Acute Deletion of the Glucocorticoid Receptor in Hepatocytes Disrupts Postprandial Lipid Metabolism in Male Mice

Abstract

Hepatic lipid metabolism is highly dynamic, and disruption of several circadian transcriptional regulators results in hepatic steatosis. This includes genetic disruption of the glucocorticoid receptor (GR) as the liver develops. To address the functional role of GR in the adult liver, we used an acute hepatocyte-specific GR knockout model to study temporal hepatic lipid metabolism governed by GR at several preprandial and postprandial circadian timepoints. Lipidomics analysis revealed significant temporal lipid metabolism, where GR disruption results in impaired regulation of specific triglycerides, nonesterified fatty acids, and sphingolipids. This correlates with increased number and size of lipid droplets and mildly reduced mitochondrial respiration, most noticeably in the postprandial phase. Proteomics and transcriptomics analyses suggest that dysregulated lipid metabolism originates from pronounced induced expression of enzymes involved in fatty acid synthesis, β-oxidation, and sphingolipid metabolism. Integration of GR cistromic data suggests that induced gene expression is a result of regulatory actions secondary to direct GR effects on gene transcription.

glucocorticoid receptor, lipid metabolism, gene expression, liver

Dynamic hepatic lipid metabolism involves, but is not limited to, breakdown of numerous lipid species for energy production, lipid export as lipoproteins and synthesis of fatty acids (FAs) and triacylglycerol (TAG) for storage within lipid droplets (1-4). In addition, lipid metabolism ensures homeostasis of lipid species necessary for normal structure and function of the cells, including phospholipids and sphingolipids that constitute organelles and the cell membrane (5, 6). Since lipids are obtained from diet, either directly or indirectly through de novo lipogenesis, or mobilized from adipose tissue during periods of fasting, the nature of hepatic lipid metabolism is intrinsically temporal, with half of the hepatic lipidome showing marked diurnal oscillations, and TAGs being the main lipid class exhibiting temporal regulation (7-10). Specifically, long-chain polyunsaturated FA-enriched TAGs accumulate in the liver and are secreted in very low–density lipoproteins (VLDLs) in between meals in the active phase and during early fasting in the resting phase (10). As fasting is extended, the liver accumulates even longer and more unsaturated TAGs that are stored preferentially in lipid droplets (10-12). Upon food intake, the fast-feeding transition exhibits increased synthesis of TAGs rich in long-chain saturated and monounsaturated FAs from de novo lipogenesis in response to increased glucose uptake (10). Finally, in the late postprandial phase, membrane lipids are synthesized leading to an increase in phospholipid species (13).

Under normal physiological conditions, glucocorticoid secretion is synchronized to the circadian rhythm and peaks at the transition between the resting/fasted phase and the active/fed phase (14-16). The metabolic function of temporal glucocorticoid receptor (GR) signaling in hepatocytes is attributed to the preprandial transcriptional control of genes involved in a range of metabolic process. This includes gluconeogenesis and metabolism of amino acids in the fasted state and early hepatic postprandial glucose uptake as well as bile acid synthesis and urea metabolism in hepatocytes (16-20). Moreover, several lines of evidence suggest that genetic GR disruption in hepatocytes leads to dysregulated serum TAG levels resulting in accumulation of TAGs in the liver (21, 22). These studies were based on CRE-loxP–mediated knockout (KO) of GR expression in hepatocytes, using CRE expression controlled by promoters and/or enhancers from the albumin and/or alpha-fetoprotein genes (albCRE or alfpCRE), both active during liver development (23, 24). Thus, developmental defects could explain the steatotic phenotype. To address this, we have used an acute hepatocyte-specific GR KO (hepGRKO) approach (16) and evaluated serum and liver TAG levels and quantified general lipid accumulation in the liver, which partly confirmed previous findings. Interestingly, general lipid accumulation was specifically augmented in the fed condition, and lipidomics analysis showed dysregulated postprandial levels of specific TAG species rich in very long–chain unsaturated FAs, normally abundant in the fasted condition. Also, we found increased hepatic levels of very long–chain nonesterified FAs (NEFAs) in the fed condition and generally elevated levels of specific ceramides (Cers) and sphingomyelins (SMs) in parallel with reduced mitochondrial respiration. Transcriptomic and proteomic analyses showed increased expression of several enzymes involved in FA synthesis, β-oxidation, and sphingolipid synthesis, which likely represents indirect effects of GR disruption in hepatocytes.

Materials and Methods

Animal Experiments

FVB/N-Nr3c1^fl/fl mice supplied by Professor Dr. Jan Tuckermann, Institute of Comparative Molecular Endocrinology, Ulm, Germany, containing loxP sites flanking exon 3 (Nr3c1m2Gsc) (25), were backcrossed with C57BL/6N mice and intercrossed. All mice were housed in groups of 2 to 4 littermates in ventilated cages with a controlled environment and a set temperature of ∼23 °C. The light and dark periods alternated with a duration of 12 hours each. The mice were fed regular chow (Altromin cat: 1324) and water was available ad libitum. All experimental procedures were approved by the Danish Animal Experiments Inspectorate (License no. 2019-15-0201-00384) in agreement with local guidelines and ethics considerations. During the 4 days preceding the day of sacrifice, all mice were single-caged and entrained to “night-restricted feeding” (NRF) to simulate their natural feeding behavior when in the wild. Throughout this period, the mice were fed at ZT12 (∼6 Pm) and their food was taken away at ZT0 (∼6 Am). To test the effects of GR in the liver, 10¹¹ genome copies of an adeno-associated virus (AAV) expressing either green fluorescent protein (GFP, Penn Vector Core, cat: AAV8.TBG.PI.eGFP.WPRE.bGH, purchased from Addgene) as a wildtype (GR^fl/fl) control or Cre recombinase (Penn Vector Core, cat: AAV8.TBG.PI.Cre.rBG, purchased from Addgene) were delivered via tail vein injection to acutely induce hepatic KO of GR. GR-floxed male mice were injected with AAV-GFP (GR^fl/fl group) or AAV-Cre (hepGRKO group) at 7-10 weeks of age. Two weeks later, the mice were trained to the NRF setup described above and sacrificed by cervical dislocation at 9-12 weeks old at preprandial timepoints ZT10 and ZT12 and postprandially (after being fed at ZT12) at ZT14, ZT18, and ZT22. Each experimental group was composed of 5 or 6 mice. The liver was perfused with cold sterile phosphate-buffered saline via either the inferior vena cava or the portal vein (depending on the accessibility), dissected, and snap-frozen in liquid nitrogen. Blood glucose was measured using a glucometer (FreeStyle Freedom Lite) and blood was collected and allowed to clot at 4 °C to obtain serum by centrifugation for 10 minutes at 10 000 rcf (4 °C). Liver tissue and serum were stored at −80 °C preceding further analysis.

Primary Hepatocyte Isolation

GR-floxed male mice (C57BL/6, Nr3c1^fl/fl) were injected with AAV-GFP (GR^fl/fl group) or AAV-Cre (hepGRKO group) via the tail vein, at 6-9 weeks of age. The mice were kept with free access to food up to sacrifice by cervical dislocation 2 weeks after injection with AAV. The liver was perfused via either the inferior vena cava or the portal vein (depending on accessibility), using a peristaltic pump, with approximately 80 mL perfusion buffer (1 × Hanks’ balanced salt solution + 0.5 mM ethylene glycol-bis-(β-aminoethyl ether)-N,N,N′,N′-tetraacetic acid [EGTA]) at 37 °C with a flow rate of 1 to 2 mL/minutes (6 rpm), which was then gradually increased to 7 mL/minutes (35 rpm), and then with digestion buffer (DMEM low glucose containing 15 mM HEPES, 4.9 mM CaCl₂ and 35.6 µg/mL liberase) for approximately 7 minutes. The digested liver tissue was collected in a Petri dish with 20 mL of digestion buffer (Dulbecco’s modified Eagle’s medium [DMEM] low glucose, Gibco cat. no. 11880-028, containing 15 mM HEPES, 4.9 mM CaCl₂, and 35.6 µg/mL liberase, Roche cat. no. 5401127001) and stripped off the Glisson capsule so the cells would diffuse into the solution. The next steps were performed in a sterile laminated air flow bench. The cell suspension was sequentially filtered through a prewet 400-µm mesh and a prewet 70-µm cell strainer and collected in a 50-mL Falcon tube. The cells were spun down for 2 minutes at 50 rcf, 4 °C. The pellet was resuspended in 20 to 25 mL of cold washing medium (DMEM low glucose with 10% fetal bovine serum and 1% Penicillin-Streptomycin). The washing step was repeated for a total of 2 washes, each followed by centrifugation at 50 rcf, 4 °C, for 2 minutes. The final pellet was resuspended in 10 to 20 mL warm plating medium (same as washing medium), and viable cells (suspension with >85% live cells) were plated in collagen-coated plates at the appropriate dilution. After 1 hour, the attached cells were washed with 1 × phosphate-buffered saline, and the medium was changed to maintenance medium (DMEM low glucose with 1% P/S). The cultured primary hepatocytes were left overnight before onset of experiments.

