m⁶A mRNA Methylation Controls Functional Maturation in Neonatal Murine β-Cells

Mature β-cells have the unique ability to secrete insulin in response to extracellular glucose concentrations (1). Defects in either β-cell number and/or function can lead to loss of functional β-cell mass and, eventually, to diabetes (2). Establishment of adequate functional β-cell mass is essential for glycemic control in adulthood, which is mainly accomplished during neonatal period before weaning by β-cell differentiation, self-replication, and functional maturation (3,4). Neonatal β-cells are immature and are unable to secrete insulin appropriately in response to a glucose challenge (5). During the early postnatal period (P0-P14), neonatal β-cells undergo dramatic changes to acquire glucose-responsive insulin secretion ability and gain functional maturity (3,6,7). Previous studies have identified orchestrated mechanisms including molecular pathways (8), transcriptional signals (9–12), and epigenetic regulators (5,13) in shaping the transcriptional networks that reinforce the functional identity of mature β-cells. Epigenetic regulators, such as DNA methylation and histone modifications, have profound effects on β-cell development and functional maturation (5,13). We and others have reported that DNMT3A directs functional maturation in murine β-cells (5,8). Meanwhile, genome-wide analysis of histone marks during β-cell development further underscored the importance of histone modifications in β-cell maturation (13). N⁶-methyladenosine (m⁶A) modification of mRNA is emerging as an important regulator of gene expression (14,15). However, its role in neonatal β-cells is still unknown.

Being the most prevalent internal modification in mRNA, the m⁶A mRNA modification functionally modulates mRNA splicing, export, localization, translation, and stability and has crucial roles in various normal and pathological processes (16,17). m⁶A methylation is dynamic and is accomplished by the orchestrated action of methyltransferase complex methyltransferase-like 3 (Mettl3), methyltransferase-like 14 (Mettl14), and Wilms tumor 1-associated protein (WTAP), while it is removed by demethylases obesity-associated protein (FTO) and AlkB homolog 5 (ALKBH5) (18). In particular, reversible m⁶A mRNA methylation plays critical roles in modulating dynamic transcriptome switching in response to diverse signals during embryonic development of various organs and organisms (17). As the key writers of m⁶A, Mettl3 and/or Mettl14 have been identified to regulate Drosophila neuronal function and sex determination (19), mouse stem cell pluripotency (20), naive T cell differentiation (21), cortical neurogenesis (22), spermatogenesis (23,24), and hematopoietic stem/progenitor cell specification (25). Mettl3 and or Mettl14 were also reported to participate in the pathology and progression of multiple human cancer diseases, i.e., myeloid leukemia (26), liver cancer (27), lung cancer (28), and breast cancer (29). Very recently, m⁶A modification was shown to regulate mature β-cell insulin secretion and survival (30,31). It is currently unknown whether and how m⁶A RNA modifications regulate embryonic and early postnatal islet development.

To investigate the role of Mettl3/14-mediated m⁶A modifications in islet development, we generated mice in which Mettl3 and/or Mettl14 were deleted in Ngn3⁺ endocrine progenitors. Loss of Mettl3/14 in endocrine progenitors (Mettl3/14^nKO) caused hyperglycemia at 2 weeks of age, with reduced cell number and impaired functional maturation of neonatal β-cells. Transcriptome and m⁶A methylome analyses of primary islets from P14 wild-type (WT) and Mettl3/14^nKO mice identified m⁶A-dependent mechanisms in regulating neonatal β-cell identity and insulin secretion. Loss of Mettl3/14 silenced MafA expression, possibly via modulating its mRNA stability and, thus prevented neonatal β-cell to fulfill functional maturation. Importantly, as early as P0, reduced MafA protein abundance and impaired glucose-stimulated insulin secretion (GSIS) were already present in mutant islets, which were prior to β-cell loss and changes in other transcription factors, i.e., Pdx1 and Nkx6.1. The findings that Mettl3 and Mettl14 expression were significantly decreased in pancreatic β-cells of db/db mice and patients with type 2 diabetes (T2D) further suggested the possible role of m⁶A in T2D in both rodents and humans.

