IntroductionBrown adipose tissue (BAT) is thermogenic fat that dissipates energy as heat, whereas white adipose tissue (WAT) stores energy. Activated BAT contributes to non-shivering thermogenesis and increases energy expenditure by utilizing free fatty acids and glucose1,2. As substantial evidence indicates that metabolically active BAT is present in adult humans, BAT may be a therapeutic target for metabolic disease3,4,5,6. Cold-induced thermogenesis in BAT is mediated through the β3-adrenergic signaling pathway, in which mitochondrial uncoupling protein 1 (UCP1) plays a central role. Upon cold exposure, norepinephrine binds to the β3-adrenergic receptor to increase the level of cAMP, which activates protein kinase A (PKA). PKA activates cAMP-response element-binding protein (CREB), which translocates into the nucleus to enhance Ucp1 transcription. PKA also activates p38 and hormone-sensitive lipase, which further enhances thermogenesis by increasing thermogenic gene expression and providing energy substrates7.Post-translational modification of proteins by the small ubiquitin-related modifier (SUMO) modulates the function, localization and stability of target proteins8, thereby regulating various cellular processes9,10,11. SUMO modification can be reversed by SUMO-specific proteases (SENPs) that remove SUMO from target proteins. Previously, we showed that SUMO-specific protease 2 (SENP2) plays important roles in various metabolic contexts. In skeletal muscle, SENP2 increases expression of fatty acid oxidation-associated enzymes, such as carnitine palmitoyl transferase-1 and long-chain acyl-CoA synthetase 1, by deSUMOylating peroxisome proliferator-activated receptor (PPAR)-δ and PPAR-γ. In transgenic mice overexpressing muscle-specific SENP2, high-fat diet (HFD)-induced obesity was ameliorated12. In pancreatic β cells, SENP2 improves mitochondrial function and insulin secretion upon metabolic stress by deSUMOylating dynamin-related protein 113.Although SENP2 is highly expressed in adipose tissue14, our current understanding of the role of SENP2 in adipose tissue metabolism, particularly in BAT, remains incomplete. SENP2 expression and its role in early adipogenesis have been documented15,16; however, the role of SENP2 in mature brown adipocytes under various metabolic stressors has not yet been fully explored. In WAT, SENP2 is intimately involved in adipogenesis and maintaining white adipocyte identity by stabilizing CCAAT/enhancer binding protein β14,15. Specifically, Senp2 knockout in both BAT and WAT using Adipoq-Cre (Senp2-aKO) induces browning in WAT, which exerts a beneficial effect on whole-body metabolism. Intriguingly, while the metabolic phenotype improves in the Senp2-aKO due to the browning of WAT, whitening of BAT was observed upon HFD feeding17. These data suggest that SENP2 plays an essential role in the metabolic adaptability of BAT, particularly in lipid utilization during energy overload.In this study, using brown adipocyte-specific SENP2 knockout (Senp2-BKO) mice, we demonstrated that SENP2 controlled BAT thermogenesis during acute cold stimulation and positively regulated energy balance during energy overload. We also showed that SENP2-mediated deSUMOylation of estrogen-related receptor alpha (ERRα) enhanced Ucp1 expression upon exposure to cold stimuli. Collectively, our data suggest that SENP2 plays an important role in maintaining the healthy metabolic phenotype of BAT.Materials and methodsGeneration of brown adipocyte-specific Senp2 knockout miceSenp2flox/+ mice were generated in the inGenious Targeting Laboratory, as described previously18. Senp2 brown adipocyte-specific knockout (Senp2-BKO) mice were generated by sequentially mating Senp2flox/flox with Ucp1-Cre transgenic mice (Jackson lab).Mice and metabolic analysisAll aspects of animal care and experiments were conducted in accordance with the Guide for the Care and Use of Laboratory Animals of the National Institutes of Health and approved by the Institutional Animal Care and Use Committee of Seoul National University Bundang Hospital, Korea (permit no. BA-2107-324-070). All animals were housed at 22–24 °C, with a 12:12 h light–dark cycle and ad libitum access to standard pelleted chow or HFD (58 kcal% fat with sucrose, D12331; Research Diets) and water. Eight-week-old mice were fed a HFD for 12 weeks. Body weight was measured weekly, and body composition was measured by body composition analyzer (Minispec LF50, Bruker). For the glucose tolerance test (GTT), glucose (1 g/kg body weight) was administered via intraperitoneal injection (i.p.) after a 16 h fasting period. For the insulin tolerance test (ITT), mice that had fasted for 6 h received an injection of human insulin (1 U/kg body weight). Serum glucose levels were measured using an OneTouch Ultra glucometer (LifeScan). For the acute cold exposure experiment, mice were kept at room temperature or at 4 °C, as specified, and body temperatures were recorded rectally every hour for 5 h using a digital thermometer. To simulate cold exposure, 10 mg/kg of CL316,243 (Sigma-Aldrich) was administered, and the mice were euthanized for BAT dissection and RNA extraction 6 h post injection. Metabolic cage experiments were conducted using a Comprehensive Lab Animal Monitoring System (Columbus Instruments, Columbus).Histological analysisTissues were fixed in 4% formaldehyde, embedded in paraffin and sectioned. Sections were subjected to hematoxylin and eosin (H&E) staining. Immunohistochemistry was performed with antibodies against UCP1 and Perilipin 1 (Abcam).mRNA sequencing and data analysisTotal RNA was obtained from BAT of control or Senp2-BKO mice. RNA-sequencing libraries were prepared by TruSeq Stranded mRNA LT Sample Prep kit (Illumina Inc.). The libraries were sequenced, and the reads were aligned to mouse transcriptome (UCSC gene) and genome (mm10) references, respectively, using HISAT2 version 2.1.0 and Bowtie2 2.3.4.1. Trimming tasks for Illumina paired-end and single-ended data for each sample’s FASTQ files were performed using Trimmomatic 0.38. Transcript assembly was performed using the StringTie program and relative log expression normalization was processed after filtering the genes with low quality. Specifically, genes with more than 50% of 0 read counts were excluded from the analysis. From the read counts and transcripts per kilobase million, differential gene expression analysis was performed with DESeq2 R statistical package using the criteria of |log2 fold change | ≥2 and nbinomWaldTest raw P