DOT1L Regulates Thermogenic Adipocyte Differentiation and Function via Modulating H3K79 Methylation

Obesity occurs when energy intake exceeds energy expenditure and results in a significant risk of insulin resistance, type 2 diabetes, cardiovascular disease, and even cancer. Along with the growing obesity epidemic, adipocytes have been recognized as crucial regulators of energy balance and glucose homeostasis (1). Among three different types of adipocytes, brown and beige adipocytes are characterized as thermogenic adipocytes that express mitochondrial uncoupling protein 1 (UCP1) and have attracted great attention because of their capacity to use glucose and free fatty acids as fuels to generate heat (2). Brown adipocytes mainly exist in the interscapular brown adipose tissue (BAT), whereas beige adipocytes emerge within some white adipose tissue (WAT) depots under certain stimuli, such as cold and adrenergic signaling (3,4). In particular, accumulating evidence indicates that the recruitment and activation of brown and beige adipocytes could potently counteract obesity and obesity-related metabolic disorders by enhancing energy expenditure and improving glucose/lipid homeostasis in both mice and humans (5–7). Thus, a better understanding of the molecular mechanisms that control brown and beige adipocyte development and function will provide useful therapeutic ideas for obesity and obesity-related metabolic diseases.

An increasing number of transcriptional and epigenetic regulators have been identified as being involved in brown and beige adipocyte development and thermogenic function (8,9). Among transcriptional regulators, PRDM16 is considered as the dominant regulator of brown and beige cell fate and function (5,10,11). Notably, a number of histone methyl–modifying enzymes as epigenetic regulators are implicated in the development and function of brown and beige adipocytes, including site-specific histone methyltransferases or demethylases for lysine 4 of histone H3 (H3K4), lysine 9 of histone H3 (H3K9), lysine 27 of histone H3 (H3K27), and lysine 36 of histone H3 (H3K36). For example, the H3K9 methyltransferase EHMT1 positively regulates brown adipocyte fate and thermogenic function through H3K9 methylation and PRDM16 (12). Demethylase JMJD3-mediated H3K27me3 dynamics control brown fat development and beige fat formation (13). NSD2-mediated H3K36 methylation has a role in brown fat development (14). However, the role of H3K79 methylation and its modifying enzyme in brown and beige adipocyte development and function remains unclear.

H3K79 is located on the nucleosome surface, and the H3K79 methylation (H3K79me1/2/3) in mammals is catalyzed by the conserved histone methyltransferase disruptor of telomeric silencing-1 like (DOT1L). Recently, DOT1L and H3K79 methylation have been reported to play substantial roles in both normal physiological processes and disease progression, such as embryonic development, cerebral cortex development, and leukemia and breast cancer progression (15–18). Nevertheless, the role of DOT1L and H3K79 methylation in obesity and obesity-related metabolic diseases remains unexplored.

In this study, we found that DOT1L expression was abnormally increased in BAT and inguinal subcutaneous WAT (iWAT) from genetically obese mice, and thermogenic adipocytes-specific deletion of DOT1L potently protected against diet-induced obesity, improved obesity-related metabolic dysfunction such as insulin resistance and hepatic steatosis, and protected mice from hypothermia in vivo. Furthermore, deletion of DOT1L promotes brown and beige adipogenesis and thermogenesis of the stromal vascular fractions (SVFs) from BAT and WAT depots in a cell-autonomous way. Further mechanism studies revealed that H3K79 methylation, in particular H3K79me2 modification, correlates with the BAT-selective gene expression program. Collectively, our data establish DOT1L as a new epigenetic regulator of energy homeostasis and metabolic health in mice and provide new insight into the epigenetic mechanisms underlying thermogenic adipocyte differentiation and function.

pGMLV-CAG-CRE lentiviruses (GM-0220CR02-M) and vector pGMLV-CAG-WPRE lentiviruses (GM-4371LV18–400) were from Genomeditech. The chemical EPZ5676 (S7062) was from Selleck Chemicals. Dot1L siRNA (sc-77175) was from Santa Cruz Biotechnology. Details for antibodies used in this article are available in the Supplementary Material.

Conditional knockout mice were established by the classic Cre/loxP recombination system. The Dot1L^flox/flox mouse strain was generated by traditional embryonic stem cell–targeting technology. Generally, exon 5 of the Dot1L gene was chosen as the target locus, which encodes most of the methyltransferase active site (19,20), and was flanked by two loxP sites (Shanghai Biomodel Organism Science & Technology Development Co., Ltd.). The first obtained mice were C57BL/6 and 129sv hybrid background and were backcrossed with homozygous C57BL/6 mice for five generations before subsequent studies. Primer sequences (5′-CTGGGCCTCCTTTCACCCTT-3′ and 5′-GCAGGCAGTCCAGCAACACT-3′) were used to amplify and identify the targeting locus from genomic DNA. The Ucp1-Cre mouse strain (stock number 024670) was from The Jackson Laboratory and crossed with the Dot1L^flox/flox mouse strain to generate Dot1L^ucp1 knockout mice.

