulin stimulation (17 nM, 20 min), and assayed for glucose uptake. Cell lysates were also prepared for Western blotting analysis of components of the insulin signaling cascade.
Functional Rescue Experiments-Prior to the addition of FFAs, each well in a 12-well culture plate was supplemented with DN TNF-␣ peptide (2 g), control peptide (2 g), TNF-␣-neutralizing antibody (2 g), TNF-␣ receptor RI and RII neutralizing antibodies (1 g each), or control Armenian IgG (2 g), as indicated. The final concentrations of peptide and antibodies were 2 g/ml. Experiments were also performed at final peptide and antibody concentrations of 10 g/ml and peptide concentration of 25 g/ml.
Calculations-In 2-DOG uptake assays, the percentage rescue value was calculated as the ratio of the difference in glucose uptake between insulin-stimulated, untreated cells and insulin-stimulated FFA-treated cells over the difference in glucose uptake between insulin-stimulated, untreated cells and unstimulated, FFA-treated cells.
Data Analysis-Values presented are expressed as means Ϯ S.E. The statistical analysis was performed using GraphPad Prism version 3 (GraphPad Software Inc., San Diego, CA). The statistical significance of the differences between various treatments was determined by one-way analysis of variance with the Bonferroni correction. The ␣ level was set at 0.05.

RESULTS
To better define how lipid infusion and high fat feeding cause insulin resistance, we used cultured 3T3-L1 adipocytes as a model system to investigate the mechanisms whereby FFAs induce cellular insulin resistance. We treated 3T3-L1 adipocytes with a mixture of saturated (lauric and myristic acid) and unsaturated (arachidonic, oleic, and linoleic acid) FFAs at various concentrations and for various amounts of time and examined the impact on insulin signaling and adipocyte function.
Down-regulation of IR-mediated Signaling by FFAs-Serum-starved 3T3-L1 adipocytes were incubated with various concentrations of FFAs, ranging from 300 M to 1.0 mM, or vehicle control for 1-6 h in the presence of 0.1% BSA, with or without insulin stimulation (1.7 nM). Activation of the insulin signaling cascade was assessed in lysates of FFA-treated and control cells by SDS-PAGE and immunoblotting. Fig.  1A shows representative results obtained from a 1-h treatment with 1 mM FFA. Similar results were obtained with 500 and 300 M FFA treatments for 3 and 6 h, respectively, in the presence of 0.1 and 2.0% BSA and with 1.7 and 17 nM insulin stimulation (data not shown).
Incubation with FFAs inhibited insulin-stimulated tyrosine phosphorylation of IR␤ and also caused a decrease in IR␤ protein levels (ϳ40%). Similarly, insulin-stimulated IRS-1 tyrosine phosphorylation and protein levels (ϳ40%) were reduced by FFAs. Downstream of IRS-1, insulin-stimulated Akt and PKC phosphorylation was inhibited by FFA treatment, indicating an insulin-resistant state. Total Akt and PKC protein levels were unaffected.
Activation of Stress Kinases by FFAs-Since inflammation has been linked to insulin resistance (31), we examined whether the stress/inflammatory kinases JNK and IKK are activated in 3T3-L1 adipocytes treated with FFAs. As demonstrated in Fig. 1B, regardless of insulin stimulation, FFA treatment dramatically increased phosphorylated JNK, despite lowering total JNK protein levels. IKK␤ was also phosphorylated upon FFA treatment. Note the FFA treatment did not activate FIGURE 1. Effects of FFA treatment on intracellular signaling. Serum-and glucose-starved 3T3-L1 adipocytes were pretreated with 1 mM FFA for 1 h and stimulated with 1.7 nM insulin as indicated. Cell lysates were prepared and analyzed by SDS-PAGE and immunoblotting. The activated state of these signaling molecules was assessed using phosphotyrosine (pY) and phospho (p)-specific antibodies. A, FFAs inhibit the IR signaling cascade by down-regulating the activation of key signaling intermediates IR␤, IRS-1, Akt, and PKC. B, FFAs activate stress and inflammatory kinases. FFA treatment caused the phosphorylation of JNK and IKK␤ but not IKK␣ and the subsequent phosphorylation of the inhibitor of NF-B, IB␣. NS, nonspecific. Representative results from six independent experiments. IKK␣ but decreased IKK␣ protein levels. This is in agreement with a previous report (32) showing that IKK␣ protein levels were lowered by 1 h of treatment with TNF-␣. Consistent with IKK␤ activation, IB␣, the inhibitor of NF-B, was phosphorylated after 1 h of treatment with 1 mM FFA. The IB␣ and NF-B protein levels were not altered by this short incubation, in contrast to longer treatments (data not shown).
