Coordination Among Lipid Droplets, Peroxisomes, and Mitochondria Regulates Energy Expenditure Through the CIDE-ATGL-PPARα Pathway in Adipocytes

The rapid increase in the global prevalence of obesity is highly deleterious, resulting in the deregulation of glucose and lipid metabolism and the development of other metabolic diseases, including insulin resistance, hyperglycemia, hepatic steatosis, dyslipidemia, and chronic inflammation (1,2). A severe reduction of white adipose tissue (WAT) with ectopic lipid accumulation in other tissues and organs, which is called lipodystrophy, can also cause these metabolic disorders (3,4). Thus, excessive lipid storage (obesity) and the failure to store lipids in adipose tissue (lipodystrophy) both contribute to metabolic disorders, although these disorders are associated with opposite fat storage phenotypes. In addition to maintaining an appropriate amount of adipose tissue, fat tissue types and their interactions, including WAT as an energy storage site and brown adipose tissue (BAT) and beige adipose tissue as energy-burning sites, are important in systemic lipid metabolism (5). At the cellular level, the interaction between lipid droplets (LDs) and other organelles in adipocytes also contributes to cellular lipid homeostasis, including metabolic flux and the availability of fatty acids (6,7). Therefore, the balance and coordination of adipocyte types and subcellular organelles are important in maintaining systemic metabolic homeostasis.

The CIDE (cell death-inducing DNA fragmentation factor 45-like effector) proteins, including Cidea, Cideb, and Cidec/Fsp27, have emerged as important regulators in the maintenance of lipid metabolic homeostasis, particularly in adipose tissues and the liver (8). CIDE proteins are expressed mainly on the surface of LDs and induce the formation of large LDs by promoting lipid exchange between LDs that come into contact with each other in adipocytes (Cidea and Cidec) and hepatocytes (Cideb) (9–11). Cidea is abundantly expressed in BAT, and a deficiency of this protein promotes energy expenditure by increasing AMPK activity (12,13). Cideb is stably expressed in the liver and small intestine to promote lipid storage and VLDL maturation for secretion (14–16). Cidec is highly expressed in WAT and moderately expressed in BAT. Cidec deficiency results in multilocular LDs and markedly reduced lipid storage in WAT, with increased energy expenditure (17,18). However, Cidec deficiency in obese animals promotes a lipodystrophy phenotype, with fatty liver, insulin resistance, and dyslipidemia (19,20), which is similar to the phenotype of a patient carrying a point mutation in the Cidec gene (21). Both Cidea and Cidec are inducible and contribute to fatty liver development, although these proteins are not expressed in the normal liver (22,23). Based on these observations, we investigated what changes would occur in mice with a double deficiency of Cidea and Cidec under obesity-inducing conditions and whether this double deficiency would improve metabolic disorders.

Hence, four genetically engineered mouse models, including ob/ob, ob/ob/Cidea^−/−, ob/ob/Cidec^−/−, and ob/ob/Cidea^−/−/Cidec^−/− mice, were developed and compared. We observed a lean phenotype in ob/ob/Cidea^−/−/Cidec^−/− mice, with reduced fat storage and improved insulin sensitivity. Increased energy consumption occurred in the WAT and BAT of ob/ob/Cidea^−/−/Cidec^−/−mice. Interestingly, a dramatic increase in energy consumption was observed through the coordinated interplay of multiple organelles, including LDs, peroxisomes, and mitochondria; this effect was mediated by the adipocyte triglyceride lipase (ATGL)-peroxisome proliferator–activated receptor α (PPARα) pathway.

Double-deficient ob/ob/Cidea^−/− and ob/ob/Cidec^−/− mice were generated by crossing Cidea^−/− or Cidec^−/− mice with leptin^+/− mice on a C57BL/6 background, as previously described (20,23). Triple-deficient ob/ob/Cidea^−/−/Cidec^−/− mice were generated by crossing leptin^+/−Cidea^−/− mice with Cidec^−/− mice or leptin^+/−Cidec^−/− mice with Cidea^−/− mice. All mice used for studies were male. Three-month-old mice that were subjected to random feeding were sacrificed during the time period 2:00–5:00 p.m. for sample collection for the analyses presented in Figs. 1, 2, and 4–6. Gonadal WAT was used in this study. Two- and 4-month-old mice were used in the analyses presented in Fig. 3. For the high-fat diet (HFD) treatments presented in Supplementary Fig. 2, 2-month-old mice were provided with an HFD for 2 months. For mouse studies, biological replicates were from independently tested individual mice. Mouse experiments were performed in the animal facility of Tsinghua University (Beijing, People’s Republic of China). All animal experiments were approved by the Institutional Animal Care and Use Committee of Tsinghua University.

