Adipocyte Glucocorticoid Receptor Deficiency Attenuates Aging- and HFD-Induced Obesity and Impairs the Feeding-Fasting Transition

Energy homeostasis requires the integration of multiple signals between the central nervous system and the periphery to adjust substrate distribution in accordance with metabolic demands. The steroid hormones glucocorticoids (GCs) are important integrators in the body’s adaptation to energetic stress by regulating several components of energy homeostasis, including glucose and lipid metabolism (1,2). Consequently, states of chronic GC exposure are associated with metabolic dysfunctions (1,2). The cellular effects of GCs are mediated in large part through activation of the glucocorticoid receptor (GR). The understanding of how tissue-specific functions of the GC-GR axis contribute to systemic energy metabolism has been substantially extended by mouse models of discrete GR deficiency in liver and muscle (1,3,4).

The most abundant energy reservoir in mammals is white adipose tissue (WAT), which allows nonadipose tissues to function normally under conditions of overnutrition or fasting (5,6). Thus, an appropriate control in WAT to store and release energy in response to changes in nutrient availability is critical for metabolic homeostasis. GCs were shown to induce lipolysis (7–9), to stimulate lipogenesis in the presence of insulin (10,11), and to promote lipid storage, uptake, and mobilization (2,9,12). An increase of local GC regeneration by the enzyme 11β-hydroxysteroid dehydrogenase 1 (11β-HSD1) within WAT suggests a role of the adipocyte GC-GR axis in common obesity (2,12–16). These lines of evidence are consistent with a central role of GCs in regulating energy metabolism. However, the exact physiological importance of the adipocyte GR for the maintenance of systemic metabolic homeostasis is yet to be determined.

To delineate this undefined role of the GR, we used an adipocyte-specific gene knockout strategy in mice. We demonstrate that adipocyte GR significantly impacts systemic nutrient partitioning in different energetic states. While GR deficiency disrupts the feeding-fasting transition, it ameliorates obesity and its associated metabolic disorders, thus exerting opposing roles in the regulation of metabolic homeostasis.

Adipocyte-specific GR-deficient mice (Nr3c1^tm2GscTg(Adipoq-cre)1Evdr: GR^∆Adip; C57BL/6 x FVB/N) were generated by crossing Nr3c1 floxed (17) with Adipoq-cre mice (18). Adipoq-cre negative littermates served as controls (Nr3c1^tm2Gs). Animals were housed under standardized conditions (12-h dark/12-h light cycle) and fed a regular diet (Ssniff EF, R/M Kontrolle; Ssniff GmbH, Soest, Germany). For high-fat diet (HFD) experiments, mice received either Ssniff EF acc.D12492 (34.6% crude fat) or Ssniff EF D12450B mod. Animal studies were approved by the Austrian government and the Medical University of Vienna (BMWF-66.009/0132-II/3b/2013) and by the Regional Commission Tuebingen, Germany (TVA1126). Except for HFD and aging experiments, 8-week-old male mice were used.

Body composition was determined using EchoMRI-100H (EchoMRI LLC, Houston, TX). µ-CT was performed with in vivo X-ray microtomograph Skyscan 1176 (RJL Micro & Analytic GmbH, Karlsdorf-Neuthard, Germany). Fat volume was reconstructed and calculated using NRecon (version 1.6.9.18) and CTAn (version 1.14.4.1+).

Glucose and pyruvate tolerance tests (GTTs and PTTs) were performed in 12–16 h–fasted mice; insulin tolerance tests (ITTs) were performed in 4 h–fasted mice. Glucose (2 g/kg) was administered orally or by intraperitoneal injection (HFD experiments); pyruvate (2 g/kg) and insulin (0.75 units/kg) were given intraperitoneally. Blood glucose levels were determined from the tail vein using a glucometer.

Cold tolerance was determined in 4 h–fasted animals. Mice were housed separately with free access to water at 4°C. Rectal temperature was measured at indicated time points using a BIO-TK9882 thermometer (Bioseb, Vitrolles, France).

Energy expenditure (EE), locomotor activity, respiratory exchange ratio (RER), and food intake were measured by combined indirect calorimetry over 93.4 h (PhenoMaster; TSE Systems, Bad Homburg vor der Höhe, Germany) as described previously (19).

