Differential Effects of Tumor Necrosis Factor-α on Protein Kinase C Isoforms α and δ Mediate Inhibition of Insulin Receptor Signaling

Tissue culture media and serum were purchased from Biological Industries (Beit HaEmek, Israel). TNF-α was purchased from Sigma (St. Louis, MO). Antibodies to various proteins were obtained from the following sources: monoclonal antibodies to IRβ were purchased from Transduction Laboratories (Lexington, KY). Monoclonal antiphosphotyrosine was obtained from Upstate Biotechnology (Lake Placid, NY). Antibodies to IRS-1, IRS-2, and specific PKC isoforms were purchased from Santa Cruz Biotechnology (Santa Cruz, CA; polyclonal) and from Transduction Laboratories (monoclonal). Horseradish peroxidase and anti-rabbit and anti-mouse IgG were obtained from BioRad (Israel). Leupeptin, aprotinin, PMSF (phenylmethylsulfonyl fluoride), orthovanadate, and pepstatin were purchased from Sigmas. GÖ6976 was purchased from Calbiochem (San Diego, CA).

Skeletal muscle cultures were freshly prepared for each experiment from the limb muscles obtained from 1- to 2-day neonatal mice as described (15,25,34). Fusion of plated myoblasts into mature myotubes occurs 3–4 days after plating. On day 5 in culture, myotubes were transferred to low-glucose (4.5 mmol/l) serum-free DMEM containing 1% BSA for 24 h before study.

Culture dishes (90 mm; Nunc) containing the muscle cultures were washed with Ca²⁺/Mg²⁺-free PBS and then mechanically detached in RIPA (radioactive immunoprecipitation assay) buffer (50 mmol/l Tris HCl, 150 mmol/l NaCl, 1 mmol/l EDTA, 10 mmol/l NaF, 1% Triton X-100, 0.1% SDS, and 1% Na deoxycholate; pH 7.4) containing a cocktail of antiproteases (20 μg/ml leupeptin, 10 μg/ml aprotinin, 0.1 mmol/l PMSF, and 1 mmol/l dithiothreitol) and antiphosphatases (200 mmol/l orthovanadate and 2 μg/ml pepstatin). After scraping, the preparation was centrifuged at 20,000g for 20 min at 4°C. The supernatant was used for immunoprecipitation as described (15,25).

Crude and fractionated lysates of control and treated cultures were subjected to SDS-PAGE and electrophoretic transfer to Immobilon-P (Millipore) membranes. The membranes were subjected to standard blocking procedures and incubated with monoclonal and polyclonal antibodies (25).

Glucose transport was evaluated by measuring 2-deoxy-d-glucose (2DG) uptake as described (15,25). After appropriate treatment, cells were washed three times with 0.5 ml PBS, and the final wash was replaced immediately with 0.5 ml PBS containing 2 mmol/l glucose and tracer amounts (1 μCi/ml) of [³H]-2DG. Cells were then incubated for 15 min at 37°C, washed four times with 0.5 ml cold PBS, and then detached from the wells by addition of 300 μl Triton X-100 (1%) and incubation for 30 min. The contents of each well were transferred to counting vials, and 23.5 ml scintillation fluid was added to each vial. Samples were counted in the [³H] window of a Tricarb scintillation counter. Nonspecific uptake was determined in the presence of excess (100 mmol/l) d-glucose. Experiments were carried in triplicate. Net specific uptake was then calculated as the difference between the total and nonspecific values.

The recombinant adenovirus vectors were constructed as described (35). The dominant-negative mutant of mouse PKCδ was generated by substitution of the lysine residue at the ATP binding site with alanine (36). The mutant δ cDNA was cut from the SRD expression vector with EcoRI and ligated into the pAxCA1w cosmid cassette to construct the Ax vector. Dominant-negative activity was demonstrated by abrogation of autophosphorylation (36).

After differentiation of cultured rat myoblasts into myotubes, the culture medium was aspirated and cultures were infected with medium containing PKC recombinant adenoviruses as described (25).

Specific PKC activity was determined in freshly prepared immunoprecipitates from mature muscle cultures after appropriate treatments as described (15,25). These lysates were prepared in RIPA buffer without NaF. Activity was measured using the SignaTECT PKC Assay System (Promega, Madison, WI). The kit contains phosphatidylserine and diacylglycerol and uses Neurogranin as substrate.

