Adipose tissue mTORC2 regulates ChREBP-driven de novo lipogenesis and hepatic glucose metabolism

To evaluate the role of mTORC2 specifically in adipose tissue, we generatedmice (hereinmice). We previously confirmed that adiponectin-Cre targets mature adipocytes with high specificity and efficiency in our colony. Note that significant differences exist betweenand mice in whichwas targeted with. We provide a detailed comparison in the Discussion and in. Adiponectin-cre;Rictor Rictor Rictor Rictor aP2-cre fl/fl Adipoq-cre Adipoq-cre ^{23 16 17} Supplementary Table 1

Deletingwith adiponectin-Cre greatly reduces RICTOR, pAKTand pAKT(a growth factor insensitive mTORC2 target site) in the major visceral (that is, perigonadal or pgWAT), subcutaneous (that is, inguinal subcutanteous or sWAT), and brown fat (that is, interscapular BAT) depots (and) and the residual signal is from stromal vascular fraction (SVF) cells because RICTOR and pAKTis undetectable in purified adipocytes (). In contrast, pAKTis intact (), which maintains AKT's ability to phosphorylate pFOXO1, pGSK3βand pPRAS40(). SVF and hepatic RICTOR levels are normal (and) confirming targeting specificity. Rictor S473 T450 S473/pT450 T308 T24 S9 T246 Fig. 1a Supplementary Fig. 1a Fig. 1b Fig. 1a,b Fig. 1a Fig. 1b Supplementary Fig. 1b

The body weight () and food intake () ofmice consuming a standard chow diet does not significantly differ from controls through 20 weeks. The mass of the major fat depots is also unaffected in both male and female mice (and) and theadipocytes appear normal by haematoxylin and eosin staining (and). Liver mass increases inmice by 18% and 13% in both males and females, respectively (and). By haematoxylin and eosin stain the liver appears normal (); however, despite no observable difference in Oil Red O staining in the liver between the control andKO mice () there is a measurable increase in total hepatic TAG content () corresponding with a 16% decrease in the number of nuclei per field (Supplementary Fig. 1f) (indicative of increased cell size) that may in part explain the overall increase in liver mass. Heart mass also increases inKO mice; however, kidney, spleen, lung, thymus and muscle mass is normal (and). Total pancreas and pancreatic β-cell mass also does not significantly differ between KO and control (and). These observations are in contrast to those reported formice, which have increased total body size because of a global increase in lean tissue mass (including heart, kidney, spleen and pancreas but not the liver) that is attributed to high circulating IGF1 (ref.). We find no difference in circulating IGF1 levels betweenmice and controls () and conclude that in 20-week-old mice living under standard conditions, losing adipose tissuedoes not affect overall fat mass or increase whole-body lean tissue growth in this model. Fig. 1c Fig. 1d Fig. 1e Supplementary Fig. 1c Fig. 1f Supplementary Fig. 1d Fig. 1g Supplementary Fig. 1e Fig. 1h Fig. 1h Fig. 1i Fig. 1g Supplementary Fig. 1e Fig. 1j Supplementary Fig. 1g,h ¹⁶ Supplementary Fig. 1i Rictor Rictor Rictor Rictor Rictor Rictor Rictor Rictor Adipoq-cre Adipoq-cre Adipoq-cre Adipoq-cre Adipoq-cre aP2-cre Adipoq-cre

Blood glucose concentrations in fasting and fedmice do not differ from controls; however, this requires approximately threefold higher plasma insulin suggesting insulin resistance (). Insulin intolerance ofmice was supported by insulin tolerance tests (ITTs), showing an 89% increase in AUC (and); glucose tolerance is normal (). Acutely inhibiting mTORC2 by treatingmice with tamoxifen also ablates RICTOR and pAKTwithin 3-week of treatment (), and this induces insulin intolerance similar to the congenital KOs without affecting glucose tolerance (Supplementary Fig. 2c–e) indicating insulin resistance occurs rapidly withloss. Rictor Rictor Rictor Rictor Adipoq-cre Adipoq-cre Adipoq-creERT2 S473 Fig. 2a,b Fig. 2c Supplementary Fig. 2a Fig. 2d Supplementary Fig. 2b

