METTL3 is essential for postnatal development of brown adipose tissue and energy expenditure in mice

Brown adipose tissue (BAT) undergoes rapid postnatal development and then protects against cold and obesity into adulthood. However, the molecular mechanism that determines postnatal development and maturation of BAT is largely unknown. Here we show that METTL3 (a key RNA methyltransferase) expression increases significantly in interscapular brown adipose tissue (iBAT) after birth and plays an essential role in the postnatal development and maturation of iBAT. BAT-specific deletion of Mettl3 severely impairs maturation of BAT in vivo by decreasing m6A modification and expression of Prdm16, Pparg, and Ucp1 transcripts, which leads to a marked reduction in BAT-mediated adaptive thermogenesis and promotes high-fat diet (HFD)-induced obesity and systemic insulin resistance. These data demonstrate that METTL3 is an essential regulator that controls iBAT postnatal development and energy homeostasis.

O besity and its associated diseases affect billions of people around the world. Obesity results from an imbalance between energy intake and energy expenditure. Increasing energy expenditure is an efficient way to treat obesity [1][2][3] . Brown adipose tissue (BAT) dissipates the mitochondrial electrochemical gradient to generate heat through uncoupling pro-tein1 (UCP1) 1,4 . It has been shown that adult humans also have functional UCP1-positive brown adipocytes [5][6][7] , and activation of BAT by cold stimuli leads to a decrease in fat mass 8 . Brown adipocytes undergo postnatal development to mature and gain function 9,10 . Thus, identification of targetable factors that promote postnatal development and function of BAT is an attractive strategy for treating obesity.
It has been demonstrated that PRDM16 and PPARγ are early key transcriptional factors in the fate-determination of brown fat [10][11][12][13] . In addition, PGC-1α expression is induced by cold challenge or by β3-adrenergic agonists 14,15 . PGC-1α then triggers mitochondrial biogenesis, oxidative metabolism, and thermogenesis by inducing expression of Ucp1 and many other genes 14,15 . Whether their mRNA N6-methyladenosine (m 6 A) modification regulates BAT development and energy expenditure is largely unknown. m 6 A is one of the most prevalent mRNA modifications in eukaryotes 16 . RNA m 6 A modification can be catalyzed by m 6 A writer proteins (METTL3/METTL14/WTAP) 17,18 , recognized by m 6 A reader proteins (YTHDF1-3) 19,20 , and removed by m 6 A eraser proteins (FTO and ALKBH5) 21,22 . Recent studies have demonstrated that m 6 A modification regulates most RNA processing steps, including mRNA splicing 23 , mRNA stability 24 , translation efficiency 25 , microRNA maturation 26 , and X chromosome imprinting 27 by m 6 A writer, reader, and eraser proteins, which further modulate a variety of biological processes such as circadian rhythms 28 , DNA damage response 29 , stem cell differentiation 30,31 , and white fat cell differentiation 32,33 . Methyltransferase-like 3 (METTL3), a key RNA methyltransferase, has been demonstrated to regulate neurogenesis 34 , spermatogenesis 35,36 , early embryonic development 31 , stem cell pluripotency in mice 30,31 , and white fat cell differentiation in vitro 37 . However, whether METTL3-mediated m 6 A modification of mRNA regulates interscapular brown adipose tissue (iBAT) postnatal development and thermogenesis is not investigated.
Here, we report that METTL3 is an essential iBAT-enriched RNA methyltransferase and controls iBAT postnatal development and maturation by regulating m 6 A modification and expression of Prdm16, Pparg, and Ucp1 transcripts. BAT-specific knockout of Mettl3 leads to a marked reduction in BAT-mediated adaptive thermogenesis and further promotes high-fat diet (HFD)-induced obesity and systemic insulin resistance. These data demonstrate that METTL3 is an essential regulator that controls iBAT postnatal development and energy homeostasis.