TNF-α has been shown to inhibit insulin-induced phosphorylation of the receptor tyrosine kinase in several cell types (6, 30,37). Figure 1 shows the effects of TNF-α on insulin-induced tyrosine phosphorylation in cultured mouse skeletal muscle. Under control conditions, immunoprecipitated IR displayed an increase in 2tyrosine phosphorylation within 1 min of insulin stimulation. This remained elevated for at least 5 min. IR phosphorylation then decreased toward prestimulation levels by 15 min. As the maximum level of IR tyrosine phosphorylation occurred within 1–5 min of insulin stimulation, this time period of insulin stimulation was used in all subsequent experiments. Effects of TNF-α on insulin-induced tyrosine phosphorylation of IR are shown in Fig. 1B. Pretreatment of cells with TNF-α (1 nmol/l) for 5 min completely inhibited insulin-induced phosphorylation of IR. Longer pretreatment (15–30 min) with TNFα not only blocked the insulin-induced phosphorylation of IR but also reduced the level of phosphorylation to below pretreatment levels.

The first element downstream of IR is the docking family of IRS proteins. In cultured mouse skeletal myotubes, the primary IRS protein involved in insulin signaling is IRS-1, which is also rapidly phosphorylated on tyrosine. As shown in Fig. 1C, immunoprecipitated IRS-1 is tyrosine phosphorylated within 5 min after insulin stimulation. Pretreatment of cells with TNF-α (1 nmol/l for 5 min) markedly reduced phosphorylation of IRS-1 by insulin. In addition, TNF-α inhibited insulin-induced association of IRS-1 with the p85 subunit of PI3-K (Fig. 9, left panel). Effects of TNF-α on the uptake of 2-deoxy-d-glucose (2-DG) were also studied (Fig. 1D). Treatment of myotubes with TNF-α for 5 min did not alter basal 2-DG uptake but significantly reduced the stimulatory effect of insulin. Treatment with TNF-α for 30 min caused a slight but significant increase in basal 2-DG uptake, which was not further increased by insulin. Thus, these findings in cultured mouse skeletal myotubes confirm that TNF-α inhibits early upstream insulin signaling as well as glucose transport.

Several studies indicate that TNF-α inhibition of insulin signaling is mediated by other, as yet unknown, elements. Among the possible candidates is the PKC family of serine-threonine kinases, because TNF-α is known to affect certain members of this family, and some PKC isoforms can modulate IR autophosphorylation. In preliminary studies, we found that cultured mouse skeletal muscle cells express PKC isoforms α, β2, δ, ε, θ, and ζ, in agreement with other studies on mammalian skeletal muscle in vivo and in culture (15,38–40). In a recent study on rat skeletal muscle in primary culture, we showed that insulin induces a rapid and specific physical association between IR and PKCδ and that this physical linkage is essential for the continuation of the IR signaling cascade (26). We therefore reasoned that interference of PKC interactions with upstream elements might be involved in TNF-α effects on IR signaling. Accordingly, in this series of experiments, we sought to determine whether TNF-α affects any physical interactions induced by insulin between various PKC isoforms and IR, IRS-1, and IRS-2. In these studies, we immunoprecipitated IR, IRS-1, and IRS-2 from control cells, insulin-stimulated cells, and insulin-stimulated cells after pretreatment with TNF-α. After SDS-PAGE and transfer, the immunoprecipitated proteins were probed with specific anti-PKC antibodies. Of the PKC isoforms examined (PKCs α, βII, δ, ε, and ζ), only PKCs α and δ were found to coimmunoprecipitate with upstream elements. PKCδ was constitutively associated with both IR and IRS-1, and this association was increased within 5 min by insulin stimulation. PKCα was found to be constitutively associated with IRS-1 only, and this association was decreased within 5 min by insulin stimulation (Figs. 3 and 4). None of the other PKC isoforms examined coassociated with IR, IRS-1, or IRS-2. Moreover, IRS-2 did not coimmunoprecipitate with either PKCδ or -α under basal or insulin-stimulated conditions (not shown). Therefore, all subsequent studies were done on IR and IRS-1 interactions with PKCs δ and α.

Treatment of muscle cells with TNF-α for 5 min before insulin stimulation had opposite effects on PKC association with IR and IRS-1. Thus, on one hand, as shown in Fig. 2, TNF-α increased the association between IR and IRS-1 with PKCδ and reduced the effect of insulin to increase coassociation of these elements with PKCδ. On the other hand, as illustrated in Fig. 3, this cytokine, while also inducing an increase in PKCα-IRS-1 association, prevented the ability of insulin to cause IRS-1 to dissociate from PKCα.