To better assess insulin action we performed hyperinsulinemic-euglycemic clamps in conscious mice. The steady-state glucose infusion rate required bymice to maintain euglycemia is 46% lower than controls indicating severe insulin resistance (). Insulin stimulated whole-body glucose uptake and glycolysis is diminished by 31 and 32%, respectively (), and there is no difference in glycogen plus lipid synthesis (). Interestingly, insulin stimulated glucose uptake into skeletal muscle—the main site of glucose clearance—is normal (), while in contrast, insulin stimulated glucose uptake into adipose tissue decreases by 69% inmice (). Basal HGP is unaltered between both cohorts; however, insulin fails to suppress HGP inmice (). Themice also express 2.2-fold more hepatic() () and they show less tolerance to a pyruvate bolus () consistent with increased hepatic gluconeogenesis. Hepatic() is not significantly elevated (). Rictor Rictor Rictor Rictor glucose 6-phosphatase G6P phosphoenolpyruvate carboxylase Pepck Adipoq-cre Adipoq-cre Adipoq-cre Adipoq-cre Fig. 2e Fig. 2f,g Fig. 2h Fig. 2i Fig. 2j Fig. 2k Fig. 2l Supplementary Fig. 2f Fig. 2l

We also examined genes involved in hepatic lipid regulation. Althoughlivers have elevated TAG content, most lipogenesis genes including hepatic(,and),, andexpress normally in theKO livers (). We do detect increasedexpression; however, expression of the ChREBP targetsanddoes not significantly differ () suggesting ChREBPβ may have a role in gluconeogenesis. We also find increased expression of the TAG synthesis genes(), whose product converts lysophosphatidic acid to phosphatidic acid in the second step ofphospholipid synthesis, and(), whose product catalyzes the synthesis of diacylglycerols, inKO livers (); other TAG synthesis genes are normal. The lipid uptake genealso expresses normally inKO livers; however,increases (). Among genes that encode β-oxidation regulators, hepatic(also increases, but(and(express normally (). Collectively, these data suggest that adipose tissue mTORC2 governs production of an adipocyte-derived signal that regulates HGP and lipid handling. Rictor Srebps Srebf1a Srebf1c Srebf2 Chrebpα Lxrα Rictor Chrebpβ Acly, Acc Fasn 1-acylglycerol-3-phosphate O-acyltransferase 2 Agpat2 de novo monoacylglycerol O-acyltransferase 1 Mgat1 Rictor lipoprotein lipase (Lpl) Rictor CD36 carnitine palmitoyltransferase 1 Cpt1) peroxisome proliferator-activated receptor α Pparα) medium-chain acyl-CoA dehydrogenase Mcad) Adipoq-cre Adipoq-cre Adipoq-cre Adipoq-cre Supplementary Fig. 2g Supplementary Fig. 2h Supplementary Fig. 2i Supplementary Fig. 2j Supplementary Fig. 2k

As expected, insulin fails to stimulate AKTphosphorylation inpgWAT while insulin-stimulated pAKTis not significantly different (). Interestingly, activating phosphorylation of the insulin receptor (pIR) is higher in the KO fat () possibly suggesting loss of an inhibitory feedback mechanism. Moreover, downstream insulin-stimulated AKT substrate phosphorylation to AS160, FOXO1, GSK3β and PRAS40 does not significantly differ from controls () indicating normal insulin-stimulated pan-AKT signalling. Overall, the sWAT gave similar results but with a few noteworthy differences (). In the mutant sWAT pAKTis more resistant to changes in insulin levels (—top) resulting in increased and decreased phospho-AKTsignal relative to controls in the fasted and insulin-stimulated state respectively (—bottom). Phospho-PRAS40 is also reduced only in the mutant sWAT; however, AS160, FOXO1 and GSK3β phosphorylation are normal (). We confirmed these findings in isolated mature adipocytes (). The fact that insulin-stimulated IR phosphorylation and AKT signalling is largely intact in the mutant WATs yet insulin fails to efficiently stimulate glucose uptake () indicates thatmice have selective adipose tissue insulin resistance. S473 Adipoq-cre T308 Y1150/1151 T308 T308 Adipoq-cre Rictor Rictor Fig. 3a Fig. 3a Fig. 3a Fig. 3b Fig. 3b Fig. 3b Fig. 3c Fig. 3c Fig. 2j