Results
METTL3 is enriched in iBAT and associated with postnatal development of iBAT. First, we observed that METTL3 was highly enriched in interscapular BAT (iBAT) compared with that in inguinal white adipose tissue (iWAT) and epididymal white adipose tissue (eWAT) (Fig. 1a), indicating that METTL3 may regulate the function of iBAT. Next, to determine whether METTL3 is associated with the postnatal development of iBAT, we measured METTL3 expression in iBAT at different ages after birth. METTL3 protein levels were dramatically increased in iBAT at 1 day of age, reached the maximal level at 5 days of age, and began a slight decline after 10 days of age (Fig. 1b). As expected, the UCP1 levels were highly induced in iBAT at 1 day of age, reached the maximal level between 10 and 20 days of age, and began a slight decline after 20 days of age (Fig. 1b). These data demonstrate that METTL3 expression shows a similar expression pattern with UCP1 in iBAT during early development after birth, indicating that METTL3 may have an important role in the postnatal development of iBAT.
METTL3 is essential for postnatal development of iBAT. To further determine whether METTL3 regulates iBAT postnatal development, we generated BAT-specific Mettl3 knockout (BKO) mice by crossing Mettl3-floxed mice ( Supplementary Fig. 1a-c) with Ucp1-iCre transgenic mice, in which IRES-Cre was inserted between exon 6 and the 3′ untranslated region (UTR) to allow Ucp1 and iCre expression at the same time with lower levels 38 . Ucp1 expression is induced after 18 days of gestation and shows a rapid increase until 10 days after birth 9 . Thus, Ucp1-iCre is supposed to delete genes in iBAT shortly after birth. METTL3 protein levels in iBAT of BKO mice began to decline at 5 days of age (see below). METTL3 was specifically deleted in iBAT but not in other tissues such as eWAT, liver, and skeletal muscle in 8week-old BKO mice ( Supplementary Fig. 1d). We also noted that METTL3 was highly enriched in iBAT compared with that in eWAT ( Supplementary Fig. 1d). We did not observe any difference in body weight ( Supplementary Fig. 2a), iBAT morphology ( Supplementary Fig. 2b), iBAT weight ( Supplementary Fig. 2c), or cold challenge ( Supplementary Fig. 2d) between Mettl3 flox/flox and Ucp1-iCre mice. Therefore, we used Mettl3 flox/flox mice as the control for BKO mice in the following experiments. Surprisingly, the morphology of iBAT in BKO mice appeared abnormal, enlarged and "whitening" roughly after 5 days of age (Fig. 2a-c). Consistently, the weight of iBAT in BKO mice was significantly higher than that of flox/flox mice (Fig. 2d), which did not affect the body weight gain of BKO mice during postnatal development (Fig. 2e). The enlarged iBAT was attributable to large cytosolic lipid droplet accumulation, resulting in an increase in average adipocyte size (steatotic hypertrophy) (Fig. 2c) postnatal development in iBAT of flox/flox mice, whereas its induction was blocked in iBAT of BKO mice (Fig. 2f). The blockage of UCP1 induction in iBAT of BKO mice during postnatal development was associated with the decline of METTL3 in BKO mice after 5 days of age (Fig. 2f). These data indicate that METTL3 is necessary for postnatal development of iBAT.
