Inhibition of Mitochondrial Calcium Overload by SIRT3 Prevents Obesity- or Age-Related Whitening of Brown Adipose Tissue

The prevalence of obesity has doubled in >70 countries and has continuously increased, leading to 4.0 million deaths globally (1). Obesity also promotes alterations in other intermediate risk factors such as hypertension, dyslipidemia, and glucose intolerance, etc. (2). Previous studies have shown that obesity promotes the accumulation of white adipose tissue (WAT) to stimulate immune cell infiltration, contributing to systemic metabolic dysfunction (3). Compared with WAT, the distribution of brown adipose tissue (BAT) is more specific, mainly located in cervical, supraclavicular, paravertebral, mediastinal, and perirenal regions in humans (4,5). As the amount and function of BAT in humans are decreased with aging and obesity, the decline in BAT function may also facilitate energy storage and metabolic dysfunction under these conditions (5,6). However, the molecular mechanisms that account for the reduced BAT function in obesity and its physiological implications remain elusive.

During obesity, loss of normal structure and function of BAT, also referred to as “BAT whitening,” involves mitochondrial dysfunction as featured by altered oxidative function, ultrastructure abnormalities, and increased oxidative stress (7,8). As mitochondrial Ca²⁺ plays a critical role in regulating mitochondrial activity (9), an aberrant modulation of mitochondrial calcium level, especially continuous calcium overload, increases mitochondrial reactive oxygen species (ROS) production, which is directly implicated in mitophagy, a process leading to progressive mitochondria loss (10,11). The most critical channel mediating Ca²⁺ uptake is the mitochondrial calcium uniporter (MCU) (12). It has been recently reported that the expression levels of MCU complex members were increased during obesity in adipose tissues of mice and humans (13). These findings all suggest a critical pathophysiological role of enhanced MCU-dependent mitochondrial Ca²⁺ uptake in the development of obesity; however, the underlying mechanism is still largely unknown.

Nowadays, some active natural ingredients from food, such as capsaicin, cinnamaldehyde, and menthol, have been receiving much attention as an effective antiobesity lifestyle intervention approach (14). Since we first reported that capsaicin prevented obesity in mice in a transient receptor potential vanilloid 1 (TRPV1)–dependent manner (15), emerging evidence from laboratory and clinical studies supports a role of capsaicin as an antiobesity agent (16). Activation of TRPV1 by capsaicin increased intracellular Ca²⁺ level in adipocytes, which enhances the activity of sirtuin 1 (SIRT1) by activating cytosolic AMPK. Subsequently, the enhanced SIRT1 activity promoted browning of WAT by increasing the binding activity of PR domain-containing 16 (PRDM16), a critical transcription factor promoting the differentiation and maintenance of BAT (17). However, whether capsaicin also exerts a protective role in mitochondrial calcium overload in brown adipocytes to counteract loss of BAT remains unclear.

Unlike SIRT1 that is mainly located in the nucleus, another member of the sirtuin family, SIRT3, is preferentially located in mitochondria and highly expressed in BAT (18). Knockout (KO) of SIRT3 in mice on high-fat diet (HFD) resulted in accelerated obesity and metabolic syndrome by increasing mitochondrial oxidative stress (19,20). Both activity and stability of SIRT3 were decreased by hyperacetylation during obesity and aging (21); thus, the reduction of SIRT3 expression and activity might result in the loss of BAT in these conditions. Nevertheless, whether SIRT3 is also involved in the regulation of mitochondrial calcium homeostasis and the effect of capsaicin on obesity need to be confirmed.

In this study, we aimed to determine the role of SIRT3 in the protective effect of capsaicin on obesity and aging-induced whitening of BAT and clarify the possible mechanism. Our results show that SIRT3 plays a crucial role in the maintenance of BAT and mediates the protective effect of capsaicin by repressing MCU-dependent mitochondrial Ca²⁺ overload in brown adipocytes, which might shed light on the mechanism of loss of BAT during obesity and aging.