For obesity studies, 5-week-old male knockout mice and control littermates were fed with high-fat diet (HFD) (60% fat; D12492; Research Diets, Inc.) for 5 months. Body weight and food intake were recorded weekly. Body composition was analyzed with nuclear magnetic resonance spectroscopy (Minispec LF90 II; Bruker). The glucose tolerance test (2 g/kg glucose, intraperitoneally) and insulin tolerance test (0.75 unit/kg insulin, intraperitoneally) were performed after 6-h fasting, respectively. For cold tolerance test, mice were single-caged and placed in a 4°C refrigerator. Rectal temperatures were detected by a BAT-12 Microprobe digital thermometer (Physitemp Instruments) at the indicated time points, and the skin surface temperatures were measured by an infrared camera (FLIR Systems) before the cold challenge began and after 6 h of exposure. Whole-body energy expenditure and O₂ consumption were measured with a TSE PhenoMaster caging system (TSE Systems) and recorded for 24 h after a 24-h adaption and calculated as previously described (21). All animal experiments were performed according to procedures approved by the Animal Ethics Committee of Shanghai Institute of Materia Medica.

Primary SVF cells were isolated from the indicated adipose depots according to methods described previously (22). SVF cells were cultured with DMEM/F12 supplemented with 10% FBS. For brown and beige adipogenesis, confluent SVF cells were first cultured for 2 days in DMEM/F12 supplemented with 10% FBS, 10 nmol/L T3, 0.5 mmol/L isobutylmethylxanthine, 125 nmol/L indomethacin, 1 μmol/L dexamethasone, 1 μmol/L rosiglitazone, and 850 nmol/L insulin and then cultured for an additional 6 days in DMEM/F12 supplemented with 10% FBS, 10 nmol/L T3, 1 μmol/L rosiglitazone, and 850 nmol/L insulin only. Lipid formation was determined by Oil Red O (Sangon Biotech) staining and visualized by microscopy. Cellular oxygen consumption was measured by an XFe analyzer (Seahorse Bioscience).

Protein expression analysis and mRNA expression analysis were performed according to normal procedures. Primers including forward sequence 5′-TGGAGAACTATGTCCTGATCGAC-3′ and reverse sequence 5′-GTGCCGCAGAAGTCCATTG-3′ were used to quantify the Dot1L mRNA level, which are specific for deleted exon 5 of the Dot1L gene. Primers for profibrosis genes were described in a previous report (23). Other primers for quantitative RT-PCR are listed in Supplementary Table 1.

Hematoxylin and eosin (H&E) staining and immunohistochemistry (IHC) assays for BAT and WATs were performed by Shanghai Zuocheng Biological Technology Co., Ltd. Adipocyte size was determined with AdipoCount software (24), and IHC quantification was determined with Image-Pro Plus software. H&E staining for liver samples was performed by Wuhan Servicebio Co., Ltd. Plasma triglyceride and total cholesterol were measured with chemical reagents from Shanghai Fosun Long March Medical Science Co. Ltd. Nonesterified fatty acid (NEFA) was measured with reagents from Wako Chemicals. LDL-cholesterol (LDL-C) was measured with reagents from Sichuan XinJianKangCheng Biological Co., Ltd. Insulin and leptin were measured with ELISA kits from Millipore. Sirius Red staining for WATs was performed by Shanghai Zuocheng Biological Technology Co., Ltd.

Differentiating BAT-SVF cells at day 6 were collected for RNA sequencing (RNA-seq) and chromatin immunoprecipitation (ChIP)–quantitative PCR (qPCR) validation. RNA-seq analysis was performed by BGI Genomics. ChIP assays were performed using an ab500 ChIP kit (Abcam). All enrichment changes were normalized to the input. Genome-wide mapping of H3K79me2 (antibody catalog number 39143) in BAT samples via ChIP followed by sequencing and data processing were performed by Active Motif China according to proprietary methods. Primers for ChIP-qPCR validation are listed in Supplementary Table 1.

Results are shown as mean ± SEM. Group differences were analyzed with the two-tailed Student t test, except ChIP-qPCR data, in which a one-tailed distribution was used. P values <0.05 were considered as statistically significant.

We declare that all of the data are available upon request.

