The scaffold protein p62 regulates adaptive thermogenesis through ATF2 nuclear target activation

During β-adrenergic stimulation of brown adipose tissue (BAT), p38 phosphorylates the activating transcription factor 2 (ATF2) which then translocates to the nucleus to activate the expression of Ucp1 and Pgc-1α. The mechanisms underlying ATF2 target activation are unknown. Here we demonstrate that p62 (Sqstm1) binds to ATF2 to orchestrate activation of the Ucp1 enhancer and Pgc-1α promoter. P62Δ69-251 mice show reduced expression of Ucp1 and Pgc-1α with impaired ATF2 genomic binding. Modulation of Ucp1 and Pgc-1α expression through p62 regulation of ATF2 signaling is demonstrated in vitro and in vivo in p62Δ69-251 mice, global p62−/− and Ucp1-Cre p62flx/flx mice. BAT dysfunction resulting from p62 deficiency is manifest after birth and obesity subsequently develops despite normal food intake, intestinal nutrient absorption and locomotor activity. In summary, our data identify p62 as a master regulator of BAT function in that it controls the Ucp1 pathway through regulation of ATF2 genomic binding. Beta-adrenergic stimulation of brown adipose tissue leads to thermogenesis via the activating transcription factor 2 (ATF2) mediated expression of the thermogenic genes Ucp1 and Pgc-1α. Here, the authors show that the scaffold protein p62 regulates brown adipose tissue function through modifying ATF2 genomic binding and subsequent Ucp1 and Pgc-1α induction.

H aving long been considered dispensable, brown adipose tissue (BAT) in adult humans is now recognized today as an organ of physiological relevance for energy metabolism [1][2][3] . In adult humans, the mass and activity of BAT inversely correlates with body mass index (BMI) and percent body fat [4][5][6] , and sympathetic nervous system (SNS) activation by cold exposure [7][8][9] , or by β 3 -adrenergic receptor agonism 10 enhances BAT thermogenesis and energy expenditure. The mitogen-activated protein kinase (MAPK) p38α, the key regulator of adaptive thermogenesis, modulates uncoupling protein 1 (Ucp1) and peroxisome proliferator-activated receptor gamma coactivator 1-alpha (Pgc-1α) expression via phosphorylation, and action of the activating transcription factor 2 (ATF2) 11,12 . Upon activation by p38α, p-ATF2 translocates to the nucleus and initiates expression of Ucp1 and Ppargc1a (Pgc-1α) via binding to cAMP responsive elements (CRE) in the Ucp1 enhancer and Pgc-1α promoter 11 . Highlighting the key role of this pathway in UCP1 action, coldinduced β-adrenergic receptor stimulation fails to promote Ucp1 expression in mouse BAT primary cells pretreated with the p38α/β MAPK inhibitor SB202190 12 . While nuclear entry and action of ATF2 is crucial for BAT adaptive thermogenesis 11,12 , the mechanism underlying ATF2 target activation is unknown 13 .
