m6A methylation promotes white‐to‐beige fat transition by facilitating Hif1a translation

Abstract

Obesity mainly results from a chronic energy imbalance. Promoting browning of white adipocytes is a promising strategy to enhance energy expenditure and combat obesity. N6‐methyladenosine (m6A), the most abundant mRNA modification in eukaryotes, plays an important role in regulating adipogenesis. However, whether m6A regulates white adipocyte browning was unknown. Here, we report that adipose tissue‐specific deletion of Fto, an m6A demethylase, predisposes mice to prevent high‐fat diet (HFD)‐induced obesity by enhancing energy expenditure. Additionally, deletion of FTO in vitro promotes thermogenesis and white‐to‐beige adipocyte transition. Mechanistically, FTO deficiency increases the m6A level of Hif1a mRNA, which is recognized by m6A‐binding protein YTHDC2, facilitating mRNA translation and increasing HIF1A protein abundance. HIF1A activates the transcription of thermogenic genes, including Ppaggc1a, Prdm16, and Pparg, thereby promoting Ucp1 expression and the browning process. Collectively, these results unveil an epigenetic mechanism by which m6A‐facilitated HIF1A expression controls browning of white adipocytes and thermogenesis, providing a potential target to counteract obesity and metabolic disease.

FTO loss in adipose tissue protects against HFD‐induced obesity via promoting energy expenditure. FTO loss increases HIF1A m6A levels and protein abundance. HIF1A activates thermogenic gene transcription thereby promoting the browning process.

Introduction

Obesity and its associated metabolic diseases, including type 2 diabetes (T2D), nonalcoholic fatty liver disease (NAFLD), and cardiovascular diseases, have been rapidly prevailed over the past decades worldwide (Ghaben & Scherer, 2019; Sung et al, 2019), which has become a global public health problem. Obesity, defined as abnormal or excessive adipose tissue accumulation, is mainly caused by energy disequilibrium (Deshpande et al, 2019). Adipose tissue is an important regulator of systemic energy homeostasis by storing and releasing energy (Wang & Seale, 2016). White adipose tissue (WAT) stores chemical energy by packing triglycerides into large lipid droplets, while brown adipose tissue (BAT) dissipates energy to produces heat through non‐shivering thermogenesis in response to systemic demands (Rosen & Spiegelman, 2014). Notably, emerging studies have identified a specific subset of “brown‐like” beige adipocytes reside within subcutaneous WAT that also possess thermogenic properties (Wu et al, 2012). The thermogenic activity of beige fat cells mainly depends on the abundant cristae‐dense mitochondria that highly express uncoupling protein 1 (UCP1) (Ikeda et al, 2018). Accordingly, beige adipocytes play a critical role in regulating the systemic energy homeostasis in mammals. Recently, beige adipocytes have attracted much mainstream attention as a promising and potential target for obesity therapy, due to their capability of energy expenditure and relevance to adult humans (Ikeda et al, 2018). Importantly, the biogenesis of beige adipocytes in WAT is highly inducible by environmental stimuli or external cues, also called browning of white adipocytes, which has been suggested to have strong anti‐obesity and anti‐diabetic benefits (Wu et al, 2012; Harms & Seale, 2013). Thus, promoting our understanding of the molecular regulatory mechanism of white adipocytes browning could help us to specifically manipulate these cells to prevent obesity.

Emerging evidence has revealed various molecular regulators that mediate browning of white adipocytes and beige adipogenesis, including PR domain zinc‐finger protein 16 (PRDM16), peroxisome proliferator‐activated receptor gamma (PPARG) and PPARG co‐activator‐1A (PPARGC1A, also known as PGC1A) (Kajimura et al, 2015). PRDM16, a 140 kDa zinc‐finger nuclear protein, induces beige adipocyte biogenesis and activates thermogenic program in subcutaneous WAT (Seale et al, 2011). PRDM16 and PPARG are key transcriptional factors in browning process that initiate thermogenic gene expression (Inagaki et al, 2016). Chronic treatment with PPARG agonist is reported to strongly induce browning of white fat (Ohno et al, 2012). PPARGC1A is identified as a master regulator in regulating mitochondrial biogenesis and thermogenesis in brown and beige adipocytes (Puigserver et al, 1998; Tiraby et al, 2003). PPARGC1A expression is activated by cold exposure or β3‐adrenergic agonists, and it then stimulates the expression of Ucp1 and other genes by interacting with various transcriptional regulators (Inagaki et al, 2016). PPARGC1A binds to complexes of PPARA or PPARG and retinoid X receptor (RXR), which both activate Ucp1 expression (Bartelt & Heeren, 2014). Thus, the ectopic expression of PPARGC1A in white adipocytes facilitates genes involved in mitochondrial oxidation phosphorylation and thermogenesis (Puigserver et al, 1998). Studies have dissected various transcriptional regulators of white fat browning and thermogenesis; however, the post‐transcriptional regulators and epigenetic mechanisms remain uncovered. The study of epigenetics in thermogenic gene expression will help us to illuminate the molecular mechanisms underlying white‐to‐beige fat transition and adipose tissue plasticity, further control adipose cell fate and stimulate thermogenesis.

N ⁶‐methyladenosine (m⁶A) is the most prevalent and abundant mRNA modification in eukaryotic cells and has recently emerged as a significant epigenetic regulator (Wu et al, 2016; Roundtree et al, 2017; Zhao et al, 2017). m⁶A modification is catalyzed by methyltransferase complex composed of methyltransferase‐like 3 (METTL3) and 14 (METTL14) subunits and other cofactors, removed by demethylases fat mass and obesity‐associated protein (FTO) and alkB homologue 5 (ALKBH5), and recognized by various m⁶A‐binding proteins, including YT521‐B homology (YTH) protein families (YTHDF1/2/3 and YTHDC1/2), (IGF2BP) protein families (IGF2BP1/2/3), and heterogenous nuclear ribonucleoproteins (HNRNP) protein families (HNRNPC, HNRNPG, and HNRNPA2B1). m⁶A methylation can dynamically and reversibly regulate RNA metabolism, including mRNA stability, translation, splicing, and export, and further influence gene expression. For example, YTHDF1 and YTHDC2 facilitate the translation of m⁶A‐modified mRNAs (Wang et al, 2015; Hsu et al, 2017; Mao et al, 2019), while YTHDF2 mainly promotes mRNA decay of its targeted transcripts (Wang et al, 2014). Recent studies have found that m⁶A methylation plays a crucial role in mediating many important biological processes such as adipogenesis, neurogenesis, embryogenesis, and carcinogenesis (Zhang et al, 2017; Frye et al, 2018; Liu et al, 2019; Wu & Wang, 2021). Notably, great progress has been made in the significant roles of m⁶A in regulating white adipocyte differentiation. However, whether m⁶A regulates white adipocytes browning and thermogenesis is largely unknown.

In this study, we report that adipose tissue‐specific knockout of Fto reduces high‐fat diet (HFD)‐induced obesity through enhancing WAT‐mediated thermogenesis. FTO deficiency in primary beige adipocytes increases the expression of hypoxia inducible factor 1 subunit alpha (HIF1A) in an m⁶A‐dependent and YTHDC2‐mediated manner, and further promotes thermogenic gene expression. Our results reveal that m⁶A modification plays a critical role in the epigenetic regulation of white fat browning and thermogenic function, which provide an attractive therapeutic strategy to fight against obesity and related metabolic disease.

Results

Adipose‐specific knockout of Fto prevents HFD‐induced obesity

To investigate the physiological function of FTO in adipose tissue and energy homeostasis, we generated an adipose‐specific Fto knockout (Fto ^AKO) mouse by crossing Fto ^flox/flox mice with the transgenic mice harboring Cre recombinase driven by Adipoq promotor (Adipoq‐Cre) (Fig EV1A). Littermates lacking the Cre gene (homozygous Fto ^flox/flox) were used as controls. As expected, Western blot analysis revealed a dramatic reduction of FTO protein level in WAT and BAT but not in other tissues compared to Fto ^flox/flox mice (Fig EV1B), verifying the specificity of the Fto ^AKO mice.

Figure EV1. Adipocyte‐specific Fto‐deleted mice accumulate less fat under chow‐fed conditions.