BAT-specific deletion of Mettl3 results in dramatically decreased expression of BAT-selective genes. To further explore the molecular mechanisms of the impaired postnatal development of iBAT in BKO mice, we examined the whole transcriptional profiles of iBAT in both BKO and flox/flox mice by performing RNA sequencing (RNA-seq) analysis. As shown in Fig. 3a, 530 genes were downregulated, and 1552 genes were upregulated. Gene Ontology analysis showed that genes related to developmental maturation, respiratory electron transport chain, adaptive thermogenesis, and energy deprivation were dramatically downregulated, whereas genes associated with inflammation, muscle system process, and muscle cell development were significantly upregulated (Fig. 3b). Quantitative PCR analysis further supported these RNA-seq data. The general markers of iBAT, such as adiponectin, aP2, Pparg, Glut4, and Cebpα, were  Striated muscle development Enrichment score (-log 10 (P-value)) Enrichment score (-log 10 (P-value)) Striated muscle cell differentiation  dramatically decreased in BKO mice (Fig. 3c). Lipogenesis-, lipolysis-, and fatty acid oxidation-related genes, including Srebp1, Fasn, ATGL, MgII, Pparα, Cpt2, and Cpt1b, were also decreased in BKO mice (Fig. 3c). Thermogenic genes, including Pgc-1α, Ucp1, Prdm16, Adrb3, and Cidea, were significantly downregulated (Fig. 3c). UCP1, PGC-1α, PPARγ, and PRDM16 protein levels were also dramatically reduced in iBAT of BKO mice (Fig. 3d). Decreased PGC-1α levels also resulted in reduced mitochondrial number (Fig. 3e). In addition, most of genes related to mitochondrial oxidative phosphorylation (OXPHOS), including the components of complexes I, II, III, IV, and V, were dramatically reduced in BKO mice (Fig. 3c, f). Interestingly, the genes associated with muscle system process and muscle cell development, including Acta1, Casq2, Tnnc1, Myoglobin, MHC-1β, MHC-IIα, and MEF2c, were dramatically increased in iBAT of BKO mice (Fig. 3c), indicating that BAT-specific deletion of Mettl3 may promote the brown fat-myoblast conversion. These data also suggest that BAT-specific deletion of Mettl3 impairs postnatal development of iBAT by decreasing expression of BATselective genes. Lipid uptake has an important role in lipid accumulation and BAT function 39,40 . CD36 is the main fatty acid transporter in iBAT 39 . The lipoprotein lipase (LPL) and its inhibitor angiopoietin-like protein 4 (ANGPTL4) in iBAT controls coldinduced triglyceride uptake in BAT 39,40 . We measured the expression of Cd36, Lpl, and Angptl4. As shown in Supplementary   Fig. 3, their mRNA levels were not altered in the iBAT of BKO mice. These data indicate that lipid uptake is not changed in the iBAT of BKO mice, which less likely contributes to the whitening of iBAT in BKO mice.
METTL3 is necessary for m 6 A modification and expression of Prdm16, Pparg, and Ucp1 transcripts. To further determine whether METTL3 regulates m 6 A modification of mRNA related to brown fat differentiation and thermogenesis, we performed m 6 A RNA immunoprecipitation sequencing (m 6 ARIP-seq) analysis in iBAT of BKO and control mice. Each sample was pooled from five mice for each group. Consistent with published m 6 ARIP-seq results 41 , the m 6 A peaks identified in iBAT of flox/flox control mice were enriched at stop codon and 3′-UTR and were characterized by the canonical GGACU motif (Fig. 4a, b). However, the m 6 A peaks in iBAT of BKO mice were dispersed at transcription start site, 5′-UTR, and stop codon, and the enriched motif was CCAUG (Fig. 4a, b). In the iBAT of flox/flox mice, we identified about 9989 significant m 6 A peaks (false discovery rate < 0.05) in~6845 transcripts ( Fig. 4c and Supplementary Fig. 4). There were 5383 transcripts exhibiting decreased m 6 A levels in the iBAT of BKO mice ( Fig. 4c and Supplementary Fig. 4). Importantly, m 6    To further test whether the m 6 A modification at last exon and 3′-UTR of Prdm16, Pparg, and Ucp1 transcripts were associated with the expression of Prdm16, Pparg, and Ucp1. We cloned the region into pMIR-REPORT Luciferase vector and performed luciferase assays in primary brown preadipocytes. Knockdown of Mettl3 by two individual siRNAs (siMettl3-1 or siMettl3-2) in primary brown preadipocytes resulted in a significant reduction of METTL3 protein levels (Fig. 5g). Either siMettl3-1 or siMettl3-2 significantly decreased Prdm16, Pparg, and Ucp1 luciferase activity, respectively (Fig. 5h-j). To identify which m 6  . g-j Primary brown preadipocytes seeded in 24-well plates were co-transfected with pMIR-REPORT Luciferase vectors (Prdm16, Pparg, Ucp1), siRNAs (Scramble siRNA, siMettl3-1, siMettl3-2) and β-galactosidase (β-Gal) reporter plasmid by X-tremeGENE siRNA Transfection Reagent for 24 h. Cells were then induced to differentiate for 48 h. METTL3 and GAPDH protein levels were measured by immunoblotting g. Relative Prdm16, Pparg, and Ucp1 luciferase activity were measured and normalized to β-Gal levels h-j (Scr in h, n = 4; others, n = 5 biologically independent cell samples for each group). Data represent the mean ± SEM. Significance was determined by unpaired two-tailed Student's t test analysis. *p < 0.05. **p < 0.01. Source data are provided as a Source Data file.
