Peroxisomal metabolism of branched fatty acids regulates energy homeostasis

Abstract

Brown and beige adipocytes express uncoupling protein 1 (UCP1), a mitochondrial protein that dissociates respiration from ATP synthesis and promotes heat production and energy expenditure. However, UCP1^-/- mice are not obese^1-5, consistent with the existence of alternative mechanisms of thermogenesis^6-8. Here we describe a UCP1-independent mechanism of thermogenesis involving ATP-consuming metabolism of monomethyl branched-chain fatty acids (mmBCFA) in peroxisomes. These fatty acids are synthesized by fatty acid synthase using precursors derived from catabolism of branched-chain amino acids⁹ and our results indicate that β-oxidation of mmBCFAs is mediated by the peroxisomal protein acyl-CoA oxidase 2 (ACOX2). Notably, cold exposure upregulated proteins involved in both biosynthesis and β-oxidation of mmBCFA in thermogenic fat. Acute thermogenic stimuli promoted translocation of fatty acid synthase to peroxisomes. Brown-adipose-tissue-specific fatty acid synthase knockout decreased cold tolerance. Adipose-specific ACOX2 knockout also impaired cold tolerance and promoted diet-induced obesity and insulin resistance. Conversely, ACOX2 overexpression in adipose tissue enhanced thermogenesis independently of UCP1 and improved metabolic homeostasis. Using a peroxisome-localized temperature sensor named Pexo-TEMP, we found that ACOX2-mediated fatty acid β-oxidation raised intracellular temperature in brown adipocytes. These results identify a previously unrecognized role for peroxisomes in adipose tissue thermogenesis characterized by an mmBCFA synthesis and catabolism cycle.

Fig. 1. The peroxisomal β-oxidation enzyme ACOX2 increases with thermogenic stimuli and promotes catabolism of mmBCFA in brown adipocytes.

a, Heatmap of genes involved in various peroxisomal pathways in BAT of mice maintained at thermoneutrality (30 °C) or gradually cold adapted. n = 4. b, Western blot analysis in BAT of warm or cold-treated mice. n = 2. c, ACOX2 gene expression after 2 h of treatment with norepinephrine or vehicle in differentiated mouse UCP1^−/− brown adipocytes (norepinephrine 1 μM) or mature adipocytes isolated from fresh pig iWAT (norepinephrine 2 μM). n = 3 biological replicates per group. d, ACOX2 gene expression in cultured human brown-like adipocytes treated with or without forskolin (10 μM, 3 days). n = 3 biological replicates per group. e, Western blot analysis in BAT SVF cells collected on various days during adipogenesis. f, Immunofluorescence of ACOX2 and PMP70 in differentiated WT BAT SVF cells. g, Western blot analysis of ACOX2 knockdown using CRISPR–Cas9 in brown adipocytes. h, Volcano plot depicting log₂[fold change] (FC) values of fatty acids. n = 4. i, Quantification of various mmBCFA species. n = 4. j, Atom-transition map illustrating flow of carbons from [U¹³C₆]glucose into newly synthesized mmBCFA. ¹³C carbons are indicated by closed circles. Created in Adobe Illustrator. k, Measurement of ¹³C-labelled iso-C15:0 in control and sgACOX2 brown adipocytes after CL316,243 treatment. n = 3. l, OCR in sgACOX2 and control brown adipocytes treated with oligomycin and iso-C17:0, followed by FCCP. n = 7. Data with error bars are reported as mean ± s.e.m. Data in a–d,h,i,k,l are from biologically independent samples. Images in e,f,g are representative of two separate experiments. Two-sided unpaired Student’s t-test in c, d and i. The data in h were adjusted for multiple comparisons using the Benjamini–Hochberg method. Two-way ANOVA with Fisher’s least-significant difference test in k and two-way ANOVA with Sidak’s multiple comparisons test in l. Scale bars, 10 μm. Source Data

Fig. 2. FASN translocates to peroxisomes, drives mmBCFA synthesis and regulates thermogenesis in BAT.