Insulin-stimulated pAKTand pAKTare both attenuated inlivers () consistent with hepatic insulin-resistance. Curiously, inmuscle IR phosphorylation is slightly higher in the unstimulated state, while insulin-stimulated pAKTis blunted () suggesting some muscle signalling might also be altered in the mutant mice. However, AKTphosphorylation is stimulated normally inmuscle (), which is consistent with the clamp data showing normal insulin-stimulated muscle glucose clearance (). These data support a model in whichloss in fat most negatively affects hepatic function. Interestingly, primary hepatocytes isolated frommice maintain insulin resistance even when cultured(). Thus, the ‘damage' imposed on hepatic insulin signalling byloss in fat is not easily reversible. T308 S473 Adipoq-cre Adipoq-cre S473 T308 Adipoq-cre Adipoq-cre Rictor Rictor Rictor Rictor Rictor ex vivo Rictor Fig. 4a Fig. 4b Fig. 4b Fig. 2i Fig. 4c

Consistent with the model that DNL in WAT regulates insulin sensitivitywe find that adipose tissueloss dramatically reducesandmRNA and protein expression (and). The expression ofandexpression is unchanged in the mutant fat; however,induction is almost completely blocked in both pgWAT and sWAT (). This is consistent withexpression driving DNL in adipocytes. Indeed, overexpressing recombinant ChREBPβ, constitutively active ChREBP(ChREBP-CA), and to a lesser extent ChREBPα rescues expression of ACLY, ACC and FASN in-deficient adipocytes supporting this notion (). These data implicate adipocyte mTORC2 as a key upstream regulator of ChREBPβ-driven DNL. ^{5 6 24} Figs 1a 5a,b Fig. 5c,d ^{5 25} Fig. 6 Rictor Acly, Acc Fasn Chrebpα Srebf1c Chrebpβ Chrebpβ Rictor

Although under standard dietary conditions the adipose tissues predominantly obtain free FAs (FFAs) from the liver and diet rather than DNL, we wondered whether losingin WAT alters FA composition. The ratio of C16:0 (palmitate) to the essential FA C18:2n6 (linoleate)—an index of DNL—is slightly but significantly decreased insWAT () suggesting that despite the normal size of mutant WATs (though trending smaller) () DNL is decreased. Moreover, C18:0 (stearate) levels decrease while C16:1n7 (palmitoleate) and C18:1n9 (oleate) levels increase inWATs (). These findings are noteworthy because thesynthesis of palmitoleate by adipose tissue has been linked to improved systemic insulin action. Diets rich in oleate are also reportedly metabolically healthy. Thus, it does not appear that a palmitoleate or oleate deficiency in fat is causing insulin resistance inmice. Rictor Rictor Rictor de novo Rictor Adipoq-cre Adipoq-cre Adipoq-cre Fig. 5e Fig. 1e Fig. 5f,g ^{26 27 28}

The altered abundance of very long-chain FAs suggested mTORC2 might additionally regulate FA elongation and/or desaturation (). Elongation of very long-chain FAs 6 (ElOVL6) elongates C16:0 FAs to C18:0 FAs, while steroyl CoA desaturase (SCD1) desaturates C16:0 and C18:0 FAs to C16:1n7 and C18:1n9, respectively. We calculated the ELOVL6 and SCD1 activity ratios by dividing their products by their substrates (), which indicates a decrease and increase respectively in ELOVL6 and SCD1 activity supporting this hypothesis. A broad survey of elongase and desaturase gene expression further indicates thatexpression in the WAT of chow-fed mice requiresexplaining the reduced C18:0/C16:0 ratio (and). In contrast,expression is unchanged ().andare reportedly co-regulated ChREBP targets in the liver. Thus, in adipose tissueandare not necessarily co-regulated genes and the altered very long-chain FA profile in the WAT ofmice likely reflects defective elongation. Supplementary Fig. 3a Fig. 5h,i Fig. 5j,k Supplementary Fig. 3b–d Fig. 5j,k ²⁹ Elovl6 Rictor Scd1 Elovl6 Scd1 Elovl6 Scd1 Rictor Adipoq-cre

