2-hydroxyglutarate mediates whitening of brown adipocytes coupled to nuclear softening upon mitochondrial dysfunction

Abstract

Mitochondria have a crucial role in regulating cellular homeostasis in response to intrinsic and extrinsic cues by changing cellular metabolism to meet these challenges. However, the molecular underpinnings of this regulation and the complete spectrum of these physiological outcomes remain largely unexplored. In this study, we elucidate the mechanisms driving the whitening phenotype in brown adipose tissue (BAT) deficient in the mitochondrial matrix protease CLPP. Here we show that CLPP-deficient BAT shows aberrant accumulation of lipid droplets, which occurs independently of defects in oxygen consumption and fatty acid oxidation. Our results indicate that mitochondrial dysfunction due to CLPP deficiency leads to the build-up of the oncometabolite D-2-hydroxyglutarate (D-2HG), which in turn promotes lipid droplet enlargement. We further demonstrate that D-2HG influences gene expression and decreases nuclear stiffness by modifying epigenetic signatures. We propose that lipid accumulation and altered nuclear stiffness regulated through 2HG are stress responses to mitochondrial dysfunction in BAT.

Fig. 1. Loss of CLPP leads to cell-autonomous BAT whitening.

a, Representative whole-BAT images from wild-type (WT), ubiquitously-CLPP-deficient mice (KO), adipose-specific CLPP-KO (AKO) mice or BAT-specific CLPP-KO (BKO) mice. b, LD size measurement from WT and CLPP-KO BAT. n = 2,289 LDs for WT and n = 724 LDs for KO. c, Representative transmission electron microscopy images of BAT tissue from WT and CLPP-KO mice. Scale bars, 50 μm, 5 μm and 0.5 μm (from top to bottom). d,e, Steady-state levels of OXPHOS supercomplexes (SC) in the BAT mitochondria of WT and CLPP-KO, CLPP-AKO and CLPP-BKO mice, analysed by blue native–PAGE followed by western blot for CI (NDUFA9) (d) and CV (ATP5A) (e) (n = 3). f, OCRs of whole BAT lysates from WT and CLPP-KO mice (n = 3). g, Numerical density of mitochondria in WT and CLPP-KO BAT (n = 48 fields per genotype obtained from three different mice from each genotype). h, Representative images of cultured and in-vitro-differentiated WT and CLPP-KO mBA cells, stained with DAPI (nuclei) and Bodipy (LDs). Scale bars, 100 μm. Images represent results obtained from four independent experiments. i, Flow cytometry analysis of lipid accumulation (Bodipy) in WT and CLPP-KO mBA cells. Data represent one out of four performed experiments. j,k, OCRs (j) and ECARs (k) of WT and CLPP-KO preADs and mBAs (n = 8). b, Data are presented using Tukey’s box plot with the middle line marking the median, and whiskers show variability within 1.5 × IQR. Anything beyond is an outlier presented as the individual value. f,g,j,k, Data are presented as mean ± s.d. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001, as determined by unpaired two-tailed Student’s t-test in b, g, j and k, and one-way analysis of variance (ANOVA) with multiple comparisons in f. In j and k, the statistical significance between WT and knockout cells of the same cell types were plotted either above (mBA cells) or below the graph lines (preAD). a,h, Schematics were created using Biorender.com. Source data

Fig. 2. BAT is remodelled at multiple levels following CLPP loss.