Scaffold proteins are key mediators of selective and efficient cell signal transduction, which they achieve through direct and specific interaction with their target proteins. The scaffold protein p62 (sequestosome 1; Sqstm1) is a multimodular adaptor protein involved in key metabolic processes like tissue inflammation, cell differentiation, cell growth, and tumorigenesis 14 . Mice with global 15 or adipose-specific 16 deletion of p62 have a severe obese phenotype with normal food intake but decreased energy expenditure and impaired BAT function. Global p62 −/− mice also have enhanced adipogenesis with hyperphosphorylation of the extracellular signalregulated kinase (ERK1/2) in the white adipose tissue (WAT) 15 . Loss of adipogenic capacity by ERK1/2 deletion can prevent obesity in global p62 −/− mice 17 , emphasizing uncertainty as to whether the dysregulated energy metabolism in p62-deficient mice originates from enhanced adipogenesis and/or impaired energy expenditure 16 . The aim of this paper was to dissect the molecular foundations underlying energy metabolism control by p62 and to identify the molecular mechanisms of how p62 regulates BAT thermogenesis. Our data demonstrate that mice that lack the amino acids (aa) 69-251 of the p62 protein (p62 Δ69-251 mice) have normal protein levels of p-ERK1/2 in WAT and show no changes in adipogenesis or adipocyte differentiation. However, these mice develop a severe obese phenotype that is accompanied by impaired energy expenditure and dysfunctional BAT. In a series of in vitro and in vivo experiments using p62 Δ69-251 mice, global p62-deficient mice (p62 −/− ) and Ucp1-Cre p62 flx/flx mice, we demonstrate that p62 is a key signaling node of the UCP1 pathway. p62 directly binds to ATF2 to orchestrate its genomic binding to and activation of the Ucp1 enhancer and Pgc-1α promoter. As demonstrated in p62 Δ69-251 mice, global p62 −/− mice and Ucp1-Cre p62 flx/flx mice, lack of p62 action leads to failure of ATF2 to activate its nuclear targets Ucp1 and Pgc1α and results in impaired BAT function and increased body weight. The cell autonomous effect of p62 to modulate ATF2 nuclear target activation is confirmed in BAT primary cells from p62 Δ69-251 mice and is verified in cultured BAT cells of global p62 −/− mice and Ucp1-Cre p62 flx/flx mice. Our data establish p62 as a key regulator of adaptive thermogenesis in that it regulates the UCP1 pathway via modulation of ATF2 genomic target activation.

Results
Generation of p62 Δ69-251 mice. To dissect the role of p62 in regulating systems metabolism, we generated mice in which the amino acids 69-251 of the p62 protein have been deleted (p62 Δ69-251 mice). Mice were bred on the C57BL/6J background and were designed to yield a truncated p62 protein of ∼37 kDa that lacks the zinc finger (zz) domain, the TB1 domain and one of the two p38 interacting motifs, but to otherwise maintain regular p62 function ( Supplementary Fig. 1a). Consistent with this, we see no difference in protein levels of phosphorylated protein kinase C (p-PKC) in the liver (Supplementary Fig. 1b) and of p-ERK1/2 in WAT of p62 Δ69-251 mice ( Supplementary Fig. 1c, d). Also, hepatic protein levels of microtubule-associated protein 1 light chain 3 (LC3) are, as expected, unchanged in p62 Δ69-251 mice (Supplementary Fig. 1e), which is consistent with demonstration of preserved p62 binding to LC3 and to p38 using immunoprecipitation analysis in HEK293FT cells ( Supplementary Fig. 1f). Notably, preserved p62 binding to p38 is not unexpected given that only one of the two p38 binding motifs is deleted in the p62 Δ69-251 mice ( Supplementary Fig. 1a) and no difference in p38/p-p38 is observed in BAT of p62 Δ69-251 mice and global p62 −/− mice ( Supplementary Fig. 1g). We further see, as expected, no difference in autophagy, as assessed by hepatic protein level of ATG7 ( Supplementary Fig. 1h), or the proteasome activity in BAT of p62 Δ69-251 mice ( Supplementary Fig. 1i). Collectively, these data indicate that the p62 motifs not affected by the p62 Δ69-251 mutation remain functional.