We also examined whether deleting adipose tissueaffects hepatic lipid composition. Interestingly,livers have a lipid profile and ELOVL6/SCD1 ratios that parallel the WATs but without changes in hepaticorexpression (). The livers frommice also have high levels of C14:0, C16:0, C18:1n9, C20:1n9 and C20:2 FAs with corresponding decreases in C18:2n6 and C20:4n6 FAs (). The C16:0/C18:2n6 ratio is elevated inliver () consistent with increased DNL. But as indicated above, most of the core hepatic lipogenic genes express normally in the mutant mice () indicating that hepatic lipid remodelling cannot easily be explained by differences in lipid synthesis genes. The hepatic lipid composition ofmice could also be remodeled in part by changes in exogenous lipid uptake as suggested by high hepaticexpression (). Regardless, the hepatic lipid profile ofmice is remarkably similar to that observed in livers of mice consuming a HFD despite the fact that they are consuming normal chow. This led us to hypothesize that HFD and adipose tissueloss might target a common pathway that antagonizes hepatic metabolism. Rictor Rictor Elovl6 Scd1 Rictor Rictor Rictor CD36 Rictor Rictor Adipoq-cre Adipoq-cre Adipoq-cre Adipoq-cre Adipoq-cre Fig. 5l–n Fig. 5l Fig. 5e Supplementary Fig. 2g,h Supplementary Fig. 2j ³⁰

To explore this further, we compared the effects of HFD to deletingin adipose tissue by feeding both WT andmice a HFD for 12 weeks. Interestingly,mice consuming HFD are resistant to weight gain (). This is due to a defect in adipose tissue expansion that may partly result from a slight reduction in food consumption (). This is in stark contrast tomice, which gain more total body weight, more pgWAT mass and more lean tissue mass on HFD. Rictor Rictor Rictor Rictor Adipoq-cre Adipoq-cre aP2-Cre Fig. 7a Fig. 7b,c ¹⁶

Importantly, there is no significant difference in insulin tolerance betweenmice consuming chow and age-matched WT mice consuming HFD (). Feeding WT mice a HFD also decreases,,,andexpression in both the pgWAT and sWAT to the same level as that which is observed in the chow-fedmice (). This is also reflected in the ACLY, ACC and FASN protein expression patterns particularly in the sWAT (). In addition, the HFD does not further exacerbate insulin resistance inmice despite these mice developing additional hepatic steatosis (and). Thus, HFD and adipose tissueloss have similar affects on gene expression and insulin sensitivity suggesting they may target a common pathway. Rictor Chrebpβ Acly Acc Fasn Elovl6 Rictor Rictor Rictor Adipoq-cre Adipoq-cre Adipoq-cre Fig. 7e–g Fig. 7h,i Supplementary Fig. 4a,b Fig. 7f,g Supplementary Fig. 4c

To test whether the adipocyte mTORC2-DNL pathway might be an early target of obesity-induced insulin resistance, we placed WT mice on HFD for only 2 weeks, which induces mild insulin resistance () and examined lipogenic gene expression. After only 2 weeks HFD,and, but notgene expression decreases in pgWAT and even more dramatically in the sWAT of WT mice (). These data are consistent with adipose tissue mTORC2 regulating DNL and insulin sensitivity by a mechanism that may be an early target of obesity. Fig. 8a Fig. 8b,c Chrebpβ, Acly, Acc Fasn Srebf1c

We next placedmice on a high carbohydrate/zero-fat diet (ZFD) for 12 weeks to maximize effects caused by a DNL deficiency.mice consuming a ZFD maintain a body weight similar to controls and consume the same amount of food (). However, the adipose tissues are smaller () indicating thatis also more critical for fat growth with increasing carbohydrate load. Liver mass also increases in ZFD-fed mutants () explaining why total body mass is unchanged relative to controls despite the decrease in fat mass. Rictor Rictor Rictor Adipoq-cre Adipoq-cre Fig. 7j,k Fig. 7l Fig. 7m

Interestingly, consuming a ZFD restores insulin tolerance inmice back to the level observed in the benchmark chow-fed controls (). This correlates with restoration of most lipogenic genes (for example,,,and) back to the benchmark expression level (that is, not significantly different from chow-fed controls) (). ZFD similarly restores sWAT lipogenic gene expression, and for,and, to levels even higher than in the sWAT of chow-fed controls (). These increases are mirrored by increases in ACLY, ACC and FASN protein expression (). Notably, while ZFD increases lipogenic gene expression inmice to chow-fed levels, lipogenic gene expression still remains lower relative to the ZFD-fed control group (). In fact, lipogenic gene expression responds robustly in the ZFD-fed control group exemplified by ∼11- and 28-foldinduction in pgWAT and sWAT, respectively, over chow-fed controls (). Thus, even though a ZFD improves insulin sensitivity inmice (presumably by ‘forcing' more glucose into adipocytes), the lipogenic genes do not respond at full capacity. Notably, ZFD also increases hepatic lipogenic gene expression and steatosis in both the control andmice (). This is consistent with high ChREBP expression in the liver increases hepatic lipogenic gene expression and steatosis without impairing insulin sensitivity. Collectively, these data support a model in which adipose tissue mTORC2 regulates lipogenic gene expression to produce an insulin-sensitizing signal. Rictor Chrebpβ Acly Acc Elovl6 Chrebpβ Acly Elovl6 Rictor Chrebpβ Rictor Rictor Adipoq-cre Adipoq-cre Adipoq-cre Adipoq-cre Fig. 7g,n Fig. 7o Fig. 7p Supplementary Fig. 5a,b Fig. 7o,p Fig. 7o,p Supplementary Fig. 5c,d ²⁹

