Activation of METTL3 Promotes White Adipose Tissue Beiging and Combats Obesity

Adipose tissue is a central metabolic organ, the dysfunction of which disrupts energy homeostasis and induces the pathogenesis of metabolic disorders, such as obesity and type 2 diabetes (1). Three different types of adipocytes have been identified: white, brown, and beige (2). White adipocytes are responsible for fat storage, whereas beige and brown adipocytes mediate adaptive thermogenesis by expressing uncoupling protein 1 (UCP1) (2,3). Although beige and brown adipocytes share similar thermogenic features, they have distinct origins and molecular identities (4). Beige adipocytes originate from white adipose tissue (WAT), a process that could be induced by various environmental stimuli, such as cold exposure, exercise, and the agonists of peroxisome proliferator–activated receptor-γ (PPARγ) (2,4). The inducible feature of beige adipocytes highlights an attractive way to treat obesity-associated disorders (4,5).

The beiging process relies on the activation of a beige adipocyte–specific gene program, which is regulated by several transcriptional and/or epigenetic factors (5). Transcriptional regulators, such as PPARγ, C/EBPβ, PGC1α, and PRDM16, lie at the center of the beige adipocyte fate decision (5,6). In addition to transcriptional regulation, gene expression could be regulated at the posttranscriptional level, but the posttranscriptional regulation of the beiging process remains largely uncharacterized.

N6-methyladenosine (m⁶A), the most abundant modification on mRNAs, has been defined as a new layer of gene expression regulation (7). The formation of m⁶A is catalyzed by the RNA methyltransferase complex METTL3/METTL14/WTAP (8–10), and demethylation is mediated by ALKBH5 or FTO (11,12). The metabolism of m⁶A-modified mRNAs, including mRNA splicing, degradation, and translation, is regulated by reader proteins, including YTH domain proteins (13,14).

Evidence indicates m⁶A has a role in adipocyte metabolism. Epidemiologic studies demonstrate that the single nucleotide polymorphism of FTO is highly correlated with the incidence of obesity in humans (15). Depletion of Fto in mice shows obvious reduction in adipose tissue mass (16). Moreover, Fto deletion in vitro promotes thermogenesis and white-to-beige adipocyte transition (17). In contrast, METTL3 is reported to positively regulate the development and energy expenditure of brown adipose tissue (BAT) (18). Entacapone, a recently identified FTO inhibitor, dramatically enhances thermogenesis in adipose tissue (19). However, whether targeting METTL3 could improve adipose tissue metabolism remains to be explored.

In this study, we demonstrate an essential role of METTL3 in WAT beiging. METTL3-mediated m⁶A installation promotes the stability of mRNAs involved in thermogenesis. Pharmaceutical activation of the METTL3 complex induces WAT beiging and corrects obesity-associated disorders. These results highlight METTL3 as a promoter of the beiging process and a potential target for obesity treatment.

Eight-week-old male mice on the C57BL/6J background were used for experiments unless mentioned otherwise. Mice were fed with either a standard chow (CD) (10% fat as kcal; cat. no. D12450J; Research Diets, Inc., New Brunswick, NJ) or a high-fat diet (HFD) (60% fat-derived calories; cat. no. D12492; Research Diets Inc.) for 8 weeks. Mice containing a loxP-flanked Mettl3 allele (Mettl3^fl/fl) were provided by Ming-Han Tong (Shanghai Institute of Biochemistry and Cell Biology, Chinese Academy of Sciences). Mettl3 adipose tissue conditional knockout mice were generated by mating Mettl3^fl/fl mice with Adipoq-Cre mice (The Jackson Laboratory). The primers used for genotyping are listed in Supplementary Table 1. For cold exposure, mice were individually housed in plastic cages at 4°C for 7 days. After anesthetization with 2% isoflurane, 6-week-old mice were injected with recombinant adeno-associated virus (AAV) (1.0 × 10¹² vg per 20 μL PBS) into inguinal WAT (iWAT) according to the method provided by Zhu et al. (20). For compound treatment, mice were intraperitoneally injected with methyl piperidine-3-carboxylate (MP3C) (Shanghai Aladdin Biochemical Technology) at 2.5 mg/kg every 3 days for 2 weeks. The metabolic parameters of each mouse were monitored using the TSE Systems PhenoMaster (Berlin, Germany). Oxygen (O₂) consumption, carbon dioxide (CO₂) generation, and energy heat generation were recorded for each mouse and normalized to body weight. iWAT, epididymal WAT (eWAT), and interscapular BAT from mice were surgically removed and immediately frozen at −80°C for additional analyses.

