Adipose triglyceride lipase (ATGL; #2138), caveolin-1 (#3267S), FABP4 (#3544S), hormone sensitive lipase (HSL) pS563 (#4139), CREB pS133 (#9191S) and pPKA motif specific [P-RXX(S/T); #9624S] antibodies were from Cell Signaling Technologies (Danvers, United States). Heat shock protein (HSP) 90 (#610418) antibody was from BD Transduction Laboratories (Franklin Lakes, United States). Perilipin-1 total (#4854) and pS522 (#4856) antibodies from Vala Science (San Diego, United States). Antibody against EHD2 (#154784) was from Abcam (Cambridge, United Kingdom). Total HSL antibody was kindly provided by Cecilia Holm (Lund University). Fluorescence-conjugated (Alexa Fluor) secondary antibodies, Bodipy 493/503 (D3922), and MitoTracker RED CMXRos (M7512) were purchased from Molecular Probe (Waltham, United States). Rosiglitazone (R2408) was from Sigma-Aldrich (St. Louis, United States). The PPRE-x3-TK-Luc reporter plasmid was a kind gift from Bruce Spiegelman (Addgene plasmid #1015) and Renilla Luciferase Control Reporter (pRLnull; #E227A) from Promega (Madison, United States). Reagent for determination of glycerol content (F6428) and dibutyryl-cAMP (DcAMP; #D0627) were from Sigma-Aldrich (St. Louis, United States). PDE3 inhibitor OPC3911 and PDE4 inhibitor Rolipram were from Otsuka Pharmaceuticals (Tokyo, Japan) and PDE3B antibody was produced in-house.

C57BL6/N mice with global deletion of exon 3 of the Ehd2 gene (Ehd2^−/−) as described (Matthaeus et al., 2020) were kindly provided by Oliver Daumke (MDC, Germany). Genotyping was carried out to confirm deletion of the Ehd2 gene. Ehd2^−/− and Ehd2^+/+ C57BL6/N mice (the latter used as WT control) were bred in parallel to generate sufficient number of animals for cellular analysis. Animals were on a 12h light cycle with non-restricted food and water. Mice were fed either chow or HFD (D12492 60 E% fat content; Research Diets, New Brunswick, United States).

All animal procedures were approved by the Malmö/Lund Committee for Animal Experiment Ethics, Lund, Sweden, and were carried out in accordance with the relevant guidelines and regulations.

Primary adipocytes were isolated from different adipose tissue depots using previously established protocol (Rodbell, 1964). Isolated, primary adipocytes were suspended 10% (v/v) in Krebs Ringer Bicarbonate HEPES (KRBH) buffer, pH 7.4, containing 200nM adenosine and 3% (w/v) bovine serum albumin (BSA).

Following incubations, cells were washed with KRH medium without BSA and subsequently lysed in a buffer containing 50mM Tris/HCl pH 7.5, 1mM EGTA, 1mM EDTA, 0.27M sucrose, 1% NP-40, and complete protease- and phosphatase inhibitor cocktail (Roche, Basel, Switzerland). Lysates were centrifuged for 10min at 13,000×g and protein concentrations were determined using the Bradford method. Samples were subjected to polyacrylamide gel electrophoresis and electro-transfer to nitrocellulose membranes. Membranes were blocked with non-fat dry milk [10% (w/v)] and probed with the indicated antibodies. Detection was performed using horseradish peroxidase conjugated secondary antibodies and enhanced chemiluminescence reagent. The signal was visualized using a BioRad Image camera (Biorad, Hercules, United States).

