Molecular Mechanisms Underlying Fasting Modulated Liver Insulin Sensitivity and Metabolism in Male Lipodystrophic Bscl2/Seipin-Deficient Mice

Abstract

Bscl2^−/− mice recapitulate many of the major metabolic manifestations in Berardinelli-Seip congenital lipodystrophy type 2 (BSCL2) individuals, including lipodystrophy, hepatomegly, hepatic steatosis, and insulin resistance. The mechanisms that underlie hepatic steatosis and insulin resistance in Bscl2^−/− mice are poorly understood. To address this issue, we performed hyperinsulinemic-euglycemic clamp on Bscl2^−/− and wild-type mice after an overnight (16-h) fast, and found that Bscl2^−/− actually displayed increased hepatic insulin sensitivity. Interestingly, liver in Bscl2^−/− mice after a short term (4-h) fast had impaired acute insulin signaling, a defect that disappeared after a 16-hour fast. Notably, fasting-dependent hepatic insulin signaling in Bscl2^−/− mice was not associated with liver diacylglyceride and ceramide contents, but could be attributable in part to the expression of hepatic insulin signaling receptor and substrates. Meanwhile, increased de novo lipogenesis and decreased β-oxidation led to severe hepatic steatosis in fed or short-fasted Bscl2^−/− mice whereas liver lipid accumulation and metabolism in Bscl2^−/− mice was markedly affected by prolonged fasting. Furthermore, mice with liver-specific inactivation of Bscl2 manifested no hepatic steatosis even under high-fat diet, suggesting Bscl2 does not play a cell autonomous role in regulating liver lipid homeostasis. Overall, our results offered new insights into the metabolic adaptations of liver in response to fasting and uncovered a novel fasting-dependent regulation of hepatic insulin signaling in a mouse model of human BSCL2.

Congenital generalized lipodystrophy, also called Berardinelli-Seip congenital lipodystrophy (BSCL), is an autosomal recessive disorder characterized by a near-total absence of body fat from birth or infancy, associated with a muscular appearance and enlarged internal organs (1, 2). Affected individuals develop metabolic abnormalities that include hyperinsulinemia, hypertriglyceridemia, insulin resistance, and type 2 diabetes (3). Hepatosteatosis occurs early, often culminating in cirrhosis and liver failure. Other clinical manifestations of BSCL include acanthosis nigricans, hyperandrogenism, and polycystic ovary syndrome (1–4).

Mutations in the BSCL2 (also called Seipin) gene underlie type 2 BSCL, a more severe form of BSCL (5). The function of Bscl2 is poorly understood. Recently, three independent groups (including ours) have generated Bscl2-knockout (Bscl2^−/−) mice (6–8), which present with generalized loss of adipose tissues accompanied by visceromegaly, particularly hepatomegaly (6–8). They recapitulate most of the metabolic manifestations in type 2 BSCL patients, including hyperinsulinemia, insulin resistance, and hepatosteatosis (6–8). They have hyperphagia and their postprandial glucose and lipid clearance are severely impaired (6–8). However, in contrast with other lipodystrophic mouse models such as A-ZIP-F (9), aP2-SREBP-1c (10), and Agpat2^−/− (11) mice, which develop overt fasting hyperglycemia and severe hypertriglyceridemia, Bscl2^−/− mice are unique in that they display lower glucose levels compared with controls after a short fast (7, 8). They also exhibit hypotriacylglycerolemia upon fasting (7, 8). Their fasting serum free fatty acids and glycerol levels are also significantly lower than those in wild-type controls (7, 8). These unique metabolic phenotypes suggest a markedly different regulation of glucose and lipid metabolism in lipodystrophic Bscl2^−/− mice.

The liver is central in glucose and lipid homeostasis under fasting conditions. Prior work has provided partial insights into the development of hepatosteatosis and lack of hypertriglyceridemia in Bscl2^−/− mice (8). However, the mechanisms that underlie hepatosteatosis and insulin resistance in Bscl2^−/− mice in the fed and/or short fasting state remain obscure. There is also no report detailing the metabolic adaptations in liver of human type 2 BSCL patients or lipodystrophic Bscl2^−/− mice after a prolonged fast. In addition, knockdown of Bscl2 in Hela cells increases triacylglycerol (TAG) synthesis, suggesting that Bscl2 may regulate lipogenesis in nonadipocytes (12). So far, it is not known whether the severe hepatosteatosis in Bscl2^−/− mice is caused by a cell-autonomous role of Bscl2 in regulating liver lipid metabolism.

In this study, we dissected the molecular underpinnings of hepatosteatosis, insulin sensitivity, and metabolic adaptations in liver of Bscl2^−/− mice in the fed, short-term, and prolonged fasting states. We also generated mice with liver-specific deletion of Bscl2 to study the possible autonomous role of Bscl2 in hepatic metabolism. Our experiments documented the absence of any significant abnormalities in liver lipid homeostasis in mice with liver-specific deletion of Bscl2. Instead, we uncovered an unusual modification of liver insulin sensitivity and metabolism in response to prolonged fasting in global-knockout (Bscl2^−/−) mice.