To investigate the role of DOT1L in brown and beige adipocyte development and function, we first examined the expression levels of Dot1L in adipose tissues and found that Dot1L was expressed in both interscapular BAT and iWAT from adult male C57BL/6J mice (Supplementary Fig. 1A). Furthermore, the expression of Dot1L was abnormally increased in both BAT and iWAT from obese ob/ob transgenic mice (Fig. 1A). iWAT is an adipose depot enriched with beige adipocytes. Based on these results, we speculated DOT1L was implicated in the regulation of brown and beige adipocyte function and might play a role in the obesity phenotype.

To test this hypothesis, we generated a conditional Dot1L allele (Dot1L^flox/flox) by introducing loxP sites flanking exon 5, which encodes most of the methyltransferase active site (19,20), and crossed this model with Ucp1-Cre mice (25) to obtain brown and beige adipocyte–specific Dot1L knockout mice (Dot1L^ucp1 knockout mice) for in vivo obesity studies. Because the available antibodies show poor specificity for detection of mouse DOT1L, instead, levels of H3K79 methylation were assessed to reflect the knockout efficiency and significantly reduced in BAT of knockout mice (Supplementary Fig. 1B).

When fed a normal chow diet, there was no significant difference in body weight between the knockout mice and control littermates, while the fat mass of knockout mice was slightly lower (Supplementary Fig. 1C and D). However, the differences between Dot1L^ucp1 knockout mice and control mice became more dramatic when fed with HFD. The Dot1L mRNA levels were significantly decreased in BAT of Dot1L^ucp1 knockout mice and also mildly decreased in some WAT depots, such as iWAT, perirenal WAT (rWAT), and epididymal WAT (eWAT) (Fig. 1B). As room temperature (22°C) causes a mild cold challenge in mice (26), it is reasonable that some Ucp1-positive beige adipocytes exist in these WAT depots. Strikingly, we found that Dot1L^ucp1 knockout mice fed with HFD gained significantly less body weight than control mice without any change in food intake (Fig. 1C and D). Moreover, Dot1L^ucp1 knockout mice showed less adiposity, as determined by greatly reduced fat mass (Fig. 1E) and decreased tissue weights of BATs and WATs (Fig. 1F). Consistently, histological morphology further revealed that the BAT of Dot1L^ucp1 knockout mice contained apparently less lipid storage than those of control mice, and adipocytes of above-mentioned WAT depots were significantly smaller in Dot1L^ucp1 knockout mice than in control mice (Fig. 1G and H). In addition, both plasma triglyceride and total cholesterol levels in Dot1L^ucp1 knockout mice were markedly lower than those in control mice (Fig. 1I and J).

Together, these results demonstrate that DOT1L in brown and beige adipocytes negatively affect obesity phenotypes, and deletion of DOT1L in these thermogenic adipocytes potently protects mice from HFD-induced obesity.

Food intake and energy expenditure balance the body weight. The weight loss of Dot1L^ucp1 knockout mice fed with HFD was obviously not due to the unchanged food intake (Fig. 1D) or physical activity (Fig. 2A). Subsequently, we found that the whole-body energy expenditure and oxygen consumption of Dot1L^ucp1 knockout mice fed with HFD in both day and night periods were significantly increased (Fig. 2B and C), which was in agreement with the weight loss.

Since the HFD-induced obesity model exhibits impaired glucose homeostasis and insulin resistance, we further observed that the Dot1L^ucp1 knockout mice fed with HFD exhibited significantly improved glucose tolerance (Fig. 2D and E) and increased insulin sensitivity (Fig. 2F and G). In accordance with the improved insulin resistance, Dot1L^ucp1 knockout mice also showed a marked decrease in the fasting plasma insulin level (Fig. 2H). The leptin level, which reflects the body weight and lipid content in mice and humans (27,28), was also significantly reduced in knockout mice (Supplementary Fig. 2A), suggesting diminished leptin resistance in Dot1L^ucp1 knockout mice. The adiponectin level, which is also associated with insulin resistance, was unchanged (Supplementary Fig. 2B). Interestingly, the Dot1L^ucp1 knockout mice fed with normal chow diet also exhibited significantly increased whole-body energy expenditure and oxygen consumption (Supplementary Fig. 2C and D), as well as improved glucose tolerance and insulin sensitivity (Supplementary Fig. 2E–H), suggesting that the metabolic effects of DOT1L are HFD independent.

Hepatic steatosis is another metabolic phenotype associated with obesity. Notably, livers of Dot1L^ucp1 knockout mice fed with HFD contained less lipid storage, as determined by markedly lower amounts of lipid droplets and triglyceride content (Fig. 2I and J), and showed improved insulin signaling as determined by phosphorylation of AKT even under a basal condition (Fig. 2K). Besides, the expression levels of genes involved in fatty acid oxidation, including Pparα, Cpt1a, and Acox1, and lipolysis genes, including Lipe and Pnpla2, were significantly increased, whereas expression levels of lipogenic genes, such as Me1 and Acc1, were decreased in the livers of Dot1L^ucp1 knockout mice (Fig. 2L), which indicates elevated lipid utilization and reduced lipogenesis in the livers of Dot1L^ucp1 knockout mice.