The SIRT3-KO mice (22) were kindly provided by Professor De-Pei Liu from the Chinese Academy of Medical Sciences and Peking Union Medical College (Beijing, China). The TRPV1-KO mice (003770) were purchased from The Jackson Laboratory. All mice were on a C57BL/6 background and housed in cages at a controlled temperature (22 ± 1°C) and relative humidity (55 ± 5%) in a 12-h light/12-h dark cycle with standard laboratory chow and tap water ad libitum. At the age of 6 weeks, wild-type (WT), SIRT3-KO, or TRPV1-KO male mice were randomized into three groups and fed standard chow (normal diet [ND]; 10% kcal from fat, 70% kcal carbohydrate, and 20% kcal protein), HFD (45% kcal fat, 35% kcal carbohydrate, and 20% kcal protein), or HFD plus 0.01% capsaicin (HFCD) for 32 weeks. In another experiment, we used 2-month-old male WT or SIRT3-KO mice as young and 18- to 20-month-old male mice as aged mice, and they were fed ND or ND plus 0.01% capsaicin (CD) for 12 weeks. We also fed a small number of young WT and SIRT3-KO mice with CD for Western blot. The approximate capsaicin intake in the CD or HFCD group was 1.64 μmol/day/mouse. Given the drug dosage ratio between mouse and human is 12.3:1 (23), the converted human dose would be almost equal to that reported in a clinical trial (24). At the end of the experiment, the mice underwent physiological tests before they were sacrificed. Tissues were stored in liquid nitrogen or fixed in 10% formalin for hematoxylin and eosin (H&E) staining and ROS detection. All experimental procedures were performed in accordance with protocols approved by the institutional animal care and research advisory committee at Daping Hospital, Third Military Medical University.

The Comprehensive Laboratory Animal Monitoring System (CLAMS; Columbus Instruments) was used to measure oxygen consumption, carbon dioxide generation, energy expenditure (EE), respiratory exchange ratio (RER), and physical activity of mice as previously described (25). EE is calculated as (3.815 + 1.232 × RER) × VO₂ × 0.001 kcal/kg^0.75/h, and RER is calculated as VO₂/VCO₂, where VO₂ is the volume of oxygen consumed per hour and VCO₂ is the volume of carbon dioxide produced per hour. Because of the huge difference in body weight (BW) between mice on ND and HFD, we compared their metabolic rates by normalizing to the metabolic size, as reflected by BW^0.75 for each mouse (25).

Primary BAT stromal vascular fraction cells were isolated using collagenase digestion followed by density separation as previously described (26). Adipocytes were stimulated with different concentrations of palmitic acid (PA) (P5585; Sigma-Aldrich) at the same time as drug treatment and incubated for 48 h. The sources and doses of the drug are as follows: capsaicin (100 μmol/L) (211275; Sigma-Aldrich), AICAR (1 mmol/L) (ab120358; Abcam), compound C (40 μmol/L) (ab120843; Abcam), curcumin (10 μmol/L) (08511; Sigma-Aldrich), and MCU inhibitor Ru360 (10 μmol/L) (557440; Sigma-Aldrich). Compound C was added 1 h before capsaicin treatment to maintain a better cell status, and Ru360 was incubated for 12 h. RNA interference and plasmid-mediated overexpression were conducted using Lipofectamine 3000 according to the manufacturer’s instructions (Thermo Fisher Scientific, Waltham, MA). The cell number plated was 1 × 10⁶ per flask. The amount of nucleotide used was 5 μg, and the volume of Lipofectamine 3000 was 3.75 μL. Mouse siRNAs against Sirt3 (sc-61556), Mcu (sc-142052), AMPKα1/2 (sc-45313), Pgc-1α (sc-38885), and negative control siRNA were purchased from Santa Cruz Biotechnology. The full-length p300 expression vector was a kind gift from Professor De-Pei Liu’s laboratory. Replication-defective adenoviral vectors expressing mouse SIRT3 (Ad-SIRT3), MCU (Ad-MCU), and peroxisome proliferator–activated receptor γ coactivator-1α (Ad–PGC-1α) were generated by OBiO Technology (Shanghai) Corp., Ltd. Adipocytes were infected with the above adenovirus (multiplicity of infection of 100). Transfection of siRNA or plasmid and adenovirus infection were performed 24 h prior to treatment with drugs or PA.

The ROS levels were measured using a dihydroethidium (DHE) fluorescent probe for cytosolic ROS detection or MitoSOX Red (Thermo Fisher Scientific) for mitochondrial ROS detection as previously described (27) using a Fluoroskan Ascent Fluorometer (Thermo Fisher Scientific). To visualize the staining, the sections or specimens were placed on an inverted fluorescence microscope (TE2000; Nikon).