were responsible for this regulation, we performed siRNAmediated m 6 A reader proteins (Ythdf1, Ythdf2, or Ythdf3) knockdown and luciferase assays. As shown in Supplementary  Fig. 5a-h, knockdown of Ythdf1 by two individual siRNAs (siYthdf1-1 or siYthdf1-2) did not affect the luciferase activity of Prdm16, Pparg, or Ucp1, whereas knockdown of Ythdf2 by two individual siRNAs (siYthdf2-1 or siYthdf2-2) significantly reduced their luciferase activity. In addition, knockdown of Ythdf3 by siYthdf3-2 but not siYthdf3-1 significantly decreased Prdm16 and Ucp1 luciferase activity ( Supplementary Fig. 5i-l). Moreover, the expression of Ythdf2 and Ythdf3 were significantly increased in mature adipocytes compared with preadipocytes ( Supplementary  Fig. 5m-o), indicating that YTHDF2 and YTHDF3 may regulate brown adipocyte differentiation and BAT development. These data also indicate that METTL3 and YTHDF2/3 may coordinate with each other and regulate BAT development.
METTL3 is essential for brown adipogenesis in vitro. Mettl3 mRNA levels were significantly higher in mature primary brown adipocytes compared with preadipocytes ( Fig. 6a), suggesting that METTL3 may directly regulate differentiation of brown adipocytes. To further determine whether METTL3 directly regulates the differentiation of brown adipocytes, we harvested the stromal-vascular fraction from the brown fat pads of Mettl3 flox/flox mice. Primary brown preadipocytes were infected with Ad-βGal and Ad-Cre adenovirus and differentiated to mature brown adipocytes. As expected, Cre adenovirus infection caused the deletion of Mettl3 in primary brown adipocytes (Fig. 6b, f). Deletion of Mettl3 significantly impaired the differentiation of precursor cells, as revealed by decreased Oil Red O staining (Fig. 6c) and the expression of general brown adipocyte markers (Fig. 6d). In addition, deletion of Mettl3 in primary brown adipocytes significantly decreased the expression of genes involved in differentiation (Prdm16 and Pparg), thermogenesis (Pgc-1α, Ucp1, Adrb3, and Cidea), lipogenesis (Srebp1 and Fasn), lipolysis (ATGL, HSL, and MgII), fatty acid oxidation (Ppara, Cpt2, and Cpt1b) and oxidative phosphorylation (Cox5b, Ndufa8, Cox7a1, and Cox6b1) (Fig. 6d, e). UCP1, PGC-1α, PPARγ, and PRDM16 protein levels were also dramatically reduced in primary METTL3-deficient brown adipocytes (Fig. 6f). Consistent with decreased expression of PGC-1α, the relative mitochondria number was dramatically decreased by 33% in METTL3 knockout brown adipocytes (Fig. 6g). These results suggest that METTL3 is necessary for brown adipogenesis in vitro.