a, mRNA levels of mmBCFA β-oxidation and synthesis genes in BAT of WT mice maintained at thermoneutrality (30 °C) or 4 °C for 7 days. n = 4. b, Immunofluorescence analysis and quantitative analysis of FASN peroxisomal localization in BAT of WT mice maintained at room temperature (RT) or after cold treatment for 2 h. n = 3. c, Western blot analysis following subcellular fractionation in differentiated BAT SVF cells after norepinephrine (0.2 μM) treatment for 2 h. d, Western blot analysis of FASN in BAT and iWAT of FASN-BKO mice. n = 3. e, Mass spectrometry analysis of a StCFA C16:0 and an mmBCFA iso-C17:0 in BAT of FASN-BKO and control mice housed at 22 °C or 4 °C for 6 h. n = 4. f, Core body temperature of control (n = 13) and FASN-BKO (n = 12) mice after cold exposure (4 °C). g, Energy expenditure of control (n = 4) and FASN-BKO (n = 6) mice after CL316,243 treatment. Data in a,b (right),c–g are from biologically independent samples. Images in b (left) and c are representative of two separate experiments. Data are presented as mean ± s.e.m. Statistical analyses were performed using two-sided unpaired Student’s t-test (a,b right), one-way ANOVA (e) or two-way ANOVA (f–g), followed by Fisher’s least-significant difference post hoc test (e–g). Scale bars, 10 μm. Source Data

Fig. 3. Adipose-specific ACOX2 knockout impairs thermogenesis and promotes diet-induced obesity and insulin resistance.

a, Gene targeting strategy using CRISPR–Cas9 to insert LoxP sites into the ACOX2 locus. The floxed mice were crossed with an adiponectin-Cre mouse to generate ACOX2-AKO mice. b, ACOX2 gene expression analysis in adipose tissue depots of control and ACOX2-AKO mice. n = 4. c, Infrared thermal imaging of mice after a 6 h of cold exposure. Quantification shows average surface temperature of two mice per genotype. d, Core body temperature of control and ACOX2-AKO male mice after cold exposure (4 °C). n = 13. e, VO₂ of control (n = 9) and ACOX2-AKO (n = 8) male mice after CL316,243 treatment. f, Western blot analysis in BAT of control (n = 4) and ACOX2-AKO (n = 3) mice. g, Body weight of control and ACOX2-AKO male mice fed HFD. n = 8. h, Body composition analysis of HFD-fed male control and ACOX2-AKO mice. n = 8. i, Regression analysis of energy expenditure with body weight as a covariate in HFD-fed ACOX2-AKO and control male mice. n = 7. j,k, Glucose tolerance testing (GTT) (j) and insulin tolerance testing (ITT) (k) in control and ACOX2-AKO mice after HFD feeding. The mice were dosed with glucose (j) or insulin (k) based on lean body mass; n = 8 for GTT; n = 5 for ITT. Data in b,c (right), d–k are from biologically independent samples. Images in c are representative of two mice per group. Data with error bars are reported as the mean ± s.e.m. Two-sided unpaired Student’s t-test in b,h or two-way ANOVA followed by Fisher’s least-significant difference test in d,e,g,j,k or two-way analysis of covariance (ANCOVA) with Tukey’s test in i. Panel a was created using BioRender (https://biorender.com). Source Data

Fig. 4. Adipose-specific ACOX2 overexpression promotes thermogenesis and counteracts diet-induced obesity and insulin resistance.

