Impairment of Insulin Signaling in Human Skeletal Muscle Cells by Co-Culture With Human Adipocytes

Reagents for SDS-PAGE were supplied by Amersham Pharmacia Biotech (Braunschweig, Germany) and by Sigma (München, Germany). BSA (fraction V, fatty acid free) was obtained from Boehringer (Mannheim, Germany), protein A trisacryl beads were from Pierce (Rockford, IL), and human recombinant TNF-α was from Sigma. Polyclonal anti-IRS-1 antiserum was a gift from Dr. J.A. Maassen (Leiden, the Netherlands). Polyclonal antibodies anti-phospho-Akt (Ser473) and anti-Akt were supplied by Cell Signaling Technology (Frankfurt, Germany). The anti-phosphotyrosine antibody (RC-20) coupled to horseradish peroxidase (HRP) was from Transduction Laboratories (Lexington, KY). HRP-conjugated goat anti-rabbit IgG antibody was from Promega (Mannheim, Germany). Collagenase CLS type 1 was obtained from Worthington (Freehold, NJ), and culture media were obtained from Gibco (Berlin, Germany). Primary human skeletal muscle cells and supplement pack for growth medium were obtained from PromoCell (Heidelberg, Germany). Sterile plasticware for tissue culture was purchased from Falcon (Becton Dickinson, Heidelberg, Germany). Membrane inserts for six-well culture dishes with a pore size of 0.4 μm (pore density 1.6 × 10⁶ per cm²) and insert companion plates were supplied by Falcon. All other chemicals were of the highest analytical grade commercially available and were purchased from Sigma.

Satellite cells were isolated from M. rectus abdominis by enzymatic digestion with trypsin followed by a purification step with fibroblast-specific magnetic beads to prevent contamination with fibroblasts. After two passages, the myoblasts are characterized by the manufacturer (PromoCell) using immunohistochemical detection of sarcomeric myosin in differentiated cultures at 100% confluence (8 days). Primary human skeletal muscle cells of three healthy Caucasian donors (male, 9 and 28 years of age [M9, M28] and female, 10 years old [F10]) were supplied as proliferating myoblasts (5 × 10⁵ cells). For all donors, chronic diseases, including diabetes and a family history of diabetes, were excluded. These cells were grown to confluence in 25-cm² flasks, trypsinized, and subsequently seeded in 175-cm² flasks at a density of 1 × 10⁶ cells. After two passages, ∼5–7.5 × 10⁷ cells were harvested and stored until further use as frozen aliquots containing 2 × 10⁶ myoblasts. For an individual experiment, myoblasts were then seeded in six-well culture dishes (9.6 cm²/well) at a density of 10⁵ cells per well and cultured in α-modified Eagle’s/Ham’s F-12 medium containing Skeletal Muscle Cell Growth Medium Supplement Pack (PromoCell) to near confluence. The cells were then differentiated and fused by culture in α-modified Eagle’s medium supplemented with 2% FCS for 7 days. The myocytes were cultured in differentiation medium without FCS for 24 h before being used for any experiment.

Adipose tissue samples were obtained from the mammary fat of 12 normal-weight or moderately overweight women (BMI 22–29.4 kg/m², between 29 and 50 years of age) undergoing surgical mammary reduction. All subjects were healthy and had no evidence of diabetes according to routine laboratory tests. Adipose tissue samples were dissected from other tissues and minced in pieces of ∼10 mg. Preadipocytes were isolated by collagenase digestion as previously described (16). Isolated cell pellets were resuspended in Dulbecco’s modified Eagle’s/Ham’s F12 medium supplemented with 10% FCS, seeded on membrane inserts (3.5 × 10⁵/4.3 cm²) in a six-well culture dish, and kept in culture for 16 h. After washing, culture was continued in an adipocyte differentiation medium, as detailed previously (16). After a period of 15 days, 60–80% of seeded preadipocytes developed to differentiated adipose cells, as defined by cytoplasma completely filled with small or large lipid droplets. This population of cells, subsequently designated as differentiated adipocytes, was used for the co-culture experiments. Preliminary data in our laboratory have shown a similar hormonal control of adipose differentiation in preadipose cells from mammary adipose tissue and the subcutaneous abdominal depot. Furthermore, the release of leptin and plasminogen activator inhibitor 1 (PAI-1) was found to be comparable from the two depots.