β-Ketones were measured from tail vein blood (Freestyle Precision Xceed, Abbott, Alameda, CA). Nonesterified fatty acids (NEFAs) were determined with the NEFA-HR (2) kit (Wako Chemicals, Neuss, Germany), glycerol with the Free Glycerol Reagent (Sigma-Aldrich, St. Louis, MO), and triglycerides (TGs) and cholesterol were measured with a Reflotron Plus analyzer (Roche, Basel, Switzerland). Insulin, corticosterone, and FGF21 were determined by ELISA (Ultra-Sensitive Mouse Insulin ELISA [Crystal Chem, Downers Grove, IL], Corticosterone ELISA [Enzo Life Sciences, Farmingdale, NY], and Mouse/Rat FGF-21 Quantikine-ELISA [R&D Systems, Minneapolis, MN]). Liver TG content was determined using a Triglyceride Colorimetric Assay (Cayman Chemical, Ann Arbor, MI).

Epididymal WAT (eWAT) was surgically removed and lipolysis was measured as previously described (20). Stimulations were performed with insulin (30 ng/mL), isoproterenol, forskolin, formoterol, or CL-316,243 (all 10 μmol/L) for 120 min. NEFA was determined as previously described, and cAMP content in WAT explants by the cAMP complete ELISA (Enzo Life Sciences). Measurements were normalized to tissue weights.

Tissues were fixed in 4% buffered formalin, paraffin embedded, sectioned, and stained with hematoxylin-eosin (H-E) using standard procedures. Adipocyte sizes were quantified from at least three to five different fields per mouse and at least 20–60 cells/field using ImageJ (Rasband, W.S., ImageJ; NIH, Bethesda, MD, https://imagej.nih.gov/ij/). Histological evaluation of livers was performed by a board-certified pathologist (J.H.).

RNA was extracted using commercial kits (Qiagen, Hilden, Germany; Peqlab, Erlangen, Germany). RNA (1 μg) was reverse transcribed using cDNA synthesis kits (Thermo Fisher Scientific, Waltham, MA; Applied Biosystems, Foster City, CA). Quantitative real-time PCR (qRT-PCR) was performed on an Eppendorf Realplex system using the Taq DNA Polymerase Kit (Eppendorf, Hamburg, Germany) or on a ViiA7RT-PCR System using the Platinum Sybr Green qPCR Supermix-UDG (Life Technologies, Carlsbad, CA). Gene expression was normalized to Gapdh or Actb (β-actin) mRNA. Primer sequences are provided in Supplementary Table 1.

Western blot analyses (40 μg protein) were performed as previously described (21). Primary antibodies against HSC70 (sc-7298), GR (sc-1004), AKT (sc-8312), G_Sα (sc-823), peroxisome proliferator–activated receptor α (PPARα; sc-9000) (all Santa Cruz Biotechnology, Santa Cruz, CA), pS473-AKT (no. 9271), uncoupling protein 1 (UCP1; no. 14670), adipose triglyceride lipase (ATGL; no. 2439), pSer563-HSL (no. 4139), pSer660-HSL (no. 4126), hormone-sensitive lipase (HSL; no. 4107), pPKA substrate (RRXS*/T*; no. 9624) (all Cell Signaling Technology, Danvers, MA) and β-actin (A1978; Sigma-Aldrich) were used.

Animals were anesthetized with ketamine/xylazin (i.p., 100 mg/kg and 10 mg/kg, respectively) or CO₂ inhalation. Blood was drawn (500–700 µL) via cardiac puncture, filled into lithium-heparin– or EDTA-coated tubes (Greiner Bio-One, Kremsmünster, Austria), and centrifuged in <30 min after collection (15 min, 7,000 rpm, 4°C). Plasma was snap frozen and stored at −80°C.

Metabolites were analyzed by targeted liquid chromatography (LC)–high resolution mass spectrometry (MS) metabolomics according to Bajad et al. (22) by hydrophilic interaction LC at the HEALTH Institute for Biomedicine and Health Sciences, JOANNEUM RESEARCH (Graz, Austria) as described previously (23). Samples were processed according to Yuan et al. (24). Raw data were converted into mzXML by msConvert (ProteoWizard Toolkit v3.0.5), and metabolites were targeted-searched by the in-house developed tool PeakScout, with a reference list containing accurate mass and retention times in agreement with standards outlined by Sumner et al. (25). Pure substances of all analytes, except lipids, were run on the same system to obtain exact reference retention times and fragmentation spectra. The results for six samples had to be removed from analysis due to considerable outlier behavior in principal component analysis (PCA; two from GR^ΔAdip HFD cohort and one from each of the fasted control, fasted GR^ΔAdip, fed control, and GR^ΔAdip cohorts).