Together, these results indicate that deletion of DOT1L in thermogenic adipocytes protects mice from obesity by increasing energy expenditure and promisingly improves obesity-related metabolic dysfunction, such as insulin resistance and hepatic steatosis.

To understand the molecular basis of the anti-obesity effects observed in Dot1L^ucp1 knockout mice fed with HFD, we next examined the molecular characteristics of BAT and WATs. We found that the BAT of Dot1L^ucp1 knockout mice had significantly more UCP1 and PRDM16 IHC staining (Fig. 3A and B). Western blot analysis further confirmed the greatly elevated UCP1 and PRDM16 expression in BAT of Dot1L^ucp1 knockout mice (Fig. 3C). Furthermore, expression levels of other BAT-selective genes, including Pgc-1α, Elovl3, Dio2, Pparα, Cox8b, Bmp8b, and Zic1, were all significantly increased in BAT of Dot1L^ucp1 knockout mice (Fig. 3D). These findings suggest that Dot1L^ucp1 knockout mice have an enhanced BAT identity and function.

Furthermore, the levels of BAT-selective genes, such as Ucp1, Cidea, Cox8b, and Pparα, and beige adipocyte-specific genes, such as Sp100 and Tbx1, were significantly increased in iWAT of Dot1L^ucp1 knockout mice (Fig. 3E). The inflammatory markers Mcp-1 and F4/80 were also markedly decreased (Fig. 3E), suggesting that HFD-induced inflammation was ameliorated in iWAT of Dot1L^ucp1 knockout mice. In addition, PRDM16 expression was augmented in iWAT of Dot1L^ucp1 knockout mice (Fig. 3F and G), which further supports the enhanced thermogenic program in iWAT. Similar results were observed in rWAT, another adipose depot that has the potential to brown (29,30) (Supplementary Fig. 3A–C), as well as in eWAT (Fig. 3H–J), an adipose depot that poorly transitions to brown (3). Besides, as the browning of WAT also results in the repression of adipose tissue fibrosis (23), the Sirius Red staining showed that there was a decreasing trend on the fibrosis of both iWAT and eWAT from knockout mice, and the expression of profibrosis genes was significantly downregulated in both iWAT and eWAT from knockout mice, indicating repressed WAT fibrosis in Dot1L^ucp1 knockout mice (Supplementary Fig. 3D–I). These results obtained from different WATs suggested an overall browning of WATs in Dot1L^ucp1 knockout mice.

Together, these observations suggest that loss of DOT1L in thermogenic adipocytes leads to enhanced BAT identity and function, as well as activation of the thermogenic program in WATs, thereby resulting in increased energy expenditure.

Given the significantly enhanced thermogenic function in BAT and WATs due to loss of DOT1L under HFD, it is not hard to imagine that the knockout mice might resist hypothermia. Accordingly, we placed them in a 4°C cold environment to examine the role of DOT1L in adaptive thermogenesis in vivo. As expected, Dot1L^ucp1 knockout mice fed with HFD exhibited higher rectal temperatures during a 6-h cold exposure (Fig. 4A). Moreover, skin surface temperatures at the interscapular BAT area of knockout mice were significantly higher than those of control littermates under both basal conditions and after 6 h of cold exposure (Fig. 4B and C), suggesting more activated thermogenesis in BAT from knockout mice. We further extended the challenge time to 48 h and analyzed changes in the histological morphology and molecular characteristics of BAT. Consistently, BAT of Dot1L^ucp1 knockout mice contained more multilocular adipocytes and much more UCP1 and PRDM16 staining compared with control mice (Fig. 4D and E). The increased UCP1 and PRDM16 expression was also confirmed by Western blot analysis (Fig. 4F and G). Notably, the mRNA levels of abundant genes were significantly elevated in BAT of Dot1L^ucp1 knockout mice, including thermogenic genes, adipogenic genes, lipolysis and fatty acid oxidation genes, as well as BAT-specific identity genes as indicated (Fig. 4H), suggesting greatly enhanced recruitment and activation of BAT due to loss of DOT1L.

