Ablation of very long acyl chain sphingolipids causes hepatic insulin resistance in mice due to altered detergent-resistant membranes§


  • Potential conflict of interest: Nothing to report.

  • Supported by the Minerva Foundation and the Israel Science Foundation (0888/11). J.-W.P was supported by National Research Foundation of Korea Grant NRF-2010-357-C00069 from the Korean Ministry of Education, Science, and Technology.

  • Joo-Won Park is currently affiliated with the Department of Biochemistry, School of Medicine, Ewha Womans University, Seoul, South Korea.


Sphingolipids are important structural components of cell membranes and act as critical regulators of cell function by modulating intracellular signaling pathways. Specific sphingolipids, such as ceramide, glucosylceramide, and ganglioside GM3, have been implicated in various aspects of insulin resistance, because they have been shown to modify several steps in the insulin signaling pathway, such as phosphorylation of either protein kinase B (Akt) or of the insulin receptor. We now explore the role of the ceramide acyl chain length in insulin signaling by using a ceramide synthase 2 (CerS2) null mouse, which is unable to synthesize very long acyl chain (C22-C24) ceramides. CerS2 null mice exhibited glucose intolerance despite normal insulin secretion from the pancreas. Both insulin receptor and Akt phosphorylation were abrogated in liver, but not in adipose tissue or in skeletal muscle. The lack of insulin receptor phosphorylation in liver correlated with its inability to translocate into detergent-resistant membranes (DRMs). Moreover, DRMs in CerS2 null mice displayed properties significantly different from those in wild-type mice, suggesting that the altered sphingolipid acyl chain length directly affects insulin receptor translocation and subsequent signaling. Conclusion: We conclude that the sphingolipid acyl chain composition of liver regulates insulin signaling by modifying insulin receptor translocation into membrane microdomains. (HEPATOLOGY 2013)

Insulin resistance is the main pathological feature of type 2 diabetes mellitus, one of the most prevalent metabolic disorders worldwide. Insulin resistance is defined as a condition in which a physiological concentration of insulin is not able to properly elicit its anabolic responses in its peripheral target tissues, including liver, skeletal muscle, and adipose tissue. Several factors, such as obesity and inflammation, induce insulin resistance, although the exact mechanism of insulin resistance is not fully understood. Recently, sphingolipids (SLs) have emerged as important mediators of insulin resistance, since metabolic factors that cause insulin resistance (e.g., tumor necrosis factor-α, free fatty acids, and glucocorticoids) were shown to regulate SL metabolism.1

Ceramide, the backbone of all complex SLs, is synthesized in mammals by a family of six ceramide synthases (CerS).2 Each CerS generates ceramide with a different acyl chain length, with CerS1 and CerS5/6 generating long acyl chain ceramides (C18- and C16-ceramides, respectively) and CerS2 generating very long acyl chain ceramides (C22-C24-ceramides). The CerS display a distinct tissue distribution, with CerS2 found at particularly high levels in liver, kidney, and lung.3-5 Recently, it has become apparent that the acyl chain length of ceramides, and of downstream SLs, is much more important than once thought, with different ceramides playing discrete roles in signaling pathways. For example, C18-ceramide and C16-ceramide play opposing roles as proapoptotic and prosurvival molecules in endoplasmic reticulum stress,6 and overexpression of CerS2 protects HeLa cells from radiation-induced apoptosis, while overexpression of CerS5 promotes radiation-induced apoptosis.7

Both ceramide and complex glycosphingolipids have been implicated in regulating insulin signaling pathways.8, 9 Ceramide interferes with insulin signaling by inhibition of protein kinase B (Akt) activation, which is a downstream mediator of the insulin receptor (IR).8 Akt inhibition by ceramide can occur by three different mechanisms, namely activation of protein phosphatase 2A, activation of the atypical protein kinase C isoform, PKCζ, or recruitment of phosphatase and tensin homolog deleted on chromosome 10 to membrane rafts.8, 10 In addition, ganglioside GM3 causes insulin resistance by inhibiting IR phosphorylation11, 12 due to dissociation of the IR-caveolin-1 complex in membrane microdomains.13

In the present study, we investigate the role of the ceramide acyl chain length on glucose metabolism and insulin resistance using a CerS2 null mouse, which is unable to synthesize very long acyl chain SLs.3, 14 Interestingly, IR and Akt phosphorylation were abrogated in liver due to altered IR translocation into membrane microdomains. We suggest that altering the acyl chain composition of ceramides may be a novel way of modulating insulin resistance and glucose metabolism.