Inhibition of Insulin-stimulated GLUT4 Translocation and Glucose Uptake by FFAs-To further define the effects of FFA treatment on insulin action, we examined the membrane targeting kinetics of the insulin-regulated GLUT4 transporter. We employed a single-cell GLUT4 translocation assay (30) to directly quantify the amount of GLUT4 transporter targeted to the plasma membrane upon insulin addition. This technique relies on the transient transfection of a GLUT4 construct tagged with an HA epitope in the first extracellular loop and GFP in the intracellular C-terminal region (HA-GLUT4-GFP) into cells. For a given cell, the extent of GLUT4 translocation to the plasma membrane can be quantitated by normalizing the immunofluorescent labeling of the extracellular HA epitope against total cellular GFP content.
As illustrated in Fig. 2A, control cells responded to insulin by translocating HA-GLUT4-GFP to the cell surface in a biphasic mode, as previously reported; their cell surface GLUT4 increased by ϳ2.8-fold at 5 min poststimulation, which persisted up to 25 min after insulin addi- In recovery experiments, FFA-containing medium was removed, and 3T3-L1 adipocytes were fed with fresh medium. Glucose uptake was assessed 12 h later in the presence of insulin, as above. Glucose uptake was expressed relative to uptake levels found in untreated, unstimulated adipocytes (white bars). White bars, basal; light gray bars, insulin; dark gray bars, FFA; black bars, FFA ϩ insulin; hatched bars, FFA washout ϩ insulin. C, FFAs act upstream of PI 3-kinase and/or in a PI 3-kinase-independent pathway to inhibit glucose transport. 3T3-L1 adipocytes were infected with an adenovirus encoding a constitutively active form of the catalytic subunit of PI 3-kinase (p110 CAAX) or a control adenovirus 2 days prior to FFA treatment (1 mM, 1 h), and glucose uptake (1.7 nM insulin stimulation) was assayed as above. All bar graphs are representative of 4 -6 independent experiments for each condition in quadruplicate. a.u., arbitrary units.
tion. At 15 min, a maximal 3.8-fold increase in membrane GLUT4 was observed. By contrast, FFA-pretreated cells (500 M for 3 h) failed to translocate GLUT4 at 5 and 10 min poststimulation and only showed a 50% increase in cell surface GLUT4 at later time points. Thus, FFA treatment interfered with both phases of insulin-induced GLUT4 translocation. Note that FFA treatment, while impairing insulin-stimulated GLUT4 translocation, increased basal translocation by ϳ2-fold (data not shown).
Because FFA-treated cells exhibited defective insulin-stimulated GLUT4 translocation, we predicted that glucose uptake would also be impaired. We therefore assayed both basal and insulin-stimulated (1.7 nM) glucose uptake in these cells. Fig. 2B shows that FFA treatments (1 mM for 1 h, 0.5 mM for 3 h, and 0.3 mM for 6 h) inhibited ϳ70 -90% of the glucose uptake stimulated by 1.7 nM insulin. A similar inhibition was observed at 17 nM insulin (data not shown). This inhibition was fully reversed following FFA washout (Fig. 2B, hatched bars), indicating that the FFA treatment was not causing a nonspecific toxic effect. Of note is the FFA-induced increase in basal glucose transport (by ϳ70%), which is consistent with the GLUT4 translocation results ( Fig. 2A).