High-Resolution Respirometry of Liver Tissue

GR-floxed male mice (C57BL/6, Nr3c1^fl/fl) were injected with AAV-GFP (GR^fl/fl group) or AAV-Cre (hepGRKO group) via the tail vein at 8-10 weeks of age. After 2 weeks, the mice were trained to the standard NRF setup and sacrificed by cervical dislocation at the postprandial timepoint ZT13. Each experimental group was composed of 8 mice. A portion of fresh liver (20-50 mg) was collected and kept in BIOPS buffer (10 mM Ca-EGTA buffer, 0.1 µM free calcium, 20 mM imidazole, 20 mM taurine, 50 mM K-MES, 0.5 mM DTT, 6.56 mM MgCl2, 5.77 mM ATP, 15 mM phosphocreatine, pH 7.1) until oxygraphy analysis. The liver tissue was transported on wet ice in BIOPS buffer (26), and then dissected carefully in small tissue blocks on ice in BIOPS, before the tissue was washed for 10 minutes in mitochondrial respiration buffer (MiR05) while kept on ice (26). After the 10-minute wash the tissue was weighed and 2 to 3 mg was transferred to the high-resolution respirometer (Oxygraph-2k, Oroboros Instruments, Innsbruck, Austria). The procedure is described in detail by Lund and colleagues (27). The following protocol was applied: Malate (2 mM), glutamate (10 mM), and pyruvate (5 mM) were added to assess state 2 leak respiration; this was followed by adenosine monophosphate (5 mM) and magnesium (3 mM) to measure state 3 respiration with complex I–linked substrates. Cytochrome c (0.01 mM) was added to check the integrity of the outer mitochondrial membrane. Maximal mitochondrial respiratory capacity was measured by adding succinate (10 mM). Finally, maximal electron transfer chain capacity was tested by stepwise adding carbonylcyanide p-trifluoromethoxyphenylhydrazone in 0.5 µM steps. The concentrations reported here are final concentrations in the respiration chambers. OXPHOS capacity is expressed as picomoles per second per microgram of wet weight tissue.

Serum Lipoprotein and NEFA Measurements

Lipoproteins were separated from individual mice by sequential ultracentrifugation (28) using 60 μL of plasma. In brief, 60 μL of phosphate-buffered saline was placed in a thick wall ultracentrifugation tube (Beckman #343621) and 60 μL of plasma was placed below with a Hamilton syringe. Then, centrifugation was performed using a 42Tl rotor in a Beckman Optima L90k ultracentrifuge for 2.5 hours at 38 000 rcf. Then, the top fraction containing VLDL was harvested. The remaining high-density lipoprotein (HDL)/low-density lipoprotein (LDL)–containing fraction was mixed with a density solution (1.12 g/mL) and placed below a density solution with 1.019 g/mL. A second centrifugation step was performed at 38 000 rcf at 4 °C for 6 hours. The top fraction containing LDL and the lower fraction containing HDL were harvested. Levels of TAG and cholesterol were determined in the different lipoprotein fractions using colorimetric commercial kits (DiaSys) adapted to microtiter plates.

NEFAs were measured in vitro by enzymatic colorimetric assay using 3.5 to 7 μL of serum from preprandial (ZT12) or postprandial mice (ZT14 and ZT18). The assay was essentially carried out according to the manufacturer's instructions, using 7 μL of oleic acid standard (#270-7700, Fujifilm Wako Chemicals) ranging between 25 and 800 μM, 150 μL of R1 reagent (#434-91795, Fujifilm Wako Chemicals), and 75 μL of R2 reagent (#436-91995, Wako, Fujifilm Wako Chemicals). Blank and initial (660 and 546 nm) absorbance values, measured using a CLARIOStar Plate Reader, were subtracted from final end point (660 and 546 nm absorbance) values and concentration of NEFAs in serum samples from mice were calculated based on the standard curve.

Measurement of β-Hydroxy-Acyl-CoA-Dehydrogenase and Citrate Synthase Activities

Citrate synthase and 3-hydroxyacyl-CoA dehydrogenase (HADH) activities were measured as previously described (29), with minor changes to optimize the analysis in mouse liver. Approximately 15 to 16 mg of wet weight liver tissue was homogenized in 800 µL of buffer containing K₂HPO₄ (0.3 M), 0.05% bovine serum albumin (BSA), pH 7.7, for 2 minutes on a Tissuelyzer (Qiagen, Venlo, Limburg, The Netherlands). Ten percent triton was added to a final concentration of 0.1% triton, and the samples were left on ice for 15 minutes before they were stored at −80 °C for later analysis. The enzyme activities are expressed as micromoles substrate per minute per gram of wet weight tissue.

Coherent Anti-Stokes Raman Scattering Microscopy

Frozen liver pieces were embedded in optimal cutting temperature compound, frozen over methylbutane cooled by liquid nitrogen, and stored at −80 °C. Slides (2 or 3 per sample) containing multiple sections were obtained by cutting the mounted tissue in 20-μm sections using a cryostat (Thermo Scientific Shandon Cryotome FSE Cryostat Microtome, cryobar at −50 °C, chamber at −15 °C, specimen at −7 to −9 °C). Prepared slides were stored at −80 °C and imaged as described below (2 images per section, 4 sections per slide, 3 slides per liver, 2 mice per condition). Coherent anti-Stokes Raman scattering (CARS) microscopy was employed for a label-free visualization of lipid droplets on a microscopic scale (30, 31). A Leica SP8 confocal microscope (Leica Microsystems, Germany) set up for CARS imaging was used to characterize the populations of lipid droplets in the liver sections. The pump laser (PicoEmerald, APE, Germany) wavelength was set at 816.4 nm and the Stokes laser at 1064 nm to generate a CARS signal from the C-H vibrations in the lipid-filled droplets at 662.3 nm (ṽ_vib of 2850.4 cm⁻¹) which was collected in the forward direction through a 545 to 755 nm bandpass filter using a nondescanned photomultiplier tube (PMT) detector. Lipid droplets were captured in image stacks of 5 µm thickness with a z step size of 1 µm. The field of view was 96.9 × 96.9 μm² (1024 × 1024 pixels) and the pixel size was 94.7 × 94.7 nm². Image analysis was performed using General Analysis 3 in the NIS Elements software (ver. 5.30.02, Nikon Laboratory Imaging, 2020). Max intensity projections of the 5-µm stacks were generated to provide a higher lipid droplet signal to background and a better representation and quantification of the lipid droplets. Segmentation via thresholding was improved via preprocessing the image projections using a median filter with 2-px radius and background subtraction performed via the rolling ball algorithm using a 3.8-µm radius—more than twice the size of the largest droplets. Binary objects with a size below the resolution limit (0.2 µm) were discarded (resolution of the microscope was set at 180 nm) to avoid artifacts. The number of lipid droplets per mm² and the area fraction of lipid droplets (total area of droplets/frame area) was measured for each image projection, whereas the individual droplet area and circle-equivalent diameter was measured for each lipid droplet. Size distribution graphs were constructed in GraphPad Prism, ver. 9.2.0.

Triacylglycerol Measurement

Frozen liver weighing approximately 35 mg was homogenized in 500 μL of isopropanol using a FastPrep-24TM 5G homogenizer (MP Biomedicals) in the preset mouse liver program. The resulting mixture was centrifuged (10 000 rcf, 2 minutes), and the supernatant was loaded into a NUNC 96-well white optic plate with transparent bottom (10 μL of each sample). Glycerol standards ranging from 25 to 2000 μg/mL were also loaded into the plate (1 μL per standard) and 100 μL of glycerol reagent (Sigma cat no. F6428) was added to all standards and samples. After 5 minutes, absorbance was measured at 540 nm on the CLARIOstar Plus Multimode Microplate Reader (BMG Labtech). Then 25 μL of TAG reagent was added, followed by an incubation at 37 °C for 30 minutes, after which absorbance at 540 nm was measured on the CLARIOstar. Determination of total TAGs was done by subtracting the calculated amount of free glycerol from the total amount of glycerol obtained after enzymatic hydrolysis of TAGs upon adding the TAG reagent.