• A
  The schematic illustration of the locations of Fto exons and loxp sites flanking exon 3, and the deletion of the exon 3 by Adipoq‐Cre recombinase.
• B
  Western blotting analysis of FTO in BAT, WAT, muscle, liver, and brain from Fto ^flox/flox and Fto ^AKO mice fed a chow diet.
• C
  A representative photograph of 16‐week‐old male Fto ^flox/flox and Fto ^AKO mice fed a chow diet. Scale bar, 2 cm.
• D
  The growth curve of Fto ^flox/flox and Fto ^AKO mice fed a chow diet.
• E
  Food intake of Fto ^flox/flox and Fto ^AKO mice during chow diet feeding.
• F
  The ratio of weight of iWAT, eWAT, and BAT to body weight in Fto ^flox/flox and Fto ^AKO mice.
• G
  Representative photos of iWAT, eWAT, and BAT from Fto ^flox/flox and Fto ^AKO mice. Scale bar, 1 cm.
• H
  Representative H&E staining of iWAT, eWAT, and BAT from Fto ^flox/flox and Fto ^AKO mice. Scale bar, 100 μm.
• I
  Adipocyte sizes of iWAT from Fto ^flox/flox and Fto ^AKO mice.
• J, K
  Serum levels of glucose (J) and triglyceride (K) in Fto ^flox/flox and Fto ^AKO mice.

Data information: The data were presented as the mean ± SD of triplicate tests (n = 6). Statistical analyses were performed using Student t‐test. **P < 0.01, ***P < 0.001.

We first analyzed the body weight of Fto ^flox/flox and Fto ^AKO mice fed a HFD or chow diet for 13 weeks (age 16 weeks). The results showed that Fto ^AKO mice gained less body weight and fat accumulation than Fto ^flox/flox mice under HFD condition (Fig 1A and B) or chow‐fed conditions (Fig EV1C and D), while the food intake showed no significant difference (Figs 1C and EV1E). Furthermore, the subcutaneous inguinal WAT (iWAT) and visceral epididymal WAT (eWAT) of Fto ^AKO mice were smaller and weighed less than those of Fto ^flox/flox mice fed a HFD (Fig 1D and E) or chow‐fed, respectively (Fig EV1F and G). In agreement with these results, histological analysis showed that Fto ^AKO mice had markedly reduced lipid contents in WATs (Figs 1F and EV1H). Especially in iWAT, the adipocyte size was dramatically reduced (Figs 1G and EV1I), indicating that adipose‐specific Fto knockout was protected against adipocyte hypertrophy of WAT. Moreover, the morphology of interscapular BAT in Fto ^flox/flox mice, but not Fto ^AKO mice, appeared abnormal, enlarged and “whitening” under HFD condition (Fig 1D–F), Of note, the difference was not significant under chow diet condition (Fig EV1F–H). Compared to Fto ^flox/flox mice, Fto ^AKO mice exhibited smaller and healthier liver and are protected against HFD‐induced fatty liver (Fig 1H). The serum of HFD‐fed Fto ^AKO mice showed lower levels of glucose, triglyceride, and free fatty acid (FFA) than those of Fto ^flox/flox mice under fasting condition (Fig 1I–K). Similar results were obtained in chow diet‐fed Fto ^AKO (Fig EV1J and K). We next tested the glucose tolerance and insulin response of HFD‐fed Fto ^flox/flox and Fto ^AKO mice. After glucose injection, the blood glucose declined markedly faster in Fto ^AKO mice when compared with that in Fto ^flox/flox mice (Fig 1L). The insulin tolerance test (ITT) also showed that Fto ^AKO mice exhibited enhanced insulin sensitivity (Fig 1M). Improved glucose and insulin tolerance tests demonstrate that adipose‐specific Fto knockout ameliorated HFD‐induced glucose intolerance and insulin resistance. These results demonstrate that adipose‐specific Fto knockout mice are protected against HFD‐induced obesity and fat accumulation.

Figure 1. Ablation of Fto in adipose tissue resists HFD‐induced obesity.

• A
  Representative photographs of 16‐week‐old male Fto ^flox/flox and Fto ^AKO (Fto ^flox/flox; Adipoq‐Cre) mice fed a HFD. Scale bar, 2 cm.
• B
  The growth curve of Fto ^flox/flox and Fto ^AKO mice fed a HFD (n = 8).
• C
  Food intake of Fto ^flox/flox and Fto ^AKO mice fed a HFD.
• D
  The ratio of weight of iWAT, eWAT, and BAT to body weight in 16‐week‐old Fto ^flox/flox and Fto ^AKO mice fed a HFD.
• E
  Representative photos of iWAT, eWAT, and BAT from Fto ^flox/flox and Fto ^AKO mice fed a HFD. Scale bar, 1 cm.
• F
  Representative H&E staining of iWAT, eWAT, and BAT from Fto ^flox/flox and Fto ^AKO mice fed a HFD. Scale bar, 200 μm.
• G
  Adipocyte sizes of iWAT from Fto ^flox/flox and Fto ^AKO mice fed a HFD were estimated from the H&E staining results in (F) using ImageJ. The percentage of cells with the indicated sizes is shown.
• H
  A representative image (upper) and H&E staining (lower) of liver from Fto ^flox/flox and Fto ^AKO mice fed a HFD. Upper image, scale bar, 0.5 cm. Lower image, scale bar, 50 μm.
• I–K
  Serum levels of glucose (I), triglyceride (J), and free fatty acid (FFA) (K) in Fto ^flox/flox and Fto ^AKO mice fed a HFD followed by fasting.
• L, M
  The blood glucose level of Fto ^flox/flox and Fto ^AKO mice fed a HFD after intraperitoneal injection of glucose or insulin for glucose (GTT) (L) and insulin tolerance tests (ITT) (M), respectively.

Data information: The data were presented as the mean ± SD (n = 6). Statistical analyses were performed using Student t‐test. *P < 0.05, **P < 0.01, ***P < 0.001.

FTO deficiency enhances energy expenditure and induces browning of white adipocytes

Because adipose‐specific Fto depletion significantly reduced WAT mass and levels of blood glucose and serum triglyceride under HFD or chow diet treatment, without a change in food intake, we speculated that FTO deficiency might influence the energy expenditure of WAT. To test our hypothesis, gas exchange and energy expenditure were monitored by housing HFD‐fed Fto ^flox/flox and Fto ^AKO mice in metabolic cages. We found that Fto ^AKO mice showed promoted oxygen consumption, carbon dioxide generation, and energy heat generation during both light and dark cycles (Fig 2A–C). Consistent with an increased metabolic rate, Fto ^AKO mice fed a HFD had higher core body temperature at room temperature when compared with the Fto ^flox/flox counterparts (Fig 2D). Since cold exposure is a classical inducer of adaptive thermogenesis, we examined the protein level of FTO in iWAT and BAT following cold exposure treatment. FTO protein levels in both iWAT and BAT were unchanged after cold exposure (Fig EV2A and B), indicating that FTO is not controlled by cold exposure. Also, Fto ^AKO mice maintained body temperature against acute cold exposure more efficiently than Fto ^flox/flox mice (Fig 2E). A similar phenomenon was observed in chow diet condition (Fig EV2C and D), indicating the enhanced energy expenditure and thermogenesis contribute to attenuated adiposity in Fto ^AKO mice. Furthermore, we test the non‐shivering thermogenic (NST) capacity by adapting Fto ^flox/flox and Fto ^AKO mice to thermoneutral conditions or injecting norepinephrine (NE) and measuring the acute oxygen consumption. Fto ^AKO mice showed higher oxygen consumption than Fto ^flox/flox mice (Fig EV2E), indicating that FTO deficiency increased non‐shivering thermogenesis. To investigate whether FTO‐mediated thermogenesis in iWAT is responsible for the enhanced energy expenditure rather than an effect in BAT, we surgically removed interscapular BAT in Fto ^flox/flox and Fto ^AKO mice and determine the phenotype under room temperature, cold, and thermoneutral conditions (30°C). After BAT removal, Fto ^AKO mice still had higher core body temperature at room temperature (Fig EV2F) and better maintained body temperature against cold exposure (Fig EV2G), but not under thermoneutral conditions (Fig EV2H), when compared with the Fto ^flox/flox mice. These results suggest that FTO‐regulated thermogenesis in WAT plays a dominating role in overall systemic energy expenditure. We, therefore, examined the expression of genes involved in thermogenesis, mitochondrial biogenesis, and adipogenesis in iWAT harvested from Fto ^flox/flox and Fto ^AKO mice. FTO deficiency resulted in an increased expression of thermogenic genes, including Ucp1, Ppargc1a, Prdm16, Cidea, Pparg, and Nrf1, while the general markers of WAT were decreased (Fig 2F). Consistently, the protein abundance of UCP1, PPARGC1A, and PPARG was upregulated in iWAT of Fto ^AKO mice (Fig 2G). Additionally, FTO deficiency decreased mRNA levels of several inflammation‐related and fibrosis‐related genes in iWAT (Fig EV2I). Immunohistochemical and immunofluorescence analysis also revealed facilitated UCP1 expression in iWAT of Fto ^AKO mice (Fig 2H and I). Similar results were obtained in the BAT of Fto ^flox/flox and Fto ^AKO mice fed a HFD (Fig EV2J–M). On a chow diet, Fto knockout promoted the mRNA levels of thermogenic genes in iWAT (Fig EV2N). The protein expression of UCP1 and PPARGC1A were also significantly enhanced in iWAT of Fto ^AKO mice fed a chow diet (Fig EV2O). However, loss of FTO decreased thermogenic gene expression in BAT (Fig EV2P), probably due to compensatory effects, while the protein expression of UCP1 and PPARGC1A were unchanged (Fig EV2Q). These data suggest that the increased thermogenic energy expenditure of iWAT is the major cause of the attenuated adiposity in Fto ^AKO mice.