Mettl3- and Mettl14-floxed mice (Mettl3^flox/flox and Mettl14^flox/flox) were generated and provided by Prof. Ming-Han Tong (University of Chinese Academy of Sciences) (24). Ngn3-cre mice were a kind gift from Prof. Wei-Zhen Zhang. The Ngn3-cre;Mettl3^flox/flox (Mettl3^nKO) or Ngn3-cre;Mettl14^flox/flox (Mettl14^nKO) mice were generated by crossing Mettl3^flox/flox or Mettl14^flox/flox mice with Ngn3-Cre mice. Mettl3- and Mettl14- double knockout (KO) mice (Mettl3/14^nKO) were generated by crossing Mettl3^flox/flox Mettl14^flox/flox mice with Ngn3-cre;Mettl3^flox/+ Mettl14^flox/+ mice. Ngn3-Cre, Mettl3^flox/flox, Mettl14^flox/flox, or Mettl3^flox/flox Mettl14^flox/flox mice were used as their littermate controls. All mice were housed in the animal facility on a 12-h/12-h light/dark cycle. Normal chow and water were available ad libitum. Blood glucose concentrations were measured by glucometers, and plasma insulin concentrations were determined by using an ELISA kit (Mouse Ultrasensitive Insulin ELISA kit; Alpco). Intraperitoneal glucose tolerance tests were performed on 8-week-old mice after overnight fasting as previously described (8). Insulin tolerance tests were performed on 4-week-old Mettl3/14^nKO mice and WT mice after 6 h of fasting. Male mice were used in all the experiments in the current study, unless otherwise stated. All animal experiments were approved by the Animal Care Committee of Shanghai Jiao Tong University.

Paraffin sections of pancreas far from the margin of the pancreatectomy were collected from our previous research (32). In brief, all patients with partial pancreatectomy performed in Ruijin Hospital between 2013 and 2017 were enrolled. Those who had been reported as having a malignant tumor were excluded, and then five patients with T2D and five age- and BMI-matched subjects who did not have diabetes (nondiabetic [ND]) were finally included in this study. Detailed information and clinical characteristics for each patient were listed in Supplementary Table 1. This study was approved by the Institutional Review Board of the Ruijin Hospital affiliated to Shanghai Jiao Tong University School of Medicine and was in accordance with the principles of the Declaration of Helsinki II.

The mouse insulinoma cells (MIN6 cell line) purchased from CAMS Cell Culture Center (Beijing, China) were grown in DMEM medium (Gibco) containing 25.0 mmol/L glucose, 15% FBS, 100 IU/mL penicillin, 100 μg/mL streptomycin, 10.2 mmol/L L-glutamine, and 2.5 mmol/L β-mercaptoethanol at 37°C in a humidified 5% CO₂ atmosphere. For knocking down Mettl3 and Mettl14, shRNA lentiviruses that targeted Mettl3, Mettl14, or a control lentivirus were constructed, packaged, purified, and titrated at GeneChem Co. Ltd. MIN6 cells were infected with purified lentivirus at 50 multiplicity of infection for 48 h. After infection, cells were harvested for RNA extraction and further analysis.

Pancreas sections were dissected, fixed, and processed as described before (33). For the immunostaining analysis, entire pancreatic tissues were continuously sectioned at 5-μm thickness. Immunochemistry procedures were performed on continuous sections (selected every 150-μm apart, 10–12 sections per animal) to obtain representative β-cell mass information of the whole pancreas. Immunochemistry staining of insulin for β-cell mass analysis was performed using a diaminobenzidine peroxidase substrate kit (Vector Laboratories, Burlingame, CA) counterstaining with eosin. Digital images of whole pancreas were captured by MZ 100 microscope (Nikon Corp., Tokyo, Japan). Total pancreatic and insulin-positive areas of each section were measured using Meta-Morph version 7.1 (Molecular Devices, Sunnyvale, CA).

P0 mouse islets and MIN6 cells were lysed, quantified, and blotted as described before. Primary antibodies are listed as following: rabbit anti-MAFA (1:1,000; Bethyl); rabbit anti-METTL3 (1:1,000; Proteintech); rabbit anti-METTL14 (1:1,000; Proteintech); rabbit anti-Pdx1 (1:1,000; CST); mouse anti-Nkx6.1 (1:1,000; DSHB); and mouse anti-GAPDH (1:10,000; Proteintech) was used as an internal control to normalize band intensity.

Total cell RNA was extracted using the TRIzol reagent (Invitrogen). Reverse transcription and quantitative real-time PCR were performed as previously described (8). PCRs were performed in duplicate. The expression levels were normalized to individual β-actin. Primers used in this study were listed in Supplementary Table 2.

Islets were isolated from 8-week-old Mettl3^nKO, Mettl14^nKO, Mettl3/14^nKO, and WT mice, and then, total RNA was extracted using a RNeasy Micro kit (Qiagen) following the manufacturer’s protocol. The change of global m⁶A levels in mRNA was measured using EpiQuik m⁶A RNA Methylation Quantification Kit (Colorimetric) (Epigentek) following the manufacturer’s protocol; 200-ng poly-A-purified RNA was used for each sample analysis.

The mRNA half-life measurements were performed according to previous publications (34). In brief, MIN6 cells with or without Mettl3/14 knockdown were treated with 5-μg/mL actinomycin D (Sigma) for 1 or 3 h at the end of culture and then collected for RNA extraction and real-time PCR analysis. Since actinomycin D treatment results in transcription stalling, the change of mRNA concentration at a given time (dC/dt) is proportional to the constant of mRNA decay (K_decay) and mRNA concentration (C).