To better define the mechanism by which adipose tissueloss alters adipocyte function, we generated primary white preadipocytes harboring a() inducible-KO () system to examine the acute effects of inhibiting mTORC2 on differentiation and function. When compared with their isogenic controls, primarypreadipocytes differentiate normally based on PPARγ, ChREBPα,,,andexpression (and). In high-glucose culture medium, DNL is the primary driver of lipid droplet formation and under these conditionscells have smaller lipid droplets ().cells also fail to upregulatemRNA andandmRNA and protein (), consistent with a DNL deficiency. Moreover, both basal and insulin-stimulated glucose uptake as well as glucose incorporation into FFAs and TAGs are blunted incells (). As expected, RICTOR is completely ablated in thecells by day 6 of differentiation at which point AKTphosphorylation is also undetectable (). In contrast, phosphorylation of AKT, AS160and S6K(a measure of mTORC1 activity) is unchanged relative to the isogenic control (). Thus, decreased glucose uptake and DNL is primary consequence ofloss. Rictor Rictor UBC-CRE Rictor-iKO Rictor-iKO C/ebpα C/ebpβ Ap2 Adiponectin Rictor-iKO Rictor-iKO Chrebpβ Acly, Acc Fasn Rictor-iKO Rictor-iKO Rictor ERT2 S473 T308 T642 T389 Fig. 9a Supplementary Fig. 6a Fig. 9b Fig. 9a,c Fig. 9d,e,f Fig. 9a Fig. 9a

GLUT4 is the major glucose transporter for insulin-stimulated glucose uptake into adipocytes. Therefore, we next examined whether GLUT4 regulation is defective in-deficient adipocytes. Indeed, in primarycells,mRNA and protein fail to induce normally during differentiation (and). Similarly, in mice consuming the lipogenic ZFD,is also reduced in both the sWAT and pgWAT ofmice (and). Thus, in highly lipogenic conditions such as in culture medium and a high-carbohydrate diet,is required for maximalgene expression. In mice consuming normal chow however,expression does not significantly differ in either depot from controls (and). Moreover, in HFD-fed miceexpression only decreases in the mutant sWAT and not pgWAT (and). Thus, additional mechanism(s) of glucose regulation by/mTORC2 in adipocytes likely exist. Rictor Rictor-iKO Glut4 Glut4 Rictor Rictor Glut4 Glut4 Glut4 Rictor Fig. 9g Supplementary Fig. 6b Fig. 9h Supplementary Fig. 6c Fig. 9h Supplementary Fig. 6c,d Fig. 9h Supplementary Fig. 6c Adipoq-cre

We also considered other proposed modulators of insulin resistance, such as increased lipolysis. Primaryadipocytes show no difference in basal or isoproterenol-stimulated glycerol release () arguing against increased lipolysis as a primary target ofloss. However, pgWAT depots resected frommice show a modest increase in basal glycerol release; isoproterenol-stimulated glycerol release is normal (). When administered at a controlled concentration, insulin is also ineffective at suppressing circulating FFA levelsin fastedmice (), which is consistent with insulin resistance but could reflect liver dysfunction. High cholesterol is detected in the mutants, however, circulating FFA and TAG levels are not significantly different between fasted control andmice possibly due to the high basal insulin levels (–h). Moreover,mice have normal hormone sensitive lipase (HSL) levels in both depots and increased HSL phosphorylation is not detected (). Higher levels of adipose triglyceride lipase (ATGL) associate with increased lipolysis; however, ATGL levels are also normal in pgWAT and lower in sWAT (). Thus, defective lipolysis may contribute to insulin resistance in the prolonged absence of/mTORC2 in fat but it does not appear to be a primary effect. Rictor-iKO Rictor Rictor ex vivo in vivo Rictor Rictor Rictor Rictor Fig. 9i Fig. 9j Supplementary Fig. 6e Supplementary Fig. 6f Fig. 1a Fig. 1a Adipoq-cre Adipoq-cre Adipoq-cre Adipoq-cre