Figure 2. Adipose‐specific Fto depletion enhances energy expenditure and induces browning of white adipocytes.

1. O₂ consumption of 16‐week‐old Fto ^flox/flox and Fto ^AKO mice fed a HFD. White and gray areas in the graphs indicate day and night, respectively.
2. CO₂ generation of 16‐week‐old Fto ^flox/flox and Fto ^AKO mice fed a HFD. White and gray areas in the graphs indicate day and night, respectively.
3. Energy expenditure of 16‐week‐old Fto ^flox/flox and Fto ^AKO mice fed a HFD. White and gray areas in the graphs indicate day and night, respectively.
4. The rectal temperature of Fto ^flox/flox and Fto ^AKO mice fed a HFD at room temperature (n = 6).
5. The rectal temperature of Fto ^flox/flox and Fto ^AKO mice fed a HFD during acute cold exposure (4°C) and fasting for 6 h.
6. qPCR analysis of genes involved in thermogenesis, mitochondrial biogenesis and adipogenesis in iWAT of Fto ^flox/flox and Fto ^AKO mice.
7. Western blot analysis of the protein levels of UCP1, PPARGC1A and PPARG in iWAT of Fto ^flox/flox and Fto ^AKO mice.
8. Immunohistochemical detection of UCP1 in iWAT from Fto ^flox/flox and Fto ^AKO mice following 6 days cold exposure (4°C). Scale bar, 200 μm.
9. Immunofluorescence detection of UCP1 in iWAT from Fto ^flox/flox and Fto ^AKO mice following 6 days cold exposure (4°C). Scale bar, 20 μm.
10. Western blot analysis of FTO protein level in differentiated beige adipocytes from Fto ^flox/flox mice infected with retroviruses MSCVhygro expressing vector (Vec) or Cre.
11. The O₂ consumption rate (OCR) in differentiated beige adipocytes described in (J). The average basal and maximal respiration rates are shown in lower panel.
12. qPCR analysis of genes involved in thermogenesis in differentiated beige adipocytes described in (J).
13. Western blot analysis of UCP1, PPARGC1A, and PPARG protein levels in differentiated beige adipocytes described in (J).
14. Immunofluorescent staining of UCP1 in differentiated beige adipocytes described in (J). Scale bar, 20 μm.

Data information: The data were presented as the mean ± SD of n = 6 mice per group or triplicate tests (n = 3). Statistical analyses were performed using Student t‐test. **P < 0.01, ***P < 0.001.

Figure EV2. Adipose‐specific Fto knockout mice exhibits promoted energy expenditure.

1. Western blot analysis of the protein levels of UCP1 and FTO in iWAT from 8‐week‐old male mice housed in cages at room temperature (RT) or at 4°C overnight with free access to food and water.
2. Western blot analysis of the protein levels of UCP1 and FTO in BAT from 8‐week‐old male mice housed in cages at RT or at 4°C overnight with free access to food and water.
3. The rectal temperature of Fto ^flox/flox and Fto ^AKO mice fed a chow diet at room temperature.
4. The rectal temperature of Fto ^flox/flox and Fto ^AKO mice fed a chow diet during acute cold exposure (4°C) and fasting for 6 h (n = 6).
5. Time course of O₂ consumption during basal and upon CL administration upon acclimation to thermoneutrality.
6. The rectal temperature of BAT‐removal Fto ^flox/flox and Fto ^AKO mice at room temperature (n = 6).
7. The rectal temperature of BAT‐removal Fto ^flox/flox and Fto ^AKO mice during acute cold exposure (4°C) and fasting for 4 h (n = 6).
8. The rectal temperature of BAT‐removal Fto ^flox/flox and Fto ^AKO mice at thermoneutral conditions (n = 5).
9. qPCR analysis of genes involved in inflammation and fibrosis in iWAT of Fto ^flox/flox and Fto ^AKO mice fed a HFD.
10. qPCR analysis of genes involved in thermogenesis, mitochondrial biogenesis and adipogenesis in BAT of Fto ^flox/flox and Fto ^AKO mice fed a HFD.
11. Western blot analysis of the protein levels of UCP1, PPARGC1A, PPARG, and HIF1A in BAT of Fto ^flox/flox and Fto ^AKO mice fed a HFD.
12. Immunohistochemical detection of UCP1 in BAT from Fto ^flox/flox and Fto ^AKO mice fed a HFD following 6 days cold exposure (4°C). Scale bar, 200 μm.
13. Immunofluorescence detection of UCP1 in BAT from Fto ^flox/flox and Fto ^AKO mice fed a HFD following 6 days cold exposure (4°C). Scale bar, 20 μm.
14. qPCR analysis of genes involved in thermogenesis and mitochondrial biogenesis in iWAT of Fto ^flox/flox and Fto ^AKO mice fed a chow diet.
15. Western blot analysis of the protein levels of UCP1 and PPARGC1A in iWAT of Fto ^flox/flox and Fto ^AKO mice fed a chow diet.
16. qPCR analysis of genes involved in thermogenesis and mitochondrial biogenesis in BAT of Fto ^flox/flox and Fto ^AKO mice fed a chow diet.
17. Western blot analysis of the protein levels of UCP1 and PPARGC1A in BAT of Fto ^flox/flox and Fto ^AKO mice fed a chow diet.

Data information: The data were presented as the mean ± SD of triplicate tests (n = 6). Statistical analyses were performed using Student t‐test. *P < 0.05, **P < 0.01, ***P < 0.001.

We next focused on iWAT to explore the underlying mechanism. To determine whether FTO affects the thermogenic program and browning of primary white adipocytes, we established beige adipocyte differentiation from primary white preadipocytes isolated from iWAT of Fto ^flox/flox mice and treated with retroviruses‐expressing Cre recombinase (Cre) or retroviruses‐expressing vector (Vec) as a control. Ectopic expression of Cre efficiently diminished FTO expression in primary differentiated beige adipocytes (Fig 2J). Consistent with in vivo data, knockout of FTO promoted mitochondrial respiration and oxygen consumption of differentiated beige adipocytes (Fig 2K). In agreement with this observation, FTO depletion increased the mRNA levels of genes involved in thermogenesis, including Ucp1, Ppargc1a, Prdm16, Cidea, and Pparg (Fig 2L). We observed that the protein levels of UCP1, PPARGC1A, and PPARG were also dramatically upregulated in FTO‐depleted differentiated adipocytes (Fig 2M). Immunohistochemistry analysis revealed FTO deficiency markedly increased the UCP1⁺ beige adipocyte population (Fig 2N). These results demonstrate that FTO depletion enhances thermogenesis and browning of white adipocytes.

Knockout of FTO induces thermogenic gene expression through activating HIF1A

To further elucidate the underlying molecular mechanism of FTO depletion‐mediated thermogenic gene activation, we analyze the published m⁶A‐seq data of 3T3‐L1 white preadipocytes with or without FTO knockdown (Zhao et al, 2014). The Gene Ontology analysis showed that the cellular response to hypoxia and temperature homeostasis were significantly upregulated in FTO‐depleted cells (Fig 3A). Various studies have shown that HIF1A, a vital regulator of cellular adaptation to oxygen stress functioning as a direct transcription factor, is involved in the regulation of obesity and energy metabolism (Mylonis et al, 2019). These results prompted us to hypothesize that FTO might regulate thermogenic gene expression through mediating HIF1A expression. Indeed, we found that the gene expression of Hif1a was not significantly changed (Fig 3B), while the HIF1A protein level was dramatically increased in iWAT of Fto ^AKO (Fig 3C), but not BAT (Fig EV2G), indicating FTO might control HIF1A expression at the translational level. Consistently, knockout of FTO in differentiated beige adipocytes facilitated protein abundance, but not gene expression of HIF1A (Fig 3D and E). In addition, knockdown of HIF1A restored mRNA and protein expression of thermogenic genes in FTO‐depleted beige adipocytes (Fig 3F and G), further confirming that the FTO deletion‐mediated enhancement in browning of white adipocyte is modulated by HIF1A.

Figure 3. FTO regulates thermogenic gene expression via targeting HIF1A.