p62 Δ69-251 mice are obese with no difference in adipogenesis. p62 Δ69-251 mice develop an obese phenotype of identical dimension relative to global and adipose-specific p62 −/− mice when fed with either regular chow or a high-fat diet (HFD) (Fig. 1a, Supplementary Fig. 2). Since p-ERK1/2 is not changed in WAT of either 12-or 32-wk-old p62 Δ69-251 mice ( Supplementary  Fig. 1c, d), ERK1/2 mediated adipogenesis can thus not explain the obese phenotype of the p62 Δ69-251 mice. Consistent with this, we see adipocyte size and volume increased but the total number of adipocytes decreased in the iWAT, eWAT, and rpWAT of young 14-wk-old non-obese p62 Δ69-251 mice (Supplementary Fig. 3a-d). In none of the analyzed white fat depots (Supplementary Fig. 4), nor in iWAT or BAT primary cells of p62 Δ69-251 mice ( Supplementary Fig. 5c-f, i-l), do we see major expression differences indicative of enhanced adipogenesis or adipocyte differentiation (Pparg, Fasn, Fabp4, Adipoq). This is further supported by unaltered Oil Red O staining during differentiation of BAT and iWAT primary cells ( Supplementary Fig. 5a,b,g,h). These data collectively show that obesity in p62 Δ69-251 mice does not result from changes in adipogenesis.
Obesity in chow-fed p62 Δ69-251 mice (Fig. 1a) is primarily the result of increased body fat with only a slight (but significant) increase in lean tissue mass (Fig. 1b), and is accompanied by glucose intolerance (Fig. 1c), elevated fasting levels of insulin (Fig. 1d) and insulin resistance (Fig. 1e). The increased adiposity is not related to changes in body length (Fig. 1f) or to circulating levels of growth hormone (Fig. 1g), nor to alterations in food intake (Fig. 1h) or the amount of nutrients that are made available for metabolism (Fig. 1i, j). Since intake and absorption of nutrients is unchanged in the p62 Δ69-251 mice, these data suggest that obesity in p62 Δ69-251 mice results from alterations in energy expenditure. dark phase activity when exposed to either 10 or 5°C (Fig. 2c). Consistent with failure of the obese p62 Δ69-251 mice to increase their energy expenditure when exposed to cold (Fig. 2b), we see a blunted increase in energy expenditure when p62 Δ69-251 mice are treated with norepinephrine (NE) (Fig. 2d). These data suggest that the impaired energy expenditure of p62 Δ69-251 mice resides from inadequate activation of the BAT and that the p62 Δ69-251 mice increase their locomotor activity upon cold exposure to compensate for BAT dysfunction. Consistent with this, BAT glucose uptake, measured by positron emission tomography/ magnetic resonance imaging (PET/MRI) following administration of 18 F-FDG, is reduced in p62 Δ69-251 mice (Fig. 2e) despite normal activity of cytochrome c oxidase (COX) and thus functional mitochondrial complex IV (Fig. 2f). The impaired BAT function of p62 Δ69-251 mice is associated with decreased expression of the key thermogenic genes Ucp1, Pgc-1α, diodinase 2 (Dio2), and cell death-inducing DNA fragmentation factor alpha subunit-like effector A (Cidea) in BAT (Fig. 2g). Expression levels Pgc-1α also decreased in iWAT of obese p62 Δ69-251 mice with very low to borderline expression of Ucp1 in both groups (Fig. 2h). Notably, despite~50% decreased expression of Ucp1 and~80% decreased expression of Pgc-1α in the BAT (Fig. 2g), we see protein levels of the upstream Ucp1 and Pgc-1α key activator p-ATF2 (Thr69/71) increased in BAT of obese p62 Δ69-251 mice, with unchanged levels of p38 and p-p38 (Fig. 2i). Since p-ATF2 induces expression of Ucp1 and Pgc-1α via direct binding to CRE elements in the promoter/enhancer regions of these genes 11 , these data suggest that p-ATF2 requires a critical p62 motif either for nuclear translocation, or for binding and activating the CRE elements of the Ucp1 enhancer and Pgc-1α promoter.
