Impaired adipose tissue insulin signalling is a critical feature of insulin resistance. Here we identify a pathway linking the lipolytic enzyme hormone-sensitive lipase (HSL) to insulin action via the glucose-responsive transcription factor ChREBP and its target, the fatty acid elongase ELOVL6. Genetic inhibition of HSL in human adipocytes and mouse adipose tissue results in enhanced insulin sensitivity and induction of ELOVL6. ELOVL6 promotes an increase in phospholipid oleic acid, which modifies plasma membrane fluidity and enhances insulin signalling. HSL deficiency–mediated effects are suppressed by gene silencing of ChREBP and ELOVL6. Mechanistically, physical interaction between HSL, independent of lipase activity, and the isoform activated by glucose metabolism ChREBPα impairs ChREBPα translocation into the nucleus and induction of ChREBPβ, the isoform with high transcriptional activity that is strongly associated with whole-body insulin sensitivity. Targeting the HSL–ChREBP interaction may allow therapeutic strategies for the restoration of insulin sensitivity.
Insulin resistance is a pathogenic mechanism involved in a wide array of diseases. The importance of adipose tissue metabolism as a determinant of systemic insulin sensitivity has been shown in transgenic mouse models1,2,3,4 and clinical studies5,6,7,8. In this context, improvement of adipose tissue insulin action seems to be an important target for recovery of whole-body systemic insulin sensitivity. Excessive circulating levels of fatty acids are important contributors to insulin resistance through development of fatty acid–induced lipotoxicity in insulin-sensitive tissues9. Lowering plasma nonesterified fatty acid levels through inhibition of fat cell lipolysis has been proposed as a way to improve insulin sensitivity. However, human data call into question the association between production of fatty acids from adipose tissue lipolysis and insulin resistance in obesity10. Partial deficiency in hormone-sensitive lipase (HSL, encoded by LIPE), one of the neutral lipases expressed in adipocytes, results in greater whole-body insulin sensitivity in obese mice without changes in plasma fatty acid levels11.
Here, in a series of in vitro and in vivo studies in humans and mice, we identified a pathway linking HSL to insulin resistance through interaction with the glucose-responsive transcription factor ChREBP. The physical interaction between HSL and ChREBP impairs nuclear translocation and activity of the transcription factor. In fat cells, the lipogenic enzyme ELOVL6 is a preferential target of ChREBP. Inhibition of HSL promotes activity of ELOVL6 and enhances insulin signalling through enrichment of plasma membrane phospholipids in oleic acid.
Reduced adipocyte HSL expression favors glucose metabolism
In adipocytes differentiated from human multipotent adipose-derived stem (hMADS) cells12,13, HSL gene silencing (Supplementary Fig. 1a–c) increased insulin-stimulated glucose transport (Fig. 1a), glucose oxidation (Fig. 1b) and de novo lipogenesis (DNL) (Fig. 1c). Insulin signalling was enhanced in adipocytes with decreased HSL expression, as shown by enhanced activating phosphorylation of insulin receptor substrate 1 (IRS1), V-Akt murine thymoma viral oncogene homolog (AKT) and AKT substrate of 160 kDa (AS160) after insulin treatment (Fig. 1d–g). As adipose tissue DNL is associated with insulin sensitivity in humans14,15, we tested whether direct inhibition of DNL affects the modulation of insulin signalling induced by HSL depletion. Treatment of human adipocytes with a selective inhibitor of fatty acid synthase (Supplementary Fig. 1d) blunted the induction of insulin-mediated phosphorylation of AKT observed in HSL-deficient fat cells (Supplementary Fig. 1e). Furthermore, HSL inhibition decreased the proportion of palmitic acid and palmitoleic acid but increased that of oleic acid in fat cell triglycerides and phospholipids (Fig. 1h,i). We thereby analyzed gene expression of enzymes catalyzing key steps in the synthesis of the main saturated and monounsaturated fatty acids derived from glucose in human fat cells (Supplementary Fig. 1d). In hMADS adipocytes with decreased HSL expression, the most robust induction was observed for ELOVL6 (Fig. 1j). The increase in the ELOVL6 mRNA level was mirrored by an increase in enzyme activity (Supplementary Fig. 1f) and an increase in the fatty acid elongation ratio attributable to ELOVL6 activity (Supplementary Fig. 1g). ELOVL6 also showed the highest induction among DNL genes in human preadipocytes differentiated in primary cultures (Supplementary Fig. 1h). Next, we evaluated the effect of adipose triglyceride lipase (ATGL, which is encoded by PNPLA2) through PNPLA2 gene silencing in hMADS adipocytes (Supplementary Fig. 1i,j). ATGL precedes HSL in the sequential breakdown of triglycerides during adipocyte lipolysis. ATGL knockdown had no effect on fat cell DNL (Supplementary Fig. 1k,l). Together, the results show that HSL depletion improves insulin signalling and promotes DNL and modification in fatty acid composition. These changes are associated with induction of the fatty acid elongase ELOVL6.
HSL inhibition results in increased adipose Elovl6 in vivo
To probe changes in insulin sensitivity upon decreased HSL expression in vivo, we investigated different mouse transgenic models, genetic backgrounds and diets. Compared to obese wild-type littermates fed with a 60% high-fat diet, B6D2/F1 transgenic mice with Lipe haploinsufficiency11 showed no difference in body weight or fat mass (Supplementary Fig. 2a,b). During euglycemic-hyperinsulinemic clamp, the glucose infusion rate tended to increase (Fig. 2a), there was no change in the rate of glucose disappearance (Fig. 2b) and insulin-mediated suppression of hepatic glucose production increased (Fig. 2c). In Lipe haploinsufficient B6D2/F1 mice fed a 45% high-fat diet, insulin tolerance increased, whereas body weight was not modified (Fig. 2d and Supplementary Fig. 2c). In Lipe haploinsufficient C57BL/6J mice fed a 60% high-fat diet, enhanced insulin sensitivity was confirmed (Fig. 2e) and adipose Elovl6 gene expression level was higher (Fig. 2f). As in human adipocytes (Fig. 1j and Supplementary Fig 1h), the induction was more pronounced for Elovl6 than for other lipogenic genes (Supplementary Fig. 2d). Using zinc finger nuclease–mediated gene editing, we generated a mouse model with HSL knockdown in adipose tissue and unaltered expression in liver (Fig. 2g,h and Supplementary Fig. 2e–g). Mice with zinc finger nuclease–mediated deletion of Lipe exon B (LipeexonB–/–) that were fed a high-fat diet showed increased glucose tolerance (Fig. 2i) without alteration of body weight (Supplementary Fig. 2h). Adipose Elovl6 gene expression was higher in these mice than in wild-type littermates (Fig. 2j). Furthermore, chronic treatment with a specific inhibitor of HSL did not alter body weight (Supplementary Fig. 2i), improved insulin sensitivity (Supplementary Fig. 2j) and increased adipose Elovl6 gene expression (Supplementary Fig. 2k). Therefore, both genetic and pharmacologic inhibition of HSL results in improved insulin sensitivity and enhanced Elovl6 expression in adipose tissue in vivo.