BAT-specific Knockout of Mettl3 results in decreased energy expenditure. Next, we tested the need for METTL3 in thermogenesis in vivo. Mettl3 mRNA levels were significantly increased after acute cold exposure (4°C for 6 h) (Fig. 7a), indicating that METTL3 may promote thermogenesis. BKO mice displayed lower oxygen (O 2 ) consumption and CO 2 production rates during both light and dark cycles (Fig. 7b-e), with similar amounts of food intake and physical activity compared with flox/flox controls ( Supplementary Fig. 6a, b), indicating that energy expenditure was decreased in BKO mice. Furthermore, BKO mice were less able to defend against acute cold exposure (Fig. 7f). In addition, serum free fatty acid levels were much higher in BKO mice after cold challenge (Fig. 7g), which further supports impaired thermogenesis in BKO mice. The morphology of iBAT in BKO mice under the cold exposure condition still appeared abnormal, enlarged and "whitening" (Supplementary Fig. 7a-c). These data suggest that METTL3 in BAT is necessary for thermogenesis. METTL3 was exclusively expressed in iBAT, and BAT-specific deletion of Mettl3 impaired acute cold-induced thermogenesis, indicating that METTL3 may regulate browning of WAT in response to chronic cold exposure or the β-adrenergic agonist. To further test this hypothesis, BKO mice and flox/flox controls were exposed for chronic cold challenge (4°C 7d) or injected with CL-316,243 for 4 days to induce browning of WAT. At room temperature, the weight of iWAT was similar between BKO and flox/flox mice (Supplementary Fig. 8a, c), and the thermogenesis related genes such as Ucp1 and Pgc-1α were not changed in iWAT of BKO mice ( Supplementary Fig. 8b, d). However, after either chronic cold exposure or multiple injection of CL-316,243, the weight of iWAT showed higher in BKO mice, and the expression of Ucp1 and Pgc-1α were downregulated under both conditions ( Supplementary Fig. 8a-d). These data indicate that BAT-specific deletion of Mettl3 impairs the browning of WAT in response to chronic cold exposure or the β-adrenergic agonist.
BAT-specific knockout of Mettl3 predisposes to high-fat dietinduced obesity. Reduced BAT thermogenesis contributes to obesity in both rodents and humans 42,43 . We next determined whether BAT Mettl3 expression was associated with obesity by measuring BAT Mettl3 expression in two obese mice models (high-fat diet-induced obese (DIO) and leptin-deficient ob/ob mice). METTL3 protein levels were lower in both DIO and ob/ob mice (Fig. 8a). Mettl3 mRNA levels were also significantly reduced in both DIO and ob/ob mice (Fig. 8b, c). These data suggest that METTL3 in iBAT might regulate energy metabolism and obesity.
To determine whether BKO mice are sensitive to HFD-induced obesity, BKO and control mice were fed with HFD (45% fat), and body weight was measured weekly. As shown in Fig. 8d, BKO mice gained more body weight than flox/flox mice after feeding with HFD. Fat mass was significantly increased in BKO mice, whereas the lean mass did not change (Fig. 8e). BKO mice under a HFD feeding condition also showed severely impaired cold tolerance (Fig. 8f). The relative weight of iBAT and liver was much higher in BKO mice (Fig. 8g). Consistently, the size of iBAT and liver were larger in BKO mice (Fig. 8h). Furthermore, hematoxylin and eosin staining showed adipocyte hypertrophy and hepatosteatosis in BKO mice (Fig. 8i). Serum triglyceride levels were significantly increased in BKO mice (Fig. 8j). These data demonstrate that BKO mice were more prone to HFDinduced obesity.
Obesity leads to impaired glucose metabolism and insulin resistance. To determine whether BAT METTL3 regulates systemic glucose homeostasis and insulin resistance, we performed glucose tolerance tests and insulin tolerance tests on mice fed with HFD for 8 weeks (16 weeks old). As shown in Fig. 9a, b, BKO mice displayed glucose intolerance and insulin resistance. Serum insulin levels in BKO mice were also higher (Fig. 9c), further supporting that BKO mice showed insulin resistance. Moreover, insulin-induced p-AKT(S473) levels were significantly decreased in the livers of BKO mice (Fig. 9d). These data suggest that decreased expression of Mettl3 in iBAT impairs systemic energy homeostasis, contributing to HFD-induced obesity and metabolic syndrome.