Materials and Methods

Mice

Global Bscl2^−/− mice [previously generated in the lab (7) and backcrossed five times to C57BL/6J background] and liver-specific Bscl2-deficient (Bscl2^Li−/−) mice generated by crossing Bscl2^fl/fl mice with Albumin-Cre transgenic mice, were maintained under standard conditions with controlled 12-hour/12-hour light-dark cycle and 21°C room temperature. Mice were fed with a chow diet ad libitum. Bscl2^Li−/− and their littermate Bscl2^fl/fl (Bscl2^Li+/+) control mice were also fed with a high-fat diet (42% of kcal from fat, Harlan Teklad TD88137) from 4 weeks of age. If not specified, 11–13-week-old male mice were used in most studies. Mice were killed under fed (1000 h) or after 4–6-hour fasts (0900 h to 1300–1500 h; we used 4-hour fast [4hF] throughout the text considering the negligible differences between 4-h and 6-h fasting) or overnight 16-hour fast (16hF; 1700–0900 h). All animal experiments were performed using protocols approved by the Institutional Animal Care and Use Committee (IACUC) at Georgia Regents University and Baylor College of Medicine.

Insulin tolerance test and hyperinsulinemic-euglycemic clamp

Insulin tolerance test (ITT) was performed in mice fasted 4–6 hours or overnight (16 h) and then injected ip with human insulin (Humulin, Novo Nordisk) at 0.75 U/kg. Blood glucose levels were measured by One-touch Ultra glucose meter before and at 15, 30, 60, and 120 minutes after injection. For clamp, a jugular vein catheter was implanted 6–7 days before the experiment. To assess whole-body glucose turnover, we administered to overnight-fasted, conscious mice a priming dose (10 μCi) and then a constant iv infusion (0.1 μCi/min) of HPLC-purified [3-³H] glucose (NEN). We collected blood samples from the tail vein at 0, 50, 55, and 60 minutes to determine basal glucose production. After the basal period, hyperinsulinemic-euglycemic clamps were conducted for 120 minutes with a primed (40 mU/kg) followed by a continuous infusion of humulin (3 mU/Kg/min), a continuous infusion of [3-³H] glucose (0.1 μCi/min) and a variable infusion of 25% glucose to maintain the blood glucose level at 100–140 mg/dl. Plasma samples were obtained at 100, 110, and 120 minutes to determine hepatic glucose production and peripheral glucose disposal rate. Data were calculated as previously described (13, 14). At the end of the clamp, mice were anesthetized with pentobarbital sodium injection and tissues were taken and snap-frozen in liquid nitrogen, and stored at −80°C for analysis.

Liver insulin signaling analysis

In vivo portal vein injection of insulin to examine acute hepatic insulin signaling was performed as previously described (15). Briefly, mice were either fasted for 4 hours or overnight 16 hours followed by anesthesia. Recombinant human humulin (0.75 U/kg body weight [BW]) was administered directly into the portal vein. Exactly 5 minutes later, tissues were exercised and immediately snap frozen in liquid nitrogen. Protein extracts from tissues were analyzed by Western blot. PKCϵ membrane translocation was assessed as previously described in fractionated liver samples (16).

Liver metabolite measurement

TAG in liver or isolated hepatocytes was extracted according to Bligh and Dyer (17), and dissolved in chloroform. A small aliquot (5–30 μl) was removed and dried. The TAG concentration in this aliquot was determined using a triacylglycerol assay kit (Thermo DMA) and normalized to tissue weights or total cellular protein levels (18). Liver diacylglycerol (DAG) was analyzed by Lipidomics Shared Resource, Hollings Cancer Center at Medical University of South Carolina. Lipids were extracted and measured as previously described (19). Data were normalized to total protein levels and expressed as pmole/mg protein. Hepatic ceramide was analyzed as previously described (20). Briefly, total lipids were extracted by Folch method and dissolved in CHCl3: MeOH (2:1) and then applied on thin-layer chromatography plate. The bands were resolved using chloroform: methanol: acetic acid 95:5:0.5 (v/v/v) and visualized after iodine absorption followed by quantification using densitometry and Image J software. C16-ceramide (Avanti Polar Lipids) was used as standard.

Fatty acid oxidation and TAG synthesis in isolated primary hepatocytes

We isolated primary hepatocytes from 4hF mice by collagenase perfusion and plated them in triplicates in the DMEM complete media and changed to Williams complete media after 4 hours' incubation. For the oxidation of fatty acids, we incubated cells in Williams media without fetal bovine serum supplemented with 400 μM ³H-palmitate conjugated on BSA and L-Carnitine for 1.5 hours. The released ¹⁴CO₂ was captured and counted using a scintillation counter. For TAG synthesis, plated hepatocytes were incubated with 360 μM ³H-oleic acid for 2 hours. Total lipids were extracted and resolved by thin-layer chromatography as previously described (21). The corresponding TAG spots were scraped from the plate and counted by scintillation.

Very low-density lipoprotein secretion

Mice were fasted for 4 hours and ip injected with pluronic F-127 (also called poloxamer 407) at 2 mg/g BW. Blood samples were collected before and at 1 and 2 hours after injection. Plasma TAG was measured by Infinity TAG assay kit (Thermo DMA).

