Lysine-specific demethylase-2 is distinctively involved in brown and beige adipogenic differentiation

Abstract

Transcriptional and epigenetic regulation is fundamentally involved in initiating and maintaining progression of cellular differentiation. The 2 types of thermogenic adipocytes, brown and beige, are thought to be of different origins but share functionally similar phenotypes. Here, we report that lysine-specific demethylase 2 (LSD2) regulates the expression of genes associated with lineage identity during the differentiation of brown and beige adipogenic progenitors in mice. In HB2 mouse brown preadipocytes, short hairpin RNA–mediated knockdown (KD) of LSD2 impaired formation of lipid droplet–containing adipocytes and down-regulated brown adipogenesis–associated genes. Transcriptomic analysis revealed that myogenesis-associated genes were up-regulated in LSD2-KD cells under adipogenic induction. In addition, loss of LSD2 during later phases of differentiation had no obvious influence on adipogenic traits, suggesting that LSD2 functions during earlier phases of brown adipocyte differentiation. Using adipogenic cells from the brown adipose tissues of LSD2-knockout (KO) mice, we found reduced expression of brown adipogenesis genes, whereas myogenesis genes were not affected. In contrast, when LSD2-KO cells from inguinal white adipose tissues were subjected to beige induction, these cells showed a dramatic rise in myogenic gene expression. Collectively, these results suggest that LSD2 regulates distinct sets of genes during brown and beige adipocyte formation.—Takase, R., Hino, S., Nagaoka, K., Anan, K., Kohrogi, K., Araki, H., Hino, Y., Sakamoto, A., Nicholson, T. B., Chen, T., Nakao, M. Lysine-specific demethylase-2 is distinctively involved in brown and beige adipogenic differentiation.

Transcriptional and epigenetic mechanisms underlie cellular differentiation processes and tissue maintenance. During differentiation, transcription factors (TFs) and chromatin modifiers synergistically enable cell type–specific gene expression in a spatially and temporally controlled manner (1). Given that DNA and histone modifications are central components of epigenetic gene regulation, understanding the function of individual epigenetic factors is crucial for deciphering complex differentiation processes (2, 3).

Based on the understanding that obesity is linked to an increased risk of type 2 diabetes, cardiovascular disease, and other disorders (4–6), mechanisms underlying adipogenic differentiation and subsequent lipid metabolism have attracted great interest. The 2 distinct classes of adipocytes, white and brown, differentially contribute to energy homeostasis. White adipocytes reside in the intraperitoneal and subcutaneous white adipose tissues (WATs) and are specialized for the storage of energy in the form of triglycerides, whereas brown adipocytes residing in brown adipose tissues (BATs) dissipate energy in the form of heat, a process known as nonshivering thermogenesis (7). Brown adipocytes are rich in mitochondria, contain small lipid droplets, and express key thermogenic genes such as uncoupling protein 1 (UCP1) (8). They are derived from myogenic factor 5 (MYF5)–positive precursors in dermomyotome, sharing a common developmental origin with skeletal myocytes (9). In addition, brown-like cells called beige adipocytes appear in WAT depots in response to chronic exposure to low temperature, β3-adrenergic receptor agonists, and peroxisome proliferator-activated receptor (PPAR)γ agonists (10). Previous reports have shown that bipotent progenitors in WAT depots give rise to both white and beige cells (11, 12). Brown and beige adipocytes share similar regulatory mechanisms of differentiation (13, 14). In addition to general adipogenic TFs such as PPARγ and CCAAT/enhancer-binding proteins, TFs such as PR domain containing 16 (PRDM16) and early B-cell factor 2 contribute to the formation of brown and beige cells. Certain epigenetic factors cooperate with these TFs to regulate brown and beige cell identity and plasticity (15–17).

Lysine-specific demethylases (LSDs) 1 and 2, also known as lysine-specific histone demethylase 1A and lysine-specific histone demethylase 1B, respectively, comprise the flavin adenine dinucleotide–dependent amine oxidase family of histone demethylases (18). These enzymes demethylate histone (H3) lysine 4 and 9 (H3K4 and H3K9, respectively), thus contributing to both repression and activation of gene expression (19). LSD2 has been shown to regulate a number of biologic processes, such as inflammatory responses and the establishment of maternal genomic imprinting in oocytes (20, 21). We previously reported that LSD2 represses the expression of genes associated with lipid transport and metabolism through H3K4 demethylation in hepatic cells, thereby preventing excess lipid flux and metabolism (22). Previous studies have reported that LSD1 plays pivotal roles in the metabolic adaptation of adipocytes in response to nutritional or hormonal cues (23, 24). Moreover, a recent report demonstrated that LSD1 is an essential regulator of brown adipogenesis (17). Although LSD1 is often implicated in adipocyte differentiation and metabolism, the roles of LSD2 in regulating brown adipogenesis and the plasticity of white adipocytes have not been documented.