O₂ consumption, CO₂ production, the respiratory exchange rate, and energy expenditure were determined using a PhenoMaster/LabMaster System (TSE Systems GmbH, Bad Homburg, Germany). Three-month-old mice were individually monitored for 48 h, and data were collected at intervals of 27 min after a 1-day adaptation period. Parameters were compared between different genotypes without considering body weight differences (24).

The concentrations of serum triacylglycerol (TAG) and glycerol were measured using a serum triglyceride determination kit (catalog #TR0100; Sigma-Aldrich, St. Louis, MO) and a free glycerol reagent (catalog #F6428; Sigma-Aldrich), respectively, and the concentration of serum nonesterified fatty acids (NEFAs) was measured using a LabAssay NEFA kit (catalog #294–63601; Wako Pure Chemical Industries, Ltd., Osaka, Japan). Serum hormone and cytokine levels were detected using a rat insulin radioimmunoassay kit (catalog #RI-13K; Millipore Sigma, Burlington, MA), mouse tumor necrosis factor-α (TNF-α) and interleukin-6 (IL-6) ELISA eBioscience Ready-SET-GO! Kits (catalog #88-7324-86 and #88-7064-88; Thermo Fisher Scientific, Waltham, MA), an adiponectin kit (catalog #ab108785; Abcam, Cambridge, U.K.), and hyaluronidase (HAase) and laminin kits (catalog #SEB217Mu and #SEA082Mu; Cloud-Clone Corp., Wuhan, People’s Republic of China). Tissue or cell catalase activity was determined using a catalase assay kit (Beyotime Biotechnology, Shanghai, People’s Republic of China).

The lipid contents of tissues were determined by thin-layer chromatography, as previously described (20). The cellular content of TAG was measured by an enzymatic reaction, according to the manufacturer instruction manual (catalog #290-63701; Wako Pure Chemical Industries, Ltd.). The tissues were lysed, and NEFA levels were determined using a LabAssay NEFA kit (catalog #294-63601; Wako Pure Chemical Industries, Ltd.). The NEFA profile was measured using a Q Exactive Mass Spectrometer with an Orbitrap Analyzer (Thermo Fisher Scientific) in the Lipidomics Center at Tsinghua University.

The mice were fasted for 6 h and intraperitoneally injected with glucose (0.5 g/kg body mass) for glucose tolerance tests (GTTs). Following a 4-h fast, an intraperitoneal injection of insulin (2 units/kg body mass) was administered for insulin tolerance tests (ITTs). Blood glucose concentrations were measured using a blood glucose monitor system (ACCU-CHEK Advantage II; Roche, Basel, Switzerland). To examine in vivo insulin signaling, 4-month-old mice were fasted for 2 h, anesthetized and injected with insulin (5 units/kg body mass). After 5 min, the mice were sacrificed, and tissues were collected for Western blotting.

Primary BAT stromal vascular cell isolation and differentiation were performed as previously described (17,25). Briefly, the interscapular brown fat pad was removed from neonates on postnatal day 2 to obtain primary brown adipocytes. The confluent cells were induced to undergo differentiation by DMEM with 10% FBS supplemented with 1 μg/mL insulin, 1 nmol/L triiodothyronine (T3), 0.125 mmol/L indomethacin, 2 μg/mL dexamethasone, and 0.5 mmol/L 3-isobutyl-1-methylxanthine (IBMX) for 2 days, followed by culture with DMEM with 10% FBS supplemented with 1 μg/mL insulin and 1 nmol/L T3 for 4 days. On day 6, the cells were trypsinized and recultured for imaging or Seahorse Analyzer (Seahorse Bioscience, Billerica, MA) experiments.