Mitochondrial respiratory function was analyzed using a two-channel titration injection respirometer (Oxygraph-2k; Oroboros Instruments) as previously described (28). A total of 2 × 10⁶ adipocytes were harvested, suspended in MiR05 solution, and transferred separately to oxygraph chambers. Routine respiration was assessed while respiration was stabilized, and then the plasma membrane was permeabilized with digitonin (16 μg/10⁶ cells). Complex I–dependent oxidative phosphorylation (CI OXPHOS) was measured after addition of glutamate (G), malate (M), and ADP (D). Subsequently, succinate (S) was added to induce maximal OXPHOS capacity with convergent input through CI + complex II (CII) OXPHOS. After uncoupling with 2-([4-(trifluoromethoxy) phenyl] hydrazinylidene) propanedinitrile (FCCP) in the noncoupled state, CI+II supported maximal convergent respiratory capacity of the CI+II electron transfer system (ETS) was measured. The addition of rotenone allowed the determination of CII ETS. Residual oxygen consumption was evaluated after the inhibition of complex III with anti–mycin A.

Tissue mitochondria were isolated by gradient centrifugation using a commercial kit (ab110169; Abcam) according to the manufacturer’s instructions. Isolated BAT mitochondria (50 μg protein) were resuspended in 100 μL high KCl buffer: 110 mmol/L KCl, 0.5 mmol/L KH₂PO₄, 1 mmol/L MgCl₂, 20 mmol/L HEPES, 0.01 mmol/L EGTA, 5 mmol/L succinate, and 0.004 mmol/L rotenone (pH 7.2) with 1 μmol/L Rhod-2 AM. The cytosolic and mitochondrial Ca²⁺ concentrations were measured using Fura-2 AM and Rhod-2 AM (Thermo Fisher Scientific), respectively, as previously described (29,30). Fluorescent signal was recorded at room temperature using a Varioskan Flash microplate reader (Thermo Electron Corporation).

Total RNA was extracted using TRIzol reagent (Invitrogen) according to the manufacturer’s protocol. Two micrograms total RNA was used for the first-strand synthesis with cDNA M-MuLV Reverse Transcriptase (New England Biolabs) using random primers. The QuantiTect SYBR Green RT-PCR Kit (Qiagen) was used for the amplification reactions using the three-step protocol described by the manufacturer (LightCycler 96; Roche). The fluorescence curves were analyzed using LightCycler 96 software (version 1.1). The primers are listed in Supplementary Table 1.

Tissue or cell samples were lysed in RIPA lysis buffer (65 mmol/L Tris-HCl, pH 7.5, 150 mmol/L NaCl, 1 mmol/L EDTA, 1% Nonidet P-40, 0.5% sodium deoxycholate, and 0.1% SDS) with protease inhibitor cocktail tablets (04693132001; Roche) and phosphatase inhibitor tablets (4906837001; Roche). Equal amounts of protein (20 μg/lane) were separated by SDS-PAGE and transferred onto polyvinylidene difluoride membranes (Millipore). After blocking with 0.1% Tween-20 containing 5% nonfat dry milk (Blotto; sc-2324; Santa Cruz Biotechnology), the filters were incubated with the following primary antibodies: GAPDH (ab9485; Abcam), PRDM16 (ab106410; Abcam), PGC-1α (ab54481; Abcam), SIRT3 (sc-99143; Santa Cruz Biotechnology), AMPK (#2532; Cell Signaling Technology), phosphorylated (p-)AMPK (#2535; Cell Signaling Technology), UCP1 (SAB1404511; Invitrogen), and MCU (ab121499; Abcam). After washes and incubation with the appropriate horseradish peroxidase–conjugated secondary antibody (Santa Cruz Biotechnology), the immune complexes were visualized using a chemiluminescence horseradish peroxidase substrate (WBKLS0100; Millipore).

Chromatin immunoprecipitation (ChIP) assays were carried out as previously described (31). Briefly, ∼1 × 10⁷ brown adipocytes were cross-linked with 1% formaldehyde, and nuclear extracts were sonicated on a Bioruptor Plus sonication system (Diagenode). Chromatin lysate was precleared with Dynabead protein A (Invitrogen) and subjected to immunoprecipitation with ChIP-grade antibody against normal IgG (#2729; Cell Signaling Technology), c-fos (#2250; Cell Signaling Technology), c-Jun (#9165; Cell Signaling Technology), H3 (ab1791; Abcam), H3K27ac (ab4729; Abcam), H3K27me3 (ab6002; Abcam), or PGC-1α (NBP1–04676; Novus Biologicals). The DNA–protein complex was precipitated with Dynabead protein A (Invitrogen), eluted in washing buffers, and treated with proteinase K and RNase A in turn to reverse cross-links. DNA was purified and analyzed by quantitative RT-PCR (LightCycler 96; Roche) with primers that targeted interesting DNA sequences. Primer sequences were listed in Supplementary Table 1.

