Cross-talk mechanisms in the development of insulin resistance of skeletal muscle cells

Palmitate rather than tumour necrosis factor inhibits insulin-dependent protein kinase B (PKB)/Akt stimulation and glucose uptake


G. Müller, Institute for Cell Biology and Immunology, University of Stuttgart, Allmandring 31, D-70569 Stuttgart, Germany. Fax: + 49 711685 7484, Tel.: + 49 711685 7485, E-mail:


Insulin resistance in skeletal muscle is one of the earliest symptoms associated with non-insulin-dependent diabetes mellitus (NIDDM). Tumour necrosis factor (TNF) and nonesterified fatty acids have been proposed to be crucial factors in the development of the insulin-resistant state. We here show that, although TNF downregulated insulin-induced insulin receptor (IR) and IR substrate (IRS)-1 phosphorylation as well as phosphoinositide 3-kinase (PI3-kinase) activity in pmi28 myotubes, this was, unlike in adipocytes, not sufficient to affect insulin-induced glucose transport. Rather, TNF increased membrane expression of GLUT1 and glucose transport in these muscle cells. In contrast, the nonesterified fatty acid palmitate inhibited insulin-induced signalling cascades not only at the level of IR and IRS-1 phosphorylation, but also at the level protein kinase B (PKB/Akt), which is thought to be directly involved in the insulin-induced translocation of GLUT4, and inhibited insulin-induced glucose uptake. Palmitate also abrogated TNF-dependent enhancement of basal glucose uptake, suggesting that palmitate has the capacity to render muscle cells resistant not only to insulin but also to TNF with respect to glucose transport by GLUT4 and GLUT1, respectively. Our data illustrate the complexity of the mechanisms governing insulin resistance of skeletal muscle, questioning the role of TNF as a direct inhibitor of glucose homoeostasis in this tissue and shedding new light on an as yet unrecognized multifunctional role for the predominant nonesterified fatty acid palmitate in this process.


non-insulin-dependent diabetes mellitus


tumour necrosis factor


insulin receptor


insulin receptor substrate


phosphoinositide 3-kinase


protein kinase B


TNF receptor


glycogen synthase kinase-3


palmitic acid methyl ester

Insulin resistance is defined as the reduced ability of cells or tissues to respond to physiological levels of insulin and is a characteristic condition of early-stage non-insulin-dependent diabetes mellitus (NIDDM) [1,2]. Obesity is the strongest risk factor for NIDDM. Two candidates, nonesterified fatty acids and tumour necrosis factor (TNF), have been proposed as mediators associated with the above disease state. Non-esterified fatty acids are one important possible link between obesity, insulin resistance and NIDDM. Plasma nonesterified fatty acid levels are chronically elevated during obesity (reviewed in [3]), promote hepatic glucose overproduction [4], inhibit pancreatic β-cell function in insulin secretion [5] and, most importantly, impair glucose utilization in skeletal muscle cells by the glucose fatty acid cycle (reviewed in [6]).

Numerous animal models of obesity and insulin resistance based on spontaneous genetic mutations or experimentally created gene deletions have been investigated to clarify the molecular mechanisms and role of different signal pathways and molecules involved in the development of diabetes with often unexpected results (reviewed in [1,7,8]). Knock-out mice of the fatty acid-binding protein developed obesity without any signs of insulin resistance or diabetes. Most notably, these mice showed no overexpression of TNF in adipose tissue or muscle cells [9]. TNF expression is usually increased in adipose tissue and skeletal muscle cells of obese diabetic patients [10,11], and TNF alters lipid metabolism in these tissues (reviewed in [12]). Furthermore, there is evidence that TNF participates in dysregulation of hepatic glucose output [13] and inhibits glucose-induced insulin secretion [14]. Knock-out mice devoid of either the TNFα gene or both TNF receptor (TNFR) genes did not develop insulin resistance in a genetically determined (ob/ob) obesity model associated with leptin deficiency [15]. Thus, these findings were all in favour of a causal relationship or essential contribution of TNF to obesity–diabetes syndromes. Only recently, however, the role of TNF has been questioned, as TNFRI/II double knock-out mice showed no significant difference to wild-type mice in disease development in a diet-induced obesity model. Of note, in this model, even TNFR double knock-out mice developed hyperinsulinaemia, suggesting a protective rather than a deleterious role for TNF in early stages of the development of NIDDM [16]. Accordingly, from the various animal models, at present no final conclusions as to the role of TNF in NIDDM can be drawn.