Adipose ELOVL6 is associated with insulin sensitivity
Knockdown of ELOVL6 in human adipocytes led to a significant decrease in ELOVL6 mRNA level and activity (Supplementary Fig. 3a,b). The increases in IRS1 (Fig. 3a) and AKT (Fig. 3b and Supplementary Fig. 3c) phosphorylation observed in HSL-deficient adipocytes were abrogated after concomitant gene silencing of ELOVL6. In agreement with data in human adipocytes, in vivo insulin-stimulated Akt phosphorylation was decreased in adipose tissue of Elovl6-null mice (Fig. 3c and Supplementary Fig. 3d,e). In mice fed a high-fat diet, C57BL/6J mice showed higher insulin tolerance (Fig. 3d and Supplementary Fig. 3f) and higher induction of adipose tissue Elovl6 gene expression during refeeding than DBA/2J mice (Fig. 3e). In humans, adipose ELOVL6 gene expression was lower in visceral adipose tissues of obese subjects with metabolic syndrome than in lean, insulin-sensitive individuals (Fig. 3f,g and Supplementary Fig. 3g). In morbidly obese subjects, weight loss observed 2 years after bariatric surgery (Supplementary Fig. 3h) was associated with greater insulin sensitivity (Fig. 3h) and an increase in subcutaneous adipose ELOVL6 mRNA level (Fig. 3i). A strong positive correlation was found between ELOVL6 mRNA levels and insulin sensitivity (Supplementary Fig. 3i). Taken together, the data identify ELOVL6 as the mediator of the beneficial effects of HSL inhibition and show a positive association between adipose ELOVL6 expression and insulin sensitivity in vivo.
ELOVL6 controls plasma membrane fluidity
In triglycerides (Fig. 4a) and phospholipids (Fig. 4b) of human adipocytes, diminished ELOVL6 expression led to an increase in palmitic acid and palmitoleic acid at the expense of oleic acid. ELOVL6-deficient adipocytes showed a decrease in the proportion of oleic acid (for example, 36:2) and an increase in the proportion of palmitic acid (for example, 32:0) and palmitoleic acid (for example, 32:2) in phosphatidylcholines and phosphatidylethanolamines (Supplementary Fig. 4a–d). In vivo, the lack of Elovl6 in mouse adipose tissue resulted in increased palmitic and palmitoleic acid and decreased oleic acid content (Supplementary Fig. 4e). Considering that ELOVL6 mediates the positive effect of HSL gene silencing on insulin signalling (Fig. 3a,b and Supplementary Fig. 3c), we determined whether this effect was dependent on oleic acid. There are two enzymatic steps between palmitic acid and oleic acid (Supplementary Fig. 1d). The first is the elongation of palmitic acid into stearic acid catalysed by ELOVL6, and the second is the desaturation of stearic acid into oleic acid catalyzed by stearoyl-CoA desaturase (SCD)16,17,18,19. Treatment of human adipocytes with an inhibitor of ELOVL6 (ref. 20) resulted in the changes we expected in fatty acid composition with a decrease of the C18/C16 fatty acid ratio (Fig. 4c). Concordant with data obtained using gene silencing (Fig. 3b), pharmacological inhibition of ELOVL6 abrogated the enhancement of insulin-induced AKT phosphorylation observed in HSL-deficient adipocytes (Fig. 4d). A specific SCD inhibitor21 decreased C16 and C18 fatty acid desaturation (Fig. 4e) and had the same effect as the ELOVL6 inhibitor on AKT phosphorylation (Fig. 4f). The data suggest that SCD is necessary but does not have a rate-limiting role, as ELOVL6 does, in increased insulin signalling induced by HSL inhibition. Accordingly, SCD mRNA levels are much higher than ELOVL6 mRNA levels in human adipocytes (Supplementary Fig. 4f). The content of oleic acid in phospholipids was directly modified by incubation of adipocytes with the fatty acid (Fig. 4g). Exposure of HSL and ELOVL6 double-deficient adipocytes to oleic acid rescued insulin-induced AKT phosphorylation to levels comparable with those observed in fat cells with diminished HSL expression (Fig. 4h). The composition of fatty acid in phospholipids may influence insulin signal transduction through modification of plasma membrane properties16,22,23. Adenovirus-mediated overexpression of ELOVL6 in human adipocytes (Supplementary Fig. 4g,h) increased the C18/C16 fatty acid ratio (Fig. 4i) and insulin-induced IRS1 phosphorylation (Fig. 4j). Comparison of the mobile fractions derived from fluorescence recovery after photobleaching (FRAP) data revealed an increase in plasma membrane lateral mobility of cholera toxin–bound glycolipids in ELOVL6-overexpressing adipocytes (Fig. 4k,l). Collectively, the data suggest that enhanced adipocyte ELOVL6 activity increases the proportion of oleic acid in phospholipids and positively influences insulin signalling through modulation of plasma membrane fluidity.