Plasma metabolite and hormone measurement

Plasma 3-β-hydroxybutyrate levels were measured using β-hydroxybutyrate (ketone body) colorimetric assay kit (Caymen Chemicals). Plasma Fgf21 levels (Millipore) were measured by ELISA according to manufacturer's instructions.

Immunoblot analysis

Snap-frozen liver was homogenized and lysed in lysis buffer containing 25mM Tris-HCl (pH, 7.4), 150mM NaCl, 2mM EDTA, 1% Triton X-100, and 10% glycerol with freshly added protease inhibitor cocktail (Sigma) and 50mM NaF, 10mM sodium pyrophosphate, 1mM sodium vanadate. The protein concentration was determined by BCA protein assay (Bio-Rad). The same amount of proteins were loaded and immunoblot analysis was carried out according to the standard protocol. The following antibodies were used: Rabbit antibodies against p-Akt-Ser473, Akt, IRβ, insulin receptor substrate (IRS) 1, and IRS2 were from Cell signaling technology, Gapdh (Fisher Scientific), PKCϵ (Santa-Cruz), Na⁺/K⁺ ATPase α1 (Santa Cruz). Image J was used to quantify the intensity of bands.

Reverse transcription and real-time PCR

Total RNA was isolated from tissues or cultured cells with TRIzol (Invitrogen) and reverse transcribed using MLV-V reverse transcriptase using random primers (Invitrogen). Real-time quantitative RT-PCR was performed on the Strategene MX3005. Data were normalized to two housekeeping genes (Cyclophilin A and β-actin) based on Genorm algorithm (medgen.ugent.be/genorm/) and expressed as fold changes relative to wild-type controls. All primers were included in Supplemental Table 1.

Statistical analysis

Quantitative data were presented as mean ± SEM. Other than clamp study, in vivo animal experiments were performed on at least two independent cohorts. Representative results were shown from three independent experiments for in vitro experiments performed with primary hepatocytes. Differences between groups were examined for statistical significance with two-tailed Student t test. P < 0.05 was considered statistically significant.

Results

Bscl2^−/− mice display increased hepatic insulin sensitivity as revealed by hyperinsulinemic-euglycemic clamp

Hepatosteatosis is often accompanied by hepatic insulin resistance. To determine whether Bscl2^−/− mice have insulin resistance, we performed hyperinsulinemic-euglycemic clamp in overnight 16-hour fasted (16hF) mice. We measured hepatic glucose production under basal conditions and under hyperinsulinemic clamped-glucose conditions (with blood glucose levels of Bscl2^+/+ and Bscl2^−/− mice shown in Figure 1A). We observed no difference in hepatic glucose production between the two genotypes under basal conditions (Figure 1B, left bars); however, under insulin-clamp conditions, hepatic insulin sensitivity was markedly increased, with a drastically diminished hepatic glucose output in Bscl2^−/− mice (Figure 1B, right bars) compared with Bscl2^+/+ mice (Figure 1, B and C). This was associated with an elevated glucose infusion rate in Bscl2^−/− mice during clamp (Figure 1D) despite similar glucose disposal rates between the two genotypes (Figure 1E). We next analyzed insulin signaling in the clamped liver of Bscl2^−/− mice which indeed displayed increased Akt phosphorylation at Ser473 (Figure 1F), consistent with the suppressed liver gluconeogenesis under clamp. Immunoprecipitation of IRS1 and IRS2 from total liver extracts confirmed up-regulated tyrosine phosphorylation of these two proteins in Bscl2^−/− mice, further supporting enhanced hepatic insulin signaling under insulin clamp in overnight-fasted Bscl2^−/− mice (Supplemental Figure 1A).

Figure 1.

Bscl2^−/− mice display enhanced hepatic insulin sensitivity under hyperinsulinemic-euglycemic clamp. A, Clamped glucose levels; B, basal and clamped hepatic glucose production; C, hepatic glucose suppression rate; D, glucose infusion rate; and E, glucose disposal rate in overnight (16-h) fasted 13–15-wk-old male Bscl2^+/+ (+/+) and Bscl2^−/− (−/−) mice under hyperinsulinemic-euglycemic clamp (n = 6 per genotype). F, Western blot analysis of liver insulin–dependent Akt signaling under hyperinsulinemic-euglycemic clamp in overnight-fasted Bscl2^+/+ and Bscl2^−/− mice. Three representative animals were shown for each group (n = 6 total). Ratios of p-Akt^Ser473 vs total Akt were calculated after densitometry analyses using Image J software. *, P < .05; **, P < .005 between Bscl2^+/+ and Bscl2^−/− mice.