Adipocytes were transfected as previously described (Lizunov et al., 2013). In short, isolated adipose cells were suspended [40% (v/v)] in DMEM supplemented with Gentamicin (20 µg/ml) and 200nM phenylisopropyladenosine (PIA). Cells were electroporated using a square-wave pulse; 400V, 12ms, one pulse (Harvard Apparatus, Holliston, United States) with PPRE reporter and pRLnull (normalization) plasmids (10:1). After electroporation, the cells were transferred into DMEM with Gentamicin (20μg/ml), PIA and [3.5% (w/v)] BSA and cultured for 24h at 37°C in 5% CO₂. Thereafter, the cells were lysed in Promega passive lysis buffer, and the lysates centrifuged at 10,000×g for 10min at 4°C to separate the protein lysate (infranatant) from the fat. Luciferase activity was measured in a Glomax luminometer (Promega) using the Dual Luciferase Reporter system (Promega), where data were normalized to Renilla reporter (pRLnull).

Imaging was performed using a Nikon A1 plus confocal microscope with a 60× Apo DIC oil immersion objective with a NA of 1.40 (Nikon Instruments) and appropriate filter sets. Images were acquired with NIS-Elements, version: 4.50.02 (Laboratory Imaging). Isolated cells were fixed using 4% paraformaldehyde (PFA) and labeled with antibodies in a buffer containing 1% BSA, 1% goat serum, and 0.05% saponin 1–2h. For neutral lipid staining, cells were stained with Bodipy 493/503 and for mitochondria we used Mitotracker. For imaging of primary adipocytes, we used previously described protocol (Wasserstrom et al., 2018).

For standard trans-electron microscopy, adipose tissue was fixed overnight using 1% PFA and 3% glutaraldehyde, followed by dehydration using methanol series (20, 40, 60, 80, and 100%) 30min each, de-lipidated using dichloromethane for 1h, and then rehydrated with reverse methanol series. Methanol was diluted in KRBH without BSA. The samples were fixed using osmium-tannic acid-osmium (OTO) for 1h each, washed in KRBH without BSA, dehydrated using ethanol series and embedded in EPON.

Lipolysis was determined by measuring glycerol release, as described previously (Hansson et al., 2016). In short, cells [10% (v/v) cell suspension] were treated with or without isoprenaline (ISO; 10nM), or a combination of insulin (1nM) and ISO (10nM) for 30min. Cell medium was subsequently removed for enzymatic determination of the glycerol content. In short, 100μl free glycerol reagent was added to 30μl sample. Absorbance at 540nm was measured after incubating for ~15min at room temperature. Each sample was analyzed in duplicate. In a separate set of experiments, cells were pre-incubated (10min) with OPC3911 (10μM), Rolipram (10μM) or DMSO 0.1% (control), followed by 30min stimulation with or without ISO (10nM) or a combination of insulin (1nM) and ISO (10nM). For experiments where DcAMP was used, cells were pre-incubated for 10min and then treated with or without ISO (10 or 1,000nM) or DcAMP (100 or 1,000μM) for 30min. Glycerol release was measured as stated above.

Total RNA was extracted using miRNeasy mini kit (Qiagen #74104). RNA purity and concentration were assessed using a Nanodrop spectrophotometer (Thermo Scientific). PCR-reactions were performed using the Quantifast SYBR Green RT-PCR kit (Qiagen #204156) and Quantitect primer assays for 18S, Adrb3, Ehd2, and Plin1. Primer sequences are considered proprietary information by Qiagen. mRNA expression levels were measured using a StepOnePlus real-time thermal cycler (Applied Biosystems Waltham, United States) and quantitated using the ΔΔC_T method as described by Livak and Schmittgen (2001). 18S rRNA was used for normalization throughout.

Adipose tissue samples were obtained from epididymal (EPI) and ING adipose depots. The adipose cell-size distributions were obtained using a Beckman-Coulter counter after osmium fixation as described previously (Li et al., 2016).

To determine the liver triglyceride concentration, ~100mg frozen liver sample was homogenized using a motorized homogenizer in buffer containing 5% (w/v) NP-40. The homogenized samples were heated at 90°C for 3min, cooled down for 15min in room temperature, reheated at 90°C for 3min, and then centrifuged at maximum speed for 2min at 20°C. The supernatant was collected and placed in a 96-well plate with triglyceride reagent (Thermo Scientific) for triglyceride concentration determination.