Impaired BAT function in young non-obese p62 Δ69-251 mice. If p62 is required for adequate BAT function, then lack of p62 signaling should already affect BAT thermogenesis before the onset of obesity. In line with this notion, we see BAT temperature and whole-surface temperature already decreased in 7-d-old   p62 Δ69-251 pups relative to wt controls, as assessed by infrared thermography (Fig. 3a). Consistent with this, in 12-wk-old nonobese p62 Δ69-251 mice we see, despite yet unchanged body weight (Fig. 3b), energy expenditure already decreased (Fig. 3c). Similar to the obese p62 Δ69-251 mice, locomotor activity is unchanged under room temperature also in the young non-obese p62 Δ69-251 mice (Fig. 3d), but expression of key thermogenic genes Ucp1, Pgc-1α, PR/SET domain 16 (Prdm16), Dio2, and Cidea are decreased in BAT (Fig. 3e). These data imply that the abrogated BAT function is a direct result of p62 deficiency and is not a consequence of the obese phenotype. Consistent with this, gene signatures indicative of decreased thermogenesis (Ucp1, Pgc-1α, Prdm16) are confirmed also in BAT and iWAT primary cells harvested from young non-obese p62 Δ69-251 mice (Supplementary Fig. 6a-h).
generated mice with targeted deletion of p62 specifically in the BAT (Ucp1-Cre p62 flx/flx ). In BAT of these mice, expression of p62 is decreased by~75% relative to wt controls (Fig. 4a). Despite showing 25% residual expression of p62 in BAT, these Ucp1-Cre p62 flx/flx mice show decreased BAT surface temperature after birth (Fig. 4b, c) and show a greater body weight gain relative to wt controls (Fig. 4d, e). Notably, while mice with global p62 deletion 15 , WAT/BAT-specific p62 deletion 16 , global p62 Δ69-251 mutation ( Fig. 2a, Supplementary Fig. 2a,d), and BAT-specific p62 deletion (Fig. 4) all become obese with impaired BAT function, mice with CNS-specific p62 deletion do not become obese 16 . These data collectively demonstrate that the metabolic dysfunction residing in p62 deficiency originates from p62 signaling in the BAT and is independent of p62 signaling in the CNS. Consistent with the p62 Δ69-251 mice, also the increased body weight of the Ucp1-Cre p62 flx/flx mice is not accompanied by changes in food intake (Fig. 4f), but expression of key thermogenic genes (Ucp1, Pgc-1α, Cidea) are decreased in in vitro cultured brown adipocytes of the Ucp1-Cre p62 flx/flx mice (Fig. 4g-i). In line with these data, isoproterenol fails to induce Ucp1 mRNA levels in BAT primary cells of the Ucp1-Cre p62 flx/flx mice (Fig. 4j).
p62 regulation of BAT function is cell autonomous. We also assessed energy expenditure using Seahorse analysis and see basal respiration and proton leak respiration decreased in BAT primary cells of p62 Δ69-251 mice (Fig. 5a, b). In agreement with our observation that p-ATF2 is increased and expression of Ucp1 decreased in the p62 Δ69-251 mice (Fig. 2g, i), we see increased levels of p-ATF2, yet lower protein and transcript levels of Ucp1 also in BAT primary cells of p62 Δ69-251 mice; even when cells are stimulated with isoproterenol ( Fig. 5c, d). These in vitro data again suggest that p-ATF2 requires functional p62 to access or activate the CRE elements in the promoter/enhancer region of Ucp1 and Pgc-1α, and they consequently further demonstrate that the effects of p62 on BAT function are cell autonomous and are not driven by CNS effects.
