Abstract

   N^6-methyladenosine (m^6A) is among the most abundant mRNA
   modifications, yet its cell-type-specific regulatory roles remain
   unclear. Here we show that m^6A methyltransferase-like 14 (METTL14)
   differentially regulates transcriptome in brown versus white adipose
   tissue (BAT and WAT), leading to divergent metabolic outcomes. In
   humans and mice with insulin resistance, METTL14 expression differs
   significantly from BAT and WAT in the context of its correlation with
   insulin sensitivity. Mettl14-knockout in BAT promotes prostaglandin
   secretion, improving systemic insulin sensitivity. Conversely,
   Mettl14-knockout in WAT triggers adipocyte apoptosis and systemic
   insulin resistance. m^6A-seq and RNA-seq integration revealed
   upregulated prostaglandin biosynthesis pathways in BAT and apoptotic
   pathways in WAT with Mettl14 deficiency. Stable METTL14-knockout
   hBAs/hWAs show METTL14-mediated m^6A promotes mRNA decay of PTGES2 and
   CBR1 in hBAs and TRAIL and TNFR1 in hWAs. These data shed light on the
   ability of m^6A to impact metabolism in a cell-type-specific manner
   with implications for influencing the pathophysiology of metabolic
   diseases.

   Subject terms: Insulin signalling, Apoptosis, RNA metabolism
     __________________________________________________________________

   N6 - methyladenosine (m6 A) is a prevalent and abundant mRNA
   modification, yet its cell-type specific roles remains unclear. Here,
   the authors show that m6 A modification differentially regulates the
   transcriptomes of brown and white adipose tissue, resulting in distinct
   impacts on systemic metabolism.

Introduction

   N^6-methyladenosine (m^6A) modification is a pivotal
   post-transcriptional regulatory process in mammalian cells, controlled
   by writers (METTL3, METTL14, WTAP), erasers (ALKBH5, FTO), and readers
   (YTHDC, YTHDF, IGF2BP2)^[48]1,[49]2. m^6A methylation of mRNA can
   affect its expression, stability, and translation efficiency, leading
   to altered protein abundance to determine diverse biological outcomes
   related to cell fate determination, organ development, cell
   proliferation, and/or survival^[50]1,[51]2. Indeed, this ability to
   modify mRNA to regulate intricate cell processes can be an advantage
   considering the human body comprises more than 200 distinct cell types,
   each with specialized roles in physiological processes. Nevertheless,
   m^6A has also shown divergent functions; for example, METTL14-mediated
   m^6A modification exhibits both oncogenic^[52]3–[53]5 and anti-tumor
   effects^[54]6,[55]7 across cancer types. A lingering question in the
   field of m^6A biology pertains to the precise molecular underpinnings
   by which m^6A governs cell-type-specific modulation of gene expression
   in a contingent fashion, thereby affording precise control over diverse
   cellular functionalities. Therefore, understanding cell-type-specific
   m^6A regulation is crucial for unraveling the complexities of cellular
   processes and for developing targeted therapeutic strategies for
   various pathophysiological states including metabolic diseases^[56]8.

   Brown adipose tissue (BAT) and white adipose tissue (WAT) are two major
   types of tissues that share some common features but also have distinct
   properties. BAT is specialized for energy expenditure through
   thermogenesis and is acknowledged as a secretory organ capable of
   producing various batokines that play roles in energy metabolism,
   inflammation, and/or whole-body insulin sensitivity^[57]9,[58]10. WAT
   primarily functions as an energy storage depot, accumulating excess
   energy in the form of triglycerides (TGs)^[59]9,[60]11,[61]12. In
   addition, WAT also secretes endocrine hormones that can regulate the
   metabolism of distal organs. Dysregulation of WAT, specifically in the
   context of perturbed signaling pathways involving cytokines of the
   tumor necrosis factor (TNF) family, underlies the initiation of
   cellular apoptosis and inflammation within the adipose tissue,
   eventually resulting in lipid spill-over, glucotoxicity, and insulin
   resistance^[62]13. Hence, the context-specific regulation within
   distinct cell types such as BAT and WAT holds relevance for the
   modulation of whole-body metabolism.

   The involvement of m^6A modification in BAT and WAT has recently
   emerged as a nascent yet critical realm of investigation, particularly
   within the scope of addressing obesity and metabolic disorders. While
   METTL3, METTL14, and WTAP are classified as major components of the
   m^6A writer complex, previous research including our own argues for
   unique roles for each. Some studies have investigated the impact of
   m^6A by focusing on METTL3 and WTAP on adipose tissue biology and
   function, with a particular focus on BAT or beige
   adipocytes^[63]14–[64]18. Our previous study indicates that Mettl14
   deficiency in BAT promotes the secretion of beneficial endocrine
   factor(s) to protect mice from diet-induced obesity (DIO) and insulin
   resistance independent of BAT thermogenesis^[65]19. In this manuscript,
   we directly address the cell-type specific regulation involving m^6A by
   examining the roles of METTL14 in BAT versus WAT in the modulation of
   whole-body metabolism.

   In this study, we build on our preliminary observations of distinct
   expression patterns of METTL14 in BAT versus WAT from humans and mice
   with insulin resistance. The BAT-specific (Ucp1-cre-driven)
   Mettl14-knockout (M14^KO) mouse model exhibits enhanced secretion of
   endocrine factor(s) with improved systemic metabolic homeostasis that
   is independent of BAT thermogenic function. Conversely, an independent
   mouse model with Adipoq-cre-driven Mettl14-knockout shows white
   adipocyte apoptosis that consequently reduces its lipid storage
   capacity, impairing whole-body insulin sensitivity and energy
   expenditure. m^6A-MeRIP-sequencing (m^6A-seq) and RNA-sequencing
   (RNA-seq) in Mettl14-deficient BAT versus Mettl14-deficient WAT
   revealed different methylomes and transcriptomes in the two cell types.
   Mechanistically, Mettl14-mediated m^6A modification specifically
   destabilizes critical BAT transcripts associated with phospholipid
   metabolic processes, while in WAT, it negatively regulates TNFα and
   apoptosis pathway-related genes via m^6A-dependent and independent
   mechanisms. These findings underscore the divergent roles of METTL14 in
   BAT versus WAT-mediated metabolism, emphasizing cell-type and
   context-specific m^6A regulation as a significant consideration for
   addressing metabolic disorders.

Results

Differential abundance of METTL14 in BAT versus WAT obtained from humans or
mice with insulin resistance

   To distinguish the m^6A methylome landscapes between BAT and WAT, we
   compared RNA-seq and m^6A-MeRIP-seq analyses between fully
   differentiated human brown adipocytes (hBAs) versus human white
   adipocytes (hWAs)^[66]20. To substantiate that m^6A selectively
   modifies transcripts expressed in hBAs and hWAs, we first filtered the
   transcripts unique to hWAs or hBAs and identified 25,563 transcripts
   that were ubiquitously expressed in both cell types (Supplementary
   Fig. [67]1a). Next, to specifically focus on the m^6A-modified
   transcripts, we intersected these 25,563 transcripts with those that
   are m^6A-modified, to reveal a total of 7866 m^6A-modified transcripts
   (Supplementary Fig. [68]1b). Principal component analysis (PCA) of the
   m^6A-seq data revealed a distinct separation between hBAs and hWAs
   (Supplementary Fig. [69]1c). About a quarter (2086) of the
   7866 m^6A-modified transcripts showed comparable m^6A methylation
   levels between hBAs and hWAs (Supplementary Fig. [70]1d). On the other
   hand, 2681 and 3099 hypermethylated m^6A transcripts were exclusively
   identified in hBAs and hWAs, respectively (Supplementary Fig. [71]1d).
   These differentially m^6A-modified transcripts were notably enriched in
   pathways associated with VEGFA-VEGFR2, TGFβ, mRNA processing, insulin,
   TNFα, or apoptosis modulation signaling pathways (Supplementary
   Fig. [72]1e). Thus, brown and white adipocytes manifest highly distinct
   m^6A modification patterns that strongly align with cell-type-specific
   gene expression programs, pointing to functional specificity of m^6A at
   the cellular/tissue level. These preliminary results provided us with
   the rationale to explore the direct impact of m^6A modulators in these
   two different metabolic cell types.

   METTL14, as a key component of the m^6A writer complex, has been
   reported to exhibit pronounced cell-type-specific roles within the
   context of metabolism^[73]19,[74]21. For example, our previous studies
   revealed that m^6A hypomethylation, and more specifically METTL14
   downregulation, led to impaired β-cell function and identity^[75]21.
   Conversely, METTL14 showed higher gene expression in BAT from humans
   with obesity, and BAT-specific Mettl14-knockout mice presented improved
   systemic insulin sensitivity^[76]19. Considering these observations, we
   hypothesized that METTL14 has divergent roles in the transcriptional
   programs to regulate the functions of brown versus white adipocytes.