Prolonged fasting sensitizes hepatic insulin sensitivity in Bscl2^−/− and Bscl2^+/+ mice in vivo

Lipodystrophic Bscl2^−/− mice were found to have severe insulin resistance as assessed by ITT (6, 8). Confirming the previous findings (6, 8), ITT in our Bscl2^−/− mice revealed lower baseline glycemia after 4hF but the glucose levels in Bscl2^−/− mice were resistant to the injected insulin (Figure 2A, upper panel), indicating insulin resistance compared with wild-type mice; the difference in glycemic response to insulin was more evident when the blood glucose was normalized to % change in glycemia (Figure 2A, lower panel). The impaired insulin sensitivity in ITT is in direct contrast with the increased insulin action in hyperinsulinemic-euglycemic clamp, the gold standard for assessing insulin sensitivity in vivo. We next scrutinized the factors that might have contributed to such discrepancy and noticed a difference in the fasting time before ITT (short 4–6 hours fast, 0900 to 1300–1500 h) and hyperinsulinemic-euglycemic clamp (prolonged overnight 16hF, 1700–0900 h). We therefore performed the same ITT in overnight-fasted mice. Surprisingly, in these 16hF mice, both genotypes revealed a very similar hypoglycemic response to acute insulin injection (Figure 2B). When directly comparing the insulin-dependent signaling in the liver after acute injection of insulin through the portal vein, we found that 4hF Bscl2^−/− mice had impaired hepatic insulin-dependent Akt phosphorylation at Ser473 compared with Bscl2^+/+ mice (Figure 2C). It is interesting that after a prolonged (16-h) fast, acute insulin treatment produced comparable levels of Akt phosphorylation in the liver of Bscl2^+/+ and Bscl2^−/− mice (Figure 2D). We did not detect significant difference in terms of basal hepatic insulin signaling as represented by Akt phosphorylation at Ser473 in 4hF and 16hF Bscl2^+/+ and Bscl2^−/− mice before acute insulin stimulation (Figure S1B). Total Akt protein expression was also similar at different fasting times in the two genotypes (Figure S1B). Taken together, these data underscore the importance of the duration of the antecedent fast in hepatic insulin action in lipodystrophic Bscl2^−/− mice.

Figure 2.

Prolonged fasting sensitizes hepatic insulin signaling in Bscl2^−/− mice. ITTs in A, 4–6-h fasted (4hF); and B, overnight 16-h fasted (16hF) 13-wk-old male mice (n ≥ 7). Western blot analysis of liver insulin–dependent Akt signaling after portal vein injection of humulin (0.75 U/kg BW) in C, 4-h fasted; and D, 16-h fasted Bscl2^+/+ and Bscl2^−/− mice. Three representative animals were shown for each group (n = 6 total). Ratios of p-Akt^Ser473 vs total Akt were calculated after densitometry analyses using Image J software. *, P < .05 compared with Bscl2^+/+ mice.

Hepatic insulin sensitivity is not associated with changes in major signaling lipid metabolites in Bscl2^−/− mice

To examine the molecular basis of the fasting-modulated switch in hepatic insulin signaling in lipodystrophic Bscl2^−/− mice, we analyzed hepatic lipid profiles in 4hF and 16hF mice. The liver of Bscl2^−/− mice remained steatotic after 4hF (Figure 3A). Interestingly, 16hF induced the appearance of numerous tiny lipid droplets that were barely discernable under low magnification in Bscl2^+/+ liver whereas 16hF markedly reduced the number and size of lipid droplets in Bscl2^−/− liver (Figure 3A). Quantitation revealed a 3.8-fold more accumulation of TAG in Bscl2^+/+ liver but a 40% reduction of TAG content in Bscl2^−/− liver after 16hF, resulting in similar levels between two genotypes (Figure 3B). Liver TAG is thought not to be detrimental to hepatic insulin sensitivity (22). Rather, accumulation of signaling lipids such as ceramides and DAG in steatotic liver is proposed to contribute to the development of hepatic insulin resistance (23, 24). Interestingly, total hepatic ceramide content was slightly lower in 4hF Bscl2^−/− mice compared with wild-type mice. After 16hF, ceramide levels remained similar between the two genotypes (Supplemental Figure 2A). In contrast, hepatic total DAG concentrations as well as the major liver DAG species (C16:0/C18:1, C16:1/18:1, and Di-C18:1) were all markedly elevated in Bscl2^−/− mice compared with Bscl2^+/+ mice at both 4hF and 16hF conditions (Figure 3C and Supplemental Figure 2B). On the other hand, we failed to observe an overt difference in PKCϵ translocation from cytosol to membrane in Bscl2^−/− liver at both fasting conditions compared with wild-type controls (Figure 3D). The ratios between membrane and cytosol PKCϵ under both fasting conditions did not differ in the two genotypes (Supplemental Figure 2C). PKCϵ directly interferes with insulin receptor kinase activity to modulate downstream tyrosine phosphorylation of IRS (25). Consistent with similar PKCϵ activation under both fasting conditions between Bscl2^+/+ and Bscl2^−/− liver, basal tyrosine phosphorylation rates of IRS1 and IRS2 were not perturbed (Supplemental Figure 2D). Tyrosine phorsphorylation of IRS1 and IRS2 under hyperinsulinemic-euglycemic clamp were even increased in Bscl2^−/− liver (Supplemental Figure 1A). Collectively, these results suggest that ceramides and DAG are not involved in the development of insulin resistance in steatotic Bscl2^−/− liver after 4hF and they also do not play a role in the improved hepatic insulin sensitivity in the liver of Bscl2^−/− after a prolonged fast.

Figure 3.

Lipid signaling metabolites are dissociated with hepatic insulin signaling in Bscl2^−/− mice. A, Liver histology. Sections of paraffin fixed tissues were stained with hematoxylin-eosin and examined by light microscopy. Scale bar = 10 μm. B, Liver TAG under 4hF and 16hF (n = 6–7 in each group). Liver DAG C, after a 4hF and an overnight fast (n = 3 pooled from 6 mice). D, PKCϵ membrane and cytosol localization after a 4hF and 16hF. *, P < .05; **, P < .005 compared with Bscl2^+/+ mice; #, P < .05 compared between same genotype.