Islet total RNA was prepared as previously described. The m⁶A mRNA immunoprecipitation sequencing (MeRIP-seq) and RNA sequencing (RNA-seq) were performed by Cloudseq Biotech Inc. (Shanghai, China) according to published procedures. Briefly, m⁶A RNA immunoprecipitation was performed with the GenSeq m⁶A-MeRIP Kit (GenSeq Inc., Shanghai, China) following the manufacturer’s instructions. Both the input sample without immunoprecipitation and the m⁶A IP samples were used for library generation with NEBNext Ultra II Directional RNA Library Prep Kit (New England Biolabs, Inc.). The library quality was evaluated with BioAnalyzer 2100 system (Agilent Technologies, Inc.). Library sequencing was performed on an illumina Hiseq instrument with 150 base pair (bp) paired-end reads. Paired-end reads were harvested from the Illumina HiSEq 4000 sequencer, and they were quality controlled by Q30 after 3′ adaptor trimming and removal of low-quality reads by cutadapt software (v1.9.3). First, clean reads of all libraries were aligned to the reference genome (MM10) by Hisat2 software (v2.0.4). Methylated sites on RNAs (peaks) were identified by MACS software. Differentially methylated sites were identified by diffReps. These peaks identified by both software overlapping with exons of mRNA were figured out and chosen by homemade scripts. For RNA-seq, total RNA was used for removing the rRNAs with NEBNext rRNA Depletion Kit (New England Biolabs, Inc.) following the manufacturer’s instructions. RNA libraries were constructed, and libraries were controlled for quality and quantified. Library sequencing was performed on an illumina Hiseq instrument with 150 bp paired end reads. Then, guided by the Ensembl gtf gene annotation file, cuffdiff software (part of cufflinks) was used to get the gene-level FPKM (fragments per kilobase of exon model per million mapped reads) as the expression profiles of mRNA. Fold change and P value were calculated based on FPKM.

The exact sample size for each experiment was indicated in the figure legends. All statistics comparing two groups used two-sided Student t tests. ANOVA was used for multiple groups. Statistical analyses were performed with GraphPad Prism 7. P < 0.05 was considered as statistically significant.

The P0 and P14 RNA-seq and m⁶A MeRIP-seq data are deposited at the Gene Expression Omnibus (GEO) database under the accession numbers: GSE149193, GSE132323, and GSE132319.

To examine whether Mettl3/14-mediated m⁶A mRNA methylation participates in β-cell development, we first checked the expression patterns of Mettl3 and Mettl14 in mouse pancreas at E17.5, P0, P4, P8, P14, and P56 by immunostaining (Fig. 1A). We found relatively weak Mettl3 and Mettl14 expressions in β-cells at E17.5 and P0, whereas the immunofluorescence intensities of both the two methyltransferases increased dramatically during β-cell maturation window P4-P8, reached maximal levels at P14, and maintained at high levels throughout adulthood (Fig. 1B and C). These data suggested that Mettl3- and Mettl14-mediated m⁶A mRNA methylation might participate in postnatal β-cell functional maturation.

We then generated mice in which Mettl3 (Ngn3-cre;Mettl3^flox/flox, named as Mettl3^nKO), Mettl14 (Ngn3-cre;Mettl14^flox/flox, named as Mettl14^nKO) or both Mettl3/Mettl14 (Ngn3-cre;Mettl3^flox/flox/Mettl14^flox/flox, named as Mettl3/14^nKO) were specifically ablated in Ngn3⁺ endocrine progenitors (Fig. 1D). The Ngn3-Cre, Mettl3^flox/flox, Mettl14^flox/flox, and Mettl3^flox/flox/Mettl14^flox/flox mice were all healthy and phenotypically normal, and thus, they were jointly used as WT littermate controls in the following experiments. Successful KO of Mettl3 and Mettl14 were confirmed by immunostaining: both METTL3 and METTL14 proteins were selectively absent in endocrine cells of P14 Mettl3/14^nKO mice (Fig. 1E). Moreover, m⁶A/A% levels in RNA extracted from Mettl3^nKO and Mettl14^nKO islets were significantly reduced compared with WT controls, and this decrease was more pronounced in Mettl3/14^nKO islets (Fig. 1F).