a, Schematic diagram of a transgene construct used to generate adipose-specific ACOX2 overexpression (ACOX2^Adipo-OE) mice. b, ACOX2 gene expression in adipose tissue depots of ACOX2^Adipo-OE and WT mice. n = 4. c, Core body temperature of ACOX2^Adipo-OE and WT mice after cold exposure. n = 10. d, VO₂ of ACOX2^Adipo-OE and WT mice after CL316,243 treatment. n = 5. e, Western blot analysis of ACOX2 in BAT of ACOX2^Adipo-OE and WT mice. n = 2. f, Body weight of HFD-fed ACOX2^Adipo-OE and WT female mice. n = 7. g, Body composition of HFD-fed ACOX2^Adipo-OE and WT female mice. n = 7. h, Haemotoxylin and eosin staining in adipose depots of HFD-fed ACOX2^Adipo-OE and WT female mice. i, GTT analysis in HFD-fed ACOX2^Adipo-OE and WT female mice. n = 9. j, ITT analysis in HFD-fed ACOX2^Adipo-OE and WT female mice. n = 8. Data in b–g,i,j are from biologically independent samples. Images in h are representative of three mice per group. Data are reported as the mean ± s.e.m. Two-sided unpaired Student’s t-test in b,g or two-way ANOVA followed by Fisher’s least-significant difference test in c,d,f,i,j. Scale bars, 75 μm. Source Data

Fig. 5. ACOX2 promotes UCP1-independent thermogenesis.

a, Western blot analysis of ACOX2 and UCP1 in BAT of WT, ACOX2^Adipo-OE, UCP1^−/− and UCP1^−/−/ACOX2^Adipo-OE mice. n = 2. b, Body temperature of male mice after cold exposure (4 °C). n = 10. c–g, Body weight (c), body composition (d), adipose tissue weights (e), gross BAT morphology (f) and haemotoxylin and eosin staining of adipose tissue (g) from HFD-fed mice housed at thermoneutrality (30 °C). n = 5 for c–e. Images in f and g are representative of three mice per genotype. h, Western blot analysis of UCP1^−/− brown adipocytes transfected with lentivirus overexpressing ACOX2 or LacZ. n = 2. i, OCR measurement using Seahorse in ACOX2 or LacZ-expressing UCP1^−/− brown adipocytes. Oligo, oligomycin; AA + R, antimycin A and rotenone. n = 6. j, Measurement of ATP in WT brown adipocytes transfected with LacZ or ACOX2. n = 12. k, Western blot analysis for pAMPK, AMPK, pACC, ACC, ACOX2 and FASN in BAT of ACOX2^Adipo-OE and WT mice cold-treated for 6 h. ACC antibodies recognize ACC1 and ACC2. n = 2. l, Schematic of Pexo-TEMP, a peroxisome-localized temperature sensor. m, Fluorescence microscopy analysis of Pexo-TEMP and a peroxisome marker (mCherry-PTS1) in brown adipocytes. Images representative of two separate experiments. n, Fluorescence intensity ratio in response to temperature changes in WT brown adipocytes expressing Pexo-TEMP. n = 16 separate wells. o, Fluorescence intensity ratio in WT brown adipocytes expressing Pexo-TEMP together with LacZ or ACOX2, followed by the indicated treatment. n = 16 separate wells. Data in a–e,h–k,n–o are from biologically independent samples. Data are reported as the mean ± s.e.m. Two-sided unpaired Student’s t-test in d,e,j; two-way ANOVA followed by Fisher’s least-significant difference test in b,c,i; or one-way ANOVA followed by Fisher’s least-significant difference test in o. Scale bars, 75 μm (g), 10 μm (m). Source Data

Extended Data Fig. 1. Characterization of ACOX2 function and regulation in adipocytes.

