SIRT5 Regulates Brown Adipocyte Differentiation and Browning of Subcutaneous White Adipose Tissue

There has been an increase in interest in adipocytes accompanied by the obesity epidemic and the realization that adipose tissue plays a vital role in metabolic regulation (1). Generally, there are two distinct adipose tissues in mammals, namely white adipose tissue (WAT) and brown adipose tissue (BAT). WAT is mainly composed of white adipocytes that function in energy storage and release, whereas BAT is mainly composed of brown adipocytes that can dissipate energy via adaptive nonshivering thermogenesis. Notably, some WAT depots such as subcutaneous fat depots, contain another type of adipocyte that have similar traits to brown adipocytes: the so-called beige adipocytes (2–5). It is conceivable that brown and beige adipocytes play a major role in the regulation of energy balance by enhancing energy expenditure. They have been intensively studied due to their potential in therapies against obesity and type 2 diabetes (6,7).

Adipocyte differentiation is a broadly studied area in adipose biology and is regulated by a complex network of transcription factors, cofactors, and other signaling factors, including peroxisome proliferator–activated receptor γ (PPARγ) and the C/EBP family members (C/EBPα, C/EBPβ, and C/EBPδ) (8,9). The nuclear receptor PPARγ is considered the master regulator of adipocyte differentiation and is required for adipose tissue development (10,11). Brown adipocyte differentiation needs other crucial factors to drive the brown-characteristic program. In particular, the zinc finger protein PRDM16 determines and controls brown fat development by interacting with other transcription factors (12–14). In addition, PRDM16 promotes browning of subcutaneous WATs by activating a brown fatlike gene program (15). Given the importance of PRDM16 in brown and beige fat development, various factors regulate brown/beige fat formation and thermogenesis through their impact on PRDM16.

Significant attention has been paid to epigenetic regulators in adipocyte development. Several function in brown adipocyte development, such as the histone methyltransferase EHMT1 and the acetyltransferase GCN5 (16,17). Sirtuins, which are class III Histone deacetylases, are an evolutionarily conserved family of NAD-dependent lysine deacetylases that play important roles in the regulation of aging, tumorigenesis, energy metabolism, and stress resistance (18–20). Among the seven members of sirtuins, several participate in the regulation of adipocyte differentiation or thermogenesis. For instance, SIRT1 promotes browning of WAT by deacetylating PPARγ (21). SIRT2 inhibits adipocyte differentiation by deacetylating FOXO1 and enhancing its repressive interaction with PPARγ (22). SIRT3 regulates mitochondrial function and thermogenesis in brown adipocytes by enhancing the expression of the uncoupling protein PCG-1α (23). SIRT6 is essential for adipogenesis because it regulates mitotic clonal expansion (24). SIRT7 promotes adipogenesis in the mouse by inhibiting autocatalytic activation of SIRT1 (25). However, whether SIRT5 plays a role in adipocyte differentiation remains unknown.

SIRT5 is a relatively unique member of the sirtuins family; it is located primarily in mitochondrial and was originally implicated in the urea cycle by activating carbamoyl phosphate synthetase 1 (26). SIRT5 modifies lysine succinylation, malonylation, and glutarylation both inside and outside of the mitochondria and affects multiple enzymes involved in diverse mitochondrial metabolic pathways (27–30). It is easy to imagine that SIRT5 has crucial effects on cellular metabolite flux. Because most chromatin-modifying enzymes use cellular metabolites as substrates or cofactors to modify both histones and DNA, alteration of metabolite concentrations could affect the activities of chromatin-modifying enzymes (31). It has been proposed that a subtle alteration in cellular metabolite levels may control the differentiation or thermogenesis of brown and beige fat (32). AMPK is essential for BAT development through the elevation of a key metabolite of the tricarboxylic acid cycle, α-ketoglutarate (α-KG), which facilitates ten-eleven translocation (TET)–mediated DNA demethylation of the Prdm16 promoter, committing precursor cells to brown adipogenic differentiation (33). Notably, SIRT5 can desuccinylate and deglutarylate IDH2 and glucose-6-phosphate dehydrogenase, respectively, thereby activating both enzymes to maintain cellular NADPH homeostasis and redox potential during oxidative stress (34). As IDH2 catalyzes the oxidative decarboxylation of isocitrate and produces α-KG, it would be intriguing to study whether SIRT5-mediated α-KG alteration affects brown adipocyte differentiation.