TNF-α is undetectable inmice and leptin, resistin and PAI-1 levels are normal (). We do detect reduced plasma adiponectin by ∼32% (), which could contribute to insulin resistance. However, reducing adiponectin levels reportedly has no or only a mild defect in insulin sensitivity in chow-fed micesuggesting this alone likely does not explain the severe insulin resistance ofmice. Thus, based on the collectiveandprimary cell data, we propose that a primary function of mTORC2 in adipocytes is to control ChREBP activity by regulating glucose flux, which promotes DNL and the production of a signal(s) that promote insulin sensitivity and possibly adipogenesis (). However, prolonged/mTORC2 loss may lead to secondary metabolic changes that exacerbate the phenotype. Rictor Rictor in vivo in vitro Rictor Adipoq-cre Adipoq-cre Supplementary Fig. 6i–k Supplementary Fig. 6l ^{31 32} Fig. 10

-floxed mice were described previouslyand backcrossed with C57BL/6 for 10 generations. Floxed mice were crossed with mice expressing either adiponectin-Cre or adiponetin- Cre, or with Ubc-Cre(JAX #007001) mice to generate conditional or inducible KO models. Floxed Cre-negative mice were used as wild-type controls. Mice were kept on a daily 12 h light/dark cycle and fed a normal chow diet (Prolab Isopro RMH 3000) from Lab Dietat 22 °C. All animal experiments were approved by the University of Massachusetts Medical school animal care and use committee. For the inducible CreER models,(WT) and(iKO) male mice at 8 weeks old were treated with 3 mg Tamoxifen per day (i.p.) for 6 constitutive days. The age of the mice used for all studies were 8–20 weeks old. For 2 weeks HFD experiment, mice were randomly placed into cages with chow or HFD. No other randomization was used while conducting experiments. No animals were excluded from any experiments, unless they displayed obvious wounds from fighting as determined by our veterinarians Rictor ad libitum Rictor Adiponetin-CreERT2;Rictor ⁵³ ERT2 ERT2 fl/fl fl/fl

All animal studies were designed to minimize and control for confounding variables such as mouse gender and age. Based on our previous studies, we use a minimum of six animals per treatment group to achieve statistical power to detect significant differences when measuring RNA, tissue mass, body weight and blood metabolites. Researchers were not blinded to the genotype.

AS160 (07-741) was purchased from Millipore. PPARγ (sc-7196) and p70 S6K (sc-9027) are from Santa Cruz. ChREBP (NB400-135) is from Novus Biologicals. All other antibodies including Rictor (2140), Raptor (2280), HSL (4107), ATGL (2439), PRAS40 (2691), AKT (9272), GSK3β (9315), ACC (3676), ACLY (4332), FASN (3180), IR (3025), S473-AKT (4058), T308-AKT (4056), T24-FoxO1 (9464), T389-S6K (234), p-IR (3024), T246-pPRAS40 (2997), S660-pHSL (4126) and T642-pAS160 (4288), were purchased from Cell Signaling Technologies. 4-hydroxy-tamoxifen (4-OHT) was obtained from Toronto Research Chemicals. Rapamycin was purchased from LC Laboratories. Dexamethasone, 3-isobutyl-1-methylxanthine (IBMX), Tamoxifen, and all other reagents were from Sigma-Aldrich.

Hyperinsulinemic-euglycemic clamps were performed following an overnight fast, a 2-h hyperinsulinemic (insulin at 150 mU kgbody weight priming followed by 2.5 mU kgmin)-euglycemic clamp was conducted in awake mice using (3-3H)-glucose and 2-deoxy-D-(1-14C)-glucose to assess glucose metabolism in individual tissues as described previously. At 8 weeks of age, male mice were fed a normal chow diet (Prolab Isopro RMH 3000) from Lab Diet, 60% HFD (D12492 Harlan Laboratories) or high-carbohydrate ZFD (TD.03314 Harlan Laboratories) and monitored for 12 weeks. Body weight was recorded weekly. The analysis of blood metabolites was performed by at the Joslin Diabetes Center (Boston). For glucose tolerance tests and pyruvate tolerance tests, mice were fasted overnight (16 h) and then administrated 2 g kgof body weight of glucose or sodium pyruvate by intraperitoneal (i.p.) injection. For insulin tolerance tests, mice were fasted for 6 h before i.p. administration of 0.75 unit kgof body weight of insulin. Blood glucose concentrations were measured before and after the injection at the indicated time points. −1 −1 −1 −1 −1 ⁵⁴