At 2 weeks after birth, Mettl3^nKO and Mettl14^nKO had comparable random blood glucose levels as WT controls (Fig. 1G). However, Mettl3^nKO and Mettl14^nKO began to show significant increases in random blood glucose levels at 4 weeks (14.34 ± 3.26 vs. 9.95 ± 1.18 mmol/L in WT, P < 0.01) and 3 weeks (13.22 ± 1.28 vs. 9.77 ± 1.70 mmol/L in WT, P < 0.01) of age, respectively; and their blood glucose rose gradually with age (Fig. 1G). At 8 weeks of age, both Mettl3^nKO and Mettl14^nKO showed a dramatic increase in glycemia after intraperitoneal glucose injection (Fig. 1H and I). Notably, a more severe phenotype was observed in double KO mice: at 2 weeks of age, Mettl3/14^nKO mice already showed a slight increase in random blood glucose levels (9.00 ± 1.45 vs. 7.88 ± 1.22 mmol/L in WT, P = 0.07), and they developed severe hyperglycemia as early as 3 weeks of age (random blood glucose levels >20 mmol/L) and had persistently severe hyperglycemia (>30 mmol/L) thereafter (Fig. 1G). Moreover, plasma insulin levels in 2-week-old Mettl3/14^nKO mice was reduced by 70%, suggesting β-cell failure predominated in these diabetic mutants (Fig. 1J). None of the mutants showed differences in body weight, compared with their age-matched controls (Fig. 1K). Our data also showed that insulin sensitivity was not impaired in 4-week-old male Mettl3/14^nKO mice compared with WT (Supplementary Fig. 1A). Female Mettl3/14^nKO mice were also diabetic at 2 weeks of age and manifested hyperglycemia thereafter (Supplementary Fig. 1B).

We performed immunostaining against four endocrine hormones including insulin (β), glucagon (α), somatostatin (δ), and pancreatic polypeptide (PP) on P0 pancreas from WT and Mettl3/14^nKO. We found similar cord-like islet structures in newborn Mettl3/14^nKO and WT mice (Fig. 2A). Then, we calculated the number of four different islet endocrine cells per section and found the absolute number of β, α, δ, and PP cells remained constant at birth between the two groups (Fig. 2B). Interestingly, after 14 days, Mettl3/14^nKO islets exhibited significantly reduced β-cell number per section (369.6 ± 45.1 vs. 819.4 ± 217.2 in WT, P < 0.05) (Fig. 2C and D). On the contrary, no differences were detected in the number of α, δ and PP cells per section between the two groups (Fig. 2C and D). The above data indicated a preferential role of Mettl3/14 in β-cell development during early postnatal period (P0-P14). In parallel, β-cell mass was 30% lower in Mettl3/14^nKO mice than in WT, while α-cell mass remained unchanged (Fig. 2E and F). The reduction in β-cell mass was attributed to changes in both proliferation and apoptosis: 40% reduction in the proportion of Ki67⁺insulin⁺ cells (Fig. 2G and Supplementary Fig. 2A) and a twofold increase in the percentage of TUNEL⁺insulin⁺ cells (Fig. 2H and Supplementary Fig. 2B) were found in 2-week-old Mettl3/14^nKO. β-Cell size was unchanged in mutant islets (Fig. 2I and Supplementary Fig. 2C). Taken together, Mettl3/Mettl14 specifically regulated β-cell number via controlling neonatal β-cell proliferation and survival and, thus, determined postnatal β-cell mass formation during P0-P14.

Neonatal β-cells are immature and undergo a dramatic change in the early postnatal period to become functional mature and acquire glucose responsiveness (3). We then checked expression levels of the critical transcription factors required for β-cell identity and function in WT and Mettl3/14^nKO at P14. We found that Pdx1, MafA, and Nkx6.1 were absent in a vast majority of Mettl3/14-deficient β-cells at P14 (Fig. 3A–C). We then calculated the percentages of Pdx1⁺insulin⁺, MafA⁺insulin⁺, and Nkx6.1⁺insulin⁺ β-cells in P14 Mettl3/14^nKO and WT pancreas and found that they decreased about 63.3%, 49.0%, and 78.8%, respectively, in P14 mutants (Fig. 3D–F). In parallel, the expression of Urocortin 3 (Ucn3) (3), a molecular marker of mature β-cells, was present in a vast majority of insulin⁺ β-cells in WT, but it was remarkably decreased in P14 mutant β-cells (Fig. 3G). We also detected reduced expression of GLUT2 (35) in P14 Mettl3/14^nKO (Fig. 3H). Then, we isolated primary islets from P14 WT and Mettl3/14^nKO to evaluate their function in vitro. P14 Mettl3/14^nKO islets displayed impaired insulin release at both basal (2.8 mmol/L) and high glucose (16.7 mmol/L) levels (Fig. 3I). We further studied β-cell ultrastructure in P14 Mettl3/14^nKO and WT pancreas using transmission electron microscopy (Fig. 3J). On the basis of the recognized criterion (36), we quantified the secretory vesicle subsets including mature (with dense core granules, red arrow), immature (with light core granules, green arrow), and empty vesicles (with no granules, blue arrow) (Fig. 3J). The proportion of mature vesicles was significantly decreased and the ratios of immature and empty vesicles were significantly increased in Mettl3/14-deficient β-cells (Fig. 3K). These results indicated that Mettl3/14 deficiency led to the loss of β-cell identity, impaired glucose-induced insulin release, and defections in secretory vesicle maturation in neonatal pancreatic β-cells. Moreover, we detected substantial Aldh1a3⁺ cells in 8-week-old Mettl3/14^nKO islets (Fig. 3L). Meanwhile, we also performed immunostaining against chromogranin A (CGA) and four endocrine cell types (insulin [ins], glucagon [Gcg], somatostatin [Sst], and PP) in 8-week-old Mettl3/14^nKO and WT pancreas (Fig. 3M). We observed a dramatic increase in the percentage of dedifferentiated cells (islet hormone-negative and CGA-positive cells: 30.8 ± 9.0% vs. 3.1 ± 1.5%, Mettl3/14^nKO vs. WT, P < 0.001, yellow arrows in Fig. 3M) in 8-week-old Mettl3/14^nKO pancreas sections. These results indicated that β-cells underwent dedifferentiation in 8-week-old Mettl3/14^nKO mice.