Statistics were performed with GraphPad Prism or R (26) (v3.2.1, packages stats, missMDA, and nlme) using TibcoSpotfire (v7.0.0). All data except LC-MS metabolomics are presented as mean ± SEM. Two-tailed Student t test or Wilcoxon rank-sum test was used for comparing two groups and one-way ANOVA followed by Tukey, Dunns, or Bonferroni post hoc tests for multiple comparison. Tolerance tests and body growth curves were analyzed with repeated-measures two-way ANOVA followed by Bonferroni post hoc tests. P < 0.05 was considered statistically significant. Data for EE were analyzed using ANCOVA with body weight and body composition (fat and lean mass) as covariates as previously suggested (27). For LC-MS metabolomics, PCA analysis was performed centered and scaled to unit variance (R function prcomp). Missing values were imputed by a regularized expectation-maximization (R function imputePCA and estim_ncpPCA). Log₁₀-transformed data were found to be sufficiently normally distributed according to Kolmogorov-Smirnov (86% of all metabolites were normally distributed) and sufficiently homoscedastic according to Levene (84% of all metabolites were homoscedastic). Differences between independent groups were analyzed by ANOVA (R function aov) followed by Benjamini-Hochberg (R function p.adjust) post hoc test.

Deletion of Nr3c1 was confirmed by mRNA and protein expression analyses of inguinal WAT (iWAT), eWAT, and brown adipose tissue (BAT) (Fig. 1A and B). GR^∆Adip mice displayed no overt alterations in weights or morphology of eWAT, iWAT, and BAT compartments (Supplementary Fig. 1A and B). Accordingly, adipogenic markers (Cebpb, Pparg, and Fabp4) and adipocyte-specific genes (Adiponectin, Leptin, and Ucp1) were not differentially expressed among genotypes (Supplementary Table 2).

To initially illustrate the basal metabolic signature of GR^ΔAdip mice, plasma samples of ad libitum–fed mice were analyzed by targeted LC-MS metabolomics (Supplementary Table 3; PCA results: Supplementary Fig. 2A). A total of 157 metabolites were of suitable quality for multivariate statistical analysis, 20 metabolites only for univariate statistical analysis. In GR^ΔAdip mice compared with controls, 59 metabolites were significantly decreased (P < 0.05, ANOVA) and 22 were decreased by trend (0.05 < P < 0.1, ANOVA). The majority of decreased metabolites were related to fatty acid (FA)/lipid metabolism (46, most pronounced in long-chain FA), amino acid (AA) metabolism (14, mainly proteogenic and branched-chain AAs, which have been implicated as independent risk factors for diabetes [28]), and nucleotide metabolism (10, mainly pyrimidine metabolites) (Fig. 1C and Supplementary Table 3). This demonstrates a considerable impact of adipocyte GR deficiency on the plasma metabolome and, thus, systemic substrate partitioning, which likely affects whole-body glucose and lipid metabolism in GR^ΔAdip mice (6,29).

First, we analyzed glucose homeostasis in GR^ΔAdip mice. In 4 h– and 16 h–fasted mice, blood glucose and plasma insulin levels were not significantly altered (Fig. 2A). GTTs revealed no differences in glucose clearance between the genotypes (Supplementary Fig. 2B). The hypoglycemic response during ITTs was similar in both genotypes (Fig. 2B), whereas blood glucose starting from 30 min after insulin administration was lower in GR^ΔAdip mice. This delay in posthypoglycemic recovery suggests a reduction in counter-regulatory mechanisms such as hepatic glucose production (HGP). Insulin-induced AKT phosphorylation (Ser473) was greater in GR^ΔAdip livers than in controls (Fig. 2C). This was associated with a lower TG content in GR^ΔAdip livers (Fig. 2D). Suppression of WAT lipolysis can indirectly reduce HGP by decreasing the influx of glycerol and NEFA (30). In the absence and presence of insulin, NEFA release from eWAT explants of 4 h–fasted GR^ΔAdip mice was clearly reduced compared with controls (Fig. 2E). Yet, consistent with a comparable insulin-stimulated AKT phosphorylation in eWAT (Fig. 2C), insulin-mediated suppression of lipolysis was similar between genotypes (49% vs. 51%). NEFA release from 16 h–fasted GR^ΔAdip eWAT explants was also reduced, which was mirrored in decreased plasma NEFA (indicator for WAT lipolysis) and blood β-ketones (indicator for hepatic FA oxidation [FAO]) in 16 h–fasted GR^ΔAdip mice (Fig. 2F). Blood glucose levels of 16 h–fasted GR^ΔAdip mice during PTTs were lower compared with controls (Fig. 2G), reflecting reduced glucose production from the carbon precursor pyruvate. Thus, the lipolytic capacity of GR^ΔAdip mice is diminished, which correlates with a decrease in glucose production (Supplementary Fig. 2C).