Based on the above clues and given that SIRT5 is expressed at higher levels in BAT than WAT (23), we hypothesize that SIRT5 plays an important part in brown adipocyte formation and thermogenesis. This study shows that SIRT5 is essential for brown adipocyte differentiation in vitro and has an impact on browning of subcutaneous adipose tissue in vivo.

For in vitro experiments, the lentiviral SIRT5 shRNA plasmid (m) (sc-63027-SH) was purchased from Santa Cruz Biotechnology for knockdown assay. The CDS sequence of mouse SIRT5 was cloned into the pLV-GFP-C1 viral expression plasmid for overexpression assays. For gene and protein expression analysis, Western blot and quantitative real-time PCR assay were performed as previously described (35). Primer sequences are shown in Supplementary Table 1. Details for antibodies used in this study are available in the Supplementary Data.

For adipocyte differentiation, confluent C3H10T1/2 cells were treated with inducing regents for 2 days, including 1 nmol/L triiodothyronine (T3), 0.5 mmol/L 3-isobutyl-1-methylxanthine (IBMX), 125 nmol/L indomethacin, 1 μmol/L dexamethasone, 850 nmol/L insulin, and 1 μmol/L rosiglitazone. Then, cells were changed to the growth medium supplemented with 1 nmol/L triiodothyronine, 850 nmol/L insulin, and 1 μmol/L rosiglitazone. The medium was changed every 2 days. Differentiated cells at day 8 were performed further analysis. Differentiated cells were stained with Oil Red O (Sangon Biotech) and visualized by microscope. Cellular respiration assay was performed using an XFe analyzer (Seahorse Bioscience).

For quantification of α-KG, differentiating cells at day 4 were resuspended by 80% methanol aqueous solutions. Approximately 50 mg tissues was weighted and homogenized in PBS. Metabolite content was determined by liquid chromatography–tandem mass spectrometry, and the assay was performed by the Servier Laboratories. Briefly, 100 μL homogenates was added into a plate followed by adding 50 μL O-benzylhydroxylamine hydrochloride (O-BHA) solution (1 mol/L) and 50 μL N-(3-dimethylaminopropyl)-N′-ethylcarbodiimide (EDC) solution (1 mol/L). The plate was put in a shaking incubator for 1 h to complete the derivative reaction. Then, 300 μL ethyl acetate was added into the mixture and incubated for 20 min. After centrifugation at 1,900 rpm for 5 min, the plate was placed at −80°C for 40 min until the liquid on the lower layer was frozen. Then, all of the liquids were extracted on the upper layer to another plate, and the samples were dried under the Organomation. The samples were suspended in 150 μL MeOH/H₂O solution. The above samples were centrifuged at 4,000 rpm for 10 min and analyzed by liquid chromatography–tandem mass spectrometry. α-KG standard was from Sigma-Aldrich (#K3752).

For determination of IDH activity, cell lysates were collected by RIPA buffer without SDS. Reaction substrates were diluted by assay buffer containing 25 mmol/L Tris-HCl (pH 7.0), 1 mmol/L MgCl₂, 25 mmol/L NaCl, and 0.025% BSA. The final reaction concentration of each substrate was 75 μmol/L. The reaction was triggered by adding 25 μL substrate diluted mixture into 25 μL cell lysate in a 384-well plate. The NADPH-producing fluorescence (excitation 355 nmol/L and emission 460 nmol/L) was detected by an EnVision microplate reader.

The histone extraction assay was performed as previously described (35).

Differentiating cells at day 4 were collected for chromatin immunoprecipitation (ChIP) assay. The assay was performed using an ab500 ChIP kit (Abcam) and carried out according to the product manual. All enrichment changes were normalized to the input. H3K9 methylation changes were also normalized to total histone H3. The 18s rRNA was used for nonspecific binding sites. Primer sequences are shown in Supplementary Table 2.

For the TET hydroxylase activity assay, differentiating cells at day 4 were collected for preparing nuclear extracts. Nuclear protein was extracted by using a nucleoprotein extraction kit (C500009; Sangon Biotech). Then, the assay was performed using a TET hydroxylase activity quantification kit (ab156912; Abcam). TET activity was calculated using the formulas provided in the kit instructions.

Differentiating cells at day 4 were collected for the methylated DNA immunoprecipitation assay. The assay was performed according to protocols described in chapter 5 of the book DNA Methylation: Methods and Protocols (36). Primer sequences are listed in Supplementary Table 2.