To investigate the underlying mechanisms, we performed RNA-seq analysis on P14 WT and Mettl3/14^nKO islets. We detected significant differences in the expression levels of 544 genes (fold change >1.5, P < 0.05), among which 294 were upregulated and 250 were downregulated (Fig. 4A). We further analyzed alterations of known pathways using the ingenuity pathway analysis and listed the top downregulated and upregulated signaling pathways in P14 Mettl3/14^nKO islets (Fig. 4B). Pathways related to insulin secretion, ion transportation, and response to glucose were among the top downregulated cellular biological processes, while the Wnt-signaling pathway, negative regulation of intracellular signal transduction, and negative regulation of cell proliferation pathway were significantly upregulated in mutant islets (Fig. 4B and C). Interestingly, several crucial β-cell transcription factors Pdx1, Nkx6.1, and MafA were downregulated at P14, among which MafA was the most significantly repressed gene after loss of Mettl3/14 (Fig. 4C). Moreover, a cluster of genes involved in ion transportation and insulin secretion were reduced in mutant islets (i.e., Fkbp1b, Ero1lb, Sytl4, Maob, Nnat, and Trpm5) (Fig. 4C). Loss of Mettl3/14 also affected expressions of genes important for β-cell proliferation, including downregulation of several positive proliferation regulators (i.e., Nkx6.1, Rtkn2, Nasp, and Plk3) and upregulation of some proliferation inhibitors (i.e., Sfrp5, lims2, and Klf11) (Fig. 4C). The dramatic reduction in Pdx1, Nkx6.1, MafA, and Trpm5 mRNA levels were still observed in 8-week-old Mettl3/14^nKO islets (Supplementary Fig. 3A). The above results reinforced that methyltransferase-like Mettl3/14 indeed regulated the expression of those critical β-cell genes.

To identify the direct effects of Mettl3/14, we treated MIN6 cells with Mettl3 and Mettl14 shRNA lentiviruses. Successful knockdown of Mettl3 and Mettl14 significantly reduced mRNA expression of β-cell transcription factors Pdx1, Nkx6.1, and MafA (Supplementary Fig. 4A), increased expression of proliferation inhibitors (Sfrp5, Tlr2, Frzb, Lims2, and Klf11) (Supplementary Fig. 4B), diminished expression of genes related to ion transportation and insulin secretion (Ero1lb, Nnat, Pcsk9, Trpm5, Fkbp1b, Arrb1, Syt3, Flna, Sytl4, and Kcng3) (Supplementary Fig. 4C). Taken together, both in vivo and in vitro, we showed that loss of Mettl3/14 impaired genes important for β-cell identity, insulin secretion, and proliferation.

To obtain insights into the transcriptome-wide m⁶A distribution in neonatal β-cells, we performed MeRIP-seq analysis using P14 mouse islets. Compared with the unbound fractions, 8,987 methylation sites (corresponding to 4,873 transcripts; fold change >2) were significantly enriched in m⁶A antibody-bound fractions and, thus, were identified as high-confidence m⁶A targets. m⁶A peaks in islet transcripts were significantly enriched in GGACU motif (Fig. 4D). m⁶A was distributed throughout islet mRNA transcripts and were abundant predominantly in the coding sequences (Fig. 4E). Gene Ontology (GO) analysis showed m⁶A-containing mRNAs were enriched in biological processes related to RNA Polymerase II Transcription, cell projection organization, RNA Metabolism, protein ubiquitination, mRNA metabolic process, and cell morphogenesis involved in differentiation (Fig. 4F).

To identify the direct m⁶A-modified targets, we correlated MeRIP-seq results with RNA-seq data identified from P14 islets. The correlation analysis eventually identified 170 differentially expressed genes with m⁶A modifications (Fig. 4G). Several m⁶A-modified genes which were important for β-cell function, such as MafA (11), Maob (37), Fkbp1b (38), and Kcng3 (39) were identified as direct m⁶A targets, and their expressions were significantly downregulated in neonatal Mettl3/14^nKO islets (Fig. 4H and Supplementary Fig. 5A–C). On the contrary, Sfrp5 (40), a well-established proliferation inhibitor in β-cells, was upregulated in mutants (Supplementary Fig. 5D).