In this study, we evaluated the role of LSD2 in brown and beige adipocyte differentiation. We found that LSD2 knockdown (KD) severely impaired lipid accumulation in murine BAT-derived HB2 adipocytes. LSD2-KD HB2 cells showed reduced expression of genes associated with brown adipogenesis and thermogenesis, whereas the expression of myogenic genes increased, suggesting that LSD2 was important for maintaining brown adipose cell lineage in HB2 cells. When brown adipogenic cells isolated from LSD2-knockout (KO) mice were induced to differentiate, expression of BAT-associated genes was reduced but that of myogenic genes was unaffected. Conversely, in progenitors derived from inguinal WAT (iWAT) of LSD2-KO mice, expression of myogenic genes was dramatically increased under beige induction. Collectively, these results suggest that LSD2 contributes to the maintenance of brown adipocyte lineage and the plasticity of white adipose cells.

MATERIALS AND METHODS

Cell culture

HB2 brown preadipocyte cell line derived from BAT of p53-KO mice was generously provided by Dr. Kazuhiro Kimura (Hokkaido University, Sapporo, Japan) (25). HB2 cells were maintained in DMEM supplemented with 10% heat-inactivated fetal bovine serum (FBS). For differentiation, confluent cells were cultured for the first 2 d in induction medium composed of DMEM supplemented with 10% FBS, 0.5 mM 3-isobutyl-1-methylxanthine (Calbiochem, San Diego, CA, USA), and 1 µM dexamethasone (MilliporeSigma, Burlington, MA, USA) (Fig. 1A). Then, cells were cultured for an additional 3–5 d in differentiation medium composed of DMEM supplemented with 10% FBS, 10 μg/ml insulin (MilliporeSigma), and 50 nM 3-3′-5-triiodo-L-thyronine (T3; MilliporeSigma). Differentiation was assessed by Oil Red O staining as previously described in Hino et al. (23).

Figure 1.

Increased expression of LSD2 during BAT differentiation of HB2 cells. A) Differentiation protocol for HB2 cells. B) Lipid accumulation in HB2 cells during differentiation. Unstained (d 0, 2, and 7) and Oil Red O–stained cells (d 7) are shown. Scale bars, 100 µm. C–E) Gene expression changes during differentiation. Genes for adipogenesis (C), pan-adipocyte (D), and BAT-selective (E) are shown. Real-time qPCR values were normalized to values for the 36B4 gene and are shown as fold differences against d 0. IBMX, 3-Isobutyl 1-methylxanthine; Cebp, CCAAT-enhancer binding protein; Elovl3, elongation of very long chain fatty acids (FEN1/Elo2, SUR4/Elo3, yeast)-like 3a; Cidea, Cidea cell death-inducing DNA fragmentation factor, alpha subunit-like effector A; Fabp4, fatty acid binding protein 4; DEX, dexamethasone. Values are mean ± sd (n = 3). *P < 0.05, ^†P < 0.01. F, G) LSD2 and LSD1 expression during differentiation. The levels of relative mRNAs (F) and proteins (G) are shown. β-Tubulin was used as a loading control.

Animals

Animal experiments were conducted in accordance with the guidelines of the Animal Care and Use Committee of Kumamoto University (Kumamoto, Japan). Lsd2-deficient (Lsd2/Aof1^1lox) mice were generated as reported in Ciccone et al. (20). These mice were backcrossed to C57BL/6J mice for several generations before use. Because Lsd2^1lox/1lox females were infertile, Lsd2^wt/1lox mice were mated to generate Lsd2-null mice. To evaluate the adiposity of these mice, body weight and adipose tissue weights (BAT and epididymal WAT) were measured. Tissue sections were then fixed with paraformaldehyde, embedded in paraffin, and subjected to hematoxylin and eosin staining.

Isolation and culture of mouse stromal vascular fraction cells

Stromal vascular fraction (SVF) cells from mouse adipose tissues were used for assessing adipogenic differentiation. Interscapular BAT and iWAT were dissected and minced with scissors and a scalpel blade. Minced tissues were incubated at 37°C for 30 min in digestion buffer containing 3 mg/ml collagenase type I (Thermo Fisher Scientific, Waltham, MA, USA), 2 mg/ml dispase II (MilliporeSigma), 4% bovine serum albumin (Fujifilm Wako Pure Chemical, Osaka, Japan), 15 mM NaHCO₃, and 25 μg/ml gentamicin sulfate (Fujifilm Wako Pure Chemical) in Krebs-Ringer solution (MilliporeSigma). After filtration through a 100-μm nylon mesh, dispersed cells were centrifuged at 700 g for 10 min at 25°C. The supernatant containing mature adipocytes was discarded, and cells from the SVF were pelleted. Cells were maintained in DMEM containing 10% FBS, 0.5 μg/ml insulin, and 1 nM T3. For induction of brown and beige differentiation, confluent cells were treated with 125 mM indomethacin (Fujifilm Wako Pure Chemical), 0.5 mM 3-isobutyl-1-methylxanthine, 1 µM dexamethasone, 10 mg/ml insulin, 10 μM T3, and 0.5 µM rosiglitazone (Rosi; Fujifilm Wako Pure Chemical) for 2 d. Thereafter, the cells were cultured for an additional 3–5 d in differentiation medium (DMEM supplemented with 10% FBS, 0.5 μg/ml insulin, 1 nM T3, and 1 µM Rosi).