With regard to molecular mechanisms of insulin resistance and potential targets of TNF cross-talk with insulin signal pathways, it is now apparent that the activation of phosphoinositide 3-kinase (PI3-kinase) is necessary and in some cases sufficient to elicit many of the insulin effects on glucose and lipid metabolism [17]. PI3-kinase activation by insulin is accomplished via phosphorylation of the immediate targets of the IR kinase, the IR substrates (IRS). One of several signal molecules bound by activated IRS is the p85 regulatory subunit of phosphoinositol 3-kinase, resulting in recruitment and activation of the p110 catalytic domain [18]. The lipid messengers produced by PI3-kinase bind to the pleckstrin homology (PH) domains of phosphoinositide-dependent protein kinase B (PKB/Akt), a multifunctional serine/threonine-specific kinase considered important in mediating downstream insulin signals, in particular GLUT4 translocation to the plasma membrane [19,20]. One of three potential in vivo substrates of PKB, namely glycogen synthase kinase-3 (GSK3), is inhibited by PKB and is thought to contribute to the insulin-induced dephosphorylation of glycogen synthase [19]. IR phosphorylation also activates the Ras/MAP kinase cascade which, however, appears not to be necessary for glucose transport [21]. Accordingly, aside from the IR and IRS themselves, PI3-kinase and downstream molecules of the signal pathway to GLUT4 activation are potential targets of interfering factors.

Most of the previous in vitro investigations showing an effect of TNF on insulin signal cascades and glucose utilization have been performed with adipocytes and hepatocytes, but skeletal muscle cells, the largest storage organ for glucose and considered to play the major role in glucose homoeostasis, have not been studied in greater detail. To elucidate the role of TNF and nonesterified fatty acids in the insulin response of muscle cells, we used a recently established cellular model, the murine myoblast cell line pmi28 which differentiates in vitro into multinucleated myotubes with markers characteristic of skeletal muscle [22]. pmi28 cells are, upon differentiation, responsive to both TNF and insulin and represent a suitable model for studying the cross-talk between TNF and insulin signalling cascades [23]. Evidence is presented here that nonesterified fatty acids may play a more important role in the development of muscle insulin resistance than previously thought. Palmitate has long been known to downregulate metabolic routes of insulin-regulated pathways (reviewed in [6]). We here show that it also interferes with insulin-induced signal cascades controlling glucose uptake, by inhibition of IR, IRS-1 and PKB/Akt kinase activity. In contrast, the inhibitory cross-talk of TNF funnels downstream to targets different from insulin-induced glucose uptake, suggesting that TNF is, despite downregulation of IR and IRS activation, not the prime mediator interfering with glucose homoeostasis in muscle cells.

Materials and methods

Cell culture

Myoblasts from a primary culture designated pmi28 were used [22]. The culture was established from the hind-leg muscles of a 7-day-old male Balb/c mouse. Initial expansion and enrichment in myoblasts was achieved by repeated preplating and growth in minimum essential medium with d-valine (Gibco) containing 20% fetal bovine serum. When the expanded culture was highly enriched (100% myoblasts as evaluated by immunofluorescent staining for desmin), the cells were further propagated in Ham F10 medium (Biochrom, Berlin, Germany) supplemented with 20% fetal bovine serum. Pmi28 cells have been propagated in vitro over 60 passages without loss of proliferative capacity or differentiation properties [23]. For differentiation into myotubes, pmi28 cells (between passages 30 and 50) were grown to confluence in tissue culture plates, and medium was changed into low-glucose (1 g·L−1) Dulbecco’s modified Eagle’s medium (Biochrom) supplemented with 10% horse serum for 2 days (Gibco-BRL). Before the experiments, differentiated pmi28 cells (pmi28 myotubes > 95%) were serum starved and, as indicated, preincubated with TNF (50 ng·mL−1), palmitic acid or palmitic acid methyl ester (PAME) for 2 h and stimulated with insulin (200 nm).

Cytokines, antibodies and reagents

Recombinant human TNF was a gift from BASF (Ludwigshafen, Germany) and insulin from bovine pancreas was from Sigma. Antibodies against IRβ (C-19), IRS-1 (C-20), phosphotyrosine (PY99) and p110 β subunit of PI3-kinase (S-19) were from Santa Cruz. Antibodies against Akt/PKBα and phospho-Akt (Ser473) were from New England Biolabs, those against GSK3β from Transduction Laboratories (San Diego, CA, USA), and those against Akt-1 PH domain and p85 subunit of PI3-kinase from Upstate (Lake Placid, NY, USA). Akt-specific substrate peptide, phosphoglycogen synthase peptide-2 and protein kinase A inhibitor peptide were from Upstate. Polyclonal GLUT1-specific antibodies were a gift from A. Schürmann, RWTH, Aachen, Germany. Secondary alkaline phosphatase-linked antibodies, goat anti-(mouse IgG and IgM) (H + L) and goat anti-(rabbit IgG) (H + L) were from Dianova (Hannover, Germany). Nonesterified fatty acids (palmitic acid, lauric acid, stearic acid, myristic acid), PAME, phosphatidylinositol and a vinculin-specific antibody were from Sigma. All inhibitors of proteases and phosphatases were from Biomol (Hannover, Germany) or Sigma. 3H-labelled 2-deoxyglucose was from Amersham.