Polar and non-polar metabolites were extracted from tissue using methanol/water/chloroform and derivitized as previously described. Briefly, polar metabolites were derivatized to form methoxine-TBDMS derivatives by incubation with 2% methoxylamine hydrochloride dissolved in pyridine at 37 °C for 1 h followed by addition of N-tert-butyldimethylsilyl-N-methyltrifluoroacetamide (MTBSTFA) with 1% tert-butyldimethylchlorosilane (TBDMCS) incubated at 37 °C for 30–60 min. Nonpolar metabolites were saponified to FFAs and transesterified to form FA methyl esters by incubation with 2% H2SO4 in methanol at 50 °C for 1 h. Derivatized polar samples were analysed by gas chromatography–mass spectrometry using a DB-35MS column (30 m × 0.25 mm i.d. × 0.25 um) installed in an Agilent 7890B gas chromatograph (GC) interfaced with an Agilent 5977 A mass spectrometer. Lipid samples were analysed by gas chromatography–mass spectrometry using a Select FAME column (100 m × 0.25 mm i.d.) installed in an Aglient 7890A GC interfaced with an Agilent 5975C mass spectrometer. ⁵⁵

Adipose tissue depots were carefully dissected to avoid contamination from surrounding tissue. Samples for RNA or protein were frozen down immediately in liquid nitrogen and then stored at −80 °C for further analysis. For histology, tissue pieces were fixed by 10% formalin. Embedding, sectioning and Hematoxylin & Eosin (HE) and PAS staining was done by the UMass Medical School Morphology Core. For Oil Red O staining, liver samples were embedded in OCT before sectioning and staining. For cell size measurements, 9–12 images were taken from each mouse (=3 wild-type and 3 conditional KOs). Image J was used to measure cell size and the distribution of cell size as percentage of total counted cells was analysed. n

Insulin stimulated signalling in each tissue was determined in mice that were fasting for 6 h before injection with 0.75 unit kgof body weight of insulin for 15 min. Each tissue was collected and frozen down immediately in liquid nitrogen and then stored at −80 °C for subsequent lysis and western blots analysis with the indicated antibodies. Cells were lysed in a buffer containing 50 mM Hepes, pH 7.4, 40 mM NaCl, 2 mM EDTA, 1.5 mM NaVO4, 50 mM NaF, 10 mM sodium pyrophosphate, 10 mM sodium β-glycerophosphate and 1% Triton X-100 typically 16 h after the cells were replenished with fresh culture medium. Tissues were homogenized using a TissueLyser (Qiagen) in the same lysis buffer but additionally supplemented with 0.1% SDS, 1% sodium deoxycholate. An equal amount of total protein was loaded into acrylamide/bis-acrylamide gels and transferred to polyvinylidene fluoride membranes for detection with the indicated antibodies. Briefly, membranes were incubated with primary antibodies (1:1,000 dilution) in 5% milk/PBST or 5% BSA/PBST overnight. Horese radish peroxidase-conjugated secondary antibodies (1:3,000 dilution) were given for 1 h. Western blots were developed by enhanced chemiluminescence (PerkinElmer) and detected by X-ray films. −1

SVF cells were isolated by digesting the sWAT ofmice in digestion buffer (NaCl (123 mM), KCl (5 mM), CaCl2 (1.3 mM), glucose (5 mM), HEPES (100 mM), P/S (1%), BSA (4%), pH 7.4) containing collagenase A at 1.5 mg ml(Roche). After 45 min of digestion, the digested tissue was filtered through 100 μm cell strainers (BD Falcon). Cells were collected by centrifugation (at 300, 5 min) and cultured in DMEM (Invitrogen) supplemented with 10% FBS and penicillin/streptomycin at 37 °C. To inducedeletion,SVF cells were treated with 1 μM 4-OHT or vehicle for 2 days when start to differentiate by adding differentiation medium containing 2 μg mldexamethasone, 100 nM insulin, 1 μM rosiglitazone and 0.5 mM IBMX for 2 days, and then induced with 100 nM insulin for another 2 days before changing back to regular medium. At different time points, the differentiated adipocytes were collected for protein, mRNA or Oil-Red-O staining analysis. For Oil Red O staining, the differentiated cells were washed three times with PBS and fixed with 10% buffered formalin at 4 °C overnight. Cells were then stained for 10 min at 37 °C with a filtered Oil Red O solution (0.5% Oil Red O in isopropyl alcohol), washed three times with distilled water, and visualized. Ubc;rictor g Rictor Ubc;rictor −1 −1