p62 regulates ATF2 nuclear target activation. We next tested whether p62 is a direct ATF2 interaction partner and whether p62 is required either for ATF2 nuclear translocation or for binding to and activation of its nuclear targets. Immunoprecipitation analysis in HEK293T cells showed direct binding of p62 to ATF2 (Fig. 6a) without ability of p62 to directly activate (phosphorylate) ATF2 (Fig. 6b). Notably, p62 binding to ATF2 is observed under both baseline conditions and after stimulation with isoproterenol and is preserved even in the p62 Δ69-251 mutants (Fig. 6a). We next tested whether p62 is required for nuclear translocation of ATF2 and quantified the cytosolic and nuclear content of ATF2. In line with the observation that p62 binding to ATF2 is preserved in p62 Δ69-251 mutant cells (Fig. 6a), nuclear presence of p-ATF2 is preserved in BAT of the p62 Δ69-251 mice under both baseline conditions and after β-adrenergic receptor stimulation (Fig. 6c). Yet, despite normal activation (phosphorylation) and nuclear presence of ATF2 in response to β-adrenergic receptor stimulation, p-ATF2 fails to increase protein levels of UCP1 in the p62 Δ69-251 mice (Fig. 6c, d). Preservation of p-ATF2 nuclear entry with failure to induce UCP1 levels are confirmed also in BAT primary cells of Ucp1-Cre p62 flx/flx mice (Fig. 6e) Fig. 2i and Supplementary Fig. 1g), we see levels of p38 and p-p38 also unchanged in the nuclear/cytosolic fractions of BAT in young non-obese p62 Δ69-251 mice (Fig. 6d) and this is confirmed also in cultured BAT cells of the global p62 −/− mice (Fig. 6f).
Collectively, these data indicate that the observed changes in p-ATF2 are not related to changes in p38 and further suggest that functional p62 is required for ATF2 to bind and activate its nuclear targets. We next performed chromatin immunoprecipitation (ChIP) assays in BAT of the p62 Δ69-251 mice. There was abundant endogenous genomic binding of ATF2 at canonical target sites in BAT from wt mice (Fig. 7a). These ATF2 target sites include promoters of Jun, Fos, and Atf3 19 , as well as previously described BAT-specific ATF2 target sites within the Ucp1 enhancer (CRE2) and promoter (CRE4), and the Pgc-1α promoter 11 . Importantly, ATF2 binding to target loci was severely impaired in BAT from p62 Δ69-251 mice (Fig. 7a) and this was confirmed also in BAT from global p62 −/− mice (Fig. 7b). These data demonstrate that failure of ATF2 to induce Ucp1 and Pgc-1α expression in the p62 mutant mice is associated with impaired genomic binding of ATF2 to the promoter/enhancer regions of target genes, and implies that functional p62 is required for ATF2-mediated transcriptional activation. Consistent with a previously described transcriptional co-regulator role for p62 [20][21][22] , we observed also reduced chromatin binding of p62 to the promoter/enhancer regions of Ucp1 and Pgc-1α in BAT of the p62 −/− mice, and this was also confirmed in the BAT-specific p62 ko mice ( Supplementary Fig. 8a,b). To further validate this finding, we assessed ATF2 target activation in vitro using a luciferase-based CRE reporter assay. We co-transfected HEK293T cells with an ATF2 sensor (pCRE-luc), along with the dominant positive p38/ATF2 activator MKK6(K82A) and the respective wt p62 or mutant p62 Δ69-251 . Consistent with our finding that lack of p62 leads to impaired binding of ATF2 to target genes (Fig. 7a, b), there was reduced transcriptional activation in cells transfected with the p62 Δ69-251 plasmids relative to the p62 wt controls (Fig. 7c). Collectively, these data demonstrate that a motif encoded by the p62 amino acids 69-251 is required for ATF2 to bind and activate its nuclear targets, and suggests that its absence in the p62 Δ69-251 mice leads to impaired energy expenditure and BAT dysfunction due to failure of ATF2 to   activate key thermogenic targets in the nucleus, including the promoter/enhancer regions of Ucp1 and Pgc-1α. Notably, while BAT function is severely impaired in all three analyzed p62 mutant mouse models, no ATF2-mediated transcriptional changes are observed on stress and DNA damage, ER stress, inflammation, cell cycle, cell death, and autophagy in either liver, muscle, kidney, and spleen of p62 Δ69-251 mice ( Supplementary  Figs. 9,10).