Thermogenesis in BAT requires mobilization of lipid stores, induction of β-oxidation, and mitochondrial uncoupling. To evaluate BAT functionality upon adipocyte GR deficiency, we subjected mice to a brief 4-h fast followed by a 4-h exposure to 4°C. Body temperatures of GR^ΔAdip were slightly higher when housed at 23°C (4-h fast; P = 0.072), whereas their ability to maintain stable body temperatures at 4°C was mildly reduced (Fig. 3A). Despite the impaired cold adaptation, UCP1 and PPARα levels were similar in BAT of cold-exposed GR^ΔAdip and control mice (Fig. 3B). While BAT uses FA stores to fuel thermogenesis, WAT provides energy in the form of FA for utilization in BAT. The histology of BAT revealed no differences between genotypes when housed at 23°C (Supplementary Fig. 1B). Yet, upon cold exposure, the amount of lipid droplets decreased to a higher extent in GR-deficient BAT (Fig. 3C), suggesting that BAT lipolysis is not impaired. In contrast, plasma NEFA and β-ketone concentrations, indicators of WAT lipid mobilization and redistribution, were reduced in cold-exposed GR^ΔAdip mice (Fig. 3D). Considering the requirement of external FA supply for BAT-mediated temperature maintenance (31), these data suggest that the reduced cold-induced thermogenesis is related to impaired WAT lipolysis and NEFA flux.

Progressing age is linked to abnormalities of carbohydrate and lipid metabolism. Thus, we analyzed whether GR deficiency would lead to beneficial metabolic effects at older age. Aging-associated weight gain was attenuated in GR^ΔAdip mice (Fig. 4A) and body fat was reduced compared with controls at 52 weeks of age (Fig. 4B). Accordingly, GR^ΔAdip mice had smaller WAT compartments (Fig. 4C), in which adipocyte hypertrophy was less frequently observed (Supplementary Fig. 3A). GR^ΔAdip mice that were 4 h fasted displayed lower blood glucose levels and presented an overall reduction of total TG, cholesterol, NEFA, and β-ketones in the circulation, whereas corticosterone levels were comparable between genotypes (Supplementary Fig. 3B). Weights of GR^ΔAdip livers were reduced (Fig. 4C) and steatosis scores of >5% were found only in control livers (Fig. 4D).

To further characterize the phenotype of aged GR-deficient mice, we determined their metabolic parameters in a PhenoMaster metabolic cage system. GR^∆Adip mice had higher RER during light and dark cycles, indicating an overall preference for carbohydrates as metabolic substrate. Locomotor activity, EE, and food intake were not significantly different between genotypes (Fig. 4E and Supplementary Fig. 3C), although GR^∆Adip mice consumed a slightly higher amount of food. There was no significant effect of GR deficiency on EE after ANCOVA with body weight or lean or fat mass as covariate (data not shown). These data indicate that neither increased physical activity/metabolic rates nor decreased food intake are determinants of reduced adiposity in aged GR^∆Adip mice. However, we cannot fully exclude subtle increases in daily EE and/or locomotor activity that might have contributed to reduced fat mass gain during the aging process.

To investigate whether reduced susceptibility to metabolic dysfunctions also occurs upon diet-induced obesity, mice were subjected to HFD feeding for 20 weeks. HFD-fed GR^∆Adip mice gained significantly less weight than HFD-fed controls, despite similar cumulative food intake (Fig. 5A and Supplementary Fig. 4A). Reduced weight gain manifested early in HFD-fed GR^∆Adip mice (Fig. 5A), accompanied by lower levels of several circulating FA species and a trend toward reduced plasma NEFA (Supplementary Table 3 and Supplementary Fig. 4B). After 2 weeks of HFD feeding, mRNA expression of several genes critical for FA storage were either significantly reduced or reduced by trend in GR^∆Adip eWAT compared with controls (acyl-CoA carboxylase [Acaca], glycerol-3-phosphate acyltransferase 3 [Gpat3], diacylglycerol-acyltransferase-2 [Dgat2], phosphoenolpyruvate carboxykinase [Pck1], and FA transport protein 1 [Slc27a1]) (Fig. 5B). Acaca and Pck1 mRNA levels showed a similar trend in GR-deficient iWAT, whereas there was no change in expression of FA synthase (Fasn) and the FA transporter Cd36 (platelet glycoprotein 4) in both WAT compartments. At termination, HFD-fed GR^∆Adip mice displayed lower WAT weights than HFD-fed controls (Supplementary Fig. 4C), and µ-CT analysis confirmed significantly decreased subcutaneous and visceral fat volume of GR^∆Adip mice (Fig. 5C). Accordingly, adipocyte hypertrophy in GR^∆Adip eWAT and iWAT was reduced (Fig. 5D and Supplementary Fig. 4D). Consistent with the aging cohort, plasma corticosterone levels were not significantly different between HFD-fed control and GR-deficient mice (Supplementary Fig. 4E).