Previous results from our laboratory have shown that insulin stimulation of cultured rat skeletal muscle cells results in activation of specific PKC isoforms (15). We therefore considered the possibility that the inhibitory effect of TNF-α on IR tyrosine phosphorylation might involve modulation of the activity of PKCs δ and α. Accordingly, we examined the effects of insulin and TNF-α on the activity of these PKC isoforms. Figure 4 shows the effects of insulin and TNF-α, separately and in combination, on the activity of specific PKC isoforms. In these studies, PKC isoforms were immunoprecipitated with specific antibodies from control cells and from cells treated with insulin, or with insulin after TNF-α, and the PKC activity of the immunoprecipitated protein was measured by an activity assay, as described in research design and methods. As can be seen, both insulin and TNF-α increased the activity of each of the PKC isoforms, the effect of the latter being at least as effective as that of the former. However, the activity of each of the enzymes from cells treated with TNF-α before insulin was less than that in cells treated with each substance alone.

Activation of PKC isoforms in skeletal muscle by insulin is associated with an increase in the tyrosine phosphorylation state (15). Accordingly, we examined effects of insulin and TNF-α on tyrosine phosphorylation of PKCα and -δ. In one series of experiments, we immunoprecipitated specific PKC isoforms and performed Western blotting with antiphosphotyrosine antibodies. The results are exemplified in Fig. 5, which shows that within 5 min after insulin stimulation, there was a strong increase in tyrosine phosphorylation of PKCs α and δ. Similar to insulin, TNF-α also induced phosphorylation of the α and δ isoforms within 5 min. As both insulin and TNF-α induced tyrosine phosphorylation of PKC isoforms α and δ, we expected that addition of TNF-α followed by insulin might produce an additive effect on these isoforms. However, when myotubes were treated with TNF-α before insulin, the level of tyrosine phosphorylation of each of the PKC isoforms induced by insulin stimulation was lower than that induced by insulin alone. In another series of studies, we immunoprecipitated with antiphosphotyrosine antibodies and immunoblotted with specific anti-PKC antibodies to each of the isoforms. The results were essentially the same as those obtained with immunoprecipitation of the PKC isoforms.

The upstream locations of PKC isoforms α and δ in the insulin signaling cascade suggest that these isoforms may play a direct role in the regulation of IR function. Indeed, our results so far have shown that TNF-α not only affects the interaction of the α and δ isoforms with IR and IRS-1 but also modifies the effects of insulin on these interactions. Thus, it is possible that TNF-α modulation of the activity of these PKC isoforms may affect IR autophosphorylation. We therefore examined more directly whether the ability of TNF-α to reduce insulin-induced activity of PKCs α and δ might play a role in the effects of this cytokine on IR autophosphorylation. To this end, we overexpressed these isoforms using an adenovirus expression system. When infected with recombinant adenovirus constructs containing cDNA for wild-type PKCδ, kinase-inactive dominant-negative PKCδ, or wild-type PKCα, myotubes displayed elevated protein expression of the transfected isoform compared with the expression of the endogenous isoforms (Fig. 6). Overexpression of each isoform, with the exception of dominant-negative PKCδ, also resulted in elevated kinase activity of the specific PKC isoform transfected (data not shown).

Next, we studied the effect of this specific overexpression of PKCα or -δ on insulin-induced IR tyrosine phosphorylation. As shown in Fig. 7A, overexpression of PKCδ resulted in an increase in tyrosine phosphorylation of IR in control and TNF-α treated cells (lanes 1, 2, 7, and 8). This effect was not additive to that of insulin, since insulin stimulation in cells overexpressing PKCδ failed to further increase the phosphorylation state (lanes 2–4). Furthermore, overexpression of PKCδ prevented the inhibitory effect of TNF-α on insulin-induced IR tyrosine phosphorylation (lanes 5 and 6). To confirm that PKCδ overexpression increases IR tyrosine phosphorylation, we infected cells with kinase-inactive dominant-negative PKCδ. The infected cells displayed high protein levels of the inactive PKCδ (Fig. 6). Overexpression of dominant-negative PKCδ abrogated both basal and insulin-induced IR tyrosine phosphorylation (Fig. 7B, lanes 1–4). In addition, the effect of TNF-α on both basal and insulin-induced tyrosine phosphorylation in cells overexpressing dominant-negative PKCδ was not significantly greater than either TNF-α or dominant-negative PKCδ alone (Fig. 7B, lanes 5– 8).