Discussion
We here demonstrate that the scaffold protein p62 is a key regulator of the UCP1 pathway. Our data demonstrate that p62 regulates the UCP1 pathway in a cell autonomous manner by directly binding to ATF2 and by regulating its binding to the Ucp1 enhancer and Pgc-1α promoter. As demonstrated in p62 Δ69-251 mice, global p62 −/− mice and Ucp-Cre p62 flx/flx mice, lack of p62 action results in severely impaired BAT function. We attribute this to failure of ATF2 to bind and activate the promoters/enhancers of its downstream thermogenic targets Ucp1 and Pgc-1α. These data are consistent with the observation that lack of p62 in adipocytes, but not in the liver, muscle, myeloid lineage cells or the CNS, results in impaired BAT function and an obese phenotype 16 . A recent report suggested that the obese phenotype of p62 deficiency originates from the lack of p62 action in the CNS, leading to hyperphagia and leptin resistance 23 . However, while changes in energy expenditure are consistently observed in mice lacking p62 either globally 15 or in an adiposespecific manner 16 , no major difference in food intake has been observed in any of these mice. Consistent with this, we also report here that p62 Δ69-251 mice have impaired energy expenditure but no difference in food intake. Moreover, also CNS-specific p62 −/− mice have no obese phenotype and normal food intake 16 . Also, as occurs in the adipocyte-specific p62 −/− mice, the impaired energy expenditure of p62 Δ69-251 mice cannot be explained by changes in locomotor activity. Since BAT thermogenesis is the only observed dysfunctional component of the energy balance equation, we consequently conclude that impaired energy expenditure of BAT is the causal factor leading to increased body weight in the p62deficient mice. We also assessed whether p62 Δ69-251 mice have ATF2 mediated disturbances in other metabolic processes, like stress and DNA damage, ER stress, inflammation, cell cycle, cell death, and autophagy. None of these pathways seem overtly changed in either liver, muscle, kidney, and spleen of p62 Δ69-251 mice. Notably, the impaired energy expenditure of p62 Δ69-251 mice becomes most apparent when the mice are exposed to cold; when the BAT is most active. In light of this notion, it seems possible that other ATF2 signaling events only become apparent in the p62 mutant mice after specific stimuli. In summary, we demonstrate that the scaffold protein p62 is an essential key regulator of adaptive thermogenesis and BAT function in that it regulates Ucp1 and Pgc-1α via modulation of ATF2 genomic binding. On an important note, the frequently observed increase in p-ATF2 in BAT of the p62 mutant mice does not seem to be an integral part of the pathological mechanism leading to the impaired BAT function. Accordingly, while we consistently see across all three analyzed p62 mutant mouse lines that p-ATF2 fails to activate its nuclear targets Ucp1 and Pgc-1α (Fig. 6c, e, f), we see that levels of p-ATF2 are often (Figs. 2i, 5c, and 6e, f) but not always (Fig. 6c) increased in BAT. In summary, our data establish p62 as an important regulator of the UCP1 pathway. Additional studies are warranted to further characterize nuclear/cytosolic localization of p62, and to determine whether p62 regulates BAT function also via UCP1-independent mechanisms.

Methods
Animal studies. Mice were kept at a constant environment with ambient temperature set to 22 ± 2°C with constant humidity (45-65%) and a 12 h/12 h light/ dark cycle. Mice had free access to water and were fed ad libitum with either a standard chow (Type 1314, Altromin GmbH, Lage, Germany) or HFD (58% kcal fat; Research Diets, New Brunswick, NJ, USA; # D12331). Animal experiments were performed in accordance with European guidelines under permission of the local animal ethics committee of the state of Bavaria or the Regional Council Tübingen. Whole-body composition (fat and lean tissue mass) was measured using nuclear magnetic resonance technology (EchoMRI, Houston, TX, USA). For glucose and insulin tolerance test, mice were fasted for 6 h and then challenged with 1.75 g glucose per kg body weight or 0.75 U insulin (Humalog, Eli Lilly, Bad Homburg, Germany) per kg body weight. Generation of full-length p62 knock-out mouse lines p62 −/− and aP2-cre + ;p62 loxp/loxp is indicated elsewhere 15,16 . p62 Δ69-251 mice were generated by Prof. Andre Gessner (University of Regensburg) using a targeting vector that replaces exons two to four of p62 with a neomycin-resistance cassette. Ucp1-Cre p62 flx/flx mice were generated by pairing p62 flx/flx mice with mice expressing Cre recombinase under control of the Ucp1 promoter.