Quantification of Genomic DNA

Genomic DNA isolation protocol was performed as previously described (32). In short, frozen liver weighing 10 to 30 mg was incubated overnight with 600 μL of lysis buffer (100 mM NaCl, 10 mM EDTA, 0.5% sodium dodecyl sulfate [SDS], 20 mM Tris-HCl pH 7.4) and 6 μL of proteinase K (20 mg/mL), with shaking at 1400 rpm at room temperature (RT). The samples were then added to RNase A (100 μg/mL) and incubated at 37 °C for 30 minutes. After adding 250 μL of ammonium acetate (7.5 M) and 600 μL of isopropanol (0.7 v/v) and centrifuging at 15 000 rcf for 10 minutes at 4 °C, the supernatant was removed and the DNA pellet was washed with 500 μL of 70% ethanol and resuspended in 100 μL of TE buffer (10 mM Tris-HCl pH 7.4, 1 mM EDTA) once dry. DNA concentration was measured on a 2-μL sample on the CLARIOstar Plus Multimode Microplate Reader (BMG Labtech), and the samples were diluted in milliQ water to a final concentration of 10 ng/μL for quantitative polymerase chain reaction (qPCR) probing. Real-time qPCR analysis consisted of a master mix containing 0.5 μL of milliQ-H₂O, 0.5 μL of the 5′ primer (10 pmol/μL), 0.5 μL of the 3′ primer (10 pmol/μL), and 3.5 μL of SYBR Green Mastermix (FastStart Essential DNA Green Master, Roche cat. no. 06402712001) per reaction, 5 μL of which and 2 μL of the cDNA samples were loaded into each well in a 384-well skirted qPCR plate. The loaded plate was sealed, spun down for 2 minutes at 1200 rpm (RT) and run in a LightCycler480 qPCR machine under a program with the following steps: 2 minutes at 95 °C, 45 cycles of 95 °C/10 seconds, 60 °C/15 seconds and 72 °C/15 seconds, and heating from 60 to 90 °C for 10 minutes for melting curve analysis. The following primers were used: 16S rRNA mtDNA: forward, CCG CAA GGG AAA GAT GAA AGA C; reverse, TCG TTT GGT TTC GGG GTT TC; Nd1 mtDNA: forward, CTA GCA GAA ACA AAC CGG GC; reverse, CCG GCT GCG TAT TCT ACG TT; NDUFV1 nDNA: forward, TTC TGC CCC AAT CCC TCA TG; reverse, CCG GTC TTC ATC CTT CAG TGA.

Lipid Metabolism Tracing Setup

The cultured primary hepatocytes were stimulated with 500 nM dexamethasone for 1 hour and subsequently treated with 200 mM BSA-conjugated ¹³C₁₈-oleic acid and 10 nM insulin for 1 or 2 hours. The cells were then gently washed twice with 10 mM ammonium acetate and harvested in 200 µL of mass spectrometry (MS)–grade water, for lipid extraction and lipidomics analysis.

Lipid Extraction From Liver and Hepatocytes

Frozen liver pieces from GR^fl/fl and hepGRKO groups (preprandial timepoints ZT10, ZT12, and postprandial, fed timepoints ZT14, ZT18, and ZT22) weighing approximately 50 mg were shaken at 1000 rpm, 4 °C for 30 minutes, with 1 mL of ice-cold 1:2 methanol/chloroform solvent containing SPLASH LIPIDOMIX MS standard internal standard mix (Merck KGaA, cat. no. 330707-1EA) and 200 µL of milliQ-H₂O. The samples were then homogenized by sonication for 8 cycles of 30 seconds ON/30 seconds OFF each, totaling 15 minutes, using a BioRuptor (Diagenode Diagnostics), and centrifuged for 10 minutes at 16 000 rcf (4 °C). Primary hepatocytes harvested in 200 µL of milliQ-H₂O were added the methanol/chloroform solvent containing standard, shaken, and centrifuged as described for the liver samples (no sonication). For all samples, both the metabolite-containing supernatant (aqueous, upper phase) and the lipid-enriched lower phase (organic phase) were saved in separate tubes, while the protein-containing pellet was discarded. The aqueous phase was re-extracted with 350 µL of 86:14:1 chloroform/methanol/water solvent, after shaking at 1000 rpm, 4 °C for 20 minutes, and centrifuging at 16 000 rcf, 4 °C for 10 minutes. The supernatant (aqueous phase) enriched in metabolites was collected in new tubes, dried in the SpeedVac (Heto Lab Equipment HetoVac VR-M with VR-I rotor and CT-IIO cooling trap) overnight and saved at −20 °C for further analysis. The lower, organic phase containing remaining lipids was added to the previous organic phase, and this lipid extract was saved at −20 °C overnight, dried under a stream of nitrogen (N₂) and stored at −20 °C until the day of analysis.

Lipidomics Analysis and Data Processing

Lipid samples were resuspended in 25 µL of eluent A (5:1:4 isopropanol/methanol/water with 5 mM ammonium acetate and 0.1% acetic acid), shaken for 20 minutes (max. speed), and centrifuged for 5 minutes (max. speed, RT). The supernatant was transferred to high-performance liquid chromatography vials, with 5 μL of each sample saved for a pooled QC sample. Six microliters was injected using an Agilent 1290 Ininity UPLC system (Agilent Technologies, Santa Clara, CA) and compounds separated on a Zorbax Eclipse Plus C18 guard/column (2.1 × 5/150 mm, 1.8 μm particle size, Agilent Technologies, Santa Clara, CA, USA) kept at 50 °C. The analytes were eluted using a flow rate of 400 μL/minutes and the following composition of eluent A (5:1:4 isopropanol/methanol/water with 5 mM ammonium acetate and 0.1% acetic acid) and eluent B (99:1 isopropanol/water with 5 mM ammonium acetate and 0.1% acetic acid): 0% B from 0 to 1 minutes, 0 to 25% B from 1 to 1.5 minutes, 25% to 95% B from 1.5 to 12 minutes, 95% B from 12 to 14 minutes, and 95% to 0% B from 14 to 15 minutes before equilibration for 3 minutes with the initial conditions. The flow from the UPLC was coupled to a 6530B quadrupole time of flight mass spectrometer (Agilent Technologies, Santa Clara, CA) for mass spectrometric analysis in both positive and negative ion mode using the following general settings for MS1 mode: scanning range from 50 to 1700 m/z, 2 scans/second, gas temp at 325 °C, drying gas at 8 L/minutes, nebulizer at 35 psig, sheath gas temp at 350 °C, sheath gas flow at 11 minutes/L, VCap at 3500 V, fragmentor at 125 V, and skimmer at 65 V. Each spectrum was internally calibrated during analysis using the signals of Hexakis ¹H,¹H,³H-tetrafluoropropoxy phosphazine, which was delivered to a second needle in the ion source by an isocratic pump running with a flow of 20 µL/minutes. Iterative data-dependent analysis MS2 mode fragmenting was applied with the following general settings, scanning range from 50 to 1700 m/z, 1 scans/second, collision energy at 40 V, precursor threshold at 5000 counts, and active exclusion after 2 spectra, which was released again after 0.5 minutes. Raw data were converted to .mzML format using ProteoWizard (v 3.0.21062) (33). Lipid species were annotated in MS-Dial (v. 4.70) (34) with 0.005/0.01 Da MS1/MS2 mass tolerance and identification score cut of at 70% as general setting. A list of the annotated compounds was exported to PCDL manager B.08.00 (Agilent Technologies, Santa Clara, CA) to generate a database, which was used in Profinder 10.0 (Agilent Technologies, Santa Clara, CA) where areas were extracted. Features were removed if their average signal were less than 5 times more abundant in the QC samples than the blanks (water extraction). The signals were normalized to the closest (based on retention time) internal standards before correction for signal drift using statTarget (35). Mass isotopologues were also extracted in the Profinder software. Analysis of lipidomics data was done using the MetaboAnalyst 5.0 webtool (36).

Protein Extraction and Western Blotting

Protein extracts obtained from primary hepatocytes harvested in 1 mL of lysis buffer or from frozen mouse liver samples homogenized with a FastPrep-24TM 5G bead beating grinder and lysis system (MP Biomedicals) in 1 mL of lysis buffer containing phosphoproteinase inhibitors (16) were treated with 2 µL of benzonase for 15 minutes, then centrifuged so that the protein-containing supernatant was stored at −80 °C. The bicinchoninic acid assay kit (Pierce BCA Protein Assay Kit, Thermo Fisher Scientific cat:23225) was used to determine protein concentrations. The manufacturer's guidelines were followed, with measurements performed in 10 µL of diluted samples and a BSA standard curve ranging from 0 to 4 mg/mL. The colorimetric determination was performed on 96-well Nunc MicroWell Optical-Bottom plates measured on the CLARIOstar Plus Multimode Microplate Reader (BMG Labtech).