To define the loci in the insulin signaling cascade where FFA-induced inhibition of insulin-stimulated glucose transport occurs, we utilized an adenovirus encoding a constitutively active form of the catalytic subunit of PI 3-kinase, termed p110 CAAX (28), to augment PI 3-kinase-mediated glucose uptake. In 3T3-L1 adipocytes, p110 CAAX adenovirusmediated gene transfer leads to p110 CAAX expression and stimulates 2-DOG uptake by ϳ4-fold in the absence of insulin (28). As illustrated in Fig. 2C, p110 CAAX expression had the expected effect of stimulating glucose uptake by 4-fold, with insulin treatment producing an additive effect to 9-fold stimulation. When the p110 CAAX-infected cells were treated with FFAs and stimulated with insulin, only the insulin-stimulated component of glucose uptake was substantially inhibited. This inhibitory effect was comparable with that seen in control cells infected with control adenovirus. These results suggest that the effects of FFAs on insulin-stimulated glucose transport are independent of PI 3-kinase.
FFAs Decrease Adiponectin Secretion-Since low circulating levels of adiponectin also correlate with inflammatory states and insulin resistance (33), we hypothesized that FFA treatment down-regulates adiponectin secretion directly in adipocytes.
Conditioned medium was harvested from FFA-treated (1 mM, 1 h) and untreated cells for SDS-PAGE and adiponectin immunoblotting analysis (Fig. 3A). As shown in Fig. 3B, FFA-treated cells secreted ϳ35 and 15% less adiponectin without and with insulin stimulation, respectively, despite similar intracellular adiponectin protein content (Fig.  3A), indicating that FFA treatment inhibits adiponectin release at the secretory stage. Moreover, FFA treatment inhibited the increase in adiponectin secretion induced by insulin stimulation, decreasing it from 155 to 85% in the presence of FFAs. Similar results were obtained with 300 M (6-h) and 500 M (3-h) FFA treatments (data not shown).
Inhibition of FFA-induced JNK Activation-To test whether FFAinduced JNK kinase activation mediates FFA-induced cellular insulin resistance, we blocked the FFA-induced activation of the JNK pathway via RNA interference. We electroporated 3T3-L1 adipocytes with two independent siRNAs to mouse JNK2 (mJ2A and mJ2B) and measured JNK2 protein levels 48 h later (Fig. 4A, 24 and 72 h data not shown). Both siRNAs effectively decreased endogenous JNK2 protein levels without altering intracellular levels of actin or other intracellular proteins (not shown); they also decreased JNK1 protein levels. To exclude potential nonspecific effects due to the transfection of siRNA, we transfected two different scrambled siRNAs and showed that they had no effect on lowering JNK protein levels (data not shown). mJ2B was more effective than mJ2A at down-regulating JNK1/2 protein levels and was thus utilized for subsequent JNK knockdown experiments unless otherwise specified.
As expected, markedly reduced FFA-induced JNK activation was observed in the siRNA-treated cells (Fig. 4B, lanes 2 and 3 versus lane 1).
Inhibition of JNK Activation Rescues the FFA-induced Impairment in Insulin-stimulated Glucose Transport-We next examined the effect of siRNA-mediated JNK knockdown on FFA-induced cellular insulin resistance. Adipocytes electroporated with control suspension buffer or scrambled or JNK siRNA were starved and treated with 1 mM FFA for 1 h with or without insulin stimulation prior to the glucose transport assay. As demonstrated in Fig. 5A, mJ2B siRNA reversed the FFA-induced impairment in insulin response by ϳ95-100% (p Ͻ 0.01) in 3T3-L1 adipocytes compared with cells electroporated with suspension buffer or scrambled siRNA. In other words, glucose transport was greatly increased in FFA ϩ insulin-treated JNK knockdown adipocytes compared with FFA ϩ insulin-treated control cells. mJ2B produced a more complete rescue than mJ2A (ϳ75% rescue; data not shown), consistent with its higher efficiency at lowering intracellular JNK protein levels.
Inhibiting JNK Activation Improves Insulin Signaling in FFA-treated Adipocytes-We further investigated the effects of JNK activation on insulin action by studying insulin signaling in siRNA-transfected 3T3-L1 adipocytes. Fig. 5B compares the activation state of two proximal components of the insulin signaling cascade (IR␤ and Akt) in mocktransfected (left column) and mJ2B-transfected (right column) cells following FFA and insulin treatments.