In contrast to the effects of PKCδ overexpression, overexpression of PKCα inhibited both basal and insulin-induced IR tyrosine phosphorylation, as shown in Fig. 8A (lanes 1–4). Interestingly, insulin did appear to increase tyrosine phosphorylation of IR in PKCα -overexpressing cells that had been treated initially with TNF-α (lanes 6 and 8), but the level was considerably less that that in control noninfected myotubes (lane 1). To determine whether PKCα is necessary for the inhibitory function of TNF-α on IR autophosphorylation, we used a selective inhibitor of the α and βI PKC isoforms, GÖ6976 (Fig. 8B). (Because PKCβI is not detected in primary cultures of mouse skeletal muscle, we used this inhibitor as a selective inhibitor against PKCα in this system.) In contrast to effects of PKCα overexpression, inhibition of this isoform increased basal tyrosine phosphorylation of IR (lanes 1 and 2). In addition, inhibition of PKCα abrogated the ability of TNF-α to inhibit IR autophosphorylation (lanes 5 and 6). The results indicate that one component of the effects of TNF-α on IR signa ling involves the activity of PKCα and its association with IRS-1. In parallel, TNF-α reduced responses to insulin by inhibiting insulin-induced activity of PKCδ and its association with IR and IRS-1.

The results therefore show that overexpression of PKCδ and inhibition of PKCα were able to reverse the deleterious effects of TNF-α on IR phosphorylation. To determine whether these treatments of the PKCs δ and α might also influence progression of the IR signal, we examined the association between IRS-1 and PI3-K which is one of the downstream steps in the IR signaling pathway. As shown in Fig. 9 (left panel), insulin induced PI3-K to associate with IRS-1, and TNF-α reduced this effect. Overexpression of PKCδ increased basal association between these two proteins, even in the absence of insulin, and insulin did not further increase this effect. Moreover, the inhibitory effect of TNF-α on IRS-1/PI3-K association was completely abrogated (Fig. 9, middle panel). Effects of inhibition of PKCα by GÖ6976 were essentially the same as those obtained with overexpression of PKCδ (Fig. 9, right panel). Thus, either PKCδ overexpression or PKCα blockade effectively antagonizes the inhibitory effects of TNF-α on signaling from IR to PI3-K.

In this study, we have shown that TNF-α has two distinct mechanisms to oppose insulin action on upstream elements in the IR signaling pathway. In each case, the effect is related to interaction with specific PKC isoforms. These interactions are summarized in Fig. 10. Thus, on one hand, insulin induced a physical interaction of PKCδ with IR and IRS-1 (Fig. 10A), and TNF-α prevented this interaction (Fig. 10B). On the other hand, insulin disrupted the constitutive association that occurs in this preparation between IRS-1 and PKCα (Fig. 10A), and TNF-α caused this association to increase (Fig. 10B). The effects of insulin and TNF-α on both PKCα and -δ were accompanied by tyrosine phosphorylation and activation. Although it has been shown that stimulation of PKCs α and δ can inhibit IR autophosphorylation, and although various effects of TNF-α on PKC have been reported (17), a link between these effects and TNF-α inhibition of IR tyrosine phosphorylation has not been established.

This is the first report to demonstrate that PKCα may actually be constitutively associated with IRS-1 and that insulin causes these elements to physically dissociate. Others have shown, however, that activation of PKCα not only inhibits IR signaling (41,42) but also that PKCα may require IRS-1 for inhibition of IR tyrosine kinase activity (27). Activation of PKCα and its increased physical association with IRS-1 in response to TNF-α, as reported here, are thus consistent with the involvement of IRS-1 in PKCα inhibition of IR signaling. Moreover, the ability of insulin to cause dissociation between IRS-1 and PKCα may, by reducing serine phosphorylation, allow the insulin signal to continue. The results demonstrating that TNF-α activation of PKCα is involved in inhibition of IR signaling are also consistent with an earlier study conducted on a line of CHO cells overexpressing PKCα. Activation of PKCα in these cells by the phorbol ester, TPA (tetradecanoylphorbol acetate), also inhibited tyrosine phosphorylation of IRS-1 in response to insulin stimulation (42). The results obtained in studies on overexpression and blockade of PKCα indicate that this isoform is essential for the inhibitory effect of TNFα on insulin signaling. Thus, overexpression of PKCα alone was sufficient to block both basal and insulin-induced IR tyrosine phosphorylation, and treatment with a PKCα inhibitor elevated basal IR autophosphorylation and blocked the inhibitory effect of TNFα. Moreover, the fact that PKCα associates with IRS-1 and not with IR further indicates that IRS-1 is required for the inhibitory effect of PKCα and that it plays a role in reduction of insulin signaling, as previously suggested (27, 29).