a), Classical peroxisomal BCFA β-oxidation pathway. For 2-methyl BCFAs, propionyl-CoA is released after the first round of β-oxidation. Created in PowerPoint. b, ACOX2 gene expression in different tissues of wild type mice housed at 4 °C or 22 °C for 7 days. N = 4. c, Gene expression analysis in sgACOX2 and control BAT SVF cells. N = 3. d, Oil Red O staining in ACOX2 KO and control brown adipocytes. Scale bar, 300 mm. e, Structures of C16:0 (palmitate), iso-C17:0, and anteiso-C17:0. f, Mass spectrometry analysis of various straight chain fatty acids in Acox2 KO and control brown adipocytes. n = 4. g, Volcano plot depicting Log₂FC values of fatty acids in brown adipocytes overexpressing ACOX2 or GFP. N = 3. h, OCR measurement using Seahorse assay in ACOX2 or GFP-expressing WT brown adipocytes treated with the indicated fatty acid or BSA. N = 7. i, Gene expression analysis in mouse brown adipocytes overexpressing ACOX2 or GFP. N = 3. j, mtDNA copy number normalized to nuclear DNA. N = 6. k, ACOX2 gene expression following treatment with BSA, C16:0 or iso-C17:0 in differentiated mouse UCP1^−/− BAT SVF cells (1 mM) or mature adipocytes isolated from fresh pig iWAT (2 mM) for 2 h. N = 3. Data are reported as the mean ± SEM. Data in b, c and f-k are from biologically independent samples. Two-sided unpaired Student’s t test in b, c, f, and i. Comparisons between groups in g were made with a two-tailed unpaired Student’s t-test adjusted for multiple comparisons using the Benjamini-Hochberg method. Two-way ANOVA with Tukey’s multiple comparison’s test in h. One-way ANOVA with Fisher’s LSD in k. Source Data

Extended Data Fig. 2. Localization of the mmBCFA synthetic proteins FASN and CRAT to peroxisomes.

a, Schematic of mmBCFA biosynthesis using short branched acylCoA (BrCoA) derived from mitochondrial catabolism of BCAA. b, Western blot analysis of CRISPR/Cas9-mediated KO of BCKDHα in brown adipocytes. N = 2. c-d, Mass spectrometry analysis of mmBCFA (c) and conventional fatty acids (d) in control and BCKDHA KO brown adipocytes. N = 4. e, mRNA levels of mmBCFA β-oxidation and synthesis genes in iWAT of WT mice maintained at thermoneutrality or 4 °C for 7 days. N = 4. f, FASN gene expression in control or norepinephrine (NE)-treated differentiated mouse UCP1^−/− brown adipocytes. N = 3. g, FASN gene expression in cultured human brown-like adipocytes treated with or without forskolin. N = 3. h, Immunofluorescence analysis of peroxisomal localization of FASN in wild-type brown adipocytes after NE treatment for 0 or 2 h. N = 3. Scale bar, 10 mm. i, Immunofluorescence analysis of peroxisomal localization of FASN in BAT of WT mice treated with or without CL316,243. N = 3. Scale bar, 10 mm. j, Co-immunoprecipitation of HA-PEX7 with FLAG-FASN in HEK293T cells. k, Immunofluorescence analysis of peroxisomal localization of CRAT in wild-type brown adipocytes after NE treatment for 2 h. Scale bar, 10 mm. l, Western blot analysis of CRISPR/Cas9-mediated KO of CRAT in brown adipocytes. N = 3. m, Oil Red O staining in differentiated CRAT KO and control adipocytes. Scale bar, 300 mm. n,o, Mass spectrometry analysis of mmBCFA (n) and stCFA (o) in control and CRAT KO brown adipocytes. N = 3. p, OCR measurement in sgCRAT and control brown adipocytes at baseline and after sequential treatment with leucine, NE, and iso-C17:0. N = 8 (sgControl); N = 7 (sgCRAT). q, Gene expression analysis in sgCRAT and control brown adipocytes overexpressing ACOX2 or GFP. N = 3. r, OCR measurement in sgCRAT and control brown adipocytes overexpressing ACOX2 or GFP. Oligo, oligomycin; FCCP, carbonyl cyanide-p-trifluoromethoxyphenylhydrazone; AA + R, antimycin A and rotenone. N = 9. Data with error bars are mean ± SEM. Data in b–i, k, l, and n–r are from biologically independent samples. Statistical significance was determined by two-sided unpaired Student’s t test (c–i, k, n–o), two-way ANOVA with Sidak’s (p) or Tukey’s test (r), and one-way ANOVA with Fisher’s LSD (q). Representative images in h–k, and m are from two independent experiments. Panel a was created using BioRender (https://biorender.com). Source Data