Akt, protein kinase B; CerS, ceramide synthase; DRM, detergent-resistant membrane; IR, insulin receptor; nSMase2, neutral sphingomyelinase 2; SL, sphingolipid; WT, wild-type.

Materials and Methods

Details are provided in the Supporting Information.


CerS2 null mice were generated as described.3, 14 All mice were treated in accordance with the Animal Care Guidelines of the Weizmann Institute of Science Animal Care Committee and the National Institutes of Health's Guidelines for Animal Care.

Statistical Analysis.

Values are expressed as the mean ± SEM. Statistical significance was calculated using a Student t test or repeated-measurements two-way analysis of variance with a post hoc Student t test using Prism software (Graphpad).


Hypoglycemia and Glucose Intolerance in CerS2 Null Mice.

To investigate the role of the SL acyl chain length on glucose metabolism, blood glucose levels were measured during feeding and fasting in CerS2 null mice and compared to wild-type (WT) littermate controls. Glucose levels were significantly lower in CerS2 null mice (Fig. 1A). Body weight decreased to a small but significant extent (Fig. 1B), but body composition (i.e., percent of fat and muscle) was unaltered (Fig. 1C). Analysis using metabolic cages showed no significant changes between CerS2 null mice and their littermate controls (Supporting Fig. 1).

Figure 1.

Low blood glucose levels in CerS2 null mice. (A) Blood glucose levels were measured during feeding and fasting. (B) Body weight was measured in WT and CerS2 null mice. (C) Body composition was determined using a Bruker Mini-Spec Analyzer (EchoMRI, Houston, TX). Data are expressed as the mean ± SEM (n = 16). *P < 0.05, ***P < 0.001.

Despite their lower basal blood glucose levels (Fig. 1A), CerS2 null mice displayed impaired glucose tolerance after intraperitoneal glucose injection (Fig. 2A). In addition, an insulin tolerance test revealed a delayed glucose clearance in CerS2 null mice compared to WT littermates, whose blood glucose levels dropped immediately (Fig. 2B). No significant changes were observed in glucose-stimulated insulin levels in plasma (Fig. 2C) or in insulin secretion from isolated pancreas islets (Fig. 2D). We therefore conclude that glucose intolerance is caused by insulin resistance rather than by impaired insulin secretion.

Figure 2.

Insulin resistance in CerS2 null mice. (A-C) Mice were starved for 8 hours followed by injection of (A) glucose (2.0 g/kg) or (B) insulin (0.75 IU/kg) (n = 16). (C) Plasma insulin levels were measured after glucose (2.0 g/kg) injection (n = 16). (D) Secreted insulin levels were analyzed from isolated pancreas islets after glucose treatment (n = 4). Data are expressed as the mean ± SEM. *P < 0.05, **P < 0.01, ***P < 0.001.

Insulin Signaling Is Impaired in the Liver of CerS2 Null Mice.

To elucidate the molecular basis of insulin resistance, insulin signaling was assessed in liver, adipose tissue, and skeletal muscle. After intravenous insulin injection, tyrosine phosphorylation of the IR was significantly reduced in liver, but not in adipose tissue or skeletal muscle of CerS2 null mice (Fig. 3A-C). Akt phosphorylation, which is downstream to IR signaling, was also significantly reduced in liver (Fig. 3A). Likewise, isolated primary hepatocytes showed altered tyrosine phosphorylation of the IR after incubation with 1 nM (Fig. 3D) or 10 nM insulin (Supporting Fig. 2). Akt phosphorylation was similarly reduced in hepatocytes from CerS2 null mice after insulin treatment (Fig. 3D and Supporting Fig. 2).