SDS-polyacrylamide gel electrophoresis was run at maximum voltage (20 mA per gel) in Tris-Glycine-SDS running buffer (1× TGS) for 2 to 2.5 hours on samples containing 40 µg of protein, to which bromophenol blue, β-mercaptoethanol, and lysis buffer were added, and loaded onto polyacrylamide gels (10% separation gel, 5% stacking gel). Separated proteins were transferred to activated polyvinylidene fluoride membranes in transfer buffer (1 × TGS containing 20% ethanol) under a current of 400 mA (constant voltage of 100 V) applied for 1 hour. The membranes were incubated with AmidoBlack dye to check for protein bands and blocked in 5% BSA dissolved in 1× Tris-buffered saline (TBS)-T for 1 hour at RT. The membranes were washed 3 × in TBS and 3 × 10 minutes in TBS-T, and incubated overnight at 4 °C with the appropriate primary antibodies (in 1 × TBS-T + 5% BSA) and subsequently with the corresponding horseradish peroxidase–conjugated secondary antibody (in 1 × TBS + 5% BSA) for 1 hour at RT. After each incubation, we performed washes as described above. For detection of horseradish peroxidase, the membranes were incubated with the luminol-based enhanced chemiluminescent substrate SuperSignal West Pico PLUS Chemiluminescent Substrate (Thermo Fisher Scientific cat. no. 34580), for 5 minutes at RT. The resulting signals were captured by the Amersham Imager 680 (GE Life Sciences). Primary antibodies: anti-GR (sc-1004, Santa Cruz, RRID:AB_2155786), anti-β-tubulin (05-661, Merck Millipore, RRID:AB_309885).

Proteomics Analysis

For quantitative comparison of liver proteomes of hepGRKO and GR^fl/fl mice sacrificed at zeitgeber times 12, 14, and 18, approximately 40 mg of perfused liver tissue was split in 2 and homogenized in denaturing buffer (either 8 M guanidine hydrochloride in 25 mM ammonium bicarbonate or 5% SDS) by 3 to 10 seconds of sonication at 20% intensity, and heated to 95 °C for 5 minutes. Protein concentrations were measured using the Pierce BCA Protein Assay Kit (Thermo Fisher Scientific) and 10 μg of protein (5 μg from each lysate) was pooled together in a 1:1 volume of denaturing buffer before diluting 4 × with water for a final concentration of 1 M guanidine hydrochloride (and 0.625% SDS). Protein aggregation capture on microparticles (37) was accomplished using 70% (v/v) acetonitrile (ACN), 40 μg of magnetic HILIC beads (ReSyn Biosciences (Pty) Ltd) and 30 minutes of incubation with shaking at RT. The bead-bound protein aggregates were retained using a magnet while washing once with pure ACN and once with 70% (v/v) ethanol. Protein aggregates were resuspended in 100 μL of 50 mM ammonium bicarbonate, reduced with 2 mM DTT for 30 minutes, and alkylated with 11 mM chloracetamide for 30 minutes in the dark. After proteolytic digestion with 1:200 of LysC (Wako) for 1 hour at 37 °C, samples were digested with 1:100 of trypsin (Promega) overnight at 37 °C, before adding trifluoroacetic acid (final pH 2.0–2.5). Samples were loaded onto C18 StageTips for desalting and eluted using 60% ACN in 0.5% (v/v) acetic acid. Eluates were vacuum-dried to 2 to 3 μL in a speed-vac and finally rediluted to 10 μL using 0.5% (v/v) acetic acid. Approximately 800 ng of digested and purified protein extract was loaded onto an liquid chromatography MS/MS system consisting of an Orbitrap Exploris 480 mass spectrometer (Thermo Fisher Scientific) coupled with an EASY-nLC 1000 nanoflow liquid chromatograph (Thermo Fisher Scientific). For reverse-phase chromatography we used a 20-cm analytical column with an inner diameter of 75 μm of packed in-house with ReproSil Pur C18-AQ 1.9 μm resin (Dr Maisch GmbH) and an ACN/water solvent system containing 0.5% acetic acid at a flow rate of 0.25 μL/minutes. Peptide separation was achieved using a step gradient from 3% to 8% ACN for 7 minutes, followed by a slow gradient to 35% ACN for 86 minutes, a ramp to 45% ACN for 15 minutes, and a plateau of 95% ACN for 5 minutes. The Orbitrap Exploris 480 was operated in data-independent acquisition (DIA) mode by sequential window acquisition of all theoretical fragment ion spectra (38). Briefly, after a full MS survey scan from 350 to 1400 m/z at 120 000 resolution with a maximum ion injection time of 45 ms and a normalized automatic gain control target of 300%, precursor ions in the 360.5 to 1000.5 m/z range were fragmented by higher-collisional dissociation at a collision energy of 28% in 50 sequential scans using a sliding 13 m/z isolation window, a 1 m/z overlap between windows, a resolution of 30 000 and a maximum ion injection time of 54 ms. The acquired DIA raw files were analyzed in the Spectronaut 14.10 software (Biognosys) using default settings, single-hit protein filtering, and our in-house–generated spectral library (39). Finally, the report table containing quantitative values for identified proteins in each run was exported for subsequent data analysis.

Analysis of Enriched Reactome Pathways

Previously published RNA-seq and GR ChIP-seq data was used to annotate direct GR target genes in mouse liver tissue as previously described (16). This dataset was used together with the proteomics data described above to identify enriched Reactome pathways using clusterProfiler 4.6 (40).

Statistical Analysis

Temporal- and genotype-regulated lipids, identified by MS-based lipidomics, were identified using 1-way analysis of variance (ANOVA) with post hoc Fisher's least significant difference test. For temporal analysis, any 2 timepoints were compared within the GR^fl/fl condition. For genotype comparative analysis, GR^fl/fl samples were compared with hepGRKO samples for each timepoint. Differentially regulated proteins, identified by proteomics analysis, were computed using a robust linear mixed model framework using MSqRob (41) with multiple testing correction using the Benjamini–Hochberg method. Differentially expressed genes, identified by RNA-seq, were identified by DESeq2 as described previously (42). Differences in specific lipids, mRNAs, proteins, or respiration were evaluated by Student's t-test or 1- or 2-way ANOVA with indicated P values as stated in the figure legends. Threshold for statistical significance is set at P = .05 unless otherwise stated. Nonsignificance is indicated as ns. The P values for specific Tukey pairwise comparisons are indicated in each figure. Error bars in all barplots indicate standard deviation.

Results

Acute GR Disruption in Hepatocytes Leads to Hepatic Steatosis

To acutely disrupt GR expression in hepatocytes from Nr3c1^fl/fl mice (termed GR^fl/fl) we used AAVs to deliver the gene coding for Cre recombinase controlled by the hepatocyte-specific Tbg promoter. As a control, we used AAVs expressing GFP controlled by the same promoter. Ten days after intravenous injection with 10¹¹ genome copies of AAV per mouse, we subjected all mice to a NRF regimen for 4 days, whereafter mice were sacrificed at different preprandial and postprandial timepoints (Fig. 1A). The GRKO strategy (from here on termed hepGRKO) resulted in robust depletion of GR expression in mouse liver tissue (Fig. 1B). Residual expression is likely a result of GR expression in nonhepatocytes. Measurement of total liver TAG (Fig. 1C) showed most pronounced TAG accumulation in the late postprandial timepoint, which confirms previous studies reporting accumulation of TAGs in the liver when GR expression is blunted by albCRE (43). We did not observe any fasting-mediated effects on liver TAG levels, although previous murine and human data suggest that fasting leads to hepatic steatosis (11, 16). This may be explained by the different mouse models and the relatively short fasting period used in the NRF setup (12 hours during the resting period at lights on) compared with the steatosis observed from extended overnight fasting periods in rodents and fasting periods of 36 hours in humans. Hepatic steatosis in hepGRKO animals was not accompanied by increased serum TAGs in very VLDLs, LDLs, or HDLs (Fig. 1D-1F), in agreement with 1 other GRKO approach (albCRE) (43) but disagreeing with another (alfpCRE) (21). Moreover, serum NEFA levels were reduced in the early postprandial phase but unchanged in hepGRKO compared with controls (Fig. 1G), in agreement with other hepatocyte GR KO studies (21, 44). This indicates that lipolysis from adipose tissue is unaffected by hepGRKO in agreement with undisrupted levels of serum insulin and corticosterone levels reported previously using a similar experimental setup to evaluate acute effects of GRKO in hepatocytes (16).

Figure 1.

Acute GR disruption leads to hepatic steatosis. (A) Experimental setup for acute hepatocyte-specific knockout of GR. Male Nr3c1(GR)^fl/fl mice, approximately 8 weeks old, were tail vein injected with AAV-Tbg-Cre or AAV-Tbg-GFP (hepGRKO and GR^fl/fl, respectively), then sacrificed for serum and liver collection 2 weeks later at the indicated preprandial, early postprandial, and late postprandial time points. The mice were entrained to a night-restricted feeding (NRF) routine for 4 days prior to serum and liver collection. (B) GR protein expression in GR^fl/fl and hepGRKO livers evaluated by immunoblotting. (C) Total temporal triacylglycerol (TAG) levels in livers from GR^fl/fl and hepGRKO. Statistical testing by 2-way ANOVA with Tukey's multiple comparison test (timepoint: ns, genotype: P < .0001, interaction: ns). Tukey pairwise comparisons: **P < .01. (D-F) Temporal TAG levels in serum VLDL, LDL, and HDL from GR^fl/fl and hepGRKO animals. No statistical difference was observed between GR^fl/fl and hepGRKO. For main time-dependent effects on VLDL levels, P < .0001 (2-way ANOVA with Tukey's multiple comparison). Tukey pairwise comparisons: **P < .01, ****P < .0001. (G) Temporal NEFA levels in serum from GR^fl/fl and hepGRKO animals. No statistical difference was observed between GR^fl/fl and hepGRKO. For main time-dependent effects on NEFA levels, P < .0001 (2-way ANOVA with Tukey's multiple comparison). Tukey pairwise comparisons: **P < .01. (H) Representative CARS microscopy images of frozen liver sections from GR^fl/fl and hepGRKO collected at ZT12 and ZT18. Scale bar corresponds to 10 µm. (I) Area of imaged liver sections from GR^fl/fl and hepGRKO covered by detected lipid droplets per unit of frame area (n = 39-54 sections for each condition). Statistical testing by 2-way ANOVA with Tukey's multiple comparison test (timepoint: P < .0001, genotype: P < .0001, interaction: P < .0001). Tukey pairwise comparisons: ****P < .0001. Error bars in barplots indicate standard deviation.