1. Gene Ontology analysis of up‐regulated genes of siCtrl and siFto 3T3‐L1 cells. P value was calculated by Fisher's exact test in DAVID (The Database for Annotation, Visualization and Integrated Discovery).
2. Western blot analysis of the protein level of HIF1A in iWAT of Fto ^flox/flox and Fto ^AKO mice (n = 6).
3. qPCR analysis of mRNA level of Hif1a in iWAT of Fto ^flox/flox and Fto ^AKO mice (n = 6).
4. qPCR analysis of mRNA level of Hif1a in differentiated beige adipocytes from Fto ^flox/flox mice infected with retroviruses MSCVhygro expressing Vec or Cre.
5. Western blot analysis of the protein level of HIF1A in differentiated beige adipocytes described in (D).
6. qPCR analysis of mRNA levels of Ucp1, Ppargc1a, Prdm16, and Pparg in differentiated beige adipocytes from Fto ^flox/flox mice infected with retroviruses MSCVhygro expressing Vec, Cre or transfected with Hif1a siRNA.
7. Western blot analysis of the protein levels of FTO, HIF1A, UCP1, PPARGC1A, and PPARG in differentiated beige adipocytes described in (F).

Data information: The data were presented as the mean ± SD of triplicate tests (n = 3). Statistical analyses were performed using Student t‐test or ANOVA. **P < 0.01, ***P < 0.001.

HIF1A positively regulates thermogenic gene expression through mediating their transcription

To explore the role of HIF1A in thermogenesis and browning of white adipocytes, we conducted gain‐ and loss‐of‐function studies in 3T3‐L1 cells. Overexpression and knockdown efficiency of HIF1A in cells was validated by Western blot (Fig 4A and B). The protein abundance of FTO was unchanged by HIF1A, according with our proposed upstream–downstream relationship between HIF1A and FTO. Using a Seahorse XF96 Extracellular Flux Analyzer, we found that HIF1A overexpression increased mitochondrial oxygen consumption rate (OCR) of differentiated beige adipocytes (Fig 4C), while silencing of HIF1A suppressed mitochondrial respiration (Fig 4D), indicating a positive correlation between HIF1A and thermogenesis in beige adipocytes. Consistently, the protein and mRNA levels of thermogenic genes were significantly upregulated in HIF1A‐overexpressing cells (Fig 4E and G). In contrast, knockdown of HIF1A markedly decreased protein abundance and mRNA expression of thermogenic genes (Fig 4F and H). In addition, immunofluorescence assays showed that HIF1A deficiency attenuated UCP1 expression in beige adipocytes (Fig 4I).

Figure 4. HIF1A facilitates browning of white adipocytes by activating transcription of thermogenic genes.

1. Western blot analysis of HIF1A and FTO protein levels in 3T3‐L1 cells overexpressing Vec or HIF1A plasmid.
2. Western blot analysis of HIF1A and FTO protein levels in siCtrl or siHif1a 3T3‐L1 cells.
3. The O₂ consumption rate (OCR) in differentiated beige adipocytes from 3T3‐L1 cells overexpressing Vec or HIF1A plasmid. The average basal and maximal respiration rates are shown in right panel.
4. The O₂ consumption rate (OCR) in differentiated beige adipocytes from siCtrl or siHif1a 3T3‐L1 cells. The average basal and maximal respiration rates are shown in right panel.
5. Western blot analysis of UCP1, PPARGC1A, and PPARG in differentiated beige adipocytes described in (C).
6. Western blot analysis of UCP1, PPARGC1A, and PPARG in differentiated beige adipocytes described in (D).
7. qPCR analysis of genes involved in thermogenesis in differentiated beige adipocytes described in (C).
8. qPCR analysis of genes involved in thermogenesis in differentiated beige adipocytes described in (D).
9. Immunofluorescent staining of UCP1 in differentiated beige adipocytes described in (D). Scale bar, 20 μm.
10. ChIP‐qPCR analysis of the association of HIF1A with the promoter of thermogenic genes in 3T3‐L1 cells differentiated into beige adipocytes. IgG was used as control. Consensus HIF1A binding sites were obtained from JASPAR database (a database of transcription factor binding profiles).

Data information: The data were presented as the mean ± SD of triplicate tests (n = 3). Statistical analyses were performed using Student t‐test. *P < 0.05, **P < 0.01, ***P < 0.001.

Since HIF1 is a heterodimeric transcription factor that functions as a master regulator of cellular and systemic homeostatic response to hypoxia by activating transcription of target genes, we speculated that HIF1A facilitated thermogenic gene expression by regulating their transcription. As expected, the chromatin immunoprecipitation‐qPCR (ChIP‐qPCR) assay indicated that HIF1A was associated with Ppargc1a, Prdm16, and Pparg (Fig 4J), while Vegf was chosen as the positive control. However, HIF1A was not associated with Ucp1 (Fig 4J), indicating that Ppargc1a, Prdm16, and Pparg, but not Ucp1, were direct targets of HIF1A. Collectively, these results demonstrate that HIF1A positively modulates thermogenic gene activation through promoting transcription of Ppargc1a, Prdm16, and Pparg and further activating Ucp1 expression.

FTO decreases the protein expression of HIF1A transcript via m⁶A modification

To examined if the m⁶A demethylase activity of FTO is required for suppressing HIF1A and thermogenic gene expression, we added back wild type (WT) or demethylase catalytic mutant (R96Q, MUT) FTO to FTO‐depleted beige adipocytes. The results showed that ectopic expression of FTO‐WT, but not FTO‐MUT, restored the protein expression of HIF1A (Fig 5A). Importantly, a similar trend was observed for UCP1 and PPARGC1A (Fig 5A), indicating that m⁶A demethylase activity of FTO is required for suppressing HIF1A and thermogenic gene expression in beige adipocytes. Consistently, overexpression of FTO‐WT significantly inhibited protein levels of HIF1A, UCP1, and PPARGC1A in beige 3T3‐L1 adipocytes, while no significant difference was observed in FTO‐MUT‐overexpressing cells (Fig EV3A). Compared with FTO‐MUT or vector, ectopically expressed FTO‐WT also decreased mRNA levels of thermogenic genes and Hif1a (Fig EV3B). Immunofluorescence assays showed that beige adipocytes overexpressing FTO‐WT, rather than FTO‐MUT, exhibited significantly alleviated UCP1 (Fig EV3C). As expected, ectopic expression of the FTO‐WT, but not FTO‐MUT, repressed mitochondrial OCR in beige adipocytes (Fig EV3D). These data demonstrate that FTO modulates thermogenic gene expression and browning of white adipocytes in an m⁶A demethylase activity‐dependent manner.

Figure 5. FTO regulates the expression of HIF1A in an m⁶A‐dependent manner.

1. Western blot analysis of the protein levels of FTO, HIF1A, UCP1, and PPARGC1A in differentiated beige adipocytes from Fto ^flox/flox mice infected with retroviruses MSCVhygro expressing Vec, Cre or together with lentiviruses pCDHpuro expressing wild‐type (WT)‐FTO or mutant (MUT)‐FTO.
2. HPLC‐QgQ‐MS/MS quantification of the m⁶A/A in mRNA of siCtrl or siFto 3T3‐L1 cells. FTO knockdown efficiency was measured by Western blot.
3. Distribution of m⁶A peaks in Hif1a mRNA transcript in siCtrl and siFto 3T3‐L1 cells. Blue boxes represent exons, and blue lines represent introns. The m⁶A peaks were in the dashed red box.
4. Methylated RNA immunoprecipitation (MeRIP)‐qPCR analysis of m⁶A levels of Hif1a mRNA in siCtrl and siFto 3T3‐L1 cells.
5. Western blot analysis of FTO in input and HA‐RIP groups.
6. RIP‐qPCR analysis of the association of Hif1a with HA in 3T3‐L1 cells overexpressing HA‐tagged Vec or FTO. Enrichment of Hif1a with HA was measured by qPCR and normalized to input.
7. Synonymous mutation sites in Hif1a.
8. Western blot analysis of the protein levels of HIF1A, UCP1, PPARGC1A, PPARG, and GFP in 3T3‐L1 cells infected with lentiviruses pCDHpuro expressing Vec, WT‐ or MUT‐HIF1A.

Data information: The data were presented as the mean ± SD of triplicate tests (n = 3). Statistical analyses were performed using Student t‐test. **P < 0.01, ***P < 0.001.

Figure EV3. The demethylase activity of FTO suppresses the expression of thermogenic gene through HIF1A.

1. Western blot analysis of the protein levels of FTO, HIF1A, UCP1, and PPARGC1A in 3T3‐L1 cells infected with lentiviruses pCDHpuro expressing Vec, WT‐FTO, or MUT‐FTO and differentiated into beige adipocytes.
2. qPCR analysis of genes involved in thermogenesis and Hif1a in cells described in (A).
3. Immunofluorescent staining of UCP1 in cells described in (A). Scale bar, 20 μm.
4. The O₂ consumption rate (OCR) in cells described in (A). The average basal and maximal respiration rates are shown in lower panel.