Figure 3.

IR phosphorylation is impaired in CerS2 null mouse liver. (A-C) Representative western blots (upper panels) and quantification (lower panels) (n = 5) of insulin-stimulated phosphorylation of IR and Akt in (A) liver, (B) adipose tissue, and (C) skeletal muscle. *P < 0.05, **P < 0.01. (D) Tyrosine phosphorylation of IR and serine phosphorylation of Akt upon insulin (1 nM) stimulation in primary hepatocytes from WT and CerS2 null mice. Results are representative of three independent experiments that gave similar results.

Changes in Activity of Other Enzymes in the SL Metabolic Pathway Cannot Explain the Effects on IR Phosphorylation.

In addition to changes in the SL acyl chain length, levels of a number of other enzymes of SL metabolism, or levels of other SLs, are also altered in CerS2 null mouse liver.3 Thus, C16-ceramide levels are elevated, apparently to compensate for the lack of very long acyl chain ceramides, sphinganine levels are elevated, and neutral sphingomyelinase 2 (nSMase2) and glucosylceramide synthase activities are increased.3 Potentially, any of these changes could play a role in impaired insulin signaling in CerS2 null mice. To address these possibilities, Hep3B cells were transfected with CerS5 (which synthesizes C16-ceramide) or with nSMase2. Overexpression of CerS5 or of nSMase2 led to an inhibition of insulin-induced Akt phosphorylation (Fig. 4A,B) with no effect on IR phosphorylation (Fig. 4A,B), consistent with previous studies.8, 15 Likewise, CerS2 transfection (which elevates very long acyl chain SLs) inhibited insulin-stimulated Akt activation but had no effect on IR phosphorylation (Fig. 4A,B). Sphinganine levels can be increased by incubation with the CerS inhibitor, fumonisin B1,16 but this had no effect on either Akt or IR phosphorylation in either Hep3B cells (Fig. 4A,B) or in cultured hepatocytes (not shown). Finally, neither GW4869 nor AMP-DNM, nSMase2 and glucosylceramide synthase inhibitors, had any effect on IR and Akt phosphorylation (Fig. 4C). We conclude that at least two mechanisms are at play in altered insulin signaling, with Akt phosphorylation affected by either elevation of C16-ceramide levels or by increased nSMase activity, whereas impaired IR phosphorylation is altered by the loss of very long acyl chain SLs.

Figure 4.

Altered sphingolipid metabolism in CerS2 null mice had no affect on IR activation. (A) Western blot analyses of insulin-stimulated phosphorylation of IR and Akt in Hep3B cells treated with fumonisin B1 (20 μM) or transfected with plasmids expressing CerS2, CerS5, or nSMase2. (B) Quantification of western blots by densitometry. Data are expressed as the mean ± SEM (n = 4). (C) Insulin-stimulated phosphorylation of IR and Akt in primary hepatocytes after incubation with GW4869 (20 μM) or AMP-DNM (10 μM) for 48 hours. Data are representative of four independent experiments that gave similar results.

Impaired IR Translocation into Detergent-Resistant Membranes.

Since IR activation depends on the biophysical properties of membranes, and in particular, the ability of the IR to partition into lipid rafts,17-19 we isolated detergent-resistant membranes (DRMs) by flotation using an Optiprep density gradient. DRMs from CerS2 null mouse liver were obtained in higher density fractions compared with WT DRMs (Fig. 5A,B), and flotillin-1, a lipid raft marker,20 was redistributed (Fig. 5B). Similarly, protein, cholesterol, and lipid phosphate were also found in higher density fractions (Fig. 5C-E), although their total levels were not changed. The peak of alkaline phosphatase, a DRM marker (Supporting Fig. 3A), and of cholesterol/protein (Supporting Fig. 3B), were also moved to higher density fractions.

Figure 5.