Quantitative results are expressed as the means ± SD. The differences among three or more groups were analyzed using one-way ANOVA followed by the Bonferroni adjustment for multiple comparisons. Two-way ANOVA was used for BW, CLAMS, and intraperitoneal glucose tolerance test (IPGTT) data with repeated measurements. Graphs were created using Prism 6.0 (GraphPad Software), and statistical analysis was performed with GraphPad Prism. A P value <0.05 was considered to be statistically significant.

The data sets and noncommercial materials generated and/or analyzed during the current study are available from the corresponding author upon reasonable request.

To investigate whether SIRT3 participated in the process of BAT whitening, we first detected the expression levels of SIRT3 in all types of WAT and BAT in mice and found that the expression of SIRT3 was much higher in BAT (Fig. 1A and B). After 4-month HFD feeding, both mRNA and protein levels of SIRT3 in BAT were significantly reduced (Fig. 1C and D), whereas in WAT, the expression of SIRT3 was almost unchanged (Supplementary Fig. 1A and B). Thus, we next determined whether the antiobese effect of capsaicin was dependent on SIRT3. To this end, 2-month-old TRPV1- and SIRT3-KO mice were fed with HFD for 8 months. As expected, mice on HFD displayed an obvious BW gain compared with those on ND, and capsaicin significantly slowed down the increase of BW and accumulation of WAT mass in HFD-fed mice in a TRPV1-dependent manner (Supplementary Fig. 1C–F). We also observed that SIRT3-KO mice showed a faster weight gain process under HFD, and even in the ND group, SIRT3 KO also resulted in an obvious weight gain without affecting their food intake amount (Fig. 1E–G and Supplementary Fig. 1F). More importantly, the antiobese effect of capsaicin almost totally disappeared in SIRT3-KO mice in both the ND group and the HFD group, indicating that SIRT3 is indispensable for the inhibitory effect of capsaicin on obesity (Fig. 1E–G). Consistently, the protective effect of capsaicin treatment against high fat–impaired glucose tolerance was blocked by KO of SIRT3 (Fig. 1H). Similar to TRPV1-KO mice, although the weight of BAT was almost equal in all groups (Fig. 1I and Supplementary Fig. 1G), H&E staining showed that capsaicin could significantly inhibit the whitening of BAT in mice on HFD, as evidenced by the reduction of lipid droplet size and elevation of lipid droplet number per field of view (Fig. 1J–L). KO of SIRT3 obviously promoted lipid droplet accumulation in BAT and blocked the inhibitory effect of capsaicin on HFD-induced BAT whitening (Fig. 1J–L). However, although KO of SIRT3 erased the preventive role of capsaicin in WAT accumulation, it failed to directly increase the lipid droplet size in WAT (Supplementary Fig. 1H–J), emphasizing a fundamental role of SIRT3 in the maintenance of BAT structure. These results suggest that the inhibitory effect of capsaicin on BAT whitening is dependent on the presence of SIRT3.

Next, the consumption of oxygen, generation of carbon dioxide, RER, and EE of all groups of mice were assayed. The EE and physical activity of mice on HFD were significantly reduced, along with the decrease of O₂ consumption and CO₂ production, especially at nighttime, when the mice were more active, suggesting a decline of BAT function (Fig. 2A and B and Supplementary Fig. 2A–D). Capsaicin significantly increased EE, O₂ consumption, CO₂ production, and physical activity in these mice, implying that capsaicin might increase BAT activity and aerobic respiratory function; however, all of these effects disappeared in SIRT3-KO mice (Fig. 2A and B and Supplementary Fig. 2A–D). Also, we assessed ROS levels in the frozen slices of BAT from these mice. The results showed that HFD remarkably induced an increase of total ROS level measured by DHE in BAT, which was significantly suppressed by capsaicin treatment (Fig. 2C and D). However, the ROS generation in the BAT of HFCD-fed SIRT3-KO mice was similar to that of HFD-fed mice, indicating that the suppressive effect of capsaicin on high fat–induced ROS generation was impaired by SIRT3 KO (Fig. 2C and D). The results of MitoSOX staining also demonstrated a similar pattern (Fig. 2C and D). In addition, the promotional effect of capsaicin on the expression of BAT-related markers in BAT of HFD-fed mice, including PRDM16, PGC-1α, and UCP-1, was also significantly reduced in SIRT3-KO mice (Fig. 2E and Supplementary Fig. 2E). In parallel, mitochondrial respiration and oxygen consumption of the isolated brown adipocytes from BAT were measured. Measurement showed that mitochondrial respiration, which was represented by CI and CII proton leak (LEAK), CI, CII, and CI+CII OXPHOS, as well as CI+CII ETS, was markedly reduced by HFD and elevated by capsaicin (Fig. 2F). Again, all of these parameters were almost equal in the BAT of HFCD- and HFD-fed SIRT3-KO mice (Fig. 2F), suggesting a crucial role of SIRT3 in the beneficial effect of capsaicin on improving mitochondrial function in BAT.