   To investigate this hypothesis, we first examined METTL14 protein
   abundance in the fully differentiated hBAs and hWAs under physiological
   conditions. We observed a higher relative METTL14 protein level in hWAs
   compared to hBAs cells (Supplementary Fig. [77]1f, g). In C57BL/6 N
   male mice fed a chow diet (CD), METTL14 protein abundance was
   significantly higher in inguinal white adipose tissue (iWAT) than
   epididymal white adipose tissue (eWAT) or interscapular brown adipose
   tissue (iBAT) (Supplementary Fig. [78]1h, i). These data support the
   notion that METTL14 exhibits adipocyte- and adipose depot-specific
   expression patterns under physiological conditions.

   Next, we examined whether METTL14 exhibits differential expression
   across distinct human WATs, such as individuals with varying levels of
   body mass index (BMI) or insulin sensitivity (Supplementary
   Table [79]1). First, a comparison between lean, insulin-sensitive
   individuals and those with obesity but who are also insulin-sensitive
   revealed no significant differences in METTL14 protein abundance
   (Supplementary Fig. [80]1j, [81]k). Likewise, when comparing
   individuals who are lean and insulin-resistant to those with obesity
   and insulin-resistant, no significant differences were detected in the
   levels of METTL14 protein (Supplementary Fig. [82]1l, [83]m). This lack
   of distinction may be attributable to patient heterogeneity and it is
   possible that METTL14 abundance is not intrinsically linked to human
   BMI.

   Next, to isolate the influence of insulin sensitivity from BMI, we
   compared individuals with obesity who are insulin-sensitive
   (HOMA-IR < 2.9) to individuals with obesity who are insulin-resistant
   (HOMA-IR > 2.9). Remarkably, METTL14 protein level was significantly
   lower in both subcutaneous white adipose tissue (scWAT) and visceral
   white adipose tissue (vsWAT) from humans with insulin resistance
   (Fig. [84]1a, [85]b). Spearman correlation analysis revealed a robust
   negative correlation between METTL14 abundance and HOMA-IR in scWAT or
   vcWAT (Fig. [86]1c, [87]d).

Fig. 1. METTL14 exhibits differential protein abundance in BAT or WAT from
insulin-resistant humans and mice.

   [88]Fig. 1
   [89]Open in a new tab

   a, b Western blots (a) and quantification (b) of METTL14 and VINCULIN
   in subcutaneous and visceral white adipose tissues from individuals
   with obesity who are insulin-sensitive or insulin-resistant (n = 11 for
   insulin-sensitive individuals with obesity in scWAT and n = 26 for
   insulin-resistant individuals with obesity in scWAT; n = 11 for
   insulin-sensitive individuals with obesity in vcWAT and n = 13 for
   insulin-resistant individuals with obesity in vcWAT). IS,
   insulin-sensitive; IR, insulin-resistant. scWAT from insulin-sensitive
   individuals with obesity was loaded in both the upper and lower panels
   of (a) to accommodate the insulin-resistant samples. c, d Spearman
   correlation between METTL14 relative abundance in scWAT (c) or vcWAT
   (d) and insulin sensitivity, as indicated by HOMA-IR (n = 11 for
   insulin-sensitive individuals with obesity in scWAT and n = 26 for
   insulin-resistant individuals with obesity in scWAT; n = 11 for
   insulin-sensitive individuals with obesity in vcWAT and n = 13 for
   insulin-resistant individuals with obesity in vcWAT). e, f Western
   blots (e) and quantification (f) of METTL14 and VINCULIN in iWAT, eWAT,
   and iBAT from control (ob + ) or ob/ob mice (n = 6/group). iWAT,
   inguinal white adipose tissue; eWAT, epididymal white adipose tissue;
   iBAT, interscapular brown adipose tissue. g–i Spearman correlation
   between METTL14 relative abundance in iWAT (g), eWAT (h), or iBAT (i)
   and insulin sensitivity indicated by HOMA-IR (n = 6/group). j, k
   Western blots (j) and quantification (k) of METTL14 and VINCULIN or
   GAPDH in iWAT, eWAT, and iBAT from control (db + ) or db/db mice
   (n = 6/group). l–n Spearman correlation between METTL14 relative
   abundance in iWAT (l), eWAT (m), or iBAT (n) and insulin sensitivity
   indicated by HOMA-IR (n = 6/group). o, p Western blots (o) and
   quantification (p) of METTL14 and VINCULIN or GAPDH in iWAT, eWAT, and
   iBAT from LFD- or HFD-fed mice (n = 6/group). q–s Spearman correlation
   between METTL14 relative abundance in iWAT (q), eWAT (r), or iBAT (s)
   and insulin sensitivity indicated by HOMA-IR (n = 6/group). All samples
   in each panel are biologically independent. Data are presented as
   means ± SEM from two independent experiments by two-tailed unpaired
   t-test (b, f, k, p) and one-tailed Spearman correlation (c, d, g–i,
   l–n, q–s). See also Supplementary Fig. [90]1. Source data are provided
   as a Source Data file.

   Consistent with the human data we detected similar changes in three
   different mouse models. First, we examined METTL14 protein levels in
   the fat tissues from the ob/ob (Lep^ob) mouse, which is a model for T2D
   and obesity that is characterized by hyperinsulinemia (Supplementary
   Fig. [91]1n), glucose intolerance (Supplementary Fig. [92]1o), and
   insulin resistance (Supplementary Fig. [93]1p). The METTL14 protein
   levels were lower in iWAT and eWAT from ob/ob mice compared to ob+ mice
   (Fig. [94]1e, [95]f). On the other hand, the METTL14 protein level was
   higher in the ob/ob-iBAT as compared to ob/+-iBAT (Fig. [96]1e, [97]f).
   Thus, in the ob/ob mouse model, METTL14 in WATs is positively
   correlated with insulin sensitivity, and its protein abundance in BAT
   is negatively correlated with insulin sensitivity (Fig. [98]1g–[99]i).
   In the db/db (Lepr^db) mouse model, which exhibits hyperinsulinemia
   (Supplementary Fig. [100]1q) and peripheral tissue insulin resistance
   (Supplementary Fig. [101]1r), METTL14 protein level was lower in iWAT
   and eWAT as compared to the db+ controls (Fig. [102]1j, [103]k).
   Conversely, elevated protein abundance of METTL14 was observed in iBAT
   dissected from db/db mice (Fig. [104]1j, [105]k). METTL14 displayed an
   inverse relationship with WAT-mediated insulin sensitivity and a
   positive correlation with BAT-mediated whole-body insulin sensitivity
   (Fig. [106]1l–[107]n). Comparable outcomes were evident in the high-fat
   diet (HFD)-induced obese (DIO) mice compared to the low-fat diet
   (LFD)-fed group which comprised lean insulin-sensitive mice
   (Fig. [108]1o, [109]s). The lack of a significant correlation between
   METTL14 protein levels in iBAT and whole-body insulin sensitivity could
   be related, in part, to the short duration of HFD feeding
   (Fig. [110]1s). Collectively, these data argue that METTL14 plays a
   differential role in BAT versus WAT to regulate whole-body metabolism.

Mettl14 deficiency promotes contrasting modulation of BAT- versus
WAT-mediated systemic insulin sensitivity in mice

   To comprehensively interrogate the divergent roles of METTL14-mediated
   m^6A modifications in BAT versus WAT and their respective effects on
   whole-body metabolic processes, we employed a dual approach.
   Specifically, we generated two distinct lines of knockout mice by
   breeding Mettl14-floxed (M14^fl/fl) mice^[111]21, with either a)
   Adipoq-cre mice^[112]22, which facilitated exploration of METTL14’s
   role in the overall functionality of adipose tissue, or b) Ucp1-cre
   mice^[113]23, enabling a focused analysis of BAT-specific Mettl14
   knockout and its repercussions on systemic metabolism (knockouts
   verified in Supplementary Fig. [114]2a, [115]b). Both knockout-out
   lines displayed a typical female and male ratio and normal birth
   counts. Under CD feeding, both M14^fl/fl-Ucp1^cre and
   M14^fl/fl-Adipoq^cre male and female mice exhibited similar body weight
   trajectories compared to the M14^fl/fl control group (Fig. [116]2a for
   males, Supplementary Fig. [117]2c for females). Intriguingly,
   evaluation of glucose homeostasis revealed that M14^fl/fl-Ucp1^cre male
   and female mice displayed enhanced insulin sensitivity and improved
   glucose tolerance, while M14^fl/fl-Adipoq^cre mice exhibited diminished
   insulin sensitivity and similar glucose tolerance as the control
   M14^fl/fl mice (Fig. [118]2b, [119]c for males, and Supplementary
   Fig. [120]2d and e for females).