Western blot analysis

To prepare total cell lysates, cells were collected and suspended in 2 times sample buffer (0.1 M Tris-HCl, pH 6.8, 4% SDS, 0.1 M DTT, 20% glycerol, and 0.2% bromophenol blue). Following sonication and centrifugation at 17,800 g for 10 min at 4°C, supernatant was collected and used for Western blotting. For detecting modified histones, histone extracts were solubilized in 0.2 N HCl. Protein samples were electrophoresed on an SDS-polyacrylamide gel and then transferred to a nitrocellulose membrane (Amersham Protran Premium; GE Healthcare, Waukesha, WI, USA) using a semidry method. After blocking for 1 h using 5% skim milk in PBS and 0.3% Tween 20, the membrane was incubated overnight at 4°C with primary antibodies in Can Get Signal solution (Toyobo, Osaka, Japan). The following primary antibodies were used: anti-LSD2 [rabbit polyclonal (22); 1:1000], anti-LSD1 (rabbit polyclonal, ab17721, 1:500; Abcam, Cambridge, MA, USA), anti-UCP1 (rabbit polyclonal, ab10983, 1:500; Abcam), anti-β-tubulin (mouse monoclonal, T4026, 1:1000; MilliporeSigma), anti-histone (H3) (rabbit polyclonal, ab1791, 1:1000; Abcam), anti–monomethylated H3K4 (rabbit polyclonal, 07-436, 1:1000; MilliporSigma), anti-dimethylated H3K4 (rabbit polyclonal, 07-030, 1:1000; MilliporeSigma), and anti-trimethylated H3K4 (rabbit polyclonal, 07-473, 1:1000; MilliporeSigma). The secondary antibodies used were anti-mouse IgG and anti-rabbit IgG–horseradish peroxidase (NA931V and NA934V, respectively; GE Healthcare). Blots were incubated for 1 min with Western Lightning Plus-ECL solution (PerkinElmer, Waltham, MA, USA) and visualized using ImageQuant LAS 4000 Mini (GE Healthcare). For quantification, band densities were determined using ImageJ (National Institutes of Health, Bethesda, MD, USA) software.

Quantitative RT-PCR

Trizol reagent (Thermo Fisher Scientific) was used to extract total RNA from cells and tissues. cDNA was synthesized from 500 ng of total RNA using ReverTra Ace quantitative PCR (qPCR) RT Kit (Toyobo). SYBR Green–based qPCR was done using Thunderbird reagents (Toyobo) on a StepOnePlus Real-Time PCR instrument (Thermo Fisher Scientific). Fold differences among groups were calculated using the ∆∆C_t method. The 36B4 (Rplp0: ribosomal protein, large, P0) gene was used as an internal control. Primer sequences are shown in Supplemental Table S1.

Lentiviral expression of short hairpin RNA

Lentiviral vectors for tetracycline (tet)-inducible short hairpin RNA (shRNA) expression were constructed using plasmids obtained from Riken BioResource Research Center (Saitama, Japan) according to the guidelines (http://cfm.brc.riken.jp/Lentiviral_Vectors_J). Double-stranded DNA, 71 bp in length, harboring shRNA sequence was inserted into an entry vector, pENTR4-H1tetOx1. Target sequences of shRNA were as follows: sh control (targeting firefly luciferase) 5′-CTTACGCTGAGTACTTCGA-3′, shLSD2_#1 5′-GAACACTTCTGCAATGAAT-3′, and shLSD2_#2 5′-CAAACAGGCATTTCCCACA-3′. Using the Gateway recombination method (developed by Thermo Fisher Scientific), a tet-inducible shRNA cassette was cloned into a lentiviral vector, CS-RfA-ETBsd. For the production of viral particles, lentiviral packaging (pCAG-HIVgp) and envelope (pCMV-VSV-G-RSV-Rev) plasmids were transfected into Lenti-X 293T cells (Takara Bio, Kusatsu, Japan) using FuGENE 6 reagent (Promega, Madison, WI, USA). At 48 hr after transfection, culture supernatant was filtered and subjected to titration. For viral transduction, subconfluent HB2 cells were infected with control, shLSD_#1, or shLSD_#2 viruses at 1.5 × 10⁸ plaque-forming units per well. Cells were subjected to blasticidin selection to establish stable lines. For induction of shRNA expression, doxycycline (Dox) was added to DMEM at 500 ng/ml.