Depletion of very long chain SLs affects the properties of DRMs. (A,B) Livers from (A) WT and (B) CerS2 null mice were homogenized and fractionated on Optiprep density gradients. Each fraction was immunoblotted for clathrin, caveolin-1, flotillin-1, and lyn kinase. Cholera toxin B conjugated to horseradish peroxidase was used to detect GM1. Representative images of four independent experiments, which gave similar results, are shown. (C,D) The amounts of (C) protein, (D) cholesterol, and (E) lipid phosphate were determined in each fraction; total levels are shown in the right-hand column. Data are expressed as the mean ± SEM (n = 4). *P < 0.05, **P < 0.01, ***P < 0.001.

Upon insulin treatment, the IR translocated into DRM fractions in WT liver (Fig. 6A),17, 21 but this was totally abolished in CerS2 null mice, where the IR remained in high-density fractions (Fig. 6B). Similar results have been observed upon cholesterol depletion,17 but this cannot explain the lack of IR translocation in the CerS2 null mouse since cholesterol levels were unaltered (Fig. 5D and 6B). Likewise, ganglioside GM3 inhibits tyrosine phosphorylation of the IR11, 12; however, GM3 levels were not altered in the livers of CerS2 null mice (data not shown). To elucidate whether the lack of IR translocation into DRM fractions inhibits IR phosphorylation, phosphorylated IR levels were determined along the Optiprep density gradient (Fig. 6A-C). Although the IR in non-DRM fractions was phosphorylated, IR phosphorylation in DRM fractions was much more robust than in non-DRM fractions from WT mice (Fig. 6C and D), whereas IR phosphorylation in CerS2 null liver was similar to that in the non-DRM fractions of WT livers (Fig. 6D). Together, these data suggest that the altered biophysical properties of membranes from CerS2 null mice prevent IR translocation into lipid rafts, which inhibits its phosphorylation and subsequent downstream signaling.

Figure 6.

Insulin receptor translocation into DRMs was inhibited in CerS2 null mice. Livers were isolated after mice were treated with insulin (10 IU/kg) and, after homogenization, fractionated on an Optiprep density gradient. Protein was collected by precipitation with trichloroacetic acid and used for western blotting. (A,B) The IR translocated into DRMs in (A) WT mice but not in (B) CerS2 null mice. (C,D) Insulin-stimulated phosphorylation of IR was quantified (C) in each fraction and (D) in DRMs and non-DRMs. Data are expressed as the mean ± SEM (n = 4). *P < 0.05, **P < 0.01, ***P < 0.001.

Impaired glycogen Storage in the Liver of CerS2 Null Mice.

Finally, we examined glycogen storage in CerS2 null mouse liver since the mice were not hyperglycemic despite displaying glucose intolerance. Glycogen storage was significantly reduced in the livers of CerS2 null mice (Fig. 7A),14 and mRNA expression of brain glycogen phosphorylase, which has a role in glycogen degradation, was increased 15.8 ± 4.2-fold (Fig. 7B). mRNA expression of the gluconeogenic genes, phosphoenolpyruvate carboxykinase 2 and fructose bisphosphatase 2, was elevated 4.9 ± 1.6-fold and 6.4 ± 1.5-fold, respectively (Fig. 7C). A pyruvate tolerance test demonstrated increased gluconeogenesis in CerS2 null mice upon provision of sufficient substrates for gluconeogenesis (Supporting Fig. 4). Finally, CerS2 null mice were glucose-intolerant after an oral glucose tolerance test, and expression of genes related to glucose absorption in the small intestine was unaltered (Supporting Figs. 5A,B).

Figure 7.

Reduced glycogen storage in CerS2 null mouse liver. (A) Glycogen content in the liver was determined using the periodic acid-Schiff reagent. The image is representative of three independent experiments (magnification ×10). (B,C) Real-time polymerase chain reaction of genes involved in (B) glycogen degradation and (C) gluconeogenesis (n = 3). Fbp1, fructose bisphosphatase 1; Fbp2, fructose bisphosphatase 2; G6pc, glucose-6-phosphatase; Gys2, glycogen synthase 2; Pepck1, phosphoenolpyruvate carboxykinase 1; Pepck2, phosphoenolpyruvate carboxykinase 2; Pygb, brain glycogen phosphorylase; Pygl, liver glycogen phosphorylase. Data are expressed as the mean ± SEM (n = 3). *P < 0.05.