Fig. 2. Ablation of Mettl14 in BAT or WAT results in opposite systemic
metabolic phenotypes in mice.

   [121]Fig. 2
   [122]Open in a new tab

   a Body weight trajectories of CD-fed male mice (n = 10/group). b
   Insulin tolerance test of CD-fed male mice (n = 6 for M14^fl/fl group,
   n = 6 for M14^fl/fl-Ucp1^cre group, and n = 9 for M14^fl/fl-Adipoq^cre
   group). c Glucose tolerance test of CD-fed male mice (n = 14 for
   M14^fl/flgroup, n = 8 for M14^fl/fl-Ucp1^cre group, and n = 9 for
   M14^fl/fl-Adipoq^cre group). d Body weight trajectories of high-fat
   diet (HFD)-fed male mice (n = 11 for M14^fl/flgroup, n = 8 for
   M14^fl/fl-Ucp1^cre group, and n = 11 for M14^fl/fl-Adipoq^cre group). e
   Insulin tolerance test of HFD-fed male mice (n = 11 for M14^fl/fl
   group, n = 8 for M14^fl/fl-Ucp1^cre group, and n = 11 for
   M14^fl/fl-Adipoq^cregroup). f Glucose tolerance test of HFD-fed male
   mice (n = 18 for M14^fl/fl group, n = 8 for M14^fl/fl-Ucp1^cre group,
   and n = 14 for M14^fl/fl-Adipoq^cre group). g Fasting glucose levels in
   the plasma of M14^fl/fl, M14^fl/fl-Ucp1^cre, and M14^fl/fl-Adipoq^cre
   male mice (n = 19 for M14^fl/fl group, n = 8 for M14^fl/fl-Ucp1^cre
   group, and n = 16 for M14^fl/fl-Adipoq^cre group). h Fasting insulin
   levels in the plasma of M14^fl/fl, M14^fl/fl-Ucp1^cre, and
   M14^fl/fl-Adipoq^cre male mice (n = 15 for M14^fl/fl group, n = 8 for
   M14^fl/fl-Ucp1^cre group, and n = 13 for M14^fl/fl-Adipoq^cre group). i
   Fasting adiponectin levels in the plasma of M14^fl/fl,
   M14^fl/fl-Ucp1^cre, and M14^fl/fl-Adipoq^cre male mice (n = 17 for
   M14^fl/fl group, n = 8 for M14^fl/fl-Ucp1^cre group, and n = 13 for
   M14^fl/fl-Adipoq^cregroup). j Representative images of the gross
   appearance of iBAT, iWAT, and eWAT from M14^fl/fl, M14^fl/fl-Ucp1^cre,
   and M14^fl/fl-Adipoq^cremale mice on HFD. k–o Representative images of
   H&E stained iWAT (k), eWAT (I), iBAT (m), liver (n), and muscle (o)
   from M14^fl/fl, M14^fl/fl-Ucp1^cre, and M14^fl/fl-Adipoq^cremale mice
   on HFD. All samples in each panel are biologically independent. Data
   are presented as means ± SEM from two independent experiments by
   Two-way ANOVA (a–f), and One-way ANOVA (g–i). Also see Supplementary
   Figs. [123]2 and [124]3. Source data are provided as a Source Data
   file.

   Upon challenging the mice with an HFD comprising 60% fat for 16 weeks,
   M14^fl/fl control mice exhibited the expected body weight gain and
   demonstrated compromised insulin sensitivity and glucose intolerance
   when compared to their low-fat diet (LFD) fed counterparts (data not
   presented). Both M14^fl/fl-Ucp1^cre and M14^fl/fl-Adipoq^cre male mice
   displayed significantly lower body weights compared to M14^fl/fl mice
   (Fig. [125]2d). M14^fl/fl-Ucp1^cre female mice showed consistently
   lower body weight, while M14^fl/fl-Adipoq^cre females had similar body
   weight when each group was compared to the control females
   (Supplementary Fig. [126]2f). Consistent with our earlier
   observations^[127]19, M14^fl/fl-Ucp1^cre mice were shielded from
   HFD-induced systemic insulin resistance and glucose intolerance
   (Fig. [128]2e, [129]f for males, Supplementary Fig. [130]2g, h for
   females). Conversely, M14^fl/fl-Adipoq^cre mice manifested pronounced
   insulin resistance, while their glucose tolerance was comparable to the
   controls (Fig. [131]2e, [132]f for males, Supplementary Fig. [133]2g, h
   for females). Additionally, M14^fl/fl-Ucp1^cre male mice exhibited
   substantially lower fasting glucose and insulin levels, along with
   elevated adiponectin levels, whereas M14^fl/fl-Adipoq^cre male mice
   displayed a different plasma profile characterized by a tendency
   towards elevated insulin and decreased adiponectin levels
   (Fig. [134]2g–[135]i). Notably, both KO mouse models exhibited reduced
   mass in iWAT and eWAT and increased mass in iBAT compared to controls;
   the latter was especially pronounced in M14^fl/fl-Adipoq^cre mice
   (Fig. [136]2j for males, and Supplementary Fig. [137]2i for females).

   We next examined the characteristics of the different fat mass depots
   and changes in the liver and skeletal muscle. Hematoxylin and eosin
   (H&E) staining revealed diminished adipocyte size in both iWAT and eWAT
   of both knockout mouse models (Fig. [138]2k, [139]l for males,
   Supplementary Fig. [140]2j, [141]k for females). Consistent with prior
   observations, M14^fl/fl-Ucp1^cre mice were protected from HFD-induced
   immune cell infiltration in eWAT, hypertrophy of brown adipocytes, and
   hepatic lipid accumulation (Fig. [142]2k–[143]n for males,
   Supplementary Fig. [144]2j–[145]n for females). In contrast,
   M14^fl/fl-Adipoq^cre mice displayed heightened immune cell infiltration
   in both iWAT and eWAT, brown adipocyte hypertrophy, and liver lipid
   accumulation (Fig. [146]2k–[147]n for males, Supplementary
   Fig. [148]2j–[149]n for females). Strikingly, while both M14^fl/fl
   M14^fl/fl-Ucp1^cre mice displayed normal skeletal muscle morphology, we
   observed the presence of both inter- and intra-cellular lipid
   accumulation in the muscle from M14^fl/fl-Adipoq^cre male mice
   indicating ectopic lipid deposition (Fig. [150]2o). Thus, manipulation
   of the expression of Mettl14 leads to contrasting regulatory roles in
   terms of insulin sensitivity and systemic metabolism in Ucp1^cre- and
   Adipoq^cre-mediated knockout mouse models.

   We next considered that Mettl14 was depleted in both WAT and BAT in the
   M14^fl/fl-Adipoq^cre mice (Supplementary Fig. [151]2a). To specifically
   investigate the role of Mettl14 in WAT functionality and overall
   metabolic control that is independent of iBAT, we surgically removed
   iBAT from M14^fl/fl and M14^fl/fl-Adipoq^cre mice (Supplementary
   Fig. [152]3a). Comparing the iBAT-depleted M14^fl/fl-Adipoq^cre mice to
   iBAT-depleted M14^fl/fl mice fed regular CD, revealed that the former
   persisted in exhibiting insulin resistance (Supplementary
   Fig. [153]2b). Of note, iBAT removal exacerbated insulin resistance
   observed in M14^fl/fl-Adipoq^cre mice as compared to iBAT-depleted
   M14^fl/fl mice (Supplementary Fig. [154]2b). Additionally, it was noted
   that in certain M14^fl/fl-Adipoq^cre mice following iBAT removal, there
   was a conspicuous absence of eWAT and pronounced atrophy in their iWAT,
   whereas iBAT-depleted M14^fl/fl mice exhibited normal gross appearance
   in their WAT depots (Supplementary Fig. [155]2c). These observations
   suggest that while WAT likely plays a dominant role in the phenotype of
   M14^fl/fl-Adipoq^cre mice, brown adipose tissues, including iBAT, may
   contribute positively, potentially acting as a metabolic sink for
   excess triglycerides or FFAs released from WAT. The specific
   contribution of other BAT depots in addition to iBAT in this context
   requires further investigation.