In addition, we observed similar promoted browning phenomena in iWAT, rWAT, and eWAT of Dot1L^ucp1 knockout mice, as determined by increased expression of genes involved in thermogenesis, and increased expression of BAT-selective proteins, such as PRDM16, PPARα, and DIO2 (Supplementary Fig. 4). Furthermore, since BAT prefers to use lipids as fuel for heat generation (31), we also found that the plasma triglyceride, total cholesterol, LDL-C, and NEFA levels were all significantly reduced in Dot1L^ucp1 knockout mice after a cold challenge (Fig. 4I).

Besides, we also found that expression levels of thermogenic genes and fatty acid oxidation-related genes were significantly increased in BAT from knockout mice on a normal chow diet (Supplementary Fig. 5A), which suggests that the enhanced thermogenic function of BAT also exists in knockout mice on a normal chow diet. Moreover, when exposing to cold, the Dot1L^ucp1 knockout mice fed with normal chow diet also exhibited resistance to hypothermia, as determined by higher body temperature and higher interscapular skin surface temperature compared with control mice (Supplementary Fig. 5B and C).

Together, these results demonstrate that loss of DOT1L in thermogenic adipocytes greatly facilitates adaptive thermogenesis under cold challenge and protects mice from hypothermia.

Next, to obtain an initial insight into the mechanisms by which DOT1L regulates thermogenic adipocyte identity and function, we used ChIP sequencing (ChIP-seq) to map the genome-wide H3K79me2 locations in BAT samples from Dot1L^ucp1 knockout mice and control mice and then compared the differences between samples to identify the target genes. Consistent with previously described results (32), the H3K79me2 was widely distributed within various genomic regions and peaked just downstream of the transcription start site (Fig. 5A and B). Remarkably, ChIP-seq data showed that H3K79me2 enrichment was significantly decreased in the merged regions of a number of BAT-selective genes in BAT from Dot1L^ucp1 knockout mice, such as key BAT transcriptional regulators Pparα, Prdm16, and Pgc-1α, while enrichment was not altered in common fat genes, such as Fabp4 and Pparg (Fig. 5C and D).

Since the expression levels of these BAT-selective genes were all notably elevated in BAT of conditional Dot1L^ucp1 knockout mice (data in Figs. 3 and 4), we speculated that the H3K79me2 mark was negatively associated with the expression of these BAT-selective genes. Of note, it does not rule out the negative regulator of brown fat development and function among these BAT-selective genes. For instance, H3K79me2 enrichment was significantly decreased in the merged region of the Kcnk3 gene, which is highly expressed in thermogenic fat and acts as a negative regulator of thermogenesis (33), while, in line with the above speculation, the expression of Kcnk3 was also significantly increased in BAT from knockout mice (Supplementary Fig. 9). Furthermore, the Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analysis of differentially expressed genes suggested that H3K79me2 targets were highly associated with metabolic pathways, peroxisome proliferator–activated receptor (PPAR) and insulin signaling, thermogenesis, lipid metabolism, as well as fatty acid metabolism and adipocytokine signaling (Fig. 5E).

As mentioned regarding adipocytokine signaling, we surprisingly found that the expression of BAT-enriched secreted factor Nrg4 was also highly correlated with H3K79 methylation, as determined by significantly decreased H3K79me2 peaks at the locus of the Nrg4 gene (Fig. 5C and Supplementary Fig. 6A), and mRNA expression of Nrg4 was increased in BAT of Dot1L^ucp1 knockout mice (Supplementary Fig. 6B). NRG4 has been identified as an endocrine link between BAT and the liver and protects against diet-induced insulin resistance and hepatic steatosis by attenuating hepatic lipogenesis and activating fatty acid oxidation (34–36). Notably, we found that the plasma NRG4 levels were markedly elevated in Dot1L^ucp1 knockout mice (Supplementary Fig. 6C). In combination with the previously observed ameliorated hepatic steatosis in Dot1L^ucp1 knockout mice (Fig. 2I–L), we speculate that DOT1L affects hepatic lipid homeostasis likely through the actions of NRG4.

These findings indicate that DOT1L-mediated H3K79 methylation controls the BAT-selective gene and metabolic regulatory gene program, which is likely to be responsible for the striking phenotypes due to loss of DOT1L in thermogenic adipocytes in mice.