Mice were maintained in specific pathogen-free conditions at the Animal Center of Zhejiang University. All animal studies were performed in compliance with the Guide for the Care and Use of Laboratory Animals by the Medical Experimental Animal Care Commission of Zhejiang University. All animal studies used the protocol that has been approved by the Medical Experimental Animal Care Commission of Zhejiang University (AP code ZJU20220513).

For the glucose tolerance test, HFD-fed mice were intraperitoneally injected with 100 mg/mL d-glucose at a dose of 0.8 g/kg body weight after overnight fasting. For the insulin tolerance test, HFD-fed mice were intraperitoneally injected with human insulin at a dose of 1.0 units/kg body weight after 4-h fasting. Glucose level was measured in tail blood at 0, 15, 30, 60, 90, and 120 min after glucose or insulin injection using a glucometer (Accu-Chek; Roche). Core body temperature was measured intrarectally at approximately 4:00 p.m.

O₂ consumption rates (OCRs) were measured using Clark-type oxygen electrodes (Strathkelvin Instruments). Freshly isolated tissue (∼30 mg) was minced in respiration buffer (1.5 mmol/L pyruvate, 25 mmol/L glucose, and 2% BSA) and placed in electrode chambers, and the OCR was detected (21).

3T3-L1 preadipocytes, HEK293T cells, and AAV293 cells were obtained from ATCC and cultured in DMEM medium (HyClone Laboratories, Logan, UT), supplemented with 10% FBS (Thermo Fisher Scientific, Scoresby, Victoria, Australia).

To construct shRNA-expressing plasmids, shRNA oligos were cloned to the lentiviral vector pLKO.1. The coding sequence (CDS) of Krüppel-like factor 9 (Klf9) was cloned into the pAAV-MCS plasmid. To construct and purify FLAG-tagged protein, the CDS of METTL3 or METTL16 tagged at the C terminus with FLAG was cloned into the lentivector pCDH. To construct luciferase reporters, the 3′ end of the Klf9 CDS region was first amplified from cDNA of iWAT and cloned into pGL3-control vector in frame with the luciferase gene (Promega, Madison, WI). The oligos used for plasmid construction are listed in Supplementary Table 1.

The lentivirus was packaged in HEK293T cells with helper plasmids pMD2.G and psPAX2. The AAV was packaged in AAV293 cells with helper plasmids pHelper and RC. After 72-h culture, the supernatant was collected for infection.

Klf9 luciferase reporters were cotransfected with pRL-TK into 3T3-L1 cells in a 12-well plate using Lipofectamine 2000 (Invitrogen, Waltham, MA). Cells were harvested at 24 h posttransfection. Firefly and Rinella luciferase activities were measured using the Dual-Luciferase Assay Kit (Promega). The activities of the luciferase were expressed as normalized relative light units to the Rinella internal control.

HEK293 cells were transfected with METTL3 or METTL16 plasmids using Lipofectamine 2000. METTL3 and METTL16 were isolated by lysis of HEK293 cells at 48 h posttransfection, and the lysate was purified with anti-FLAG M2 magnetic beads (Sigma-Aldrich, St Louis, MO).