Ablation of Mettl14 in BAT or WAT differentially impacts insulin sensitivity
in distinct peripheral metabolic tissues

   Next, we examined tissue-autonomous and endocrine effects of Mettl14
   ablation on whole-body insulin sensitivity by injecting mice with
   insulin via the vena cava and assessing changes in insulin signaling
   proteins (pIRβ/IGF1Rβ and pAKT[S473]) in target metabolic tissues
   (Fig. [156]3a). In control mice, HFD decreased insulin-stimulated (1
   U/mouse) pIRβ/IGF1Rβ and pAKT[S473] in iWAT and eWAT compared to LFD
   (Fig. [157]3b, [158]c). Insulin resistance induced by the HFD was
   mitigated in M14^fl/fl-Ucp1^cre mice, evident in both iWAT and eWAT
   (Fig. [159]3b, [160]c). However, the insulin response was notably
   blunted in M14^fl/fl-Adipoq^cre mice in both iWAT and eWAT
   (Fig. [161]3b, [162]c).

Fig. 3. Ablation of Mettl14 in BAT or WAT differentially impacts insulin
sensitivity of peripheral metabolic tissues.

   [163]Fig. 3
   [164]Open in a new tab

   a Diagram illustrating the timeline for collecting metabolic tissues
   following saline/insulin injection into the vena cava of M14^fl/fl,
   M14^fl/fl-Ucp1^cre, and M14^fl/fl-Adipoq^cre male mice fed either a
   low-fat diet (LFD) or a high-fat diet (HFD) (n = 6/group) as indicated
   (Created with BioRender.com). b–f Western blot analysis and
   quantification of insulin-stimulated phosphorylation of IRβ/IGF1Rβ and
   AKT[S473] in iWAT (b), eWAT (c), iBAT (d), liver (e), and muscle (f)
   after injection of 1U insulin into the vena cava (n = 6/groups). g
   Western blot analysis of METTL14 protein abundance in sgNTC-, sgM14#1-,
   sgM14#2-, sgM14#3-hBA preadipocytes. h, i Western blot analysis (h) and
   quantification (i) of insulin-stimulated phosphorylation of IRβ/IGF1Rβ
   and AKT[S473] in differentiated sgNTC-, sgM14#1-, sgM14#2-, sgM14#3-hBA
   cells (n = 6/group). j Glucose uptake of differentiated sgNTC-,
   sgM14#1-, sgM14#2-, sgM14#3-hBA cells treated with or without insulin
   stimulation (n = 4/group). k Western blot analysis of METTL14 protein
   abundance in sgNTC-, sgM14#1-, sgM14#2-, sgM14#3-hWA preadipocytes. l,
   m Western blot analysis (l) and quantification (m) of
   insulin-stimulated phosphorylation of IRβ/IGF1Rβ and AKT[S473] in
   differentiated sgNTC-, sgM14#1-, sgM14#2-, sgM14#3-hWA cells
   (n = 6/group). n Glucose uptake of differentiated sgNTC-, sgM14#1-,
   sgM14#2-, sgM14#3-hWA cells treated with or without insulin stimulation
   (n = 4/group). All samples in each panel are biologically independent.
   Data are presented as means ± SEM from two or three independent
   experiments by Two-way ANOVA (b–f, i, j, m, n). Also see Supplementary
   Fig. [165]4. Source data are provided as a Source Data file.

   Furthermore, M14^fl/fl-Ucp1^cre mice on HFD exhibited elevated levels
   of pIRβ/IGF1Rβ and pAKT[S473] in iBAT, liver, and skeletal muscle
   tissues (Fig. [166]3d–[167]f). Conversely, M14^fl/fl-Adipoq^cre mice
   displayed reduced insulin signaling in their liver and skeletal muscle,
   excluding iBAT (Fig. [168]3d–[169]f). These findings highlight the
   distinct effects of Mettl14 ablation in different adipose tissue depots
   and underscore its role in modulating insulin sensitivity in diverse
   metabolic tissues.

   To investigate the translational significance of our findings in murine
   models, we capitalized on immortalized human brown and white adipocytes
   and established stable METTL14-knockout cell line colonies in hBAs and
   hWAs utilizing a single guide RNA targeting METTL14 (sgM14)
   (Fig. [170]3g, [171]k, respectively). Notably, the ablation of METTL14
   yielded contrasting effects on insulin sensitivity and glucose uptake
   capacity in fully differentiated hBAs (Fig. [172]3h–[173]j) versus hWAs
   (Fig. [174]3l–[175]n). These observations support the notion that
   METTL14 regulates insulin sensitivity in a context- and
   cell-type-dependent manner.

Mettl14 deficiency in BAT or WAT differentially impacts food intake and
energy expenditure in mice

   Next, we investigated whether METTL14 differentially modulates BAT- or
   WAT-mediated whole-body energy balance. Mettl14 deficiency utilizing
   Ucp1^cre exhibited an enhanced cold tolerance (Fig. [176]4a) without
   significant alterations in food intake (Fig. [177]4b), oxygen
   consumption (Fig. [178]4c), or energy expenditure (Fig. [179]4d).
   Conversely, M14^KO driven by Adipoq^cre resulted in a significant
   reduction in thermogenic capacity (Fig. [180]4a), food intake
   (Fig. [181]4b), oxygen consumption (Fig. [182]4e), and energy
   expenditure (Fig. [183]4f). Similarly, at the cellular level, METTL14
   deficiency showed no significant impact on the basal
   (Fig. [184]4g–[185]j) or β3-adrenergic agonist-stimulated
   (Figs. [186]4k–[187]n) mitochondrial and glycolytic functions of hBAs,
   while, the absence of METTL14 in hWAs negatively regulated these
   functions (Fig. [188]4o–[189]r). Previous studies have reported that
   METTL3 positively regulates thermogenesis in both BAT-specific
   (Ucp1^cre driven-knockout) and beige-specific (Adipoq^cre-mediated
   knockout) mouse models^[190]15,[191]17. Our findings highlight the
   intricate and diverse roles of the m^6A writer protein, METTL14, in the
   regulation of thermogenesis and metabolic control in adipose tissues,
   and unlike METTL3, METTL14 can regulate the thermogenesis of adipose
   tissues in a cell-specific manner.

Fig. 4. Mettl14 Deficiency Differentially Regulates Thermogenic Capacity of
BAT and WAT.

   [192]Fig. 4
   [193]Open in a new tab

   a Core body temperature of M14^fl/fl, M14^fl/fl-Ucp1^cre, and
   M14^fl/fl-Adipoq^cre mice after acute cold exposure (n = 15 for
   M14^fl/fl group, n = 8 for M14^fl/fl-Ucp1^cre group, and n = 6 for
   M14^fl/fl-Adipoq^cre group). b Daily food consumption measured by
   indirect calorimetry (n = 11 for M14^fl/fl group, n = 5 for
   M14^fl/fl-Ucp1^cre group, and n = 11 for M14^fl/fl-Adipoq^cre group).
   c, d Oxygen consumption (c) and energy expenditure (d) of
   M14^fl/fl-Ucp1^cre mice versus M14^fl/fl mice (n = 6 for M14^fl/fl
   group, n = 5 for M14^fl/fl-Ucp1^cre group). e, f Oxygen consumption (e)
   and energy expenditure (f) of M14^fl/fl-Adipoq^cre mice versus
   M14^fl/fl mice (n = 11/group). g–j Seahorse analysis of cellular oxygen
   consumption rate (OCR) (g, h, n = 5/group) and extracellular
   acidification rate (ECAR) (i, j, n = 5 for sgNTC, and n = 10 for sgM14)
   in differentiated sgNTC- and sgM14-hBA cells. k–n Seahorse analysis of
   cellular oxygen consumption rate (OCR) (k, l, n = 5/group) and
   extracellular acidification rate (ECAR) (m, n, n = 4 for sgNTC, and
   n = 5 for sgM14) in differentiated sgNTC- and sgM14-hBA cells
   stimulated with 10 µM CL-316,243 for 5 h before seahorse assay. o–r
   Seahorse analysis of cellular oxygen consumption rate (OCR) (o, p,
   n = 9 for sgNTC, and n = 10 for sgM14) and extracellular acidification
   rate (ECAR) (q, r, n = 5/group) in differentiated sgNTC- and sgM14-hWA
   cells (Data are representative of 2 independent experiments). All
   samples in each panel are biologically independent. Data are presented
   as means ± SEM from two or three independent experiments by Two-way
   ANOVA with Dunnett’s multiple comparisons test (a, b), Two-way ANOVA
   (c–g, i, k, m, o, q), and Two-tailed unpaired t-tests (h, j, l, n, p,
   r). Source data are provided as a Source Data file.

Mettl14 Deficiency Differentially Impacts m^6A Methylome and Transcriptome in
Mouse iBAT and iWAT

   The distinct roles of METTL14 in BAT- versus WAT-mediated whole-body
   metabolism prompted the hypothesis that METTL14 differentially
   modulates the transcriptional profiles of BAT and WAT by selectively
   methylating specific target transcripts. To test this hypothesis, we
   conducted m^6A-MeRIP-seq and RNA-seq analyses on iBAT from M14^fl/fl
   and M14^fl/fl-Ucp1^cre mice, as well as iWAT from M14^fl/fl and
   M14^fl/fl-Adipoq^cre mice.