For in vivo experiments, all animal experiments were performed according to procedures approved by the Animal Ethics Committee of the Shanghai Institute of Materia Medica. SIRT5 knockout (KO) mice were a gift from Prof. David Lombard, University of Michigan (Ann Arbor, MI), and were bred in-house. All animals were housed in a temperature-controlled room (22 ± 2°C) with a 12-h light/dark cycle. Body composition was measured using nuclear magnetic resonance spectroscopy (Bruker). For cold exposure, mice were single-caged and exposed to 4°C for either 6 h or 2 days. Rectal temperature was monitored using a microprobe digital thermometer (Physitemp Instruments) at indicated time points. Whole-body energy expenditure and O₂ consumption were measured using a TSE Phenomaster caging system (TSE Systems). For the glucose tolerance test, mice were injected intraperitoneally with glucose (2.5 g/kg) after 6-h fasting.

Data were plotted using GraphPad Prism software and are presented as the mean ± SEM. Statistical comparisons between groups were assessed by two-tailed Student t test, except rescue experiments and animal experiments, in which a one-tailed distribution was used, and Fig. 7I and J, in which Tukey multiple comparison was used. A P value <0.05 was considered statistically significant.

Cold exposure is a classic model to induce browning. To explore the potential physiological role of SIRT5, we placed 8-week-old male mice at 30°C or 10°C for 10 days and then tested the expression levels of sirtuins in BAT and subcutaneous inguinal WAT (IngWAT) with browning capacity. Ten-day cold exposure markedly upregulated the thermogenic gene Ucp1 expression in both BAT and IngWAT (Fig. 1A). Importantly, Sirt5 expression was significantly upregulated in BAT after cold exposure and showed higher expression than other sirtuins (Fig. 1B). Moreover, in subcutaneous IngWAT, only Sirt5 expression was significantly increased after cold exposure (Fig. 1C). These observations suggest that SIRT5 may play an important role in adaptive thermogenesis of brown/beige adipocytes.

Next, we directly investigate the biological role of SIRT5 in brown adipocyte differentiation in vitro. Lentiviruses expressing shRNAs targeting SIRT5 were transduced into C3H10T1/2 multipotent mesenchymal cells. Cells were induced to undergo brown adipocyte differentiation thereafter. We found that knockdown of SIRT5 prior to the induction of differentiation resulted in a severe brown adipocyte differentiation defect in C3H10T1/2 cells. Lipid accumulation was significantly reduced in SIRT5 knockdown cells (Fig. 2A). In addition, basal and uncoupled oxygen consumption rates were both significantly reduced in SIRT5 knockdown cells (Fig. 2B). Furthermore, not only expression levels of the common adipogenic makers Fabp4, Pparγ, and the C/ebp family but also BAT-selective genes were remarkably downregulated, including Prdm16, Pgc-1α, Cox7a1, and Cox8b (Fig. 2C). Consistently, we observed a reduction in protein expression levels of adipogenic and thermogenic markers (Fig. 2D). We established stable knockdown cell lines to support the lentiviral experiments. The stable SIRT5 knockdown cell line also showed defective brown adipocyte differentiation, as evidenced by reduced lipid accumulation and low expression levels of adipogenic and thermogenic genes (Supplementary Fig. 1A and B). Furthermore, we used a GFP-fused expression vector to construct a SIRT5 overexpression plasmid and packaged lentiviruses to transfect C3H10T1/2 cells. Notably, the expression levels of PPARγ, PRDM16, and UCP1 were significantly augmented in the SIRT5 overexpression cells (Fig. 2E), which was consistent with the increased lipid accumulation (Supplementary Fig. 1C). We further examined whether ectopically expressed SIRT5 could rescue the phenotypes in SIRT5 knockdown cells. As expected, UCP1 expression was significantly increased in SIRT5 overexpression groups, and overexpression of SIRT5 diminished the difference of UCP1 expression between SIRT5 knockdown cells and scramble control cells (Supplementary Fig. 1D and E).