Metabolic analysis using an extracellular flux analyzer

Mitochondrial oxygen consumption rate (OCR) and extracellular acidification rate (ECAR) were measured using an XF24 extracellular flux analyzer (Agilent Technologies, Santa Clara, CA, USA). After trypsinization of the cells, 40,000 cells per well were seeded in a 24-well assay plate and incubated for 12 h until the cells adhered. Subsequently, measurement was carried out in unbuffered assay medium (25 mM glucose, 1 mM sodium pyruvate, and 2 mM L-glutamine in DMEM, pH 7.4).

Poly(A) RNA sequencing

Total RNA from HB2 cells and primary SVF was extracted using Trizol reagent. mRNA was purified using NEBNext Poly(A) mRNA Magnetic Isolation Module (New England Biolabs, Ipswich, MA, USA). For sequencing, a cDNA library was synthesized using the NEBNext Ultra DNA Library Prep Kit for Illumina (New England Biolabs). Sequencing was performed on a NextSeq 500 Sequencer (Illumina, San Diego, CA, USA) with 75-bp single-end reads. Resulting reads were mapped to the University of California–Santa Cruz (Santa Cruz, CA, USA) mm10 reference genome using HISAT2 v.2.0.5.2 (Center for Computational Biology, Johns Hopkins University, Baltimore, MD, USA). Number of reads was calculated using the HTSeq Python (http://htseq.readthedocs.io/) module, and comparative quantification of gene expression was analyzed by DESeq2 pipeline (Galaxy package v.2.11.39; https://toolshed.g2.bx.psu.edu/view/iuc/deseq2/24a09ca67621). Gene set enrichment analysis (GSEA) was performed using GSEA v.2.0 from the Broad Institute (Cambridge, MA, USA; http://www.broadinstitute.org/gsea/).

Chromatin immunoprecipitation

Chromatin immunoprecipitation (ChIP) analyses for detecting histone modifications were done as previously described in Hino et al. (23). In brief, cells were fixed in 1% formaldehyde prior to nuclear extraction. Chromatin fragmentation was done using a water bath sonicator (Bioruptor; Cosmo Bio, Tokyo, Japan). Solubilized chromatin was incubated with antibodies against methylated H3K4 as listed above. DNA was purified by ethanol precipitation and subjected to qPCR analyses. Primers used are listed in Supplemental Table S1.

Statistical analyses

Equality of variance was examined using an F test. Statistical analyses between 2 groups were performed using either Student’s t test for equal variance or Welch’s t test for unequal variance.

RESULTS

Increased expression of LSD2 during brown adipose differentiation

To elucidate the function of LSD2 in brown adipocytes, we used HB2, a clonal preadipocyte line derived from BAT in p53-KO mice (25). Upon differentiation induction, HB2 cells underwent morphologic changes at an early phase and finally accumulated lipid droplets in the cytoplasm at d 7 (Fig. 1A, B). Accordingly, key adipogenic genes such as CCAAT/enhancer-binding protein and Pparγ and pan-adipocyte marker genes such as Elovl3, fatty acid binding protein 4 (Fabp4), and Cidea showed increased expression after induction (Fig. 1C, D). Acquisition of BAT character was confirmed by the increased expression of BAT-selective genes such as Pgc1α, Prdm16, and Ucp1 (Fig. 1E). Notably, we found that LSD2 expression was up-regulated during differentiation, whereas LSD1 expression was not (Fig. 1F, G). A similar increase in LSD2 expression was observed during the differentiation of SVF cells from mouse BAT (Supplemental Fig. S1). These results suggest that LSD2 is dynamically controlled in the differentiation process of brown adipocytes.

Loss of LSD2 impairs the differentiation of HB2 brown preadipocytes

To investigate the function of LSD2 in brown adipogenesis, we generated HB2 cells harboring lentiviral vectors that allow tet-inducible LSD2-KD. A total of 2 different shRNAs were used to specifically suppress Lsd2 gene expression (shLSD2_#1 and shLSD2_#2). We confirmed that 48 h after Dox addition, the expression of LSD2 declined substantially, whereas that of LSD1 was not affected (Fig. 2A, B and Supplemental Fig. S2B, C). To test the effect of LSD2-KD on brown differentiation, Dox was added to the medium 48 h before the induction of differentiation (Supplemental Fig. S2A). In Dox-treated shLSD2_#1 and shLSD_#2 cells, marked reduction in fat accumulation was observed, as revealed by Oil Red O staining (Fig. 2C). Consistent with these results, induction of both adipogenesis and BAT-selective genes was significantly attenuated by the LSD2 depletion (Fig. 2D and Supplemental Fig. S2D). These results clearly indicate that the loss of LSD2 interfered with the brown adipogenesis program in HB2 cells.

Figure 2.