To further validate that DOT1L regulates brown and beige adipocyte development and function in a cell-autonomous way, we separately isolated primary SVF cells from the interscapular BATs or WATs of conditional Dot1L^flox/flox mice and induced them into brown or beige adipocytes in vitro. We used lentiviruses, which express Cre, to stably delete DOT1L. The efficient deletion of Dot1L and H3K79 methylations in Cre-treated BAT-SVF cells was confirmed by mRNA expression analysis and H3K79 methylation expression analysis (Fig. 6A and B). Strikingly, we found that deletion of DOT1L significantly promoted brown adipocyte differentiation, as determined by Oil Red O staining (Fig. 6C and D), and increased both the basal and uncoupled oxygen consumption rates, which suggest enhanced thermogenesis (Fig. 6E). Moreover, DOT1L depletion markedly elevated levels of the BAT-enriched protein UCP1 and those important transcriptional regulators involved in BAT development and function, including C/EBPβ, PRDM16, and PGC-1α (Fig. 6F). Notably, mRNA levels of BAT-selective genes, including Ucp1, Prdm16, Pgc-1α, Pgc-1β, Dio2, Cidea, Cox5b, and Cox7a1 (Fig. 6G), genes involved in fatty acid oxidation, such as Pparα, Acox1, Mcad, and Lcad (Fig. 6H), and BAT-specific identity genes, such as Eva1, Fbxo31, and Oplah (Fig. 6I), were broadly elevated in differentiated DOT1L-deficient BAT-SVF cells. Similar results were observed in both SVF cells isolated from iWAT (Fig. 6J–O) and SVF cells isolated from eWAT (Fig. 6P–U), which suggests that deletion of DOT1L also significantly promoted beige adipogenesis and thermogenesis. In addition, examination of a time course of differentiating BAT cells showed that DOT1L depletion significantly promoted the expression of Ucp1 gene starting from day 2 of differentiation and increased protein expression levels of key adipogenic and thermogenic markers at days 4 and 6 of differentiation (Supplementary Fig. 10A–C).

The seeming contradiction between downregulated thermogenic gene expression suggested by Yi et al. (37) and our observations on DOT1L-deficient BAT cells can be explained by the distinct time point of DOT1L loss and effects on differentiation. To further demonstrate it, we established two methods to knockdown Dot1L by transfecting with siRNA targeting Dot1L at day −3 before initiation of differentiation or at day 2 of differentiation, and then we examined lipid accumulation and gene expression in differentiated brown adipocytes. Consistent with results in the article by Yi et al. (37), knockdown of DOT1L did not affect brown adipogenesis when siRNA was transfected at day 2 of differentiation, whereas knockdown of DOT1L significantly increased brown adipocyte differentiation when siRNA was transfected before initiation of differentiation (Supplementary Fig. 10D and E). These results indicate that loss of DOT1L enhances terminal brown adipocyte differentiation and functions indispensably in the early phase of differentiation. Besides, we used a selective and potent DOT1L inhibitor, EPZ5676 (38), and tested its role on mesenchymal C3H10T1/2 cells to validate the importance of histone methyltransferase enzymatic activity for DOT1L-mediated gene expression and function. During the entire time course of differentiation, 10 μmol/L EPZ5676 was added. As expected, the inhibitor EPZ5676 strongly inhibited H3K79 methylations (Supplementary Fig. 7A) and enhanced brown adipocyte differentiation and thermogenesis in C3H10T1/2 cells (Supplementary Fig. 7B–D), which was consistent with previous report (39). Moreover, EPZ5676 promoted the transcriptional activity of Ucp1, as determined by a luciferase assay using a reporter construct containing the Ucp1 promoter (Supplementary Fig. 7E). However, it must be mentioned that in the article by Yi et al. (37), EPZ5676 was treated on immortalized BAT cells and 3T3L1 cells at day 4 of differentiation. Hence, we suppose the difference between our work and the work by Yi et al. (37) in cells can be explained by the effect of DOT1L in brown/beige adipocyte differentiation, which should not to be ignored.

Taken together, our observations demonstrate that DOT1L and H3K79 methylation negatively regulates brown and beige adipocyte differentiation and function in a cell-autonomous way.

To further explore the molecular mechanisms of DOT1L-mediated gene expression on thermogenic adipocytes, we performed RNA-seq analysis on differentiating BAT-SVF cells. Gene expression profiles show that loss of DOT1L resulted in a number of differentially expressed genes (812 upregulated genes and 1,056 downregulated genes) (Fig. 7A). Gene ontology enrichment analysis revealed that upregulated genes were significantly enriched in those biological process associated with brown adipocyte differentiation and function, including mitochondrial respiratory process, brown fat cell differentiation, lipid and fatty acid metabolic process, and cold-induced thermogenesis (Fig. 7B). For instance, genes involved in brown adipocyte differentiation and thermogenic function, such as Cebpβ, Ucp1, Ucp2, Cidea, Cox7a1, and Bmp8b, were induced in BAT-SVF cells transfected with Cre viruses (Fig. 7C). Furthermore, KEGG pathway analysis of upregulated genes revealed significant enrichment of metabolic pathways, thermogenesis, glucose and fatty acid metabolism, and PPAR signaling (Fig. 7D), which was highly overlapped with the enrichment pathways revealed by ChIP-seq on BAT samples from Dot1L^ucp1 knockout mice (Fig. 5E).

