The important role of m6A RNA modification in β-cell function has been established; however, how it regulates pancreatic development and endocrine differentiation remains unknown. Here, we generated transgenic mice lacking RNA methyltransferase-like 3 (Mettl3) specifically in Pdx1+ pancreatic progenitor cells and found the mice with the mutation developed hyperglycemia and hypoinsulinemia at age 2 weeks, along with an atrophic pancreas, reduced islet mass, and abnormal increase in ductal formation. At embryonic day 15.5, Mettl3 deletion had caused a significant loss of Ngn3+ endocrine progenitor cells, which was accompanied by increased Sox9+ ductal precursor cells. We identified histone deacetylase 1 (Hdac1) as the critical direct m6A target in bipotent progenitors, the degeneration of which caused abnormal activation of the Wnt/Notch signaling pathway and blocked endocrine differentiation. This transformation could be manipulated in embryonic pancreatic culture in vitro through regulation of the Mettl3-Hdac1-Wnt/Notch signaling axis. Our finding that Mettl3 determines endocrine lineage by modulating Hdac1 activity during the transition of bipotent progenitors might help in the development of targeted endocrine cell protocols for diabetes treatment.
Loss of Mettl3 in pancreatic progenitor cells reduced the size of the endocrine progenitor pool and postnatal islet mass and caused early-onset diabetes in mice.
Mettl3 deletion disturbed the fate determination of bipotent progenitors toward endocrine or ductal lineage during pancreatic development.
The Mettl3-Hdac1-Wnt/Notch axis was crucial for endocrine cell formation both in vivo and in vitro.
Introduction
The prevalence of diabetes in children and adolescents has been steadily rising globally over the last few decades (1). Genome-wide association studies have implicated a growing list of genetic variants associated with age at onset among patients with diabetes (2–6), indicating that individuals with these genetic mutations might be more susceptible to early-onset diabetes. Islet cells are formed from neurogenin 3+ (Ngn3+) endocrine progenitors, and the nascent islet cells proliferate during the end of embryogenesis to constitute the bulk of the prenatal β-cell mass (7). The transient activation of Ngn3 from embryonic day 12.5 (E12.5) to E16.5 in mice (7,8) and from week ∼8 to 10 in humans (9) drives the specification of bipotent progenitor cells to endocrine progenitors (7,8). It has been documented that intrauterine growth restriction results in inadequate embryonic β-cell mass and therefore impairs glucose homeostasis in adulthood (10,11). Endocrine cell differentiation and proliferation are tightly controlled by epigenetic mechanisms during pancreatic endocrine development (12), including DNA methylation, histone modifications, and noncoding RNAs (13).
RNA methyltransferase-3 (Mettl3) plays a major catalytic role in the m6A addition process (14,15), and m6A mRNA modification is crucial for various physiological processes (16) through regulation of mRNA splicing, export, localization, translation, and stability (14,15). Our group previously reported that Mettl3/14 regulated neonatal β-cell functional maturation by affecting MafA mRNA stability (17). Moreover, De Jesus et al. (18), Liu et al. (19), and Men et al. (20) reported that m6A modification affected insulin secretion and survival of mature β-cells. However, it is currently unknown whether Mettl3-mediated m6A methylation regulates pancreatic development or endocrine differentiation.
In the current study, we generated transgenic mice (Mettl3pKO) in which Mettl3 was specifically deleted in Pdx1+ pancreatic progenitor cells and found that these mice developed severe hyperglycemia and hypoinsulinemia at age 2 weeks, with reduced islet mass, decreased number of endocrine cells, and increased number of CK19+ ductal cells. In addition, Mettl3 deletion caused a significant loss of Ngn3+ endocrine progenitor cells at E15.5, accompanied by an abnormal accumulation of Sox9+ ductal precursor cells. Combined with single-cell technology and RNA methylation studies, we demonstrated that the fate choice between pancreatic ductal and endocrine lineages could be modulated by Mettl3/histone deacetylase 1 (Hdac1) in an m6A-dependent manner in vivo. This transformation could be manipulated in embryonic pancreatic culture in vitro, reinforcing the importance of the Mettl3-Hdac1-Wnt/Notch axis in endocrine lineage determination. Our finding has led to promising strategies for the targeted generation and expansion of sufficient islets from human pluripotent stem cells for the reversal of diabetes.
Research Design and Methods
Mice
Mice were maintained in a pathogen-free animal facility on a 12 h–12 h light–dark cycle at an ambient temperature of 25°C, with normal food and water provided ad libitum. Mettl3-floxed mice (Mettl3flox/flox) were generated and provided by Ming-Han Tong (University of Chinese Academy of Sciences). Mettl3flox/flox mice were crossed with Pdx1-Cre mice (21) to obtain Mettl3 complete pancreatic knockout mice (Pdx1-cre;Mettl3flox/flox [i.e., Mettl3pKO]). Cre-negative littermate controls (Mettl3flox/flox) were used as their littermate controls. Male mice were used in all experiments in the current study, and embryo sex was not determined.
Immunostaining Analysis
The pancreas were dissected, fixed, and processed as described previously (17). For immunostaining analysis, the entire pancreatic tissue was dissected and then continuously sectioned at a thickness of 5 μm. Immunochemistry was performed on continuous sections (selected every 150 μm, with at least six sections per animal) to obtain representative cell number information for the whole pancreas. Immunofluorescence of paraffin-embedded sections of the pancreas was performed with antibodies to insulin (1:50; Dako), glucagon (1:200, Abcam), somatostatin (1:200; Abcam), pancreatic polypeptide (PP; 1:400; Proteintech), Pdx1 (1:100; Cell Signaling Technology), Ngn3 (1:1,000; Developmental Studies Hybridoma Bank [DSHB]), Nkx2.2 (1:100; DSHB), Nkx6.1 (1:100; DSHB), Sox9 (1:100; Abcam), and Hdac1 (1:5,000; Abcam).
Circulating Insulin Measurement
Plasma insulin concentrations were determined using an ELISA kit (Mouse Ultrasensitive Insulin ELISA Kit; ALPCO) as previously described (22).