Prolonged fasting up-regulates hepatic insulin receptor and IRS in Bscl2^−/− mice

We next analyzed the basal protein expression level of key insulin signaling components, including insulin receptor IRβ, IRS1, and IRS2, in the liver of lipodystrophic Bscl2^−/− mice after a short 4hF and prolonged 16hF. IRβ, IRS1, and IRS2 proteins in 4hF Bscl2^−/− liver were significantly down-regulated compared with 4hF Bscl2^+/+ liver (Figure 4A, and quantitated in Figure 4B). However, after an overnight fast, IRβ protein levels in both Bscl2^+/+ and Bscl2^−/− liver were increased, with the latter reaching that of 4hF Bscl2^+/+ liver. Marked up-regulation of IRS1 protein was only obvious in 16hF Bscl2^+/+ liver relative to that at 4hF. The IRS2 protein in 16hF Bscl2^−/− liver increased to a level similar of that in 16hF Bscl2^+/+ liver (Figure 4, A and B). When comparing the hepatic mRNA expression of these three proteins between 4hF and 16hF animals, we found a grossly similar pattern of changes in expression as observed in Western blots (Figure 4C). Overall, these differences strongly suggest that down-regulation of crucial insulin signaling molecules underlies hepatic insulin resistance in Bscl2^−/− mice under a short fast; while the up-regulation of IRβ and IRS2 may partially contribute to the improvement of liver insulin sensitivity in Bscl2^−/− mice after a prolonged fasting.

Figure 4.

Prolonged fasting up-regulates crucial insulin signaling components in liver of Bscl2^−/− mice. A, Protein expressions of IRβ, IRS1, and IRS2 under 4hF and 16hF in liver of Bscl2^+/+ (+/+) and Bscl2^−/− (−/−) mice. Representative figures were shown. A total of four animals were analyzed in each group. Gapdh was used as loading control. B, Densitometric analysis of protein expression after normalization to Gapdh. Data were expressed as fold changes as relative to Bscl2^+/+ liver after 4hF (n = 4 each). C, qRT-PCR analysis of livers from 4hF and 16hF mice (n = 6 each). Data were normalized to Cyclophilin A and β-actin. *, P < .05; **, P < .005 compared with Bscl2^+/+ mice; #, P < .05, ##, P < .005 compared between same genotype.

Increased de novo lipogenesis and decreased β-oxidation contribute to hepatosteatosis in vitro

The development of hepatosteatosis in Bscl2^−/− mice has been partially attributed to increased lipoprotein uptake by the liver (8). We further analyzed the possible mechanisms in primary hepatocytes isolated from Bscl2^+/+ and Bscl2^−/− mice. Under basal conditions (5-h serum starvation), Bscl2^−/− hepatocytes contained twice as much TAG as wild-type hepatocytes, reflecting the steatotic liver in vivo (Figure 5A). De novo lipogenic gene transcripts including Fasn and Scd1 were significantly increased in Bscl2^−/− hepatocytes compared with wild-type cells (Figure 5B). Among the transcripts for the key enzymes involved in TAG synthesis, Lipin1β mRNA was increased whereas no significant difference in the mRNA expression of Dgat1 and Dgat2 was found. Mgat1, an enzyme that mediates an alternate pathway of TAG synthesis, was drastically up-regulated (Figure 5B). CD36, a gene involved in lipid flux, was also markedly up-regulated, consistent with the increased influx of free fatty acid to the liver in the absence of adipose tissue (Figure 5B). Meanwhile, the expression of genes responsible for gluconeogenesis such as Pepck was not changed but there was a 30% down-regulation of G6Pase mRNA in Bscl2^−/− hepatocyte. Not surprisingly, when incubated with ³H-labeled oleic acid, Bscl2^−/− hepatocytes displayed an enhanced rate of TAG synthesis (Figure 5C). Measurement of ¹⁴C-labeled CO₂ released from complete oxidation of ¹⁴C-palmitate in isolated primary Bscl2^−/− hepatocytes revealed a reduced fatty acid oxidation rate compared with wild-type controls (Figure 5D). These data suggest that up-regulated de novo lipogenesis together with decreased fatty acid oxidation contributes to the increased hepatic steatosis in Bscl2^−/− hepatocytes in vitro.

Figure 5.

Increased de novo lipogenesis and decreased β-oxidation contribute to hepatic steatosis in vitro. A, TAG content; and B, gene expression in primary hepatocytes isolated from 13-wk-old male Bscl2^+/+ and Bscl2^−/− mice. Data were normalized to cyclophilin A and expressed as relative to wild type. Data were representative of three independent experiments from six total animals in each group. C, TAG synthesis rate measured in isolated primary hepatocytes by tracing hepatic ³H-TAG synthesized during a 2-h interval from ³H-labeled oleic acid. D, Fatty acid oxidation rate measured in isolated primary hepatocytes. After incubation of the cells with the indicated amount of ¹⁴C-palmitate for 120 min, ¹⁴C labeled CO₂ was measured by scintillation counting. Dpm, disintegrations per minute. *, P < .05; **, P < .005 vs Bscl2^+/+ mice.