As excessive mitochondrial Ca²⁺ level is a chief culprit of mitochondrial dysfunction (32), we tested Ca²⁺ uptake in mitochondria isolated from BAT of mice on HFD. As expected, mitochondria isolated from BAT of HFD-fed mice showed higher Ca²⁺ uptake ability than those on ND (Fig. 3A). In addition, these changes were reversed by dietary capsaicin (Fig. 3A). However, KO of SIRT3 failed to affect the overall Ca²⁺ intake of primary brown adipocytes from mice, but significantly increased the level of Ca²⁺ influx in mitochondria, especially in mice fed HFD (Fig. 3B and C). Accordingly, the inhibitory effect of capsaicin on HFD-induced mitochondrial Ca²⁺ overload in BAT was also blocked in SIRT3-KO mice (Fig. 3A and C). To further determine the direct effect of SIRT3 on mitochondrial Ca²⁺ level, the expression level of SIRT3 was manipulated by its siRNA- or Ad-mediated overexpression vector in primary brown adipocytes from WT mice. The brown adipocytes incubated with small interference (si-)Sirt3 displayed a noticeable increased mitochondrial Ca²⁺ uptake, which was more obvious after treatment with PA, without affecting cytosolic Ca²⁺ level or Ca²⁺-related signaling component calmodulin-dependent protein kinase II (Fig. 3D and Supplementary Fig. 3A–C). Also, si-Sirt3 significantly attenuated the inhibitory effect of capsaicin on mitochondrial Ca²⁺ uptake and blocked the promotional effect of capsaicin on mitochondrial respiration (Fig. 3D and Supplementary Fig. 3D). In contrast, Ad-SIRT3 inhibited mitochondrial calcium overload induced by PA in brown adipocytes (Fig. 3E and Supplementary Fig. 3E). Accordingly, cells treated by si-Sirt3 displayed a higher mitochondrial ROS level, whereas Ad-SIRT3 exerted a reversed effect (Supplementary Fig. 3F and G). These results imply that SIRT3 exerts its beneficial effect on the maintenance of BAT function through inhibiting mitochondrial Ca²⁺ overload.

We next examined the expression of some ion channels mainly controlling mitochondrial calcium intake, such as MCU and MICU1 (32), in BAT of mice. HFD significantly increased the expression of MCU and MICU1 in BAT, while other channels, such as MCUb and MICU2, were not affected (Fig. 4A and B). Moreover, SIRT3 KO directly increased the expression of MCU and MICU1 in BAT, and this trend was more obvious under HFD (Fig. 4A and B). Capsaicin displayed an inhibitory effect on the expression of MCU and MICU1, which almost disappeared in SIRT3-KO mice (Fig. 4A and B), suggesting that SIRT3 was required for the inhibition of capsaicin on mitochondrial calcium overload.

As MICU1 acts as a gatekeeper restricting MCU-mediated calcium influx (33), the expression of MCU was chosen to be knocked down by si-Mcu in primary SIRT3-KO brown adipocytes or si-Sirt3–treated WT cells. It was found that si-Mcu not only completely blocked the increase of mitochondrial calcium uptake caused by SIRT3 KO or si-Sirt3 in brown adipocytes (Fig. 4C and D and Supplementary Fig. 4A and B) but also reduced the excessive ROS generation and improved mitochondrial activity in these cells (Supplementary Fig. 4C–E), even though knockdown or inhibition of MCU alone also impaired basal mitochondrial activity, probably due to insufficient mitochondrial Ca²⁺ level (Supplementary Fig. 4F), suggesting that the increased MCU expression was a main reason for mitochondrial calcium overload and dysfunction in BAT of SIRT3-KO mice. In contrast, the inhibitory effects of Ad-SIRT3 on PA-induced mitochondrial calcium overload and dysfunction were also blocked by Ad-MCU (Fig. 4E and Supplementary Fig. 5). However, neither overexpression nor knockdown of MCU failed to affect the expression of SIRT3 (Fig. 4F), indicating that MCU acts as a downstream target of SIRT3.