As shown in Fig. 1, FFA treatment led to inhibition of insulin-stimulated tyrosine phosphorylation of IR␤ and Ser 473 and Thr 308 phospho- rylation of Akt. In contrast, in siRNA-treated adipocytes, where FFAinduced JNK activation was inhibited, these FFA effects on IR␤ and Akt phosphorylation were markedly attenuated.
FFAs Increase the Secretion of TNF-␣ by 3T3-L1 Adipocytes-We further searched for a downstream effector of JNK involved in FFAinduced insulin resistance. Studies have suggested a correlation between high FFA levels, low adiponectin levels, and high TNF-␣ levels (22), and TNF-␣ gene expression has been shown to be regulated by JNK activity (34). Given that FFAs impair IR-mediated signaling and activate JNK, we assessed whether FFAs induce insulin resistance by augmenting the secretion of TNF-␣. Secreted TNF-␣ in conditioned medium from FFA-treated (1 mM, 1 h) and control 3T3-L1 adipocytes was measured by ELISA. Fig. 6 shows that FFA treatment increased TNF-␣ secretion by ϳ80% (p Ͻ 0.001 as calculated by one-way analysis of variance). Increases were also detected after 500 M (3-h) and 300 M (6-h) FFA treatments (data not shown). This raises the possibility that TNF-␣ is responsible for FFA-induced inhibition of insulin action in 3T3-L1 adipocytes.
Blocking TNF-␣ Signaling Partially Reverses FFA-induced Inhibition of Insulin-stimulated Glucose Uptake-We determined whether the decreased insulin sensitivity could be rescued by blocking TNF-␣ signaling. To accomplish this, we assessed the effect of FFAs (1 mM, 1 h) on glucose transport in 3T3-L1 adipocytes that were pretreated with antibodies (2 g/ml) that neutralize either TNF-␣ or its receptors or with nonimmune hamster immunoglobulin (2 g/ml) as control. Fig. 7A shows that neutralizing antibodies against either TNF-␣ or its receptors restored the insulin responsiveness of glucose transport by ϳ50% in FFA-treated cells. Similar results were obtained using 10 g/ml neutralizing antibodies (data not shown). The percentage of rescue was calculated as described under "Experimental Procedures." Interestingly, the neutralizing antibodies also slightly increased basal glucose uptake regardless of FFA treatment, suggesting an inhibitory autocrine/paracrine effect of TNF-␣ in the basal state.
We confirmed these findings in complementary rescue experiments using a TNF-␣ peptide with dominant negative (DN) properties. DN TNF variants are novel TNF-␣-blocking agents engineered to structurally resemble TNF-␣ but contain mutations that alter their ability to bind and signal through TNF receptors (35). DN TNF peptides inhibit TNF-␣-mediated signal transduction by rapidly heterotrimerizing with soluble TNF-␣, sequestering native ligand into inactive complexes. XENP345, the DN TNF peptide utilized in our experiments, encompasses two point mutations (A145R and I97T) that disrupt receptor interactions and a polyethylene glycol modification for increased stability. This protein was shown to poorly bind to TNF receptors and lack biological activity in vivo (35). We used a DN RANKL peptide as control. RANKL is a member of the TNF superfamily of proteins, whose structure is similar to that of TNF-␣. However, RANKL does not trimerize with TNF-␣; thus, DN RANKL cannot block TNF-␣ signaling.
As above, the effect of FFAs (1 mM, 1 h) on glucose transport in 3T3-L1 adipocytes pretreated with DN TNF or control peptide (2 g/ml) was assessed. As shown in Fig. 7B, incubation with DN TNF peptide rescued the FFA-induced inhibition of insulin-stimulated glucose transport by ϳ50%. By comparison, no rescue was observed with  control peptide. Similar results were obtained using 10 and 25 g of peptide/ml (data not shown).
Blocking TNF-␣ Signaling and Insulin Action-We examined the effect of blocking TNF-␣ signaling on the activation and protein levels of downstream intermediates of the insulin signaling cascade IR␤, IRS-1, and Akt. As shown in Fig. 7C, blocking TNF-␣ signaling using a neutralizing antibody to TNF-␣ did not prevent the inhibition of insulin-stimulated tyrosine phosphorylation of IR␤ and IRS-1 by FFAs. Similarly, reduced insulin-stimulated Akt phosphorylation by FFAs was not prevented by antibody pretreatment. However, anti-TNF-␣ pretreatment prevented the decrease in IR␤ and IRS-1 protein levels caused by FFA treatment, suggesting that one mechanism whereby TNF-␣ induces insulin resistance involves degradation of components of the insulin signaling cascade.