   First, a comparison of the m^6A methylome between
   M14^fl/fl-Ucp1^cre-iBAT versus M14^fl/fl-iBAT, and between
   M14^fl/fl-Adipoq^cre-iWAT versus M14^fl/fl-iWAT (comparison approach
   shown in Fig. [194]5a) revealed 2465 and 2838 m^6A-hypomethylated genes
   (Fig. [195]5b), as well as 4854 and 8007 hypermethylated genes by
   Mettl14 deficiency in iBAT and iWAT respectively (Fig. [196]5c).
   Mettl14 deficiency resulted in 1375 and 1748 m^6A-hypermethylated genes
   (Fig. [197]5b), as well as 2010 and 5163 m^6A-hypermethylated genes
   exclusively in iBAT and iWAT, respectively (Fig. [198]5b). Next, the
   intersection of the differentially m^6A-modified genes in
   M14^fl/fl-Ucp1^cre-iBAT and M14^fl/fl-Adipoq^cre-iWAT revealed 1090
   (25.9%) and 2844 (28.4%) commonly hypomethylated and hypermethylated
   genes compared to M14^fl/fl mice, respectively (Fig. [199]5b, [200]c).

Fig. 5. Mettl14 Deficiency Results in Different m^6A Methylome and
Transcriptome of Mouse iBAT and iWAT.

   [201]Fig. 5
   [202]Open in a new tab

   a Schematic diagram of comparison approach for analyzing RNA-seq and
   m^6A-seq data derived from iWAT in M14^fl/fl and M14^fl/fl-Adipoq^cre
   mice, and from iBAT in M14 ^fl/fl and M14^fl/fl-Ucp1^cre mice (Created
   with BioRender.com). b Venn diagram representation of the exclusively
   m^6A-hypomethylated genes in M14^fl/fl-Ucp1^cre-iBAT versus
   M14^fl/fl-iBAT, commonly m^6A-hypomethylated genes, and exclusively
   m^6A-hypomethylated genes in M14^fl/fl-Adipoq^cre-iWAT versus
   M14^fl/fl-iWAT, statistical analyses were performed using the
   Benjamin-Hochberg procedure and genes were filtered for p < 0.01. c
   Venn diagram representation of the exclusively m^6A-hypermethylated
   genes in M14^fl/fl-Ucp1^cre-iBAT versus M14^fl/fl-iBAT, commonly
   m^6A-hypermethylated genes, and exclusively m^6A-hypermethylated genes
   in M14^fl/fl-Adipoq^cre-iWAT versus M14^fl/fl-iWAT, statistical
   analyses were performed using the Benjamin-Hochberg procedure and genes
   were filtered for p < 0.01. d Venn diagram representation of the
   exclusively upregulated genes in M14^fl/fl-Ucp1^cre-iBAT versus
   M14^fl/fl-iBAT, commonly upregulated genes, and exclusively upregulated
   genes in M14^fl/fl-Adipoq^cre-iWAT versus M14^fl/fl-iWAT, statistical
   analyses were performed using the Benjamin-Hochberg procedure and genes
   were filtered for FDR < 0.05. e Venn diagram representation of the
   exclusively downregulated genes in M14^fl/fl-Ucp1^cre-iBAT versus
   M14^fl/fl-iBAT, commonly downregulated genes, and exclusively
   downregulated genes in M14^fl/fl-Adipoq^cre-iWAT versus M14^fl/fl-iWAT,
   statistical analyses were performed using the Benjamin-Hochberg
   procedure and genes were filtered for FDR < 0.05. f, g Enriched
   pathways of the upregulated (f) or downregulated g genes in
   M14^fl/fl-Ucp1^cre-iBAT versus M14^fl/fl-iBAT. h, i Enriched pathways
   of the upregulated (h) or downregulated (i) genes in
   M14^fl/fl-Adipoq^cre-iWAT versus M14^fl/fl-iWAT. j Venn diagram
   representation of the m^6A-hypomethylated (p < 0.01) and upregulated
   genes (FDR < 0.05) in M14^fl/fl-Ucp1^cre-iBAT versus M14^fl/fl-iBAT. k
   Top 10 enriched pathways of the m^6A-hypomethylated and differentially
   upregulated genes in M14^fl/fl-Ucp1^cre-iBAT versus M14^fl/fl-iBAT. l
   Venn diagram representation of the m^6A-hypomethylated (p < 0.01) and
   upregulated genes (FDR < 0.05) in M14^fl/fl-Adipoq^cre-iWAT versus
   M14^fl/fl-iWAT. m Top 10 enriched pathways of the m^6A-hypomethylated
   and differentially upregulated genes in M14^fl/fl-Adipoq^cre-iWAT
   versus M14^fl/fl-iWAT. p-values of pathway enrichment analysis were
   calculated according to the hypergeometric test based on the number of
   physical entities present in both the pre-defined set and
   user-specified list of physical entities (f–i, m). Source data are
   provided as a Source Data file.

   Similar comparisons were performed between the differentially expressed
   genes in M14^fl/fl-Ucp1^cre-iBAT and M14^fl/fl-Adipoq^cre-iWAT
   (Fig. [203]5a). Only 106 genes (2.8% of total upregulated genes)
   exhibited overlap between M14^fl/fl-Ucp1^cre iBAT and
   M14^fl/fl-Adipoq^cre iWAT (Fig. [204]5d), and 91 (1.8%) genes were
   commonly downregulated (Fig. [205]5e). Gene set enrichment analysis
   revealed significant divergence in the transcriptional programs of
   METTL14 between BAT and WAT (Fig. [206]5f–[207]i). Upregulated genes of
   M14^fl/fl-Ucp1^cre iBAT were enriched in pathways related to lipid
   metabolism, prostaglandin synthesis, and phospholipid biosynthetic
   processes (Fig. [208]5f), while downregulated genes were associated
   with insulin resistance, white fat cell differentiation, and apoptosis
   pathways (Fig. [209]5g). In contrast, M14^fl/fl-Adipoq^cre iWAT showed
   upregulation of pathways involving immune response, TNFα signaling, and
   apoptosis (Fig. [210]5h), with downregulated genes enriched in lipid
   metabolism, lipolysis, insulin signaling, and phospholipid metabolism
   pathways (Fig. [211]5i).

   Considering that METTL14 is responsible for installing m^6A, the
   depletion of METTL14 is expected to result in the hypomethylation of
   its target transcripts. Moreover, it is established that reduced m^6A
   methylation correlates with enhanced mRNA stability, as it protects the
   target transcripts from degradation^[212]19,[213]21,[214]24. To clarify
   if these transcriptomic changes were governed by METTL14-mediated m^6A
   modification, we intersected m^6A-hypomethylated and differentially
   upregulated genes. These analyses confirmed that lipid and
   phospholipid-related pathways were m^6A-dependently upregulated in
   M14^fl/fl-Ucp1^cre iBAT (Fig. [215]5j, [216]k), while apoptosis and
   TNFα signaling pathways were upregulated in an m^6A-dependent manner in
   M14^fl/fl-Adipoq^cre iWAT (Fig. [217]5l, [218]m).

METTL14 selectively methylates key target transcripts in hBAs versus hWAs

   Given the marked cellular heterogeneity of BAT and WAT
   tissues^[219]25,[220]26, we harnessed hBA and hWA models to delve into
   the cellular-level transcriptional mechanism(s) influenced by METTL14.
   Using the same comparison strategy described above, the intersection of
   m^6A hypomethylated genes in sgM14-hBAs and sgM14-hWA cells revealed
   only 160 (6.3%) commonly hypomethylated and 912 (14.2%) hypermethylated
   genes, along with 559 (11.8%) commonly upregulated and 554 (11.9%)
   commonly downregulated genes (Fig. [221]6a–[222]d).

Fig. 6. METTL14 Selectively Methylates Key Target Transcripts in hBA versus
hWA cells.