ChREBP is the link between HSL and ELOVL6
During dual knockdown of HSL and ChREBP, encoded by MLXIPL (Supplementary Fig. 5a), the increase in glucose transport and DNL observed in adipocytes with low HSL expression was diminished (Fig. 5a–c). A similar pattern was observed for glucose and acetate carbon incorporation into fatty acid (Fig. 5b,c), showing that the upregulation of fatty acid synthesis resulted not only from increased glucose uptake but also from specific induction of DNL. MLXIPL gene silencing also mitigated the increase in insulin-induced IRS1 and AKT phosphorylation (Fig. 5d,e). Similar to what we observed for ELOVL6 knockdown (Fig. 4a,b), MLXIPL gene silencing led to an increase in palmitic acid and palmitoleic acid and a decrease of oleic acid (Fig. 5f,g). MLXIPL gene silencing potently suppressed ELOVL6 gene expression but had a weak or no effect on other lipogenic genes, suggesting that ELOVL6 is a preferential target of ChREBP in human fat cells (Fig. 5h). Accordingly, in adipose tissue of Mlxipl-null mice (Supplementary Fig. 5b), Elovl6 was the lipogenic gene whose expression was most severely impaired (Fig. 5i). Two isoforms of ChREBP have been identified: ChREBPα, whose transcriptional activity is regulated by glucose, and ChREBPβ, a transcriptionally superactive and unstable isoform that is a direct transcriptional target of ChREBPα (Supplementary Fig. 5c)24,25. In human adipocytes with knockdown of HSL, the levels of MLXIPL transcripts, notably that encoding the β-isoform of ChREBP, were increased (Fig. 5j). In chromatin immunoprecipitation assays on hMADS adipocytes, more binding events were detected on ELOVL6 functional carbohydrate response element (ChoRE)26 than on positive control regions in RORC and TXNIP, a well characterized target of ChREBP (Supplementary Fig. 5d). ChREBP recruitment onto the ELOVL6 promoter was markedly enhanced in HSL-deficient adipocytes (Fig. 5k). ELOVL6 was strongly associated with gene expression of ChREBPβ isoform transcripts in human adipocyte and human adipose tissue (Fig. 5l,m and Supplementary Fig. 5e,f). Short-term elevation in plasma glucose and insulin levels during a hyperglycemic hyperinsulinemic clamp led to pronounced induction of adipose ChREBPβ isoform transcript (Fig. 5n) and ELOVL6 (Fig. 5o) gene expression, showing the importance of glucose flux into fat cells in the control of ELOVL6 expression in humans. Altogether, our results show that ChREBPβ mediates the effect of HSL deficiency on glucose metabolism and insulin signalling through transcriptional activation of ELOVL6.
HSL modifies ChREBP activity via protein-protein interaction
Considering that HSL catalyzes one of the rate-limiting steps in fat-cell triglyceride hydrolysis, we investigated whether lipolysis per se contributed to the induction of ChREBP. In the culture conditions used to study DNL, the release of glycerol and fatty acid in the culture medium was low and was not influenced by HSL gene silencing (Supplementary Fig. 6a,b). Treatment of adipocytes with triacsin C, which inhibits fatty acid resterification12 did not influence the upregulation of MLXIPL and ELOVL6 mRNA in adipocytes with HSL knockdown (Supplementary Fig. 6c–e). These findings led us to hypothesize that physical interaction between HSL and ChREBP may influence ChREBP activity. Co-immunoprecipitation of HSL and ChREBPα was first shown in HEK293 cells transfected with vectors encoding the two proteins (Supplementary Fig. 6f)27,28. Co-immunoprecipitation was observed using ChREBPα with FLAG epitope tag immobilized on magnetic beads and recombinant HSL (Fig. 6a). Surface plasmon resonance assays supported direct binding between ChREBPα and HSL (Fig. 6b). Interaction between endogenous proteins in adipocytes was shown through immunoprecipitation with antibodies to ChREBP and HSL (Fig. 6c and Supplementary Fig. 6g). In line with the lack of effect of ATGL knockdown on DNL (Supplementary Fig. 1k,l), ATGL showed no interaction with ChREBPα. This further indicates that HSL interaction with ChREBPα is independent of lipolysis and specific to this neutral lipase (Fig. 6c). Furthermore, interaction of HSL with ChREBPα was shown using in situ proximity ligation assays29. Specific and robust fluorescence signals were observed in the cytosol of fat cells from subcutaneous adipose tissue (Supplementary Fig. 6h). Such signals were also seen in differentiated hMADS adipocytes (Supplementary Fig. 6i). Little signal was detected in undifferentiated fibroblasts that do not express HSL. Various negative controls supported the specificity of the interaction (Supplementary Fig. 6j,k). Human HepG2 hepatocytes, which express significant levels of ChREBP but minute amounts of HSL, showed few fluorescent spots (Supplementary Fig. 6l,m). The respective expression of ChREBPα and HSL in mouse liver and adipose tissues is consistent with a fat-specific interaction of the two proteins (Supplementary Fig. 6n). In mouse adipose tissue and human adipocytes, co-immunoprecipitation and proximity ligation assay signal for HSL and ChREBP were diminished in fat cells with reduced HSL expression (Fig. 6d,e and Supplementary Fig. 7a). Human adipocytes with low HSL expression showed higher immunofluorescence of ChREBP in nuclei, indicating that ChREBPα nuclear translocation is facilitated when interaction with HSL is diminished (Fig. 6f). The increased nuclear translocation of ChREBPα was confirmed using subcellular fractionation in human adipocytes and mouse adipose tissues with low HSL expression, whereas no significant differences were observed in the cytosolic fraction (Fig. 6g,h and Supplementary Fig. 7b,c). In mice, there was no difference in ChREBPα protein content in fat pads with high and low HSL contents (Supplementary Fig. 7d). To evaluate the effect of HSL on ChREBPα transcriptional activity, HEK293 cells were transfected with a vector containing the luciferase reporter gene under the control of a promoter containing functional ChoREs30. Promoter activity increased when cells expressed ChREBPα and decreased when cells coexpressed increasing amounts of HSL (Fig. 6i and Supplementary Fig. 7e). When HEK293 cells expressing HSL and ChREBP were treated with a HSL inhibitor, less interaction between HSL and ChREBP was observed (Supplementary Fig. 7f). In human adipose tissue, we have previously identified a short form of HSL produced by in-frame skipping of exon 6 (Supplementary Fig. 2e)31,32. As exon 6 encodes the catalytic site serine, the short form of HSL is devoid of enzymatic activity (Supplementary Fig. 7g). However, it retains the capacity to bind ChREBP (Fig. 6j). An adenovirus expressing the short form of HSL was used to transduce human adipocytes transfected with control or LIPE small interfering RNA (siRNA) (Supplementary Fig. 7h). The induction of ELOVL6 in adipocytes with diminished levels of HSL was blunted when the short form of HSL was expressed (Fig. 6k and Supplementary Fig. 7i). The catalytically inactive form also diminished the increase in IRS1 phosphorylation mediated by HSL downregulation (Fig. 6l). Notably, the changes in fat cell insulin signalling were not associated with variations in the amount of ChREBPα protein (Supplementary Fig. 7j,k). Together, our data suggest that HSL, independently of fat cell triglyceride hydrolysis, represses ChREBP activity via direct interaction with the transcription factor.