Next, we asked whether GR deficiency improves deteriorated glucose metabolism associated with HFD. Blood glucose levels trended lower in HFD-fed GR^∆Adip mice compared with controls (Fig. 6A), and their fasting plasma insulin levels were significantly lower (Fig. 6B). In agreement, HOMA-IR was decreased in HFD-fed GR^∆Adip mice, suggesting improved insulin sensitivity (Fig. 6C). Consistent with the lower HOMA-IR, systemic glucose tolerance of HFD-fed GR^∆Adip mice was improved compared with obese controls (Fig. 6D), and insulin-stimulated AKT phosphorylation (Ser473) in muscle, eWAT, and liver of HFD-fed GR^ΔAdip mice was increased (Fig. 6E and Supplementary Fig. 4F).

Similar to the aging cohort, adipocyte GR deficiency attenuated HFD-induced hepatic lipid accumulation and liver weights, which was reflected in lower steatosis scores (Fig. 6F and G). Hepatic mRNA expression of Fasn and Gpat1, two rate-limiting enzymes of lipogenesis, was reduced in HFD-fed GR^∆Adip mice, accompanied by diminished expression of Pparg and two FA transporters, Cd36 and Slc27a4 (long-chain FA transport protein 4) (Fig. 6H). This suggests that decelerated FA uptake and lipogenesis might contribute to lower steatosis scores in GR-deficient mice. Thus, adipocyte GR deficiency reduces obesity, diminishes hepatic steatosis, and improves glucose tolerance in HFD-fed mice.

Having demonstrated that adipocyte GR promotes metabolic dysfunctions in aged and fat-fed mice, we lastly evaluated the requirement of adipocyte GR for energy homeostasis under prolonged fasting conditions. Upon 48-h fasting, total body weight loss was similar among genotypes (Supplementary Fig. 5A). Notably, all mice survived 48-h fasting and did not present symptoms of a hypoglycemic shock (i.e., trembling, seizures, or unconsciousness). Body fat mass was approximately twofold higher in fasted GR^ΔAdip mice compared with controls, whereas their lean mass was decreased (Fig. 7A). Accordingly, GR^ΔAdip WAT compartments were enlarged, whereas their gastrocnemius, heart, and liver weights were reduced (Supplementary Fig. 5B). Consistent with their preserved fat mass, GR^ΔAdip mice presented increased adipocyte sizes, lower plasma NEFA and glycerol concentrations, decreased hepatic TG accumulation, and reduced circulating TG levels (Fig. 7B and C, Supplementary Fig. 1B, and Supplementary Fig. 5C and D). Blood β-ketones were substantially lower (Fig. 7C), indicating reduced FAO and ketogenesis. Hepatic PPARα and FGF21 are key mediators of fasted states and contribute to lipid utilization by increasing FAO in liver (32). Fasting-induced upregulation of Ppara and Fgf21 mRNA expression was impaired in GR^ΔAdip livers, accompanied by decreased plasma FGF21 levels (Fig. 7D and Supplementary Fig. 5E). In contrast to shorter fasting periods, 48-h fasting blood glucose was unexpectedly increased in GR^ΔAdip mice compared with controls (Fig. 7C), suggesting that the observed breakdown of lean mass might provide substrates for glucose production. In support of this assumption, circulating corticosterone, which potently induces protein catabolism and HGP (1), increased to a higher extent in 48 h– but not in 16 h–fasted GR^ΔAdip mice (Fig. 7E). Additionally, fasting-induced upregulation of gluconeogenic genes Pgc1a and Pck1 was similar in livers of both genotypes (Supplementary Fig. 5F). To characterize the fasting metabolome of GR^ΔAdip mice in more detail, plasma samples were analyzed by targeted LC-MS. From 177 metabolites, 42% were significantly different in 48 h–fasted compared with fed GR^∆Adip mice (Fig. 7F and Supplementary Table 3). Strong increases in AAs and related metabolites, several FA species, and nucleotides were observed. Conversely, in fasted controls, abundance of most AAs and nucleotides was either unchanged or lower than in fed controls. Eighteen percent of metabolites were significantly different between 48 h–fasted GR^∆Adip mice and controls. These included hexose, pentose, nucleotides, and AAs, all of which, in contrast to the fed state, displayed a greater quantity in GR^ΔAdip mice (Fig. 1C, Fig. 7F, and Supplementary Table 3). This reveals that the GR in adipocytes is indispensable for normal substrate mobilization and energy metabolism under prolonged fasting conditions.