Single-Cell Transcriptome Capture, Library Construction, and Sequencing
E15.5 mouse pancreata were dissected and digested with 0.25% trypsin (cat. no. T4799; Sigma-Aldrich) for 5 min at 37°C. Digestion was terminated with equal volumes of FBS. Cells were firstly stained with two fluorescent dyes, Calcein AM and Draq7, for precise determination of cell concentration and viability via BD Rhapsody Scanner (BD Biosciences). Cells were loaded in one BD Rhapsody microwell cartridge based on the report by Fan et al. (23). Cell capture beads were then loaded excessively to ensure that nearly every microwell contained one bead, and the excess beads were washed away from the cartridge. After lysing cells with lysis buffer, cell capture beads were retrieved and washed before performing reverse transcription. Microbead-captured single-cell transcriptomes were generated and introduced into a cDNA library containing cell label and unique molecular identifier information. All procedures were performed with the BD Rhapsody cDNA Kit (BD Biosciences) and BD Rhapsody Targeted mRNA & AbSeq Amplification Kit (BD Biosciences), strictly following the manufacturer’s protocol. All libraries were sequenced in the PE150 mode (paired end for 150-bp read) on the NovaSeq platform (Illumina).
m6A MeRIP-Seq and Data Analysis
mRNA Stability Assay
The turnover rate of mRNA half-life measurements was performed according to a previous publication (24). In brief, E13.5 pancreatic cells with or without Mettl3 knockdown were treated with 5 μg/mL actinomycin D (Sigma-Aldrich) for 1 or 3 h at the end of culture and then collected for RNA extraction and RT-PCR analysis.
Pancreatic Culture and Treatments
The E13.5 mouse pancreatic buds were dissected from embryonic control and Mettl3pKO mice. Pancreata were cultured on Whatman Nuclepore Track-Etched Membranes (cat. no. 110414; Sigma-Aldrich) in a 12-well plate with 1 mL DMEM medium (cat. no. 11885; Gibco) per well containing 10% FBS and 1% penicillin/streptomycin (25). Hdac1 inhibitor parthenolid or Hdac agonist ITSA-1 was added to the culture medium at a final concentration of 10 μmol/L, and the embryonic mouse pancreas was then cultured in vitro for another 2 days. The treated pancreas was then collected for paraffin embedding, sectioning, and immunofluorescence staining. To knock down Mettl3, shRNA lentiviruses targeting Mettl3 were constructed, packaged, purified, and titrated at GeneChem Co., Ltd. Pancreata were infected with purified lentivirus at 50 multiplicities of infection for 48 h. After infection, they were harvested for RNA extraction and further analysis.
Statistics and Reproducibility
The exact sample size for each experiment is indicated in the figure legends. All statistical comparisons of two groups used the two-sided Student t test; ANOVA was used for multiple groups. Statistical analyses were performed with GraphPad Prism 8. P < 0.05 was considered statistically significant.
Data and Resource Availability
m6A MeRIP-seq and single-cell RNA sequencing (scRNA-seq) data were deposited in the Gene Expression Omnibus database under accession numbers GSE229502 and GSE234256.
Results
Early Onset of Diabetes and Impaired Islet Formation in Mettl3pKO Mice
To examine whether Mettl3-mediated m6A mRNA methylation affects pancreatic development, Mettl3flox/flox mice (26) were crossed with Pdx1-Cre mice (21) to obtain Mettl3pKO mice (Fig. 1A). Immunostaining results demonstrated that the Mettl3 protein was selectively absent in the Pdx1+ pancreatic epithelial cells of E11.5 Mettl3pKO pancreatic buds (Fig. 1B). We found that Mettl3 expression was lost in most (∼81%) E11.5 Mettl3pKO Pdx1+ pancreatic cells (Fig. 1C), indicating that recombination is extremely efficient during early pancreatic organogenesis. Moreover, there was a reduction (P = 0.051) in the m6A levels of total RNA extracted from E15.5 Mettl3pKO embryonic pancreata (Fig. 1D), indicating the absence of Mettl3-reduced m6A RNA methylation levels in E15.5 Mettl3pKO pancreatic cells.
Metttl3pKO mice developed hypoinsulinemia and early-onset diabetes with atrophic pancreas. A: Model illustrating the generation of Mettl3pKO. B: Representative pancreatic sections immunostained for Mettl3 (red) and Pdx1 (green) in E11.5 WT and Mettl3pKO mice. C: Ratios of Mettl3+ cells to total Pdx1+ pancreatic cells in E11.5 Mettl3pKO and WT pancreata were determined (n = 3). D: m6A amounts relative to adenosine (A) in mRNA extracted from E15.5 Mettl3pKO and WT pancreata were quantified (n = 3). E: Body weights of Mettl3pKO, Mettl3pHET, and WT mice were monitored weekly (n = 4–9). F: Random blood glucose levels of Mettl3pKO, Mettl3pHET, and WT mice were monitored weekly (n = 4–9). G: Plasma insulin levels of P14 Mettl3pKO and WT mice were determined (n = 3–4). H: Morphology of P0 Mettl3pKO and WT pancreata. I–K: Pancreatic weights of P0 (I), E13.5 (J), and E15.5 (K) Mettl3pKO and WT mice were determined (n = 3–5). Data are presented as mean ± SEM of independent experiment indicated above. Student t test was used for two groups and ANOVA for multiple groups. Nuclei were counterstained with DAPI (blue). Scale bars, 50 μm. *P < 0.05, **P < 0.01, ***P < 0.001.
