Plasma Membrane Subdomain Compartmentalization Contributes to Distinct Mechanisms of Ceramide Action on Insulin Signaling

All reagent-grade chemicals, insulin, methyl-isobutylmethylxanthine, palmitate, dexamethasone, protein A–Sepharose 4B, and BSA were purchased from Sigma-Aldrich. C₂-ceramide was obtained from Tocris and OKA from Calbiochem. Complete protein phosphatase inhibitor tablets were obtained from Boehringer-Roche Diagnostics. Antibodies against caveolin-1 were purchased from BD Biosciences; native PKB/Akt, Ser473-PKB, Thr308-PKB, Ser21/9–glycogen synthase kinase 3α/β (GSK3α/β), and Thr410/403-PKCζ/λ were from Cell Signaling (New England Biolabs); and Tyr307-PP2A, PP2A A-subunit, PKCζ, and hemagglutinin (HA) were from Santa Cruz Biotechnology. Horseradish peroxidase anti-rabbit, -mouse, and -sheep/goat IgGs were from Jackson ImmunoResearch Laboratories and the enhanced chemiluminescent substrate was from Pierce-Perbio Biotechnology. [³H]2-deoxy-d-glucose was obtained from PerkinElmer.

Adenovirus containing the cDNA of wild-type PKCζ (WT-PKCζ), myristoylated PKCζ (myr-PKCζ), or kinase-dead PKCζ (KD-PKCζ) (24) was prepared as previously described (25). All PKCζ constructs contain an HA tag for monitoring their expression. Caveolin-1 enhanced green fluorescent protein (EGFP) adenovirus was constructed according to He et al. (26). Fully differentiated adipocytes were infected with either construct at 150 multiplicity of infection.

3T3-L1 cells were maintained in Dulbecco's modified Eagle's medium (DMEM; Life Technologies) and differentiated into mature adipocytes as described previously (22). Differentiated cells were usually harvested at days 10–12 after confluence. Human fibroblasts were cultured in DMEM/F-12 (Life Technologies) until confluence.

Cells were lysed after experimental manipulation (see figure legends) in an appropriate volume of lysis buffer (27).

Detergent-resistant membranes (DRMs) were prepared as described previously (22). Cells were homogenized into 25 mmol/l MES (2-[N-Morpholino]ethanesulfonic acid), pH 6.0, 150 mmol/l NaCl, 1% (wt/vol) Triton X-100, complete inhibitor tablet, and lysate ran on a sucrose gradient. The gradient was centrifuged at 120,000g for 20 h at 4°C. DRM fractions were then collected and frozen at −20°C until required.

Cell lysates and membrane fractions were subjected to SDS-PAGE on polyacrylamide gels and immunoblotted as previously reported (27). Nitrocellulose membranes were probed with various antibodies as described in the figure legends. Detection of primary antibodies was performed using appropriate peroxidase-conjugated IgGs, and protein signals were visualized using enhanced chemiluminescence (Thermo Scientific Pierce) by exposure to Kodak autoradiographic film.

Adipocyte plasma membrane lawns were prepared as described (28). The PKCζ antibody was detected with cyanin 3 anti-rabbit antibody.

3T3-L1 adipocytes were incubated in serum-free DMEM 5 h before transport studies and then were exposed to 100 μmol/l C₂-ceramide for 2 h, before being treated with 100 nmol/l insulin in the last 30 min. The uptake of glucose was assessed as described before (4).

HA-WT-PKCζ or PKB/Akt was immunoprecipitated from 500 μg of 3T3-L1 adipocyte lysates. Immunocomplexes were captured by incubation with protein A–agarose beads and solubilized in Laemmli sample buffer prior to SDS-PAGE and immunoblotting as described above.

Statistical analysis was carried out using a Student t test. Data were considered statistically significant at P values ≤0.05.