Partial inhibition of the fat-cell neutral lipase HSL alleviates insulin resistance without increasing body weight (ref. 11 and present work). The evolution of plasma fatty acid levels and variation in insulin sensitivity were dissociated in this model. Here, we deciphered the mechanisms behind HSL inhibition–mediated improvement of glucose metabolism and identified interactions between prototypical metabolic pathways of the adipocyte (Supplementary Fig. 8).
Adipose DNL is under the control of the glucose-responsive transcription factor ChREBP. In humans, insulin sensitivity is positively associated with adipose DNL and expression of ChREBP, notably its transcriptionally superactive β-isoform14,25,33,34. Knockdown of ChREBP in human adipocytes counteracted effects of HSL gene silencing to increase insulin sensitivity. In this context, the fatty acid elongase ELOVL6 was the main target of ChREBPβ. We observed that ELOVL6 expression in fat is lower in insulin-resistant than in insulin-sensitive subjects, in line with previous reports14,35,36. ELOVL6 catalyzes a critical step in the elongation of C16 fatty acid19,37. In adipocytes, enhanced ELOVL6 activity favored oleic acid synthesis whereas ELOVL6 knockdown had the opposite effect. In vivo and in vitro studies reveal a protective effect of oleic acid on insulin sensitivity and signalling16,23,38,39. As ELOVL6 induced an increase of oleic acid in major classes of phospholipids, we postulated that it may alter plasma membrane fluidity owing to its conformational plasticity22,40,41. The plasma membrane lateral mobility of glycolipids was increased in fat cells overexpressing ELOVL6. We therefore propose that ELOVL6-mediated increase in phospholipid oleic acid content improves fat cell insulin signalling through alteration of plasma membrane properties. In mice, Elovl6 deficiency impaired white adipose tissue insulin signalling, whereas the opposite, or lack of alteration, has previously been reported in the liver, suggesting tissue-specific differences in ELOVL6-mediated modulation of insulin action37,42.
HSL is a multifunctional enzyme with a broad range of substrates43,44. As ChREBP activity is influenced by metabolites and other transcription factors in the liver, products of HSL enzymatic activity could directly or indirectly influence ChREBP-mediated modulation of gene transcription24. Although we cannot rule out the possibility that, in some conditions, upregulation of ChREBPα protein expression partially contributes to the phenotype of adipocytes depleted in HSL, we bring a solid body of evidence showing that physical interaction between HSL and ChREBPα controls the intracellular location and activity of the transcription factor in fat cells. The catalytic activity of HSL is dispensable for the interaction with ChREBP and the ChREBP-mediated effect on ELOVL6 expression and insulin signalling. A specific HSL inhibitor diminished the interaction between HSL and ChREBP and enhanced adipose Elovl6 expression in mice, suggesting that small molecules may be designed to disrupt the interaction. The interaction between HSL and ChREBP, which controls ChREBP nuclear translocation and transcriptional activity, provides a molecular basis for the differences between liver and adipose tissue DNL. The pathway is generally considered as detrimental in the liver, as it is activated during the development of fatty liver disease, whereas it is seen as beneficial in adipose tissue45,46. Because of the low level of HSL expression, the interaction between HSL and ChREBP is not found in human hepatocytes. Alleviation of HSL-mediated inhibition of ChREBP activity may constitute a fat cell–specific mechanism to enhance DNL and insulin signalling.
In conclusion, our work identifies a pathway critical for optimal insulin signalling in fat cells that links the neutral lipase HSL to the glucose-responsive transcription factor ChREBP and its target gene, the fatty acid elongase ELOVL6. Inhibition of the HSL-ChREBP interaction may constitute an adipose-specific strategy to reduce insulin resistance.
General experimental approaches
No samples, mice, human research participants or data points were excluded from the reported analysis. Randomization was not performed except when noted below. Analyses were not blinded except when noted below. Detailed information and common techniques are described in Supplementary Methods as indicated below.
Culture of human adipocytes and in vitro measurements
Culture of adipocytes
hMADS cells were expanded in DMEM with 5.5 mM glucose (Lonza) supplemented with 10% FBS (Lonza), 2 mM l-glutamine (Invitrogen), 10 mM HEPES buffer (Lonza), 50 units ml–1 penicillin (Invitrogen) and 50 mg ml–1 streptomycin (Invitrogen), supplemented with 2.5 ng ml–1 fibroblast growth factor 2 (Sigma). At confluence, fibroblast growth factor 2 was removed from proliferation medium. On the next day (day 0), the cells were incubated in differentiation medium (DMEM and Ham’s F-12 medium containing 7.8 mM glucose, HEPES, l-glutamine, penicillin-streptomycin, 10 µg ml–1 transferrin (Sigma), 10 nM insulin (Sigma), 0.2 nM triiodothyronine (Sigma), 100 µM 3-isobutyl-1-methylxanthine (Sigma), 1 µM dexamethasone (Sigma) and 100 nM rosiglitazone (Sigma)). At days 3 and 10, respectively, dexamethasone and 3-isobutyl-1-methylxanthine, and then rosiglitazone, were removed from culture medium. The experiments were carried out between days 12 and 15.