To identify the causes of the differentiation defect in SIRT5 knockdown cells, we further examined the expression of adipogenic markers during differentiation. We found that the induction of the master adipogenic transcription factors (Pparγ and C/ebpα) and a brown cell fate driver gene (Prdm16) were significantly blocked in SIRT5 knockdown cells on day 2, while the activation of other regulators was unaltered in the early stage of differentiation, including C/ebpβ, C/ebpδ, and Pgc-1α (Fig. 2F–I and Supplementary Fig. 1F–H). Protein levels of PPARγ, C/EBPα, and PRDM16 were also significantly downregulated in the later stage of differentiation (Fig. 2J). Moreover, expression of thermogenic protein UCP1 was also downregulated in the later stage of differentiation (Fig. 2J). Collectively, these results confirm that SIRT5 is required for brown adipocyte differentiation and suggest that SIRT5 functions by affecting the activation of the brown adipogenic gene program.

SIRT5 can affect IDH2 enzyme activity through the succinylation modification (34). IDH2 catalyzes the oxidative decarboxylation of isocitrate and produces α-KG. To further explore the mechanisms by which SIRT5 functions in brown adipocyte differentiation, we examined alterations of α-KG concentration in SIRT5 knockdown cells. As expected, the IDH activity was markedly attenuated in SIRT5 knockdown cells (Fig. 3A). In accordance with decreased IDH activity, α-KG concentration was significantly reduced in SIRT5 knockdown cells (Fig. 3B).

To further determine whether reduced α-KG concentration was responsible for impaired brown adipocyte differentiation in SIRT5 knockdown cells, we tried to rescue this phenotype by supplementing α-KG. A membrane-permeable α-KG, dimethyl-α-KG, was added during the entire course of differentiation. As shown by Oil Red O staining, enhanced differentiation was observed in SIRT5 knockdown cells, even though ectopic supplementation of α-KG was not sufficient to fully recover the differentiation deficiency (Fig. 3C). Furthermore, mRNA expression levels of Pparγ, C/ebpα, Prdm16, and Ucp1 were significantly increased in SIRT5 knockdown cells when supplemented with α-KG (Fig. 3D). Although protein expression of markers such as PPARγ and UCP1 in SIRT5 knockdown cells did not recover to those in control cells, the expression of PRDM16 was comparable to that in control cells (Fig. 3E and F). These results indicate that the alteration of α-KG at least partially contributes to the differentiation deficiency resulting from knockdown of SIRT5.

Among various chromatin-modifying enzymes, histone and DNA demethylases such as the JMJD histone demethylases and the TET family of 5mC hydroxylases use α-KG as a cofactor (37,38). Moreover, when progenitor cells differentiate, JMJD histone demethylases remove repressive histone methylation marks (H3K9me3 and H3K27me3) and activate the expression of differentiation-related genes (31). As α-KG concentration was significantly reduced in SIRT5 knockdown cells, we proposed that the activities of histone demethylases were downregulated. To explore reasonable changes to histone modifications that blocked the activation of major brown adipogenic regulators, we examined methylations on repressive H3K9 and H3K27 marks in SIRT5 knockdown cells. Strikingly, we found that levels of repressive histone marks H3K9me2 and H3K9me3 were both augmented in SIRT5 knockdown cells (Fig. 4A and B), whereas methylations on H3K27 were not significant changed; this suggests that H3K9me2 and H3K9me3 are involved in repressing expression of brown adipogenic genes.

To further confirm the roles of altered H3K9me2 and H3K9me3 in repressing expression of master brown adipogenic genes in SIRT5 knockdown cells, we performed ChIP assays. The ChIP results revealed that binding levels of H3K9me2 and H3K9me3 at the promoter regions of Pparγ and Prdm16 were significantly increased, whereas they were unaltered at the promoter region of C/ebpα (Fig. 4C–H). In addition, binding levels of RNA polymerase II at the promoter regions of Pparγ and Prdm16 were decreased, while binding was unchanged at the promoter region of C/ebpα (Fig. 4I–K). In contrast, although α-KG concentration can also impact the activities of the TET family of 5mC hydroxylases, knockdown of SIRT5 had no effect on DNA methylations at the promoter regions of Pparγ and Prdm16 (Supplementary Fig. 2A–D), which was correlated with unchanged Tet expression and TET activity in the nuclei (Supplementary Fig. 2E and F). The above findings indicate that knockdown of SIRT5 results in enrichment of H3K9me2 and H3K9me3 at the promoter regions of Pparγ and Prdm16, which inhibited the expression of Pparγ and Prdm16, thereby leading to a differentiation defect.