Taken together, these results indicated that TNF-␣ is not the sole downstream effector of JNK-mediated FFA-induced insulin resistance.
Blocking TNF-␣ Signaling Rescues the FFA-induced Decrease in Adiponectin Secretion-Since FFA treatment inhibited adiponectin secretion (Fig. 3), we assessed whether the inhibition could also be prevented by blocking TNF-␣ signaling. We added TNF-␣ neutralizing or control antibody to 3T3-L1 adipocytes and determined the effect of FFA treatment (1 mM, 1 h) on adiponectin secretion by blotting conditioned medium with an adiponectin antibody. Fig. 7D shows that the pretreatment prevented FFAs from inhibiting adiponectin secretion both in the absence and presence of insulin. Intracellular adiponectin levels were not altered by these pretreatments. Similar rescuing effects were obtained in cultures pretreated with TNF-␣ receptor-neutralizing antibodies (data not shown).
Effect of FFAs on SOCS-3-SOCS proteins have been implicated by recent reports as regulators of insulin action, and SOCS-1 and -3 expression and activation have been correlated with insulin resistance, obesity, and high TNF-␣ levels (24). We therefore assessed the effects of FFAs on SOCS-1 and SOCS-3. We found that FFA treatment altered the pattern of SOCS-3 protein expression. FFA increased SOCS-3 total protein by 31 Ϯ 16%, particularly elevating the levels of high molecular weight SOCS-3, which is probably the phosphorylated and activated form of the protein (Fig. 8A). In contrast, FFA treatment did not appear to affect SOCS-1 expression. To our knowledge, this is the first demonstration of a stimulus or treatment regulating SOCS-3 protein expression/activity in 3T3-L1 adipocytes.
This SOCS-3 effect of FFAs was not inhibited when TNF-␣ signaling was abrogated using neutralizing antibodies against TNF-␣ (lanes 5 and  6 versus lanes 3 and 4), indicating that FFAs modulate SOCS-3 through a TNF-␣-independent pathway. Interestingly, this change in SOCS-3 protein expression was markedly inhibited in JNK siRNA-treated cells (Fig. 8B). These FFA-induced decreases in SOCS-3 protein are specific, since scrambled siRNA did not have such an effect, and JNK downregulation by RNA interference did not affect the levels of actin or other proteins (data not shown), as above.
JNK Is Upstream of TNF-␣ in Mediating the Effects of FFAs-The SOCS-3 data in conjunction with the results from Figs. 5 and 7 suggest that JNK is a central mediator of the FFA effects and that TNF-␣ is one downstream effector of JNK. In other words, JNK activation by FFAs is upstream of TNF-␣ action, and there are JNK-mediated effects on insulin resistance that are TNF-␣-independent. To test this idea, we examined the effect of TNF-␣ blockade on JNK activation in FFA-treated adipocytes. 3T3-L1 adipocytes were starved and pretreated with neutralizing antibody to TNF-␣ (2 g/ml) prior to FFA addition (1 mM, 1 h). Cell lysates were harvested and analyzed by SDS-PAGE and immunoblotting. The results in Fig. 9A show that inhibition of TNF-␣ signaling did not prevent activation of JNK by FFAs. Similar results were obtained using neutralizing antibodies to TNF-R and DN TNF peptide (data not shown).
We also determined the effect of JNK inhibition on TNF-␣ secretion. 3T3-L1 adipocytes were electroporated with JNK siRNA and treated with FFA (1 mM, 1 h; 500 M, 3 h) 48 h later. Conditioned medium was collected for quantification of TNF secretion by ELISA. The results in Fig. 9B show that siRNA-mediated inhibition of JNK activation markedly decreased secretion of TNF-␣.