Isolated SVF from Ubc;rictor mouse were cultured and differentiated into adipocytes for 7 days. The cells were then incubated in KRH buffer supplemented with 2.5% BSA and 2 μCi per ml D-[U-14C]-glucose (PerkinElmer) for 4.5 h with or without the presence of 150 nM insulin. The cells were lysed with Doley's solution (isopropyl alcohol:hexane:1N H2SO4 (v:v:v)=4:1:1), and newly synthesize lipids were extracted with hexane. About 1/3 of hexane phase was used for analysing triglycerides (TAGs), and the remaining were dried, deacylated with ethanol:water:KOH (v:v:v)=20:1:1 at 80 °C for an hour, neutralized with sulfuric acid and dissolved the neutral FFAs in hexane. The samples containing TAGs and FFAs were dried, reconstituted in scintillation fluid, and incorporation of 14C was determined by counting counts per minute (CPM). Each condition was done in duplicate, and the experiment was repeated with three independently isolated SVF. The incorporation was then normalized with protein content measured by BCA protein assay kit (Bio-Rad).

Isolated SVF from Ubc;rictor mouse were cultured and differentiated into adipocytes for 7 days. The cells were incubated in KRH buffer supplemented with 0.5% BSA and 2 mM sodium pyruvate, with or without 150 nM insulin stimulation for 15 min. [1, 2-H] 2-deoxy-D-glucose (PerkinElmer) was added to the samples and incubate at 37 °C for 10 min, and the assays were terminated by three-time KRH wash. The cells were lysed with 1% Triton, dissolved in scintillation buffer and uptakenH was determined by counting CPM level. The 2-DOG uptake level was normalized with protein concentration of each sample. Each condition was done in triplicate. 3 3

Hepatocytes were isolated from mice using a modified two-step perfusion method using Liver Perfusion Media and Liver Digest Buffer (Invitrogen). Cells were seeded on plates (pre-coated (1 h) with collagen I (BD Biosciences)) in DMEM plus 10% FBS, 2 mM sodiumpyruvate, 1 μM dexamethasone and 100 nM insulin plus 2% penicillin/streptomycin. After attachment (3 h), the medium was removed and the hepatocytes were incubated (22 h) in maintenance medium (DMEM (4.5 g lglucose) supplemented with 10% FBS, 0.2% BSA, 2 mM sodium pyruvate, 2% Pen/Strep, 0.1 μM dexamethasone, 1 nM insulin) before stimulated with insulin at 100 nM insulin for 30 min. −1

For measurement of lipolysis, differentiated adipocytes at day7 and pgWAT from mice were cultured in DMEM with or without isoproterenol at 10 μM for 4 or 6 h, respectively, before collecting medium to measure glycerol concentration using commercial kit (Sigma). The glycerol level was normalized with protein concentration of differentiated adipocytes and tissue mass of pgWAT.

Paraffin-embedded pancreatic sections were immunohistochemical stained with insulin. β-Cell mass of five mice per group of each genotype was measured in insulin-stained pancreas section using ImageJ (NIH, Bethesda, MD).

Cells or tissues were lysated with Qiazol (Invitrogen) and total RNA was isolated with the RNeasy kit (Invitrogen). Equal amounts of RNA were retro-transcribed to cDNA using a high-capacity cDNA reverse transcription kit (#4368813, Applied Biosystems). Quantitative real-time PCR was performed in 10 μl reactions using a StepOnePlus real-time PCR machine from Applied Biosystems using SYBR Green PCR master mix (#4309156, Applied Biosystems) according to manufacturer instructions. Relative mRNA expression was determined by the ΔCt method and Tbp expression was used as a normalization gene in all conventional PCR with reverse transcription experiments. Primer information is listed in (Supplementary Table 2).

Unless otherwise stated, values given are mean±s.e.m. Two-way analysis of variance was performed where indicated. For most experiments, unpaired two-tailed Student's-tests was used to determine statistical significance among two groups (*<0.05; **<0.01;***<0.001). t P P P