For primary culture and differentiation of human preadipocytes, subcutaneous adipose tissue samples were obtained from five women (age 39 ± 9 years; body mass index (BMI) 28 ± 4 kg m–2) undergoing elective plastic surgery in the abdominal or dorsal region at Rangueil Hospital, Toulouse, France. Adipose tissue was cleaned from blood vessels and fibrous material, minced into pieces and digested in 1 volume of collagenase I (300 units ml–1, Sigma) for 90 min in 37 °C shaking water bath. Digested tissue was filtered through a 250-μm strainer, diluted with PBS-gentamycin and centrifuged at 1,300 r.p.m. for 5 min. The pellet was incubated in erythrocyte lysis buffer for 10 min at room temperature. Cells were filtered, centrifuged and suspended in PM4 medium with 132 nmol l–1 insulin for differentiation and collected at day 13 (refs. 47,48). The study was approved by the Ethics Committee of Toulouse University Hospitals (Comité de Protection des Personnes Sud Ouest et Outre Mer 2, DC-2014-2039). The volunteers signed informed consent for anonymous use of samples.
HEK293 and HepG2 cell cultures
RNA interference was achieved by siRNA. Briefly, on day 7 and day 4 of differentiation, hMADS and primary preadipocytes, respectively, were detached from culture dishes with trypsin-EDTA (Invitrogen) and counted. Control siRNA against green fluorescent protein (GFP) (siCTR) and gene-specific siRNA for LIPE, MLXIPL, ELOVL6 and PNPLA2 (Eurogentec) were delivered into adipocytes using a microporator (Invitrogen) with the following parameters: 1,100 V, 20 ms and 1 pulse. The targeted sequences are provided in Supplementary Methods.
Under the control of a cytomegalovirus promoter, adenoviruses encoding ELOVL6 in tandem with GFP (ADV-207862), the short form of HSL in tandem with GFP or encoding GFP alone (catalog no. 1060) were obtained from Vector Biolabs. Adenoviral particles (multiplicity of infection, 200) were added in the culture medium for 24 h at day 11–12 of hMADS cell differentiation. Medium was changed and experiments were carried out 48 h later.
Oleic acid supplementation in human adipocytes
Treatments with enzyme inhibitors
For fatty acid synthase, SCD and ELOVL6 inhibition, hMADS adipocytes were treated, with 1 µM of compound AZ12756122 (example 117 from WO2008075070A1, synthesized at AstraZeneca), 75 nM of compound A939572 (ref. 21) (Tocris Biosciences) and 1 µM compound 1w20 (provided by AstraZeneca), respectively, in culture medium for 48 h. To study the effect of bioactive fatty acid on the induction of ChREBP, cells were treated for 8 h with 10 µM of triacsin C (Sigma), an inhibitor of acyl-CoA synthase, in the culture medium.
Gene expression analysis
Characterization of human ChREBPβ-specific exon
Western blot analysis
Triacylglycerol hydrolase activity was measured on cell extracts12. For other metabolic measurements, insulin was removed from culture medium the day before the assay. To determine glucose uptake, cells were incubated for 50 min at 37 °C with or without 100 nM insulin. Then, 125 µM of cold 2-deoxy-d-glucose and 0.4 µCi 2-deoxy-d-[3H]glucose (PerkinElmer) per well were added for 10 min incubation. Culture plates were put on ice and rinsed with 10 mM glucose in ice-cold PBS and then with ice-cold PBS. Cells were scraped in 0.05 N NaOH, and radioactive 2-deoxy-d-glucose uptake was measured by liquid scintillation counting of cell lysate. To determine glucose oxidation, cells were incubated for 3 h in Krebs Ringer buffer supplemented with 2% BSA, 10 mM HEPES, 2 mM glucose and 1 µCi d-[14C(U)]glucose (PerkinElmer) with or without 100 nM insulin. A 2- × 2-cm Whatman 3M paper was placed on top of each well and soaked with 120 µl 1 N NaOH. After incubation, filter-trapped 14CO2 was measured by liquid scintillation counting. Medium was acidified with 1 M sulfuric acid and 14CO2 in the culture medium was trapped by benzethonium hydroxide during a 2-h incubation. Benzethonium-trapped 14CO2 was measured by liquid scintillation counting. Specific activity was counted and used to determine the quantity of oxidized glucose equivalent. To assess glucose incorporation into fatty acids, cells were then washed twice in PBS and then scraped in STED (0.25 mM sucrose, 10 mM Tris, 1 mM EDTA, 1 mM dithiothreitol, pH 7.4). Neutral lipids were extracted in methanol-chloroform (1:2). The organic phase was dried under nitrogen and hydrolyzed in 1 ml 0.25 N NaOH in methanol-chloroform (1:1) for 1 h at 37 °C. The solution was neutralized with 500 µl 0.5 N HCl in methanol. Fatty acids and glycerol were separated by adding 1.7 ml chloroform, 860 µl water and 1 ml methanol-chloroform (1:2). Incorporation of 14C into fatty acids was measured by liquid scintillation counting of the lower phase. Specific activity was counted and used to determine the quantity of incorporated glucose equivalent. DNL was also measured using acetic acid-sodium salt-[1-14C] (PerkinElmer). Cells were incubated for 3 h in Krebs buffer supplemented with 10 mM HEPES, 2 mM glucose, 2% BSA and 2 µCi ml–1 radiolabelled acetate stimulated with or without 100 nM insulin. Cells were then washed twice and harvested in PBS with 0.1% SDS. Neutral lipids were extracted in methanol–chloroform (1:2). Incorporation of 14C into neutral lipids was measured by liquid scintillation counting of lower phase. Results from metabolic measurements were normalized to total protein content of cell extracts.