Primary stromal vascular cells were isolated from the white fat tissues of 1-month-old mice (26). Briefly, the WAT was cut into small pieces and digested in buffer (Hanks’ buffer, catalog #14025-092; Thermo Fisher Scientific) with 2% BSA and collagenase (catalog #C6885, 2 mg/mL; Sigma-Aldrich) for 40 min at 37°C. The confluent cells were induced to differentiate for 2 days in DMEM with 10% FBS, 5 μg/mL insulin, 2 μg/mL dexamethasone, 0.5 mmol/L IBMX, and 1 μm rosiglitazone, followed by culture for an additional 4 days in DMEM with 10% FBS supplemented with 5 μg/mL insulin. On day 6, the cells were trypsinized and recultured for imaging or Seahorse Analyzer experiments.

For mouse embryonic fibroblast (MEF) isolation, 13.5- to 14.5-day mouse embryos were used (17). The isolated MEFs were differentiated at high density in high-glucose DMEM plus 10% FBS, 0.8% biotin, 0.4% pantothenate, 100 units/mL penicillin/streptomycin, 2 mmol/L l-glutamine, 10 μg/mL human transferrin, 1 μg/mL insulin, 0.1 μmol/L cortisol, 2 nmol/L T3, 0.25 μmol/L dexamethasone, 0.5 μmol/L IBMX, and 2 μmol/L rosiglitazone. The cells were then cultured without dexamethasone, IBMX, and rosiglitazone for 4 days. Small interfering RNAs (siRNAs) were introduced into differentiated MEFs by electroporation using Amaxa Nucleofector II (Lonza, Basel, Switzerland) with program A-033, according to the manufacturer instructions. The sequence of the siRNA used to target ATGL was 5′-GCACATTTATCCCGGTGTA-3′. The sequences of the siRNA used to target PPARα were 5′-GCTTCTTTCGGCGAACTAT-3′ and 5′-GCTAAAGTACGGTGTGTAT-3′. The cells were cultured for 48 h before harvesting. Triplicate samples were tested independently with MEFs isolated from three pregnant mice.

The oxygen consumption rates (OCRs) of tissue explants were measured using an XF24 Analyzer (Seahorse Bioscience). Approximately 4 mg of WAT or BAT pieces were washed with Seahorse Bioscience assay buffer containing 25 mmol/L glucose and 25 mmol/L HEPES (pH 7.4) and subsequently were cultured for 1 h in the center of a Seahorse Bioscience XF24 islet capture microplate containing 500 μL of Seahorse Bioscience assay buffer with 25 mmol/L glucose at 37°C without CO₂. Oxygen consumption was measured six times and normalized to tissue weight. For cellular Seahorse Analyzer experiments, ∼1 × 10⁴ cells/well (96-well plates) were maintained in XF assay medium supplemented with 1 mmol/L sodium pyruvate, 4 mmol/L glutamine, and 25 mmol/L glucose. The cells were subjected to a mitochondrial stress test by the addition of oligomycin (2 μmol/L), followed by carbonyl cyanide 4-(trifluoromethoxy) phenylhydrazone (FCCP; 5 μmol/L), and antimycin/rotenone (1 μmol/L/1 μmol/L).

After differentiation, adipocytes were trypsinized and cultured on coverslips for 12 h, followed by immunostaining. Briefly, the cells were incubated with MitoTracker (catalog #M7512; Thermo Fisher Scientific) before fixation, followed by permeabilization with 0.4% Triton X-100. A PMP70 antibody (catalog #ab3421; Abcam), as a peroxisome marker, and a secondary antibody (catalog #A11008; Thermo Fisher Scientific) were used for peroxisome staining. LDs and nuclei were stained with LipidTOX (catalog #H34477; Thermo Fisher Scientific) and Hoechst, respectively. For tissue immunofluorescence, deparaffinization, rehydration of paraffin-embedded sections, and antigen retrieval were required. Nonspecific antigens were blocked with 5% BSA at 4°C overnight, and the slides were incubated with PMP70 (catalog #ab3421; Abcam) and Cox4 (catalog #11967; Cell Signaling Technology, Danvers, MA) antibodies overnight at 4°C. The slides were then incubated with a secondary antibody (catalog #A11008 and #A11031; Thermo Fisher Scientific). Nuclei were stained with DAPI for 10 min. Images of immunostained cells were collected on a Nikon (Tokyo, Japan) A1R laser-scanning confocal microscope with a CFI Plan Apo 100× oil immersion objective (numerical aperture 1.45) and 405/488/561/640-nm lasers. Images of the cells after oil-red staining were collected on a Nikon Eclipse 90i system.