Data information: The data were presented as the mean ± SD of triplicate tests (n = 3). Statistical analyses were performed using ANOVA. *P < 0.05, **P < 0.01, ***P < 0.001.

Next, we investigated whether FTO regulated HIF1A protein expression through m⁶A modification. High‐performance liquid chromatography coupled with triple quadrupole tandem mass spectrometry (HPLC‐QqQ‐MS/MS) analysis revealed that knockdown of FTO increased the cellular m⁶A levels when compared with control cells (Fig 5B). Based on the published m⁶A‐seq data (GSE53244) of 3T3‐L1 cells (Zhao et al, 2014), we found that m⁶A peaks were located at CDS of Hif1a transcript and sharply increased upon FTO depletion (Fig 5C). Consistently, methylated RNA immunoprecipitation‐qPCR (meRIP‐qPCR) analysis showed that the m⁶A level in Hif1a transcript was dramatically increased in FTO‐deficient cells (Fig 5D), which might contribute to the augmented protein expression of HIF1A. Furthermore, RNA immunoprecipitation‐qPCR (RIP‐qPCR) analysis revealed that Hif1a interacted with HA‐tagged FTO in 3T3‐L1 cells (Fig 5E and F), indicating Hif1a is a direct target of FTO. To further investigate whether the m⁶A methylation at CDS is associated with the protein expression of HIF1A, we introduced synonymous mutation at the m⁶A site on CDS (HIF1A‐MUT) (Fig 5G). The results showed that HIF1A protein abundance was higher in HIF1A‐WT‐overexpressing cells than that in HIF1A‐MUT‐overexpressing cells (Fig 5H), suggesting that m⁶A enhances HIF1A protein expression and the mutation significantly decreased the protein levels of HIF1A. As expected, less HIF1A protein led to relatively lower levels of UCP1, PPARGC1A, and PPARG in HIF1A‐MUT‐overexpressing cells than that in HIF1A‐WT‐overexpressing cells (Fig 5H). Together, these results illustrate that FTO deficiency promotes the protein expression of HIF1A by increasing the m⁶A modification in Hif1a mRNA.

m⁶A reader YTHDC2 facilitates Hif1a translation

The next obvious question is how m⁶A modification regulates protein expression of HIF1A. Numerous studies have reported that m⁶A methylation on mRNAs regulates translation, which is mediated by specific m⁶A binding proteins, including YTHDF1, YTHDF2, IGF2BP1, IGF2BP2, IGF2BP3, and YTHDC2 (Wang et al, 2015; Zhou et al, 2015; Huang et al, 2018; Mao et al, 2019). To identify which m⁶A readers were responsible for this regulation, we performed siRNA‐mediated knockdown of Ythdf1, Ythdf2, Igf2bp1, Igf2bp2, Igf2bp3, or Ythdc2. As shown in Figs 6A and EV4A, knockdown of YTHDF1, YTHDF2 in 3T3‐L1 cells did not affect the protein abundance and mRNA expression of Hif1a. The gene expression of Hif1a was decreased in IGF2BP1‐, IGF2BP2‐, and IGF2BP3‐deficient cells (Fig EV4B). However, depletion of IGF2BP1, IGF2BP2, or IGF2BP3 did not influence protein level of HIF1A (Fig 6B), which could not explain the m⁶A‐promoted HIF1A protein expression. Notably, knockdown of YTHDC2 significantly reduced protein expression, but not mRNA level of HIF1A (Figs 6C and EV4C). On the contrary, overexpression of YTHDC2 increased protein level of HIF1A (Fig 6D). Using RIP‐qPCR analysis, we found that Hif1a transcript interacted with FLAG‐tagged YTHDC2 in 3T3‐L1 cells (Fig 6E and F), confirming that YTHDC2 directly binds Hif1a transcript.

Figure 6. m⁶A reader YTHDC2 mediates mRNA translation of Hif1a .

1. Western blot analysis of the protein levels of YTHDF1, YTHDF2, and HIF1A in siCtrl, siYthdf1 and siYthdf2 3T3‐L1 cells.
2. Western blot analysis of the protein levels of HIF1A in siCtrl, siIgf2bp1, siIgf2bp2 and siIgf2bp3 3T3‐L1 cells.
3. Western blot analysis of the protein levels of YTHDC2 and HIF1A in siCtrl and siYthdc2 3T3‐L1 cells.
4. Western blot analysis of the protein levels of YTHDC2 and HIF1A in 3T3‐L1 cells overexpressing Vec or YTHDC2 plasmid.
5. Western blot analysis of YTHDC2 in input and FLAG‐RIP groups.
6. RIP‐qPCR analysis of the association of Hif1a with FLAG in 3T3‐L1 cells overexpressing FLAG‐tagged Vec or YTHDC2. FLAG was measured by qPCR and normalized to input.
7. Polysome profiling of 3T3‐L1 cells overexpressing Vec or YTHDC2 plasmid.
8. qPCR analysis of mRNA levels of Hif1a in 40–80S or polysome fraction in 3T3‐L1 cells overexpressing Vec or YTHDC2 plasmid.
9. Polysome profiling of siCtrl and siYthdc2 3T3‐L1 cells.
10. qPCR analysis of mRNA levels of Hif1a in 40–80S or polysome fraction in siCtrl and siYthdc2 3T3‐L1 cells.

Data information: The data were presented as the mean ± SD of triplicate tests (n = 3). Statistical analyses were performed using Student t‐test. **P < 0.01, ***P < 0.001.

Figure EV4. HIF1A is not a target of Ythdf1‐2 nor Igf2bp1‐3 .

1. qPCR analysis of mRNA level of Ythdf1, Ythdf2, and Hif1a in siCtrl, siYthdf1 or siYthdf2 3T3‐L1 cells.
2. qPCR analysis of mRNA level of Igf2bp1, Igf2bp2, Igf2bp3 and Hif1a in siCtrl, siIgf2bp1, siIgf2bp2 or siIgf2bp3 3T3‐L1 cells.
3. qPCR analysis of mRNA level of Ythdc2 and Hif1a in siCtrl or siYthdc2 3T3‐L1 cells.
4. mRNA stability analysis of Hif1a in preadipocytes from Fto ^flox/flox mice infected with retroviruses MSCVhygro expressing Vec or Cre.

Data information: The data were presented as the mean ± SD of triplicate tests (n = 3). Statistical analyses were performed using Student t‐test. **P < 0.01, ***P < 0.001.

To investigate whether YTHDC2 mediates HIF1A expression through promoting mRNA translation, we performed polysome fraction analysis in 3T3‐L1 cells upon HIF1A overexpression or knockdown, separating cellular fractions along a 10%‐50% sucrose gradient. The results showed that forced expression or knockdown of YTHDC2 in 3T3‐L1 cells increased or decreased, respectively, the Hif1a mRNA level in the polysome portion (Fig 6G–J), whereas there was no significant change in the 40–80S portion, suggesting that YTHDC2 promotes translation efficiency of Hif1a mRNA. In addition, we validated that FTO deficiency did not affect mRNA stability of Hif1a (Fig EV4D). Taken together, these results indicate that FTO deficiency promotes HIF1A and thermogenic gene expression through m⁶A‐YTHDC2‐dependent mRNA translation.

Discussion

Our results provide compelling evidence that FTO depletion promotes browning of white adipocytes and thermogenic gene expression through increasing m⁶A methylation of Hif1a transcripts. FTO is the first identified m⁶A demethylase, as well as the first discovered obesity gene, which is strongly associated with obesity and adipocyte differentiation (Loos & Bouchard, 2008; Jia et al, 2011). Studies have shown the regulatory role of FTO in browning of white adipocytes and energy metabolism. Global FTO knockout or FTO C‐terminal mutation in mice decreases body weight through enhancing energy expenditure (Church et al, 2009; Fischer et al, 2009). FTO‐deficient mice under HFD condition exhibited promotes browning process of WAT and upregulated mRNA expression of thermogenic genes, such as Ucp1, Ppargc1a, Pparg, and Prdm16 (Tews et al, 2013; Ronkainen et al, 2016). However, the underlying mechanism is unclear. In our study, we focus on the specific role of FTO in adipose tissue and find that FTO is a key negative regulator of white adipocytes browning by mediating m⁶A methylation. Deletion of FTO in vivo or in vitro dramatically increases browning of white adipocytes by promoting the expression of HIF1A and thermogenic genes, including Ucp1, Ppargc1a, Pparg, and Prdm16. Thus, Fto ^AKO mice show enhanced thermogenesis and resist HFD‐induced obesity and metabolic syndrome.