DISCUSSION
In the present study, we used 3T3-L1 adipocytes as an in vitro model system to assess the cellular mechanisms of insulin resistance initiated by elevated FFA levels. As such, these studies provided insights into the mechanisms of insulin resistance induced by lipid infusion and high fat feeding.
We treated 3T3-L1 adipocytes with a mixture of FFAs rather than individual FFAs to better recapitulate the metabolic consequences of high fat feeding and lipid infusion in animal models at concentrations comparable with those found in vivo. Although 3T3-L1 adipocytes can secrete fatty acids into the medium through lipolysis of intracellular triglyceride stores, this cell type is amenable for studies of FFA-induced cellular insulin resistance. Indeed, compared with the concentrations of exogenous FFAs used in our studies and compared with circulating physiologic FFA concentrations, the amount of FFAs released by 3T3-L1 adipocytes is negligible. We measured the concentrations of FFA secreted into the 3T3-L1 adipocyte culture medium in the absence of exogenous FFAs and found that these were less than 0.1 mM. Thus, essentially all of the effects observed in our studies were the result of exogenously added FFAs.
We found that in 3T3-L1 adipocytes, FFA treatment impaired insulin signaling at multiple steps. Phosphorylation of key downstream components, such as IR␤, IRS-1, Akt, and PKC, was inhibited, reducing the amplitude of the insulin signal. IR and IRS-1 protein levels were also down-regulated by FFA treatment, further dampening signal transduction. Consistent with these defects in insulin signaling, FFA treatment caused decreased insulin-stimulated GLUT4 translocation and decreased glucose transport. These effects were fully reversed after removing FFAs from the medium (Fig. 2B), demonstrating that FFA treatment does not produce nonspecific, toxic effects. The rapid reversibility of these effects further suggests that FFA-induced resistance may not require changes in gene expression. With respect to mechanisms, we found that adenovirus-mediated expression of a constitutively active FIGURE 6. FFAs increase the secretion of TNF-␣. Starved 3T3-L1 adipocytes were treated with 1 mM FFA for 1 h. Conditioned medium was collected and analyzed for TNF-␣ content using an ELISA. Levels of secreted TNF-␣ were expressed relative to basal TNF-␣ levels from non-FFA-treated adipocytes. 12 samples were analyzed for each condition, and statistical significance was achieved (p Ͻ 0.001). FIGURE 7. Blocking TNF-␣ signaling and insulin action. Serum-and glucose-starved 3T3-L1 adipocytes were pretreated with 2 g/ml neutralizing antibodies to TNF-␣ or TNF receptors (A) or a DN TNF peptide (B). Cells were then treated with 1 mM FFA for 1 h, with or without subsequent insulin stimulation (1.7 nM). A, glucose transport was assessed by 2-DOG uptake as above. The percentage of rescue is calculated as described under "Experimental Procedures." Neutralizing antibodies to TNF receptors and TNF-␣ rescued the FFA-induced defect in insulin-stimulated glucose transport by ϳ50 -56% (*, p Ͻ 0.005). This was not observed with control IgG. B, similarly, a DN TNF peptide rescued the defect in insulin-stimulated glucose transport by ϳ50% (*, p Ͻ 0.005). Control peptide did not produce this effect. The bar graphs are representative of four independent experiments for each condition in triplicate. a.u., arbitrary units. White bars, basal; light gray bars, insulin; dark gray bars, FFA; black bars, FFA ϩ insulin. C, cell lysates were prepared following the FFA and insulin treatment for SDS-PAGE and immunoblotting analysis using phosphotyrosine (pY), phosphospecific, and protein-specific antibodies. Blocking TNF-␣ signaling rescued the FFA-induced decrease in IR␤ and IRS protein levels but did not prevent FFA-induced decrease in IR␤, IRS, and Akt phosphorylation. D, blocking TNF-␣ signaling prevents the FFA-induced decrease in secreted adiponectin. 2 g/ml neutralizing antibodies to TNF-␣ or control IgG were added to starved 3T3-L1 adipocytes prior to FFA treatment. Conditioned medium was collected, and cells were lysed for Western blot analysis of secreted and intracellular adiponectin, as above. All results shown are representative of three-six independent experiments. form of the catalytic subunit of PI 3-kinase, p110 CAAX, stimulated glucose transport in the absence of insulin and that FFA treatment inhibited the insulin effect on glucose transport above the p110 CAAX effect. These results suggest that the FFA effects are exerted independently of PI 3-kinase. Concomitantly, FFA treatment caused activation of the stress/inflammatory kinases JNK and IKK␤, and activation of the stress/inflammatory pathways is associated with insulin resistance (9,10,31).