Terminal serum samples were collected and blood glucose levels measured using OnetouchUltra2 (Lifescan, Milpitas, United States), and insulin levels assayed in serum using ELISA from Mercodia (Uppsala, Sweden). Serum concentration of free fatty acid (FFA) was measured using FFA assay kit (#ab65341) from Abcam (Cambridge, United Kingdom).

Cut frozen sections (5μm) were fixed with 4% PFA and incubated for 10min at room temperature. Excess of 4% PFA was removed by three rinses in PBS. Subsequently, the sections were rinsed with 60% isopropanol and then stained with freshly and filtrated oil Red-O working solution [30ml of 0.5% ORO (O0625, Sigma; Saint Louis, United States) stock solution+20ml distilled water] for 20min at room temperature. After the sections were rinsed with 60% isopropanol and then washed with distilled water and covered with a coverslip using permanent aqueous mounting medium (BUF058A, Biorad). Images were captured using an Olympus BX60 microscope.

Analysis was performed by unpaired two-tailed Student’s t-test or one-way ANOVA, Tukey’s post hoc test using GraphPad Prism 6 (Graphpad Software Inc.) software. Significance was determined according to ^*p<0.05, ^**p<0.01, and ^***p<0.001.

The data that support the findings of this study are available from the corresponding author upon reasonable request.

To examine adipocyte function in the EHD2-deficient state, we made use of Ehd2 whole-body knockout (Ehd2^−/−) mice (Matthaeus et al., 2020). Throughout the study, we used corresponding Ehd2^+/+ (WT) mice as control. Western blot analysis from various tissues confirmed global deletion of the EHD2 protein (Figure 1A). Age-matched (~11–12weeks) Ehd2^−/− mice displayed lower body weight (BW), but similar blood glucose and serum insulin levels compared with WT (Figures 1B–D). We found Ehd2^−/− mice to have elevated liver triglyceride levels (p=0.064; Figures 1E,F), a decreased amount of EPI adipose tissue, an increased amount of inguinal (ING) adipose tissue, and trends toward an increased amount of retroperitoneal (RETRO) adipose tissue compared with WT (Figures 1G–I). Further, adipocyte size distribution analysis of EPI and ING adipose tissue revealed similar distribution curves comparing Ehd2^−/− and WT mice (Figures 1J,K). These data pinpoint that Ehd2^−/− mice not only have a trend toward increased liver triglycerides, in line with previous report (Claudia Matthäus et al., 2019), but also have reduced visceral fat mass compared with WT.

To test if a HFD further aggravated the shifted fat distribution, animals were fed 2weeks of HFD, which has proven sufficient to induce weight gain and systemic insulin resistance in C57Bl6/J mice (Hansson et al., 2018).

Both Ehd2^−/− and WT mice rapidly gained weight to the same extent (Figure 2A), and no differences in food intake, blood glucose, serum insulin, or serum FFAs levels were observed (Figures 2B–E). Glucose and insulin levels increased significantly, and to the same extent, with HFD-feeding compared with chow in both genotypes (shown in Figures 1C,D, 2C,D), which confirms that C57Bl6/N mice also are prone to rapidly develop obesity and insulin resistance on a high fat diet. There was a trend toward increased liver triglyceride levels in Ehd2^−/− mice compared with WT (Figure 2F). Further, Ehd2^−/− mice fed HFD had a significantly reduced amount of EPI fat tissue, while the ING fat depot was similar with WT (Figures 3A,B). Correlation analysis demonstrated a significant difference in expansion of fat depots, comparing EPI and ING from WT and Ehd2^−/− (Figure 3C), whereas RETRO adipose tissue expansion was similar in Ehd2^−/− compared with WT (Supplementary Figure S1). Cell size distribution analysis revealed a clear shift in adipocyte size distribution toward larger sizes in response to HFD-feeding in both WT and Ehd2^−/− mice [Figures 3D,E (EPI) and Figures 3H,I (ING)]. Even though, mature adipocytes from Ehd2^−/− mice expanded in response to HFD, the distribution curve displayed lower frequency of very small adipocytes and an accumulation of medium-sized adipocytes in Ehd2^−/− compared with WT in the HFD-fed state (Figures 3F,G,J,K).