Few studies investigate the function of m⁶A modification in regulating energy metabolism of adipocytes. A recent study revealed that METTL3‐mediated m⁶A methylation was required for BAT postnatal development and thermogenesis (Wang et al, 2020). BAT‐specific knockout of METTL3 impaired postnatal development of BAT by decreasing m⁶A levels and expression of thermogenic genes, leading to alleviated energy expenditure and promoted HFD‐induced obesity. However, whether m⁶A modification regulates browning process and energy metabolism in WAT is unknown. In this study, we reveal that FTO‐mediated m⁶A demethylation of Hif1a acts as an important mechanism in inhibiting thermogenesis in iWAT. Adipose‐specific FTO depletion did not changed the expression of HIF1A in BAT. Moreover, FTO knockout only enhanced protein expression of thermogenic genes in BAT under HFD condition, but not chow diet condition. These results suggest that the FTO‐HIF1A pathway‐mediated thermogenesis in iWAT is the major cause of the metabolic improvements in Fto ^AKO mice. Similar with our results, entacapone, a chemical inhibitor of FTO m⁶A demethylation activity, was recently reported to reduced body weight and fat mass in diet‐induced obese mice by upregulated thermogenesis in iWAT and downregulated gluconeogenesis in the liver (Peng et al, 2019). Further study demonstrated that entacapone treatment enhanced thermogenesis by decreasing FTO‐catalyzed m⁶A demethylation of Foxo1 mRNA. Since Foxo1 and Hif1a are not the only two targets of FTO, other target genes might also play a role in m⁶A‐mediated white adipocytes browning and energy metabolism. Together, these works implicate an emerging role of m⁶A in the regulation of thermogenesis, which will be a new direction to explore in the future.

Our results show that Hif1a transcript is a critical substrate of FTO. HIF1A is a master regulator of cellular and systemic homeostatic response to hypoxia through activating transcription of genes involved in energy metabolism and other genes whose protein products increase oxygen delivery or facilitate metabolic adaptation to hypoxia (Lee et al, 2004). Several studies have investigated the relationship between HIF1A and obesity, how HIF1A affects obesity and thermogenesis, however, remains controversial. Studies show that HIF1A activation induces obesity and impedes energy expenditure, while inhibition of HIF1A ameliorates obesity (Krishnan et al, 2012; Sun et al, 2013; Lee et al, 2014). HIF1A is reported to suppress thermogenesis in brown adipose tissue and cause obesity (Jun et al, 2017). In contrast, other studies have reported that obesity is increased by HIF1A inhibition and decreased by activation of HIF1A signaling (Mylonis et al, 2019). Adipose tissue‐specific inhibition of HIF1A predisposes mice to HFD‐induced obesity by restricting the thermogenic functions of BAT (Zhang et al, 2010). This could be explained by the enhanced angiogenesis by HIF1A in BAT. However, the role of HIF1A in regulation of white adipocyte thermogenesis is unknown. In our study, we unveil that HIF1A plays a key role in facilitating browning and thermogenesis of WAT and HIF1A‐mediated thermogenesis prevent HFD‐induced obesity. Furthermore, lipolysis is increased in adipose tissue and 3T3‐L1 adipocytes in response to hypoxia (Yin et al, 2009), which suggest that hypoxia could promote lipolysis. Consistently, we found that serum levels of FFA are increased in Fto ^AKO mice, indicating the enhanced lipolysis in WAT. Forced expression or silencing of HIF1A promotes and attenuates, respectively, the expression of thermogenic genes and mitochondrial respiration of differentiated beige adipocytes. More importantly, using Chip‐qPCR assay, we further validate that HIF1A binds and promotes transcription of Ppaggc1a, Prdm16, and Pparg and then upregulates Ucp1 expression. These data suggest that FTO‐inhibited HIF1A expression represses thermogenesis. Although the role of FTO and HIF1A in adipocytes has been extensively studied, respectively, we uncover the interplay between FTO and HIF1A in the regulation of white adipocyte thermogenesis. Therefore, selectively promoting expression of Hif1a via the inhibition of FTO could potentially be a target to treat obesity.

Numerous studies have demonstrated that mRNA transcripts with m⁶A modifications tend to be promoted to translation or degradation (Wang et al, 2014, 2015; Li et al, 2017), which is regulated by various m⁶A‐binding proteins, especially YTH domain family proteins (Yang et al, 2018). YTHDF1 and YTHDC2 mediate translation of m⁶A‐modified transcripts (Wang et al, 2015; Hsu et al, 2017; Mao et al, 2019). YTHDF2 mainly controls mRNA decay of its target genes (Wang et al, 2014), while enhances the cap‐independent translation of m⁶A‐containing transcripts in response to heat shock (Zhou et al, 2015). In our study, we find that FTO deficiency only increases the protein expression of HIF1A, whereas the mRNA level is unchanged. More importantly, the positive correlation between m⁶A level and protein abundance of HIF1A further indicates that FTO‐mediated m⁶A demethylation influences mRNA translation of Hif1a. As expected, the lifetime of Hif1a mRNA shows no significant difference in control and FTO‐depleted primary beige adipocytes, further confirming that FTO regulates HIF1A at the translational level. We find that YTHDC2, but not YTHDF1 nor YTHDF2, regulates the translation and expression of HIF1A. A recent study showed that m⁶A located within CDS promoted mRNA translation by YTHDC2 (Mao et al, 2019). Using polysome fraction analysis, we validate that YTHDC2 positively regulates translation efficiency of Hif1a. Consistent with our results, YTHDC2 is also reported to promote translation initiation of Hif1a mRNA and metastasis in colon tumor cells (Tanabe et al, 2016), indicating a conserved role of YTHDC2 in regulating HIF1A expression among different cell types. The IGF2BP proteins, including IGF2BP1, IGF2BP2, and IGF2BP3, were recently reported to enhance mRNA stability and translation (Huang et al, 2018). Although loss of IGF2BP1, IGF2BP2, and IGF2BP3 decreased the gene expression of Hif1a, the protein level was unchanged, indicating that IGF2BP proteins might not involve in FTO‐inhibited HIF1A protein expression. These results reveal a key role of YTHDC2 in the m⁶A‐mediated expression of HIF1A and thermogenic genes. However, we still cannot rule out that other m⁶A‐binding proteins might target other thermogenic transcripts and play a role in the regulation of thermogenesis, which needs to be further investigated.

In conclusion, we show that FTO deficiency in adipose tissue protects against HFD‐induced obesity via promoting energy expenditure. Loss of FTO promotes thermogenic gene expression by increasing mRNA m⁶A level and protein abundance of HIF1A (Fig 7). Our results illuminate an emerging role of m⁶A methylation in modulating browning of white‐to‐beige adipocytes transition, providing a novel insight into the epigenetic mechanism of white adipocyte browning and energy homeostasis.

Figure 7. Working model of FTO in regulating browning of white adipocytes.

In normal condition, FTO demethylases m⁶A modification of Hif1a mRNA, which in turn prevents YTHDC2‐mediated mRNA translation and thermogenic gene expression, thereby increasing white adipocytes differentiation and lipid accumulation, leading to HFD‐induced obesity. FTO deficiency in adipose tissue increases mRNA m⁶A level of Hif1a transcript, leading to YTHDC2‐mediated translation and enhanced protein expression of HIF1A. Subsequently, HIF1A promotes transcription and expression of thermogenic genes, including Ppargc1a, Prdm16, Pparg, which results in white adipocytes browning and thermogenesis, thereby resisting obesity.

Materials and Methods

Animals

Conditional knockout mice with a loxP‐flanked Fto allele (Fto ^flox/flox mice) were generated as previously reported (Gao et al, 2010). Fto ^flox/flox mice were crossed with transgenic mice carrying Adipoq promoter‐driven Cre recombinase (The Jackson Laboratory, 010803) to produce the adipose‐specific Fto knockout mice (Adipoq‐Cre Fto ^flox/flox, Fto ^AKO mice). The Fto ^flox/flox littermates were used as controls. The genotypes of the mice were determined by PCR, and deletion of FTO in tissues was confirmed with Western blot. Mice were housed and maintained at 22 ± 2°C with a humidity of 35 ± 5% with 12‐h light and dark cycles and free access to water and food. Male Fto ^flox/flox and Fto ^AKO mice were used for experiments. For the HFD experiment, mice at 3‐week‐old were fed a high‐fat diet containing 60% fat‐derived calories (Research Diets, D12492) for 13 weeks. The body weight was record every week after weaning. After the mice were sacrificed, their tissues were collected, weighed, and immediately snap‐frozen in liquid nitrogen. All animal experiments were performed according to procedures approved by the Committee on Animal Care and Use and Committee on the Ethics of Animal Experiments of Zhejiang University.

Cold exposure and body temperature monitoring

For acute cold exposure, mice were placed in a freezer (4°C) for 6 h and fasted. Core body temperature was measured using a rectal probe at indication time points. Chronic exposure experiment was performed according to published protocol (Lim et al, 2012). Mice were placed in a freezer (4°C) for 6 days with free access to food and water. After cold exposure, mice were sacrificed and adipose tissues were collected for histology and protein analysis.