To further elucidate the mechanism underlying the increased MCU expression in SIRT3-KO mice, we examined the promoter region of mouse Mcu and Micu1 genes and found potential AP-1 binding sites. ChIP results indicate that KO or knockdown of SIRT3 could not directly increase AP-1 binding on these regions, but it can be enhanced by capsaicin intervention (Supplementary Fig. 6A and B). As capsaicin actually reduces the expression of MCU and MICU1, we switched to examine the histone modification status and found a high acetylation level of histone lysine 27 on the promoters of Mcu and Micu1. SIRT3 KO or si-Sirt3 significantly increased H3K27ac in these regions, with a decrease of H3K27me3, while capsaicin intervention exerted an opposite effect (Fig. 5A and B). These results suggest that acetylation of H3K27 might play a key role in the transcription of Mcu and Micu1.

Next, as p300 is a major histone acetylase accounting for H3K27ac, we treated brown adipocytes with an inhibitor of p300 histone acetyltransferase, curcumin (34). Curcumin potently inhibited the elevation of Mcu or Micu1 expression by SIRT3 knockdown and suppressed the promotional effect of PA on elevation of mitochondrial calcium uptake, accompanied by a decrease of H3K27ac level on the Mcu or Micu1 gene promoter (Fig. 5C–E and Supplementary Fig. 6C and D). In contrast, the inhibitory effect of capsaicin on Mcu or Micu1 expression was blocked in adipocytes transfected with a p300 overexpression plasmid (Supplementary Fig. 6E and F). This evidence indicates that the increased H3K27ac level on the promoter of Mcu or Micu1 was an important factor stimulating their transcription.

The activity of p300 has been reported to be reduced by AMPK (35), a critical metabolic regulator, which could be activated by SIRT3. Thus, we explored whether AMPK mediated the promotional effect of SIRT3 KO on MCU expression. AICAR, an activator of AMPK, almost blocked the promotional effect of si-Sirt3 on both Mcu expression and the increased H3K27ac level on its promoter (Fig. 5F and G), whereas compound C, an agent inhibiting AMPK activity, or siRNA-mediated knockdown of the catalytic subunits of AMPK, AMPKα1/2, erased the inhibitory effect of capsaicin on these changes (Supplementary Fig. 6G–J). These results suggest that inactivation of AMPK by knockdown of SIRT3 might account for the increased H3K27ac level on the promoter of Mcu, which leads to overexpression of Mcu.

Next, we explored the mechanism of the declined SIRT3 expression in BAT of mice on HFD. The expression of SIRT3 has been reported to be activated by PGC-1α (36), a downstream target of AMPK (37,38). As expected, HFD reduced AMPK phosphorylation and PGC-1α expression in BAT of mice, and capsaicin intervention significantly inhibited these effects (Fig. 6A). Similarly, the inhibitory effect of PA on AMPK activity and PGC-1α expression could also be reversed by capsaicin (Fig. 6B). Both activation of AMPK by AICAR and adenovirus-mediated overexpression of PGC-1α upregulated the expression of SIRT3 in the presence of PA (Fig. 6C and D and Supplementary Fig. 7A and B). In addition, either PGC-1α siRNA or inhibition of AMPK by compound C abrogated capsaicin-induced SIRT3 expression, and PGC-1α siRNA also erased the promotional effect of AICAR on stimulating SIRT3 expression (Fig. 6E–G and Supplementary Fig. 7C and D), indicating that PGC-1α acts as a downstream target of AMPK to activate SIRT3 expression. Moreover, ChIP results also indicated that there existed an obvious positive binding signal of PGC-1α on the promoter of the Sirt3 gene, which could be reduced by PA and enhanced by capsaicin or AICAR treatment (Fig. 6H). Thus, the reduction of AMPK/PGC-1α signaling pathway by HFD might account for the lowered SIRT3 expression in BAT, and it also suggests that there exists a positive-feedback loop between AMPK and SIRT3 in BAT, which could be activated by capsaicin.