Because supplementation of α-KG enhanced the mRNA and protein expression levels of PPARγ and PRDM16 in SIRT5 knockdown cells, we examined whether binding levels of H3K9me2 and H3K9me3 at the promoter regions of Pparγ and Prdm16 were restored. ChIP assays showed that binding levels of H3K9me2 and H3K9me3 were significantly increased at the promoter regions of Pparγ and Prdm16 in SIRT5 knockdown cells, consistent with the data in Fig. 4, and α-KG treatment diminished the enrichment, as expected (Fig. 5). The binding levels of H3K9me2 and H3K9me3 at the promoter region of C/ebpα were unchanged in SIRT5 knockdown cells consistent with data in Fig. 4, but α-KG treatment also decreased the enrichment of H3K9me3 at the promoter region of C/ebpα loci (Supplementary Fig. 3). These data support the conclusion that the differentiation deficiency in SIRT5 knockdown cells is correlated with reduced α-KG concentration and enrichment alterations of H3K9me2 and H3K9me3 at the promoter regions of Pparγ and Prdm16.

Although SIRT5 KO mice do not show any apparent metabolic abnormalities (39), we still wondered whether SIRT5 deficiency has physiological functions in vivo comparable to those observed in cell culture. Sv129 mice contain higher amounts of multilocular UCP1-positive adipocytes than C57BL/6J mice at 28°C (40,41). We thus explored the adipose-related phenotype of SIRT5 KO mice on the Sv129 background.

Strikingly, we found that SIRT5 KO mice exhibited more adiposity, as evidenced by more fat mass and higher proportions of WAT, including IngWAT, epididymal WAT, and perirenal WAT, whereas body weight showed no difference compared with age-matched wild-type (WT) mice (Fig. 6A–C). Consistent with those data, we found that the subcutaneous IngWAT of SIRT5 KO mice contained larger adipocytes compared with age-matched WT mice (Fig. 6D and E). mRNA expression levels of thermogenic genes (Ucp1, Cidea, and Cox7a1) and fatty acid oxidation genes (Cpt1b and Mcad) were significantly downregulated in subcutaneous IngWAT of SIRT5 KO mice, though the expression levels of adipogenic genes (Pparγ and C/ebpα) were not altered (Fig. 6F). Analysis of protein levels revealed markedly reduced expression of thermogenic protein UCP1 in IngWAT of SIRT5 KO mice (Fig. 6G and H). However, SIRT5 KO mice showed no difference in the morphology or UCP1 content of BAT (Supplementary Fig. 4A and B), nor in the mRNA expression levels of BAT-selective thermogenic genes and other markers, including brown fat–specific genes, fatty acid oxidation genes, and adipogenic genes (Supplementary Fig. 4C). The expression of thermogenic protein UCP1 was also unchanged (Supplementary Fig. 4D and E).

In contrast, considering the function of adipose tissue in the regulation of whole-body energy expenditure, we further observed slightly decreased energy expenditure and oxygen consumption in SIRT5 KO mice (Supplementary Fig. 5A and B), but no difference in the respiratory exchange ratio compared with age-matched WT mice (Supplementary Fig. 5C). By contrast, the locomotor activity was higher in SIRT5 KO mice (Supplementary Fig. 5D). Given that brown and beige fat contribute significantly to the regulation of glucose homeostasis (42), we performed a glucose tolerance test and found that SIRT5 KO mice exhibited impaired glucose tolerance (Supplementary Fig. 5E and F).

Collectively, the above observations suggest that SIRT5 deficiency in mice causes less browning capacity in subcutaneous IngWAT and a slight imbalance in energy and glucose homeostasis.

Cold exposure induces the formation of beige adipocytes to produce heat to keep warm. Given that SIRT5 KO mice showed less browning capacity in subcutaneous IngWAT, we wondered whether SIRT5 KO mice exhibited intolerance under cold conditions. As expected, SIRT5 KO mice showed a blunted response to cold stimuli, revealed by a markedly lower core temperature following 6 h at 4°C, which suggested that these mice had a certain defect in activating adaptive thermogenesis (Fig. 7A). Furthermore, we observed fewer multilocular beige adipocytes and UCP1 content in subcutaneous IngWAT of SIRT5 KO mice (Fig. 7B and C). Moreover, mRNA expression levels of thermogenic genes were markedly reduced (Fig. 7D), and protein expression levels of PRDM16, PPARγ, and UCP1 were also significantly reduced in subcutaneous IngWAT of SIRT5 KO mice (Fig. 7E–H). Nevertheless, no differences were detected in the thermogenic function of BAT after cold exposure (Supplementary Fig. 6). We further found that α-KG levels of both IngWAT and BAT in WT mice under cold stress were markedly augmented relative to room temperature, indicating that α-KG level is relevant to cold stress (Fig. 7I and J). Of note, the α-KG level of IngWAT in SIRT5 KO mice was significantly lower than WT mice under cold stress (Fig. 7I), whereas the α-KG level of BAT was unchanged (Fig. 7J), consistent with the normal browning capacity in BAT. This implies that SIRT5 is required for the cold regulation of α-KG production in IngWAT. Taken together, these findings confirm the speculation that SIRT5 deficiency in mice leads to cold intolerance, likely due to a defect in activating the browning process in subcutaneous IngWAT under cold conditions.