H&E staining, immunohistochemistry, and immunofluorescence

Tissues were fixed in 4% paraformaldehyde and embedded in paraffin. Hematoxylin and eosin (H&E) staining, immunohistochemical, and immunofluorescent analysis were carried out on 5 μm sections. For H&E staining, sections were stained by hematoxylin and eosin. For immunohistochemistry, 5 μm sections were treated with 3% hydrogen peroxide for 15 min at room temperature, followed by antigen recover using citrate buffer and blocking with bovine serum albumin (BSA) for 1 h at room temperature. UCP1 staining was performed according to the manufacturer’s protocol. Subsequently, samples were detected by goat anti‐rabbit IgG conjugated with HRP, followed by detection using DAB chromogenesis kit (Dako). The sections were observed under light microscopy. For immunofluorescence, sections were fixed with 4% paraformaldehyde, permeablized by 0.1% Triton X‐100, and incubated with UCP1 antibody, following by detecting secondary antibodies conjugated with DyLight 594 and DAPI staining. Immunofluorescent samples were imaged by confocal laser microscope (Carl Zeiss Ltd).

Blood biochemical analysis

Blood glucose level was measured using glucometer with glucose testing strips. Serum triglyceride level was determined using a triglyceride assay kit (Nanjing Jiancheng Bioengineering Institute, A110‐1‐1). Serum FAA level was detected using a nonesterified free fatty acids assay kit (Nanjing Jiancheng Bioengineering Institute, A042‐1‐1).

Glucose and insulin tolerance tests

For glucose tolerance test, mice fasted for overnight were intraperitoneally injected with glucose (2 g/kg). For insulin tolerance test, mice fasted for overnight were intraperitoneally injected with human insulin (0.75 U/kg, Sigma‐Aldrich). The tail was cleaned using 70% ethanol. Blood glucose level was detected in tail blood at 0, 15, 30, 60, 90, and 120 min after glucose or insulin injection using a glucometer with glucose testing strips.

Energy metabolism measurement

Energy metabolism experiments were performed using a Promethion metabolic measurement system (Sable Systems International). Mice were housed individually in metabolic cages and free to water and food, while temperature and humidity levels were tightly regulated. The concentration and flow of O₂ and CO₂ and energy expenditure were constantly monitored for 48 h.

Surgical removal of interscapular BAT

As previous descriptions (Kong et al, 2018), mice received a surgical removal of interscapular BAT. Mice were anesthetized by intraperitoneal injection of 5% chloral hydrate (8 μl/g body weight: Sigma‐c8383). All mice were treated with the pain‐killer meloxicam (ZooPharm) before surgery. Under a stereomicroscope, a midline incision in the skin was performed along the upper dorsal surface. The vein draining the iBAT was located and tied off above the point of branching into the two iBAT lobes. The interscapular BAT pads were separated carefully from the surrounding tissues by blunt dissection. The incision was closed, and animals were allowed to recover. Mice were maintained under a 12‐h light/12‐h dark cycle at thermoneutral conditions (30°C) with free access to food and water after surgery.

Thermogenic capacity experiment

To test the non‐shivering thermogenic (NST) capacity, metabolic rates upon administration of the norepinephrine (NE) were studied using a laboratory animal monitoring system (PhenoMaster, TSE Systems GmbH, Bad Homburg, Germany). Prior to the analyses, mice were acclimatized for 3 weeks to thermoneutrality at 29–31°C. Metabolic rates were monitored at thermoneutral temperature (30°C) (Virtue & Vidal‐Puig, 2013). Basal metabolic rates were monitored for subsequent 20 min until reaching steady state. Then, NE was injected intraperitoneally using a dosage of 1 mg/kg body weight (Sigma‐Aldrich, A9512). The subsequent increase in metabolic rates was monitored for 50 min.

Cell culture and beige adipocyte differentiation

Isolation of primary white preadipocytes from iWAT and induction of beige adipocyte differentiation were performed as previous described with minor modifications (Seale et al, 2011; Lee et al, 2019). Briefly, the iWAT were excised from 8‐week‐old male mice, minced by scissors, and then digested with 1 mg/ml type I collagenase (Gibco) at 37°C for 30 min. Digested tissue was filtered through 40‐mm sieve to remove large pieces, and the flow through was then centrifuged at 500 g for 5 min. Pellets containing primary white preadipocytes were incubated with red blood cell lysis buffer for 5 min, following centrifuged at 500 g for 5 min. The cell pellets were cultured in high‐glucose DMEM (Gibco) supplemented with 10% fetal bovine serum (FBS; Gibco), 1% penicillin–streptomycin at 37°C in a 5% CO₂ humidified incubator. The 3T3‐L1 mouse preadipocytes were purchased from ZenBio Inc. (ZenBio). Cells were tested negative for mycoplasma contamination before use. For beige adipocyte differentiation, 2 days post‐confluent primary white preadipocytes (day 0) were induced by induction medium containing 0.5 mM IBMX, 1 µM dexamethasone, 850 nM insulin, 1 nM T₃, 125 nM indomethacin, and 1 µM rosiglitazone. Two days after induction (day 2), medium was replaced with a maintenance medium containing 10% FBS, 850 nM insulin, 1 nM T₃, and 1 µM rosiglitazone. Fresh maintenance medium was replaced every 2 days until ready for harvest (generally day 8).

Cell transfection and virus infection

The siRNA and plasmid transfection were performed using Lipofectamine RNAiMAX (Invitrogen) and Lipofectamine 2000 (Invitrogen), respectively, according to the manufacturers’ instructions. The sequence for negative control siRNA is as follows (5ʹ–3ʹ): 5ʹ‐UUCUCCGAACGUGUCACGUTT‐3ʹ. Mouse Hif1a siRNA was ordered from Genepharma: 5ʹ‐GACACAGCCTCGATATGAA‐3ʹ. Mouse Fto siRNA was ordered from Qiagen as custom synthesis which targets 5ʹ‐TTAAGGTCCACTTCATCATCGCAGG‐3ʹ. Mouse Ythdf1 siRNA was ordered from Qiagen: 5ʹ‐GATCCTTACCTGTCCAGTTAC‐3ʹ. Mouse Ythdf2 siRNA was ordered from Qiagen: 5ʹ‐AAGGACGTTCCCAATAGCCAA‐3ʹ. Mouse Igf2bp1, Igf2bp2, and Igf2bp3 siRNA were ordered from Genepharma: 5ʹ‐GCCAUCAGCGUGCAUUCAATT‐3ʹ, 5ʹ‐GGAGCAAGUCAACACAGAUTT‐3ʹ, and 5ʹ‐GCAGAGGAUUCGUAAACUUTT‐3ʹ, respectively. Mouse Ythdc2 siRNA was ordered from Genepharma: 5ʹ‐CCUGUUAGAUGAUUGCUUUTTT‐3ʹ. The HA‐tagged wild‐type FTO‐CDS expression plasmid was generated by cloning the full‐length ORF of mouse Fto gene (NM_011936.2) into a pCDHpuro expression vector. The mutant FTO R96Q‐CDS was amplified by PCR and cloned into pCDHpuro vector. The FLAG‐tagged YTHDC2 plasmid was provided by professor Bin Shen (Hsu et al, 2017). The FLAG‐tagged HIF1A (NM_010431) plasmid was purchased from YouBio. The m⁶A mutant HIF1A plasmid was generated by a Multi Site‐Directed Mutagenesis Kit (Yeasen Biotech). For retrovirus infection of primary white preadipocytes, 60% confluent cells were incubated with retroviruses MSCVhygro expressing Vec or Cre in growth medium, followed by replacement with fresh media containing hygromycin (200 μg/ml) for selection for 2 days. For lentivirus infection of cells, 60% confluent cells were incubated with lentiviruses pCDHpuro expressing Vec, WT‐, or MUT‐FTO in growth medium, followed by replacement with fresh media containing puromycin (1 μg/ml) for selection for 2 days.

Mitochondrial respiration analysis

Mitochondrial oxygen consumption rate (OCR) of differentiated beige adipocytes was measured using a Seahorse XF96 Extracellular Flux Analyzer (Agilent) and XF Cell Mito Stress test kit (Agilent). Briefly, 2 × 10⁴ preadipocytes were seeded into XF96 cell culture microplate and differentiated into beige adipocytes. Cells at day 8 of differentiation were equilibrated in Seahorse assay medium with glucose (10 mM), sodium pyruvate (1 mM), and Glutamax (2 mM) for 30 min in a 37°C incubator without CO₂ before analysis. The XF96 plates were then transferred to a temperature‐controlled (37°C) Seahorse XF96 analyzer. During assay, cells were sequentially injected to obtain final concentrations of oligomycin (1 mM), FCCP (2 mM), and rotenone/actinomycin A (0.5 mM) were then sequentially added into the microplate by automatic pneumatic injection. OCR was recorded at the time points indicated in Figure. After the assay, cells were trypsinized and total cell number from each well were counted for normalization. Basal respiration was calculated as the OCR measurement before oligomycin injection subtracted by the measurement after rotenone injection. Maximum respiration was the OCR measurement after FCCP injection subtracted by the measurement after rotenone injection. Data were analyzed by Seahorse Wave Software (Agilent).