We listed the top five downregulated and upregulated m⁶A-modified genes that were preferentially changed in neonatal Mettl3/14^nKO mice (Fig. 4H), among which MafA was identified as the most significantly reduced gene (Fig. 4H). Integrative Genomics Viewer tracks showed MafA transcripts had abundantly enriched m⁶A peaks spread its mRNA, which indeed had the highest methylation score among all enriched transcripts (Fig. 4H and I). On the contrary, the other two preferentially changed transcription factors, Pdx1 and Nkx6.1, showed no m⁶A modifications in murine β-cells as identified in our MeRIP-seq.

To determine the direct effect of Mettl3/14 on MafA protein expression, we then treated MIN6 cells with ShMettl3/14 virus for 48 h and found reduced MafA protein abundance after loss of Mettl3 and/or Mettl14 in MIN6 cells (Fig. 4J). These results supported that Mettl3/14 directly regulated MafA expression in β-cells.

It is known that m⁶A modifications on mRNA transcripts might affect mRNA stability and translation (17). We then treated control or Mettl3/14-deficient MIN6 cells with transcription inhibitor actinomycin D and checked the role of m⁶A on MafA mRNA decay. We found that knockdown of Mettl3/14 significantly decreased the mRNA stability of MafA (Fig. 4K), but had no effect on Pdx1, Nkx6.1 mRNA stability (Supplementary Fig. 6A and B). These findings indicated that Mettl3/14 directly regulates MafA expression in neonatal β-cells, at least partly, via modulating its mRNA stability.

To exclude the possibility that Mettl3/Mettl14 inactivation compromised β-cell differentiation that contributed to later β-cell functional immaturity, comprehensive gene expression analyses on P0 islets were performed. At birth, mutant mice showed comparable blood glucose levels (5.37 ± 0.98 vs. 4.23 ± 0.92 mmol/L in WT, P = 0.17) and islet composition (Fig. 2A and B) compared with WT. RNA-seq on isolated P0 islets identified 945 differentially expressed genes (fold change >1.5, P < 0.05), among which 191 were upregulated and 754 were downregulated (Fig. 5A). Pathway analysis identified that pathways related to the response to external stimulus, developmental process, metabolic process, and regulation of secretion were downregulated, while pathways like response to stress, defense response, and the apoptosis process were significantly upregulated (Fig. 5B). We further compared these genes with the MeRIP-seq database and identified 326 differentially expressed genes with m⁶A modifications (Fig. 5C and D). Among them, genes related to ion transportation and insulin secretion (i.e., Fxyd5, Slc3a2, Vamp8, and Syt9) were significantly downregulated (Fig. 5D).

Importantly, although similar expression levels of MafA, Nkx6.1, and Pdx1 mRNA were observed in RNA-seq, by in situ immunostaining, we still detected significantly reduced MafA protein abundance in insulin-positive cells of P0 Mettl3/14^nKO islets, whereas no reductions in protein levels were detected in Pdx1 and Nkx6.1 at this early time point (Fig. 5E). We also found slightly diminished GLUT2 expression in P0 Mettl3/14^nKO β-cells (Fig. 5E). However, Ucn3 was rarely expressed in new-born β-cells (Supplementary Fig. 7A). Moreover, we performed Western blot on isolated islets from P0 WT and mutant mice. Again, we found reduced MafA protein abundance in P0 Mettl3/14-deficient islets (Fig. 6F and H), whereas the expression levels of Pdx1 and Nkx6.1 were comparable to that of WT islets (Fig. 6G and H). The above data indicated that m⁶A-induced MafA change was prior to that of Pdx1 and Nkx6.1 at P0. Indeed, we found impaired glucose responsive insulin secretion ability in P0 Mettl3/14^nKO islets. The mutant islet displayed comparable basal insulin secretion at 2.8 mmol/L glucose, but they had a trend toward reduced GSIS response under high glucose (16.7 mmol/L) stimulation (P = 0.059) (Fig. 5I).

To evaluate a possible role of Mettl3/14-mediated m⁶A RNA methylation in the pathology process of T2D, we checked the expression of these two methyltransferases in islets of lean control and db/db mice at 8 weeks of age (Fig. 6A). The random blood glucose level of db/db mice was significantly higher than lean controls (23.2 ± 3.5 vs. 8.2 ± 0.6 mmol/L, db/db vs. WT, P < 0.05). We found Mettl3 and Mettl14 were abundantly expressed in both endocrine and exocrine cells in pancreas of lean control mice; however, by counting the percentage of Mettl3⁺ or Mettl14⁺ β-cells in diabetic db/db mice, both of their expressions in β-cells were significantly reduced by 31.3% and 27.1%, respectively, compared with lean control mice (Fig. 6A–C).