LSD2 is essential for BAT differentiation in HB2 cells. A, B) Dox-inducible KD of LSD2. HB2 cells were transduced with tet-inducible shLSD2 (LSD2-KD#1 and #2) or control (shCtr) lentiviral vectors followed by blasticidin selection. Cells were cultured with or without Dox for 48 h. For relative mRNA levels of LSD2 (A), real-time quantitative PCR values were normalized to values for the 36B4 gene and are shown as fold differences against Dox (−). For levels of LSD2 protein (B), β-tubulin was used as a loading control. Values are mean ± sd (n = 3). *P < 0.05, ^†P < 0.01. C) Oil Red O staining of shRNA-expressing HB2 cells at d 7. Culture details are described in Supplemental Fig. S1A. Scale bars, 100 µm. D) Expression of adipogenesis and BAT-selective genes in shLSD2_#1 cells at indicated days. Quantitative RT-PCR values were normalized to those for the 36B4 gene and are shown as fold differences against Dox (−) at d 0. Values are mean ± sd (n = 3). *P < 0.05, ^†P < 0.01 vs. Dox (−) at each time point.

LSD2 is required for metabolic reprogramming during differentiation in HB2 cells

LSD2-KD cells showed reduced expression of UCP1, a key protein that enhances oxygen consumption through respiratory uncoupling (Figs. 2D and 3A and Supplemental Fig. S2D) (26, 27). To test whether LSD2-KD affected metabolic properties, we characterized the energy metabolism profiles of these cells using an extracellular flux analyzer. Prior to differentiation (d 0), no obvious difference in OCR as well as ECAR (an index of glycolytic activity) was detected between control and LSD2-KD cells (Fig. 3B). After differentiation (d 5), LSD2-KD cells [Dox (+)] exhibited relatively lower OCR and ECAR than those of control cells [Dox (−)]. Importantly, thermogenic stimulation by an addition of isoproterenol activated oxygen consumption in control cells but not in LSD2-KD cells. These results suggest that LSD2 plays an essential role in metabolic reprogramming that enables respiratory uncoupling in response to thermogenic stimuli in HB2 cells.

Figure 3.

Decreased OCR in LSD2-KD HB2 adipocytes. A) UCP1 protein levels in LSD2-KD HB2 cells. β-Tubulin was used as a loading control. B) OCR and ECAR in LSD2-KD cells at d 0 and 5 (differentiated). shLSD2_#1 HB2 cells were cultured with or without Dox, and oxygen consumption was activated by an addition of isoproterenol for 4 h before metabolic measurements. Values are mean ± sd (n = 3). *P < 0.05, ^†P < 0.01.

LSD2 is indispensable during the early phase of differentiation in HB2 cells

Brown adipocyte differentiation progresses stepwise via the cooperation of various TFs (13). To further understand the function of LSD2 in differentiation, we added Dox to the medium 2 d after the induction of differentiation (d 2) [doxycycline-differentiation (Dox-diff)] (Fig. 4A). We confirmed that efficient LSD2 KD could be achieved in this experimental setting (Fig. 4B). In contrast to cells exposed to Dox throughout differentiation [Dox (+)], Dox-diff cells accumulated lipids to a level comparable with control cells [Dox (−)] (Fig. 4C). Consistent with this finding, Dox-diff and Dox (−) cells expressed adipogenesis and BAT-selective genes at similar levels (Fig. 4D). These results suggest that LSD2 is essential during the early phase, rather than the late phase, of HB2 cell differentiation.

Figure 4.

LSD2 is indispensable in the early phase of differentiation of HB2 cells. A) Dox was added throughout differentiation (Dox +) or during the late phase (Dox-diff.). B) Lsd2 expression in LSD2-KD (shLSD2_#1) cellsReal-time quantitative PCR values were normalized to those for the 36B4 gene and are shown as fold differences against d 0 Dox (−). Values are means ± sd, n = 3. ^†P < 0.01 vs. Dox (−). C) Oil Red O staining of LSD2-KD cells at d 7. Scale bars, 100 µm. D) Relative mRNA levels of adipogenesis and BAT-selective genes in LSD2-KD cells. Quantitative RT-PCR values were normalized to values for the 36B4 gene and are shown as fold differences against Dox (−). Values are mean ± sd (n = 3). ^†P < 0.01 vs. Dox (−).

Down-regulation of brown differentiation genes and up-regulation of myogenesis-associated genes in LSD2-KD HB2 cells

Next, to clarify the role of LSD2 at the early stages of differentiation, RNA sequencing (RNA-seq) analysis was performed using control and LSD2-KD cells 24 h after differentiation induction. A total of 24 genes showed higher expression in LSD2-KD cells than in controls with statistical significance, whereas 100 genes showed lower expression (Fig. 5A and Supplemental Table S2). GSEA of RNA-seq data showed that genes involved in brown adipogenesis were significantly down-regulated in LSD2-KD cells (Fig. 5B, C). Of note, genes associated with muscle development and differentiation were up-regulated by LSD2 KD, suggesting that LSD2 demarcates brown adipose and muscle lineages in HB2 cell differentiation.