a, Left, volcano plots of differentially expressed genes in BAT isolated from CLPP-KO, CLPP-AKO or CLPP-BKO mice, compared with the WT (n = 4). Coloured dots show significantly changed transcripts (P ≤ 0,01, two-fold-change). Right, Gene Ontology: Biological Process (GO:BP) analysis of shared significantly changed transcripts (in RNA-seq analysis) in BAT from CLPP-KO, CLPP-AKO and CLPP-BKO mice (Supplementary Table 3). b, Representative H&E staining of BAT from STING-deficient (STING-KO) or STING- and CLPP-deficient (STING/CLPP DKO) mice. Scale bars, 200 μm (results obtained from three mice per genotype). c,d, GSEA enrichment plots for gene signatures related to angiogenesis and actin cytoskeleton (c) and brown-fat differentiation and myogenesis (d). e, Heat map with relative Z-scores of significantly changed proteins (P ≤ 0.05) in SVF from WT and CLPP-KO mice. f, GSEA enrichment plots for gene signatures related to mitochondrial function. g, Top, heat map showing relative Z-scores of ATF4-target genes in the transcriptome dataset from BAT tissue of CLPP-BKO mice compared with those in WT mice. Bottom, GSEA enrichment plots for ATF4-target gene signatures. h, GSEA enrichment plots for gene signatures related to lipid metabolism. i, Venn diagram depicting overlapping and unique changes in significantly changed transcripts (P ≤ 0.05) in CLPP-BKO BAT compared with changes in mice housed at 4 °C and those initially housed at 4 °C for 1 week and subsequently moved to thermoneutrality (30 °C) for 4 weeks (rewarm versus cold) and LSD1-KO mice (Supplementary Table 2). j, GO:BP analysis of common significantly changed BAT proteins (P ≤ 0.05) from CLPP-KO, CLPP-AKO and CLPP-BKO mice (Supplementary Table 3). k, Average changes of individual OXPHOS complexes isolated from CLPP-KO cytoplasmic mitochondria (CM) or the whole BAT tissue lysate (WTL). Each data point represents average fold change value (CLPP-KO/WT, n = 4) of individual OXPHOS subunits obtained from proteomics analyses (Supplementary Tables 3 and 4). The number of individual OXPHOS complex subunits identified in proteomics determines n (n = 40 (CI); n = 4 (CII); n = 7 (CIII); n = 11 (CIV); n = 14 (CV)). Data are presented as mean ± s.d. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001 as determined by paired two-tailed Student’s t-test. a,j, The size of the dots represents the number of genes and the colour of the dots represent the adjusted P values.

Fig. 3. d-2HG accumulates in CLPP-deficient adipocytes.

a, Volcano plots of metabolites with significantly changed levels in BAT isolated from ubiquitously-deficient (KO), adipose-specific (AKO) or BAT-specific (BKO) CLPP-KO mice, compared with the WT (n = 5); dark colored dots present significantly changed metabolites (P ≤ 0,01, two-fold-change) (as seen in Supplementary Table 5). b–d, 2HG levels in BAT tissue from WT and CLPP-KO, CLPP-AKO and CLPP-BKO mice (n = 10) (b); in WT and CLPP-KO mBA cells (left) and released in media by cultured cells (right), (n = 5) (c); and in WT and CLPP-KO preADs (n = 5). a.u., arbitrary units. e, Relative changes of proteins implicated in production of 2HG, measured in BAT tissue proteomics from CLPP-KO, CLPP-AKO and CLPP-BKO mice compared with levels in WT mice (n = 4). f, Schematic depicting specific roles of PHGDH and the inhibition by NCT503. TCA, tricarboxylic acid cycle. g, Relative percentages of the d and l isoforms of 2HG in WT, and CLPP-KO, CLPP-AKO and CLPP-BKO BAT tissue, as assessed by the derivatization of 2HG using TSCP (n = 5). a–g, Each dataset represents either a single mouse or single cell culture plate. b–e,g, Data are presented as mean ± s.d. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001, as determined by unpaired two-tailed Student’s t-test in c–e and g and one-way ANOVA with multiple comparisons in b. The schematic in f was created using Biorender.com. Source data

Fig. 4. d-2HG accumulation in adipocytes upon CLPP loss causes whitening.

a, 2HG levels as determined by targeted metabolomics in WT, CLPP-KO and NCT503-treated KO mBA cells (n = 5). b, Representative images of WT, CLPP-KO, d-2HG-treated WT and NCT503-treated CLPP-KO mBA cells, stained with DAPI (nucleus) and Bodipy (LDs). Scale bar, 100 μm (images represent results from four independent experiments). c, Lipid levels in NCT503-treated and untreated CLPP-KO mBA cells, normalized to the average MFI value of WT cells (n = 12). d, 2HG levels in WT and CLPP-KO BAT after 14 days of NCT503 treatment in vivo (n = 12 for WT, n = 7 for KO and n = 11 for KO + NCT). e, Top, representative images of whole BAT. Bottom, H&E staining of tissue from WT, CLPP-KO and NCT503-treated KO (KO + NCT) mice. Scale bars, 200 μm. Images represent results from four mice per condition. f, Lipid levels in d-2HG-treated and untreated WT mBA cells, normalized to the average MFI value of WT cells (n = 12). MFI quantification of Bodipy-stained 2HG-treated and untreated WT mBA cells, normalized to the average of the MFI value of the WT cells (n = 5 per condition). Individual data points represent either a single mouse or a single cell culture plate. Data are presented as mean ± s.d. **P < 0.01, ***P < 0.001, ****P < 0.0001, as determined by unpaired two-tailed Student’s t-test in c and f and one-way ANOVA with multiple comparisons in a and d.