After in vitro differentiation of preadipocytes on membrane inserts, the adipocytes were washed once with PBS and then incubated for 24 h in skeletal muscle cell differentiation medium containing 1 pmol/l insulin. Thereafter, adipocytes were washed twice with PBS, and individual membrane inserts were subsequently transferred to the culture plates containing differentiated (8 days) myocytes in α-modified Eagle’s medium containing 1 pmol/l insulin. This resulted in an assembly of the two cell types sharing the culture medium but separated by the membrane of the insert. Distance from the bottom of the culture dish to the membrane was 0.9 mm. Co-culture was conducted for 48 h. Integrity of the two cell types was routinely checked by light microscopy at the end of the co-culture period.

Muscle cells were treated as indicated and lysed in a buffer containing 50 mmol/l Tris/HCl (pH 7.4), 1% (vol/vol) NP-40, 0.25% (vol/vol) sodium-deoxycholate, 150 mmol/l NaCl, 1 mmol/l EGTA, 1 mmol/l Na₃VO₄, and protease inhibitor cocktail. After incubation for 2 h at 4°C, the suspension was centrifuged at 10,000g for 10 min. Thereafter the supernatant (0.3 mg protein) was used for immunoprecipitation. Anti-IRS-1 antiserum (5 μl) was preadsorbed to 20 μl protein A trisacryl beads, which were then added to the solubilized cell supernatant and incubated overnight at 4°C with gentle rotation. The immunopellet was collected by centrifugation and washed three times with ice-cold Tris-buffered saline (TBS). The pellet was incubated for 10 min at 95°C with 20 μl electrophoresis buffer and centrifuged, and the complete supernatant was used for Western blot analysis.

Immunoprecipitates or total cell lysates were separated by SDS-PAGE using gradient (8–18%) horizontal gels and transferred to polyvinylidene difluoride filters in a semidry blotting apparatus (17). For phosphotyrosine detection, filters were blocked for 60 min in TBS containing 0.05% Tween-20 and 1% BSA. Thereafter, filters were incubated overnight with a 1:10,000 dilution of the anti-phosphotyrosine antibody (RC-20) coupled to HRP and subsequently processed for enhanced chemiluminescence (ECL) detection using SuperSignal Substrate (Pierce). For detection of IRS-1, p-Akt or Akt filters were blocked with TBS containing 0.05% Tween-20 and 5% nonfat dry milk and incubated overnight with a 1:1,000 dilution of appropriate antibody. After extensive washing, filters were incubated with goat anti-rabbit HRP-coupled antibody and processed for ECL detection. Signals were visualized and evaluated on a LUMI Imager workstation using image analysis software (Boehringer Mannheim).

TNF-α and resistin secreted by adipocytes and human skeletal muscle cells into the culture medium were quantified using ELISA kits from either Research Diagnostics or Phoenix Pharmaceuticals (Belmont, CA). Differentiated adipocytes and skeletal muscle cells were incubated for 48 h either individually or under co-culture conditions using the culture medium outlined above. After that time, the cell culture supernatants were collected and concentrated using Vivaspin Columns (Sartorius, Göttingen, Germany). The assay was then performed as recommended by the manufacturer using duplicate samples for all determinations. Sensitivity of the assay procedure was ensured by addition of known amounts of TNF-α or resistin.

All data analysis was performed using Prism (GraphPad, San Diego, CA) or t-ease (ISI, Philadelphia) statistical software. Significance of reported differences was evaluated using the null hypothesis and t statistics for paired data. A P value <0.05 was considered statistically significant.

Human skeletal muscle cells in primary culture have been shown to represent a valuable tool for studies of insulin action in this tissue (18–20). We have now co-cultured these cells with differentiated human adipocytes to investigate the cross-talk between the tissues. Initial experiments indicated that the two cell types can be kept in the same medium and can be co-cultured for at least 48 h without any gross morphological alterations. Using this protocol, cells of three skeletal muscle donors were co-cultured with adipocytes of three to five different fat cell donors; after removal of the fat cells, the myocytes were acutely stimulated with insulin. As presented in Fig. 1A, a strong increase in insulin-stimulated tyrosine phosphorylation of IRS-1 was observed under control conditions.

After co-culture, this insulin response was largely reduced in skeletal muscle cells from donors M9 and F10. In cells from donor M28, a reproducible increase in the basal tyrosine phosphorylation of IRS-1 occurred after co-culture, with no additional effect of insulin (Fig. 1A). Quantification of the data for the three individuals indicated that insulin was unable to produce a significant stimulation of tyrosine phosphorylation of IRS-1 in skeletal muscle cells subjected to co-culture with adipocytes (Fig. 1B). It should be noted that the level of IRS-1 expression in the myocytes was not affected by the co-culture (see Fig. 3B).