Finally, in order to verify whether capsaicin also inhibits aging-induced loss of BAT by activating SIRT3, WT and SIRT3-KO aged mice were fed with CD for 3 months. Compared with the young mice, aged mice possessed a higher BW with lower food intake, and KO of SIRT3 resulted in an obvious increase of BW without affecting food intake (Fig. 7A–C), accompanied by a further impaired glucose tolerance (Fig. 7D). In addition, aging did not lower BAT weight but caused a significant whitening process in BAT with elevated size and decreased number of lipid droplets, which was even more obvious in SIRT3-KO mice (Fig. 7E–G). Accordingly, aged mice consumed less O₂ and produced less CO₂, and their EE was also lowered (Fig. 7H and Supplementary Fig. 8). Moreover, capsaicin intervention also had beneficial effects on all of the above-mentioned parameters to counteract the influence of aging, but the effects were much weaker than those in young mice (Fig. 7A–H and Supplementary Fig. 8). Also, these effects were almost blocked in SIRT3-KO mice (Fig. 7A–H and Supplementary Fig. 8). These results indicate that reduced SIRT3 expression facilitates the process of BAT whitening during aging, which could be partially reversed by capsaicin. Similar to mice receiving HFD, the BAT of the aged mice also displayed overall repressed expression of SIRT3, p-AMPK, and PGC-1α, but much higher MCU expression, which could be significantly reversed by capsaicin treatment. Compared with their littermates, the expression of p-AMPK and PGC-1α in the BAT of SIRT3-KO mice was further downregulated, with overexpression of MCU (Fig. 7I and Supplementary Fig. 9). As expected, the regulatory effects of capsaicin on these gene expression changes in BAT were almost eradicated by KO of SIRT3 (Fig. 7I and Supplementary Fig. 9). Consistently, mitochondrial calcium uptake in the primary brown adipocytes from aged mice was obviously elevated, and capsaicin inhibited this age-related mitochondrial calcium overload in a SIRT3-dependent manner (Fig. 7J). Together, these results support that aging can increase mitochondrial calcium overload by inhibiting the expression of SIRT3 in BAT, resulting in the whitening and loss of BAT.

In this work, we provide evidence to prove that the decreased SIRT3 expression in BAT during obesity and aging is an important reason for BAT whitening. The inhibition or KO of SIRT3 upregulates MCU expression in an AMPK-dependent epigenetic manner and via exaggerated MCU-mediated mitochondrial calcium uptake, resulting in more ROS generation and promoting whitening of BAT. Capsaicin intervention can promote the expression of SIRT3 by activating AMPK to inhibit mitochondrial calcium overload in brown adipocytes, thus resisting the whitening of BAT (Fig. 7K).

Unlike WAT, the origin of brown adipocytes in BAT is very similar to that of skeletal muscle cells, and PRDM16 is required for the differentiation and maintenance of mature brown adipocytes (39). As KO of SIRT3 directly reduced the expression of PRDM16 and blocked the promotional effect of capsaicin on its expression, the exaggerated BAT whitening caused by SIRT3 KO would also be related to the inhibition of differentiation of brown adipocytes. In addition, as the weight of BAT was not affected by KO of SIRT3, the increased BW in SIRT3-KO mice on both ND and HFD mainly depended on the accumulation of WAT—not BAT whitening. However, as KO of SIRT3 directly facilitated the whitening of BAT and blocked the antiobese effect of capsaicin, the exaggerated BAT whitening process would be critical to HFD- or age-related obesity by slowing down the metabolic rate of the whole body that helps promote the accumulation of WAT. Thus, the enhanced whitening of BAT caused by SIRT3 KO could be considered an important initial step for HFD- or aging-induced obesity. In addition, the data on physical activity support an alternative explanation that capsaicin activation of physical activity through regulation of neuronal or muscle functions as the primary driver for the phenotypes of elevated EE. Browning of adipose tissue could also be a result of muscle activity, which might be also regulated by SIRT3 (40,41).

Loss of mitochondria is the main feature of BAT whitening. Enhanced expression of SIRT3 alleviates ROS accumulation (42), stabilizes the mitochondrial membrane (43), and promotes mitochondrial fusion to sustain the number of mitochondria (44), thus exerting a fundamental protective role in mitochondrial homeostasis. In this study, we found that the role of SIRT3 in maintaining mitochondrial homeostasis is not confined to the mitochondria itself but also regulates the ionic homeostasis and epigenetic modification in the whole cell through its downstream target, AMPK. Interestingly, we also noticed that capsaicin failed to totally block age-related BAT whitening, while in young mice, it exerted a more potent antiobesity role against HFD, possibly due to the inability of capsaicin to restore SIRT3 expression to the level of young mice in aged mice, thus further emphasizing that the reduced SIRT3 is the key factor leading to BAT whitening.