Metabolic adaptations during fasting in Bscl2^−/− liver

To better understand the metabolic adaptations in Bscl2^−/− liver in the different fasting states, we further compared liver metabolic gene expression profiles in the fed (NF), and after 4hF and 16hF. As expected, under nonfasting condition, genes involved in de novo lipogenesis (Fasn, Scd1, and Acc1) as well as TAG synthesis (Lipin1β, Mgat1) and fatty acid transport (CD36), were all up-regulated (Figure 6A, upper panel), similar to what was observed in isolated Bscl2^−/− hepatocytes (Figure 5B). There was a trend toward down-regulation in Pgc1α and Pparα expression, whereas the mRNA expression of gluconeogenic enzymes (Pepck and G6pase) was reduced, possibly due to hyperinsulinemia-mediated inhibition in the fed state. After a short 4hF, increased expression of de novo lipogenic genes was only observed in Scd1, whereas there was sustained up-regulation of TAG synthesis enzymes and CD36 mRNA. Interestingly, two transcription factors that are known to regulate fasting-induced hepatic glucose and lipid gene expression, Pgc1α and Pparα, were differentially regulated, with the first being up-regulated, whereas the latter down-regulated in Bscl2^−/− liver (Figure 6A, middle panel). Prolonged fasting further minimized the increased mRNA expression of TAG synthesis enzymes and fatty acid transporter, but the down-regulation of Pparα was further potentiated in 16hF Bscl2^−/− liver (Figure 6A, lower panel). Overexpression of PGC1α in liver was known to stimulate very low-density lipoprotein (VLDL) assembly and secretion (26). Indeed, 4hF Bscl2^−/− mice had an increased rate of VLDL secretion compared with controls when injected with detergent pluronic F-127 to inhibit lipoprotein lipase activity (Figure 6B). In line with the decreased gene expression of fasting Pparα, the plasma level of Fgf21, a cognate PPARα–targeted starvation hormone mainly secreted by liver (27), was much higher in the fed state but gradually decreased to undetectable levels after a 16hF in Bscl2^−/− mice, contrary to the normal up-regulation of Fgf21 in response to fasting in Bscl2^+/+ mice (Figure 6C). Furthermore, we found a markedly reduced plasma concentration of β-hydroxybutyrate in Bscl2^−/− mice compared with wild-type in NF, 4hF, and 16hF states, corroborating the impaired mitochondrial fatty acid oxidation in Bscl2^−/− liver (Figure 5D). Interestingly, we found no differences in liver glycogen concentrations (normalized to per mg liver) between two genotypes in the fed and short-fast states (Supplemental Figure 3A). After 16hF, Bscl2^−/− liver retained high glycogen concentrations compared with the almost-depleted Bscl2^+/+ liver (Supplemental Figure 3A). Together with the significantly reduced TAG content in 16hF Bscl2^−/− liver (Figure 4A), our data underscore that, in the absence of the major lipid storage organ (adipose tissue) in Bscl2^−/− mice, the liver undergoes significant metabolic adaptation when the animal is subject to prolonged fasting.

Figure 6.

Fasting induces metabolic adaptation in liver of Bscl2^−/− mice. A, qRT-PCR analysis of hepatic gene expression in fed (NF), 4hF, and 16hF male Bscl2^+/+ and Bscl2^−/− mice (n = 5–6). Data were normalized to cyclophilin A and β-actin based on Genorm algorithm and expressed as relative to Bscl2^+/+ mice. B, VLDL secretion in 10-wk-old male Bscl2^+/+ and Bscl2^−/− mice after a 4hF. C, Plasma Fgf21 levels; and D, Plasma β-hydroxybutyrate levels in NF, 4hF, and 16hF 13-wk-old male Bscl2^+/+ and Bscl2^−/− mice (n = 6–8 in each group).*, P < .05; **, P < .005 vs wild type.

Bscl2 does not play an autonomous role in liver lipid metabolism

Bscl2 is expressed in the liver, albeit at a lower level compared with white adipose tissue (28). To determine whether hepatic Bscl2 expression directly mediates hepatic lipid metabolism, we generated liver-specific Bscl2-knockout mice (Bscl2^Li−/−) by crossing Bscl2^fl/fl mice with Alb-Cre transgenic mice. Bscl2^Li−/− mice displayed a 90% reduction of Bscl2 mRNA in the liver but no change in epididymal white adipose tissue (Figure 7A) compared with control Bscl2^fl/fl mice (designated as Bscl2^Li+/+). These mice were grossly normal with similar body weights, liver, and adipose tissue weights as their Bscl2^Li+/+ counterparts (data not shown). In contrast with the severe hepatic steatosis in global Bscl2^−/− mice, hepatic TAG content was not different between Bscl2^Li+/+ and Bscl2^Li−/− mice on normal-chow diet. When high-fat diet was fed to induce hepatic steatosis, deletion of Bscl2 in the liver did not exacerbate hepatic lipid accumulation in Bscl2^Li−/− mice compared with Bscl2^Li+/+ mice (Figure 7B). The hepatic mRNA levels of major lipid metabolism genes that were altered in global Bscl2^−/− liver also remained unchanged in Bscl2^Li−/− mice (Figure 7C). Furthermore, there was no difference in glucose tolerance and insulin tolerance tests between the two groups of mice when they were fed a regular-chow or a high-fat diet (data not shown). Overall, these data suggest that Bscl2 does not play an autonomous role in hepatic glucose and lipid metabolism and the severe hepatosteatosis observed in global Bscl2^−/− mice is secondary to the effects of lipodystrophy.