Metttl3pKO mice developed hypoinsulinemia and early-onset diabetes with atrophic pancreas. A: Model illustrating the generation of Mettl3pKO. B: Representative pancreatic sections immunostained for Mettl3 (red) and Pdx1 (green) in E11.5 WT and Mettl3pKO mice. C: Ratios of Mettl3+ cells to total Pdx1+ pancreatic cells in E11.5 Mettl3pKO and WT pancreata were determined (n = 3). D: m6A amounts relative to adenosine (A) in mRNA extracted from E15.5 Mettl3pKO and WT pancreata were quantified (n = 3). E: Body weights of Mettl3pKO, Mettl3pHET, and WT mice were monitored weekly (n = 4–9). F: Random blood glucose levels of Mettl3pKO, Mettl3pHET, and WT mice were monitored weekly (n = 4–9). G: Plasma insulin levels of P14 Mettl3pKO and WT mice were determined (n = 3–4). H: Morphology of P0 Mettl3pKO and WT pancreata. I–K: Pancreatic weights of P0 (I), E13.5 (J), and E15.5 (K) Mettl3pKO and WT mice were determined (n = 3–5). Data are presented as mean ± SEM of independent experiment indicated above. Student t test was used for two groups and ANOVA for multiple groups. Nuclei were counterstained with DAPI (blue). Scale bars, 50 μm. *P < 0.05, **P < 0.01, ***P < 0.001.
Mettl3pKO mice were born in the expected Mendelian ratio and had no difference in body weight compared with their age-matched wild-type (WT; Mettl3flox/flox) and heterozygous (Pdx1-cre; Mettl3flox/w [i.e., Mettl3pHET]) counterparts within 8 weeks after birth (Fig. 1E). At birth, Mettl3pKO mice had random blood glucose levels comparable to those of WT controls and heterozygous mice (Fig. 1F); however, mice with the mutation started to display significantly increased random blood glucose levels at age 2 weeks and had severe hyperglycemia (∼30 mmol/L) thereafter (Fig. 1F). Moreover, plasma insulin levels in 2-week-old Mettl3pKO mice decreased by 65.1% (P < 0.05) (Fig. 1G), suggesting β-cell failure predominated in these mice with the diabetic mutation.
At birth, both the weight and size of the pancreas were significantly decreased in Mettl3pKO mice (Fig. 1H and I), whereas no change was detected between the weight of the pancreas in WT or Mettl3pKO mice at E13.5 and that at E15.5 (Fig. 1J and K). At birth (P0), we observed a 52% decrease in the area of amylase+ cells (Fig. 2A and B). Moreover, the reduced acinar area was mainly attributed to defects in acinar cell proliferation; nearly half of the reduction in the proportion of Ki67+ amylase+ cells was observed in the mutated pancreas (Supplementary Fig. 1A and B). No change in the rate of apoptotic caspase 3+ acinar cells was found between mutated and WT pancreata (Supplementary Fig. 1C). Histological analysis also revealed a nearly 50% reduction in islet mass (P < 0.001) (Fig. 2C and Supplementary Fig. 2A) in pancreata of newborn Mettl3pKO mice, including reductions in both islet size (Fig. 2D) and islet number per section (Fig. 2E). These results indicate that neonatal islet formation requires Mettl3-mediated m6A methylation.
Loss of Mettl3 results in reduced acinar area, impaired endocrine formation, and abnormal increased ductal formation. A: Immunostaining against amylase (red) and synaptophysin (green) in P0 mutated and WT pancreatic buds. B: Pancreatic acinar areas of P0 Mettl3pKO and WT mice were determined (n = 4). C–E: Relative islet mass (C), islet size (D), and islet number (E) in P0 Mettl3pKO and WT pancreata were determined (n = 4). F: Representative pancreatic sections immunostained for insulin (INS; green) and glucagon (GCG; red) in Mettl3pKO and WT pancreata at P0 are shown. G: Percentages of β/α/δ/ PP cells in P0 Mettl3pKO and WT mice were determined (n = 3). H: Representative pancreatic sections immunostained for INS (green) and cytokeratin 19 (CK19; red) in Mettl3pKO and WT pancreata at P0 were shown. I: Percentages of CK19+ cells in P0 Mettl3pKO and WT mice were determined (n = 4). Data are presented as mean ± SEM of independent experiment indicated above. Student t test was used for two groups. Nuclei were counterstained with DAPI (blue). Scale bars, 50 μm. *P < 0.05, **P < 0.01, ***P < 0.001. SST, somatostatin.
Loss of Mettl3 results in reduced acinar area, impaired endocrine formation, and abnormal increased ductal formation. A: Immunostaining against amylase (red) and synaptophysin (green) in P0 mutated and WT pancreatic buds. B: Pancreatic acinar areas of P0 Mettl3pKO and WT mice were determined (n = 4). C–E: Relative islet mass (C), islet size (D), and islet number (E) in P0 Mettl3pKO and WT pancreata were determined (n = 4). F: Representative pancreatic sections immunostained for insulin (INS; green) and glucagon (GCG; red) in Mettl3pKO and WT pancreata at P0 are shown. G: Percentages of β/α/δ/ PP cells in P0 Mettl3pKO and WT mice were determined (n = 3). H: Representative pancreatic sections immunostained for INS (green) and cytokeratin 19 (CK19; red) in Mettl3pKO and WT pancreata at P0 were shown. I: Percentages of CK19+ cells in P0 Mettl3pKO and WT mice were determined (n = 4). Data are presented as mean ± SEM of independent experiment indicated above. Student t test was used for two groups. Nuclei were counterstained with DAPI (blue). Scale bars, 50 μm. *P < 0.05, **P < 0.01, ***P < 0.001. SST, somatostatin.
Loss of Mettl3 Alters Lineage Determination of Pancreatic Endocrine Cells
We performed immunostaining against endocrine hormones (insulin-β, glucagon-α, somatostatin-δ, and PP) on P0 pancreata and found the percentage of β-cells was significantly reduced in mutated pancreata at birth (Fig. 2F and G). This was accompanied by decreased α-cells, δ–cells, and PP cells (Fig. 2G and Supplementary Fig. 2B). In contrast, cytokeratin 19+ ductal cells were significantly increased (P < 0.001) in Mettl3pKO pancreata, accompanied by mild dilation (Fig. 2H and I), whereas no obvious cyst formation was detected. The above data indicate Mettl3-mediated m6A methylation affects the composition of the pancreas during pancreatic development.