Western blotting and immunoprecipitation

pmi28 myoblasts (5 × 166 per well) were seeded in tissue culture dishes (9 cm) and differentiated into myotubes after 2 days of cultivation. After stimulation, the cells were washed twice with NaCl/Pi (4 °C) and, depending on the experimental set-up, scraped into 750 µL ice-cold phosphorylation buffer (50 mm Hepes, pH 7.8, 2.5 mm EDTA, 1% Triton X-100, 150 mm saccharose, 10 mm sodium pyrophosphate, 100 mm NaF) or lysis buffer [50 mm Tris/HCl, pH 7.4, 1% (v/v) Triton X-100, 150 mm NaCl, 5 mm EDTA, pH 7.4, 1 mm NaF, 1 mm sodium pyrophosphate]. Both buffers also contained the following inhibitors: 2 mm sodium orthovanadate, 1 mm sodium molybdate, 100 nm okadaic acid, 100 nm calyculin A, 1 mmp-nitrophenyl phosphate, 1 µg·mL−1 leupeptin, 1 µg·mL−1 aprotinin and 1 mm phenylmethanesulfonyl fluoride. After 60 min cell lysis, the lysates were centrifuged (10 000 g, 15 min, 4 °C) and Western blotting or immunoprecipitation was performed as described [23]. Protein content of the samples for Western blotting or immunoprecipitations was quantified using the Bio-Rad Protein Assay using BSA as a standard.

Isolation of microsomal membranes

The preparation of microsomal membranes was performed as described previously [24]. pmi28 myoblasts (15 × 106 per well) were seeded in tissue culture dishes (14.5 cm) and differentiated into myotubes after 2 days of cultivation. After stimulation the cells were washed twice with 4 °C NaCl/Pi and scraped into 2 mL homogenization buffer (25 mm Hepes, pH 7.4, 5 mm EGTA, 50 mm NaF, 1 µg·mL−1 leupeptin, 1 µg·mL−1 aprotinin and 1 mm phenylmethanesulfonyl fluoride). After homogenization (25 strokes in a 5-mL Dounce homogenizer), cell debris was removed by centrifugation (800 g, 4 °C, 5 min) and the supernatant was centrifuged at 250 000 g and 4 °C for 30 min. The pellet was resuspended in 100 µL lysis buffer, and 100 µL reducing 2 × Laemmli buffer were added.

Measurement of 2-deoxyglucose uptake

pmi28 myoblasts (104 per well) were seeded in 24-well tissue plates and differentiated into myotubes after 2 days of cultivation (≈ 95% differentiated cells). The cells were serum starved and incubated with TNF or fatty acids for 2 h. Insulin was added for the last 30 min. After stimulation the cells were washed once with uptake buffer (20 mm Hepes, pH 7.4, 140 mm NaCl, 5 mm KCl, 2.5 mm MgSO4, 1 mm CaCl2) and incubated with 1 µCi 2-deoxy[3H]glucose per well in uptake buffer for 40 min at 37 °C. Cells were washed twice with ice-cold uptake buffer and lysed with 200 µL 0.05% NaOH for 20 min. The lysate was transferred to 3 mL scintillation liquid and radioactivity counting was performed in a scintillation counter.

PI3-kinase assay

For evaluation of in vitro kinase activity of PI3-kinase co-precipitated with IRS-1, a modified assay described by Gold et al. [25] was performed. pmi28 myoblasts (15 × 106 per well) were seeded in tissue culture dishes (14.5 cm) and differentiated into myotubes after 2 days of cultivation. Cells were stimulated as indicated and washed twice with ice-cold NaCl/Pi. The cells were lysed in lysis buffer and immunoprecipitation of IRS-1 was performed as described above. The precipitate was washed three times with ice-cold KP buffer (20 mm Tris/HCl, pH 7.4, 50 mm NaCl, 10 mm MgCl2). PI3-kinase activity was measured by adding 10 µg sonicated phosphatidylinositol in 30 mm Hepes (sodium salt), pH 7.4, to each tube and the tubes were incubated for 10 min on ice. Assay buffer (40 µL) containing 30 mm Hepes, pH 7.4, 30 mm MgCl2, 200 µm adenosine, 50 µm ATP, 20 µCi [γ-32P]ATP and inhibitors (same as used for immunoprecipitation) were added to each tube and the reactions were carried out for 15 min at room temperature and terminated by the addition of 0.1 mL 1 m HCl. The lipids were extracted with 0.2 mL chloroform/methanol (1 : 1). A 20 µL volume of the organic phase was spotted on to silica-gel TLC plates that had been developed with 1% potassium oxalate (in bidistilled water). The chromatograms were developed in chloroform/methanol/water/28% ammonia (45 : 35 : 7.5 : 2.5, v/v), dried and baked for 30 min at 80 °C. Radioactivity incorporated into PtdInsP was quantified using a PhosphoImager.