Microarray analysis data have been deposited in Gene Expression Omnibus with the number GSE100989. Briefly, equal amounts of total RNA from three mice were combined to form an RNA pool, and Affymetrix gene chips (GeneChip Mouse Gene 1.0 ST Array; Affymetrix, Santa Clara, CA) were used for hybridization and data collection. Quality control and statistical analyses of the complete microarray data were conducted using R/Bioconductor. The genes with altered expression (≥1.5-fold) were mapped to biological pathways using DAVID Bioinformatics Resources 6.8 (https://david.ncifcrf.gov/summary.jsp). The heat map presented in Supplementary Fig. 4 was generated as a selective profile of peroxisome-related genes. The expression of the represented genes was confirmed by real-time PCR or Western blotting. Related PCR primers and antibodies are listed in Supplementary Tables 4 and 5, respectively.

WAT and BAT samples were collected from mice and cut into small pieces. The tissues were washed three times with DMEM and then cultured in DMEM for 1 h. The medium was collected, and the glycerol level was determined using a glycerol determination kit (catalog #F6428; Sigma-Aldrich).

The statistical data reported include results from at least three biological replicates. All results are expressed as the mean ± SEM. All statistical analyses were performed in GraphPad Prism Version 5 (GraphPad Software, San Diego, CA). Significance was established using one-way ANOVA with Tukey multiple-comparison test (Figs. 1B–O, 2, 3A–C and F–H, 4, 5D and I, and 6 and Supplementary Figs. 1, 2, 5A and E, and 6C and D) and a two-tailed Student t test (Figs. 5C and H and 7 and Supplementary Figs. 3B, 4, 5B, C, F, and G, 6E, and 7). We used two-way repeated-measures ANOVA to evaluate the data presented in Figs. 1A and 3D, E, I, and J, and Supplementary Fig. 3C and D. In all cases, differences were considered significant at P < 0.05. P values are indicated in each figure as *P < 0.05, **P < 0.01, and ***P < 0.001.

To investigate the roles of Cidea and Cidec in lipid metabolism, we compared the metabolic phenotypes of ob/ob, ob/ob/Cidea^−/−, ob/ob/Cidec^−/−, and ob/ob/Cidea^−/−/Cidec^−/− mice. Growth performance, tissue weight, and serum parameters were similar between ob/ob and ob/ob/Cidea^−/− mice (Fig. 1 and Supplementary Fig. 1). As a lipodystrophy model, ob/ob/Cidec^−/− mice showed increased liver and BAT weights, with higher serum TAG, HAase, and laminin levels (Fig. 1H, N, and O and Supplementary Fig. 1D). Interestingly, ob/ob/Cidea^−/−/Cidec^−/− mice did not develop lipodystrophy, which occurred in ob/ob/Cidec^−/− mice. Compared with ob/ob and ob/ob/Cidea^−/− mice, ob/ob/Cidea^−/−/Cidec^−/− mice exhibited an improved metabolic status, including a low body weight, high oxygen and energy consumption, low levels of serum NEFAs and inflammatory cytokines (TNF-α and IL-6), and high adiponectin levels (Fig. 1). In addition, ob/ob/Cidea^−/−/Cidec^−/− mice had the lowest BAT and WAT weights, whereas the liver weights of these mice were similar to those of ob/ob and ob/ob/Cidea^−/− mice (Fig. 1C and Supplementary Fig. 1C and D). No significant differences in food intake per mouse were observed among these four types of genetically engineered mice, but food intake per gram of body weight was significantly increased in ob/ob/Cidec^−/− and ob/ob/Cidea^−/−/Cidec^−/− mice (Fig. 1B and Supplementary Fig. 1B).