To determine at which time point loss of Mettl3 affects pancreatic cell lineage, we examined the number of pancreatic epithelial cells staining for Ngn3+ and Sox9+ in Mettl3pKO and WT pancreata at E13.5 and E15.5 (Fig. 3A). We did not find changes in the number of Ngn3+ or Sox9+ cells at E13.5 (Fig. 3B and C). In contrast, at E15.5, the mutated embryos exhibited significantly reduced Ngn3+ populations (Fig. 3D) and increased Sox9+ populations (Fig. 3E) in the pancreatic epithelium. We also found reduced expression of endocrine-specific transcription factors Ngn3, Nkx2.2, and Nkx6.1 and increased mRNA expression of Hes1, a negative regulator of Ngn3, in E15.5 Mettl3pKO pancreata (Fig. 3F). In parallel, the numbers of Nkx2.2+ and Nkx6.1+ cells per pancreatic section in E15.5 Mettl3pKO mice were dramatically decreased compared with those in their littermates (Supplementary Fig. 3A and B). In contrast, the mRNA expression of Hnf1β, which was confined to later ductal cells, was abnormally increased in E15.5 Mettl3pKO pancreata (Fig. 3F). However, the mRNA expression of Ptf1α and the ratio of cells staining for proacinar cell marker Cpa1, which controls acinar cell differentiation in the pancreas, remained unchanged (Fig. 3F and Supplementary Fig. 3C and D). These results reveal that depletion of Mettl3 did not affect acinar cell differentiation at E15.5. Our data demonstrate that loss of Mettl3 in Pdx1+ cells at E15.5 retarded endocrine progenitor lineage but triggered Sox9+ progenitor cell differentiation.
Loss of Mettl3 caused reduced pancreatic endocrine progenitors at E15.5. A: Immunofluorescence staining of endocrine progenitor markers Ngn3 (red) and Pdx1 (green) and ductal marker Sox9 (red) in E13.5 or E15.5 WT and Mettl3pKO embryonic pancreata. B and C: Proportions of Ngn3+ endocrine progenitor cells (B) and Sox9+ progenitor cells (C) in Pdx1+ pancreatic epithelial cells at E13.5 in WT and Mettl3pKO pancreata were calculated (n = 3). D and E: Proportions of Ngn3+ endocrine progenitor cells (D) and Sox9+ progenitor cells (E) in Pdx1+ pancreatic epithelial cells at E15.5 in WT and Mettl3pKO pancreata were calculated (n = 3). F: Relative expression levels of Ngn3, Nkx2.2, Nkx6.1, Hes1, Hnf1β, and Ptf1α mRNA in E15.5 WT and Mettl3pKO pancreata were detected by RT-PCR (n = 4–5). Data are presented as mean ± SEM of independent experiment indicated above. Student t test was used. Nuclei were counterstained with DAPI (blue). Scale bars, 50 μm. *P < 0.05, **P < 0.01, ***P < 0.001.
Loss of Mettl3 caused reduced pancreatic endocrine progenitors at E15.5. A: Immunofluorescence staining of endocrine progenitor markers Ngn3 (red) and Pdx1 (green) and ductal marker Sox9 (red) in E13.5 or E15.5 WT and Mettl3pKO embryonic pancreata. B and C: Proportions of Ngn3+ endocrine progenitor cells (B) and Sox9+ progenitor cells (C) in Pdx1+ pancreatic epithelial cells at E13.5 in WT and Mettl3pKO pancreata were calculated (n = 3). D and E: Proportions of Ngn3+ endocrine progenitor cells (D) and Sox9+ progenitor cells (E) in Pdx1+ pancreatic epithelial cells at E15.5 in WT and Mettl3pKO pancreata were calculated (n = 3). F: Relative expression levels of Ngn3, Nkx2.2, Nkx6.1, Hes1, Hnf1β, and Ptf1α mRNA in E15.5 WT and Mettl3pKO pancreata were detected by RT-PCR (n = 4–5). Data are presented as mean ± SEM of independent experiment indicated above. Student t test was used. Nuclei were counterstained with DAPI (blue). Scale bars, 50 μm. *P < 0.05, **P < 0.01, ***P < 0.001.
scRNA-Seq on E15.5 Mettl3pKO Pancreatic Cells
To investigate the molecular mechanisms underlying pancreatic cell fate determination upon Mettl3 loss, we performed scRNA-seq on E15.5 embryonic pancreatic cells obtained from WT and Mettl3pKO mice (Fig. 4A) and sequenced 11,049 WT and 10,928 Mettl3pKO cells, revealing 16 transcriptionally unique subtypes (Fig. 4B). We identified expected populations based on the expression of known markers, that is, the expression of Ins1 in β-cells, Gcg in α-cells, Ngn3 in endocrine progenitor cells, Spp1 in trunk (also known as bipotent progenitor cells), and Cpa1 in acinar cells (Fig. 4C) as well as mesenchymal, mesothelial, endothelial, immune, vascular, and neuron cells (Fig. 4C). A reduction in endocrine lineage, including endocrine progenitors and endocrine cells, was noted (Fig. 4D).
scRNA-seq analysis of E15.5 WT and Mettl3pKO pancreatic cells. A: Schematic of scRNA-seq experimental approach. B: Uniform manifold approximation and projection (UMAP) showing clusters of E15.5 WT and Mettl3pKO pancreatic cells both overall (left) and separately (right). C: Characteristic genes of different cell populations in combined clusters of E15.5 WT and Mettl3pKO pancreata. D: Relative cell numbers of different populations in E15.5 WT and Mettl3pKO pancreata. E: UMAP showing pancreatic bipotent progenitors, endocrine progenitors, and endocrine cell clusters in E15.5 WT and Mettl3pKO pancreata, respectively. F: Characteristic gene profiles of pancreatic bipotent progenitors, endocrine progenitors, and endocrine cell populations in combined clusters of E15.5 WT and Mettl3pKO pancreata. G: Absolute cell numbers of pancreatic bipotent progenitors, endocrine progenitors, and endocrine cell clusters in E15.5 WT and Mettl3pKO pancreata. H: Gene ontology (GO) analysis of differentially expressed genes in BP_late cluster in E15.5 WT and Mettl3pKO pancreata. I: Volcanic map of differentially expressed genes in BP_late cluster of E15.5 WT and Mettl3pKO pancreata.