To analyze hepatic steatosis more closely, we quantified total lipid droplet content in the livers of control and hepGRKO mice using CARS microscopy. Interestingly, total lipid droplet content (droplets >200 nm) was increased specifically in livers from hepGRKO at the late postprandial timepoint ZT18 compared with preprandial time point ZT12 (Fig. 1H nd 1I). We also observed increased lipid droplet size in hepGRKO livers (Fig. S1A and B (45)) irrespective of fasted/fed status, suggesting that postprandial lipid accumulation is a result of an increased number of larger lipid droplets. Accordingly, microvesicular steatosis has been reported previously (alfpCRE driver based) using oil red O staining against neutral lipids (44). Interestingly, total hepatic lipid droplet analysis by CARS also showed that lipid droplet content in the preprandial state is similar between GR^fl/fl and hepGRKO, yet lipid droplet size is larger in hepGRKO. This suggests that lipid droplets in GR^fl/fl livers are smaller and more frequent than larger and less frequent droplets in hepGRKO. In addition, although liver TAG levels were similar between all timepoints in hepGRKO we observed considerable difference in total lipid droplet content in the livers from preprandial vs postprandial hepGRKO. This could reflect that detected lipid droplets may contain different lipid species in addition to TAGs depending on the feeding status. Alternatively, TAGs may accumulate in the cytoplasm in droplets less than 200 nm in size (detection threshold used for the CARS analysis) which may explain discrepancy between total TAG levels and lipid droplet content. Collectively, we show that acute hepGRKO leads to increased hepatic TAG levels, without any effects on serum TAG and NEFA content, and that general hepatic lipid accumulation as lipid droplets is most pronounced in the postprandial phase of hepGRKO animals.

Dynamic Lipid Metabolism at the Transition Between Fasted and Fed States

To study the accumulated lipids in control and hepGRKO animals in more detail, we used a MS-based lipidomics approach at several preprandial and postprandial timepoints. Initially, we focused the analysis on the control animals to obtain insights into the temporal lipid turnover as animals transitioned from a fasted to a fed state. We identified a total of 408 lipid species distributed over 19 lipid classes, with 126 species showing significant temporal differences (Fig. 2A; Fig. S2A (45)). The 126 lipid species, showing temporal regulation, were distributed over 15 different lipid classes, most notably TAG, phosphatidylcholines (PCs), phosphatidylethanolamines (PEs), diacylglycerols (DAGs), SMs, and FAs (Fig. 2B), collectively representing 85% of all differentially regulated lipids. For these classes, a large fraction of the individual species showed temporal regulation (Fig. 2C; Fig. S2A (45)). Interestingly, some lipid species were pronouncedly enriched in the postprandial state (eg, long-chain C20-C32 saturated FAs, but not shorter chain C16-18 FAs and unsaturated FAs), whereas others were mostly enriched in the preprandial state (eg, CARs and SMs). In contrast, different lipid species within TAGs, DAGs, PEs, and PCs were either increased or decreased in response to feeding (Fig. 2C), depending on the specific features of their FA composition (Fig. 2D).

Figure 2.

Temporal hepatic lipid metabolism in GR^fl/fl mice. (A) Temporal regulation of all detected lipid species within 19 major lipid classes. Differentially regulated lipids at FDR < 0.05 (comparing 2 of any given timepoint) are indicated (126 lipid species in total). (B) Total amount of major lipid classes showing temporal regulation at FDR < 0.05. (C) Relative number of lipid species within each lipid class regulated by the fast-feeding transition. Percentage of lipid species mostly enriched in the preprandial and postprandial phase are indicated by light and dark blue, respectively. (D) Left: levels of relatively shorter chain (<52 carbons), more saturated (0-2 double bonds) FAs within TAGs, and levels of relatively longer chain (≥52 carbons), more unsaturated (≥3 double bonds) FAs within TAGs and their distribution between preprandial and postprandial enrichment. Right: levels of relative shorter chain (<36 carbons), more saturated (0-1 double bonds) FAs within PE/PCs, and levels of relative longer chain (≥36 carbons), more unsaturated (≥2 double bonds) FAs within PE/PCs and their distribution between pre- and postprandial enrichment. (E) Hierarchical clustering of temporal regulated lipids. Specific fatty acids (FAs), sphingomyelins (SMs), and triacylglycerols (TAGs) are indicated. (F) Levels of specific SM (32:1), FA (32:0), and TAGs (51:2 and 53:4). Statistical testing by 1-way ANOVA with Tukey's multiple comparison test, *P < .05, **P < .01, ***P < .001, ****P < .0001.

Clustering analysis of all differentially regulated lipid species showed a clear separation of individual lipids enriched in the preprandial or postprandial states (Fig. 2E). Notably, the levels of specific SMs, including SM 32:1, increased from ZT10 to ZT12 and subsequently decreased gradually in the postprandial phase (Fig. 2E and 2F). In contrast, several saturated long-chain FAs were most abundant toward the late postprandial timepoints, including FA 32:0 (Fig. 2E and 2F). Interestingly, the TAGs enriched in the fasted state primarily contain longer, more unsaturated fatty acyl chains, whereas the TAGs increased in the postprandial phase predominantly contain relatively shorter, more saturated fatty acyl chains (Fig. 2D and 2F; Fig. S2B (45)). For instance, TAG 51:2 was increased in the fed state while TAG 53:4 was decreased (Fig. 2F). A similar trend could also be observed for other abundant lipid classes such as PEs/PCs (Fig. 2D; Fig. S2C (45)). These results suggest that a major fraction of hepatic lipid species, belonging to specific classes, is temporally regulated when mice transition from a fasted to a fed state, including acute changes in the FA composition in lipid classes such as TAGs, PEs, and PCs. Importantly, this agrees with a recent comprehensive analysis of the circadian hepatic lipidome (10), which qualifies the experimental approach.

Acute hepGRKO Leads to Accumulation of Specific Lipid Species in the Postprandial State

To assess the role of GR in the temporal regulation of specific hepatic lipid species, we performed the same analysis as described above on lipids extracted from livers of hepGRKO mice sacrificed at timepoints ZT10, 12, 14, 18, and 22. Differential analysis between control and hepGRKO identified a total of 25 differentially regulated lipid species in at least 1 timepoint (Fig. 3A). Most of these lipids belong to TAGs, SMs, FAs, and Cers, and differential lipid levels were most pronounced in the postprandial condition (Fig. 3A and 3B). More than 15% of the detected TAGs, SMs, and FAs were dysregulated in hepGRKO compared with control (Fig. 3C). Interestingly, the levels of the TAGs and FAs were temporally regulated in the control animals, suggesting that the temporal profile of a subset of TAGs and FAs is controlled by circadian activity of GR (Fig. 3C and 3D). Moreover, 2 subsets of TAG species showed opposite patterns of dysregulation. Specifically, TAGs composed of relatively shorter fatty acyl chains were decreased in the fed state in hepGRKO, including TAG 51:1 and TAG 50:2 (Fig. 3D; Fig. S3A (45)). The levels of these TAGs increased in the fed state in the control animals, which was not observed in hepGRKO. These TAGs are likely composed of de novo synthesized FAs in combination with dietary FAs such as palmitic acid (16:0) and oleic acid (18:1) (10). This suggests that synthesis of TAGs from de novo synthesized FAs in the fed state or uptake of dietary FAs into hepatocytes is compromised in hepGRKO. Incubation of primary hepatocytes with oleic acid suggests that reduced TAGs containing relatively shorter fatty acyl chains arise from reduced FA uptake in hepGRKO hepatocytes (Fig. S3B (45)).

Figure 3.