Fatty acid elongation activity was measured in crude microsomal extracts from hMADS adipocytes37. Briefly, cells were washed with PBS, scraped in 3 ml of ice-cold 250 mM sucrose, 20 mM HEPES, 1 mM EDTA, pH 7.5 and Dounce homogenized. Homogenate was centrifuged at 1,000g at 4 °C for 7 min. Supernatant was collected and centrifuged at 2,000g at 4 °C for 30 min. Supernatant was collected and centrifuged at 17,000g at 4 °C for 1 h. The resulting pellet was suspended in 50 µl of 100 mM Tris-HCl, pH 7.4, and used for fatty acid elongation activity after determination of protein concentration. ELOVL6 activity was assayed by the measurement of [2-14C]malonyl-CoA (PerkinElmer) incorporation into exogenous palmitoyl-CoA49. ELOVL6 inhibitor (1 µM of compound 1w; ref. 20) was preincubated for 30 min at 37 °C with microsomal protein before addition of reaction mixture. Incubation was stopped by adding 0.2 ml of 5 M KOH, 10% methanol, and saponified at 65 °C for 1 h. Then the samples were cooled and acidified with 0.2 ml of ice-cold 5 N HCl and 0.2 ml of ethanol. Free fatty acids were extracted from the mixture three times with 1 ml of hexane, 2% acetic acid. The pooled hexane fractions were dried under nitrogen, and after addition of 3 ml of scintillation cocktail, the radioactivity incorporated was counted. Blanks were carried out in parallel reactions incubated without microsomal fractions. ELOVL6 activity was obtained by subtracting [14C]malonyl-CoA molecules incorporated into fatty acids in the absence of inhibitor from the values in the presence of the ELOVL6 inhibitor.
Chromatin immunoprecipitation assays
Human adipocyte cells (107 cells per condition) were fixed with 1% formaldehyde for 15 min and quenched with 0.125 M glycine. Chromatin was isolated by the addition of lysis buffer, followed by disruption with a Dounce homogenizer. Lysates were sonicated and the DNA sheared to an average length of 300–500 base pairs. Genomic DNA (Input) was prepared by treating aliquots of chromatin with RNase, proteinase K and heat for decrosslinking, followed by ethanol precipitation (Active Motif). Pellets were resuspended and the resulting DNA was quantified on a NanoDrop spectrophotometer. Extrapolation to the original chromatin volume allowed quantitation of the total chromatin yield. Aliquots of chromatin (30 µg) were precleared with protein A agarose beads (Invitrogen). Genomic DNA regions of interest were isolated using an antibody to ChREBP (Novus, NB400-135). Positive and negative controls were designed by Active Motif. Complexes were washed, eluted from the beads with SDS buffer and subjected to RNase and proteinase K treatment. Crosslinks were reversed by incubation overnight at 65 °C, and ChIP DNA was purified by phenol-chloroform extraction and ethanol precipitation. Quantitative PCR (qPCR) reactions were carried out in triplicate using SYBR Green Supermix (Bio-Rad, 170-8882) on a CFX Connect real-time PCR system. Positive and negative control sites were tested for each factor, and so were the sites of interest. The resulting signals were normalized for primer efficiency by carrying out qPCR for each primer pair using input DNA (pooled unprecipitated genomic DNA from each sample).
Nuclear and cytosolic fractions from hMADS adipocytes were prepared using Nuclear Extract Kit (40010) from Active Motif. Cells were rinsed with PBS and immediately scrapped into 1× hypotonic buffer. For adipose tissue, tissues were ground in liquid nitrogen and lysed using the NE-PER nuclear and cytoplasmic extraction reagent kit (Thermoscientific, 78835). Subsequent steps followed the manufacturer’s protocol. Antibodies to histone H3 (4499, Cell Signaling Technology), lamin A/C (4777, Cell Signaling Technology) and α tubulin (T5168, Sigma) were used to analyze the efficiency of cellular fractionation.
Fatty acid composition of triglycerides and phospholipid
Cells were scraped in PBS and then mixed with methanol supplemented with 0.001% butylated hydroxytoluene. Lipid extraction was performed with a chloroform–methanol mixture (1:1) and KCl (0.5 M) after centrifugation (2,500 r.p.m., 10 min). Phospholipids and triglycerides were isolated by thin-layer chromatography on silica glass plates (Merck) using petroleum ether–diethyl ether–acetic acid (80:20:1) as the mobile phase. Fatty acid methyl esters were generated by transmethylation of the glycerolipids in methanol with 5% acetyl chloride at 60 °C for 1 h, extracted twice by isooctane. Analysis was carried out with a gas chromatograph (Shimadzu GC 2100) equipped with a CP-Wax 58 capillary column 50 m long with a 0.25-mm external diameter and 0.2-μm thickness of the stationary phase (Varian), with helium (1 ml min–1) as the carrier gas. Programmed temperature vaporization (PTV system) injector and flame ionization detector were used. Results are expressed in percentage of total fatty acid contained in the sample.
Fatty acid composition in phospholipid classes
Measurement of glycerol and NEFA in culture medium
Fluorescence recovery after photobleaching
Cells were labelled for 15 min with 1 µg ml–1 Alexa 555-labelled Cholera ToxinSubunit B (CTxB Molecular Probes) at 4 °C, then washed three times in chilled medium supplemented with 25 mM HEPES buffer, pH 7.4. A LSM780 confocal microscope, equipped with highly sensitive 32-channel GaAsP detectors, operated with Zen Blue software, coupled to a DPSS laser (561 nm, maximum power 20 mW) was used for excitation with a detection bandwidth of 571–624 nm (Carl Zeiss). All experiments were done at room temperature (22 °C). Cells were observed using a Plan-Apochromat 63× NA, 1.4 oil immersion objective, and the pixel dwell was set to the optimal value of 1.92 µs. The fluorescence intensity of three regions of interest of 6.4 μm × 3.2 μm was measured: the photobleached area, a region within the cell that was not photobleached to check for overall photobleaching and cell position fluctuation, and the background. After 10 prebleach scans (one scan every 200 ms) at 1% maximal laser power to determine initial fluorescence intensity, one photobleaching scan was performed at 100% laser power. Postbleach fluorescence recovery was then sampled at 1% laser power for 150 s. FRAP data analysis was done as described by Bonneau et al50.