scRNA-seq analysis of E15.5 WT and Mettl3pKO pancreatic cells. A: Schematic of scRNA-seq experimental approach. B: Uniform manifold approximation and projection (UMAP) showing clusters of E15.5 WT and Mettl3pKO pancreatic cells both overall (left) and separately (right). C: Characteristic genes of different cell populations in combined clusters of E15.5 WT and Mettl3pKO pancreata. D: Relative cell numbers of different populations in E15.5 WT and Mettl3pKO pancreata. E: UMAP showing pancreatic bipotent progenitors, endocrine progenitors, and endocrine cell clusters in E15.5 WT and Mettl3pKO pancreata, respectively. F: Characteristic gene profiles of pancreatic bipotent progenitors, endocrine progenitors, and endocrine cell populations in combined clusters of E15.5 WT and Mettl3pKO pancreata. G: Absolute cell numbers of pancreatic bipotent progenitors, endocrine progenitors, and endocrine cell clusters in E15.5 WT and Mettl3pKO pancreata. H: Gene ontology (GO) analysis of differentially expressed genes in BP_late cluster in E15.5 WT and Mettl3pKO pancreata. I: Volcanic map of differentially expressed genes in BP_late cluster of E15.5 WT and Mettl3pKO pancreata.
We then subclustered the endocrine cell lineage in the data sets (Fig. 4E) and identified 11 well-established clusters, including a β-cell cluster, a δ-cell cluster, two clusters of α-cells, three clusters of endocrine progenitor cells, four clusters of bipotent progenitor cells, and three minor uncharacterized clusters based on a previous resource of embryonic pancreatic pseudotime analysis (Fig. 4E and F and Supplementary Fig. 4A). Endocrine progenitor cells can be further classified into three subpopulations (EP_early, EP_mid, and EP_late) according to their expression of Ngn3, Chgb, and Fev (Fig. 4F and Supplementary Fig. 4A). The quantification analysis demonstrated that endocrine progenitor, β-, α-, and δ-cells were all significantly decreased in mice with the mutation (Fig. 4G). Four clusters of bipotent progenitor cells were identified according to their differential cell fate segregation and regulatory characteristics, in which BP_early 1, BP_early 2, and BP_mid cells represented earlier developmental stages, whereas BP_late cells were considered precursors of endocrine progenitors on the developmental trajectory (Supplementary Fig. 4A and B). Of note, inappropriate accumulation of BP_late cells (523 vs. 388 in Mettl3pKO vs. WT, respectively) was detected (Fig. 4G). The above data indicate loss of Mettl3 blocked the transition from the late stage of bipotent progenitor cells to endocrine progenitors at E15.5.
We compared the gene expression profiles of BP_late cells between WT and Mettl3pKO mice (Fig. 4H and I). Gene ontology analysis of BP_late subpopulations revealed that genes highly expressed in Mettl3pKO were related to positive regulation of epithelial cell migration and embryonic epithelial tube formation, whereas the downregulated genes were enriched in gene ontology terms associated with oxidative phosphorylation, ATP metabolic process, glucose metabolism, and endocrine pancreatic development (Fig. 4H). Interestingly, the Wnt signaling pathway (known as an inhibitor of endocrine differentiation) was among the upregulated terms, whereas negative regulation of the Wnt signaling pathway was found in the downregulated terms (Fig. 4H).
Loss of Mettl3 resulted in reduced expression of genes important for endocrine differentiation (i.e., Nkx6.1 [27], Hdac1 [28], Nr5a2 [29], and Kcnq1 [30]) in the BP_late cluster (Fig. 4I). In contrast, a cluster of genes involved in early epithelial cell differentiation and epithelial tube formation (Onecut2 [31], Nr2f2 [32], Sox4 [33,34], Sox11 [34], Wnt4 [35], and Hnf1β [36]) were dramatically upregulated (Fig. 4I). Notably, Sox4 is also essential for normal endocrine pancreatic development (35).
Identification of Hdac1 as an m6A Target in the E15.5 Mettl3pKO Pancreas
To identify whether the altered transcripts were m6A modified, we further performed m6A MeRIP-seq analysis on E15.5 WT and Mettl3pKO pancreatic cells. MeRIP-seq revealed 1,288 common peaks, as well as 9,228 and 2,653 special peaks in WT and Mettl3pKO mice, respectively (Fig. 5A). Among these, 1,241 hypermethylation and 6,543 hypomethylation m6A peaks were found in Mettl3pKO compared with WT mice (Fig. 5A). The m6A sequencing data demonstrate that a vast majority of m6A peaks were distributed in the coding sequence and 3′ untranslated regions of the transcripts (Fig. 5B), and the m6A peak density was mainly decreased in the coding sequence regions of Mettl3pKO (Fig. 5C). A representative motif analysis in WT and Mettl3pKO groups identified GGACU and DGGACU (D = A/G/U), respectively (Fig. 5D). The differential expression analysis of methylated transcripts revealed that 1,029 genes were hypermethylated, whereas 3,985 transcripts were hypomethylated in Mettl3-deficient pancreatic cells (Fig. 5E). Pathway analysis showed that hypomethylated transcripts were significantly enriched in transcript sets related to maturity-onset diabetes of the young, the insulin signaling pathway, and pancreatic secretion in Mettl3-deleted pancreatic cells (Fig. 5F).