Adiponectin is an adipokine that sensitizes skeletal muscle and liver toward insulin, and low levels of circulating adiponectin have also been associated with decreased whole body insulin sensitivity and insulin resistance (22,33). Our results show that FFA treatment decreased adiponectin release from adipocytes at the secretory step, without altering intracellular adiponectin levels (Fig. 3). This finding is novel, demonstrating that adipocytes, without cues from other cell types, can respond directly to FFA treatment by attenuating adiponectin secretion. If such an effect of FFAs to decrease adiponectin secretion occurred in vivo, this could potentially lead to decreased systemic insulin sensitivity.
We found that the FFA effects on glucose transport and insulin signaling are largely mediated by JNK activation. FFAs directly activated JNK to cause cellular and molecular defects leading to insulin resistance, and siRNA-mediated inhibition of JNK activation prevented these impairments. This confirms previous in vivo studies correlating JNK activation and lipid-associated insulin resistance (9,20). Interestingly, siRNA-mediated knockdown of JNK2 also decreased JNK1 protein lev-els, as previously observed (36,37), suggesting coordinated regulation of JNK1/2 protein expression/activity in intact cells.
Since exogenously added TNF-␣ can activate JNK and induce insulin resistance (22) and since the TNF-␣ gene and TNF-␣ secretion can be regulated by JNK activity in other cell types (34), we hypothesized that in adipocytes, adipocyte-derived TNF-␣ is a downstream mediator of JNK in FFA-induced insulin resistance. Consistent with this, we showed that FFA treatment increases the secretion of TNF-␣ from adipocytes by ϳ80% (from ϳ0.1 to 0.2 pg/h/l of medium, p Ͻ 0.001; Fig. 6). This secretion rate lies within the physiological range of TNF-␣ release, estimated at 0.02 pg/l/h for isolated adipocyte suspensions and 0.07 pg/mg of tissue/h for whole adipose tissue (38) and corresponds to Ͻ1% of the pharmacological doses typically used to confer insulin resistance in adipocytes (39). Although these secreted TNF-␣ levels are low, they are biologically relevant, given the instability of TNF-␣ (half-life of 6 -10.5 min (40,41)) and the restricted spatio-temporal pattern of its release, trimerization, and action within adipose tissue. These findings also establish FFA as a novel pathophysiological factor that can up-regulate TNF-␣ secretion.
These small FFA-induced increases in secreted TNF-␣ proteins are sufficient to propagate proinsulin resistance signals; this was demonstrated by showing that inhibition of secreted TNF-␣ action using neutralizing antibodies against TNF-␣ or its receptors or DN TNF partially prevented the cellular insulin resistance. These interventions prevented the effect of FFAs to impair glucose transport and insulin signaling by ϳ50% (Fig. 7, A and B). This confirmed that the FFA effects were not nonspecific and toxic. That only part of the FFA effects were blocked by TNF-␣ neutralization could mean that the inhibitory agents were only A, FFAs alter the protein expression pattern of SOCS-3, an effect that is not inhibited by blocking TNF-␣ signaling. Serum-starved 3T3-L1 adipocytes were treated with 1 mM FFA for 1 h, with or without 1.7 nM insulin stimulation, with or without pretreatment with 2 g/ml neutralizing antibodies to TNF-␣ or control IgG. Cell lysates were prepared, fractionated by SDS-PAGE, and immunoblotted with polyclonal antibodies against SOCS-1 and SOCS-3 proteins. In control IgGtreated adipocytes, as in nonpretreated cells (data not shown), FFA treatment did not affect SOCS-1 protein expression. However, it altered the SOCS-3 protein expression pattern, increasing the expression of the high molecular weight form of SOCS-3. This effect was not blocked by pretreatment with TNF-␣-neutralizing antibodies (lanes 5 and  6 versus lanes 1 and 2). Representative results from six experiments in duplicate are shown. B, adipocytes were electroporated with JNK2 siRNA (mJ2B) or scrambled siRNA (scr) 48 h prior to FFA treatment. The effect of FFA on SOCS-3 protein expression in these JNK-deficient cells was assessed as above. siRNA-mediated JNK knockdown markedly diminished the change in the SOCS-3 protein expression pattern induced by FFAs, including the increased levels of high molecular weight SOCS-3 protein. Representative results from four independent experiments are shown. partially effective at blocking TNF-␣ signaling or that a component of the FFA effect is mediated through non-TNF-␣-dependent pathways. It is unlikely that the TNF-␣-blocking agents were only partially effective, because increasing their concentration from 2 to 25 g/ml did not increase the response and because they prevented the decrease in adiponectin secretion by Ͼ90%. Thus, we conclude that the FFA effects are mediated in part by adipocyte-derived TNF-␣ and in part by TNF-␣independent mechanisms.