Figure 7.

Liver-specific inactivation of Bscl2 does not cause hepatic steatosis. A, Bscl2 mRNA was specifically down-regulated by 90% in the liver of Bscl2^Li−/− mice (n = 6 each). **, P < .005. B, Under both normal-chow diet and high-fat diet, liver TAG contents were not different between Bscl2^Li+/+ and Bscl2^Li−/− mice (n = 6–8 each). C, qRT-PCR reveals no differences in lipid metabolism genes in normal chow diet–fed Bscl2^Li+/+ and Bscl2^Li−/− liver after a 4hF (n = 5–6 each). EWAT, epididymal white adipose tissue.

Discussion

The recent creation of Bscl2^−/− mice provides an interesting lipodystrophic mouse model that recapitulates most features of type 2 BSCL in humans. Although considerable work has been performed on the metabolic derangements in lipodystrophic Bscl2^−/− mice, in this study, we focused in greater detail on how fasting durations modulate hepatic lipid homeostasis and response to insulin in Bscl2^−/− mice. We demonstrate that increased de novo lipogenesis and reduced β-oxidation contribute to hepatic steatosis in Bscl2^−/− mice. 4hF Bscl2^−/− mice have severe hepatic insulin resistance due to reduced expression of insulin receptor and IRSs. Prolonged overnight fasting modulates differential fasting-gene transcription and improves hepatic insulin sensitivity as assessed by hyperinsulinemic-euglycemic clamp and insulin tolerance test in lipodystrophic Bscl2^−/− mice. Diacylglycerol- and ceramide- mediated lipid-signaling pathways were not involved in hepatic insulin signaling in Bscl2^−/− mice. Furthermore, we discovered that mice with liver-specific deletion of Bscl2 presented no hepatic steatosis and other metabolic derangements, suggesting Bscl2 does not play a cell-autonomous role in regulating liver lipid metabolism (Figure 8).

Figure 8.

Summary of major approaches and findings. See text for detailed description. DNL, de novo lipogenesis; β-ox, β-oxidation; IR, insulin receptor; IRS1/2, insulin receptor substrates; clamp, hyperinsulinemic-euglycemic clamp; ITT, insulin tolerance test; DAG, diacylglycerol. All changes were suggested as relative to Bscl2^+/+ liver under same fasting conditions.

Previous studies clearly suggested Bscl2^−/− mice are insulin resistant based on classical ITT analysis (6, 8). Our data confirmed such finding in short-term fasted Bscl2^−/− mice (Figure 2A). But, surprisingly, when we used the insulin clamp, the gold standard for measuring insulin sensitivity, to assess hepatic insulin sensitivity in Bscl2^−/− mice that were fasted for a prolonged 16hF, we clearly observed enhanced hepatic insulin sensitivity (Figure 1). Inconsistencies with respect to fasting duration–induced altered insulin sensitivity have been previously reported in overnight fasted–chow and high-fat-fed C57BL/6J mice (29, 30). However, fasting duration–induced improved insulin sensitivity is not a general feature of lipodystrophic mice becasue overnight fasted A-Zip/F-1 fat-less mice still retain hepatic insulin resistance under hyperinsulinemic-euglycemic clamp (13). It represents another unique metabolic phenotype of Bscl2^−/−mice in addition to the previously reported lower fasting plasma glucose and lack of hypertriglyceridemia (7, 8). So far, the exact physiological relevance of our findings to human BSCL2 lipodystrophy is unclear due to limited clinical studies available. However, our results suggest Bscl2 deficiency incurs a unique fasting-mediated hepatic metabolic adaptation and reinforce the notion that controlling liver insulin sensitivity through reduced total caloric intake could be the key therapeutic option for BSCL2 patients to increase hepatic insulin sensitivity and reduce hepatic steatosis. The observed negative value for hepatic glucose production in Bscl2^−/− mice during the clamp may imply either significant formation of glycogen (Supplemental Figure 3) or significantly enhanced glucose oxidation. Nonetheless, our data also illustrates the importance of correlating the effect of fasting with the duration when evaluating insulin sensitivity in genetic mouse models. This observation also supports the use of standardized procedures during metabolic testing in mice (31, 32).

Mounting evidence suggests that abnormal accumulation of specific fatty acid metabolites such as DAG, ceramides, and fatty acyl-CoA instead of TAG itself or other inert lipids in liver may lead to defects in insulin signaling (33). Intriguingly, our data excluded both DAG and ceramides as culprits underlying the insulin resistance in the steatotic liver of lipodystrophic Bscl2^−/− mice under a normal basal fasting state (4hF); neither do these lipid mediators play a direct role for the improved hepatic insulin sensitivity after an overnight fast (Figure 3, Supplemental Figure 2). Other lipid signaling metabolites such as fatty acyl-CoA were not measured and it is also not known whether there is a difference in the intracellular localization of DAG, another crucial factor in mediating insulin resistance (34). Meanwhile, an overt inflammatory state is not likely to contribute to insulin resistance and fasting-induced switch in hepatic insulin sensitivity in Bscl2^−/− mice, as we did not observe any evident sign of inflammation in Bscl2^−/− liver following 4hF or 16hF (Figure 4A), corroborating the findings in a previous report (6).