To determine the molecular underpinning of impaired lipolysis, we compared mRNA and protein expression of lipolytic key factors in eWAT of fed and 48 h–fasted mice. HSL (Lipe) mRNA levels were similar among genotypes, whereas upregulation of ATGL (Pnpla2) and CGI58 (comparative gene identification 58; Abhd5) mRNA expression was significantly reduced in fasted GR^∆Adip eWAT (Fig. 8A). The protein level of ATGL and its coactivator CGI58 displayed a similar expression pattern in eWAT of fasted mice (Fig. 8B). Fasting-induced lipolysis requires cAMP-mediated activation of protein kinase A (PKA) with subsequent phosphorylation of HSL and perilipin, thereby indirectly activating ATGL via release of CGI58 (33). Upon fasting, total PKA substrate phosphorylation including HSL (Ser563/660) and perilipin was substantially decreased in GR^ΔAdip eWAT (Fig. 8B and C). Thus, we analyzed mRNA expression of genes critical for β-adrenergic activation of PKA. Expression of the inhibitory α2-adrenergic receptor (α2-AR) and inhibitory G-protein α-subunit isoforms was similar between genotypes (Supplementary Fig. 6A), whereas upregulation of β2-AR (Adrb2) was reduced in GR^ΔAdip eWAT under fasting conditions. Conversely, β3-AR (Adrb3) mRNA levels were increased (Fig. 8D). Notably, fasting-induced upregulation of the stimulatory G-protein α-subunit (G_Sα; Gnas), which couples β-ARs to adenylate cyclase–mediated production of intracellular cAMP (34), was significantly diminished in GR-deficient eWAT (Fig. 8D and E). To functionally identify the GR-dependent signaling step in β-adrenergic signal transduction, we applied pharmacological agonists on eWAT explants from 16 h–fasted mice. GR deficiency reduced cAMP generation upon treatment with isoproterenol (nonselective β-adrenergic agonist) but not in response to forskolin (direct adenylate cyclase agonist) (Fig. 8F). NEFA release from GR^∆Adip explants was decreased in response to isoproterenol, CL-316,243 (β3-AR agonist), and formoterol (β2-AR agonist), whereas forskolin-stimulated lipolysis was similar between both genotypes (Fig. 8G). Similar results were obtained from isoproterenol- or forskolin-treated eWAT explants of fed GR^∆Adip mice (Supplementary Fig. 6B), albeit the defect in lipolysis was less pronounced than in 16 h–fasted mice. These data demonstrate that the GR is a pivotal permissive factor for β-adrenergic signal transduction to adenylate cyclase and concomitant activation of the lipolytic cascade in WAT.

Our study provides the first genetic evidence for the GR in adipocytes as a critical component in normal and pathophysiological energy metabolism. Initial plasma metabolomics revealed a marked global effect of GR deficiency on systemic metabolite abundance, including a reduction in various FA species and proteogenic/branched-chain AAs (valine, leucine, and isoleucine). Albeit steady-state analysis of plasma represents a static net-balance between metabolite production and consumption, the broad changes indicate that systemic fuel partitioning is partly controlled by GR in adipocytes. Along this line, the diminished gluconeogenic capacity of GR^∆Adip mice during ITT/PTT may be due to increased hepatic responsiveness to insulin but could also result from a decrease in substrate availability (6). GC exposure at high concentrations is thought to stimulate lipolysis (1,2). Our data from eWAT explants of 4 h–fasted mice show that the GR is a key factor for lipolysis already in the postabsorptive state. Suppression of WAT lipolysis correlates with hepatic insulin sensitivity and reduced HGP (34,35) and indirectly counter-regulates gluconeogenesis (6,30,36), as WAT-derived substrate fluxes provide glycerol as direct substrate, and NEFA oxidation promotes acetyl-CoA–mediated activation of pyruvate carboxylase (37). Therefore, it is conceivable that reduced lipolysis in GR^∆Adip mice contributes in a similar manner to the observed reduction in gluconeogenic capacity.