When JNK activation was blocked by RNA interference, the FFAinduced increase in secreted TNF-␣ was inhibited (Fig. 9B), indicating that TNF-␣ is a downstream effector of JNK in mediating the effects of short term FFA treatment on adipocytes. That siRNA-mediated JNK inactivation did not reduce the FFA-induced levels of secreted TNF-␣ to basal levels is probably due to the residual levels of activated JNK protein resulting from incomplete protein knockdown (see Fig. 4A). It is possible that the relative roles of JNK and TNF-␣ in mediating the effects of FFA-induced insulin resistance may be different with long term or chronic FFA exposure, since additional mechanisms affecting gene expression may contribute.
We found that one of the potential TNF-␣-independent mediators of the effects of FFAs is SOCS-3. We detected several endogenous forms of SOCS-3 protein in 3T3-L1 adipocytes. These isoforms could result from cell-specific alternate splicing of the SOCS-3 gene or alternate initiation of translation of the SOCS-3 mRNA (42). When 3T3-L1 adipocytes were treated with FFAs, the pattern of SOCS-3 protein expression was altered, with increased expression of a high molecular weight form of SOCS-3. This isoform could arise from post-translational modification such as phosphorylation; SOCS-3 phosphorylation on multiple tyrosine residues in response to growth factors has been documented in SOCS-3-overexpressing A431 and NIH3T3 cells (43). This alteration in the SOCS-3 protein expression pattern by FFAs was not inhibited by TNF-␣ neutralization (Fig. 8A), indicating that SOCS-3 can act independently of TNF-␣. This alteration in the SOCS-3 protein expression pattern was markedly reduced in JNK knockdown cells, probably due to reduced SOCS-3 proteins in the basal state. This suggested that SOCS-3 expression/activity is dependent upon JNK and that SOCS-3 is a potential JNK-dependent but TNF-␣-independent mediator of the effects of FFAs.
The effect of FFAs on secreted adiponectin was reversed by blocking TNF-␣ signaling using neutralizing antibodies and DN TNF. Given that at the level of the whole organism, lowered adiponectin levels can result in liver, skeletal muscle, and whole body insulin resistance, these findings also have potential implications for the treatment of insulin resistance. The DN TNF peptide inhibited the FFA-induced TNF-␣-mediated inhibitory effect on insulin action, raising the possibility that DN TNF peptides could be of therapeutic value in treating insulin resistance. DN TNF peptides are smaller in size and have enhanced stability compared with TNF-neutralizing antibodies, and these properties might enable them to reach target tissues in a biologically active form to improve insulin sensitivity, which previous clinical studies with neutralizing antibodies have not accomplished (44,45).
The current findings support and extend previous studies of the effects of FFAs on insulin signaling, glucose uptake, and inflammation (2,4,16,22,31). More importantly, they allow us to propose a mechanistic model for FFA-induced cellular insulin resistance. FFAs cause JNK activation, thereby activating TNF-␣-independent and -dependent downstream events leading to cellular insulin resistance. Thus, JNK activation directly induces cellular insulin resistance via downstream effectors. In addition, JNK activation increases the secretion of proinflammatory TNF-␣, which acts in an autocrine/paracrine fashion to produce and amplify proinsulin resistance signals.