SIRT5 is expressed broadly in tissues, including brain, heart, liver, skeletal muscle, and BAT, and it is expressed more abundantly in BAT than in WAT (23). It has not been reported that SIRT5 plays a role in adipose tissues yet.

In this study, we demonstrate that SIRT5 was required for brown adipocyte differentiation. We first speculate that SIRT5 played an important role in adaptive thermogenesis, because we found Sirt5 expression was significantly increased in BAT and subcutaneous IngWAT after cold exposure. In vitro, we further discovered that knockdown of SIRT5 impaired brown adipocyte differentiation and blocked expression of adipogenic regulators, whereas overexpression of SIRT5 promoted the expression of brown adipogenic regulators such as PPARγ and PRDM16. Therefore, we propose that SIRT5 plays a role in brown adipocyte differentiation by affecting critical gene expression.

SIRT5 regulates three new types of posttranscriptional modifications including succinylation, malonylation, and glutarylation. In particular, potential substrates for SIRT5 include multiple enzymes involved in cell metabolism (43). Several lines of evidence have revealed the close linkage between cellular metabolites and cell differentiation (33,44,45). Notably, SIRT5 can desuccinylate and activate IDH2, which catalyzes the oxidative decarboxylation of isocitrate and produces α-KG (34). Based on these clues, we found repressed IDH activity and reduction of α-KG concentration in SIRT5 knockdown cells, which was consistent with previous reports. Furthermore, to get insight on whether reduced α-KG is involved in the regulation of brown adipocyte differentiation in SIRT5 knockdown cells, we supplemented α-KG to recover differentiation deficiency. We found that supplementation of α-KG rescued brown adipocyte differentiation. The ectopic supplementation of α-KG was not sufficient to fully recover the differentiation deficiency, suggesting that the reduced α-KG did not entirely explain the phenomenon. Some alternative mechanisms need to be further investigated.

Because histone demethylases, such as the JMJD histone demethylases, use α-KG as a cofactor, it is conceivable that alteration of α-KG concentration could affect the activities of histone demethylases. Histone modifications, in particular lysine methylation, play critical roles in regulating gene expression (46,47). Histone methylations are correlated with adipocyte differentiation by modulating the expression of principal adipogenic regulators (48,49). Some JMJD histone demethylases regulate adipocyte development. For instance, histone H3K9 demethylase JMJD2B facilitates adipocyte differentiation by regulating H3K9 methylation on Pparγ and C/ebpα during adipocyte differentiation (50).

Of note, in our study, consistent with the reduced α-KG level, we observed elevated H3K9me2 and H3K9me3 in differentiating SIRT5 knockdown cells, whereas H3K27 methylation levels were not changed. ChIP assays revealed that the levels of H3K9me2 and H3K9me3 at the promoter regions of Pparγ and Prdm16 were significantly increased compared with controls, which was in accordance with the repressed expression levels of Pparγ and Prdm16. Moreover, α-KG supplementation diminished the enrichment of H3K9me2 and H3K9me3 at the promoter regions of Pparγ and Prdm16 in SIRT5 knockdown cells, which was consistent with the enhanced expression levels of Pparγ and Prdm16. However, it is still unclear which histone demethylases were responsible for the elevated H3K9me2 and H3K9me3 marks. Additionally, knockdown of SIRT5 affected the enrichment of H3K9me2 and H3K9me3 at the promoter regions of Pparγ and Prdm16, but without affecting their binding levels at the promoter region of C/ebpα. There is no doubt that some specific factors exist that can recruit certain histone demethylases to selective gene loci, and these remain to be identified.