GC-stimulated lipolysis has been linked to transcriptional upregulation of ATGL and HSL (7,9,38), elevated adenylate cyclase activity (39), intracellular cAMP production, and HSL activation (7,8). Reduced lipolysis in GR^∆Adip mice largely resulted from a block in β-adrenergic activation of PKA and lipolytic downstream signaling (33). The sympathetic nervous system (SNS) regulates induction of WAT lipolysis, and PKA-mediated phosphorylation of HSL serves as surrogate marker for SNS outflow to WAT (40,41). In fasting states, systemic GR antagonism with RU486 decreased β-adrenergic signaling in WAT through suppression of angiopoietin-like 4 (Angptl4) (8). However, Angptl4 expression was not affected in GR^∆Adip eWAT (Supplementary Fig. 6C). GR deficiency decreased the expression of G_Sα, which mediates SNS-stimulated lipolysis by transducing signals from β-ARs to adenylate cyclase. Notably, adipocyte-specific deficiency in G_Sα blocks WAT lipolysis, mimicking the phenotype of the GR^∆Adip mice (34). In agreement, cAMP generation and NEFA release from GR^∆Adip eWAT explants was reduced in response to β-AR agonists but could be normalized by direct adenylate cyclase agonism. GCs were shown to positively regulate G_Sα expression in rat brain (42), suggesting that a similar regulatory mechanism might exist in adipocytes to trigger β-adrenergic lipolytic responses. Accordingly, dexamethasone treatment increased G_Sα levels in eWAT of control mice but not in liver or GN muscle, suggesting that the GC-mediated induction of G_Sα occurs in a tissue-selective manner (Supplementary Fig. 6D). However, sequence analysis of the corresponding Gnas promoter region (NM_201616.2, NM_001077510.4, NM_001310083.1; −3,000 bp of TSS) did not reveal the presence of any sequences likely to act as GC response element. Notably, the GR positively regulates transcription through several mechanisms, including DNA-independent means, by acting as cofactor for lineage-specific transcription factors (43). At this point, however, we cannot predict which molecular mechanism might account for the reduction in Gnas expression in GR-deficient WAT, and further molecular studies will be necessary to decipher the underlying causes. Nevertheless, our current data support a model where the GR is required for signal transduction from β2/3-AR to adenylate cyclase and concomitant induction of lipolysis.

In mice and humans, an increase in GC regeneration by the enzyme 11β-HSD1 within WAT compartments suggests a role of the GC-GR axis in common obesity (2,12–16). Our findings from aged and HFD-fed GR^∆Adip mice support that suppressing the adipocyte GC-GR axis restricts obesity and its underlying disease state. We observed no differences in EE and/or food consumption that would explain the attenuated obesity of GR^∆Adip mice. Recent studies established an interdependence of lipid mobilization and storage in WAT. Chronic stimulation of lipolysis was shown to coincide with an upregulation of genes critical for FA storage and induction of de novo lipogenesis in WAT (44). Conversely, blocking lipolysis in chow- and/or HFD-fed mice resulted in downregulation of genes critical for FA storage and in decreased lipogenesis/lipid synthesis in WAT (34,35). Although not functionally proven by our data, the mild, yet collective reduction in several genes related to FA storage in HFD GR^∆Adip WAT suggests that a decrease in lipid synthesis might contribute to the attenuated obesity phenotype. However, further studies are required to determine the exact mechanism(s) underlying the antiobesity effects of adipocyte GR deficiency.

Dysfunctional hypertrophic adipocytes are thought to be in part causative for metabolic dysfunctions upon progressive adiposity (5,45). Improved glucose tolerance and reduced hyperinsulinemia indicate enhanced peripheral responsiveness to insulin in HFD-fed GR^∆Adip mice. Indeed, increased insulin-stimulated AKT phosphorylation was not confined to WAT but also present in muscle and liver of HFD-fed GR^∆Adip mice. As NEFA/lipid accumulation can cause deleterious effects on insulin-sensitive organs (6), and inhibition of lipolysis improves systemic glucose metabolism (34,35), the attenuated adiposity and reduced lipolytic capacity are conceivable possibilities for the ameliorated glucose tolerance in GR^∆Adip mice. Along this line, adipocyte GR deficiency attenuated HFD-induced steatosis. HFD feeding promotes hepatic lipid accumulation partly through upregulation of genes involved in FA uptake and lipid synthesis (46–48), several of which were downregulated in the livers of HFD-fed GR^∆Adip mice. Accordingly, reduced FA influx and endogenous lipid production likely contribute to the attenuation of steatosis in HFD-fed GR^∆Adip mice. Similar to our observations, adipocyte-specific 11β-HSD1 deletion was shown to ameliorate hypercortisolism-induced glucose intolerance and hepatic steatosis (38). Along with the beneficial metabolic state of the aged GR-deficient cohort, these findings indicate that adipocyte GC-GR activity is involved in the development of systemic metabolic dysfunctions.