Discussion
In recent years, much progress has been made in identifying transcriptional factors and cofactors in the regulation of brown cell differentiation and thermogenesis. However, whether mRNA m 6 A modification regulates iBAT development and thermogenesis is largely unknown. In this study, we demonstrated that METTL3 is an essential iBAT-enriched RNA methyltransferase and controls postnatal development of iBAT and thermogenesis A key question is how METTL3 regulates the postnatal development of a brown fat gene program. METTL3 is well known for RNA m 6 A modification. It has been reported that METTL3 regulates stem cell differentiation by promoting RNA m 6 A modification of key transcriptional factors 44 . METTL3 also has been linked to white fat cell differentiation 37  decreases iBAT-specific mRNA m 6 A modification and expression, including Prdm16, Pparg, and Ucp1 transcripts. One target of METTL3 is Prdm16, which has been shown to drive the differentiation of brow fat cells and browning of WAT by activating the expression of BAT-selective genes including Ucp1 and Pgc-1α 11,45,46 . Similarly, BAT-specific deletion of Mettl3 impaired BAT development and also impaired the development of beige adipocytes in iWAT. PRDM16 also controls the myoblast/brown fat switch 47 . Knockdown of Prdm16 downregulates BAT-selective genes and upregulates skeletal muscleassociated genes 11,47 . Likewise, BAT-specific deletion of Mettl3 also downregulates BAT-selective genes and upregulates skeletal muscle-associated genes, which is likely owing to decreased PRDM16 expression. Mechanistically, BAT-specific deletion of Mettl3 decreases m 6 A peaks of the Prdm16 transcript, which may be associated with decreased Prdm16 mRNA levels with the involvement of YTHDF2/3. These data indicate that METTL3 regulates iBAT postnatal development at least partially by regulating m 6 A modification and expression of the Prdm16 transcript. However, BAT-specific Prdm16 knockout mice show normal postnatal development of iBAT at a young age and display "whitening" and enlarged iBAT in old age 10 , whereas Mettl3 BKO mice show severely impaired postnatal development of iBAT at 5 days of age, suggesting that METTL3 in iBAT has certain actions that probably do not occur solely through modulation of Prdm16 expression.
Another target of METTL3 is Pparg, which has an essential role in adipocyte differentiation. Pparg-deficient mice display no brown or white adipose tissues 48,49 , suggesting that PPARγ is necessary for development of both BAT and white adipose tissue. BKO mice show significantly decreased m 6 A modification and expression of the Pparg transcript, which contributes to the impaired postnatal development of iBAT.
Another important target of METTL3 is Ucp1, which is a key thermogenic factor in iBAT. During postnatal development of iBAT, UCP1 and METTL3 protein levels are significantly increased. BAT-specific deletion of Mettl3 decreases m 6 A peaks and expression of the Ucp1 transcript, which contributes to impaired thermogenesis. However, Ucp1 knockout mice do not show enlarged iBAT 50 , suggesting that BAT-specific deletion of Mettl3 is less likely to result in enlarged iBAT solely through modulation of Ucp1 expression. Thus, it is possible that METTL3 acts as an essential regulator of postnatal development of iBAT and thermogenesis at least partially by controlling m 6 A modification and expression of Prdm16, Pparg, and Ucp1 transcripts simultaneously. However, we cannot rule out other possible mechanisms because METTL3 regulates m 6 A modification and expression of many other BAT-selective genes, based on the m 6 ARIP-seq and RNA-seq data.
METTL3-mediated mRNA m 6 A modification can be recognized by the individual m 6 A reader proteins, which play key roles in controlling gene expression. Early studies showed that YTHDF1 binding promotes translational efficiency 51 , YTHDF2binding increases mRNA decay 19 , and YTHDF3 can assist the effects of YTHDF1 and YTHDF2 52,53 . However, other roles for these reader proteins are rapidly emerging in the regulation of specific mRNAs in different types of cells during different stimuli 29,54,55 . In addition, the profiles and function of METTL3mediated mRNA m 6 A are different depending on tissue and developmental stage 41,56,57 . For example, hypoxia increases m 6 A content of mRNA and enhances the stability of m 6 A modified mRNAs 24 , raising the possibility that m 6 A modification could also stabilize transcripts. In iSLK.219 cells, both METTL3 and YTHDF2 appear to promote lytic viral gene expression during Kaposi's sarcoma-associated herpesvirus infection 58 . Another study shows that YTHDF2 also recognizes multiple m 6 A sites in the SV40 late transcripts and strongly promotes SV40 replication 59 . Our study shows that the expression of Mettl3 and Ythdf2 are increased during brown adipocyte differentiation. BATspecific deletion of Mettl3 reduces m 6 A modification and expression of Prdm16, Pparg, and Ucp1 transcripts, indicating that m 6 A modification is required for the expression of Prdm16, Pparg, and Ucp1 transcripts in iBAT. Knockdown of either Mettl3 or Ythdf2 decreases the luciferase activity of Prdm16, Pparg, and Ucp1 in primary brown preadipocytes, indicating that METTL3 and YTHDF2 may play a similar role in brown adipogenesis. Future studies should clarify the function and the molecular mechanisms of YTHDF2 in the regulation of BAT development.