Figure 5.

Down-regulation of BAT differentiation genes and up-regulation of skeletal muscle differentiation genes in LSD2-KD HB2 cells. A) An MA plot of RNA-seq data from LSD2-KD HB2 cells (shCtr, n = 2; shLSD2_#1 and #2, n = 2) at differentiation d 1. The MA plot shows the log₂ fold changes attributable to a given variable over the mean of normalized counts for all the samples in analyzing RNA-seq data with DESeq2 package Galaxy v.2.11.39. Red dots indicate genes with the adjusted P < 0.1. B, C) GSEA results showing the gene sets significantly affected by LSD2 depletion. A total of 15 pathways were identified with statistical significance (P < 0.05 and false discovery rate, q < 0.25). Pathways related to brown adipocyte differentiation, lipid metabolism, and myogenesis are highlighted in red. Val., value.

Effect of LSD2 KD on H3K4 methylation in HB2 cells

Because LSD2 serves as an H3K4 demethylase, we examined the effect of LSD2 KD on H3K4 methylation status. Only a modest rise in the H3K4 dimethylation level was detected in the histone extracts from LSD2-KD cells (Fig. 6A), suggesting that LSD2 regulated H3K4 methylation at target genomic regions. To elucidate the involvement of LSD2-mediated epigenetic regulation in transcriptional control, we analyzed the H3K4 methylation levels at regulatory regions of myogenic genes (Fig. 6B). LSD2 KD led to an increased H3K4me1 at a putative enhancer in Myf5 loci (Fig. 6B), which is indicative of an activated enhancer activity. In addition, a promoter of Prdm16, whose expression was reduced by LSD2 KD, showed a decline in H3K4me3 level in LSD2-KD cells. These results suggest that LSD2 is involved in the epigenetic regulation of lineage-associated genes in HB2 cells.

Figure 6.

Effects of LSD2 KD on H3K4 methylation states in HB2 cells. A) Western blot analyses of methylated H3K4 using histone extracts. HB2_shLSD2_#1 cells were cultured with or without Dox for 48 h followed by differentiation induction for 24 h. Densities of each band determined using ImageJ software are indicated. B) ChIP-qPCR analyses of methylated H3K4 at myogenic gene enhancers and BAT-associated gene promoters. A total of 3 independent ChIP analyses were done. Realt-time qPCR values from each experiment were first normalized by input, and then the fold difference in Dox (+) cells over Dox (−) cells was calculated.

Differential effects of LSD2 on brown and beige adipogenesis in primary preadipocytes from LSD2-KO mice

In order to investigate the role of LSD2 in brown adipocyte differentiation, we examined the differentiation capacity of SVF cells from BAT in LSD2-KO mice (Fig. 7A) (20). Although LSD2-KO cells developed into lipid-accumulating adipocytes (Fig. 7B), the expression of adipogenesis and BAT-selective genes were down-regulated (Fig. 7D). Preadipocytes from subcutaneous iWAT can be differentiated into beige cells by Rosi treatment (Fig. 7A) (28). We found that the expression of LSD2 increased during the beige induction, suggesting its involvement in gene regulation (Supplemental Fig. S1C). LSD2-KO beige cells accumulated lipids to a lesser extent in comparison with the wild type (WT) (Fig. 7C). Notably, in LSD2-KO beige cells, expression of adipogenesis genes was not altered, but key myogenesis genes, including Myf5, myogenic differentiation 1 (Myod1), and myogenin (Myog), were dramatically up-regulated (Fig. 7E).

Figure 7.

Differential effects of LSD2-KO on BAT and beige differentiation in primary preadipocytes from mice. A) Primary preadipocytes from brown and inguinal adipose tissues were subjected to BAT and beige differentiation, respectively. SVF from WT and LSD2-KO mice were used as preadipocytes. B, C) BAT (B) and beige (C) differentiation of WT and LSD2-KO preadipocytes, assessed by Oil Red O staining. Scale bars, 100 µm. D, E) Relative mRNA levels of adipogenesis, BAT-selective, pan-adipocyte, and myogenesis genes under BAT (D) and beige (E) differentiation. Real-time qPCR values were normalized to those for the 36B4 gene and are shown as fold differences against WT. Values are mean ± sd (n = 3). *P < 0.05, ^†P < 0.01.

We further performed a transcriptomic analysis of WT and LSD2-KO iWAT cells at d 1 of differentiation (Fig. 8A). A total of 412 genes showed higher expression, whereas 206 genes showed lower expression in LSD2-KO cells than in WT cells (Supplemental Table S3). The data clearly reveal that the loss of LSD2 led to increased expression of myogenesis-associated genes (Fig. 8B, C). We noticed that expression levels of myogenic genes were much lower in BAT SVF than in iWAT SVF based on qPCR ΔC_t values (unpublished data), suggesting that repression of myogenic genes in iWAT SVF cells is relatively loose. These results suggest that LSD2 maintains both the brown and beige adipogenic programs by regulating distinct sets of genes (see Discussion).