Figure 1,A recapitulates insulin and ceramide effects on one of the most important end points of the insulin-signaling pathway in adipocytes. Glucose uptake in 3T3-L1 adipocytes was increased 4.5-fold after insulin treatment, an effect repressed in the presence of ceramide (Fig. 1,A). It is well established that both ceramide and insulin act through PKB/Akt to mediate their actions (17,22). Indeed, 100 nmol/l insulin for 10 min induced the phosphorylation of PKB/Akt on its Ser473 site in 3T3-L1 adipocytes, and this effect is completely blunted by pretreating the cells with 100 μmol/l C₂-ceramide for 2 h (Fig. 1,B). To investigate the involvement of the PKCζ and PP2A in the effects of ceramide, we examined whether these proteins were activated in the presence of insulin and ceramide. 3T3-L1 adipocytes were infected with an adenovirus vector containing a WT-PKCζ construct tagged with HA. A weak membrane association in the basal state (Fig. 1,C) was observed by confocal immunofluorescence microscopy of plasma membrane lawns. Both insulin and ceramide, added to the cells alone or together, induced an increase in the amount of PKCζ at the PM (Fig. 1,C). As observed previously in L6 muscle cells (19), treatment of 3T3-L1 adipocytes with either 100 nmol/l insulin for 10 min or 100 μmol/l ceramide for 2 h induced the phosphorylation of PKCζ (using a PKCζ/λ antibody that recognizes the kinase phosphorylation sites Thr410/403 of both PKCζ and λ) (Fig. 1,B). Insulin and ceramide added together had no additive effect on PKCζ/λ phosphorylation, indicating that these two factors targeted a similar pool of PKCζ/λ (Fig. 1,B). It has been shown in fibroblasts that insulin inhibited the constitutive activity of PP2A by phosphorylating its catalytic subunit on a Tyr307 residue (29,30). In 3T3-L1 adipocytes, we observed a similar effect of the hormone (Fig. 1,B). However, pretreatment of the cells with ceramide completely abolished the insulin-induced phosphorylation of PP2A (Fig. 1 B), indicating that the phosphatase is fully active in the presence of ceramide. These results indicate that both PKCζ and PP2A are targets for insulin and ceramide in 3T3-L1 adipocytes.

To find out whether both PKCζ and PP2A could associate with PKB/Akt in response to ceramide in our adipocyte culture system, coimmunoprecipitation experiments were performed. 3T3-L1 adipocytes were infected with an adenovirus expressing an HA-tagged WT-PKCζ construct. Infected cells were treated with insulin and ceramide, and WT-PKCζ was immunoprecipitated with an anti-HA antibody. Resulting immunocomplexes were analyzed by Western blotting for PKB/Akt. Figure 1,D shows that, as observed before (22), ceramide, alone or combined with insulin, induced an increased association between PKCζ and PKB/Akt. As negative controls, the use of nonimmune serum or untransfected lysates did not detect any PKB/Akt in the immunoprecipitates (Fig. 1,D). Interestingly, PP2A was absent from these immunocomplexes (Fig. 1,D). Subsequently, we used the same cell extracts to immunoprecipitate PKB/Akt. Figure 1,E shows that, whereas insulin did not induce any association between PP2A and PKB/Akt, ceramide, alone or in the presence of insulin, triggered it. Altogether, these results show that, in response to ceramide, both PKCζ and PP2A can associate with PKB/Akt, possibly targeting different pools of the kinase, because PP2A is not detected in a ceramide-induced PKCζ-PKB/Akt complex (Fig. 1 D).