In addition, m 6 A pathway has been shown to regulate white adipocyte differentiation 32,33,37,60,61 . The well-studied protein in m 6 A pathway is FTO, which was previously identified by human genome-wide associated studies with strong association with obesity 62 . FTO promotes white adipocyte differentiation in vitro and development of obesity in vivo 32,61 , whereas deletion of FTO impairs white fat adipogenesis both in vitro and in vivo 32,33,60,63 . The potential molecular targets include RUNX1T1, ATG5/7, and JAK2 32,33,60,61 . Interestingly, knockdown of m 6 A writer proteins (METTL3, METTL14, or WTAP) in 3T3-L1 cells results in cell cycle arrest and impaired adipogenesis by suppressing expression of cyclin A2 37 . These results indicate that m 6 A eraser and writer proteins play similar roles in white fat adipogenesis although their targets are different. With respect to WAT browning, FTO deficiency has been shown to induce UCP1 expression in white adipocytes 64 , whereas our study shows that BAT-specific deletion of Mettl3 impairs the browning of WAT in response to chronic cold exposure or the β-adrenergic agonist, indicating that m 6 A eraser and writer proteins may play opposite roles in regulation of WAT browning and BAT development. Further investigations are necessary to fully explore the mechanisms of RNA m 6 A pathway (including other writer, reader, and eraser proteins) in controlling postnatal development of iBAT and energy metabolism.
In conclusion, we have shown that METTL3 is an essential BAT-enriched RNA methyltransferase and controls postnatal development of iBAT by regulating m 6 A modification and expression of Prdm16, Pparg, and Ucp1 transcripts. BAT-specific knockout of Mettl3 leads to a marked reduction of BAT-mediated adaptive thermogenesis and results in obesity and systemic insulin resistance. These data demonstrate that METTL3 is an essential regulator that controls postnatal development of iBAT and energy homeostasis.

Methods
Animal experiments. Animal experiments were carried out in strict accordance with the Guide for the Care and Use of Laboratory Animals published by the US National Institutes of Health (NIH publication no. 85-23, revised 1996) and approved by the Institutional Animal Care and Use Committee or Animal Experimental Ethics Committee of Harbin Institute of Technology (HIT/IACUC). The permit number is IACUC-2018002. Mice were housed under controlled light (12-h light/12-h dark cycle), temperature (24 ± 2°C) and humidity (50 ± 10%) conditions and fed a normal chow diet with a free access to water. For diet-induced obesity studies, mice were fed with an HFD (MD12032, 45% fat, Medicience). Mettl3 flox/flox mice, in which the exon 2 and exon 3 of Mettl3 gene was flanked by two loxp sites, were generated by using CRISPR-Cas9 technique ( Supplementary  Fig. 1A). Southern blot data showed that the homologous recombination in Mettl3 flox/+ mice were correct (Supplementary Fig. 1B). Ucp1-iCre mice, in which IRES-Cre was inserted between exon 6 and the 3′-UTR to allow Ucp1 and iCRE expression at the same time with lower levels, have been shown previously 38 . BATspecific Mettl3 knockout mice were generated by crossing Mettl3 flox/flox mice with Ucp1-iCre mice.