In contrast to HFD feeding, adipocyte GR deficiency had adverse consequences when FA became the major energy substrate during prolonged fasting. Fasting induces robust shifts in fuel selection, during which coordinated increases in lipolysis, FAO, and ketogenesis spare glucose and preserve lean mass. Similar to models with impairments in lipolytic enzymes (33,49–51), GR deficiency disrupted the transition to lipid-based energy production (i.e., preserved fat, inefficient FA utilization, and ketogenesis). The simultaneous impaired upregulation of Ppara and Fgf21 in 48 h–fasted GR^∆Adip livers is consistent with a proposed necessity of lipolysis for the expression of both genes (51). FGF21 was shown to be required to sustain fasting hypoglycemia by stimulating corticotropin-releasing hormone in hypothalamic neurons and concomitant release of adrenal corticosterone (52). The increased blood glucose levels in 48 h–fasted GR^∆Adip mice combined with elevated corticosterone indicate a uniquely different fasting response upon complete FGF21 deficiency versus its scarcity due to defective lipolysis. GCs stimulate lipolysis, lean mass catabolism, AA utilization, and gluconeogenesis (1,4,53). Accordingly, fasting metabolism upon muscle-specific GR deficiency is opposite GR^∆Adip mice in that muscle mass is preserved and WAT depleted (3). Conversely, in fasted GR^∆Adip mice, all tissues except WAT respond to GC, visible as increased circulating AAs (proteogenic and urea cycle related), their metabolites (i.e., 2-ketobutyrate, creatine, creatinine, and carnitine), and glucose (4,53). As gluconeogenesis is intimately connected to the tricarboxylic acid cycle by both substrate supply and energy demand (6,29), lean mass–derived AAs conceivably provided both sources in 48 h–fasted GR^∆Adip mice. The elevation of purine nucleotide precursor and degradation products (i.e., xanthosine, hypoxanthine, xanthine, and urate) additionally suggest that the energetic requirements of the disturbed fasting metabolism in GR^∆Adip mice result in imbalanced ATP synthesis/degradation rates and, thus, a reduction in energy state. Thereby, our results show adipocyte GR as a critical regulator of energy homeostasis during prolonged fasting.

In conclusion, our results demonstrate that the adipocyte GR controls systemic fuel partitioning and energy metabolism. In prolonged fasting states, GR activity is vital to prevent aberrant fuel selection and lean mass wasting by permitting β-adrenergic stimulation of lipolysis and the switch to lipid-based energy production. Conversely, in models of diet- and aging-induced obesity, GR activity is a determinant of systemic metabolic dysfunctions.

Acknowledgments. The authors thank Ute Burret (Institute for Comparative Molecular Endocrinology, University of Ulm) and Safia Zahma (Institute of Animal Breeding and Genetics, University of Veterinary Medicine Vienna) for excellent technical assistance, and the animal facility staff at the University of Ulm, Medical University of Vienna, and the University of Veterinary Medicine Vienna for their support.

Funding. This work was supported by the Austrian Science Fund (FWF P26766 to T.S. and SFB F2807-B20 and SFB F4707-B20 to R.M.), the Deutsche Forschungsgemeinschaft (grants Priority Program Immunobone 1468/Tu 220/6-1, 6-2, Collaborative Research Centre 1149/C02/INST 40/492-1, and DFG-ANR/TU 220/13-1), and the European Union Research and Innovation funding program (FP7 BRAINAGE) to J.P.T. Part of this work was funded by the Österreichische Forschungsförderungsgesellschaft (COMET Program/K1 COMET Competence Center CBmed, Land Steiermark and Land Wien).

The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.

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

Author Contributions. K.M.M. and K.H. designed and coordinated the study, performed most experiments, supervised experiments, analyzed and interpreted data, and wrote the manuscript. D.K., S.V., M.B., L.M., and S.A. provided major technical support. S.J., K.F., T.D.M., and M.H.T. performed metabolic cage experiments and data analyses and interpretation. H.E. assisted in data interpretation and contributed to discussion and experimental design. C.M. performed LC-MS metabolomics and data analyses. J.H. conducted histopathological liver analyses, assisted in data interpretation, and contributed to discussion and experimental design. T.S. assisted in data interpretation, contributed to discussion and experimental design, and contributed to the writing of the manuscript. N.B. performed LC-MS metabolomics and data analyses, assisted in data interpretation, contributed to discussion and experimental design, and contributed to the writing of the manuscript. J.P.T. and R.M. designed and coordinated the study and wrote the manuscript. J.P.T. and R.M. are the guarantors of this work and, as such, had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.