It is now well accepted that in 3T3-L1 adipocytes ceramide acts mainly, if not exclusively, on PKB/Akt through the PKCζ pathway (13,17). This is confirmed in Fig. 2, which shows that preincubation of 3T3-L1 adipocytes with ceramide, and then with insulin, led to a significant reduction in the ability of the hormone to phosphorylate PKB/Akt on its Ser473 residue. In line with previous work (17), treatment of the cells with 500 nmol/l OKA for the last 30 min of the ceramide incubation time had no effect on the action of the lipid on PKB/Akt. In contrast, infection of cells with an adenovirus-mediated transfer of an inactive dominant negative PKCζ mutant (KD-PKCζ) into these cells counteracted the action of ceramide on PKB/Akt, and on one of its physiological downstream targets, GSK3α/β (Fig. 2,A). However, in contrast to data we previously obtained in muscle cells (19), overexpression of the KD-PKCζ in adipocytes did not affect the basal and insulin-stimulated PKB/Akt phosphorylation state compared with what was observed in untransfected adipocytes or in WT-PKCζ–overexpressing adipocytes (Fig. 2,B). It is interesting to note, however, that the preventive effect of the KD-PKCζ on PKB/Akt was slightly potentiated in the presence of OKA (Fig. 2 A). Overall, these data indicate that, whereas a PP2A activity is present in 3T3-L1 adipocytes, the involvement of the phosphatase to mediate ceramide action remains minimal.

To understand why the PKCζ pathway predominates on the PP2A pathway in 3T3-L1 adipocytes, we investigated the compartmentalization of these two pathways within the cell membrane, keeping in mind that adipocytes are rich in CEM subdomains where ceramides are concentrated (23,31,32). Thus, we isolated lipid raft–containing CEMs by taking advantage of the capacity of lipid rafts to resist nonionic detergent solubilization at 4°C and purified these domains on discontinuous sucrose density gradient (22). Caveolin-1 enrichment is a characteristic of these domains that are usually recovered in the low-density fractions 3–6 of the sucrose gradient (22). In contrast to caveolin-1, PP2A was not found within CEMs at the basal state or after insulin or ceramide treatment (Fig. 3,A). Insulin did not provoke any recruitment of WT-PKCζ to CEMs (Fig. 3,A). Interestingly, as already described in H9c2 cardiomyoblasts (33), the total amount of PKB/Akt did not change in response to insulin in these subdomains and the hormone induced only a weak phosphorylation of the kinase on both Ser473 and Thr308 (Fig. 3,A). In contrast to insulin, treatment of these cells with ceramide induced both the translocation and more than a threefold increase in the phosphorylation of WT-PKCζ on its Thr410 site, and the recruitment of PKB/Akt to these domains (Fig. 3 A). These data confirmed that the activation of PKCζ by ceramide takes place in CEMs. However, it appears that PP2A is absent from these subdomains.

We used a membrane-associated myr-PKCζ mutant to understand whether the presence of a constitutive PKCζ into CEMs could mimic ceramide action. As expected, Fig. 3,B showed that the myr-PKCζ construct was associated with the CEMs in the basal state and that ceramide could not induce a further increase in myr-PKCζ content in these domains (Fig. 3,B). However, a minimal phosphorylation of the kinase was observed at the basal state and ceramide needed to be added to the cells to induce both myr-PKCζ phosphorylation and PKB/Akt recruitment to CEMs (Fig. 3 B). In summary, these data show a crucial role for CEMs in the activation of PKCζ by ceramide and the subsequent sequestration of PKB/Akt.

The heavy non-CEM fractions (high-density fractions 10–12 of the sucrose gradient) do not contain caveolin-1 (data not shown) (22) and are a mixture of proteins contained in detergent-soluble cellular membranes and in the cytosol. In contrast to what was observed in CEMs, PP2A was detected in these fractions (Fig. 4,A). In addition, both PKCζ and PKB/Akt were robustly phosphorylated in response to insulin in these fractions (Fig. 4,A). The lack of phosphorylation of PKCζ by ceramide in these fractions (Fig. 4,A) confirmed that this lipid is principally present in CEMs and acts on a different PKCζ intracellular pool. Surprisingly, a more than 70% decrease in PKB/Akt basal phosphorylation on its two sites was always observed in ceramide-treated fractions (Fig. 4,A). This process was independent of PKCζ because the inactivation of PKB/Akt was still visible in cells overexpressing the KD-PKCζ mutant (Fig. 4 B).