RNA isolation and quantitative real‐time PCR (qPCR) analysis

Total RNAs from tissue or cells were purified using TRIzolTM Reagent (Invitrogen) and converted to cDNA with M‐MLV reverse transcriptase (Invitrogen). Quantitative real‐time PCR (qPCR) analysis was performed using SYBR Green PCR Master Mix (Roche) following the manufacturer’s instructions and analyzed with an ABI Step‐One Plus™ Real‐Time PCR System (Applied Biosystems). Relative expression of mRNAs was determined after normalization to 18S or Actb. All real‐time qPCR reactions were carried out in triplicate. Primer sequences are listed in Table EV1.

Chromatin immunoprecipitation‐qPCR

The chromatin immunoprecipitation (ChIP) assay was performed using a ChIP Assay kit (Beyotime Biotechnology) according to the manufacturer’s instructions. Briefly, cells were crosslinked with 1% formaldehyde in culture medium for 10 min at room temperature followed by the addition of 125 mM glycine for 5 min, after which cells were washed twice with cold PBS and scraped into SDS lysis buffer supplemented with 1mM PMSF. After that, crosslinked chromatin was shattered by sonication with 5 s on/45 s off for 10 cycles. Precleared chromatin samples were immunoprecipitated with HIF1A antibody or control IgG at 4°C overnight. 1% of the precleared chromatin samples lacking primary antibody was used as the Input. The association between HIF1A and target genes was detected by qPCR. The primers were designed based on promoter sequences and predicted site sequence on the JASPAR database and listed in Table EV1.

Quantification of m⁶A in mRNA

mRNA was isolated from total RNA using a Dynabeads mRNA DIRECT kit (Invitrogen) following the manufacturer’s protocols. Quantification of m⁶A in mRNA was performed as described in our previous study (Wang et al, 2019). Briefly, 300 ng of mRNA was digested by nuclease P1 (2 U) at 42°C for 2 h, followed by the addition of alkaline phosphatase (0.5 U) with incubation at 37°C for 2 h. The amount of m⁶A in mRNA was measured using HPLC‐QqQ‐MS/MS.

Methylated RNA immunoprecipitation‐qPCR (MeRIP‐qPCR)

MeRIP‐qPCR analysis was performed according to a reported method (Dominissini et al, 2013). Briefly, mRNA was fragmented using RNA Fragmentation reagent (Invitrogen) at 70°C for 15 min. A small amount of the fragmented sample was used as input RNA. Fragmented mRNA was immunoprecipitated with m⁶A rabbit polyclonal antibody (Synaptic System) coupled to Dynabeads (Invitrogen) in immunoprecipitation buffer (10 mM Tris–HCl, 150 mM NaCl, 0.1% Igepal CA‐630 and 400 U RNasin Plus RNase inhibitor) at 4°C for 2 h, followed by elution with free m⁶A (Sigma‐Aldrich) at 4°C for 1 h. The input and immunoprecipitated (IP) samples were reverse transcribed and determined by qPCR analysis. The sequences of primers used are presented in Table EV1.

RNA immunoprecipitation‐qPCR (RIP‐qPCR)

RIP‐qPCR analysis was performed as described in our previous study (Wang et al, 2019). Briefly, HA‐FTO or FLAG‐YTHDC2 overexpressed cells pellets were lysed in lysis buffer of 150 mM KCl, 10 mM HEPES, 2 mM EDTA, 0.5% NP‐40, 0.5 mM dithiothreitol (DTT), 1× Protease Inhibitor Cocktail, and RNasin Plus RNase inhibitor for 30 min at 4°C. The lysates were centrifuged, and the supernatant was transferred to pass through a 0.45‐μm membrane syringe filter. A small aliquot of lysate was saved as input, and the remaining sample was incubated with HA antibody‐conjugated Protein A/G beads or anti‐FLAG M2 magnetic beads (Sigma‐Aldrich) for 4 h at 4°C. Then, the beads were eluted in wash buffer containing 0.1% SDS and 10 ml proteinase K (Invitrogen) and incubated at 55°C for 30 min. The input and immunoprecipitated RNAs were isolated by TRIzol reagent and were reverse transcribed into cDNA using M‐MLV reverse transcriptase. The fold enrichment was detected by qPCR.

Polysome fraction analysis

Polysome fraction analysis was performed according to previously report with minor modifications (Garre et al, 2012). 0.1 mg/ml cycloheximide (CHX) was added to the cell medium for 10 min at 37°C before collection. Cells were rinsed in cold PBS with 0.1 mg/ml CHX and quickly frozen in liquid nitrogen before lysis. Cells were harvested in lysis buffer containing 20 mM HEPES, 100 mM KCl, 5 mM MgCl₂, 1% Triton X‐100, and 0.1 mg/ml CHX with freshly added 1:100 protease inhibitor (Beyotime Biotechnology) and 40 U/ml SUPERase in RNase Inhibitor (Ambion). Lysates were cleared by centrifugation at 12,000 g for 20 min at 4°C, after which the supernatant was collected and add Turbo DNase (Ambion) for 15 min at room temperature. 10–50% (w/v) sucrose density gradients were freshly made in SW41 Ti ultracentrifuge tubes (Backman) using Gradient Master (BioComp Instruments). 500 µl of lysates was loaded onto sucrose gradients followed by ultracentrifugation for 2.5 h at 272,000 g, 4 °C in a SW41 Ti rotor. Gradients were then fractionated from the bottom and approximately thirty 0.3 ml samples were collected. The A260 of samples were detected by Nanodrop to determine the 40–80S or polysome fraction. The RNA of each fraction was isolated from 3‐fold volume of TRIzol reagent. The reversed cDNA was analyzed by qPCR.

Western blot analysis

Protein extracts from cells and tissue were prepared with lysis in RIPA buffer (Beyotime Biotechnology) supplemented with protease and phosphatase inhibitor cocktail (Beyotime Biotechnology) and centrifuged at 13,000 g for 15 min at 4°C. The protein concentration was detected by BCA assay kit (Beyotime Biotechnology). Protein samples were separated by SDS–PAGE and transferred to PVDF membrane (Millipore). The membrane was blocked with 5% non‐fat milk at room temperature and then incubated with relevant primary antibodies at 4°C overnight, followed by HRP‐conjugated secondary antibody. After chemiluminescent reaction, blots were visualized with BeyoECL Star (Beyotime Biotechnology). Primary antibodies’ information is listed in Table EV2.

Statistics

All data were presented as the mean ± SD. Significance test between groups was performed by using Student’s t‐test or ANOVA (for comparison of three or more experimental conditions), using GraphPad Prism 6 software. Significance was established at P < 0.05. Sequence data were downloaded from the Gene Expression Omnibus (GEO) database, accession # GSE53244 (see https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE53244). For data analysis, read sequences were filtered by the FastQC (v0.11.9) and then aligned to the mouse genome (mm10) using TopHat (v2.1.1). HTSeq v0.6.1 was used to count the read numbers mapped to each gene. Differential expression genes were identified by the DESeq2 R package (v1.3.0) using a P < 0.05 and fold‐change > 1.5. Gene ontology (GO) enrichment analysis of differentially expressed genes was performed using the DAVID (https://david.ncifcrf.gov/). GO terms with P < 0.05 were considered as statistically significant. The m⁶A peaks of m⁶A‐immunoprecipitation were identified by MACS2 (v2.2.4) peak‐calling software with the corresponding input sample serving as control. MACS2 was run with default options except for–no model; –keep up all to turn off fragment size estimation and to keep all uniquely mapping reads, respectively. A stringent cutoff threshold for false discovery rate (FDR) < 0.05 was used to obtain high‐confidence peaks. Integrative Genomics Viewer (IGV) was used to visualize the distributions of the m⁶A peaks.

Author contributions

RW, YC, YL, WC, BZ, XL, and GG perform experiments under the supervision of XW. Y.L. performed bioinformatics analysis of sequence data. RW wrote the manuscript under the supervision of XW. LZ, YW, and XW designed the project and XW provided the final approval of the manuscript.

Conflict of interest

The authors declare that they have no conflict of interest.

Supporting information

Expanded View Figures PDF

Table EV1

Table EV2

EMBO reports (2021) 22: e52348.

Data availability

This study includes no data deposited in external repositories.