Our data provide strong evidence that decreased levels of the insulin receptor itself and IRS proteins, a common feature of most insulin-resistant, hyperinsulinemic states including obesity and type 2 diabetes (35, 36), contribute to the hepatic insulin resistance in 4hF Bscl2^−/− mice. Such down-regulation of insulin receptors and IRS proteins was reported to underlie the insulin resistance in the steatotic liver of Agpat2^−/− mice (11). The prolonged fasting-induced improvement in acute hepatic insulin action is at least partially associated with restoration of these important insulin signaling players, especially IRβ and IRS2, despite that the level of IRS1 is still lower in Bscl2^−/− liver after an overnight fast (Figure 4). The discrepancy between the improved hepatic insulin sensitivity under hyperinsulinemic-euglycemic clamp but similar insulin sensitivity under acute insulin stimulation in the same overnight-fasted Bscl2^−/− liver (Figure 1F and 2D) is reproducible and most intriguing. Although we lack an easy explanation for the apparent discrepancy, we speculate that the sustained hyperinsulinemia under clamp conditions may help engage more receptors and downstream insulin-signaling molecules leading to a positive feedback loop. Additional experiments are needed to further dissect the possible mechanisms involved. It is worth mentioning that we also observed up-regulation of IRβ and IRS1 in wild-type mice after a prolonged fast, which may explain the enhanced hepatic insulin sensitivity in previously reported overnight-fasted C57BL/6J mice (29).

The mechanisms underlying hepatosteatosis in lipodystrophic mice have been well documented in lipodystrophic Agpat2^−/− mice (11). Similar to Agpat2^−/− mice, we observed increased fatty acid influx, elevated de novo lipogenesis, and activation of an alternative TAG synthesis pathway through increased Mgat1 expression in liver of Bscl2^−/− mice. These pathways are possibly heightened in the NF state due to hyperphagia-caused substrate oversupply in Bscl2^−/− mice (Figures 5 and 6). Given that Bscl2^−/− liver (2.21 ± 0.22 g) is almost twice the size of wild-type liver (1.24 ± 0.048 g), the total glycogen stored per liver in fed or short-fast mice was almost twice that in Bscl2^−/− vs wild-type liver, corroborating the finding of increased glycogen storage in patients with BSCL (2). The mechanisms underlying fasting-triggered differential regulation of PGC1α and PPARα, the two critical transcriptional (co)factors known to cooperatively regulate genes involved in fasting hepatic lipid and glucose metabolism, are not known, possibly involving the presence of a much different and complicated hormonal-mediated fasting signaling in lipodystrophic Bscl2^−/− mice. Of note, substrates availability is also the rate-limiting factor in determining glucose and lipid metabolism. The unique lower fasting metabolites (especially free fatty acids) may provide a unifying framework for lower fatty acid oxidation and ketone body production as well as the lower circulating Fgf21, whose synthesis is stimulated by acetoacetate (37) after 16hF.

Like other lipodystrophic mouse models (9, 10), lipodystrophic Bscl2^−/− mice cannot survive more than 36 hours of fasting (data not shown). A 16-hour overnight fast would then be deemed as prolonged fasting in lipodystrophic Bscl2^−/− mice, which may even lead to a state resembling torpor, hibernation, and similar conditions (38). We observed dramatically reduced liver TAG in Bscl2^−/− mice in contrast with increased fat accumulation in wild-type mice after an overnight fast (Figure 3A), suggesting that the liver is able to release its stored TAG for other peripheral tissues in the energy-deficient state. In fact, in direct contrast with impaired (6) or similar (8) VLDL secretion reported by two other groups on Bscl2^−/− mouse models, we observed increased VLDL secretion in our Bscl2^−/− mice. The reason for the discrepancy is unclear but may be related to the different detergents used to block lipoprotein lipase (LPL) activity or different ages of animals used.

In addition, mice with liver-specific deletion of Bscl2 did not phenocopy abnormal lipid deposition in liver even under high fat diet–feeding, which directly argues against a cell-autonomous effect of Bscl2 in modulating liver lipid metabolism and emphasizes that the hepatic steatosis in lipodystrophic Bscl2^−/− mice is likely the result of insufficient capacity to store fatty acids in adipose depots and the presence of selective liver insulin resistance (13, 39); which is further exacerbated by hyperphagia.

In conclusion, the results reported here provide new insights into the metabolic adaptations of the liver in response to fasting in lipodystrophic Bscl2^−/− mice. More importantly, we also uncover a novel fasting-dependent regulation of hepatic insulin signaling in Bscl2^−/− mice that has never been previously reported in other lipodystrophic mouse models. The findings may enhance our understanding of the metabolic dysfunction that underlies BSCL2, the general pathophysiology of insulin action under NF and fasting conditions, as well as the hepatic metabolic adaptations that occur when extrahepatic fat storage is limiting.