Fig. 5. CLPP loss mirrors d-2HG-induced histone methylation.

a, Heatmap and coverage plots of H3K4me3 ChIP–seq in WT, CLPP-KO and 2HG-treated WT mBA cells, ±2 kb around the transcription start site (TSS) (n = 4). b, Upset plot depicting the shared and unique H3K4me3 signatures between WT, CLPP-KO and d-2HG treated WT mBA cells (WT + 2HG). c, GO:BP analysis of H3K4me3 chromatin marks shared between KO and d-2HG-treated WT mBA cells (Supplementary Table 7). The size of the dots represent the number of genes, and the colour of the dots represent the adjusted P values (n = 4). Neg. reg., negative regulation; pos. reg., positive regulation; Pol II, RNA polymerase II. d, Integrative Genomics Viewer browser views showing H3K4me3 read density on the promoters of selected genes (n = 4). e, Venn diagram depicting the number of unique and overlapping changes between transcripts differentially expressed in CLPP-KO mice, compared with distinct H3K4me3 signatures found in KO and d-2HG-treated WT mBA cells. f, GO:BP analysis of common significantly changed transcripts between CLPP-KO and d-2HG-treated WT mBA cells (Supplementary Table 8). The size of the dots represent the number of genes and the colour of the dots represent the adjusted P values (n = 4). Source data

Fig. 6. CLPP loss and d-2HG reshape lipid metabolic pathways.

a, GSEA enrichment plots for gene signatures related to lipid metabolism shared between CLPP-KO cells and WT mBA cells treated with d-2HG. b, Relative mRNA levels of genes under GO:BP analysis term ‘Lipid metabolism’ enriched in CLPP-KO cells and WT cells treated with d-2HG, obtained from RNA sequencing (RNA-seq) analysis (n = 4) (Supplementary Table 8). c, Relative mRNA levels of genes involved in lipid metabolism in CLPP-KO cells upon expression of d-2HGDH, as measured by quantitative real-time PCR. Untreated WT cells, and cells treated with d-2HG, were used as controls (n = 6). d, Relative mRNA levels of major regulators of lipogenesis genes enriched in CLPP-KO cells and WT cells treated with d-2HG obtained from RNA-seq analysis (n = 4) (Supplementary Table 8). e, Relative mRNA levels of genes under GO:BP analysis term ‘Cholesterol metabolism’ enriched in CLPP-KO cells and WT cells treated with d-2HG, obtained from RNA-seq analysis (n = 4) (Supplementary Table 8). f, Cholesterol levels in untreated WT cells, CLPP-KO mBA cells or WT cells treated with either d-2HG or simvastatin (Sim) (n = 5). b–f, Data are presented as mean ± s.d. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001, as determined by one-way ANOVA with multiple comparisons.

Fig. 7. 2HG mediates nuclear softening in CLPP-deficient adipocytes.

a, Representative TEM image from BAT of WT and CLPP-KO mice. LDs (yellow), nucleus (blue) and mitochondria (red) (n = 3 mice). Scale bars, 2 μm. b, 3D reconstruction of serial TEM images from KO BAT tissue (Supplementary Video 1). Scale bars, 1 μm. c, Quantification of nuclear dents in mitochondria (n = 176 fields for WT and n = 150 for KO, from two mice per group). d–g, Quantification of Young’s modulus from AFM-mediated force spectroscopy of mBA nuclei from: WT and CLPP-KO (n = 269 for WT and n = 260 for KO, each from four individual cell culture plates) (d); untreated WT or d-2HG-treated WT (WT + HG) (n = 772 for WT n = 642 for WT + HG, each from four cell culture plates) (e); CLPP-KO, either untreated (KO) or treated with NCT503 (KO + NCT) (n = 1,006 for KO, n = 735 for KO + NCT, each from four cell culture plates) (f); WT, CLPP-KO and WT treated with d-2HG, supplemented with cholesterol (Ch) (g). Untreated WT simvastatin (Sim)-treated cells were used as controls (n = 80 for WT, n = 169 for WT + HG, n = 143 for WT + HG+chl n = 151 for sim, n = 177 for KO and n = 171 for KO+chl, from three cell culture plates). h, 2HG levels as determined by targeted metabolomics in WT and actinonin-treated (WT + Act) mBA cells (n = 5). i, Lipid levels in WT, actinonin-treated WT cells and CLPP-KO mBAs, normalized to the average MFI value of WT cells (n = 11). j, Quantification of Young’s modulus of nucleus of WT mBAs, CLPP-KO and WT cells treated with actinonin (WT + Act) (n = 107 for WT, n = 158 for WT + Act and n = 115 for KO, from three cell culture plates). k, Schematic summarizing the major findings of the study. c,h,i, Data are presented as mean ± s.d. d–g,j, Data are presented using Tukey’s box plot with middle line marking the median, and whiskers show variability within 1.5×IQR. Anything beyond is an outlier presented as individual value. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001, as determined by unpaired two-tailed Student’s t-test in c,h and i and Kolmogorov–Smirnov test in d–g and j. The schematic in k was created using Biorender.com.