Combined analysis of MeRIP-seq and scRNA-seq identified Hdac1 as Mettl3 target. A: Venn diagram of MeRIP-seq data showing the common and unique peaks of m6A RNA methylation in E15.5 WT and Mettl3pKO pancreata (upper panel). Bar chart showing numbers of differentially methylated peaks between WT and Mettl3pKO groups (lower panel). B: Distribution density of methylated peaks across mRNA transcripts in E15.5 WT and Mettl3pKO groups. C: Pie chart depicting the fraction of m6A peaks in five transcript segments in E15.5 WT and Mettl3pKO groups. D: Representative motif analysis of methylated peaks in E15.5 WT and Mettl3pKO groups. E: Differential expression analysis of hypermethylated and hypomethylated genes in E15.5 Mettl3pKO group compared with WT group. F: KEGG pathway analysis of enriched signaling pathways of hypomethylated genes in E15.5 Mettl3pKO group. G: Venn diagram showing the overlap between genes with hypomethylation and genes that were differentially expressed in scRNA-seq of E15.5 Mettl3pKO BP_late cluster. H: Integrative Genomics Viewer tracks showing MeRIP-seq read distribution in Hdac1 mRNA in E15.5 WT and Mettl3pKO groups. I: MeRIP quantitative PCR (qPCR) analysis of the fold enrichment of Hdac1 m6A level by immunoprecipitation with specific m6A antibody in E15.5 WT and Mettl3pKO groups (n = 3). J: Quantification of RIP qPCR verifying the binding of Mettl3 protein to Hdac1 mRNA (n = 3). K: E13.5 WT pancreata were transfected with Lv-ShMettl3 or control virus for 48 h; then, Hdac1 protein abundance was assayed by immunoblot (n = 3). L: Hdac1 mRNA remaining after actinomycin D (ActD) treatment with or without ShMettl3 in E13.5 WT pancreata is shown (n = 3). M: Representative pancreatic sections from E15.5 WT and Mettl3pKO pancreata were coimmunostained for Sox9 (green) and Hdac1 (red; n = 3). N: Representative pancreatic sections from E15.5 WT and Mettl3pKO pancreata were coimmunostained for Pdx1 (green) and β-catenin/Hes1 (red; n = 3). Data are presented as mean ± SEM of independent experiment indicated above. Student t test was. Nuclei were counterstained with DAPI (blue). Scale bars, 50 μm. *P < 0.05, **P < 0.01, ***P < 0.001. CDS, coding sequence; UTR, untranslated region.
Combined analysis of MeRIP-seq and scRNA-seq identified Hdac1 as Mettl3 target. A: Venn diagram of MeRIP-seq data showing the common and unique peaks of m6A RNA methylation in E15.5 WT and Mettl3pKO pancreata (upper panel). Bar chart showing numbers of differentially methylated peaks between WT and Mettl3pKO groups (lower panel). B: Distribution density of methylated peaks across mRNA transcripts in E15.5 WT and Mettl3pKO groups. C: Pie chart depicting the fraction of m6A peaks in five transcript segments in E15.5 WT and Mettl3pKO groups. D: Representative motif analysis of methylated peaks in E15.5 WT and Mettl3pKO groups. E: Differential expression analysis of hypermethylated and hypomethylated genes in E15.5 Mettl3pKO group compared with WT group. F: KEGG pathway analysis of enriched signaling pathways of hypomethylated genes in E15.5 Mettl3pKO group. G: Venn diagram showing the overlap between genes with hypomethylation and genes that were differentially expressed in scRNA-seq of E15.5 Mettl3pKO BP_late cluster. H: Integrative Genomics Viewer tracks showing MeRIP-seq read distribution in Hdac1 mRNA in E15.5 WT and Mettl3pKO groups. I: MeRIP quantitative PCR (qPCR) analysis of the fold enrichment of Hdac1 m6A level by immunoprecipitation with specific m6A antibody in E15.5 WT and Mettl3pKO groups (n = 3). J: Quantification of RIP qPCR verifying the binding of Mettl3 protein to Hdac1 mRNA (n = 3). K: E13.5 WT pancreata were transfected with Lv-ShMettl3 or control virus for 48 h; then, Hdac1 protein abundance was assayed by immunoblot (n = 3). L: Hdac1 mRNA remaining after actinomycin D (ActD) treatment with or without ShMettl3 in E13.5 WT pancreata is shown (n = 3). M: Representative pancreatic sections from E15.5 WT and Mettl3pKO pancreata were coimmunostained for Sox9 (green) and Hdac1 (red; n = 3). N: Representative pancreatic sections from E15.5 WT and Mettl3pKO pancreata were coimmunostained for Pdx1 (green) and β-catenin/Hes1 (red; n = 3). Data are presented as mean ± SEM of independent experiment indicated above. Student t test was. Nuclei were counterstained with DAPI (blue). Scale bars, 50 μm. *P < 0.05, **P < 0.01, ***P < 0.001. CDS, coding sequence; UTR, untranslated region.
To identify the Mettl3-mediated direct m6A-modified targets controlling endocrine cell lineage, we compared preferentially changed transcripts identified by scRNA-seq and MeRIP-seq in BP_late cells between WT and Mettl3pKO mice. Importantly, the correlation analysis eventually identified 70 downregulated transcripts and 43 upregulated transcripts with diminished m6A modification (Fig. 5G). Notably, endocrine transcriptional factors Nkx6.1 and Kcnq1 were downregulated, whereas Hnf1β, which participated in pancreatic ductal formation, was differentially upregulated upon Mettl3 deletion (Fig. 5G). Among m6A-mediated downregulated transcripts, we noticed a preferential change in Hdac1 expression, which was an upstream suppressor of Wnt and Notch pathways (37) (Fig. 5H). Interestingly, Hdac1 mRNAs were significantly downregulated in bipotent progenitor cells but had comparable expression levels in endocrine progenitor cells (Supplementary Fig. 5A and B). Integrative Genomics Viewer tracks showed Hdac1 transcripts had abundantly enriched m6A peaks throughout their mRNA in WT mice but exhibited diminished m6A peaks upon Mettl3 knockdown (Fig. 5H). We further demonstrated that m6A enrichment in the mRNA of Hdac1 was decreased in E15.5 Mettl3pKO pancreatic cells by MeRIP quantitative PCR (Fig. 5I). Moreover, by conducting RIP quantitative PCR on P0 pancreata, we found that the Mettl3 protein could interact with Hdac1 mRNA (Fig. 5J).