As we have shown, PKCδ is induced by insulin to associate with both the IR and its first downstream element, IRS-1. This association, upstream in insulin signal transduction, is consistent with the suggested role for PKCδ in regulating the function of IR itself (25,26). Our results, showing that an important component of the ability of TNF-α to interfere with IR signaling involves disturbance of the insulin-induced physical association between PKCδ and IR/IRS-1, further strengthen this idea. Additional support for an upstream location of PKCδ in insulin signaling is provided by results obtained in rat skeletal muscle showing that specific inhibitors of PI3K fail to inhibit insulin-induced activation of PKCδ (15).

We found that overexpression of PKCα reduced both basal and insulin-induced tyrosine phosphorylation of IR, whereas inhibition of PKCα increased IR tyrosine phosphorylation in the basal state. These results are in agreement with a previous study (27) in which it was shown that PKC isoforms α, δ, and θ, when coexpressed with HIR (human insulin receptor) and IRS-1 in HEK293 cells, inhibited HIR tyrosine phosphorylation, probably through serine/ threonine phosphorylation of IRS-1. PKCδ, on the other hand, appears to have a dual effect on IR phosphorylation. In addition to serine phosphorylation, which was observed in cells overexpressing wild-type PKCδ even without insulin stimulation (data not shown; see also 26), overexpression of PKCδ resulted in IR tyrosine phosphorylation. This was not reported by Kellerer et al. (27). Because PKCs are serine/threonine kinases, the effect of PKCδ to increase IR tyrosine phosphorylation must be mediated by some other factor or factors. One possibility is that PKCδ, by its physical interaction with IR, may have a permissive effect on the autophosphorylation mechanism. Blockade of IR autophosphorylation by overexpression of DMPKCδ is consistent with this possibility. Alternatively, or additionally, PKCδ may affect other elements that mediate this phosphorylation. Another possibility may be related to the reported effects of PKCδ on IR routing (26,43). Recent results from our laboratory suggest that PKCδ increases the routing of IR by modulating the activity of specific proteins involved in the internalization. Internalized IR is relatively highly phosphorylated on tyrosine residues (44).

The results regarding the opposing effects of both insulin and TNF-α on PKCs α and δ suggest that the two isoforms may play opposite roles in insulin signaling. Opposing effects of these isoforms have been reported in other systems and in signaling cascades other than the insulin signaling pathway, such as induction or inhibition of apoptosis, regulation of proliferation and differentiation, and mediation of cell transformation (45,46,47). Our additional finding that blockade of PKCα abrogated TNF-α inhibition of IR signaling further demonstrates that both PKC isoforms are required for its effect.

The results of this study indicate that TNF-α inhibition of insulin-induced IR and IRS-1 phosphorylation, IRS-1/PI3 K association, and glucose transport occurs, in part, via modulation of the activity of PKC isoforms α and δ and their association with upstream elements in the insulin signaling cascade. We, therefore, propose that TNF-α activates PKCα and -δ and influences their association with IR and IRS-1 in a manner that interferes with the ability of insulin to regulate these isoforms. We further suggest that tyrosine phosphorylation, which may occur at distinct and separate sites in response to insulin than to TNF-α, is a key element in this phenomenon. This would be consistent with studies that indicate the importance of tyrosine phosphorylation in directing the substrate specificity of PKC, in particular PKCδ (16). The mechanisms and specific sites of tyrosine phosphorylation of PKCs α and δ in response to insulin and TNF-α are currently being investigated. The apparently mutually antagonistic effects of insulin and TNF-α on major elements upstream in the insulin signaling cascade indicate that there may be a delicate balance among different substrates for the satisfactory propagation of the insulin signal. Moreover, the effects of this cytokine on specific PKC isoform interactions with these upstream elements support a role for TNF-α in insulin resistance of skeletal muscle.

This study was supported in part by the Sorrell Foundation, the Ben and Effie Raber Research Fund, the Harvett-Aviv Neuroscience Research Fund, and grants from the Israel Science Foundation (founded by the Israel Academy of Sciences and Humanities) and the Chief Scientists Office of the Israel Ministry of Health. S.R.S. is the incumbent of the Louis Fisher Chair in Cellular Pathology.

This work represents an essential portion of a thesis submitted by T.R. in partial fulfillment of the PhD degree in the Faculty of Life Sciences at Bar-Ilan University.