Extended Data Fig. 1. Effects of CLPP loss show similar changes in AKO and BKO.

(a) Immunoblot of CLPP from wild type (WT), ubiquitously-deficient CLPP mice (KO), adipose-specific CLPP KO (AKO) or BAT specific CLPP KO (BKO) mice (n = 3, blots representative of four independent experiments); (b) H&E staining of BAT from WT, KO, AKO and BKO mice, scale bar = 200μm (images represent results from three animals per genotype); (c) Gain in body weight with age in AKO and BKO male and female mice, (n = 11 and 12 male of WT and AKO respectively, 10 and 6 female of WT and AKO respectively, n = 6 and 5 male of WT and BKO respectively, 6 and 7 female of WT and BKO respectively); (d-e) In-gel activity assay for (d) Complex I (CI) and (e) Complex V (CV) in WT, KO, AKO and BKO mice, (n = 3); (f) Quantification of lipolysis (glycerol release) in BAT of WT and KO mice under baseline (control) and stimulated conditions (n = 3); (g) Oil red O staining of WT and KO mBA cells (images represent results from three independent experiments). (h) Relative steady-state protein levels (log2) as obtained by label-free proteomics analysis of WT and KO preadipocytes and differentiated WT and KO mBA cells (n = 3); (f and h) Data are presented as mean ± SD. *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001 as determined by One-way ANOVA with multiple comparisons. Source data

Extended Data Fig. 2. CLPP loss in BAT shows no immune or developmental changes.

(a) Flow cytometry analysis of immune cells profile in the BAT from BAT specific CLPP KO (BKO) mice (n = 5); (b) Flow cytometry analysis of immune cells profile from bone marrow of BKO mice, graphs showing quantification of (left to right) lymphoid cells, hematopoietic progenitor cells, monocytes and mentioned progenitor cells (n = 5); (c) (left) Representative pipeline for deconvolution analysis followed; (right) Deconvoluted bulk RNA Seq data from BAT of ubiquitously-deficient CLPP mice (KO) mice, represented as UMAPs, with cell populations marked in different colours; (d) Relative proportions of designated cell types from deconvolution, reveals no quantitative differences in muscle cell population; (e) Comparison of BAT transcriptome from CLPP BKO to gene expression changes observed in BAT whitening models induced either by (left) BAT specific loss of LSD1 or (right) by thermoneutral temperatures. (a and b) Data are presented as mean ± SEM *p < 0.05, as determined by One-way ANOVA with multiple comparisons in a and b. Schematic created using Biorender.com.

Extended Data Fig. 3. Isolated mitochondria from KO BAT have OXPHOS deficiency.

(a) Heatmap of significantly changed protein abundances of mitochondrial proteins from BAT tissue of wild type (WT) and brown adipose specific CLPP deficient (BKO) mice. (b) Heat map of significantly changed protein abundances of mitochondrial proteins from isolated cytoplasmic and peri lipid droplet mitochondria of WT and ubiquitously-deficient CLPP mice (KO) mice, full table can be seen in Supplementary Table 4; (c) Steady-state levels of OXPHOS supercomplexes (SC) in BAT cytoplasmic (CM) or peri lipid droplet (PDM) mitochondria of WT and CLPP KO mice, analysed by BN-PAGE followed by western blot for (left) CI (NDUFA9) and (middle) CIII (UQCRC1) and (right) CV (ATP5A) (n = 3); (d) Relative fold changes of subunits of OXPHOS complexes in isolated CM from BAT and whole BAT tissue lysates, (n = 3 for CM and n = 4 for WTL). (d) Data are presented as mean ± SD. Red text colour represents complexes which were significantly changed in isolated cytoplasmic mitochondria and red text with # represents complexes significantly changed compared to whole BAT tissue lysates, as determined by paired two-tailed Student’s t-test in d. Source data

Extended Data Fig. 4. 2-HG accumulates in BAT and is potentially produced by PHGDH.