The experiments were conducted in reaction buffer (20 mmol/L Tris-hydrochloride [pH 7.5], 1 mmol/L dithiothreitol, 0.01% Triton X-100, and 20 units/100 μL buffer RNaseOUT) (22,23). The reaction mixture contained 2.5 μmol/L unmethylated single-stranded RNA probe with a biotin tag (5′-biotin-CGUCUCGGACUCGGACUGCU-3′; Dharmacon, Inc.), 25 μmol/L S-adenosylmethionine (Yeasen Biotechnology, Shanghai, China), and 50 ng/μL purified METTL3 or METTL16 protein. Enzymatic assay reactions were transferred to streptavidin-coated 96-well plates (Beaver Biomedical Engineering, Suzhou, China) and incubated for 20 h at 21°C on shakers. The plate was incubated with rabbit anti-m⁶A antibody (Synaptic Systems, Göttingen, Germany) for 2 h and incubated for additional 1 h with horseradish peroxidase–conjugated secondary antibody at room temperature with slow shaking. The plate was incubated with TMB (3,3′, 5,5; -tetramethylbenzidine) chromogen solution (Biosharp Life Sciences, Hefei, China) for 30 min. The reactions were stopped with 2 mol/L hydrochloride, and expression levels of m⁶A were detected at 450 nm using Varioskan Flash (Thermo Fisher Scientific).

mRNA was purified from total RNA by using Dynabeads Oligo(dT)₂₅ (Invitrogen) and spotted to a Hybond-N+ membrane (GE Healthcare, Waukesha, WI), followed by ultraviolet crosslinking at ultraviolet 254 nm (0.12 J/cm²). The membrane was blocked with 5% nonfat milk in PBS and then incubated with an anti-m⁶A antibody. The membrane was incubated with horseradish peroxidase–conjugated secondary antibody and visualized by using enhanced chemiluminescence. Methylene blue staining was used as mRNA loading control.

Total RNA from iWAT was extracted with TRIzol reagent. The polyadenylated RNA was enriched using Dynabeads Oligo(dT)₂₅ and fragmented into fragments ∼100 nucleotides long using RNA fragmentation reagent (cat. no. AM8740; Ambion). Fragmented RNA samples were used for library construction and high-throughput sequencing by Hangzhou KaiTai Biotechnology Co., Ltd.

For m⁶A sequencing (m⁶A-seq), fragmented RNA was incubated with m⁶A primary antibody–coated beads for 6 h at 4°C. The immunoprecipitation complex was digested with proteinase K at 55°C for 1 h. RNA was then extracted using TRIzol reagent and used for library construction and high-throughput sequencing.

For ribosome sequencing (ribo-seq), the iWAT of mice was homogenized in polysome lysis buffer. One of 10 clarified lysates was set aside as input. The remaining nine of 10 lysates were digested with Escherichia coli RNase I (Ambion) for 1 h. Ribosome-protected fragments were collected by sucrose density gradient ultracentrifugation and extracted using TRIzol reagent. The library was constructed with a small RNA library construction kit (New Englad Biolabs) and used for high-throughput sequencing.

For RNA sequencing (RNA-seq) and m⁶A-seq, adaptor and low-quality bases were trimmed by Fastp, and the trimmed reads with lengths <15 nucleotides were discarded (24). The remaining reads were mapped to the mouse reference genome (mm10) using HISAT2 (25). After counting the number of reads, the expression of transcripts was quantified as reads per kilobase per million mapped reads (RPKM). Read coverage of m⁶A-seq and RNA-seq on transcripts was plotted, and the m⁶A peaks were further identified based on the following criteria: 1) genes with sufficient expression level (RPKM >1), 2) m⁶A modification detected by software MACS2 (significance level ≤0.05), and 3) POI (peak reads/input reads) >2 (26).

Statistical analysis was performed using GraphPad Prism 8 software (GraphPad Software, Inc., San Diego, CA). Throughout, data are represented as mean ± SD unless stated otherwise. Asterisks denote the level of statistical significance (*P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001).

The data and resources that support the findings of this study are available from the corresponding author upon reasonable request. Sequencing data, including m⁶A-seq, ribo-seq, and RNA-seq data, are available in the Gene Expression Omnibus under accession number GSE178884 (https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE178884).