To investigate whether SIRT5 deficiency has physiological functions in vivo, we studied a SIRT5-deficient mouse model on the Sv129 background. SIRT5 deficiency does not lead to any remarkable metabolic abnormalities on the C57BL/6J background (39). Yu et al. (39) studied metabolic phenotypes of SIRT5 KO mice on the C57BL/6J background under chow-fed and high-fat diet conditions. In their study, body composition, whole-body heat production, oxygen consumption, and locomotor activity were similar between KO and control mice, whereas we found that fat mass was significantly increased, whole-body heat production and oxygen consumption were decreased, and locomotor activity was increased in Sv129-background KO mice. The cold tolerance test in C57BL/6J-background mice showed a similar tolerance between KO and control mice, whereas Sv129 background KO mice showed obvious cold intolerance. The intraperitoneal glucose tolerance test in C57BL/6J-background mice showed a similar response between KO and control mice, while Sv129-background KO mice exhibited impaired glucose tolerance. Therefore, we conclude that our results in Sv129-background mice differ from the previous data in C57BL/6J-background mice. These differences may have been caused by the higher amounts of multilocular UCP1-positive adipocytes in Sv129 than C57BL/6J mice. Moreover, in our study, SIRT5 KO mice on the Sv129 background exhibited more adiposity and less browning capacity in subcutaneous IngWAT in vivo. Furthermore, the α-KG level of IngWAT in KO mice was significantly lower than control mice under cold stress (Fig. 7I), indicating that SIRT5 is required for the cold regulation of α-KG production in IngWAT. However, the α-KG level of BAT in KO mice was not different compared with control mice either at room temperature or under cold stress (Fig. 7J), which aligns with the phenomenon that SIRT5 deletion had little influence on thermogenic capacity in BAT in vivo. To further explore the difference between these two adipose tissues, we separated the stromal vascular fraction (SVF) cells of BAT and those of IngWAT and subjected them to brown adipogenesis. As shown in Supplementary Fig. 7, thermogenic genes in differentiated IngWAT-SVF cells from KO mice were significantly downregulated (Supplementary Fig. 7A), which was consistent with observations in short hairpin SIRT5 (shSIRT5) C3H10T1/2 cells and the browning defect in IngWAT of KO mice. In contrast, Ucp1 expression in differentiated BAT-SVF cells from KO mice was upregulated (Supplementary Fig. 7B). In addition, Sirt3 expression was unchanged in IngWAT of KO mice (Supplementary Fig. 7C), whereas it was markedly upregulated in BAT of KO mice (Supplementary Fig. 7D). Because SIRT3 is highly expressed in BAT and regulates thermogenic function by enhancing PCG-1α expression, we speculated that the compensatory change of SIRT3 in BAT could partly explain why the thermogenic defect in BAT of SIRT5 KO mice disappeared. It would be intriguing to clarify the cross talk between SIRT5 and SIRT3 in BAT. In fact, because a whole-body KO may cause potential defects in other tissues that can affect adipose tissue in turn, our results should be verified in conditional KO mice by crossing conditional floxed KO mouse strains with adipose-specific Cre mice.

In summary, our work provides new insight into the biological function of SIRT5. Our current data support an important role of SIRT5 in the regulation of brown adipocyte differentiation in vitro and the browning process in vivo. Moreover, SIRT5 links cellular metabolism to adipocyte development via an indirect impact on histone modifications, thereby extending the known role of epigenetics in brown adipocyte differentiation. Considering SIRT5 has multiple targets in metabolic pathways, it may serve as a plausible pharmacological target for the treatment of obesity and other metabolic disease via improving metabolic processes.

Funding. This work was supported by the National Natural Science Foundation of China (grants 81125023 and 81470166), National Program on Key Research Project (2016YFC1305505), Shanghai Commission of Science and Technology (16JC1405000), and the K.C. Wong Education Foundation.

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

Author Contributions. L.S. contributed to the design and conduction of experiments, data collection and analyses, discussion, and writing of the manuscript. L.-N.Z. and B.-H.L. contributed to the design and conduction of experiments, data analyses, and discussion. C.-L.T. contributed to animal introduction and breeding and experimental assistance. L.-Y.W. contributed to experimental assistance. J.L. and J.-Y.L. contributed to research design and reviewed the manuscript. J.-Y.L. is the guarantor of this work and, as such, had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.