Akt/PKB kinase/GSK3 assays

The immunoprecipitates of Akt/PKB or GSK3 were washed three times with ADB (Akt kinase assays; 20 mm Mops, pH 7.2, 25 mmβ-glycerophosphate, 5 mm EGTA, 1 mm sodium orthovanadate, 1 mm dithiothreitol) or GSK buffer (GSK3 kinase assays; 40 mm Hepes, pH 7.2, 25 mmβ-glycerophosphate, 10 mm MgCl2). After removal of the supernatant, 10 µL ice-cold ADB or GSK buffer, 10 µL protein kinase A inhibitor peptide (final concentration 10 µm/sample), 10 µL specific substrate peptide (100 µm), and 10 µL ATP solution (1 µCi [γ-32P]ATP·µL−1, 75 mm MgCl2, 500 µm unlabeled ATP or 2 mm unlabelled ATP for the GSK3 assay) were added and the samples were incubated for 10 min at 30 °C. The kinase reaction was stopped on ice, the supernatant was transfered to another tube, and 10 µL spotted on to the center of a 2-cm × 2-cm piece of P81 phosphocellulose paper. The assay squares were washed three times with 0.75% phosphoric acid, once with acetone for 5 min, dried at 70 °C, and measured by Cherenkov counting in a scintillation counter. Immunoprecipitations with preimmune serum and samples without Akt-specific or GSK3-specific substrate peptide served as negative controls.


TNF inhibits insulin-induced tyrosine phosphorylation of IRβ, IRS-1 and PI3-kinase, but fails to interfere with glucose uptake

A 2-h TNF pretreatment of differentiated pmi28 cells causes downregulation of a subsequent insulin response at a time point of maximum IR kinase activity (10 min stimulation with 200 nm insulin; Fig. 1A [23]), as is evident from an ≈ 40 and 60% reduction of the tyrosine phosphorylation of IRβ and IRS-1, respectively (Fig. 1B,C). PI3-kinase has been implicated as one of the key signal molecules in insulin stimulation and insulin-induced GLUT4 activation [17]. It is a heterodimeric enzyme consisting of an 85-kDa regulatory subunit with SH2 domains capable of binding to tyrosine-phosphorylated IRS-1 and a 110-kDa catalytic subunit [18]. The functional significance of the TNF-dependent decrease in IRS-1 tyrosine phosphorylation in pmi28 cells was therefore monitored by anti-IRS-1-mediated co-precipitation of both the catalytic and regulatory subunits of PI3-kinase and subsequent Western blotting. PI3-kinase heterogeneity in the catalytic p110 domains has been reported. Until now, it was unclear which of the three known p110 subunits is predominantly associated with IRS-1 in skeletal muscle cells. We obtained evidence that the p110 β subunit (Fig. 1D) but not the α or γ subunit (data not shown) is associated with the IRS-1-bound p85. Phosphatidylinositol phosphate formation was stimulated threefold by insulin (Fig. 1E). TNF pretreatment resulted in partial inhibition of insulin-stimulated phosphatidylinositol phosphate formation (Fig. 1E), which is in the range expected from the extent of inhibition of IR-mediated IRS-1 phosphorylation (Fig. 1C). Together, these data show that the in vitro differentiated mouse muscle cell line exhibits insulin–TNF signal cross-talk at the level of IRβ, IRS-1 and its immediate downstream target PI3-kinase similar to that observed in hepatocytes and adipocytes [26].

Figure 1.

Figure 1.

    Tyrosine phosphorylation andPI3-kinase activity in pmi28 cells. (A) Kinetics of insulin-induced tyrosine phosphorylation of IRβ and IRS-1 in differentiated pmi28 cells. Differentiated pmi28 cells were serum starved for 2 h and stimulated with insulin (200 nm) for the indicated times. Cells were lysed in phosphorylation buffer and IRβ or IRS-1 were immunoprecipitated. The samples were separated by SDS/PAGE, blotted to nitrocellulose and stained with PY antibody. Tyrosine phosphorylation of IRβ chain (●) or IRS-1 (▪) is shown after densitometric evaluation; maximum activation after 10 min was set at 100%. (B,C) Influence of TNF and insulin on tyrosine phosphorylation of IRβ and IRS-1 in pmi28 muscle cells. Differentiated pmi28 cells were serum starved and preincubated with TNF (50 ng·mL−1). After stimulation with insulin (500 nm) for 10 min the cells were lysed in phosphorylation buffer. Immunoprecipitates were separated by SDS/PAGE and blotted to nitrocellulose. (B) Immunoprecipitation of IRβ. The blot was stained with PY-specific or IRβ-specific antibodies. (C) Immunoprecipitation of IRS-1. The blot was stained with PY-specific or IRS-1-specific antibodies. Experiments were performed three times with similar results. (D,E) Influence of TNF on insulin-induced PI3-kinase activity. Differentiated pmi28 cells were incubated with TNF (50 ng·mL−1) in serum-free medium for 2 h. After 10 min stimulation with insulin (500 nm), cell lysis occurred in phosphorylation buffer. (D) Co-precipitation of PI3-kinase subunits p85 and p110β with IRS-1. Immunoprecipitation of IRS-1 and detection of the blot with antibodies against IRS-1, p85 or p110β. Densitometric evaluation: p110 subunit: insulin, 100%; TNF + insulin, 53%; p85 subunit: insulin, 100%; TNF + insulin, 49%. (E) PI3-kinase assay of PI3-kinase co-precipitated with IRS-1. Immunoprecipitation of IRS-1 and performance of a PI3-kinase assay and separation of lipids on TLC plates (upper panel) as described in Materials and methods. Densitometric evaluations of TLC plates (lower panel). PIP, phosphatidylinositol phosphate.