Extended Data Fig. 3. Brown adipocyte-specific FASN knockout reduces mmBCFA levels and increases diet-induced obesity.

a, Western blot analysis of FASN in 4-hydroxytamoxifen (4-OHT)-treated FASN^Lox/Lox brown adipocytes transduced with retrovirus expressing GFP or Cre-ERT2. N = 3. b, Microscopy images of FASN-iKO and control BAT adipocytes prior to and after 4-OHT treatment. In the bottom panel, the cells were stained with Oil Red O. Scale bar, 75 mm (upper panel), 300 mm (lower panel). c, Mass spectrometry analysis of mmBCFA in control and FASN-iKO brown adipocytes. N = 3. d, Mass spectrometry analysis of various straight chain fatty acids in control and FASN-iKO brown adipocytes. N = 3. e, Body weight of HFD-fed FASN-BKO and control female mice. N = 7. f, Regression analysis of energy expenditure in HFD-fed mice. N = 7. g, Cumulative energy intake in HFD-fed FASN-BKO and control female mice. N = 4. h, Locomotor activity in FASN-BKO and control female mice. N = 4. Data are reported as mean ± SEM. Data in a and c-h are from biologically-independent samples. Images in b are representative of two separate experiments. Two-sided unpaired Student’s t test in c and d; two-way ANOVA with Fisher’s LSD in e; and two-way ANCOVA with Tukey’s test in f. Source Data

Extended Data Fig. 4. Global knockout of ACOX2 impairs thermogenesis and exacerbates diet-induced obesity and metabolic dysfunction.

a, ACOX2 gene expression in adipose tissue depots of ACOX2^−/− and WT mice. N = 3. b, Western blot analysis of ACOX2 in BAT of ACOX2^−/− and WT mice. N = 2. c, Core body temperature of ACOX2^−/− and WT male mice after cold exposure (4 °C). N = 10. d, Gene expression analysis of proteins involved in UCP1-dependent and -independent pathways of thermogenesis in BAT of ACOX2^−/− and WT mice. N = 5. e, Gene expression analysis of WT mice fed normal chow diet or HFD. N = 3. f, Body weight of HFD-fed ACOX2^−/− and WT female mice. N = 7. g, Body weight of ACOX2^−/− and WT male mice fed with HFD. N = 9. h, Body composition analysis of ACOX2^−/− and WT male mice fed HFD. N = 9. i, Food intake of ACOX2^−/− and WT mice. N = 5. j, Locomotor activity of ACOX2^−/− and WT mice. N = 5. k, Tissue weights of adipose depots of HFD-fed ACOX2^−/− and WT male mice. N = 9. l, Gross images of BAT and iWAT from HFD-fed ACOX2^−/− and WT male mice. m, Histologic analysis of H&E-stained adipose tissue sections from HFD-fed ACOX2^−/− and WT male mice. Scale bar, 75 mm. n, GTT analysis in HFD-fed ACOX2^−/− and WT female mice. N = 7. o, ITT analysis in HFD-fed WT and ACOX2^−/− female mice. N = 7. Data are reported as the mean ± SEM. Data in a-k and n-o are from biologically independent samples. Images in l and m are representative of 3 mice/group. Two-sided unpaired Student’s t test in a, d, e, and h-k; two-way ANOVA followed by Fisher’s LSD test in c, f, g and n-o. Source Data

Extended Data Fig. 5. Adipose-specific ACOX2 deletion impairs thermogenesis and promotes diet-induced obesity.