As the key metabolic organs and CIDE-expressing tissues, adipose tissues and the liver were selected for further study. The ob/ob/Cidea^−/− mice had moderately lower TAG levels in BAT and liver than did the ob/ob mice, but no differences were observed in WAT (Fig. 2A–C). The ob/ob/Cidec^−/− mice failed to store TAG in WAT, with a much lower TAG content in WAT, but higher TAG levels were observed in the BAT and liver of ob/ob/Cidec^−/− mice than in the other three types of mice (Fig. 2A–C). In contrast, ob/ob/Cidea^−/−Cidec^−/− mice had obviously reduced TAG levels in BAT and WAT and maintained liver TAG levels similar to those of ob/ob mice (Fig. 2A–C). Consistent with these findings, compared with the LD size of ob/ob mice, Cidea deficiency (ob/ob/Cidea^−/−) resulted in slightly smaller LDs in BAT and liver (Fig. 2D–H). Compared with the LD size of the ob/ob and ob/ob/Cidea^−/− mice, Cidec deficiency (ob/ob/Cidec^−/−) caused dramatically smaller LDs in WAT but larger LDs in BAT and liver (Fig. 2D–H). Compared with the LD size of ob/ob and ob/ob/Cidea^−/− mice, double deficiency of Cidea and Cidec (ob/ob/Cidea^−/−Cidec^−/− mice) caused strikingly smaller LDs in both BAT and WAT (Fig. 2D–H). The performance of wild-type, Cidea^−/−, Cidec^−/−, and Cidea^−/−/Cidec^−/− mice fed an HFD were also examined, and the observed tendency was similar to that of mice with an ob/ob genetic background, including WAT and BAT performance (Supplementary Fig. 2). Increases in liver weight and hepatic TAG levels were observed in Cidea^−/−/Cidec^−/− mice compared with those of wild-type mice.

Next, we assessed the blood glucose levels and insulin sensitivity status of ob/ob/Cidea^−/−/Cidec^−/− mice compared with those of ob/ob, ob/ob/Cidea^−/−, and ob/ob/Cidec^−/− mice by GTTs and ITTs when the mice were 2, 4, and 8.5 months old. The fasting serum glucose and insulin levels of ob/ob/Cidea^−/−/Cidec^−/− mice were lower than those of the other three types of mice at 2 and 4 months of age (Fig. 3A, B, F, and G). The fed serum glucose levels exhibited no differences among ob/ob, ob/ob/Cidea^−/−, and ob/ob/Cidea^−/−/Cidec^−/− mice, but the fed serum glucose level was slightly higher in 2-month-old ob/ob/Cidec^−/− mice and was significantly higher at age 4 months (Fig. 3C and H). Insulin sensitivity became complex at different ages in these four types of genetically engineered mice (Fig. 3D, E, I, and J, Supplementary Fig. 3C and D, and Supplementary Table 6). No difference in insulin sensitivity was observed between ob/ob and ob/ob/Cidea^−/− mice from 2 to 8.5 months of age. ob/ob/Cidec^−/− mice were more sensitive to insulin than ob/ob and ob/ob/Cidea^−/− mice, according to GTT curves, at 2 months of age (Fig. 3D). However, compared with ob/ob and ob/ob/Cidea^−/− mice, these mice exhibited similar insulin sensitivity when they were 4 months old (Fig. 3I and J) and developed severe insulin resistance when they were 8.5 months old, according to ITT curves (Supplementary Fig. 3D). Interestingly, compared with that of the other three types of genetically modified mice, the insulin sensitivity of ob/ob/Cidea^−/−/Cidec^−/− mice was significantly improved at 2, 4, and 8.5 months of age. Consistent with this finding, the phosphorylation status of AKT was dramatically increased in the BAT and WAT of ob/ob/Cidea^−/−/Cidec^−/− mice compared with those of ob/ob and ob/ob/Cidea^−/− mice (Supplementary Fig. 3A and B).