We determined the expression level of METTL3 in different types of adipose tissue. Data revealed that METTL3 is ubiquitously expressed in adipose tissue with enrichment in BAT (Supplementary Fig. 1A). Cold treatment (4°C) successfully induced the beiging of WAT, as evidenced by smaller lipid droplets and induction of thermogenesis genes, including UCP1 (Fig. 1A). Strikingly, the protein level of METTL3 increased in beige adipose tissue (Fig. 1A and B). Consistent with the upregulation of METTL3, the total m⁶A methylation level was significantly increased in beige adipose tissue but decreased after Mettl3 knockout (Fig. 1C), implying that METTL3-mediated m⁶A formation may be involved in the beiging process.

To analyze the m⁶A signature, we performed m⁶A-seq in WAT from wild-type mice treated with or without cold exposure (Supplementary Fig. 1B). m⁶A-seq data revealed the expected distribution of m⁶A peaks enriched at the stop codon (Fig. 1D). However, the pattern of the two groups was different. The m⁶A peaks in start and stop codons were higher in WAT from control mice, whereas the CDS region was significantly higher in WAT from cold-treated mice (Fig. 1D and Supplementary Fig. 1C). The m⁶A peaks in both groups were characterized by the canonic AGACA motif (Fig. 1E) (27,28). Further analysis of m⁶A-seq data identified 6,199 differential m⁶A sites within 3,853 transcripts. A higher number of sites showed elevated m⁶A methylation in beige adipose tissue compared with WAT (Supplementary Fig. 1D). m⁶A methylation alternation also reflected cold-induced thermogenesis in Gene Set Enrichment Analysis (Fig. 1F). Furthermore, Kyoto Encyclopedia of Genes and Genomes pathway and Gene Ontology analyses of the m⁶A methylome revealed that the genes with upregulated m⁶A modification during beiging were involved in a series of metabolism- and thermogenesis-related pathways, such as oxidative phosphorylation (Cox10 and Ndufa5), fatty acid metabolism (Cpt2, Elovl6, and Scd2), and carbon metabolism (Pfkp, Phgdh, and Eno1) (Fig. 1G and Supplementary Fig. 1E). Together, these data provide an overview of METTL3-mediated m⁶A reprogramming during beiging and imply a role for m⁶A mRNA modification in this biological process.

To study the function of METTL3 in adipose tissue, we generated an adipocyte-specific Mettl3 knockout mouse model under the adiponectin promoter (Adipoq^Cre/+: Mettl3^fl/fl, termed Mettl3^cKO or M3^cKO) and successfully depleted Mettl3 in adipose tissue (Fig. 2A and Supplementary Fig. 2A). Compared with wild-type mice (Mettl3^fl/fl, termed Mettl3^CTL or M3^CTL), Mettl3^cKO mice had similar weights of BAT, iWAT, eWAT, and liver, as well as similar body weights (Supplementary Fig. 2B and C). Of note, daily food intake on standard CD did not change in the two groups (Supplementary Fig. 2D). Mettl3^cKO mice had whiter BAT (Supplementary Fig. 2E), in line with the previous report that UCP1-Cre-driven Mettl3 depletion impairs BAT maturation and energy expenditure (18). Mature iWAT is characterized by high expression of complement factor D (Adipsin) and low expression of preadipocyte factor-1 (Pref-1). Mettl3 depletion did not alter the expression of Adipsin or Pref-1 mRNA in vivo, indicating the successful maturation of WAT (Supplementary Fig. 2F). Hematoxylin-eosin staining showed the similar size of lipid droplets in Mettl3^cKO and Mettl3^CTL mice (Supplementary Fig. 2G). The above data suggest that adipocyte-specific knockout of Mettl3 does not impair WAT adipogenesis.

The upregulation of METTL3 during WAT beiging prompted us to study its function in this process. We found that cold exposure failed to induce beiging in Mettl3^cKO mice (Fig. 2B). Histological analysis showed that the adipose tissue of Mettl3^cKO mice had larger lipid droplets than that of Mettl3^CTL mice (Fig. 2C). Consistently, the expression of UCP1, along with other thermogenic and lipolytic genes, was dramatically reduced in the iWAT of Mettl3^cKO mice (Fig. 2D–F), just as in the BAT of Mettl3^cKO mice (Supplementary Fig. 2H and I).