We also evaluated the expression of METTL3 or METTL14 in β-cells in pancreas sections of five patients with T2D and five age- and BMI-matched ND patients who had performed partial pancreatectomy (Supplementary Table 1). According to American Diabetes Association guidelines, all the five patients with T2D were under adequate glucose control with antidiabetic therapy (Supplementary Table 1). We found METTL3 and METTL14 were present in islet endocrine cells in the ND group, including both α-cells and β-cells. In contrast, islets from patients with T2D showed dramatic reductions in METTL3 and METTL14 expression (Fig. 6D). We further calculated the percentage of METTL3⁺ or METTL14⁺ β-cells in the two groups: in diabetic islets, the ratio of METTL3⁺/insulin⁺ cells (50.1 ± 16.2% vs. 95.6 ± 0.6%, T2D group vs. ND group, P < 0.01) (Fig. 6E) and METTL14⁺/insulin⁺ cells (55.3 ± 9.9% vs. 94.2 ± 3.0%, T2D group vs. ND group, P < 0.01) (Fig. 6F) declined substantially. These observations indicated that loss of Mettl3/14 might be involved in pathogenesis of T2D in both rodents and human.

Recent studies have reported the role of posttranscriptional mRNA modifications on organ development, cellular function, and human diseases, thus highlighting its involvement in both physiological and pathological status (19,23–25,41). Dynamic m⁶A modification was known to affect development and fate determination of various cell types (i.e., hematopoietic stem/progenitor cells or spermatogenesis) (23–25). Germ cell–specific deletion of the m⁶A RNA methyltransferase Mettl3 or Mettl14 disrupted spermiogenesis through inhibiting diploid spermatogonia stem cells proliferation/differentiation (24). Neural stem cells lacking Mettl14 displayed markedly decreased proliferation and premature differentiation (42). During zebrafish embryogenesis, Mettl3-depletion led to continuous activation of the Notch signaling pathway, ultimately resulting in the blockage of endothelial-to-hematopoietic transition to specify the earliest hematopoietic stem/progenitor cells (25). Very recently, Men et al. (31) reported that Mettl14 deficiency in β-cells resulted in defects of β-cell survival and insulin secretion (30). However, whether m⁶A modification was implicated in neonatal islet development and the detailed regulation mechanisms remained unknown.

In the current study, by using transgenic mice in which Mettl3/14 were specifically ablated in Ngn3⁺ endocrine progenitors, we observed that the composition of the four endocrine cell types (α, β, δ, and PP cells) remained unchanged in WT and Mettl3/14^nKO at birth. Strikingly, a dramatic drop in β-cell number and mass during the weaning period (P0-P14) led to the conclusion that m⁶A is essential for neonatal β-cell mass establishment. Interestingly, this regulation seems to be specific to β-cells, since Mettl3/14 deficiency had no effect on α, δ, and PP cell formation and expansion. Since Ngn3⁺ cells would also give rise to enteroendocrine cells, the interaction between enteroendocrine cells and endocrine cells should also be considered.

The postnatal weaning period is pivotal for immature β-cells to acquire their glucose-responsive insulin-producing mature phenotype (3). Previous studies have provided evidence on transcriptional factors (i.e., Nkx6.1, MafA, and Pdx1) (9–11) and cellular signals (i.e., mechanistic target of rapamycin [mTOR], connective tissue growth factor [CTGF], and estrogen-related receptor γ [ERRγ]) (8,43,44) in modulating β-cell functional maturation and/or postnatal expansion. In recent years, epigenetic modifiers were confirmed to be involved in the functional maturation of β-cells in the postnatal period to promote adequate metabolic reprogramming (5,8,13). For the first time, we demonstrated that Mettl3 and Mettl14 were essential for functional maturation of neonatal murine β-cells. First, the loss of Mettl3/14 compromised expression of critical β-cell transcription factors Pdx1, MafA, and Nkx6.1, which were reported to modulate β-cell maturation and function. Second, Mettl3/14-deficient β-cells exhibited reduced expression of mature β-cell marker Ucn3 and GLUT2. Third, impaired glucose-induced insulin release was detected in Mettl3/14^nKO β-cells, with defects in genes responsible for glucose sensing (loss of GLUT2) (35), insulin biosynthesis (loss of Ero1lb) (45), and granule docking (loss of Sytl4) (46). Fourth, the secretory vesicles in Mettl3/14^nKO β-cells were mostly immature. Interestingly, the pattern of Mettl3 and Mettl14 expression during the physiological maturation window in murine β-cells further supported our observation. The relatively low levels of Mettl3/14 at E17.5 and P0 might explain that they are dispensable for cell lineage formation during differentiation. The strong induction of Mettl3/14 during early postnatal period P0-P14 and high expression level throughout adulthood reinforced the importance of Mettl3/14 in driving β-cell functional maturation and maintaining β-cell function.