A microarray analysis was performed to observe transcriptional changes in the BAT, WAT, and liver of ob/ob and ob/ob/Cidea^−/−/Cidec^−/− mice. The enriched pathways of the altered genes were analyzed to identify the key upregulated pathways, focusing mainly on mitochondria, peroxisomes, fatty acid oxidation, and oxidative phosphorylation in both the BAT and WAT of ob/ob/Cidea^−/−/Cidec^−/− mice compared with those of their ob/ob counterparts (Supplementary Tables 1–3).

Further, a significant induction of mitochondrial and peroxisome-related proteins and genes, together with an increased number of mitochondria and peroxisomes, was observed in the BAT of ob/ob/Cidea^−/−/Cidec^−/− mice and in the WAT of ob/ob/Cidec^−/− and ob/ob/Cidea^−/−/Cidec^−/− mice (Fig. 4 and Supplementary Fig. 4). Consistent with this finding, dramatically increased oxygen consumption and catalase activity were observed in the BAT of ob/ob/Cidea^−/−/Cidec^−/− mice and in the WAT of ob/ob/Cidec^−/− and ob/ob/Cidea^−/−/Cidec^−/− mice (Fig. 4D–G), demonstrating increased activity of mitochondria and peroxisomes.

Primary stromal vascular cells from BAT or WAT were induced and differentiated into brown or white adipocytes, respectively, to observe changes in LDs, mitochondria, and peroxisomes (Fig. 5). Interestingly, compared with that of their wild-type counterparts, an increased number of mitochondria and peroxisomes was observed in brown and white adipocytes from Cidea^−/−/Cidec^−/− mice; this change was accompanied by smaller LDs and lower cellular TAG levels (Fig. 5A and F and Supplementary Fig. 5A, D, and E). Consistent with this finding, genes related to the activity and biogenesis of mitochondria and peroxisomes were significantly induced in Cidea^−/−/Cidec^−/− adipocytes (Supplementary Fig. 5B, C, F, and G). Similar to those of ob/ob/Cidea^−/−/Cidec^−/− mice, Cidea^−/−/Cidec^−/− adipocytes exhibited increased basal and maximal oxygen consumption (Fig. 5B, C, G, and H) and catalase activity (Fig. 5D and I). In addition, differentiated Cidea^−/−/Cidec^−/− MEF cells exhibited low cellular TAG contents and increased expression of genes related to mitochondria and peroxisomes (Supplementary Fig. 6). Overall, the three sets of data from ob/ob/Cidea^−/−/Cidec^−/− mice, differentiated stromal vascular cells from BAT or WAT, and MEF cells from Cidea^−/−/Cidec^−/− embryos all demonstrated that CIDE protein deficiency could activate mitochondria and peroxisomes. It is worth noting that Cidec deficiency was sufficient to activate mitochondria and peroxisomes in white adipocytes, but double deficiency of Cidea and Cidec was required for the activation of mitochondria and peroxisomes in brown adipocytes.

In addition to activated mitochondria and peroxisomes, smaller LDs were always observed in Cidea^−/−/Cidec^−/− adipocytes in vivo and in vitro (Figs. 2E and 5A and F). Consistent with this finding, protein levels of ATGL and CGI58 (a key lipolysis enzyme and regulator, respectively) were significantly increased in brown and white adipocytes when both Cidea and Cidec were deficient (Figs. 5E and J and 6A and B). The levels of the lipolysis products glycerol and NEFAs were also enhanced in the BAT and WAT of ob/ob/Cidea^−/−/Cidec^−/− mice (Fig. 6C–F), demonstrating increased lipolysis. Lipidomics analysis showed that nearly all species of free fatty acids were significantly increased, including fatty acids with different chain lengths and saturation degrees (Supplementary Fig. 7A and B). PPARα and peroxisome proliferator–activated receptor γ coactivator 1α, which are known to regulate mitochondrial and peroxisome biogenesis, were also dramatically induced in the BAT and WAT of ob/ob/Cidea^−/−/Cidec^−/− mice (Fig. 6A and B and Supplementary Fig. 7C and D). To determine whether PPARα and ATGL were involved in this increased β-oxidation, PPARα and ATGL were efficiently knocked down in MEFs from Cidea^−/−/Cidec^−/− mice (Fig. 7A–F). Consistent with those of ob/ob/Cidea^−/−/Cidec^−/− mice, PPARα and ATGL levels were increased in Cidea^−/−/Cidec^−/− MEFs compared with those of wild-type mice (Fig. 7A and D). Genes related to mitochondria and peroxisomes, including Cpt1, Acox1, the Acad family, and the Pex family, were significantly induced in Cidea^−/−/Cidec^−/− MEFs (Fig. 7B and C). When PPARα or ATGL was knocked down in Cidea^−/−/Cidec^−/− MEFs, this increased gene expression was blocked, indicating that the increased activity of mitochondria and peroxisomes was regulated by PPARα or ATGL (Fig. 7B, C, E, and F). Interestingly, the knockdown of PPARα did not affect the expression of ATGL (Fig. 7A). However, the knockdown of ATGL suppressed the expression of PPARα (Fig. 7D), suggesting that increased expression of PPARα in Cidea^−/−/Cidec^−/− MEFs was regulated by induced ATGL.