   [223]Fig. 6
   [224]Open in a new tab

   a Venn diagram representation of the exclusively m^6A-hypomethylated
   genes in sgM14-hBA versus sgNTC-hBA cells, commonly m^6A-hypomethylated
   genes, and exclusively m^6A-hypomethylated genes in sgM14-hWA versus
   sgNTC-hWA cells, statistical analyses were performed using the
   Benjamin-Hochberg procedure and genes were filtered for p < 0.01. b
   Venn diagram representation of the exclusively m^6A-hypermethylated
   genes in sgM14-hBA versus sgNTC-hBA cells, commonly
   m^6A-hypermethylated genes, and exclusively m^6A-hypermethylated genes
   in sgM14-hWA versus sgNTC-hWA cells, statistical analyses were
   performed using the Benjamin-Hochberg procedure and genes were filtered
   for p < 0.01. c Venn diagram representation of the exclusively
   upregulated genes in sgM14-hBA versus sgNTC-hBA cells, commonly
   upregulated genes, and exclusively upregulated genes in sgM14-hWA
   versus sgNTC-hWA cells, statistical analyses were performed using the
   Benjamin-Hochberg procedure and genes were filtered for FDR < 0.01. d
   Venn diagram representation of the exclusively downregulated genes in
   sgM14-hBA versus sgNTC-hBA cells, commonly downregulated genes, and
   exclusively downregulated genes in sgM14-hWA versus sgNTC-hWA cells,
   statistical analyses were performed using the Benjamin-Hochberg
   procedure and genes were filtered for FDR < 0.01. e, f Enriched
   pathways of the upregulated (e) or downregulated (f) genes in sgM14-hBA
   versus sgNTC-hBA cells. g, h Enriched pathways of the upregulated (g)
   or downregulated (h) genes in sgM14-hWAs versus sgNTC-hWA cells. i Venn
   diagram representation of the m^6A-hypomethylated and upregulated genes
   in sgM14-hWA versus sgNTC-hWA cells, statistical analyses were
   performed using the Benjamin-Hochberg procedure and genes were filtered
   for p < 0.01, and FDR < 0.05, respectively. j Top 10 enriched pathways
   of the m^6A-hypomethylated and differentially upregulated genes in
   sgM14-hWA versus sgNTC-hWA cells. k Schematic depicting experimental
   strategy to knockdown METTL14 in the differentiated hWA cells for m^6A-
   and RNA-seq (Created with BioRender.com). l Venn diagram representation
   of the intersection of the m^6A-hypomethylated and upregulated genes in
   siM14-hWA versus siNTC-hWA cells. Wildtype hWA cells were fully
   differentiated and transfected with siNTC or siM14 to knock down
   METTL14 in the mature hWA cells. Statistical analyses were performed
   using the Benjamin-Hochberg procedure and genes were filtered for
   p < 0.01 and FDR < 0.05, respectively. m Top 10 enriched pathways of
   the m^6A-hypomethylated and upregulated genes in siM14-hWA versus
   siNTC-hWA cells. n Coverage plots of m^6A peaks in TRAIL (I), TNFSF1R
   (m), RIP1K (n), CASP7 (o), DFFA (p) genes in sgM14-hWA versus sgNTC-hWA
   cells. Plotted coverages are the median of the n replicates presented.
   o, p Heatmaps of genes related to TNFα and apoptosis signaling pathways
   (o) and prostaglandins synthesis and regulation pathway (p) (filtered
   by FDR < 0.05) in sgM14-hBAcells versus sgNTC-hBA cells, or sgM14-hWA
   cells versus sgNTC-hWA cells. q A proposed model for the molecular
   mechanism of action that METTL14-mediated m^6A installation selectively
   destabilizes transcripts in brown (left panel) and white adipocytes
   (right panel). p-values of pathway enrichment analysis were calculated
   according to the hypergeometric test based on the number of physical
   entities present in both the pre-defined set and user-specified list of
   physical entities (e–h, j, m). Source data are provided as a Source
   Data file.

   In hBA cells, enrichment analyses of the DGE (differentially gene
   expression) pointed to elevated lipid metabolism and phospholipid
   processes among upregulated pathways, and apoptosis and TNFα signaling
   among the prominent downregulated pathways due to METTL14 deficiency
   (Fig. [225]6e, [226]f). Conversely, in hWA cells, METTL14 deficiency
   predominantly led to m^6A-dependent upregulation of TNFα and apoptosis
   pathways (Fig. [227]6g, i, [228]j), while resulting in the
   downregulation of pathways regulated by EGFR1, focal
   adhension-PK3K-AKT-mTOR, and insulin signaling (Fig. [229]6h).

   Importantly, while the stable knockout cell line enables the conduct of
   several phenotypical and mechanistic studies, the ablation of METTL14
   in preadipocytes may also influence the differentiation process of both
   hWAs and hBAs. To address this potential confounding factor in
   adipocyte differentiation, we performed a transient knockdown of
   METTL14 using siRNA in fully differentiated adipocytes, (Fig. [230]6k).
   Consistently, METTL14 deficiency in mature hWA cells accentuated the
   significance of TNFα and cell death signaling pathways (Fig. [231]6l,
   [232]m). These findings emphasize the ability of METTL14 to exhibit
   divergent roles in shaping the transcriptional profiles of mature brown
   adipocytes to promote a favorable metabolic phenotype. Conversely, in
   the M14^fl/fl-Adipoq^cre model, METTL14 deficiency distinctly
   upregulated genes linked to TNFα and apoptosis in mature white
   adipocytes, contributing to adipose tissue apoptosis and insulin
   resistance. Thus, at both tissue and cellular levels, METTL14
   deficiency leads to an upregulation of genes encoding prostaglandin
   synthases in BAT, and key genes related to TNFα signaling and apoptosis
   in WAT.

   As mentioned above, METTL14 deficiency is expected to induce
   hypomethylation in its target transcripts. Next, we conducted further
   analysis to explore the m^6A-hypomethylated and upregulated genes,
   which are considered METTL14-targeted transcripts, within the most
   enriched pathways of sgM14-hBAs cells and sgM14-hWA cells. Within hBA
   cells, METTL14 specifically destabilizes transcripts associated with
   prostaglandin synthesis^[233]19. In this study, we identified five
   transcripts—necrosis factor-related apoptosis-inducing ligand (TRAIL),
   TNF receptor superfamily member 1 A (TNFR1), receptor-interacting
   serine/threonine kinase 1 (RIP), caspase 7 (CASP7), and DNA
   fragmentation factor subunit alpha (DFFA), as direct targets of METTL14
   in hWA cells (Fig. [234]6n). Interestingly, while METTL14 deficiency
   led to the upregulation of several genes within the TNFα and apoptosis
   pathways in hWA cells, most of these genes exhibited downregulation in
   sgM14-hBA cells (Fig. [235]6o). Notably, transcripts previously
   identified as METTL14 targets in hBA, including PTGES2, CRB1, and
   PGC1A, were downregulated in sgM14-hWA compared to sgNTC-hWA cells
   (Fig. [236]6p). These findings provide strong evidence that METTL14
   regulates distinct transcriptional processes in BAT versus WAT, as
   summarized in Fig. [237]6q.

METTL14 deficiency promotes apoptosis in mouse and human white adipocytes

   We next sought to validate our sequencing outcome and investigate the
   molecular mechanism(s) underlying the METTL14-mediated m^6A decoration
   of target transcripts in brown versus white adipocytes/adipose tissues.

   METTL14 deficiency promotes prostaglandin synthesis in brown
   adipocytes, as reported in our previous study^[238]19. To validate our
   sequencing outcome from mouse and human white adipocytes, which
   indicate increased apoptosis (Figs. [239]5 and [240]6), we conducted
   Terminal deoxynucleotidyl Transferase dUTP Nick End Labeling (TUNEL)
   assays on the iWAT and eWAT isolated from M14^fl/fl-Adipoq^cre mice.
   Our data confirmed a higher percentage of apoptotic cells in both
   depots (Fig. [241]7a–[242]c). Notably, apoptosis was primarily observed
   in PERILIPIN-positive cells (Fig. [243]7a), suggesting
   adipocyte-specific apoptosis induced by METTL14 deficiency.

Fig. 7. METTL14 deficiency promotes apoptosis in mouse and human white
adipocytes.