Regulation of lipid temporal profiles by acute hepatocyte GR disruption.(A) Differentially regulated lipid classes in hepGRKO compared with GR^fl/fl in at least 1 timepoint (FDR < 0.05). (B) Number of differentially regulated lipids in the preprandial (fasted) and postprandial (fed) conditions. (C) Percentage of differentially regulated lipid species relative to the total detected lipids in each lipid class. (D) Hierarchical clustering of differentially regulated lipids in hepGRKO animals. (E) Temporal levels of specific FA (26:0), TAG (52:3), SMs (58:2 and 60:3), and Cers (44:4 and 58:3) in livers from GR^fl/fl and hepGRKO. Statistical testing by 2-way ANOVA with Tukey's multiple comparison test. FA (26:0), timepoint: P < .0001, genotype: P = .0072, interaction: P = .0019. TAG (52:3), timepoint: P = .0003, genotype: P < .0001, interaction: P = .0054. SM (58:2), timepoint: ns, genotype: P = .0003, interaction: ns. SM (60:3), timepoint: 0.0096, genotype: P < 0.0001, interaction: ns. Cer (44:4), timepoint: P = .0073, genotype: P < .0001, interaction: ns. Cer (58:3), timepoint: P < .0001, genotype: P < .0001, interaction: P = .0294. Tukey pairwise comparisons: *P < .05, **P < .01, ***P < .001, ****P < .0001.

In contrast, we found that TAGs with longer, more unsaturated chains, normally reduced when animals transition from fasted to fed conditions (Fig. 2D and 2F), accumulate in the postprandial phase in hepGRKO animals (Fig. 3D). This includes TAG 52:3 and TAG 52:5 (Fig. 3D and 3E; Fig. S3A (45)) and indicates that hepGRKO impairs turnover of these TAG species in the fed state. Alternatively, synthesis of TAGs with longer and more unsaturated chains may be increased in hepGRKO. Interestingly, we observed a significant increase in very long–chain FAs (26:0-32:0) at ZT18-22 (Fig. 3D and 3E), and increased levels of TAGs with very long fatty acyl chains may arise as a compensatory mechanism to incorporate these FAs into neutral lipids.

In addition to increased levels of specific TAGs and FAs we also observed constitutively increased levels of specific SM and Cer species. These include SM 58:2, SM 60:3, Cer 44:4, and Cer 58:3 (Fig. 3E). In contrast to TAGs and FAs, none of the differentially regulated SM and Cer species in hepGRKO were temporally regulated in the control animals (Fig. 3C and 3D). Additionally, these SMs and Cers are enriched for very long polyunsaturated fatty acyl chains (>C30, n > 2), contrasting with, for instance, temporally regulated SMs containing FAs with 16-20 carbons (Fig. 2D-2E). Like the postprandial accumulation of specific TAGs mentioned above, postprandial accumulation of SMs and Cers with very long polyunsaturated fatty acyl chains may arise from an increased pool of very long–chain FAs.

Hepatic GR Disruption Perturbs Expression of Genes Involved in Lipid Metabolism

To examine the underlying mechanisms for dysregulated lipid metabolism, we quantified differential gene expression in hepGRKO by transcriptomic analysis using RNAseq and proteomics analysis using label-free DIA-MS. Analysis of individual pre- and postprandial timepoints showed most pronounced dysregulated mRNA expression in hepGRKO at timepoints ZT12 and ZT14 (Fig. 4A; Fig. S4A (45)), coinciding with high serum levels of corticosteroids (16) which suggests direct genomic action of GR at the fasted-fed transition. A less striking temporal difference was observed for differentially expressed proteins (Fig. 4B; Fig. S4B (45)), yet most differentially expressed proteins were identified at ZT14, which agrees well with a time delay between direct GR-controlled mRNA expression followed by translation. Yet, we found little overlap of differentially expressed mRNAs and proteins from the same gene (Fig. 4C), indicating significant posttranscriptional regulation. In agreement, quantification of previously identified GR binding sites (16) in the vicinity of regulated genes suggests that reduced mRNA levels to a larger degree can be explained by direct genomic GR actions compared with overall reduced protein expression (Fig. 4D). Similarly, genes induced by hepGRKO also show relatively lower frequency of nearby GR binding sites, suggesting that increased gene expression to a larger extent originates from indirect effects of GR action through the transcriptional apparatus. Moreover, most differentially expressed proteins in the preprandial phase were also differentially expressed in the postprandial phase, in contrast to differentially expressed mRNAs (Fig. 4E). Collectively, this suggests significant indirect effects on gene expression in hepGRKO livers, which may contribute to the steatotic phenotype.

Figure 4.

Differentially expressed genes (DEGs) at mRNA and protein level in hepGRKO livers compared to GR^fl/fl controls. (A) DEGs (mRNA level) quantified by RNAseq at FDR < 0.05 and (B) DEGs (protein level) quantified by DIA-MS at FDR < 0.1. Bar plot at the top shows the number of DEGs at individually sampled timepoints. Pie chart at the bottom shows the number of DEGs aggregated into the preprandial and postprandial phases. (C) Percentage of differentially regulated proteins that are also differentially regulated at mRNA level. (D) DEGs harboring GR binding sites in the vicinity (100 kb) of the transcriptional start site. (E) Number of DEGs in the preprandial phase that overlap with DEGs in the postprandial phase. Top shows DEGs overlap at mRNA level and bottom shows DEGs overlap at protein level. (F) Enriched Reactome subpathways belonging to lipid metabolism (R-MMU-556833). Size of the circles indicates the number of genes. Nodes indicate the number of genes shared between connecting subpathways. Colors indicate direction of differential regulation at mRNA or protein level. Size of the circles shows the number of DEGs assigned to each pathway.

To annotate differentially expressed genes to putative biological function, we performed Reactome pathway analysis with a focus on pathways related to metabolism. This suggests major impact on a range of metabolic pathways including metabolism of lipids, glucose, amino acids, bile, nucleotides, and vitamins (Fig. S4C (45)), in agreement with previous studies (16-20, 43). Furthermore, the analysis suggests effects on the tricarboxylic acid cycle (TCA) and the electron transport chain (Fig. S4C (45)), particularly evident for proteins whose expression is decreased postprandially in hepGRKO livers. Specific analysis of Reactome subpathways of lipid metabolism (R-MMU-556833) revealed enrichment of interconnected pathways including FA oxidation, FA synthesis, bile acid synthesis, steroid metabolism, and sphingolipid synthesis (Fig. 4F; Fig. S4C (45)). Expression of proteins and mRNAs involved in FA metabolism (synthesis and β-oxidation) was generally increased in hepGRKO livers, in agreement with disrupted FA and TAG metabolism quantified by MS (Fig. 3). Additionally, we observed disrupted gene expression related to sphingolipid metabolism, in accordance with disrupted levels of specific sphingolipids (Fig. 3).

Dysregulated Fatty Acid Metabolic Pathways

To specifically analyze gene expression related to FA metabolism, we focused on all interconnected lipid metabolism pathways and found that most regulated genes assigned to FA metabolism are shared among the other enriched lipid metabolism pathways (Fig. S5A (45)). Thus, to simplify the analysis, we focused specifically on regulated genes belonging to the interconnected pathways and plotted their regulatory relationship (Fig. 5A). Interestingly, pathways involved in FA synthesis as well as β-oxidation (mitochondrial and peroxisomal) were enriched for hepGRKO-induced genes (Fig. 5A). This included a large proportion of genes involved in de novo acyl-CoA synthesis, elongation, and saturation (Fig. 5B-5D). A number of these enzymes were primarily differentially induced specifically in the preprandial phase. In addition, we observed preprandial and postprandial reduced expression of pyruvate dehydrogenase complex subunits PDHA and PDHB as well as several enzymes in the TCA cycle (Fig. 5C and 5D), suggesting that pyruvate metabolism through TCA is reduced, and acetyl-CoA is shunted towards FA synthesis. In parallel, expression of several proteins involved in FA β-oxidation were increased (Fig. 5E and 5F), suggesting continuous synthesis and β-oxidation of acyl-CoAs in hepGRKO livers. The accumulation of TAGs observed in hepGRKO indicates that acyl-CoA synthesis prevails oxidation. Moreover, expression of several ACOT (ACOT1, 3, 4 and 13) enzymes were increased, which may increase hydrolysis of acyl-CoAs at various chain lengths during β-oxidation (46). This could contribute to hepatic accumulation of FAs in the postprandial phase. Moreover, ACOT13 expression negatively impacts FA β-oxidation (47), which could drive equilibrium towards FA synthesis and TAG accumulation.

Figure 5.

Induced expression of genes involved in fatty acid synthesis and β-oxidation. (A) Assignment of induced genes in hepGRKO livers to fatty acid metabolism pathways. Seven different enriched Reactome pathways are shown and DEGs in each pathway are connected by nodes. The color indicates increased protein or mRNA expression in the preprandial and postprandial phases. (B) Major steps in fatty acid synthesis and β-oxidation. The color of the enzyme represents differential expression: increased/decreased as a result of GR disruption or regulated from the transition between the preprandial and postprandial phase. Chain length of acyl-CoAs is indicated by MC, medium chain; LC, long-chain; VLC, very long–chain. (C and D) Temporal mRNA and protein expression of enzymes involved in TCA cycle and acyl-CoA synthesis. (E and F) Temporal mRNA and protein expression of enzymes involved in acyl-CoA β-oxidation and fatty acid production.