The key novel finding in the current study was the observation that a coordinated interplay among different organelles, including LDs, peroxisomes, and mitochondria, is mediated by the CIDE-ATGL-PPARα pathway in adipose tissues. Similar to our previous reports (12,17,20), the current study demonstrates that increased lipolysis occurs in CIDE-deficient adipocytes; this lipolysis occurred on the surface of LDs and released fatty acids. The coordination of peroxisomes and mitochondria is required to complete the oxidation of long-chain acyl-CoA, the citric acid cycle, and oxidative phosphorylation and ultimately to produce CO₂ and H₂O (27), which was demonstrated in the current study. Increased energy expenditure and increased mitochondrial activity were previously reported in Cidea^−/−, Cideb^−/−, Cidec^−/−, and ob/ob/Cidec^−/− mice (8,28). The active regulation of peroxisomes, in addition to mitochondria, by CIDE proteins is proposed in the current study. Peroxisomes and mitochondria both contain fatty acid β-oxidation machinery; however, the enzymes and reaction processes used for β-oxidation in peroxisomes and mitochondria vary significantly. Peroxisomes are responsible for the oxidation of very-long-chain fatty acids to produce shortened acyl-CoAs, acetyl-CoA, and propionyl-CoA. We observed that the key enzymes Acadvl and ACOX1, which are specific to the peroxisome for the oxidation of very-long-chain fatty acids (29), were upregulated when CIDE proteins were deficient. In addition, CIDE deficiency caused increased expression of several peroxins, which are involved in peroxisome biogenesis. For example, Pex3 and Pex16 play multiple roles in the direct targeting of peroxisomal membrane proteins (PMPs) and in the synthesis of the peroxisome membrane (30,31). Pex5 and Pex7 are responsible for the targeting of matrix proteins and the import machinery. In addition to the increased mRNA and protein expression levels of these molecular markers, immunofluorescence analysis demonstrated increased contents of mitochondria and peroxisomes in CIDE-deficient adipose tissues or cells. Moreover, increased OCRs, ATP products, and catalase activity directly supported the enhanced activity of mitochondria and peroxisomes.

The current study has demonstrated that the ATGL-PPARα pathway is necessary to activate peroxisomes and mitochondria in CIDE-deficient adipocytes. Several previous studies reported ATGL activated PPARα/δ and mitochondrial metabolism in the liver, heart, small intestine, pancreas, and skeletal muscle (32–37). The current study has revealed the involvement of the ATGL-PPARα pathway in the metabolic activity of both BAT and WAT. ATGL-deficient mice showed reduced gene expression related to oxidative phosphorylation (38) and defective lipolysis, and these mice accumulated large amounts of lipids in the heart, causing cardiac dysfunction and premature death (39,40). Similar defects in lipid and energy disorders were reported in humans with ATGL mutations (41,42). This evidence suggested that lipolysis induced by ATGL is necessary to promote the expression of PPARα/δ and its target genes. The induction of increased lipolysis by ATGL in CIDE-deficient cells could be explained from two aspects. First, CIDE proteins play a predominant role in LD fusion (8,11), and double deficiency of Cidea and Cidec led to a failure to form large LDs in adipocytes. Small LDs have an enlarged total surface area, which makes the LDs accessible to lipase attack and lipolysis. Second, the double deficiency of Cidea and Cidec significantly increased the protein level or activity of ATGL in adipocytes. Cidec was reported to physically interact with ATGL and inhibit ATGL activity and lipolysis (43). Cidec was also found to interact with early growth response protein 1 to inhibit ATGL promoter activity (44). Therefore, increased lipolysis in CIDE-deficient adipocytes was potentially regulated by the decreased size of LDs and increased ATGL enzyme activity or protein expression levels, although the mechanism by which ATGL activity was regulated remains unclear.