a, Gene expression analysis of ACOX1, ACOX3 and FASN in BAT of ACOX2-AKO and control mice. N = 4. b, Volcano plot depicting Log₂FC values of fatty acids in BAT of ACOX2-AKO and control mice. N = 5. c, Total serum bile acids in ACOX2-AKO and control mice. N = 9. d, Core body temperature of ACOX2-AKO and control female mice after cold exposure (4 °C). N = 11. e, VO₂ of female ACOX2-AKO (N = 6) and control mice (N = 7) after CL316,243 treatment. f, Body weight of HFD-fed ACOX2-AKO and control female mice. N = 8. g, Histologic analysis of H&E-stained adipose tissue sections from HFD-fed ACOX2^−/− and WT male mice. Scale bar, 250 mm. h, Cumulative energy intake of ACOX2-AKO (N = 5) and control mice (N = 4). i, Locomotor activity of ACOX2-AKO (N = 5) and control mice (N = 4). j, Western blot analysis of insulin-stimulated AKT phosphorylation in BAT of liver of HFD-fed ACOX2-AKO and control mice. The bar graph shows quantification of AKT phosphorylation (N = 2). Data with error bars are reported as the mean ± SEM. Data in a-f and h-j are from biologically independent samples. Images in g are representative of 3 mice/genotype. Two-sided unpaired Student’s t test in a and c; two-sided unpaired Student’s t-test adjusted for multiple comparisons using the Benjamini-Hochberg method in b; two-way ANOVA with Fisher’s LSD in d-f. Source Data

Extended Data Fig. 6. Adipose-specific ACOX2 overexpression enhances thermogenesis and metabolic health.

a, Media glycerol levels of fresh BAT explants from ACOX2^Adipo-OE (N = 5) or WT (N = 4) mice incubated in media containing vehicle or isoproterenol for 2 h. b, Core body temperature prior to and after cold exposure in ACOX2^Adipo-OE and WT mice treated with 25 mg/kg BCAT-IN-2 or vehicle for 1 week. N = 7. c, Food intake of ACOX2^Adipo-OE and WT mice. N = 7. d, Fecal FFA content of ACOX2^Adipo-OE and WT mice. N = 6. e, Locomotor activity of ACOX2^Adipo-OE and WT mice. N = 6. f, Body weight of HFD-fed ACOX2^Adipo-OE and WT male mice. N = 8. g, Body composition analysis in HFD-fed ACOX2^Adipo-OE and WT male mice. N = 8. h, GTT analysis in HFD-fed ACOX2^Adipo-OE and WT male mice. N = 8. i, ITT analysis in HFD-fed ACOX2^Adipo-OE and WT male mice. N = 8. Data are reported as mean ± SEM and are from biologically-independent samples. One-way ANOVA with Fisher’s LSD in a and b and two-way ANOVA with Fisher’s LSD in f, h, and i; Two-sided unpaired Student’s t test in g. Source Data

Extended Data Fig. 7. ACOX2 overexpression increases energy expenditure and improves metabolic health in UCP1^−/− mice.

a, Relative FASN expression in BAT of UCP1^−/− (N = 8) and UCP1^−/−/ACOX2^Adipo-OE (N = 7) mice. b, Immunofluorescence analysis and quantitative analysis of FASN peroxisomal localization in BAT of UCP1^−/− and UCP1^−/−/ACOX2^Adipo-OE mice housed at thermoneutrality. N = 3. Scale bar, 10 mm. c-e, Indirect calorimetry analysis of (c) energy expenditure, (d) cumulative food intake, and (e) locomotor activity in ACOX2^Adipo-OE (N = 8) and WT (N = 7) mice housed at thermoneutrality. f, Body weight of HFD-fed male UCP1^−/− and UCP1^−/−/ACOX2^Adipo-OE mice fed mice housed at normal room temperature (22 °C). N = 7. g, Body composition analysis of HFD-fed UCP1^−/− and UCP1^−/−/ACOX2Adipo-OE mice housed at 22 °C. N = 7. h, Adipose tissue weights of HFD-fed UCP1^−/− and UCP1^−/−/ACOX2^Adipo-OE mice housed at 22 °C. N = 7. i, Gross images of BAT from HFD-fed UCP1^−/− and UCP1^−/−/ACOX2Adipo-OE mice fed housed at 22 °C. j, H&E staining in adipose depots of HFD-fed UCP1^−/− and UCP1^−/−/ACOX2^Adipo-OE mice housed at 22 °C. Scale bar, 75 mm. k, GTT analysis in HFD-fed UCP1^−/− and UCP1^−/−/ACOX2^Adipo-OE female mice housed at 22 °C. N = 8. l, ITT analysis in HFD-fed UCP1^−/− (n = 7) and UCP1^−/−/ACOX2^Adipo-OE (N = 8) female mice fed housed at 22 °C. Data are presented as mean ± SEM. Data in b-h, k, and l are from biologically independent samples. Images in i and j are representative of three mice per genotype. Statistical analysis was performed using two-sided unpaired Student’s t-test (a, b, g, h), two-way ANOVA with Tukey’s test (c), or Fisher’s LSD test (f, k, l). Source Data