Given the presence of PP2A in these non-CEM fractions, we then decided to investigate whether a ceramide-activated PP2A mechanism could be responsible here for the inhibitory effect of ceramide on PKB/Akt by using OKA. OKA has previously been shown to inhibit PP2A activity and therefore increase PKB/Akt basal phosphorylation in adipocytes (15). We treated 3T3-L1 adipocytes with ceramide for 2 h and added 500 nmol/l OKA during the last 30 min of the incubation before assessing the basal phosphorylation state of PKB/Akt in non-CEM fractions. Figure 5,A showed that incubation of 3T3-L1 adipocytes with OKA was able to reverse the inhibitory effect of ceramide on PKB/Akt basal phosphorylation and to induce the phosphorylation of the kinase to a level observed after OKA treatment alone, suggesting that the whole bulk of PP2A was inhibited in basal and ceramide-treated cells. Furthermore, OKA was also able to prevent an inhibitory effect of ceramide on insulin-induced PKB/Akt phosphorylation observed in non-CEM fractions (Fig. 5,B). These results show that in non-CEMs, an active PP2A pathway for PKB/Akt inhibition exists in 3T3-L1 adipocytes. The fact that the PP2A pathway is only marginally involved in 3T3-L1 adipocytes when considering the global effect of ceramide in cells (Fig. 2) may be explained by the overabundance of CEMs in these cells.

We have shown that both ceramide-activated PKCζ and PP2A pathways coexist in 3T3-L1 adipocytes, although in different membrane subdomains and only to a small extent for the latter. Our hypothesis is that the relative abundance of CEMs versus non-CEMs might determine the route by which ceramide acts on PKB/Akt. To test this hypothesis, we modulated the relative abundance of CEMs by manipulating caveolin-1 expression because this protein is a key structural component required for CEM formation (34). First, to test the effect of caveolin-1 downregulation, we used undifferentiated 3T3-L1 preadipocytes that expressed much lower quantities of caveolin-1 than fully differentiated 3T3-L1 adipocytes (35) (Fig. 6,A). Total expressions of PKCζ, PP2A, and PKB/Akt did not differ significantly in differentiated and undifferentiated 3T3-L1 (Fig. 6,A). As expected, Fig. 6,B shows a near-complete absence of CEMs in 3T3-L1 preadipocytes compared with differentiated 3T3-L1 adipocytes. Interestingly, in these cells, OKA did not potentiate the insulin effect on PKB/Akt, suggesting that both insulin and OKA were targeting the same pool of PKB/Akt into the cell (Fig. 6,C). In both adipocytes (Fig. 2) and preadipocytes (Fig. 6,D), ceramide completely inhibited the Ser473 residue phosphorylation of PKB/Akt by insulin. However, unlike in adipocytes, treatment of the preadipocytes with OKA prevented the ceramide effect on PKB/Akt by 60%, whereas overexpression of the KD-PKCζ mutant did not change the ability of ceramide to suppress the hormonal activation of PKB/Akt (Fig. 6,D). Identical results were observed on both GSK3 isoforms (Fig. 6 D). Thus, decreasing the relative abundance of CEMs in adipocytes favors the ceramide-activated PP2A pathway over the ceramide-activated PKCζ pathway.

To further substantiate the physiological importance of ceramide on PKB/Akt, we reproduced the latter experiment using palmitate (Fig. 6,E). Palmitate can produce long-chain ceramide de novo by a pathway that is dependent on serine palmitoyl transferase. Generated ceramide has been shown to promote the inactivation of the insulin-induced PKB/Akt activation by the same mechanism described above with short-chain ceramides (4). Figure 6,E showed that incubation of preadipocytes with 0.75 mmol/l palmitate for 20 h downregulated the phosphorylation of PKB/Akt by insulin. Pretreatment of the cells with a serine palmitoyl transferase inhibitor, myriocin, completely prevented the inhibitory effect of palmitate on PKB/Akt (Fig. 6,E), indicating that the lipid was acting on PKB/Akt through the synthesis of ceramides. As observed with C2-ceramide–treated cells, if OKA prevented the action of palmitate on PKB/Akt, overexpression of the KD-PKCζ mutant was not effective (Fig. 6 E).