We next assessed the effect of the thiazolidinedione compound troglitazone on the induction of insulin resistance in the co-culture system. When added for 48 h in the absence of adipocytes, troglitazone sensitized the skeletal muscle cells toward insulin, with a twofold increase in the effect of the hormone on the tyrosine phosphorylation of IRS-1 (Fig. 1A and B), without affecting IRS-1 expression. Addition of troglitazone to the co-culture 1) completely prevented the perturbation of insulin signaling at the level of IRS-1 (Fig. 1A) and 2) sensitized the skeletal muscle cells toward insulin to the same degree as in the absence of adipocytes (Fig. 1B).

As presented in Fig. 2, the co-culture conditions were also found to induce prominent insulin resistance in the skeletal muscle cells at the level of the serine/threonine kinase Akt, a central component of downstream insulin signaling to glucose transport and glycogen synthesis (21).

Under control conditions, insulin produced a three- to fourfold increase in the Ser473 phosphorylation of Akt in the myocytes, with a significant (P < 0.001) reduction of this response (by 60–70%) after co-culture with adipocytes (Fig. 2A and B). It is worth noting that in cells from donor F10, we observed a reproducible increase in the basal serine phosphorylation of Akt, with no additional effect of insulin (Fig. 2A). As found for IRS-1, the expression of Akt was not affected by the co-culture conditions. Troglitazone strongly sensitized the skeletal muscle cells toward insulin-mediated activation of Akt, with a parallel 1.5- to 2-fold increase in the protein expression of this enzyme (Fig. 2). The sensitizing effect of troglitazone was not observed under co-culture conditions; however, the thiazolidinedione compound restored a completely normal insulin response of Akt serine phosphorylation in the skeletal muscle cells (Fig. 2).

The role of TNF-α in the development of insulin resistance under co-culture conditions was assessed by culturing human skeletal muscle cells for 48 h in the presence of a high dose of the cytokine. Direct addition of TNF-α to the skeletal muscle cells was found to reduce the protein expression of IRS-1 by ∼40% (Fig. 3A), with troglitazone preventing this reduction. In contrast, unaltered IRS-1 expression was observed in myocytes subjected to co-culture (Fig. 3B).

Akt protein expression remained unaffected by TNF-α. As presented in Fig. 4, insulin signaling to both IRS-1 and Akt was completely blocked in skeletal muscle cells cultured in the presence of 2.5 nmol/l TNF-α. Troglitazone ameliorated this insulin resistance, in agreement with our findings in skeletal muscle cells subjected to co-culture with adipocytes. It should be noted that all data on protein phosphorylation were corrected for the actual amount of IRS-1 and Akt present in the immunoprecipitates and lysates, respectively.

The data obtained so far indicated that TNF-α was able to mimic the effect of co-culture on insulin signaling in human skeletal muscle cells. It was therefore of interest to study this effect of TNF-α using lower, physiologically more relevant doses. The dose-response relationship for IRS-1 and Akt expression and phosphorylation is presented in Fig. 5A and B.

As can be seen from the data, insulin signaling to IRS-1 remained unaffected at 0.15 and 1 nmol/l TNF-α, with a dose-dependent reduction of IRS-1 expression. Insulin-induced signaling to Akt was also reduced in a dose-dependent fashion by TNF-α; however, even at 1 nmol/l of the cytokine, this reduction was not more than ∼30% (Fig. 5B). Only at 2.5 nmol/l TNF-α was the perturbation of insulin signaling in the skeletal muscle cells comparable to the results obtained in the co-culture of myocytes and adipocytes.

The dose-response data described above make it unlikely that TNF-α represents a major mediator of insulin resistance in the co-culture system. To further address this issue, attempts were made to quantify the TNF-α concentration in the cell culture supernatant of skeletal muscle cells, adipocytes, and the cells in co-culture using a highly sensitive TNF-α ELISA assay (detection limit, 0.2 pmol/l). Cell culture supernatants were collected and concentrated 10-fold. Essentially no TNF-α was detected under these conditions after 48 h culture. When 0.3 pmol/l of TNF-α was added to the culture medium, this amount of the cytokine was reproducibly detected, thus validating the sensitivity of the assay.