The process of WAT beiging antagonizes obesity. One study showed that UCP1-Cre-driven Mettl3 knockout mice were prone to obesity (18). To further study the role of Mettl3 in the regulation of the metabolic capability of mice, we treated M3^CTL and M3^cKO mice with HFD feeding. Interestingly, we found that HFD-treated mice showed decreased expression of METTL3 in iWAT compared with CD-fed mice (Fig. 3A). The depletion of Mettl3 had no effect on body weight or rectal temperature under similar food intake (Fig. 3B–D). However, the BAT of M3^cKO mice was whiter and larger than that of M3^CTL mice (Fig. 3E and F). M3^cKO mice also showed impaired glucose tolerance and insulin resistance compared with M3^CTL mice (Fig. 3G and H). O₂ consumption, CO₂ generation, and energy heat generation were significantly reduced during light and dark cycles in M3^cKO mice (Fig. 3I–K). Despite the unchanged body weight that can be attributed to the low expression of METTL3 under HFD treatment, these data imply that the depletion of Mettl3 impairs the metabolic capability of diet-induced obese mice.

To identify METTL3-targeted m⁶A modifications and their function in mRNA metabolism, we performed m⁶A-seq, RNA-seq, and ribo-seq in beige adipose tissue from Mettl3^cKO and Mettl3^CTL mice (Fig. 4A and B and Supplementary Fig. 3A). We considered that m⁶A sites with a decreased signal after Mettl3 knockout were METTL3 direct targets. On the basis of this criterion, we identified 3,827 METTL3-targeted m⁶A sites located in 2,847 transcripts, which account for ∼62% of all m⁶A modifications in mRNAs in beige adipose tissue (Fig. 4C and Supplementary Fig. 3B). These METTL3-targeted transcripts showed a notable decrease at the mRNA level (RNA-seq) after Mettl3 knockout, compared with non-METTL3-targeted transcripts and non-m⁶A transcripts (Fig. 4D and E and Supplementary Fig. 3C), suggesting that m⁶A modification in beige adipocytes positively regulates gene expression. Specifically, the m⁶A distribution within UCP1 was significantly increased in beige adipose tissue but decreased after Mettl3 knockout (Supplementary Fig. 3D), which is consistent with the previous report (18). The m⁶A methylation levels of other thermogenic-associated genes, such as Pparg, Prdm16, and Zfp423, changed slightly in beige adipose tissue (Supplementary Fig. 3E–G). Ribo-seq change correlated well with RNA-seq change, and only a small group of mRNAs were regulated at the translational level (altered in ribo-seq but not RNA-seq) (Fig. 4F). These data suggest that METTL3 regulates gene expression mainly at the mRNA level during iWAT beiging.

Metagene analysis revealed a common reduction of m⁶A methylation around the stop codon of transcripts in Mettl3^cKO beige adipose tissue (Fig. 5A). Among the altered transcripts, we noticed that m⁶A modification in Klf9 mRNA showed a dramatic change (Fig. 5B). Consistently, the mRNA level of Klf9 was also dramatically reduced in Mettl3 knockout iWAT (Fig. 5C). The elevated expression of KLF9 in beige adipose tissue was also confirmed, implying a role of KLF9 in iWAT beiging (Fig. 5D and E). To decipher the functional relationship between KLF9 and METTL3, we performed rescue experiments by expressing FLAG-tagged KLF9 in Mettl3-depleted iWAT using an AAV8 system (Fig. 5F). Mettl3 knockout induced larger lipid droplets and inhibited the expression of UCP1 (Fig. 5G and H). Overexpression of KLF9 largely reversed the impaired beiging and UCP1 expression elicited by METTL3 deletion (Fig. 5G and H). These results indicate that KLF9 is a key effector in promoting beiging downstream of METTL3.