The current study helps to elucidate the functions of Cidea and Cidec in adipocytes. Both Cidea and Cidec are expressed in brown adipocytes, but white adipocytes specifically and abundantly express Cidec. Our data demonstrated that significant changes occurred in Cidec^−/− white adipocytes, even when Cidea was induced to some degree, indicating the predominant role of Cidec in white adipocytes. However, in the mouse model and in primary brown adipocytes, the double deficiency of Cidea and Cidec strikingly decreased cellular TAG levels and activated mitochondria and peroxisomes in brown adipocytes, demonstrating the redundant roles of Cidea and Cidec in brown adipocytes. It is widely regarded that inducing WAT browning/beiging is an effective way to alleviate obesity or prevent obesity development in mice and humans (45–48). Significant browning and increased energy consumption occurred in the WAT of ob/ob/Cidec^−/− mice, but these mice still exhibited metabolic disorder due to severe fatty liver and inactive BAT. This finding shows that BAT function should be considered even when browning is induced in WAT. The current study demonstrates the simultaneous activation of BAT and WAT in Cidea and Cidec double-deficient mice contributed to dramatically increased energy consumption, providing some additional strategies when considering obesity therapy.

Recently, a new isoform of Cidec, designated Cidec2/Fsp27β, was identified in fatty liver; this isoform contains 10 additional amino acids at the N terminus of the original Cidec/Fsp27α (49). Both Cidec/Fsp27α and Cidec2/Fsp27β are localized on the surface of LDs and suppress lipolysis (49). More recently, Cidec2/Fsp27β was reported to be expressed in BAT and to prevent LD fusion by inhibiting Cidea function (50). It is well known that CIDE proteins promote LD fusion and lipid storage (51). In the current study, we also observed two bands for Cidec in wild-type BAT; the upper band (designated Cidec2/Fsp27β) was more abundant than the lower band (designated Cidec/Fsp27α), which appeared as the only form in WAT (Supplementary Fig. 8). However, in the ob/ob BAT, Cidec/Fsp27α was more abundant than Cidec2/Fsp27β. The current study did not provide any direct evidence to show different functions of Cidec isoforms in BAT, which has been proposed previously (50).

In conclusion, the current study demonstrates that the coupled functional interplay among LDs, peroxisomes, and mitochondria is mediated by the ATGL-PPARα pathway in Cidea and Cidec double-deficient adipocytes.

Acknowledgments. The authors thank the members of P.L.’s laboratory at Tsinghua University for their helpful discussion. The authors also thank Libing Mu (Tsinghua University) for assistance with drawing the model. In addition, the authors thank Pengcheng Jiao for assistance with the Seahorse Analyzer experiments and Jinyu Wang and Huizhen Cao (all from Tsinghua University) for assistance with image processing.

Funding. This work was supported by grants from the National Natural Science Foundation of China (31430040, 31690103, and 31621063 to P.L. and 31771303 to L.Z.), from the National Basic Research Program (2018YFA0506901 and 2016YFA0502002), and from the China Postdoctoral Science Foundation (2012M520249 and 2013T60103 to L.Z.). L.Z. was supported by the Young Elite Scientists Sponsorship Program of the China Association for Science and Technology (CAST).

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

Author Contributions. L.Z. designed the study, performed most of the experiments, analyzed the data, and wrote the manuscript. M.Y. performed most of the experiments and analyzed the data. M.A. and W.W. performed some of the immunoblotting and analyzed the data. Y.L. performed the HFD experiment. J.G. performed the MEF isolation. Y.G. performed the microarray data analyses. P.L. and L.X. designed the study, analyzed the data, and wrote the manuscript. L.X. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.