(a) Volcano plot of significantly changed metabolites in CLPP deficient (KO) mBA cells compared to wild type (WT) cells (n = 5), dark colored dots present significantly changed metabolites (p≤0,01, two-fold-change); (b) Relative transcript abundances of PHGDH in BAT from ubiquitously-deficient (KO), adipose-specific (AKO) or BAT specific (BKO) CLPP KO mice, compared to the wild type wild type (WT) (n = 4); (c) Immunoblot and quantification of steady state levels of PHGDH from WT and KO mBA cells (n = 3, western blot analysis was repeated with additional three independent samples/group); (d) Quantification of PHGDH enzymatic activity (n = 5) (e) Relative percentages of 2HG D and L isoforms in WT and CLPP KO mBA cells, as assessed by the derivatization of 2HG using TSCP (n = 5); (f) Quantification of 2HG by targeted metabolomics in KO, D2HGDH overexpressing KO (KO + D2HGDH), NCT503 treated KO (KO + NCT) and NCT503 treated KO + D2HGDH (KO + D2HGDH + NCT) mBA cells, (n = 5); (g) Lipid levels in wild type (WT), D2HGDH overexpressing WT (WT + D2HGDH), KO and KO + D2HGDH mBA cells, normalized to the average of the MFI value of the WT cells (n = 11,11,11,14 respectively); (h) Immunoblot of steady state levels of ISR proteins from wild type (WT), ISRIB treated WT (WT + ISRIB), KO and ISRIB-treated KO (KO + ISRIB) mBA cells (n = 3) (blots representative of three independent experiments); (i) Quantification of 2HG by targeted metabolomics from WT, KO and KO + ISRIB mBA cells (n = 5); (j) Lipid levels in WT, WT + ISRIB, KO and KO + ISRIB mBA cells, normalized to the average of the MFI value of the WT cells (n = 4). (b-g, i-j) Data are presented as mean ± SD. *p < 0.05, ***p < 0.001, ****p < 0.0001, as determined by One-way ANOVA with multiple comparisons in b, e, f, h and i and Students paired two tailed t-test in d; ^##p < 0.001 represent significant difference in the ratio between D- and L-forms between WT and KO, as determined by Students paired two tailed t-test. Source data

Extended Data Fig. 5. Omics analysis shows remodelling of mBA cells upon treatments.

(a) Immunoblot of steady state levels of indicated histone modifications in wild type (WT), D-2HG treated WT (WT + HG) and CLPP deficient (KO) mBA cells, (right) quantification of levels of the histone marks normalized to H3 from the immunoblots, (n = 3); (b) Single nuclei intensity quantitation of histone marks in untreated WT, WT + HG and KO mBA (n = 798 for WT, n = 825 for WT + HG and n = 1088 for KO, collectively obtained from three individual cell culture plates); (c) Peak count frequency profile depicting enrichment of H3K4me3 peaks from CHIP Seq of WT, KO and WT + HG mBA cells relative to the TSS; (d) Feature distribution bar plot depicting relative proportions of genomic regions enriched in H3K4me3 in WT, KO and WT + HG mBA cells; (e) Relative mRNA levels of genes involved in lipid transport in CLPP KO cells. Untreated WT cells, and cells treated with D-2HG are used as controls (n = 3); (f) GSEA enrichment plots for gene signatures related to sterol biosynthesis in mBA cells and BAT tissue of KO mice compared to WT. (a, e) Data are presented as mean ± SD. (b) Data are presented using Tukey’s box plot with middle line marking the median, and whiskers show variability within 1.5×IQR. Anything beyond is an outlier presented as individual value. *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001, as determined by One-way ANOVA with multiple comparisons in a, b and e. Source data

Extended Data Fig. 6. Nuclear architecture changes upon CLPP loss in BAT.

(a) Venn diagram representing GO:CC analysis of common differentially expressed genes in CLPP deficient (KO) and 2HG treated WT (WT + HG) cells compared to untreated wild type (WT) mBA cells; (b) GSEA enrichment plots for gene signatures related to nuclear lamina in both KO mBA cells and KO BAT tissue compared to WT; (c) Representative TEM images showing mitochondria impinging on the nucleus and juxtaposed to nuclear pores (images represent results observed in 19 TEM sections from KO BAT of three animals); (d) Quantitation of the number of ER sites spotted in TEM images from WT and KO mice, represented as percentage of the total number of fields assessed; (e) Quantification of Young’s modulus in KO and ISRIB-treated KO mBA nuclei (n = 110 for KO, n = 120 for KO + ISRIB and n = 1088 for KO, collectively obtained from three individual cell culture plates). (e) Data are presented using Tukey’s box plot with middle line marking the median, and whiskers show variability within 1.5×IQR. Anything beyond is an outlier presented as individual value. No significant differences were observed as assessed by paired two-tailed Student’s t-test.