Increased Expression of Enzymes Involved in Sphingolipid Synthesis

Gene expression analysis also found enriched pathways in sphingolipid metabolism in hepGRKO livers (Fig. 4F; Fig. S4C (45)), in agreement with dysregulated SM and Cer levels (Fig. 3). Several enzymes involved in Cer synthesis were induced at mRNA level, including KDSR and CERS (Fig. 6A and 6B). CERS2, a CERS isoform that preferentially uses long-chain acyl CoAs as substrate (48), was induced at protein level at ZT14 (Fig. 6C) (48). Moreover, expression of SDS (involved in serine catabolism) was constitutively decreased at both mRNA and protein level (Fig. 6A-6C). Together with increased expression of enzymes involved in FA synthesis (Fig. 5), SDS downregulation could increase the substrate (serine, palmityl-CoA, and acyl-CoA) availability for Cer synthesis, potentially providing an alternative pathway to storing the increased pool of very long–chain FAs observed in hepGRKO livers (Fig. 3) (49). Consistently, hepatic GR disruption leads to increased levels of Cers with very long-chain unsaturated FAs (Fig. 3D-3E).

Figure 6.

Induced expression of genes involved in sphingolipid metabolism. (A) Major steps in ceramide and sphingomyelin synthesis. Serine, together with palmitoyl-CoA, is the initial precursor in ceramide synthesis. Serine is catabolized by SDS. (B) Temporal mRNA expression of enzymes involved in sphingolipid metabolism. (C) Temporal protein expression of SDS, CER2 and ALDH3A2 in GR^fl/fl and hepGRKO livers, measured by DIA-MS. Statistical testing by 2-way ANOVA with Tukey's multiple comparison test. SDS, timepoint: ns, genotype: P < 0.0001, interaction: ns. CERS2, timepoint: P = .0451, genotype: P = .0004, interaction: ns. ALDH3A2, timepoint: ns, genotype: P < .0001, interaction: ns. Tukey pairwise comparisons:**P < .01, ***P < .001, ****P < .0001. Error bars in barplots indicate standard deviation.

Expression of enzymes involved in downstream Cer metabolism was also induced in hepGRKO livers, including SPHK, SGLP, and ALDH3A2 (Fig. 6A-6C). Collectively, these enzymes convert Cers to FAs and palmitate, maybe as a compensatory mechanism to increased Cer synthesis. Moreover, increased expression of SGMS and SAND8 may increase synthesis of SM and Cer phosphoethanolamine, respectively, with release of DAG potentially containing very long–chain FAs. Interestingly, expression of LPIN1, which synthesizes DAG from phosphatidate, was downregulated in hepGRKO (Fig. S6A and 6B (45)), suggesting that Cer metabolism could provide an alternative source of DAG for TAG synthesis. In agreement, we observed little change to DAG levels in hepGRKO livers. Decreased expression of LPIN1 also suggests that PEs and PCs are preferentially synthesized through the CDP-DAG pathway and to a lesser degree through the pathways using ethanolamine or choline (Fig. S6A (45)) in hepGRKO livers (50), supported by observed increased expression of Phosphatidylserine decarboxylase proenzyme (PISD) (Fig. S6C (45)).

Attenuated Mitochondrial Respiration in Hepatocytes After GR Disruption

Increased levels of Cers and intermediates in Cer synthesis have been shown to compromise mitochondria function (51-54). This includes decreased activity of complex I, III and IV and interference with mitochondrial membrane potential. Reactome pathway analysis showed decreased expression of proteins (but not mRNAs) involved in respiratory electron transport (Fig. S4 (45) and Fig. 7A). Dysregulated expression was most striking in the postprandial phase and with an abundance of proteins belonging to complex I (Fig. 7A). This suggests a possible impairment of mitochondrial respiration or reduced level of mitochondrial content. To test this, we first measured mitochondrial DNA (mtDNA) copy number by qPCR (Fig. 7B) and determined mass-specific mitochondrial density by quantification of citrate synthase activity (Fig. 7C). We also measured 3-hydroxyacyl-CoA dehydrogenase (HADH) activity (Fig. 7D), a biomarker of mitochondrial content and oxidative capacity used as an estimate of β-oxidation. Collectively, these analyses suggest no significant difference in hepatic mitochondrial content in the absence of GR. To assess mitochondrial respiratory capacity, we performed a high-resolution respirometry assay on fresh liver tissue collected from GR^fl/fl and hepGRKO mice at a postprandial timepoint, using the Oroboros oxygraph-2k system. We found no statistically significant changes (P < .05) in basal oxidative phosphorylation (Rou), in the activity of complexes I (MGP and adenosine monophosphate substrates) and II (succinate as substrate) or in uncoupled maximal respiration (carbonylcyanide p-trifluoromethoxyphenylhydrazone treatment) (Fig. 7E). However, although hepGRKO does not result in statistically significant changes in oxidative phosphorylation, both basal respiration and OXPHOS capacity of complexes I and II showed a tendency toward impaired oxygen consumption (P < .1), which suggests that mitochondrial respiration is reduced by acute disruption of GR expression.

Figure 7.

Mitochondrial effects in hepGRKO livers. (A) Temporal expression of mitochondrial-located proteins in livers from hepGRKO animals. (B) Mitochondrial DNA copy number estimated by Reverse transcription-qPCR probing for mitochondrial genes Nd1 and 16S rRNA, normalized to nuclear DNA gene NDUFV1 (nDNA), at timepoints ZT10, 12, 14, 18 and 22 from GR^fl/fl and hepGRKO mice. Statistical testing by 2-way ANOVA with Tukey's multiple comparison test. Nd1 mtDNA, timepoint: P < .0001, genotype: ns, interaction: ns. 16S rRNA mtDNA, timepoint: P = .0014, genotype: P = .0306, interaction: P = .0036. Tukey pairwise comparisons: *P < .05. (C) Hepatic activity of citrate synthase (CS) in liver samples collected at timepoints ZT10 and ZT14, reflecting mitochondrial content in GR^fl/fl and hepGRKO mice. Statistical testing by 2-way ANOVA with Tukey's multiple comparison test (timepoint: P = .0180, genotype: ns, interaction: ns). (D) Hepatic activity of hydroxyacyl-CoA dehydrogenase (HADH) in liver samples collected at timepoints ZT10 and ZT14, reflecting β-oxidation in GR^fl/fl and L-GKRO mice. Statistical testing by 2-way ANOVA with Tukey's multiple comparison test (timepoint: P = .0058, genotype: ns, interaction: ns). (E) OXPHOS capacity in livers from GR^fl/fl and hepGRKO mice collected at ZT13. Statistical tests by 1-tailed Student's t-test, P values are indicated.

Discussion

Disruption of GR signaling leads to metabolic complications in several tissues, including immune cells, muscle, fat, and liver (43, 55-61). Many of these studies are based on GR disruption during organogenesis and the observed phenotypes may be a consequence of developmental defects. This includes hepatic steatosis observed in conditional hepatic GRKO models using Alb- and Alfp-based CRE drivers (21, 43, 44). To study GR action in adult tissue, we used a mouse model to acutely disrupt GR expression in hepatocytes and subsequently analyzed lipid metabolism at preprandial and postprandial timepoints during a standard 12 hour light/dark circadian rhythm. To control for food intake, we housed mice under a night restricted feeding schedule. Consequently, the animals experienced 12 hours of fasting in the resting/light phase, which extends the normal pre-prandial period for ad libitum fed mice (62). Thus, the night restricted feeding setup applied in this study may influence the recorded circadian lipid metabolism. However, as lipid metabolism is highly regulated by the fed/fasted status (10), we prioritized an experimental model that took this into account. Moreover, our study design was restricted to male mice and since fasting response and liver metabolism are highly affected by the sex (63, 64), future hepGRKO studies with female mice will be highly relevant.

In agreement with previous long-term GR-disruptive studies in hepatocytes (21, 43, 44) we observed TAG accumulation after acute GR KO in hepatocytes. Moreover, hepatocyte specific deletion of 11β-HSD1, the enzyme converting inactive cortisone to active corticosterone, leads to hepatic steatosis (65), collectively suggesting that disruption of GR signaling in hepatocytes results in hepatic steatosis. Interestingly, we observed that lipid accumulation was most pronounced in the postprandial phase, where the concentration of circulating corticosteroid levels is normally low relative to the circadian peak observed prior to the onset of active phase (16). This suggests that active GR signaling at the transition from fasted to fed states affects subsequent lipid metabolism in the late postprandial state. We have previously suggested similar regulation of hepatic glucose uptake (16), indicating a broad metabolic effect of GR signaling in the postprandial phase. Likewise, bile acid synthesis and reuptake in the postprandial phase may in part be regulated by GR at the preprandial to postprandial transition (18, 66).