A number of previous studies have suggested a relationship between SLs and insulin resistance. For instance, SLs accumulate in heart and muscle in type 2 diabetic patients,22, 23 inhibition of ceramide formation ameliorates glucocorticoid- or obesity-induced insulin resistance and enhances whole body oxygen consumption,24, 25 and various factors associated with insulin resistance increase SL synthesis.1, 24 We now demonstrate a role for the acyl chain composition of ceramide and SLs in hepatic insulin resistance.

A role for the SL acyl chain length was determined based on the use of a mouse that is unable to synthesize C22-C24-ceramides and downstream C22-C24-SLs.3, 14 Total ceramide levels were unaltered, since C16-ceramide levels increased.3 Apart from an increased activity of nSMase2 and glucosylceramide synthase, no other enzymes in the SL metabolic pathway were altered,3 and no changes were observed in cholesterol and glycerophospholipid levels. Interestingly, membranes from CerS2 null mice displayed a higher membrane fluidity.3 The mice display severe nonzonal hepatopathy with increased rates of hepatocyte death and proliferation, which leads to formation of multiple hepatic nodules and eventually to noninvasive hepatocellular carcinomas.14 Chronic liver disease, especially nonalcoholic fatty liver disease, is closely associated with hepatic insulin resistance mainly derived from inflammation and triglyceride accumulation.26-28 However, no significant fibrosis, inflammation, or triglyceride accumulation were observed in CerS2 null mice liver.14 Therefore, the hepatic insulin resistance of CerS2 null mice cannot be explained by their hepatopathy.

The glucose intolerance observed in CerS2 null mice was derived from hepatic insulin resistance. The reason for the specificity of IR and Akt phosphorylation in liver, compared with adipose tissue and skeletal muscle, is probably related to the high levels of CerS2 mRNA in liver. Thus, whereas CerS2 comprised >90% of total CerS mRNA in liver (Supporting Fig. 6A,B), it only comprised 20.7 ± 3.2% and 45.8 ± 4.5% of total CerS mRNA in skeletal muscle and adipose tissue (Supporting Fig. 6C,D) (see also Pewzner-Jung et al.3 and Laviad et al.4).

Hepatic insulin resistance in CerS2 null mice was caused by inhibition of IR phosphorylation. Gangliosides and cholesterol have both been reported to influence IR phosphorylation18, 29-31; however, neither of them were altered in CerS2 null mice. The recruitment of the IR into DRMs is essential for the initiation of IR signaling in liver.18 Although there is considerable controversy concerning the precise relationship between DRMs and lipid rafts, DRMs can be used as a simple biochemical tool to examine receptor translocation in specific signaling pathways.17 For the IR, previous studies have shown that cholesterol depletion using β-methylcyclodextrin32 leads to a similar abrogation of translocation into DRMs,32 which is associated with the inhibition of IR phosphorylation. Cholesterol depletion also results in the redistribution of flotillin.32 However, in the current study, the density of DRMs is also significantly altered, implying a more extensive disruption of DRMs than observed after cholesterol depletion. This disruption in the CerS2 null mouse is perhaps, in retrospect, not surprising, since lipid acyl chain length is a key factor in maintaining membrane structure. Lipid rafts consist of tightly packed, liquid-ordered SLs/cholesterol/saturated phospholipids.33 Very long chain SLs comprise >70% of the total SLs in the liver of WT mice,4 and these very long chain SLs interdigitate with the other half of the lipid bilayer, enhancing formation of the tightly packed, liquid-ordered phase.30 Upon depletion of very long chain SLs, the lipids are likely to pack less tightly, which may cause the mislocalization of DRMs into higher density fractions. Because very long chain SLs have a higher melting temperature than long chain SLs, ablation of CerS2 would affect the formation of gel domains,34 which may be responsible for impaired lipid raft formation. Similarly, an important role of the glycosphingolipid acyl chain length has been shown in lipid raft-mediated signal transduction.31, 35