    In the latter cell type, such as differentiated mouse 3T3-L1 cells, acute exposure to insulin resulted in a twofold increase in 2-deoxyglucose uptake over the basal levels which is downregulated by TNF pretreatment (data not shown). In differentiated pmi28 cells, insulin stimulated glucose uptake in a concentration-dependent manner, with maximal uptake reached between 100 and 500 nm insulin (Fig. 2A). In contrast with adipocytes, TNF by itself increased basal glucose uptake and did not block insulin-stimulated glucose uptake (Fig. 2B). Furthermore, insulin and TNF stimulation were additive. Similar effects were observed in the presence of high (50 ng·mL−1) (Fig. 2B) or low (5 ng·mL−1, data not shown) concentrations of TNF and during short-term (2 h, Fig. 2B) or long-term (16 h, data not shown) exposure, indicating that, in differentiated pmi28 cells, TNF is not an inhibitor of insulin-induced glucose transport despite clear interference with upstream IR signalling cascades (Fig. 1).

    Figure 2.

    Figure 2.

      Effect of TNF on insulin-induced2-deoxyglucose uptake in differentiated pmi28 cells. 2-deoxyglucose uptake assays were performed as described in Materials and methods. (A) Role of insulin concentration in differentiated pmi28 cells. (B) Role of human TNF (50 ng·mL−1) and insulin (500 nm) in differentiated pmi28 cells. The data are the mean ± SD from four determinations from one of five independent experiments.

      Upregulation of GLUT1 transporters by TNF

      Apart from the insulin-regulated GLUT4 transporter, muscle cells also contain GLUT1 transporters [27]. In the context of the above data, it was of interest to know by which mechanism TNF stimulates glucose transport in muscle cells. Therefore, GLUT1 protein levels in cell membrane extracts of pmi28 with or without TNF treatment were investigated. As shown in Fig. 3, cells stimulated with TNF for 1.5 h showed a 30% increase in GLUT1 proteins as revealed from Western blots. Vinculin served as an internal cellular protein standard, which remained unchanged during the time of TNF incubation (Fig. 3). These results suggest that TNF stimulates basal glucose transport by upregulation of GLUT1 transporters in skeletal muscle cells.

      Figure 3.

      Figure 3.

        Upregulation of GLUT1 protein levels in cellular membranes after TNF treatment. Differentiated pmi28 cells were treated with TNF (50 ng·mL−1, 1 h) or insulin (200 nm, 1 h) and microsomal membranes were prepared. Equal amounts of protein were separated by SDS/PAGE (10%), transferred to nitrocellulose, and bands were stained with GLUT1-specific antibodies. Vinculin was used as internal standard (left panel). Right panel, densitometric evaluation of GLUT1 bands. Experiments were performed three times with similar results.

        Role of nonesterified fatty acids as inhibitors of insulin-regulated glucose transport

        As TNF induced insulin resistance at the level of IR phosphorylation cascades, but did not result in inhibition of insulin-dependent glucose transport, we hypothesized that additional factors must be involved in muscle cell insulin resistance and investigated other potential mechanisms responsible for inhibition of insulin-dependent glucose transport, an existing defect in muscle of patients with NIDDM that has been verified by 13C/31P NMR [28]. Candidates known for several decades to induce insulin resistance in muscle cells are nonesterified fatty acids (reviewed in [6]). Therefore, the role and specificity of the effects of palmitic acid on insulin-induced glucose transport in comparison with a non-metabolizable analogue of palmitic acid, PAME, and other nonesterified fatty acids of longer and shorter chain length were tested. As shown in Fig. 4, palmitic acid, which is known to be one of the major lipids in plasma and tissue [6], was found to be a potent inhibitor of insulin-induced glucose transport, whereas all other fatty acids and analogues tested did not prevent insulin action.

        Figure 4.

        Figure 4.