Combined analysis of P14 RNA-seq and MeRIP-seq identified that MafA was the direct target of Mettl3/14 for identity maintenance and functional maturation of neonatal β-cells. In Mettl3/14^nKO pancreas, remarkably diminished MafA expression was found as early as P0, before changes of other transcriptional factors (Pdx1 and Nkx6.1) and islet morphology. Moreover, our MeRIP-seq analysis of neonatal islets demonstrated abundant m⁶A modifications in MafA mRNA, which were mostly located in the exon region, whereas Pdx1 and Nkx6.1 mRNA had rather weak or no m⁶A modifications. Mechanical studies proved that knockdown of Mettl3/14 reduced MafA mRNA stability and shortened its mRNA half-life, while had no effect on that of Pdx1 and Nkx6.1. These observations suggest m⁶A/MafA serves as a direct trigger of functional maturation process in neonatal β-cells.

To exclude the possibility that Mettl3/Mettl14 inactivation compromised β-cell differentiation that contributed to later β-cell functional immaturity, comprehensive gene expression analyses in P0 β-cells were performed. In our P0 RNA-seq, we did not detect significant differences in MafA mRNA, which might be caused by their relatively low expression level at this early time point. However, we confirmed MafA downregulation at the protein level in P0 mutant islets by both immunofluorescence staining and Western blot. The downregulation in MafA expression was accompanied with reduced GLUT2 expression and impaired GSIS in P0 mutant islets, when the islet composition and blood glucose level were unaffected. Interestingly, m⁶A-induced MafA alteration at P0 preceded and was followed by the changes in Nkx6.1 and Pdx1. It is well accepted that Pdx1 and Nkx6.1 are essential for β-cell differentiation (47–50). On the contrary, MafA is not involved in endocrine specification during embryonic development but is critical for β-cell functional maturation (51,52). After birth, MafA expression remains low, and it increases gradually with age (53). The role of MafA on regulating replication/survival and function of β-cells is gradually increased after birth, but it is not conspicuous at birth (54). Our findings provide evidence that Mettl3/14 regulates MafA expression and, thus, controls β-cell functional maturation after birth.

Indeed, our Mettl3/14^nKO mice were to some extent phenotypically similar to MafA^KO mice (11): both of them showed normal islet morphology and comparable β-cell numbers at birth but a progressive loss of β-cell identity and function with age. The link between m⁶A/MafA modulation in neonatal β-cells indicates that this finely tuned regulation network is crucial for proper β-cell maturation during a critical time window. Interestingly, like other diabetic mice models, we observed a dramatic increase in the percentage of dedifferentiated β-cells in 8-week-old Mettl3/14^nKO islets. Considering the high percentage of dedifferentiation (30%) and the relatively short hyperglycemic period (8 weeks old), the possible role of Mettl3/14 on β-cell dedifferentiation, independent of hyperglycemia, needs to be further clarified. Moreover, it would be more ideal to perform RNA-seq analyses on purified β-cells from earlier developmental times instead of P14 islets to fully clarify the identical regulatory mechanisms in Mettl3/14 KO β-cells.

Using diabetic mouse models, we and others have shown that loss of Mettl3/14 in β-cells is one of the critical mechanisms in β-cell failure during the pathology of diabetes (30,31). The current study provided observational evidence on remarkably reduced METTL3 and METTL14 expression in islets from patients with T2D compared with patients who were ND, revealing the significance of Mettl3/14-mediated m⁶A signaling in the occurrence and development of human diabetes. This is consistent with a very recent finding by De Jesus et al. (55), who reported that m⁶A mRNA methylation was involved in human β-cell biology. A deeper understanding of the role of m⁶A mRNA methylation in β-cell biology at different stages of development or disease would help to find therapeutic strategies for maintaining or regenerating functional β-cell mass. Taken together, our data showed, for the first time, that as the most abundant mRNA modification in mammalian cells, Mettl3/14-mediated m⁶A modification played critical roles in governing neonatal β-cell functional maturation and identity maintenance. Loss of Mettl3/14 in neonatal islets led to overt neonatal diabetes due to incapability of establishing adequate functional β-cell mass after birth. Our results highlight the important role of Mettl3/14-mediated m⁶A modification in driving functional maturation in neonatal β-cells, which might be a potential target for diabetes therapy.

Acknowledgments. The authors thank Prof. Ming-Han Tong (University of Chinese Academy of Sciences, Shanghai, China) for providing the Mettl3^flox/flox and Mettl14^flox/flox mice and for his valuable discussion and comments for the study. The authors thank Prof. Wei-Zhen Zhang (Department of Physiology and Pathophysiology, School of Basic Science, Peking University Health Science Center, Beijing, China) for providing the Ngn3-cre mice.

Funding. This work was supported by National Natural Sciences Foundation of China grants 81670700 and 81870527.

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

Author Contributions. Y.W. and J.S. performed all the experiments and analyzed the data. Y.W., J.S., and Q.W. wrote the manuscript. Z.L., W.Z., and S.W. contributed to the data discussion. W.W., Q.W., and G.N. designed the project, supervised research, and coordinated the execution of the experimental plan. G.N. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.