HEK293T cells were harvested in a lysis buffer containing 3% 5 M NaCl, 5% 1 M Tris-HCl (pH 7.5), 1% 500 mM EDTA, 1.3% sodium pyrophosphate and 0.02% NaF, supplemented with 1% Triton X-100 (Sigma), 2% 50× protease inhibitor cocktail (Roche) and 1% 1 mM orthovanadate (Sigma). Protein (1 mg) was immunoprecipitated overnight at 4 °C, with 40 µl of anti-FLAG M2 magnetic beads (Sigma). Beads were gently centrifuged for 1 min and washed with the lysis buffer before elution in Laemmli buffer.
For immunoprecipitation of purified proteins, FLAG-tagged ChREBP was expressed in HEK293T cells. Cells were harvested in lysis buffer described above. Protein (300 μg) was immunoprecipitated overnight at 4 °C with 40 µl of anti-FLAG M2 magnetic beads. Beads bound with ChREBP were washed with the lysis buffer and incubated with 1 μg of human recombinant HSL (Cayman) in 350 μl of lysis buffer for 3 h at 4 °C with gentle rocking. The beads were washed three times with lysis buffer.
For endogenous co-immunoprecipitation in hMADS adipocytes, cells were lysed for 15 min in 1× hypotonic buffer (Active Motif) with 4% 25× protease inhibitor cocktail (Roche) and 1 mM orthovanadate (Sigma). Cell debris and fat were discarded after 12,700 r.p.m. centrifugation at 4 °C for 15 min. Preclearing was performed at 4 °C for 30 min using 50 µl protein G and 4 µg control rabbit (2729, Cell Signaling Technology) or mouse (sc-2025, Santa Cruz) IgG. Beads were discarded and supernatants were incubated with 4 µg anti-ChREBP (NB400-135, Novus) or 2 µg anti-HSL (sc-74489, Santa Cruz Biotechnology) for 90 min at 4 °C. As a negative control of immunoprecipitation, 4 µg control rabbit (2729, Cell Signaling Technology) or 2 µg mouse (sc-2025, Santa Cruz) IgG was used. Protein A–protein G (50:50) magnetic beads were added for 1 h at 4 °C. Beads were washed in cold PBS with 4% 25× protease inhibitor cocktail (Roche) and 1 mM orthovanadate.
For ChREBP immunoprecipitation in mouse white adipose tissue, fat was cut in small pieces and lysed during 2 h in 20 mM Tris-HCl, 150 mM NaCl, 0.5% NP-40 and protease and phosphatase inhibitors; pH 8. After centrifugation at 15,000g for 20 min at 4 °C, the fat layer was removed before collecting the supernatant. For each immunoprecipitation, 0.8–1 mg protein was precleared with 50 µl of Protein A Dynabeads (ThermoFisher) for 1 h at 4 °C, then incubated overnight at 4 °C with 40 µl Protein A dynabeads coupled with 5 µg rabbit IgG or ChREBP antibody (Novus). Beads were washed four times with lysis buffer before elution in 2× Laemmli buffer.
In situ proximity ligation assay and immunofluorescence
In situ proximity ligation assay was performed using Duolink In Situ reagents (Sigma). Cells and pieces of subcutaneous adipose tissue were fixed with 4% paraformaldehyde (Sigma) for 15 min and permeabilized for 10 min at room temperature with 0.2% Triton X-100 (Sigma). Incubation of antibodies, ligation of oligodeoxynucleotides and amplification were performed according to the manufacturer’s instructions. The following primary antibodies were incubated overnight at 4 °C: anti-HSL (murine antibody, sc-74489, Santa Cruz Biotechnology), anti-ATGL (mouse antibody, NBP2-59390, Novus), anti-AKT (mouse antibody, 2920, Cell Signaling Technology) and anti-ChREBP (rabbit antibody, NB400-135, Novus). The same antibodies were used in immunofluorescence assays. Anti-mouse (Alexa-fluor 488-conjugated, A21202, and Alexa-fluor 546-conjugated, A10036, Invitrogen) and anti-rabbit (Alexa-fluor 546-conjugated, Invitrogen) secondary antibodies at 1/300 dilution were incubated for 45 min. Neutral lipids were stained using Bodipy (4-3922, Life Technologies) for 30 min. Nucleus labelling was done using Hoescht (33342, 5 mg ml–1, Invitrogen) for 5 min. Confocal microscopy was performed using Zeiss LSM780. Image processing was similar for all conditions. The same settings were applied to entire images.
Surface plasmon resonance assays
All binding studies based on surface plasmon resonance technology were performed on a BIAcore T200 optical biosensor instrument (GE Healthcare). The Fc region of anti-ChREBP (NB400-135, Novus) was immobilized to the chip surface using native Protein A sensorchip in PBS-P+ buffer (20 mM phosphate buffer, pH 7.4, 2.7 mM KCl, 137 mM NaCl, and 0.05% surfactant P20) (GE Healthcare). Immobilization steps were performed at a flow rate of 5 µl min–1 with a final concentration of 2 µg ml–1. Total amount of immobilized antibody was 11,000–12,000 response units (RU). Then all injection steps were performed at a flow rate of 20 µl min–1. Channel Fc1 was used as a reference surface for nonspecific binding measurements.
No randomization or blinding was performed. Animals from several litters were used in each protocol to avoid litter-to-litter variation.