Indirect calorimetry. Energy expenditure and home-cage activity were assessed using a climate-controlled combined indirect calorimetric system (TSE System, Bad Homburg, Germany). After acclimatization for 24 h, levels of O 2 and CO 2 were measured every 10 min for 4-5 days. Data for energy expenditure were analyzed using ANCOVA with body weight as covariate as previously suggested 18 . For ANCOVA analysis, energy expenditure raw data were blotted against the individual body weight. Every individual data point was then slighted up or down the group's own regression line until all data line up at a predefined body weight of 40 g. The result of this ANCOVA correction by body weight was then plotted as a bar graph as previously suggested 18,24 .
Cold challenge in wt and p62 Δ69-251 mice. Body weight and body composition was measured at the beginning and end of the study. Mice were acclimatized to thermoneutrality (30°C) for 1 day, followed by subsequent measurement of energy expenditure starting at 30°C for another day. On Day 3, a non-shivering thermogenesis (NST) test was performed at 30°C. For the NST test, mice were i.p. injected with a single bolus of Norepinephrine (Arterenol, Sanofi Aventis, Frankfurt, Germany; 1 mg kg −1 ) or saline control. Afterward, the temperature of the climate chamber was gradually decreased from 30 to 23°C (2 days), to 10°C (1 day), to 5°C (1 day). Tissues were collected at the end of the study.  In vivo PET and MRI studies. Animals were injected i.p. with 11.7 ± 0.3 MBq 18 F-FDG dissolved in saline. Animals were kept awake and allowed to move freely in their home cages during 18 F-FDG uptake for 30 min. Directly after that, the mice were euthanized with CO 2 . Static PET acquisitions (10 min) were performed using a dedicated micro-PET scanner (DPET, Siemens Preclinical Solutions, Knoxville, USA). The spatial resolution was 1.4 mm (center of field of view), the axial field of view was 12.7 cm. All PET scans were acquired with an energy window of 350-650 keV and a coincidence timing window of 3.432 ns. All major corrections were applied to the PET scans (e.g., normalization, decay & deadtime correction), but no attenuation correction was used as only mice on thin plastic beds were scanned. PET-images were reconstructed using the three-dimensional ordered-subsetexpectation-maximization (OSEM3D) algorithm with a 128 × 128 matrix and a final pixel size of 0.79 × 0.79 ×0.80 mm³.
PET and MRI data analysis. Inveon Research Workplace (IRW v3.0, Siemens Preclinical Solutions, Knoxville, USA) was used for data analysis. PET and MRI data were co-registered for exact, MRI-guided delineation of volumes of interest (VOI), drawn manually for BAT based on the MR image. PET data were expressed as standardized uptake value (SUV), calculated as measured activity in BAT divided by injected activity and body weight.
Infrared thermography. For surface body temperature measurement, 7-d-old pups were temporarily placed in a 6-well dish for temperature recording via a thermal imaging camera Optris, OPTPI400 (Ekomess, Heilbronn, Germany).