Figure 8.

Increased expression of skeletal muscle–associated genes in beige-induced LSD2-KO cells. A) An MA plot of RNA-seq data from WT and LSD2-KO iWAT cells at d 1 of beige induction (WT, n = 2 and LSD2-KO, n = 2). The MA plot shows the log₂ fold changes attributable to a given variable over the mean of normalized counts for all the samples in analyzing RNA-seq data with DESeq2 package Galaxy v.2.11.39. Red dots indicate the genes with the adjusted P < 0.1. B, C) GSEA results showing the gene sets affected by LSD2 KO. A total of 12 pathways were identified with statistical significance (P < 0.05 and false discovery rate, q < 0.25). Pathways related to myogenesis are highlighted in red.

LSD2-mediated myogenic suppression in beige adipogenesis irrespective of PPARγ activation

To induce the differentiation of brown and beige primary adipocytes (Fig. 7), we used Rosi, a potent agonist of PPARγ that promotes brown adipogenesis by increasing the stability of PRDM16 (27). To gain insight into the functional relationship of the PPARγ-PRDM16 axis to LSD2-mediated gene regulation, we induced the differentiation of BAT- and iWAT-derived SVF cells in the absence of Rosi (Supplemental Fig. S3A–C). Expression levels of adipogenesis genes were similar in WT and LSD2-KO brown adipocytes (Supplemental Fig. S3D), clearly contrasting with results from the Rosi-stimulation experiment (Fig. 7D). In the absence of Rosi, LSD2-KO iWAT cells had lower expression of BAT-selective genes than did WT cells, but Rosi exposure activated their expression to comparable levels in both cells (Supplemental Fig. S3E, F). This suggests that the beige cell program was suppressed by loss of LSD2, but the suppression was compensated for by the effect of Rosi. On the contrary, expression of myogenesis genes was significantly induced by LSD2-KO, regardless of Rosi treatment (Fig. 7E and Supplemental Fig. S3E, F). These results suggest that LSD2 represses myogenic gene expression via a PPARγ-independent mechanism.

DISCUSSION

Enhancing the thermogenic properties of adipose tissue has attracted much interest because of the associated benefits in preventing and controlling obesity-linked metabolic disorders (29, 30). To further this goal, efforts have been devoted to understanding the development and plasticity of thermogenic adipocytes. TFs, such as PPARγ and PRDM16, have been reported as central players in the brown and beige cell program by cooperating with epigenetic regulators (14). Given that significant heterogeneity in the adipogenic-precursor population exists (10), multiple epigenetic mechanisms might operate in maintaining lineage identity and plasticity. Here, we demonstrated that the loss of LSD2 significantly altered the gene expression profiles during brown and beige adipogenesis. Notably, LSD2-KO SVF cells from BAT showed attenuated expression of BAT-associated genes upon differentiation, whereas those from iWAT showed dramatically increased expression of myogenic genes upon the induction of beige adipocyte differentiation. These results suggest that LSD2 secures lineage identity and plasticity of both brown and beige adipocytes by regulating distinct sets of genes. Furthermore, adipose tissues in LSD2-KO mice did not exhibit clear alterations in tissue mass, cell morphology, and adipocyte marker gene expression (Supplemental Fig. S4A–E). In addition, although inactivation of Prdm16 has been implicated in the development of fibrosis in epididymal WAT (31), we did not observe changes in the expression of fibrosis-associated genes in LSD2-KO mice (Supplemental Fig. S4F). Thus, our in vitro and in vivo observations possibly reflect the potential function of LSD2 in particular environmental conditions.

We observed that the loss of LSD2 in HB2 preadipocytes led to severely impaired lipid accumulation, whereas LSD2-KO BAT SVF cells did not exhibit such defects. This may be due to the dissimilarity in differentiation stages of these cells. Because HB2 cells are immortalized cells with an ability to self-renew, these cells may be more immature than primary SVF cells with limited proliferative capacity. In addition, HB2 cells lack p53 expression, which has been reported as an essential factor for brown adipogenesis (32), suggesting that HB2 cells have evolved a unique strategy for differentiation control. Nonetheless, BAT-associated genes were down-regulated in both LSD2-KD HB2 cells and LSD2-KO BAT SVF, indicating the involvement of LSD2 in the BAT differentiation.

In this study, we found that LSD2 deficiency exerted different influences on gene expression depending on the tissue sources as well as Rosi addition (Supplemental Fig. S5). LSD2-KO BAT SVF cells showed reduced expression of BAT-associated genes in the presence of Rosi, but not in its absence. This suggests that LSD2 works downstream of PPARγ activation, possibly in cooperation with PRDM16, to induce the BAT program. In contrast, LSD2-KO iWAT-derived SVF cells expressed BAT genes at levels comparable to WT cells under beige adipocyte differentiation induction by Rosi treatment but showed reduced expression of BAT-selective genes when differentiated without Rosi. This suggests that LSD2 promotes the PPARγ-mediated BAT program in iWAT preadipocytes and that its function is dispensable under potent activation of PPARγ by Rosi. Interestingly, elevated expression of myogenic genes in SVF cells from LSD2-KO iWAT was not affected by Rosi treatment, suggesting that LSD2 suppresses the myogenic program irrespective of PPARγ activation.