It is known that m6A modifications on mRNA transcripts might affect mRNA stability and translation. We then dissected E13.5 intact embryonic pancreatic buds and treated them with ShMettl3 lentivirus or control virus for 48 h to assess the direct effect of Mettl3 on Hdac1 expression. We found that the protein abundance of Hdac1 was significantly decreased after 48-h Mettl3 knockdown in embryonic pancreata (Fig. 5K). ShMettl3 treatment on E13.5 pancreatic buds increased the decay rate of Hdac1 mRNA after addition of transcriptional inhibitor actinomycin D, indicating that Mettl3 deletion directly reduced the stability of Hdac1 mRNA (Fig. 5L). We then performed Hdac1 immunofluorescence staining in situ on E15.5 WT and Mettl3pKO embryonic pancreata. We found that the Hdac1 fluorescence intensity of Sox9+ bipotent progenitors was significantly reduced in Mettl3pKO mice (Fig. 5M). Meanwhile, the expression of β-catenin, the core protein of the Wnt signaling pathway, and that of Hes1, the effector of the Notch signaling pathway, were simultaneously upregulated in embryonic Mettl3pKO pancreatic cells (Fig. 5N).
Mettl3/Hdac1 Pathway Controls Endocrine Lineage Determination
To understand whether Hdac1 mediated the differentiation of endocrine progenitors from the bipotent trunk, we first treated E13.5 WT embryonic pancreata with Hdac1 inhibitor parthenolide or DMSO for 2 days. Parthenolide not only inhibited Hdac1 expression but also induced the activation of Wnt/β-catenin signaling and Notch/Hes1 signaling (Fig. 6A). Meanwhile, 48-h inhibition of Hdac1 activity by parthenolide treatment significantly reduced the number of Ngn3+ cells but also increased the number of Sox9+ cells in embryonic pancreatic buds (Fig. 6B and C).
Mettl3 regulates endocrine progenitor differentiation via Hdac1. A: Representative pancreatic sections from E13.5 WT pancreata cultured with parthenolide or DMSO for 48 h were immunostained for Hdac1/β-catenin/Hes1 (red; n = 5–6). B: Representative pancreatic sections from E13.5 WT pancreata cultured with parthenolide or DMSO for 48 h were immunostained for Ngn3 (green) and Sox9 (red; n = 3). C: Numbers of Ngn3+ and Sox9+ cells were calculated in E13.5 WT pancreata cultured with parthenolide or DMSO for 48 h (n = 3). D: Representative pancreatic sections from E13.5 WT and Mettl3pKO pancreata cultured with ITSA-1 or DMSO for 48 h were immunostained for Hdac1/β-catenin/Hes1 (red; n = 5–6). E: Representative pancreatic sections from E13.5 WT and Mettl3pKO pancreata cultured with ITSA-1 or DMSO for 48 h were coimmunostained for Ngn3 (red)/insulin (green) and Sox9 (red; n = 3–4). F: Numbers of Ngn3+, insulin+, and Sox9+ cells were calculated in E13.5 WT and Mettl3pKO pancreata cultured with ITSA-1 or DMSO for 48 h (n = 3–4). Data were presented as mean ± SEM of independent experiment indicated above. Student t test was. Nuclei were counterstained with DAPI (blue). Scale bars, 50 μm. *P < 0.05, **P < 0.01, ***P < 0.001.
Mettl3 regulates endocrine progenitor differentiation via Hdac1. A: Representative pancreatic sections from E13.5 WT pancreata cultured with parthenolide or DMSO for 48 h were immunostained for Hdac1/β-catenin/Hes1 (red; n = 5–6). B: Representative pancreatic sections from E13.5 WT pancreata cultured with parthenolide or DMSO for 48 h were immunostained for Ngn3 (green) and Sox9 (red; n = 3). C: Numbers of Ngn3+ and Sox9+ cells were calculated in E13.5 WT pancreata cultured with parthenolide or DMSO for 48 h (n = 3). D: Representative pancreatic sections from E13.5 WT and Mettl3pKO pancreata cultured with ITSA-1 or DMSO for 48 h were immunostained for Hdac1/β-catenin/Hes1 (red; n = 5–6). E: Representative pancreatic sections from E13.5 WT and Mettl3pKO pancreata cultured with ITSA-1 or DMSO for 48 h were coimmunostained for Ngn3 (red)/insulin (green) and Sox9 (red; n = 3–4). F: Numbers of Ngn3+, insulin+, and Sox9+ cells were calculated in E13.5 WT and Mettl3pKO pancreata cultured with ITSA-1 or DMSO for 48 h (n = 3–4). Data were presented as mean ± SEM of independent experiment indicated above. Student t test was. Nuclei were counterstained with DAPI (blue). Scale bars, 50 μm. *P < 0.05, **P < 0.01, ***P < 0.001.
We then explored whether restored expression of Hdac1 in Mettl3pKO embryonic pancreata could rescue the impaired differentiation toward endocrine cells. We dissected E13.5 WT and Mettl3pKO embryonic pancreatic buds and added Hdac1 agonist ITSA-1 or DMSO to the culture medium. After 2 days of culture, we found that ITSA-1 treatment significantly increased the expression of Hdac1 in WT pancreatic buds (Fig. 6D), without changing Wnt/β-catenin or Notch/Hes1 signaling (Fig. 6D). ITSA-1 treatment did not influence the number of Ngn3+/Ins+ or Sox9+ cells in WT pancreatic buds (Fig. 6E–G). Importantly, ITSA-1 treatment prevented the abnormal induction of Wnt/β-catenin and Notch/Hes1 signaling (Fig. 6D) and rescued the decrease in Ngn3+ endocrine cells, along with Ins+ β-cells, and the increase in Sox9+ cells (Fig. 6E–G) in Mettl3pKO pancreatic buds. The above data strongly indicate that the Mettl3/Hdac1 pathway is required for endocrine lineage determination during pancreatic development.