The same analysis was subsequently performed for the recently detected hormone resistin. This peptide has been implicated as an adipocytokine leading to insulin resistance (13). We used a resistin ELISA with a detection limit of 5 pg/ml. As shown in Fig. 6, skeletal muscle cells were found to secrete a measurable amount of resistin (10–20 pg/ml) into the culture medium, whereas the hormone remained below the detection level in the supernatant of the differentiated adipocytes. Under co-culture conditions and in the presence of troglitazone, resistin was also detected in the culture medium, with a trend for even lower values compared with skeletal muscle cells alone (Fig. 6).

Skeletal muscle accounts for ∼80% of total glucose disposal under insulin-stimulated conditions (22), and defects of insulin action in this tissue are central to the pathogenesis of type 2 diabetes. Based on animal studies, it has been postulated that the overproduction of TNF-α and resistin by adipose tissue may link obesity to type 2 diabetes (5,13). However, it remains questionable if this concept may be translated to humans (12,14). Furthermore, adipose tissue is also dispersed within muscle (15). This intramuscular fat increases with obesity (23), which may lead to a paracrine interaction between adipose and muscle cells potentially involving much lower concentrations of adipocyte factors. We report here for the first time on the co-culture of human adipocytes and human skeletal muscle cells, a generally accepted approach for studies on the paracrine interaction between two cell types (24,25). Our results support the notion of a significant cross-talk between fat and muscle cells but exclude the involvement of TNF-α and resistin in this process.

A key finding of the present investigation is the observation of a profound disturbance of insulin signaling in muscle cells subjected to co-culture. This involves the absence of insulin-stimulated phosphorylation of IRS-1 and the reduced ability of the hormone to stimulate Akt-kinase, without affecting the expression of these proteins. Recent studies on insulin signaling in human skeletal muscle of obese type 2 diabetic patients have shown that insulin was completely unable to stimulate the phosphatidylinositol (PI) 3-kinase pathway in muscle biopsies of these subjects, as measured by the absence of both insulin-regulated phosphorylation of IRS-1 and activation of PI 3-kinase (26,27). These data are in excellent agreement with the present results obtained in the co-culture system and define the IRS-1/PI 3-kinase pathway as a key step in the development of insulin resistance. An apparent difference between adolescent and adult muscle donors was observed, concerning the potential mechanism underlying the absence of insulin-stimulated phosphorylation of IRS-1 (see Fig. 1). However, several lines of evidence support the view that this difference more likely reflects an inter-individual variation in the sensitivity of the myotubes toward the modulatory crosstalk with adipocytes. First, in contrast to rodent models, the proliferative capacity and the program of myogenic differentiation were found to be identical in 9-year-old and 60-year-old humans (28,29). Second, we have determined the differentiation-dependent gene expression profile (including insulin receptor, IRS-1, and IRS-2) of myoblasts from the three donors and did not detect any difference (J.E. and T. Gromke, unpublished data). Third, a consistent response toward TNF-α was observed for the three individuals (Figs. 3 and 4). At the present stage, the precise mechanism leading to the abrogation of insulin-stimulated tyrosine phosphorylation of IRS-1 after co-culture remains unknown.

Using in vitro incubations of skeletal muscle strips from type 2 diabetic patients, Krook et al. (30) reported a marked reduction in insulin-stimulated Akt kinase at high concentrations of the hormone. In subsequent studies using in vivo insulin stimulation, however, the phosphorylation and activation of the three Akt isoforms was found to be normal in skeletal muscle from obese type 2 diabetic subjects (31), and the role of Akt for human insulin resistance has been questioned. In contrast, our data suggest that the negative cross-talk between fat and muscle may largely abrogate insulin-regulated activation of Akt. Some limitations of our in vitro system may explain these discrepancies. Thus, 1) desensitization of Akt may be transient and reversible, 2) fiber type-specific insulin signaling (32) must be considered, and 3) only a subset of skeletal muscle tissue is subject to the paracrine interaction in vivo, making potential alterations undetectable in a large muscle sample. Interestingly, in a recent study, Storgaard et al. (33) reported a tendency for reduced activation of Akt in skeletal muscle of first-degree relatives of type 2 diabetic patients after in vivo insulin stimulation. Additional work will be needed to clarify the role of Akt in skeletal muscle insulin resistance.