Body composition and energy metabolism measurement. Body weight was measured weekly. Body composition (fat and lean mass) was determined by a body composition analyzer (Brucker Minispec LF50). For metabolic studies, mice were housed individually in metabolic cages (Promethion, Sable Systems, LasVegas, NV), and free access to food and water. Oxygen consumption and CO 2 production rates were monitored for 72 hr. Food intake and physical activity monitoring were performed simultaneously with metabolic measurements. Data were collected and analyzed by MetaScreen-Data Collection Software (V2.3.15) and Expedata-P Data Analysis Software (V1.9.17), respectively.
Glucose tolerance tests and Insulin tolerance tests. For glucose tolerance test experiment, mice fasted for 6 hr were injected intraperitoneally with D-glucose (1 g/ kg). For insulin tolerance test experiment, mice fasted for 6 h were injected intraperitoneally with human insulin (Lily) (1 U/kg). Blood glucose levels were measured from the tail vein at indicated times using a glucometer 65 . Blood samples were collected from orbital sinus. Serum insulin levels were measured using insulin ELISA kits (MS100, EZassay).
Cold-stress experiments. For cold exposure experiments, individual mouse was placed in a single cage in a cold room (4°C) with free access to water. Control mice were kept at room temperature (22°C). The core body temperature was monitored using a rectal probe (7001HT, Phyritemp) at each time point. For chronic cold exposure, mice were first treated with 3 days of the intermittent cold exposure (12 h at 4°C in the light cycle and 12 h at room temperature), and then followed by 4 days of continued cold exposure at 4°C 66,67 .
Chronic CL-316,243 treatment. BKO and flox/flox mice were injected intraperitoneally with CL-316,243 at 1 mg/kg body weight or equal volume of saline daily for 4 days. Mice were killed on day 5 without addition injection.
Quantitative Real-Time PCR (qPCR). Total RNA was isolated using TriPure Isolation Reagent (94015120, Roche), and the first-strand cDNAs were synthesized using Random Primers and M-MLV reverse transcriptase (M1701, Promega) 65,68 . RNA abundance was measured using SYBR Green Mixs (4913914001, Roche) and Roche LightCycler 480 real-time PCR system (Roche, Mannheim, Germany). The expression of individual genes was normalized to the expression of 36B4, a housekeeping gene. Primers for real-time qRT-PCR were listed in Supplementary  Table 1.
Quantification of mtDNA copy number. Total DNA was isolated from mouse brown adipose tissue. mtDNA was amplified using primers specific for the mitochondrial mtND1 gene and normalized to genomic DNA by amplification of the LPL nuclear gene. The primer sequences can be found in Supplementary Table 1.
Primary brown adipocytes culture and adenovirus infection. The interscapular brown fat pad was dissected from 4-5-week-old Mettl3 flox/flox or C57BL/6 wildtype mice, minced, and then digested for 20-30 min at 37°C in PBS containing 10 mM CaCl2, 1.5 mg/ml Collagenase type II, and 1.4 U/ml Dispase II. Digested tissue was filtered through a 100 µm cell strainer to remove large pieces, and then centrifuged for 10 min at 1000 g to pellet the stromal-vascular fraction (SVF) cells. SVF cells were resuspended in complete culture medium (DMEM with 10% FBS and Pen/ Strep), and then plated on collagen-coated 24-well plates. For preadipocyte differentiation, cells grown to 100% confluence (Day 0) were exposed to induction in DMEM containing 2 μg/mL dexamethasone, 1 μM insulin, 0.5 mM isobutylmethylxanthine, 1 μM rosiglitazone, 1 nM T3, 62.5 μM indomethacin, and 10% FBS. Three days after induction (from Day 3), cells were maintained in media containing 1 μM insulin, 1 nM T3 and 10% FBS until ready for harvest (generally day 6-7 post differentiation). All chemicals for cell culture were obtained from Sigma-Aldrich. For adenoviral infection of primary SVF cells, 100% confluent cells were infected with Cyclization Recombination Enzyme (Cre) or βGal-expressing adenovirus in growth medium overnight. The medium was then switched to induction medium for 72 hr to induce adipogenic differentiation, and then cells were maintained in differentiation medium until ready for harvest. For oil-red staining, infected cells were maintained in differentiation medium for 3 days. For RT-qPCR and Western blot, infected cells were maintained in differentiation medium for 1 day.