We then examined the effects of an increase of CEM relative abundance by overexpressing caveolin-1 in human fibroblasts. In contrast to adipocytes, only 3–5% of the PM of fibroblasts is of CEM origin (36), and it has been shown that overexpressing caveolin-1 could induce the formation of more of these membrane domains (34). We used a caveolin-1–GFP adenoviral construct that is known to retain the functional characteristics and intracellular distribution of the endogenous caveolin-1 (37). Caveolin-1–GFP overexpression in human fibroblasts did not change the endogenous expression of PKCζ, PP2A, and PKB/Akt (Fig. 7,A). In human fibroblasts, like in preadipocytes, OKA did not increase further the phosphorylation of PKB/Akt by insulin, suggesting that both insulin and OKA act on the same pool of PKB/Akt (Fig. 7,B). Caveolin-1 overexpression did not affect the ability of ceramide to inhibit insulin-induced activation of either PKB/Akt or GSK3α/β (Fig. 7,C). In untransfected cells, OKA prevented ceramide from having an effect, pointing to involvement of a PP2A pathway, whereas in cells overexpressing caveolin-1, OKA was much less efficient (Fig. 7 C). Conversely, although the KD-PKCζ had no effect on control fibroblasts, it was able to partially abolish ceramide effects in caveolin-1–overexpressing cells. Thus, the differential effects of both OKA and the KD-PKCζ in cells overexpressing caveolin-1 and in control cells underscore the importance of CEMs in supporting ceramide action via PKCζ. Furthermore, the results show that the lipid can switch from one mechanism to another, in order to act negatively on PKB/Akt, depending on the submembrane domain composition of the PM of the cells.

In the present study, we investigated the mechanisms by which the sphingolipid-derived second messenger ceramide induced insulin resistance. As it is established that ceramide does not act directly on its PKB/Akt target (17,18), two distinct mechanisms have been proposed to mediate the inhibitory action of the lipid, one involving PKCζ and the other PP2A. Although not mutually exclusive, it was unclear, until now, why one mechanism was favored over the other in a given cell type. Our present data provide an explanation for the alternative use of these pathways toward PKB/Akt inhibition by ceramide. We demonstrate that compartmentalization of the PM into subdomains, and particularly the abundance of CEMs, is a crucial determinant of the pathway used by ceramide to inhibit PKB/Akt phosphorylation. We show that in CEM-enriched PM, ceramide acts through the recruitment of both PKCζ and PKB/Akt within these membrane domains, whereas in cells with low caveolin expression, hence basically devoid of these domains, ceramide inhibits PKB/Akt mainly via PP2A.

The PM has been shown to be heterogeneous in its lipid composition, and some lipid species, such as cholesterol or sphingolipids, are known to segregate into specific subdomains called lipid rafts that form more ordered and less fluid regions within the PM (38). One specific subset of lipid rafts, called CEMs, is composed of membrane regions characterized by the presence of caveolins. These CEMs are believed to act as platforms for conducting a variety of cellular functions by recruiting or excluding specific signaling molecules and also regulating the accessibility of these proteins to other regulatory or effector molecules (38). CEM abundance is largely determined by the level of caveolin expression that is detectable in most cell types but is very abundant in adipocytes and endothelial cells where caveolae invaginations can occupy nearly 30% of the cell surface (39). Ceramide has been shown to accumulate largely in these domains (40), but it is as yet unclear how they could modulate signaling. The inability of ceramide to transfer rapidly between lipid bilayers (41) implies that downstream signaling targets with which ceramide interacts could be recruited to the PM, rather than ceramide itself being translocated internally. One of the best targets of ceramide to be characterized in vitro has been PKCζ (42,43), and very recently a specific protein fragment of the kinase has been demonstrated to bind ceramide (44). Here, we have shown that whereas ceramide directs PKCζ to CEMs to be activated specifically in response to the lipid, insulin promotes the recruitment of PKCζ outside these membrane subdomains where the kinase is activated through phosphorylation by PDK1 (45). Our results suggest that the whole insulin-signaling process occurs outside these CEMs. Submembrane compartmentalization of PKCζ, depending on the stimulus, could explain the two opposite roles that this kinase plays to regulate the insulin-signaling pathway (positive with insulin and negative with ceramide). Some studies suggest that CEMs could also play a role in insulin signaling. Indeed, contradictory data have been published, some showing that insulin receptor and GLUT4 could be localized in CEMs (46), whereas others did not (47,48). More work will be necessary to solve this discrepancy.