Quantification of the lipidome revealed 2 main profiles of temporal lipid levels as mice enter the postprandial phase of a circadian cycle. One pool of preprandially accumulated TAGs and phospholipids, containing very long unsaturated acyl chains, showed decreased levels as mice are fed. This pool is synthesized from lipolysis-derived FAs from white adipose tissue in the fasted state or from dietary FAs that reach the liver within chylomicron remnants in the very late postprandial phase and is eventually secreted as VLDL particles during the early resting/fasted phase (10). Another pool consists of postprandially accumulated very long saturated FAs as well as TAGs containing long-chain saturated acyl chains originating from FA synthesis upon glucose uptake by the liver or by chylomicron remnants after feeding (10). Interestingly, GR disruption affects the abundance of specific TAGs and FAs in these 2 major groups of temporally regulated lipids and leads to accumulation of specific sphingolipids that do not show any apparent temporal regulation. Collectively, this may be the leading cause of pronounced lipid droplet formation in the postprandial phase.

Specifically, in acute absence of GR expression, feeding-induced TAG species with long-chain saturated FAs accumulated less in the postprandial phase, while very long, saturated FAs accumulated more and earlier in the postprandial phase, suggesting that synthesis of specific TAGs could be impaired by the loss of GR. We assessed TAG synthesis from a labeled long-chain FA (oleic acid, C18:1), a major component of dietary fat, and found reduced FA uptake and incorporation into TAGs in primary hepatocytes. Decreased FA uptake could be associated with the impairment of proteins involved in FA uptake by the liver, including FATPs and CD36 (4, 67). Although we do not observe any effects on the expression level of these FA transporters, activity and localization of the protein at the plasma membrane might be disrupted (68). Alternatively, previous studies have shown induced de novo lipogenesis by glucocorticoid treatment (69, 70), suggesting that hepGRKO may reduce lipogenesis in hepatocytes. However, in contrast we observed increased levels of several enzymes involved in de novo lipogenesis, suggesting no negative impact on lipogenesis. Yet, we are not able to exclude the possibility of reduced enzyme activity. Along with increased levels of lipogenic enzymes we also observed increased levels of enzymes involved in FA elongation and desaturation, suggesting that TAG synthesis from shorter chain acyl-CoAs may be compromised by continuous acyl-CoA desaturation and elongation which eventually leads to increased levels of TAGs with very long unsaturated FAs, potentially contributing to the postprandial steatotic phenotype. Increased TAG levels may also arise from disrupted lipoprotein synthesis; however, we do not observe any effect on TAG content in circulating VLDL.

In addition to disrupted FA and TAG levels, we observed increased accumulation of ultralong Cers (C26 corresponding to Cer 44:4 and C44:5, and C40 corresponding to Cer 58:3) upon disruption of GR in the liver, regardless of the circadian phase status. Dysregulated synthesis of very long–chain Cers has been reported to impair FA uptake in the liver and decrease TAG synthesis via downregulation of CD36/FAT levels and altered membrane localization (71), which may partly explain impaired FA uptake. Very long–chain Cers have also been associated with lipotoxic effects leading to mitochondrial dysfunction and consequently oxidative stress and cell death (51, 72), which may contribute to the reduced mitochondrial respiration in hepGRKO animals. Ultralong Cers do not seem to be as well described and are even occasionally grouped together with very long Cers; however, there is indication that accumulation of Cers could have significant, chain length–dependent effects on metabolism (72), including negative effects on lipid uptake and mitochondrial function, also observed in the acute hepGRKO model. Additionally, we observed an increase in saturated very long–chain FAs. The changes in very long–chain FA levels could also be associated with the accumulation of ultralong Cers, either for being used in their synthesis or by being partly originated from compensatory degradation of those Cers (72).

Collectively, our study suggests that the steatotic phenotype observed in the hepGRKO mice originates from disrupted lipid metabolism as the mice transition from a fasted to a fed state. It should be noted that, based on the data presented, we are not able to mechanistically assign the steatotic phenotype solely to cell autonomous effects in the hepatocytes. KO of GR in hepatocytes may affect metabolism in other tissues including muscle and fat, which subsequently may contribute to dysregulated hepatic lipid metabolism. Yet, we did not observe any significant changes of serum corticosterone, insulin and NEFA levels, which indicates that lipolysis in adipose tissue is not a contributing factor. In agreement, NEFA levels and adipose tissue morphology are unaffected by long term GR KO using AlfpCre driver (44).

It is well-documented that systemic administration of glucocorticoids leads to hepatic steatosis (73) and experiments with primary hepatocytes suggest that glucocorticoids augment TAG accumulation (69, 70). Thus, disruption as well as increased activity of GR in hepatocytes can lead to hepatic steatosis. These seemingly contradicting findings are likely explained by different mechanisms leading to steatosis. Systemic administration of glucocorticoids disrupts multiple interconnected metabolic pathways in several tissues, including amino acid metabolism (liver and muscle), gluconeogenesis (liver), lipolysis (adipose tissue), hyperinsulinemia/insulin resistance (liver, pancreas, muscle, and adipose tissue), lipogenesis (liver), VLDL secretion (liver), and FA β-oxidation (liver) (73). Importantly, earlier studies suggests that GR disruption (by GR or 11β-HSD1 deletion) in adipocytes alleviates glucocorticoid-induced hepatic steatosis, suggesting that adipocyte lipolysis is a major contributor to glucocorticoid-induced hepatic steatosis (55, 61, 65), supported by a recent study showing that inhibition of lipolysis protects against glucocorticoid-induced hepatic steatosis (74). This implies that augmented GR signaling in hepatocytes plays a marginal role for hepatic steatosis during systemic glucocorticoid administration. In contrast, hepatic GR signaling may protect against hepatic steatosis by the control of feeding-regulated lipid metabolism. Interestingly, a recent study suggests that selective modulation of GR can prevent liver steatosis in diet-induced obese animals, suggesting that targeting specific actions of GR can prove beneficial without metabolic side effects (75). Yet, it remains to be determined if these effects are attributed to GR in hepatocytes. The described animal model used in our study may prove beneficial.

In conclusion, we show that GR has a significant role on the regulation of mitochondrial respiration and feeding-responsive processes such as lipid metabolism in the liver. Disrupted GR signaling in the liver leads to pronounced FA, TAG, and sphingolipid accumulation in the postprandial phase accompanied by increased lipid droplet formation, indicating that the preprandial surge of corticosterone controls lipid turnover in response to feeding.

Acknowledgments

We thank professor Dr. Jan Tuckermann, Institute for Comparative Molecular Endocrinology, Ulm, Germany, for the initial FVB/N-Nr3c1^fl/fl mice. Trine V. Dam and Victor E. Goitea helped with animal work, Philip Hallenborg generated the DIA library, and Anita Lunding (DaMBIC) prepared samples for CARS imaging. Part of figure 1A was created with BioRender.com.

Funding

The research was funded by the SDU2020 initiative, the Novo Nordisk Foundation, Independent Research Fund Denmark (Sapere Aude Starting Grant) and the Danish National Research Foundation through a grant (#141) to the Center for Functional Genomics and Tissue Plasticity (ATLAS). Proteomics work was supported by the INTEGRA research infrastructure and the PRO-MS Danish National Mass Spectrometry Platform for Functional Proteomics. CARS microscopy was performed at the Danish Molecular Biomedical Imaging Center (DaMBIC).

Disclosures

The authors have nothing to disclose.

Data Availability

A link to the supplementary data (Tables S1 and S2 and Figs. S1-S6) is included in the References (45). Processed MS based proteomics and lipidomics data is available in Table S1 (45) (differentially regulated proteins) and Table S2 (45) (differentially regulated lipids). RAW proteomics data is available at PRIDE: PXD041168. We used RNA-seq and GR ChIP-seq data previously described in (16) and available at GEO: GSE173723.

Abbreviations

 
• AAV

  adeno-associated virus

 
• ANOVA

  analysis of variance

 
• ACN

  acetonitrile

 
• BSA

  bovine serum albumin

 
• CARS

  coherent anti-Stokes Raman scattering

 
• Cer

  ceramide

 
• DAG

  diacylglycerol

 
• DIA

  data-independent acquisition

 
• FA

  fatty acid

 
• GFP

  green fluorescent protein

 
• GR

  glucocorticoid receptor

 
• hepGRKO

  hepatocyte-specific GR knockout model

 
• HDL

  high-density lipoprotein

 
• KO

  knockout

 
• LDL

  low-density lipoprotein

 
• NEFA

  nonesterified fatty acid

 
• NRF

  night-restricted feeding

 
• PC

  phosphatidylcholine

 
• PE

  phosphatidylethanolamine

 
• qPCR

  quantitative polymerase chain reaction

 
• RT

  room temperature

 
• SDS

  sodium dodecyl sulfate

 
• SM

  sphingomyelin

 
• TAG

  triacylglycerol

 
• TBS

  Tris-buffered saline

 
• VLDL

  very low–density lipoprotein