In combination with these transcriptional changes and H3K79 methylation enrichment state, we performed ChIP-qPCR assays using differentiating BAT-SVF cells to validate the correlation, since BAT samples unavoidably contain other types of cells in addition to brown adipocytes. Results from normal control IgG assays showed no difference between groups (Supplementary Fig. 8). Surprisingly, the ChIP-qPCR data revealed that the binding levels of H3K79me1 were not only significantly decreased on genomic regions of both BAT-selective genes, such as Pparα, Prdm16, and Pgc-1α, but also decreased on genomic regions of common fat genes, such as Pparg and Fabp4, in differentiating DOT1L-deficient BAT-SVF cells (Fig. 7E and F), suggesting global changes of H3K79me1 modification on the genome. However, the binding levels of H3K79me2 were selectively decreased on genomic regions of indicated BAT-selective genes in differentiating DOT1L-deficient BAT-SVF cells, while remaining unchanged on genomic regions of Pparg and Fabp4 (Fig. 7G and H), which was consistent with the genome-wide H3K79me2 distribution revealed by ChIP-seq on BAT samples from knockout mice (Fig. 5C and D). The binding levels of H3K79me3 on these genes were generally lower than levels of H3K79me2, but enrichment patterns were similar with the pattern of H3K79me2 (Fig. 7I and J).

These results indicate that DOT1L-mediated H3K79 methylation, in particular H3K79me2 modification, plays a role in the regulation of the BAT-selective transcriptional program.

Thermogenic adipocytes have been extensively studied because of their great potential to treat obesity and associated metabolic diseases. An increasing number of studies have revealed that epigenetic mechanisms, such as DNA methylation and histone modifications, are involved in thermogenic adipocyte function and energy homeostasis, yet one of the core histone modifications, H3K79 methylation, has not been reported. In this study, we identified the H3K79 methyltransferase DOT1L as a novel epigenetic regulator that controls brown and beige adipocyte differentiation and function. Firstly, the increased expression of Dot1L in BAT and iWAT from the obese mouse model indicated the linkage between Dot1L expression and obesity phenotype. As germline DOT1L knockout causes embryonic lethality (15), we established conditional knockout mice for brown and beige adipocytes and revealed that thermogenic adipocyte-specific deletion of DOT1L was associated with reduced obesity and improved obesity-related metabolic dysfunction. These striking findings in genetically obese mice and conditional knockout mice strongly suggest DOT1L has promising therapeutic potential to counteract obesity and obesity-related metabolic disorders.

Brown and beige adipocytes share a number of BAT-selective genes, such as Ucp1, Pparα, Pgc-1α, Cidea, and Prdm16 (40,41). Notably, we observed a similar greatly upregulated BAT-selective gene program due to loss of DOT1L both in mice and SVF cells. Although there is a high correlation between H3K79 methylation and active transcription, there are some exceptions, such as ENaCα, which encodes a subunit of the epithelial Na⁺ channel, Aquaporin 5, and some genes involved in cortical development (18,42,43). In our study, we identified some BAT-selective genes as possible downstream targets of DOT1L that are repressed by H3K79 methylation, such as Pparα, Prdm16, and Pgc-1α. PRDM16 is well known for its dominant role in brown and beige fat development and function (11,44). PPARα is well known for its role in fatty acid oxidation and also can directly activate BAT-selective genes (45). PGC-1α is also well known for its role in energy metabolism via controlling thermogenic gene activation and adaptive metabolic gene programs, but not essential for brown adipocyte differentiation (46,47). We inferred these master transcriptional regulators of BAT function and energy homeostasis were direct targets of DOT1L; however, it is difficult to determine which regulator is more important for the regulation network of DOT1L. Besides, it cannot be ruled out that other differential metabolic regulators revealed by ChIP-seq and RNA-seq also have contributions to the whole-body phenotypes. Future validation studies should be focused on whether the role of DOT1L depends on these indicated downstream targets.

Although DOT1L catalyzes mono-, di-, and trimethylation of H3K79, we surprisingly found that H3K79 methylation, in particular H3K79me2, was selectively correlated with BAT-selective gene expression. According to other reports, it is suggested that H3K79me2 has the potential to initiate, prime, and/or maintain specific developmentally transcriptional programs during cell fate decision and tissue development, such as cardiomyocyte differentiation and cerebral cortex development (18,48). Our data suggest that DOT1L and H3K79me2 may prime the progenitors for brown and beige cell fate decision and later developmental process. But our ChIP-seq and ChIP-qPCR data are not sufficient to clarify the dynamic changes of H3K79me2 during the adipogenic and thermogenic processes and cannot explain the distinct effect patterns of mono-, di-, and trimethylation of H3K79 on transcriptional regulation of brown and beige adipose development and function. Further studies should elucidate these details.