   [244]Fig. 7
   [245]Open in a new tab

   a Representative pictures of immunofluorescence staining of apoptotic
   nuclei (TUNEL-stain, green), adipocytes (PERILIPIN-stain, red), and
   nuclei (DAPI, blue) in eWAT and iWAT sections collected from male
   M14^fl/fl and M14^fl/fl-Adipoq^cre mice fed with HFD (n = 7/group)
   (scale bar, 50 μm; insert, 2× magnification). b, c Quantification of
   TUNEL-positive nuclei as a percentage of total nuclei in eWAT (b) and
   iWAT (c) of male M14^fl/fl and M14^fl/fl-Adipoq^cre mice fed with HFD
   (n = 7/group). d qRT-PCR analysis of the indicated mRNAs in the iWAT of
   male M14^fl/fl and M14^fl/fl-Adipoq^cre mice (n = 10/group). mRNA
   levels were normalized to Actb. e, f Western blot analysis (e) and
   quantification (f) of the indicated proteins in the iWAT of male
   M14^fl/fl and M14^fl/fl-Adipoq^cre mice. VINCULIN was used as a loading
   control (n = 5/group). g, h ELISA analysis of free fatty acid (FFA) (g,
   n = 10 for M14^fl/fl, and n = 11 for M14^fl/fl-Adipoq^cre) and
   triglyceride (TG) (h, n = 10/group) levels in the serum of male
   M14^fl/fl and M14^fl/fl-Adipoq^cre mice on HFD. i ELISA analysis of
   triglyceride (TG) levels in the liver and muscle of male M14^fl/fl and
   M14^fl/fl-Adipoq^cre mice on HFD (left panels) and iBAT removed cohort
   on CD (right panels) (n = 6/group for liver and muscle from HFD-fed
   mice, and liver from iBAT removed mice; n = 5 for muscle from iBAT
   removed M14^fl/fl mice, and n = 6 for iBAT from iBAT removed
   M14^fl/fl-Adipoq^cre mice). j qRT-PCR analysis of the indicated mRNAs
   in the differentiated sgNTC-hWA and sgM14-hWA cells. mRNA levels were
   normalized to ACTINB mRNA (n = 4/group). k Western blot analysis of the
   indicated proteins in the differentiated sgNTC-hWA and sgM14-hWA cells.
   TUBULIN was used as a loading control (n = 3/group). l ELISA analysis
   of FFA (upper panels, n = 5 for sgNTC, and n = 4 for sgM14) and TG
   (lower panels, n = 7 for sgNTC, and n = 5 for sgM14) levels in the
   culture medium of differentiated sgNTC-hWAs and sgM14-hWA cells. m A
   proposed model illustrating the mechanism of systemic insulin
   resistance induced by WAT-apoptosis in the M14^fl/fl-Adipoq^cre mouse
   model (Created with BioRender.com). All samples in each panel are
   biologically independent. Data are presented as means ± SEM from two or
   three independent experiments by Two-tailed unpaired t-tests (b–d, f–i,
   j, l). Source data are provided as a Source Data file.

   In mouse iBAT and differentiated hBAs, METTL14-mediated m^6A
   installation promotes the decay of transcripts related to prostaglandin
   synthesis in a YTHDF2/3-dependent manner^[246]19. In iWAT, Mettl14
   deficiency leads to upregulation of Trail, Dr5, Tnfα, Tnfr1, Casp3,
   Casp7, and Dffa at the mRNA level (Fig. [247]7d), and increased protein
   abundance of TRAIL and TNFA (Fig. [248]7e, [249]f).

   Considering the ectopic fat deposition in iBAT, liver, and muscle
   (Fig. [250]1m–[251]o), we wondered whether apoptosis of white
   adipocytes increases circulating free fatty acids (FFAs) and/or
   triglyceride (TG) levels, leading to iBAT hypertrophy, hepatic
   steatosis and/or muscle lipid accumulation. Since Mettl14 deficiency
   downregulated genes associated with fatty acid degradation (also
   referred to as lipolysis) and regulation of lipolysis in adipocyte
   pathways (Fig. [252]5h), it is plausible that lipolysis-induced FFA
   production does not contribute to ectopic lipid accumulation. Indeed,
   FFA levels (Fig. [253]7g) were lower while TG levels (Fig. [254]7h)
   were elevated in the serum of M14^fl/fl-Adipoq^cre mice compared to
   control mice. Moreover, the TG level was higher in the liver of HFD-fed
   and CD-fed iBAT-deficient M14^fl/fl-Adipoq^cre mice; as well as higher
   in the liver of HFD-fed Adipoq^cre mice (Fig. [255]7i).

   Consistent results were obtained at the cellular level, as indicated by
   the upregulated gene expression and protein abundance of apoptotic
   markers (Fig. [256]7j, [257]k, respectively), decreased levels of FFAs,
   and heightened TGs in the conditioned media of differentiated sgM14-hWA
   cells (Fig. [258]7l). Of note, the levels of both cleaved caspase 3
   (cCASP3) and caspase 7 (cCASP7) were significantly increased in the
   sgM14-hWA cell line (Fig. [259]7k), providing further support for
   adipocyte-specific apoptosis. These data collectively indicate that
   M14^KO-WAT apoptosis-induced lipid spillover contributes to the
   observed ectopic lipid accumulation and consequent peripheral insulin
   resistance (Fig. [260]7m).

METTL14-Mediated m^6A accelerates the decay of apoptosis-related transcripts
in A YTHDF2/3-dependent manner in white adipocytes

   Next, we conducted mechanistic studies using differentiated hWA cells.
   First, we examined the effects of METTL14 deficiency on the stability
   of m^6A-targeted mRNAs (identified in Fig. [261]6n) following treatment
   with transcription inhibitor actinomycin D (ActD). METTL14 deficiency
   suppressed the degradation of TRAIL and TNFR1 mRNAs without
   significantly affecting the stability of ACTINB mRNA (Fig. [262]8a).
   These findings suggest that METTL14-mediated m^6A modification
   orchestrates hWA apoptosis by regulating the stability of key apoptotic
   transcripts.

Fig. 8. METTL14-mediated m^6A promotes decay of TNFR1 and TRAIL in a YTHDF2/3
dependent manner in white adipocytes.

   [263]Fig. 8
   [264]Open in a new tab

   a qRT-PCR analysis of the TRAIL and TNFR1 mRNA in differentiated
   sgNTC-hWA and sgM14-hWA cells after a time-course treatment with
   100 μg/mL Actinomycin D (Act D). mRNA levels were normalized to ACTINB
   mRNA (n = 3). b Representative Western blot of YTHDF1, YTHDF2, and
   YTHDF3 protein levels in differentiated wildtype hWA cells transfected
   with siNTC, siYTHDF1, siYTHDF2, or siYTHDF3 siRNA. GAPDH was used as a
   loading control (n = 3). c qRT-PCR analysis of TRAIL and TNFR1 mRNA in
   differentiated wildtype hWA cells transfected with siNTC, siYTHDF1,
   siYTHDF2, or siYTHDF3 siRNA following a time-course treatment 100 μg/mL
   Act D. TRAIL and TNFR1 mRNA levels were normalized to ACTIN (n = 4).
   The p-value indicates the significance of differences between any group
   and the siNTC group. d, e Western blot analysis (d) and quantification
   (e) of METTL14 protein abundance in the differentiated hWA cells
   transfected with scramble, wild-type METTL14 (METTL14-WT), or mutated
   METTL14 at R298 (METTL14-MUT) plasmid (n = 3/group). f Relative m^6A
   levels in the differentiated hWA cells transfected with scramble,
   wild-type METTL14-WT, or METTL14-MUT plasmid (n = 3/group). g qRT-PCR
   analysis of TRAIL and TNFR1 mRNA in the differentiated hWA cells
   transfected with scramble, wild-type METTL14-WT, or METTL14-MUT plasmid
   after a time-course treatment with 100 μg/mL Actinomycin D (Act D).
   TRAIL and TNFR1 mRNA levels were normalized to ACTINB mRNA (n = 3). The
   p-value indicates the significance of differences between any group and
   the Scr group. h Scheme of experimental approach depicting M14^fl/fl
   mice receiving AAV8 eGFP (AAV-scramble), M14^fl/fl-Adipoq^cre mice
   receiving AAV8 eGFP, M14^fl/fl-Adipoq^cre mice receiving AAV8 knocking
   down Tnfr1 (AAV-shTnfr1), and M14^fl/fl-Adipoq^cre mice receiving AAV8
   knocking down Trail (AAV-shTrail) (Created with BioRender.com). i, j
   Representative Western blot analysis (i) and quantification (j) of
   METTL14, TNFR1, and TRAIL protein levels in iWAT of the AAV-injected
   mice (n = 3). k Serum TG levels measured by ELISA assays
   (n = 5–6/group). l Intraperitoneal insulin tolerance tests of mice
   injected with AAVs (n = 7 for M14^fl/fl-AAV-Scr group; n = 5 for
   M14^fl/fl-Adipoq^cre_AAV-Scr group, n = 7 for
   M14^fl/fl-Adipoq^cre_AAV-shTnfr1 group, n = 8 for
   M14^fl/fl-Adipoq^cre_AAV-shTrail group,). The p-value indicates the
   significance of differences between any group and the
   M14^fl/fl-Adipoq^cre_AAV-Scr group. m, n Western blot analysis (m) and
   quantification (n) of pIRβ/IGF1Rβ and pAKT[S473] in iWAT 5-week
   post-AAV injections, followed by injection of 1U of insulin via the
   vena cava (n = 2 for non-stimulated groups; n = 4 for
   insulin-stimulated groups). The p-value indicates the significance of
   differences between any group and the M14^fl/fl-Adipoq^cre_AAV-Scr
   group. All samples in each panel are biologically independent. Data are
   presented as means ± SEM from two or three independent experiments by
   Two-tailed unpaired t-test (a), One-way ANOVA (c, g), and Two-way ANOVA
   (e, f, j, k, l, n). Source data are provided as a Source Data file.