Discussion
In the last few decades, the incidence of early-onset diabetes has gradually increased all around the world (38–41). Genetic factors, together with the possible pathogenic role of epigenetics (which may affect β-cell development in individuals with a family history of diabetes), might play particularly important roles in early-onset diabetes (42). Recent advances in cell purification strategies and sequencing technologies as well as novel molecular tools have revealed that epigenetic modifications, such as DNA methylation, histone modifications, and noncoding RNAs, represent an integral part of the transcriptional mechanisms regulating pancreatic development and endocrine differentiation (43–47). It has been reported that loss of DNA methyltransferase 1 in pancreatic progenitors results in disruption of all three pancreatic lineages and causes apoptosis of progenitor cells because of repression of p53 (45). Xu et al. (46) inhibited Ezh2, a histone methyltransferase member of H3K27me3, in embryonic pancreatic explants and thus increased endocrine progenitor cells. In addition, conditional deletion of the miRNA processing enzyme Dicer1 in pancreatic progenitors caused reduced numbers of endocrine progenitor as well as endocrine cells. Recently, we and others demonstrated that m6A methylation modification affects neonatal β-cell maturation, as well as mature β-cell function and apoptosis (17–20). However, the role of m6A regulation in pancreatic development and endocrine differentiation is currently unknown.
In the current study, we demonstrated that m6A mRNA modification was required for the establishment of an endocrine progenitor pool by using transgenic Mettl3pKO mice. Mettl3pKO mice developed severe hyperglycemia and insulin deficiency at age 2 weeks, accompanied by acinar atrophy, which arose from reduced proliferation, islet mass, and endocrine cell numbers. At E15.5, loss of Mettl3 led to a dramatic reduction in the number of Ngn3+ endocrine progenitor cells and reduced expression of endocrine progenitor markers like Ngn3, Nkx2.2, and Nkx6.1 in mutated pancreata. It is known that a sufficient endocrine progenitor pool has central importance in the regulation of blood glucose levels after birth (48,49). The above data indicate an important role of Mettl3 in controlling the size of the endocrine progenitor pool during pancreatic development and thus influencing the glycemia level after birth.
During pancreatic development, a variety of transcription factors form a complex gene regulatory network culminating in different cell lineages, including acinar, ductal, and endocrine components (50). At the secondary transition (E12.5–E16.5), bipotent progenitor cells were divided into either ductal cells retaining Sox9 or endocrine progenitor cells, initiated by the transient activation of Ngn3 (7,8). Interestingly, along with the decrease in endocrine progenitor cells, abnormal increases in cells expressing Sox9 at E15.5 were detected in Mettl3pKO embryonic pancreata. Meanwhile, proacinar cells staining for Cpa1 remained unchanged, revealing that depletion of Mettl3 did not affect acinar cell differentiation at the secondary transition of pancreatic development. Single-cell transcriptome analysis further confirmed that bipotent progenitor cells in the late stage, also considered precursors of endocrine progenitors based on the developmental trajectory, were inappropriately accumulated in the mutated pancreas. The absence of Mettl3 in the late stage of bipotent progenitor cells led to the upregulation of several genes required for early epithelial cell differentiation and epithelial tube formation (e.g., Nr2f2, Sox4, Sox11, Wnt4, Hnf1β, and Onecut2) and the downregulation of factors essential for endocrine cell differentiation, like Nkx6.1, Kcnq1, and Hdac1. It is known that in retinal progenitor cells, Hdac1 can suppress Wnt and Notch signaling pathways (37), both of which have been reported to negatively regulate endocrine differentiation during pancreatic development (51,52). In the current study, we demonstrated that Mettl3 could regulate Hdac1 in an m6A-dependent manner during the transition of endocrine specification. This is consistent with the findings of Chen et al. (53), who reported that Mettl3 mediated the m6A modification of Hdac1 mRNA, and knockdown of Mettl3 could inhibit Hdac1 expression in the rodent liver. Mettl3 knockdown in the embryonic pancreas via shRNA reduced Hdac1 protein abundance by regulating its mRNA stability. Hdac1 protein expression was reduced and its downstream signals Wnt/β-catenin and Notch/Hes1 were activated in Sox9+ bipotent progenitors of E15.5 Mettl3pKO pancreata. The above observations thus provide a novel m6A/Hdac1 epigenetic regulatory pathway that might determine the endocrine fate of bipotent progenitors during pancreatic development. The other altered genes might also play collaborative roles in Mettl3-deficient pancreatic cells, which merits further investigation.
In the pancreatic explant system, we demonstrate that the Mettl3-Hdac1-Wnt/Notch axis may affect the formation of endocrine progenitor pools. We found that inactivation of Hdac1 in E13.5 intact pancreatic buds for 48 h indeed activated Wnt/Notch signaling, reduced Ngn3+ cells, and increased Sox9+ cells. Moreover, the addition of Hdac1 agonist to E13.5 Mettl3pKO pancreatic buds was able to prevent the activation of Wnt/Notch signaling and partially restore the Ngn3+ endocrine pool. Our novel finding highlights a permissive role of Mettl3-Hdac1-Wnt/Notch in directing the fate of pancreatic bipotent progenitors toward Ngn3+ endocrine progenitors, creating a developmental milieu that can be extended to manipulate embryonic stem cells or induced pluripotent stem cells cells toward an endocrine fate. A deeper understanding of the role of m6A methylation in human endocrine cell biology may help in finding therapeutic strategies that improve the process of maintaining or regenerating the functional β-cell mass.
This article contains supplementary material online at https://doi.org/10.2337/figshare.24549739.
J.Su. and Y.W. contributed equally to this work.
Article Information
Acknowledgments. The authors thank Ming-Han Tong (University of Chinese Academy of Sciences, Shanghai, China) for providing the Mettl3flox/flox mice and for valuable discussion and comments on the study.
Funding. This work was supported by grants from the National Natural Sciences Foundation of China (82270841, 82070795, 81870527, and 91857205) and the National Key Research and Development Program of China (2021YFC2501601).
Duality of Interest. No potential conflicts of interest relevant to this article were reported.
Author Contributions. J.Su., Y.W., H.F., and F.K. performed experiments and analyzed the data. J.Su., Y.W., and Q.W. wrote the manuscript. J.So., M.X., and J.W. contributed to the data discussion. G.N., J.W., W.W., and Q.W. designed the project, supervised research, and coordinated the execution of the experimental plan. Q.W. 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.