H3K79 methylation is widespread at transcribed gene, and also exists at expressed miRNA genes (49). Interestingly, in our ChIP-seq data, we found that H3K79me2 peaks were significantly reduced in the locus of some BAT-enriched miRNAs, such as miR193b, miR365, miR203, and miR378 (data not shown). Among these, the miR193b-365 cluster is reported as essential for brown adipocyte differentiation (50). These findings suggest the involvement of miRNAs in the regulatory network of DOT1L in thermogenic adipocyte function.

Previous studies have revealed that brown-selective genes are premarked by both H3K27me3 and H3K4me1 in preadipocytes, and the removal of H3K27me3 is required but not sufficient for BAT gene expression and browning of WAT, while the predeposition of H3K4me1 is essential for poising brown-selective genes for expression in mature brown cells (13,51). Our data further suggest that inhibition of DOT1L also plays an important role in promoting BAT-selective gene expression. However, our understanding regarding the downstream effects of H3K79 methylation is still limited. It will be intriguing to explore whether cross talk exists between the H3K79 methylation and H3K4/K27 methylation involved in thermogenic adipocyte development and function.

Concerning the opposite phenotypes in the mice used by Yi et al. (37), the differences may be caused by distinct DOT1L floxed mouse models, the knockout efficiency between the two knockout mouse models, the different ages of mice, and different HFDs used for HFD experiments. Although the full range of causes underlying the opposite phenotypes remain elusive, these differences we have found in animal studies may provide a reason for the paradox. The differences between our work and the work by Yi et al. (37) in cellular studies can be explained by the distinct effects of DOT1L in brown/beige adipocyte differentiation due to the time point of DOT1L loss, which we have addressed in our results. It should be mentioned that Yi et al. (37) suggest DOT1L interacts with ZC3H10 to activate Ucp1 and other thermogenic genes; however, in our studies, there were no significant changes on the expression of Zc3h10 in both differentiated BAT-SVF cells and BAT of DOT1L knockout mice (Supplementary Fig. 11), indicating ZC3H10 was not involved in the DOT1L’s function in thermogenic adipocyte differentiation and function, at least in our data.

We can only conclude that H3K79 methylation correlates with the transcriptional changes observed in thermogenic adipocytes and cannot rule out other DOT1L functions than H3K79 methylation are responsible for the transcriptional regulation. Of note, DOT1L and many nuclear protein complexes have direct interactions in the regulation of gene expression (52), and it is very likely that there are other specific binders or effectors involved in the regulation network of DOT1L in brown and beige adipose tissues. For instance, among various interactors of DOT1L, there is histone deacetylase SIRT1. It has been reported that DOT1L interacts with SIRT1 and negatively affects the activity of SIRT1 in mixed-lineage leukemia–rearranged leukemia (53). In addition, high SIRT1 activity can deacetylate PPARγ to recruit PRDM16 binding to induce BAT-selective gene expression (54). It is possible that SIRT1 may serve as a coregulator of DOT1L on affecting BAT-selective gene expression. This possibility means that loss of DOT1L leads to desuppression of the positive regulator of BAT. The dynamic interplay between DOT1L and other chromatin regulators controlling thermogenic adipocyte development and function needs to be further explored.

Lastly, it is very worthwhile to reveal whether DOT1L is genetically linked to obesity and obesity-related metabolic diseases in human patients. In conclusion, our study provides a new perspective to understand the epigenetic mechanisms underlying thermogenic adipocyte development and function and raises the possibility that targeting DOT1L in brown and beige adipose tissues can lead to promising therapeutic interventions for obesity and associated metabolic diseases.

Acknowledgments. The authors thank Shanghai Biomodel Organism Science & Technology Development Co., Ltd. for the establishment of conditional knockout mice. The authors also thank Active Motif China for the ChIP-seq and data processing.

Funding. This work was supported by a China Postdoctoral Science Foundation grant (2018M642118), National Natural Science Foundation of China (81673493), National Program on Key Research Project (2016YFC1305505), National Science and Technology Major Project of China (2018ZX09711002-018), and the K.C. Wong Education Foundation.

Duality of Interest. No potential conflicts of interest were reported.

Author Contributions. L.S., B.-H.L., J.L., and J.-Y.L. contributed to the design of experiments and data discussion. L.S. and B.-H.L. contributed to the conduction of experiments and data analysis. H.-W.J., and L.Y. contributed to experimental assistance. L.S. wrote the paper. B.-H.L., J.L., and J.-Y.L. reviewed the manuscript. J.-Y.L. is the guarantor of this work and, as such, had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.