          Specificity of palmitic acid-induced inhibition of insulin-induced 2-deoxyglucose uptake in pmi28 myotubes. Differentiated pmi28 cells were serum starved and preincubated for 2 h with the indicated nonesterified fatty acids (200 µm) or with PAME (200 µm). After stimulation with insulin (200 nm) for 30 min 2-deoxyglucose-uptake assays were performed as described in Materials and methods. A representative experiment out of three is shown.

          Palmitate inhibits the IR signal pathway to PKB/Akt phosphorylation

          Although one mechanism of action of nonesterified fatty acids on glucose homoeostasis has been known for several decades, namely the inhibition of pyruvate dehydrogenase [6], an enzyme responsible for glucose metabolization, we further searched for additional inhibitory mechanisms in the insulin signal pathway. Interestingly, palmitic acid inhibited tyrosine phosphorylation of IRβ and IRS-1 by ≈ 50% (Fig. 5), similar to that found with TNF (Fig. 1). To obtain additional evidence of the specificity of palmitic acid, the non-metabolizable analogue PAME was tested in parallel. The results shown in Fig. 6 prove that the nonesterified acid is required for the inhibitory cross-talk.

          Figure 5.

          Figure 5.

            Influence of palmitic acid and insulin on tyrosine phosphorylation of IRβ and IRS-1 in pmi28 myotubes. Differentiated pmi28 cells were serum starved and preincubated with palmitic acid (200 µm). After stimulation with insulin (200 nm) for 10 min, the cells were lysed in phosphorylation buffer. Immunoprecipitates were separated by SDS/PAGE and transferred to nitrocellulose. (A) Immunoprecipitation of IRβ. The blot was stained with PY-specific or IRβ-specific antibodies. (B) Immunoprecipitation of IRS-1. The blot was stained with PY-specific or IRS-1-specific antibodies. (C) Specificity of the effect of palmitate compared with other nonesterified fatty acids on IRS-1 phosphorylation. All experiments were performed at least three times with similar results.

            Figure 6.

            Figure 6.

              Specificity of palmitic acid-inducedinhibition of insulin-induced tyrosinephosphorylation of IRβ and IRS-1 in pmi28 myotubes. Differentiated pmi28 cells were serum starved and preincubated with palmitic acid (200 µm) or PAME (200 µm). After stimulation with insulin (200 nm) for 5 min the cells were lysed in phosphorylation buffer. Immunoprecipitates were separated by SDS/PAGE and blotted to nitrocellulose. Immunoprecipitation of IRβ (left panel) or IRS-1 (right panel). The blot was stained with PY-specific antibodies. Experiments were performed three times with similar results.

              As deduced from the data on the TNF–insulin cross-talk, a partial inhibition of IR kinase activity and IRS-1 phosphorylation might not be sufficient to downregulate glucose uptake. Therefore, further candidate signal molecules involved in regulation of glucose transport were investigated. An important downstream effector of IRS-1-coupled PI3-kinase and regulator of GLUT4 translocation is the kinase PKB/Akt. Indeed, palmitate, but not TNF, was found to be operative at the level of PKB/Akt kinase (Fig. 7A). Palmitate induced a 50% reduction of the insulin-stimulated serine phosphorylation of this enzyme, whereas TNF treatment had no effect on PKB/Akt. This activation pattern is also reflected by a more than fivefold increase in insulin-induced substrate phosphorylation by PKB/Akt kinase which is inhibited by about 50% in the presence of palmitate. TNF or TNF in combination with palmitate had no effect (Fig. 7B). Probing the activity of GSK3 as potential substrate of PKB/Akt using the same experimental design, however, showed no significant changes induced by palmitate (data not shown). These results indicate that the nonesterified fatty acid palmitate interferes in multiple ways in insulin-stimulated signalling pathways in muscle cells, which lead to inhibition of insulin-stimulated glucose uptake rather than to changes in GSK3 activity.

              Figure 7.

              Figure 7.

                Serine phosphorylation of Akt kinase (PKB). (A) Differentiated pmi28 cells were preincubated in serum-free medium containing 0.01% BSA and palmitic acid (200 µm) or TNF (50 ng·mL−1) for 2 h. After 10 min stimulation with insulin (100 nm), the cells were lysed in 200 µL lysis buffer. Equal amounts of protein (40 µg per lane) were loaded on SDS/10% polyacrylamide gels. Akt kinase was detected on Western blots with Akt-specific and phospho-Akt-specific (S473) antibodies. Shown are densitometric evaluations of phospho-Akt corrected for relative intensities of Akt and the results are from a representative experiment out of five. (B) Differentiated pmi28 cells were preincubated in serum-free medium containing 0.01% BSA and palmitic acid (200 µm) or TNF (50 ng·mL−1) for 2 h. After 10 min stimulation with insulin (100 nm), the cells were lysed in 200 µL lysis buffer. Akt was immunoprecipitated and an Akt kinase assay was performed as described in Materials and methods. The substrate peptide was dotted on to P81 paper and counted by Cherenkov counting in a scintillation counter. Experiments were performed three times with similar results.