Extended Data Fig. 8. Creatine cycling is dispensable for ACOX2-induced enhancement of oxygen consumption.

a, Activities of complex II and complex IV in isolated mitochondria from BAT of ACOX2^Adipo-OE and WT mice measured using Seahorse (N = 9). b,c, Gene expression (b) and Western blot (c) analysis in sgCKB and control brown adipocytes overexpressing ACOX2 or GFP. N = 3 for panel b and N = 2 for panel c. d, OCR measurement in sgCKB and control brown adipocytes overexpressing ACOX2 or GFP and treated with iso-C17:0 or BSA (N = 12). e, OCR measurement using Seahorse in ACOX2 or GFP-expressing WT brown adipocytes treated with AA + R to block mitochondrial respiration, followed by treatment with C17:0-iso or BSA. N = 4. Data are from biologically-independent samples and are reported as mean ± SEM. One-way ANOVA with Fisher’s LSD in b and two-way ANOVA with Tukey’s test in d and e. Source Data

Extended Data Fig. 9. ACOX2 overexpression induces catalase to attenuate oxidative stress.

a, Temperature-dependent fluorescence spectra of Sirius fluorescent protein. b, Temperature-dependent fluorescence spectra of mTSapphire fluorescent protein. c, Fluorescence microscopy analysis and quantification of peroxisome-localized H₂O₂ sensor HyPer7-SKL in brown adipocytes transfected with lentivirus expressing ACOX2 or LacZ and treated with BSA, C17:0-iso, C16:0 or H₂O₂. N-3. Scale bar, 20 mm. d, Catalase gene expression in GFP or ACOX2-overexpressing brown adipocytes treated with BSA, C16:0 or iso-C17:0. N = 3. e, Western blot analysis of catalase knockout using CRISPR/Cas9 in brown adipocytes. N = 2. f, MTT assay in control and CAT KO brown adipocytes after treatment with BSA, iso-C17:0, C16:0 or H₂O₂. N = 12. g, 4-HNE levels in BAT of ACOX2^Adipo-OE and WT mice. N = 7. Data are reported as the mean ± SEM. Data in c-g are from biologically independent samples. Images in c are representative of three separate experiments. Two-sided unpaired Student’s t test in c (quantification) and f. One-way ANOVA followed by Fisher’s LSD test in d. Panels a,b adapted from ref. , PLOS, under a Creative Commons licence CC BY 4.0. Source Data

Extended Data Fig. 10. UCP1-independent thermogenesis through an ATP-consuming cycle of monomethyl branched chain fatty acid (mmBCFA) synthesis and β-oxidation.

Mitochondrial catabolism of branched chain amino acids (BCAA) generates a short-branched acylCoA (BrCoA), such as isovaleryl-CoA. Carnitine acetyltransferase (CRAT), which is localized in mitochondria and peroxisomes, catalyses interconversion of branched-chain acylcarnitine and BrCoA, which is exported to the cytosol. The de novo lipogenesis enzyme FASN translocates to peroxisomes in response to acute cold exposure or β-adrenergic receptor stimulation and mediates synthesis of mmBCFA using BrCoA as a precursor. ACOX2 promotes β-oxidation of mmBCFA, generating heat in the process. Activation of the FASN–ACOX2 thermogenic pathway depletes ATP, triggering AMPK activation to promote compensatory ATP synthesis and sustain thermogenesis. Graphic created using BioRender (https://biorender.com).