As a highly conserved cellular energy sensor regulating health and longevity, AMPK is activated by a low-energy state (45,46). Besides directly phosphorylating PGC-1α, AMPK also promotes SIRT1-mediated deacetylation of PGC-1α, both of which lead to activation of PGC-1α (37,38). In accordance with our findings, as a downstream target of PGC-1α (36), SIRT3, in turn, also activates the AMPK/PGC-1α pathway (47,48). Our results also confirm that AMPK is indispensable for not only the inhibitory effect of SIRT3 on MCU expression but also the promotional effect of capsaicin on SIRT3 expression. Moreover, activation of AMPK by capsaicin requires the participation of SIRT3. These results suggest that these two proteins form a positive-feedback loop, which could act as a link between the energy state of cells and the mitochondrial function. In addition, it also explains how brown adipocytes sense the excessive energy status caused by overfeeding or aging, which leads to mitochondrial dysfunction and clearance.

Increasing evidence has shown that the imbalance of intracellular Ca²⁺ has an important effect on adipocytes. A transient rise in cytosolic Ca²⁺ inhibits triglyceride accumulation and lipid storage, whereas sustained high levels of Ca²⁺ inhibit lipolysis (32). Consistently, we and others (15,49) have discovered that adipocytes isolated from obese humans possess an elevated cytosolic Ca²⁺ level, with a decreased ability of transient uptake of extracellular Ca²⁺. Similarly, transient mitochondrial Ca²⁺ boosts mitochondrial bioenergetics and activates the citric acid cycle (9), but permanent calcium overload increases ROS level and stimulates mitophagy (10,11). These phenomena may be related to AMPK activity, because AMPK phosphorylation could be activated when the cytosolic Ca²⁺ is transiently elevated, whereas this Ca²⁺-activated AMPK activity is also blocked by persistent high intracellular Ca²⁺ level (50). This mechanism may protect cells from apoptosis and autophagy but facilitates fat accumulation and insulin resistance. In a normal state, although an increase of mitochondrial Ca²⁺ inhibits the activity of SIRT3 (51), SIRT3 also retrospectively restricts mitochondrial Ca²⁺ overload by activating AMPK to form a negative-feedback loop. However, during obesity or aging, the activity of AMPK could be reduced by sustained elevation of cellular Ca²⁺ level and not only reduces the expression of SIRT3 but also blocks the inhibitory effect of SIRT3 on MCU, thus aggravating mitochondrial Ca²⁺ overload. The current study emphasizes an epigenetic mechanism of AMPK-dependent regulation of mitochondrial Ca²⁺ homeostasis and suggests that the regulation of epigenetic modification might be potentially applied in the intervention of metabolic diseases.

In summary, we found that capsaicin could activate the AMPK-SIRT3 positive-feedback loop to epigenetically inhibit MCU-dependent mitochondrial Ca²⁺ overload in brown adipocytes, thus maintaining the morphology and function of BAT against the whitening stimuli. This work clearly indicates a central role of mitochondrial Ca²⁺ homeostasis in BAT function and suggests that intervention in the AMPK/SIRT3 pathway may be an effective target for preventing obesity- or age-related metabolic diseases.

Acknowledgments. The authors thank Tingbing Cao from Daping Hospital, Third Military Medical University, for technical assistance. The authors also thank De-Pei Liu and Hou-Zao Chen from the Institute of Basic Medical Sciences, Chinese Academy of Medical Sciences, and Peking Union Medical College (Beijing, China) for providing the SIRT3-KO mice.

Funding. This work was supported by grants from the National Natural Science Foundation of China (81570761, 31871199, and 81721001), National Key Research and Development Project (2018YFA0800601), and Innovative Research Team in University (IRT1216).

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

Author Contributions. P.G. and Y.J. designed the experiments, analyzed data, and wrote the paper. P.G. and Y.J. performed experiments with contributions from H.W., F.S., H.H., B.W., Z.L., Y.H., X.W., Y.C., C.H., and L.W. P.G., Y.J., H.W., F.S., Y.L., H.H., B.W., Z.L., Y.H., X.W., Y.C., C.H., L.W., H.Z., G.Y., D.L., Z.Y., and Z.Z. read and approved the manuscript. P.G., Y.L., H.Z., G.Y., D.L., Z.Y., and Z.Z. critically read and revised the paper. P.G., Z.Y., and Z.Z. initiated the project. Z.Y. 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 accuracy of the data analysis.