Ligand-activated transcription factor PPARγ is essential for adipogenesis (Rosen et al., 2002). Since EHD2 has been reported to influence transcriptional activity (Torrino et al., 2018), we next assessed if reduced EPI fat tissue expansion could reflect impaired PPARγ activity in Ehd2^−/− cells by using a PPARγ response element (PPRE) reporter assay. Indeed, primary EPI adipocytes from chow-fed Ehd2^−/− mice displayed 60–70% lower PPARγ activity compared with WT, both in non-stimulated cells, and in cells treated with the PPARγ ligand Rosiglitazone (Figure 4A). Further, by western blot, we detected ~50% lower levels of FABP4 (fatty acid binding protein), one of the target genes of PPARγ (Kletzien et al., 1992), in both EPI and ING adipocytes isolated from Ehd2^−/− mice compared to WT, even though, less pronounced in ING cells following HFD-feeding (Figures 4C,D). The EHD2 knockout was confirmed by western blot using cell lysates and by confocal imaging (Figure 4B). These results suggest that EHD2 influences the transcriptional activity of PPARγ in primary adipocytes, which could explain the impaired expansion of the EPI adipose tissue depot. Possibly, this is due to lowered levels of FABP4 and intracellular fatty acids, the latter acting as endogenous PPARγ ligands, stimulating both lipogenic protein expression and differentiation of new adipocytes. In addition, the impaired response to the synthetic ligand Rosiglitazone in Ehd2^−/− cells also suggests that PPARγ itself or its co-activators are downregulated.

Further, knock-down of EHD2 is reported to affect the stability of caveolae rather than caveolae density (Moren et al., 2012; Ludwig et al., 2013; Matthaeus et al., 2020). In line with that, we could observe the presence of caveolae in intact adipose tissue from WT and Ehd2⁻/⁻ by electron microscopy (Figure 4E, arrows indicate caveolae). Western blot analysis revealed that the main core protein of caveolae, caveolin-1, was reduced ~50% in adipocytes isolated from EPI and ING adipose tissue in Ehd2^−/− mice compared to WT after HFD-feeding (Figures 4C,D). Possibly, the lower caveolin-1 expression could affect the composition or stability of caveolar structures.

So far, the data suggest that absence of EHD2 causes impaired expansion of the adipocytes per se and also leads to reduced EPI adipose tissue mass. To investigate whether lack of EHD2 affects intracellular lipid handling, we next focused on lipolysis, the hydrolysis of triglycerides. In the following, lipolysis, measured as the release of glycerol in primary adipocytes, was examined in both non-stimulated (basal) and β-adrenergic [Isoprenaline (ISO)] stimulated cells. In chow-fed mice, basal but not ISO-stimulated lipolysis was significantly lower in EPI adipocytes from Ehd2^−/− mice compared with WT (Figure 5A). In contrast, there was no significant difference in lipolysis in EPI adipocytes isolated after HFD-feeding in either non- or ISO stimulated cells comparing WT and Ehd2^−/− (Figure 5A). In ING adipocytes, no significant change was observed in chow-fed mice (Figure 5B). Yet, both basal and ISO stimulated lipolysis was significantly reduced (~50%) in cells obtained from Ehd2^−/− mice compared to WT after HFD-feeding (Figure 5B). To further elucidate mechanisms behind reduced lipolysis in ING cells obtained after HFD-feeding, we analyzed the phosphorylation of perilipin-1 and HSL, the main triglyceride lipase. We found that adipocytes from Ehd2^−/− mice displayed reduced phosphorylation levels of perilipin-1 (pS522) and HSL (pS563) following ISO stimulation (Figure 5C), which is in line with reduced lipolysis (Figure 5B).

Still, the expression of HSL and ATGL was unaltered, whereas total perilipin-1 level was significantly increased (~30%) in Ehd2^−/− adipocytes compared to WT (Figure 5D). Possibly, the latter could reflect an increased lipid droplet surface area since small surface-associated lipid droplets were visible to a larger extent in adipocytes from Ehd2^−/− mice (Figure 5E).