To resolve the direct regulation of METTL3 on KLF9 expression, we generated Mettl3 knockdown preadipocyte 3T3-L1 cell lines (Fig. 6A), which also showed decreased KLF9 protein. Consistent with the result in vivo, knockdown of METTL3 dramatically reduced the expression of Klf9 at the mRNA level (Fig. 6A). Furthermore, we examined the stability of Klf9 mRNA by blocking de novo RNA synthesis with actinomycin D. The half-life of Klf9 mRNA is longer in control cells than in METTL3-deficient cells (Fig. 6B), indicating that METTL3-mediated m⁶A modification stabilizes Klf9 mRNA.

Two m⁶A peaks were identified in Klf9 mRNA, lying in the CDS region near stop codon and the 3′ untranslated region (UTR), respectively. The m⁶A peak in the 3′UTR (1816nt-2199nt of the transcript ENSMUST00000036884) showed dynamic change, with an increase in cold-treated iWAT and a decrease in Mettl3^cKO iWAT (Fig. 6C). To examine whether the m⁶A modification in this region regulates Klf9 stability, we cloned each of the m⁶A peaks to a luciferase reporter (Fig. 6D). Erasing m⁶A by silencing METTL3 decreased Klf9-3′UTR luciferase activity but not Klf9-CDS luciferase activity (Fig. 6E), suggesting that the m⁶A modification in 3′UTR is involved in the stabilization of Klf9 mRNA. Collectively, METTL3-dependent m⁶A modification in the 3′UTR mediates the stabilization of Klf9 mRNA, which is an essential prerequisite for KLF9 expression.

We proposed that activation of m⁶A methyltransferase activity using small-molecule chemicals may induce beiging. Intriguingly, a former study identified methyl piperidine-3-carboxylate (MP3C) as a small-molecule agonist targeting the METTL3-METTL14-WTAP complex (23). Indeed, we confirmed that intraperitoneal injection of MP3C increased the mRNA m⁶A methylation level of adipose tissue (Fig. 7A and B). The in vitro addition of MP3C increased the activity of METTL3 in a dose-dependent manner but had almost no effect on another SAM-dependent m⁶A methyltransferase, METTL16 (Fig. 7C). Mice treated with MP3C showed smaller lipid droplets, elevated UCP1 expression, and increased OCR in iWAT (Fig. 7D–F). The MP3C treatment significantly improved the O₂ consumption, CO₂ generation, and energy heat generation of mice (Fig. 7G–I). These data indicate that this agonist promotes beiging and thermogenesis.

We proposed that MP3C might regulate obesity-associated metabolic dysfunction. To that end, we treated diet-induced obese mice with the agonist. Notably, MP3C hampered the body weight accumulation of HFD-fed mice, despite comparable daily food intake (Fig. 8A–C). The weight of iWAT from MP3C-treated mice was smaller than that from control mice, but the weight of BAT did not change (Fig. 8D and E). Strikingly, the MP3C-treated mice fed an HFD had more ameliorated glucose intolerance and insulin resistance than their control littermates (Fig. 8F and G). The MP3C treatment of obese mice significantly upregulated O₂ consumption, CO₂ generation, and energy heat generation (Fig. 8H–J). Thus, MP3C is considered a potential molecule targeting the m⁶A writer complex for curing metabolic diseases (Fig. 8K).

A series of transcriptional and epigenetic regulators have been characterized in WAT beiging. In this study, we demonstrate that the m⁶A writer METTL3 is induced in beige adipocytes and required for WAT beiging. Moreover, METTL3-mediated m⁶A modification posttranscriptionally regulates the expression of thermogenic mRNAs. Importantly, the agonist of METTL3, methyl piperidine-3-carboxylate, shows great potential for promoting beiging and reversing obesity-associated disorders.