Although a previous study32 implicated a correlation between hepatic IR phosphorylation and IR translocation into DRMs, we detected higher hepatic IR phosphorylation in DRM fractions compared with non-DRM fractions upon insulin treatment, indicating that the hepatic IR can be activated more efficiently in lipid rafts. The IR could also be phosphorylated in non-DRM fractions, although to a much smaller extent than in DRM fractions, indicating that alteration of IR translocation into DRMs can affect its phosphorylation rather than vice versa. Although abrogation of very long chain SLs inhibited IR phosphorylation, elevating levels of very long chain SLs upon CerS2 overexpression in Hep3B cells did not increase IR phosphorylation. Previous studies18, 21 demonstrated that both cholesterol depletion and cholesterol overload inhibit IR activation, suggesting that a delicate balance of the cholesterol and SL composition is crucial for optimal IR activation.

In accordance with impaired hepatic IR signaling, some of the phenotypes observed in the CerS2 null mouse, such as glucose intolerance, impaired glycogen storage and development of hyperplastic nodules, are similar to those observed in a liver-specific insulin receptor knockout mouse.36 Surprisingly, blood glucose levels were lower in CerS2 null mice despite hepatic insulin resistance, which could be explained by the mitochondrial dysfunction observed in CerS2 null mouse liver, which is caused by impaired activity of mitochondrial respiratory chain complexes (H. Zigdon, A. Kogot-Levin, A.H. Futerman, unpublished data). Because mitochondria are the source of most cellular ATP, mitochondrial dysfunction will result in inefficient energy production, such that the CerS2 null mouse will need more glucose to make the same amount of ATP, and glucose produced from gluconeogenesis will be rapidly consumed. Consequently, the excessive glucose consumption in the liver protects CerS2 null mice from hyperglycemia such as observed in liver-specific insulin receptor knockout mice.36 Clinically, hypoglycemia is frequently encountered in patients with respiratory chain defects,37 and the phenotype observed in the CerS2 null mouse might share similarities with the pathophysiological changes very occasionally observed in patients with concurrent hypoglycemic and glucose intolerance.

Sirtuins and AMP-activated protein kinase modulate various cellular processes involved in energy sensing and in maintaining glucose homeostasis.38 Sirtuin 1 and AMP-activated protein kinase levels were not altered in CerS2 null mouse liver (Supporting Fig. 7A), and of the sirtuin 1-regulated genes, only peroxisome proliferator-activated receptor γ and p53 mRNA levels were elevated (Supporting Fig. 7B),14 which might be a result of elevated hepatic NF-E2-related factor 2 (Supporting Fig. 7C).39 The beneficial effects of sirtuin 1 on insulin sensitivity are associated with a decrease in activation of the mammalian target of rapamycin complex 1 and inhibition of the unfolded protein response.40 In agreement with the unaltered sirtuin 1 levels, p70S6K, which is downstream to mammalian target of rapamycin complex 1, was not altered (Supporting Fig. 7D), and no changes were observed in the unfolded protein response in CerS2 null mice (data not shown).

Adipocytokines such as adiponectin, leptin, and tumor necrosis factor-α are also associated with insulin resistance. Serum adiponectin and leptin levels were elevated ∼ 2-fold in CerS2 null mice (Supporting Fig. 8). Because elevated serum adiponectin and leptin levels are associated with higher insulin sensitivity,41 these results cannot explain insulin resistance; in addition, the adiponectin/leptin ratio, which is associated with insulin resistance,42 was not altered. Serum tumor necrosis factor-α levels were below 30 pg/mL in both WT and CerS2 null mice.

In conclusion, our data suggest that the acyl chain length of SLs plays a critical role in determining the biophysical properties of membranes, regulating lipid raft function, which is important for insulin signaling in liver.


We thank Yuval Dor of the Hebrew University-Hadassah Medical School for critical comments and Tzipora Goldkorn (University of California School of Medicine, Davis) for the nSMase2 plasmid.