   To further confirm this hypothesis, we performed independent
   siRNA-mediated knockdown of the m^6A reader proteins YTHDF1, YTHDF2, or
   YTHDF3 (Fig. [265]7i), which are known to regulate mRNA
   decay^[266]24,[267]27, and examined the expression and stability of
   TRAIL and TNFR1 mRNAs after ActD treatment in wild type hWA cells.
   Knockdown of YTHDF1, YTHDF2, or YTHDF3 (Fig. [268]8b) followed by ActD
   treatment significantly increased the stability of TRAIL and TNFR1
   (Fig. [269]8c). These data collectively confirmed that METTL14-mediated
   m^6A installation negatively regulates the stability of TRAIL and TNFR1
   mRNAs in white adipocytes.

   Next, to substantiate the implication that METTL14 is linked to the
   modulation of mRNA stability of apoptosis-associated genes, we explored
   this possibility with a gain-of-function model by overexpressing
   METTL14 in hWA cells (Fig. [270]8d, [271]e). Additionally, to
   investigate whether METTL14 regulates key target mRNA stability in an
   m^6A-dependent manner, we established a hWA cell line harboring the
   METTL14 R298 mutation (Fig. [272]8d, [273]e). As expected,
   overexpression of METTL14 increased global m^6A levels, whereas mutated
   METTL14 had no impact (Fig. [274]8f). Importantly, while METTL14
   upregulation led to decreased stability of TRAIL and TNFR1 through m^6A
   hypermethylation, their stability was not affected by mutated METTL14
   in hWA cells (Fig. [275]8g). These findings support that METTL14
   directly regulates TRAIL and TNFR1 in an m^6A-dependent manner.

   Finally, to validate our identification of TRAIL and TNFR1 as drivers
   of systemic insulin resistance in M14^fl/fl-Adipoq^cre mice, we used
   AAV8 vectors to genetically knock down Trail or Tnfr1 specifically in
   iWAT by injecting AAV8-scramble, AAV8-shTrail, or AAV8-shTnfr1
   (Fig. [276]8h). Knockdown efficiency was confirmed by reduced TRAIL or
   TNFR1 protein levels in iWAT (Fig. [277]8i, [278]j). Notably, knockdown
   of each gene significantly decreased serum triglyceride (TG) levels in
   M14^fl/fl-Adipoq^cre mice (Fig. [279]8k), affirming that TRAIL and
   TNFR1 triggered adipocyte apoptosis and subsequent lipid spillover.
   Additionally, depletion of either Trail or Tnfr1 partially reversed
   insulin resistance in M14^fl/fl-Adipoq^cre mice at both the systemic
   and tissue levels (Fig. [280]8l–[281]n, respectively).

Discussion

   An accumulating body of evidence points to the importance of m^6A mRNA
   modifications in orchestrating pathways that underlie intricate
   mechanisms that contribute to metabolic disorders^[282]28. However, the
   current gap in our understanding pertains to the cell-type-specific
   regulation by m^6A within various biological domains, including
   metabolism. Here, we provide one example of the divergent roles of
   METTL14-mediated m^6A modifications in orchestrating the metabolic
   functions of brown versus white adipocytes. By integrating m^6A-seq and
   RNA-seq obtained from both mouse BAT/WAT and human brown/white
   adipocytes, we unveil cell-type specific m^6A methylomes and
   transcriptomes. Our study identifies specific transcripts through which
   METTL14 exerts dual regulatory roles: on one hand, negatively
   modulating the secretory functions of BAT while on the other
   simultaneously enhancing the lipid sequestration capabilities of WAT.
   This multifaceted role of METTL14 in shaping adipocyte-mediated
   metabolism adds a new layer of nuance to the field of metabolic
   regulation.

   Notably, while the M14^fl/fl-Ucp1^cre model was featured in our recent
   report^[283]19, the current study includes newly generated data from a
   different cohort of animals in this model, which is essential as it
   provides a critical BAT-specific knockout for comparison with the
   overall-adipose knockout (M14^fl/fl-Adipoq^cre) model. This approach
   allows us to differentiate BAT-specific effects from those in total
   adipose tissue. Furthermore, the consistency of findings in different
   cohorts of animals demonstrates the reproducibility and robustness of
   our experimental approach, further validating the significance of our
   results.

   A study recently reported by Kang and colleagues^[284]29 also used the
   M14^fl/fl-Adipoq^cre mouse model to create an adipose-specific
   Mettl14-knockout. While our study, as well as the study by Kang et al.,
   observed decreased body weight, impaired cold tolerance, decreased WAT
   mass, and increased BAT mass, several obvious differences stand out
   when comparing the two reports: First, while the major findings in the
   Kang et al. study exclusively rely on the Mettl14^fl/fl-Adipoq^cre
   model, it was surprising that no information regarding the origin of
   Mettl14^fl/fl mice and the KO targeting strategy (e.g., which exons are
   floxed) was provided, which limits the comparison between our KO model
   and theirs. On the other hand, we and several other groups have
   characterized our Mettl14^fl/fl mice^[285]21,[286]30. Second, in their
   study, Kang et al surprisingly do not use m^6A-seq but bulk RNA-seq to
   determine targets that contribute to the phenotype. In contrast, we
   used m^6A-MeRIP- and RNA-seq as drivers to determine the mechanism(s)
   that underlie the phenotype. Considering METTL14’s non-m^6A-dependent
   effects, it is likely the approach by Kang et al. introduced bias in
   the identification of m^6A-targeted genes. Third, in their study, Kang
   et al. attribute decreased body weight, enhanced insulin sensitivity,
   and improved glucose tolerance to heightened cell-autonomous lipolysis
   resulting from Mettl14 deficiency. This conclusion is inconsistent with
   the large body of literature demonstrating that excessive lipolysis,
   particularly in white adipocytes, contributes to adipose tissue- and
   systemic “insulin resistance”^[287]31–[288]34. The exacerbated insulin
   resistance is secondary to the influx of elevated circulating FFAs into
   non-adipose tissues and the release of proinflammatory
   cytokines^[289]31, both of which interfere with insulin signaling. Kang
   et al. noted increased circulating FFAs, improved hepatic insulin
   sensitivity and steatosis in their model, while reduced Tnfα and Il6
   gene expression in M14^KO adipose tissue, confounding the overall
   interpretation of the phenotype^[290]29. Finally, insulin usually
   inhibits lipolysis via Akt activation^[291]35,[292]36, yet the authors
   suggested enhanced insulin sensitivity and lipolysis synergy, creating
   interpretive challenges. In contrast, our study reveals the
   downregulation of lipolysis pathways in the Mettl14^fl/fl-Adipoq^cre
   iWAT (Fig. [293]5i), aligning with a consistent decrease in circulating
   FFAs (Fig. [294]7i). Mettl14 deficiency in WAT induces insulin
   resistance in a tissue-autonomous manner. This, in turn, leads to
   severe peripheral insulin resistance through adipocyte
   apoptosis-induced triglyceride spillover and systemic inflammation.
   Nevertheless, other factors, including diets, and differences in
   microbiome composition secondary to housing conditions should be
   considered when interpreting the discrepancies of the two studies.

   Our study identifies specific transcripts by which METTL14 negatively
   governs the secretory function within BAT and positively regulates the
   lipid sequestration capability within WAT in an m^6A-dependent manner.
   This cell-type-specific regulation by METTL14-mediated m^6A can be
   attributed to various factors. The level of m^6A modification depends
   on the expression and abundance of m^6A regulators including METTL14
   and METTL3, the intrinsic preference of METTL3/METTL14
   methyltransferase for specific nucleotide sequences, and the extrinsic
   regulation of methyltransferase complex activity by transcription
   factors, RNA-binding proteins (PBPs), RNA polymerases, histone
   modification, or posttranslational modifications^[295]37–[296]39.

   The cell-type specificity of m^6A modifications is highlighted by the
   differential expression and protein abundance of METTL14 across various
   cell types. Specifically, under physiological conditions, METTL14
   exhibits significantly lower protein levels of iBAT and hBAs compared
   to iWAT and hWAs, respectively (Supplementary Fig. [297]1f–[298]h).
   Despite METTL3 being the enzyme with catalytic activity, its function
   is critically dependent on forming a complex with METTL14^[299]40.
   Therefore, the observed variability in METTL14 abundance may lead to
   differences in the activity of the m^6A writer complex and substrate
   availability. This, in turn, could result in distinct m^6A target
   preferences, influencing the physiological functions of brown versus