                Palmitate is an inhibitor of TNF-mediated enhancement of basal glucose transport

                The action of nonesterified fatty acids is of special interest in muscle cells, because, in obese muscle tissue, they are rather abundant in the interstitial fluid [29]. As TNF showed a positive effect on basal glucose uptake in muscle cells, we also investigated whether or not nonesterified fatty acids modulate this activity in pmi28 cells. Our data reveal that palmitic acid is a potent inhibitor of TNF-mediated enhancement of basal glucose transport (Fig. 8). These results indicate that this nonesterified fatty acid not only plays a role in inhibition of insulin-dependent glucose transport via GLUT4, but also affects the TNF-sensitive GLUT1-dependent pathway of basal glucose uptake.

                Figure 8.

                Figure 8.

                  Effect of palmitic acid on TNF-induced 2-deoxyglucose uptake in differentiated pmi28 cells. pmi28 myotubes were preincubated with palmitic acid (200 µm, 2 h) and stimulated with TNF (50 ng·mL−1, 1.5 h). 2-Deoxyglucose-uptake assays were performed as described in Materials and methods. Shown is a representative experiment out of five.


                  The present study provides new insights into the mechanisms of insulin resistance in mouse skeletal muscle cells. Our data indicate that TNF, despite interference with IR proximal signal pathways, does not downregulate insulin-induced glucose transport, and even exerts a positive effect by enhancing basal glucose transport. In contrast, the nonesterified fatty acid palmitate, which has previously been implicated in insulin resistance in muscle cells via regulation of the glucose fatty acid cycle [6], was identified here as a major regulator of insulin-induced signals and glucose uptake. We show that palmitate affects IR phosphorylation cascades including serine phosphorylation of Akt kinase. Further, our results provide evidence that, unlike in adipocytes, in muscle cells the palmitate not only induces insulin resistance, but also abolishes ‘protective functions’ of TNF, namely the stimulation of basal glucose transport.

                  Several previous studies have investigated whether an impaired signalling capacity of the IR contributes to the pathogenesis of skeletal muscle insulin resistance. Most studies have reported that autophosphorylation or substrate phosphoryation of the IR kinase isolated from diabetic skeletal muscle is reduced [30]. Although this correlated with reduced insulin-dependent glucose uptake [28], no downregulation of GLUT4 was found [31]. Therefore, it was concluded that interference with insulin-dependent signalling cascades per se may be the cause for the apparent insulin resistance in human muscle tissue. Thus, the common denominator of the different models of insulin resistance was the quantitatively and/or qualitatively reduced IR kinase activity and insufficient substrate phosphorylation as the crucial initiating event [26,32,33]. In pmi28 cells, TNF caused an ≈ 50% inhibition of IR, IRS-1 and IRS-1-associated PI3-kinase activity (Fig. 1), but no impairment of glucose transport (Fig. 2). Activation of IRS-associated PI3-kinase, however, has been assumed to be a response-limiting event for stimulation of insulin-dependent glucose transport [34]. It is well known that IRS-1 contains numerous docking sites for PI3-kinases, which exert different biological roles in insulin signalling, with activation of GLUT4 being only one of them (reviewed in [17]). Thus, it may be concluded that the 50% inhibition of insulin-stimulated PI3-kinase by TNF comprises molecules which are not involved in GLUT4 translocation, or that the remaining 50% activity is enough to fully maintain insulin-induced glucose transport. A third alternative explanation may be that other subtypes of IRS may compensate TNF-induced inhibition, as proposed previously [35,36]. This opens up the new possibility that, in muscle cells, TNF-mediated cross-talk with insulin signal pathways does not affect or is functionally not apparent for those controlling GLUT4-dependent glucose transport and thus is restricted to interference with other IR-induced signals. One of the potential targets of TNF cross-talk is STAT5, suggesting that TNF interferes with the expression of insulin-responsive STAT5-regulated genes [23]. Our in vitro model suggests that TNF itself is not the factor that induces insulin resistance in muscle cells, which is in contrast with adipocytes, in which the TNF-induced insulin resistance at IR level correlating with inhibition of glucose transport is well documented [26].