In our study, we used Adipoq-Cre for the deletion of the Mettl3 gene, which induces Mettl3 knockout in mature WAT and BAT, avoiding the possible effect on embryonic development. Using this knockout mouse, we demonstrated that in addition to BAT thermogenesis, METTL3 plays an essential role in WAT beiging under cold treatment. A recent study reported that UCP1-Cre-driven Mettl3 knockout inhibits BAT maturation and thermogenesis (18). It should be noted that UCP1 is activated in both brown and beige adipocytes. Therefore, the phenotype observed in UCP1-Cre-driven Mettl3 deletion mice may be attributed to the abnormality in both WAT and BAT. Moreover, we demonstrated that MP3C, an activator of METTL3-METTL14-WTAP, combats obesity-associated disorders in mice. The weight of WAT rather than BAT was reduced under MP3C treatment, implying that the effects of MP3C occur largely through beige adipose tissue, although the function of BAT is not excluded.

During WAT beiging, not only does the total m⁶A level increases, but the landscape of m⁶A methylome is also remodeled. Profiling of m⁶A revealed that most of the metabolism-related mRNAs were dynamically harbored with m⁶A modifications during the process of beiging. Loss of Mettl3 impaired proper m⁶A modification in metabolic mRNAs, leading to decreased mRNA levels. These findings point to a model in which mRNA transcription and m⁶A-regulated mRNA stabilization orchestrate a regulatory network required for beiging.

It has been reported that KLF9 is induced by cold treatment and β-adrenergic agonists in adipose tissue by an unknown mechanism (29). We demonstrate that the expression of KLF9 relies heavily on the stabilization of its mRNA by m⁶A modifications. Multiple m⁶A sites were identified in Klf9 mRNA, but only the m⁶A in the 3′UTR dynamically responded to cold treatment and Mettl3 deletion. Luciferase assays confirmed the regulatory effect of this m⁶A modification on Klf9 expression. Of note, the sequences flanking the functional m⁶A sites are conserved between human and mouse, implying an evolutionary principle for regulation. Overexpression of KLF9 rescues impaired beiging by Mettl3 depletion, indicating that KLF9 functionally mediates METTL3-promoted adipose thermogenesis. Consistent with our findings, adipose-specific Klf9 transgenic mice display enhanced energy expenditure and are resistant to HFD-induced obesity (29). Moreover, a genome-wide association study also indicated that common variants at the KLF9 gene are associated with BMI (30). These data indicate an essential role of KLF9 in the control of adipose energy homeostasis; however, the detailed mechanism requires further investigation.

The thermogenic activity of m⁶A provides an attractive therapeutic target for obesity and other metabolic disorders. Indeed, the FTO inhibitor entacapone, which increases total m⁶A level, shows great potential in activating thermogenesis in adipose tissue, reducing body weight, and lowering the blood glucose level in obese mice (19). Here, we report that methyl piperidine-3-carboxylate, a small-molecule agonist of the METTL3-METTL14-WTAP complex (23), can also promote the thermogenesis of adipose tissue. Mice treated with this agonist showed activation of beige adipose tissue, along with reduced body weight and improved glucose metabolism. The MP3C treatment significantly improved energy consumption and energy expenditure, including upregulated O₂ consumption, CO₂ generation, and energy heat generation. In conclusion, the results indicate that m⁶A has therapeutic potential in regulating metabolic disorders.

Acknowledgments. The authors are grateful to Dr. Ming-Han Tong (Shanghai Institute of Biochemistry and Cell Biology, Chinese Academy of Sciences) for providing Mettl3^fl/fl mice and Dr. Chi Luo (Zhejiang University) for the critical reading of the manuscript.

Funding. This work was supported by grants from the National Natural Science Foundation of China (82100926 to S.W. and 82073110 and 81672847 to X.G.), the Natural Science Foundation of Zhejiang Province, China (LQ19H120012 to K.L.), and the Zhejiang Youth Talent Support Program (ZJWR0308085 to X.G.).

Duality of Interest. No potential conflicts of interest relevant to this article were reported.

Author Contributions. R.X. and S.Y. performed most experiments. R.X. and X.G. designed the experiments, performed the data analysis, and wrote the manuscript. X.Z., Y.G., Y.Q., J.H., Z.C., and K.L. assisted in the animal experiments. X.G. and S.W. conceived the project. All authors discussed the results and edited the manuscript. X.G. and S.W. are the guarantors of this work and, as such, had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.