The thiazolidinedione compound troglitazone completely prevented the insulin resistance in skeletal muscle cells when present during the co-culture with adipocytes. This observation excludes nonspecific alterations of the myocytes and shows that our co-culture approach should be useful for defining novel pharmacological interventions of the cross-talk between fat and muscle. It is generally assumed that troglitazone acts primarily on adipocytes by binding to the nuclear peroxisome proliferator-activated receptor-γ (PPAR-γ) and that the improvement of insulin responsiveness in skeletal muscle of type 2 diabetic patients is indirect (34). This may result from a decrease in elevated free fatty acid levels and a change in the body distribution pattern of adipose tissue (35). As reported by Park et al. (36), PPAR-γ is also present in primary cultured human skeletal muscle cells, with increased expression upon culture with troglitazone. We have confirmed this increased expression of PPAR-γ in our cultures of human skeletal muscle cells (data not shown) and show here a prominent sensitization of insulin signaling in skeletal muscle cells exposed to troglitazone. Our data do not exclude effects of troglitazone on the release of adipocyte factors; nevertheless, it is evident that the direct effect on human skeletal muscle cells is sufficient to prevent the insulin resistance induced by the cross-talk with adipocytes.

When added at high concentrations, TNF-α was able to mimic the disturbance of insulin signaling found after co-culture; however, concentrations <1,000 pmol/l were marginally effective. This result differs markedly from our earlier study on human adipocytes, where we reported that TNF-α at 10 pmol/l was able to attenuate insulin signaling (10). Thus, auto-/paracrine effects of TNF-α on adipocytes may be much more relevant for the induction of insulin resistance in adipose tissue than the effects of the cytokine on skeletal muscle. This notion is strongly supported by a recent study on human skeletal muscle cells showing no effect of 300 pmol/l TNF-α on insulin-mediated activation of Akt kinase (19). Importantly, we were unable to detect TNF-α in the co-culture medium despite using a highly sensitive assay with a detection limit of 0.2 pmol/l. Taking into account that the culture medium was concentrated 10-fold, the TNF-α level during the co-culture can be estimated to be <0.02 pmol/l. These considerations strongly suggest that TNF-α is not responsible for the insulin resistance observed in skeletal muscle cells after the co-culture procedure. Consistent with this suggestion, no correlation between insulin sensitivity and subcutaneous adipose tissue TNF-α expression could be detected in obese type 2 diabetic patients (37). Furthermore, TNF-α is expressed at much lower levels in human adipose tissue compared with rodents (11). However, it cannot be ruled out that TNF-α induces muscle insulin resistance in a synergistic fashion together with other yet unidentified adipocyte factors.

The peptide resistin represents such a recently detected novel adipocyte factor, and based on rodent studies, it has been postulated that resistin acts as a hormonal link between obesity and insulin resistance (13). We show here that resistin is undetectable in the supernatant of differentiated mammary human adipocytes and therefore conclude that resistin does not mediate the negative cross-talk between fat and muscle in our co-culture system. Our findings agree with two recent reports showing low/absent expression of resistin in human adipocytes, with a high interindividual variability and no correlation to BMI (14,38). Instead, resistin was found to be predominantly expressed in mononuclear cells (14). Interestingly, resistin appears to be secreted by untreated skeletal muscle cells, with a tendency to a reduction in the co-culture. This clearly rules out resistin as a mediator of insulin resistance in human skeletal muscle. In addition to cytokines, free fatty acids may also be involved in the negative cross-talk between fat and muscle (39). However, studies by Kausch et al. (40) suggest that palmitate, 2-bromo-palmitate, and linoleate do not modify insulin signaling in primary cultured human skeletal muscle cells. Initial experiments in our laboratory confirm these findings and make it unlikely that the results obtained in the co-culture can be explained by the release of free fatty acids.

In summary, we describe here the co-culture of human skeletal muscle cells and human adipocytes as a novel model system for studies on the paracrine interaction between fat and muscle. Our data show that adipocytes induce a rapid disturbance of insulin signaling in skeletal muscle cells that is not mediated by TNF-α and resistin. It is suggested that a complex array of adipocyte factors is responsible for generating muscle insulin resistance.

This work was supported by the Ministerium für Wissenschaft und Forschung des Landes Nordrhein-Westfalen, the Bundesministerium für Gesundheit, EU COST Action B 17, the Deutsche Forschungsgemeinschaft (FOR-441, TP7), and, in part, by a grant from Sankyo Co., Tokyo, Japan.

We wish to thank Prof. R. Olbrisch and his team, Department of Plastic Surgery, Florence-Nightingale-Hospital Düsseldorf, for support in obtaining adipose tissue samples. The secretarial assistance of Birgit Hurow is gratefully acknowledged.