Bomb calorimetry. For assessment of assimilated energy, food intake and feces production were assessed over a 5-d period. Feces and food were dessicated in a drying oven (50°C) for bomb calorimetric combustion. The assimilated energy was then calculated by subtracting the energy lost with feces (E out ) from the energy intake (E in ). Chow control diet was combusted to calculate energy content of the absorbed food. Bomb calorimetric analysis was performed using a 6300 Oxygen Bomb Calorimeter (Parr Instrument Technology, IL, USA).   have been analyzed using a two-sided two-tailed t-test, data in (c) have been analyzed using one-way ANOVA was treatment as co-variant and using Bonferroni multiple comparison between the treatment groups. Data represent mean ± SEM. *p < 0.05; **p < 0.01; ***p < 0. proteinase and phosphatase inhibitors (Thermo Fisher). After homogenization at 16 Hz for 2 min, samples were centrifuged at 750 × g for 5 min at 4°C. The supernatant was collected and centrifuged at 10,000 rpm, 10 min (cytoplasmic fraction). This process was repeated four times to remove nuclear contamination. The pellet was washed with fractionation buffer and centrifuged at 750 × g, 5 min to remove cytoplasmic contamination at least three times. The nuclear/membrane fraction was dissolved in RIPA buffer.
Oil Red O staining. Cells were fixed with 4% paraformaldehyde (PFA) during different days of differentiation. Adipocytes were stained with Oil Red O (Sigma-Aldrich, Munich, Germany) for 10 min and immediately washed with H 2 O. For lipid quantification, Oil Red O was retrieved from the cells by 100% isopropanol and absorbance was measured at 500 nm. Dapi signal was measured to correct for cell number.
Preparation of RNA and gene expression analysis. Total RNA was prepared using RNeasy Kit (Qiagen, Hilden, Germany) according to manufacturer's instructions. cDNA synthesis was performed with QuantiTect Reverse Transcription Kit (Qiagen, Hilden, Germany) according to manufacturer's instructions. Gene expression of cell samples (N = 3 per group) was profiled with quantitative PCR-based (qPCR) techniques using SYBR green or TaqMan Single Probes (Thermo Fisher Scientific, Erlangen, Germany). The relative expression of the selected genes was measured using the 7900HT Fast Real-Time PCR System (Thermo Fisher Scientific, Erlangen, Germany). The relative expression levels of each gene were normalized to the housekeeping gene peptidylprolyl isomerase B (Ppib) or Hypoxanthine Phosphoribosyltransferase (Hprt). TaqMan Low Density Array (Thermo Fisher Scientific, Erlangen, Germany) was performed according to instructions. The following primers were used: Ucp1 forward 5ʹ-GGCCTCTACGA CTCAGTCCA-3ʹ, Ucp1 reverse 5ʹ-TAAGCCGGCTGAGATCTTGT-3ʹ, Pgc-1α forward 5ʹ-AGCCGTGACCACTGACAACGAG-3ʹ, Pgc-1α reverse 5ʹ-GCTGC ATGGTTCTGAGTGCTAAG-3ʹ, Prdm16 forward 5ʹ-CCGCTGTGATGAGTG TGATG-3ʹ, Prdm16 reverse 5ʹ-GGACGATCATGTGTTGCTCC-3ʹ, Bmp8b forward 5ʹ-TCCACCAACCACGCCACTAT-3ʹ, Bmp8b reverse 5ʹ-CAGTAGGCAC Statistics. Statistical analyses were performed using statistical tools in GraphPad Prism8 (version 8.3.0). Differences between groups were assessed by Student's twosided two-tailed t-test, one-way ANOVA, or two-way ANOVA with time and genotype as co-variants followed by Bonferroni's post hoc multiple comparison testing for individual time points. All results are given as mean ± SEM. P < 0.05 was considered statistically significant. Differences in energy expenditure were calculated using ANCOVA with body weight as covariate using SPSS (version 24). For animal studies, sample sizes were calculated based on a power analysis assuming that a ≥5 g difference in body weight between genotypes can be assessed with a power of ≥75% when using a two-sided statistical test under the assumption of a standard deviation of 3.5 and an alpha level of 0.05.
Reporting summary. Further information on research design is available in the Nature Research Reporting Summary linked to this article.