                  The observation in pmi28 cells that TNF enhances basal glucose transport is in accordance with recent findings in other cellular models, namely L6 myoblasts [37] and normal non-obese rat skeletal muscle [38]. Interestingly, the same effect was observed in human skeletal muscle cell cultures from both non-diabetic and type-2 diabetic subjects [39]. In the cell model studied here, the observed moderate increase in glucose transport activity correlated closely with the increased GLUT1 protein expression in membranes (Fig. 3). The underlying mechanism may be stabilization of GLUT1 mRNA levels by TNF [40]. Increased basal glucose uptake by TNF was also reported from in vivo studies: infusion of TNF reduced plasma glucose levels and increased basal uptake of glucose into muscle cells, suggesting that TNF may, in addition to insulin, also regulate glucohomoeostasis [41]. Very recently, data from TNFR double knock-out mice subjected to diet-induced diabetes, which was not coupled to leptin deficiency, also indicated that TNF may have a protective rather than a deleterious role in the development of diabetes under these conditions, as these mice were hyperinsulinaemic, although still normoglycaemic [16]. The protective role of TNF in muscle cells may be an increased basal glucose transport potential, which, at least partially, could compensate insulin resistance caused by a factor other than TNF during progression of obesity. As numerous in vivo studies with NIDDM patients have shown that these patients in fact suffer from insulin resistance in skeletal muscle tissue with the typical reduced ability to take up glucose [28], but in vitro neither skeletal muscle cells nor muscle fibres [39,42] showed this phenomenon in response to TNF, an additional factor or a totally TNF-independent mechanism appears to be critically involved in muscular insulin resistance of obese NIDDM patients.

                  We have here obtained several lines of evidence for an important role for an nonesterified fatty acid as the major insulin-resistance-inducing agent in skeletal muscle cells. Apart from known metabolic interference of nonesterified fatty acid with pyruvate dehydrogenase, until now unknown additional actions of the predominant serum nonesterified fatty acid, palmitate, were revealed. In particular, palmitate was found to be a negative regulator of insulin-induced phosphorylation cascades at the level of IR, IRS-1 and Akt kinase, which is thought to be the immediate regulator of GLUT4 translocation. Brozinick et al. [43] demonstrated that insulin activates Akt/PKB in isolated rat muscle. Of note is the finding that palmitate, but neither shorter-chain or longer-chain fatty acids, induced insulin resistance, indicating a selective action. This inhibitory cross-talk with IR signalling cascades sheds new light on mechanisms of insulin resistance in muscle cells and may provide a rational explanation for discrepant results obtained in investigations of muscle cell models and muscle tissue. Interestingly, insulin-stimulated Akt kinase activity is reduced in skeletal muscle from NIDDM patients [44], but we clearly demonstrated in the in vitro cell model that TNF does not interfere with Akt kinase signalling, indicating that palmitate may be the relevant mediator in vivo.

                  The role of nonesterified fatty acids is also interesting in the context of the finding that palmitate reverses the TNF-mediated enhancement of GLUT1-dependent basal glucose transport (Fig. 8). Adapting the hypothesis that TNF may exert protective functions during diet-induced obesity [16], nonesterified fatty acids may render muscle cells fully incapable of regulating glucohomoeostasis by inhibiting both insulin- and TNF-triggered glucose transport.

                  The role of skeletal muscle cells during the development of insulin resistance and NIDDM has been put into a new perspective by an elegant study employing muscle-specific homozygous deletion of IR [45]. The surprising finding was that the resulting complete insulin resistance in muscle did not provoke diabetes, but rather altered fat metabolism associated with type-2 diabetes, favouring increased fat depots, elevated serum triacylglycerols and elevated nonesterified fatty acids. This suggests that other organ systems, particular liver and fat, can fully compensate muscular deficiency and that, at least in this model, different from previous assumptions, muscle cell-associated insulin resistance may not be the initiating cause of diabetes. However, it should be remembered that GLUT4–/– knock-out mice [46], in contrast with GLUT4+/– heterozygous animals [47], also did not develop diabetes, as complete deficiency of GLUT4, in contrast with only reduced GLUT4 levels, resulted in compensatory replacement by GLUT1 receptors. Thus quantity may determine a new quality and the tissue-specific abolition of all IR functions in muscle tissue may induce certain compensatory reactions in a similar way. The recent creation of knock-out mice deficient in white adipose tissue leads to an even more differentiated view of these complicated interactions [48]. Total lack of white adipose tissue associated with inability to esterify nonesterified fatty acids in adipocytes resulted in elevated nonesterified fatty acid and triacylglycerol levels and severe symptoms of diabetes, with hyperglycaemia and hyperinsulinaemia [48]. Thus both the lack (as in lipodystrophy) and the overproduction (as in NIDDM) of adipose tissue are associated with increased nonesterified fatty acid levels and may induce, in appropriate circumstances, pathological states culminating in diabetes. The TNF receptor knock-out mice seem to represent a similar situation; functions of TNF during obesity may tilt from one extreme to the other, from deleterious to protective, deleterious when leptin deficiency or leptin resistance is associated with obesity [15,49] and protective in normal diet-induced obesity [16].


                  This work was supported by Boehringer Ingelheim Pharma KG and by the Bundesministerium für Bildung, Wissenschaft, Forschung und Technologie (grant 0316500C, B3.10E) and the Deutsche Forschungsgemeinschaft (Mu625/5).