Whereas Myf5-positive dermomyotome cells give rise to brown adipocytes (9), heterogeneous cell populations have been reported as sources of beige cells (33). Genetic tracing experiments revealed that at least a subset of beige cells is of smooth muscle-like origin (34). Here, we observed increased expression of skeletal muscle genes in LSD2-KO iWAT SVF cells. Interestingly, LSD2-KO iWAT had reduced expression of smooth muscle marker genes (Supplemental Fig. S4). Thus, these results raise the possibility that LSD2 regulates a unique type of beige progenitor that is of non–smooth muscle origin. Given that expression levels of myogenic genes were much lower in BAT SVF than in iWAT (unpublished data), it is plausible that the repression of myogenic genes in iWAT-progenitors is relatively loose. In other words, myogenic genes are strongly suppressed in BAT SVF cells. A previous report demonstrated that myogenic genes in BAT were repressed through euchromatic histone lysine methyltransferase 1 (EHMT1)–mediated H3K9 methylation, a modification associated with tight repression (35). Thus, differences in repression mechanisms might account for the inconsistency between BAT and beige cells in the expression of myogenic genes under LSD2 depletion.

Another flavin-dependent histone demethylase, LSD1, has been implicated in the epigenetic regulation of brown adipogenesis. LSD1 forms a complex with PRDM16 and C-terminal binding proteins 1 and 2 to repress WAT-specific gene expression in BAT (17). In addition, LSD1 promotes both white and brown adipocyte differentiation by suppressing Wnt signaling in mesenchymal stem cells (36). In this study, we found that LSD2 positively regulates BAT-associated genes while repressing myogenic genes depending on the derivation of cells. LSD1 and LSD2 greatly differ in their protein structure, especially in the N-terminal domains that are involved in protein-protein interactions and enzymatic activities (37), suggesting that LSD2 may serve a unique role in the regulation of brown and beige adipocytes.

Previous reports demonstrated that LSD2 is involved in both activation and repression of transcription. Through H3K4 demethylation, LSD2 represses lipid metabolism genes in hepatic cells (22) while contributing to faithful transcription elongation in some cell types (38). Thus, the recruitment of LSD2 to target genomic regions may be mechanistically different from that of LSD1. In addition, LSD2 has been shown to play a role in H3K9 demethylation (21) and ubiquitin ligase activity (39). Therefore, it is possible that multiple biochemical features and functions of LSD2 are involved in the regulation of BAT-associated and myogenic genes.

In this study, we discovered a unique role of LSD2, a histone demethylase, in controlling brown and beige adipocyte differentiation. Our findings provide novel insights into the developmental origin and plasticity of thermogenic adipocytes and promote the understanding of how adipocytes contribute to both health and disease.

Supplementary Material

This article includes supplemental data. Please visit http://www.fasebj.org to obtain this information.

Glossary

36B4 (Rplp0)

ribosomal protein large P0

BAT

brown adipose tissue

ChIP

chromatin immunoprecipitation

Dox

doxycycline

Dox-diff

doxycycline-differentiation

ECAR

extracellular acidification rate

FBS

fetal bovine serum

GSEA

gene set enrichment analysis

H3K4

histone [³H] lysine 4

H3K9

histone [³H] lysine 9

H3K4me1

mono-methylated H3K4

H3K4me2

di-methylated H3K4

H3K4me3

tri-methylated H3K4

iWAT

inguinal WAT

KD

knockdown

KO

knockout

LSD2

lysine-specific demethylase 2

MYF5

myogenic factor 5

OCR

oxygen consumption rate

PPAR

peroxisome proliferator-activated receptor

PRDM16

PR domain containing 16

qPCR

quantitative PCR

RNA-seq

RNA sequencing

Rosi

rosiglitazone

shRNA

short hairpin RNA

SVF

stromal vascular fraction

T3

3-3′-5-triiodo-L-thyronine

tet

tetracycline

TF

transcription factor

UCP1

uncoupling protein 1

WAT

white adipose tissue

WT

wild type

Rplp0 (36B4)

ribosomal protein large P0

AUTHOR CONTRIBUTIONS

R. Takase, S. Hino, and M. Nakao designed research; R. Takase, S. Hino, K. Nagaoka, K. Anan, K. Kohrogi, H. Araki, Y. Hino, and A. Sakamoto performed research; T. B. Nicholson and T. Chen contributed knockout mice; R. Takase and S. Hino analyzed data; and R. Takase, S. Hino, and M. Nakao wrote the manuscript.