Targeted disruption of the Lipe gene and generation of Lipe+/– mice have been described elsewhere11. Before being killed, mice were fasted for 24 h or re-fed for 18 h, supplemented with 20% glucose in drinking water. The promoter upstream of exon B governs HSL expression in fat cells51. To create transgenic mice with specific deletion of Lipe exon B, mRNA coding for zinc finger nucleases specifically targeting HSL (CompoZr Custom Zinc Finger Nucleases, CSTZFN-1KT, Sigma) was injected into pronuclei of one-cell embryos from female B6D2/F1 mice. Homozygous mice (LipeexonB–/– registered as B6D2-Lipeem1Land mice) were obtained. A full description of the model will be published elsewhere. The specific inhibitor of HSL (BAY 59-9435) was synthesized by NoValix52. Transgenic mice were fed a high-fat diet (60% or 45% kcal fat, respectively, D12492 and D12451 from Research Diets) for indicated times. In pharmacological studies, C57BL/6J male mice (12–15 weeks old, Janvier Laboratories) were treated orally with dimethylsulfoxide (DMSO) or HSL inhibitor (70 mg kg–1 once daily) for 11 d. Eight-week-old DBA2/J and C57BL6/J male mice (Charles River) were fed a high-fat diet (60% kcal fat, D12492 from Research Diets) for 6 weeks before being killed. Mice were housed and manipulated according to Inserm guidelines and European Directive 2010/63/UE in the local animal care facility (agreements A 31 555 04 and C 31 555 07). Protocols were approved by the French Ministry of Research after review by local ethical committee (CEEA122).
In studies on ChREBP-null mice, 10- to 12-week-old male and female Mlxipl global knockout mice53 and wild-type littermates were maintained in a 12-h light, 12-h dark cycle with water and chow diet (65% carbohydrate, 11% fat and 24% protein). For the fasting–refeeding experiment, mice were either fasted for 24 h (fasted group) or re-fed for 18 h on chow diet with access to drinking water with 20% glucose after a 24-h fast (re-fed group). Mice were housed and manipulated according to Inserm guidelines and European Directive 2010/63/UE in the local animal care facility (agreement A751320). Protocols were approved by the French Ministry of Research after review by local ethical committee (CEEA34).
Mice homozygous for a deletion in Elovl6 and their wild-type littermates were phenotyped on a C56BL6/J background54. The research has been regulated under the Animals (Scientific Procedures) Act 1986 Amendment Regulations 2012 after ethical review by the University of Cambridge Animal Welfare and Ethical Review Body.
Gene and protein expression analyses
Measurement of fasting glucose and insulin
Glucose and insulin tolerance tests and insulin bolus injection
The nature of the groups was blinded to the investigator performing gene expression experiments.
Women with differing obese and metabolic status
Participating women (lean group mean age 37 ± 16 years; obese with metabolic syndrome group mean age, 49 ± 11 years) were scheduled to have abdominal surgery (laparoscopic or laparotomic cholecystectomy and gastric banding)55. During the surgical procedure, samples of visceral adipose tissue were obtained by surgical excision. Euglycemic-hyperinsulinemic clamp was performed at rest after an overnight fast. Each subject gave written informed consent and the study was approved by the Ethics Committee of the Third Faculty of Medicine, Charles University, Prague.
The eight participating men were 23 ± 3 years old (body mass index (BMI) 23 ± 2 kg m–2). The hyperglycemic hyperinsulinemic clamp was a modification of the hyperglycemic method used by Del Prato et al.56 combined with the original hyperinsulinemic clamp described by Defronzo et al.57 For hyperglycemia, the objective was to increase plasma glucose 5.5 mmol l–1 above fasting level by infusing 20% dextrose in two phases: (1) bolus dose to increase glycemia to the desired target and (2) continuous infusion dose adjusted every 5–10 min according to measured plasma glucose to maintain glycemia at the desired target. To obtain hyperinsulinemia, insulin was co-infused at a rate of 75 mU m–2 min–1 for 180 min. The study was approved by the Ethics Committee of University of Montreal. The volunteers gave their written consent after being informed of the nature, purpose and possible risks of the study.
Morbidly obese subjects undergoing bariatric surgery
This cohort has in part been described before58. In brief, 14 obese women (BMI > 35 kg m–2; age 48 ± 9 years) referred to the hospital for gastric bypass surgery (Roux-en-Y) were investigated before surgery and for 2 years post-operatively. According to self-report, body weight had been stable (±2 kg) for at least 3 months before both investigations. The study was approved by the regional ethics board in Stockholm and registered at http://clinicaltrials.gov as NCT01785134. Subjects were randomized to omentectomy or not and this was blinded to investigators and subjects. The procedure was explained in detail to each woman and written informed consent was obtained.
Gene expression analysis
Results from biological replicates were expressed as mean ± s.e.m. Statistical analyses were performed using GraphPad Prism (GraphPad Software version 5.0). D’Agostino and Pearson omnibus normality test was used to test normality. Fischer test was used to test for equality of variances. Data were log transformed when appropriate to reach normality and uniform distribution. Statistical tests were two-sided. Paired or unpaired Student’s t-tests, Wilcoxon test and Mann–Whitney test were performed to compare two conditions. Paired or unpaired one-way analysis of variance (ANOVA) and Friedman’s tests were performed and followed by Bonferroni’s and Dunn’s post hoc tests, respectively, to determine differences between several groups. Paired or unpaired two-way ANOVA with Bonferroni’s post hoc tests were used to compare two variables. Linear regression was used to test association between two variables.
Further information on research design is available in the Nature Research Reporting Summary linked to this article.
The data that support the plots within this paper and other findings of this study are available from the corresponding author upon reasonable request.
Publisher’s note: Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
The authors acknowledge N. Venteclef (Centre de Recherche des Cordeliers, Paris) and J. Boucher (AstraZeneca, Göteborg, Sweden) for critical reading and comments on the manuscript. E. Courty and J. Personnaz participated in mouse studies during internship at I2MC. The GenoToul Animal Care, Anexplo, Imaging-TRI (especially F. Gaits-Iacovoni for helpful discussion) and Quantitative Transcriptomics facilities contributed to the work. This work was supported by Inserm, Paul Sabatier University, Fondation pour la Recherche Médicale (DEQ20170336720 to D.L.), Agence Nationale de la Recherche (ANR-12-BSV1-0025Obelip and ANR-17-CE14-0015Hepadialogue to D.L.), Région Midi-Pyrénées (OBELIP and ILIP projects to D.L.), FORCE/F-CRIN for clinical research on obesity, EU/EFPIA Innovative Medicines Initiative Joint Undertaking (EMIF grant 115372 to P.A., A.V.P. and D.L.) and AstraZeneca France (TALIP project to D.L.). D.L. is a member of Institut Universitaire de France.