To test if the reduced lipolysis was caused by impaired signaling at the level of beta adrenergic receptor or adenylate cyclase, we treated cells with dibutyryl-cAMP (DcAMP), a cell permeable, non-degradable cAMP analog. DcAMP activates protein kinase A (PKA) which phosphorylates and activates HSL and perilipin-1, and thereby induces lipolysis. Compared to WT, Ehd2^−/− adipocytes showed lower lipolysis upon DcAMP treatment (Figure 6A). This suggests that the impaired lipolysis in Ehd2^−/− adipocytes occurs downstream of cAMP in the beta adrenergic signaling pathway. Notably, in this set of experiments, we could confirm that also at a very high dose of ISO (1,000nM), lipolysis was significantly reduced in Ehd2^−/− adipocytes. Further, these results are also in agreement with similar gene expression of adrenergic receptors (Adr3b) in adipocytes from Ehd2^−/− mice compared with WT (Figure 6B). The trends toward increased gene expression of Plin1 mirrors the increased perilipin-1 protein expression shown in Figure 5D.

Consistent with the impaired effects of ISO and DcAMP on lipolysis and phosphorylation of HSL and perilipin-1, we found reduced ISO-induced phosphorylation of several PKA targets in adipocytes from Ehd2^−/− mice using a phospho-PKA motif specific antibody [P-RXX(S/T); Figure 6C]. Further, we examined the phosphorylation of the cAMP-responsive transcription factor CREB. CREB (pS133) was lower in non-stimulated cells from Ehd2^−/− mice compared with WT but increased in response to ISO (Figure 6D). In comparison, the fold increase of phosphorylated perilipin-1 and HSL in response to ISO was lower in Ehd2^−/− adipocytes, whereas the fold increase of phosphorylated CREB was increased in Ehd2^−/− adipocytes (Figure 6E). Thus, the ability to transmit cAMP-dependent signaling was not compromised for all targets in adipocytes from Ehd2^−/− mice.

Next, we investigated if altered expression and activity of PDE3 could explain the impaired lipolysis in ING adipocytes from Ehd2^−/− mice. PDE3B is a major PDE family member expressed in adipocytes, responsible for degradation of cAMP. Thus, the expression level and activity of PDE3B have an impact on cAMP mediated signaling events, particularly the anti-lipolytic effect of insulin (Degerman et al., 2011).

The use of a PDE3 inhibitor (OPC3911) increased basal and ISO-mediated lipolysis to the same extent (~2-fold) in both genotypes (Figure 6F). We also detected similar protein expression of PDE3B in WT and Ehd2^−/− adipocytes (Figure 6G). Still, lower lipolysis in ISO-stimulated cells in Ehd2^−/− adipocytes was observed even in the presence of OPC3911, indicating that increased PDE3B activity is not causing the impaired ISO-mediated lipolysis in Ehd2^−/− adipocytes (Figure 6F).

Further, we investigated insulin’s ability to suppress lipolysis, one of insulin’s major metabolic effects in adipocytes. As expected, insulin efficiently suppressed ISO-induced lipolysis in WT (~50% reduction; Figure 6F, same data normalized to ISO are presented in Figure 6H). In contrast, much less suppression was observed in cells from Ehd2^−/− [~25% reduction (Figures 6F,H)]. Further, when co-stimulating with both OPC3911, ISO, and insulin, the anti-lipolytic effect of insulin was lost in WT as expected, but also in Ehd2^−/− adipocytes (Figure 6F). In comparison, the use of a PDE4 inhibitor, Rolipram, increased ISO-induced lipolysis, but did not abolish insulin’s effect in either genotypes (Figure 6F).

Together, alterations of PDE3B expression/activity appear not to contribute to the low lipolysis observed in adipocytes from Ehd2^−/− mice. Instead, the data suggest that a defect at the level of, or downstream of PKA, contributes to this phenotype.