We have also shown that the PP2A pathway for ceramide inhibition of PKB/Akt takes place outside the CEMs of the PM. In agreement with this, one study performed in 3T3-L1 adipocytes has shown that ceramide-activated PP2A was able to dephosphorylate a cytosolic insulin-stimulated PKB/Akt construct lacking its PH domain (49). Our present data, in which the endogenous form of PKB/Akt has been studied, demonstrate that at least in adipocytes, a ceramide-stimulated PP2A pathway coexists with the dominant ceramide-activated PKCζ pathway. The fact that the overexpression of the KD-PKCζ mutant did not completely prevent the ceramide effect on PKB/Akt in adipocytes could be explained by the fact that 1) the lack of PKCζ activity in these cells amplified the PP2A pathway activated by ceramide, and 2) the transfection efficiency of adipocytes is known to be quite poor (50). By manipulating caveolin-1 expression levels, we demonstrated that the relative abundance of CEMs within the PM represents a switch for one pathway at the expense of the other. The PP2A pathway is operating only in preadipocytes or fibroblasts with low CEM abundance and to a much smaller extent in CEM-containing cells such as adipocytes. Our previous results on mice adipocytes treated with ceramide suggested a similar shift from a PKCζ-dependent machinery in wild-type adipocytes to a mechanism involving PP2A in caveolin-1 knockout adipocytes (22). Indeed, adipocytes lacking CEMs, although partly resistant to ceramide action, still displayed a 60% reduction in insulin-induced PKB/Akt phosphorylation, suggesting that the ceramide-activated PP2A pathway substituted for the one involving PKCζ in these cells.

In summary, this study points to the importance of membrane composition and CEM abundance to determine the molecular mechanisms by which ceramide inhibits insulin-stimulated PKB/Akt. In adipocytes, the stimulation of glucose transport by insulin requires the activation of both PKCζ and PKB/Akt (Fig. 8,A). In these cells, the preferential sublocalization of ceramide within abundant CEMs drives its negative action on PKB/Akt nearly exclusively through the mechanism involving PKCζ (Fig. 8,B). However, in cells lacking CEMs, ceramide inhibits insulin-activated PKB/Akt through its dephosphorylation by PP2A (Fig. 8 C). These data highlight that redundant pathways exist within cells to mediate negative actions of ceramide on insulin sensitivity. This clearly indicates that targeted approaches to a single pathway (either PKCζ or PP2A) would not be efficient strategies to fight ceramide-induced insulin resistance. Instead, efforts to prevent ceramide accumulation in insulin-sensitive tissues would be a more accurate approach.

This work was supported by INSERM. C.M.B. was supported by the French ministry for research.

No potential conflicts of interest relevant to this article were reported.

Part of the study was presented at the 44th annual meeting of the European Association for the Study of Diabetes, Rome, Italy, 7–11 September 2008.

We thank Isabelle Hainault, Soazig Le Lay, and Xavier Le Liepvre (INSERM U872) for their expert technical support. We are grateful to Dr. Alex Toker (Harvard Medical School, Boston, MA) for providing myr-PKCζ, KD-PKCζ, and WT-PKCζ constructs and Dr. Lucas Pelkmans (Institute of Molecular Systems Biology, Zürich, Switzerland) for the caveolin-1 EGFP plasmid.