METTL3 and METTL14 determine human neural fate specifications

Abstract

METTL3/METTL14, the canonical methyltransferase complex modifying N⁶-methyladenosine (m⁶A) on mRNAs, plays critical roles in development and various diseases. However, its precise functions in specifying neural fate from human embryonic stem cells (hESCs) remain poorly understood. Here, we demonstrate, using an inducible knockout system, that METTL3/METTL14 deletion impairs the generation of neural progenitor cells (NPCs) from hESCs. Furthermore, inducible METTL3/METTL14-deficient NPCs exhibit compromised long-term proliferation and fail to differentiate into neurons. Mechanistically, METTL3 is enriched in gene loci essential for neurogenesis and chromatin remodeling in human NPCs, thereby promoting chromatin accessibility at these sites. Importantly, forced expression of BRM, a catalytic component of the chromatin-remodeling BAF complex, in METTL3-deficient hESCs rescues the defects in neural fate specifications caused by METTL3 loss. Our study systematically defines the essential requirement of METTL3 and METTL14 in ensuring the fidelity and neuronal specification of human NPCs, and underscores the coordination among distinct epigenetic regulators in neural development.

Graphical Abstract

Graphical Abstract.

Introduction

During embryonic development, cell lineage fate determination is successively orchestrated by multiple factors, including lineage-specific transcription factors (TFs), epigenetic regulators, and signaling molecules [1, 2]. Neurogenesis, a pivotal event in early human embryonic development, exemplifies a highly ordered process that encompasses the specification of cell fates from totipotency towards neural lineages, ultimately generating neural progenitor cells (NPCs) and diverse functional subtypes of neurons and glial cells [3, 4]. In model organisms such as zebrafish and mice, cell fate specification during neurogenesis is precisely modulated by these regulators at multiple levels to ensure the proper generation of NPCs and their neural derivatives within the nervous system [1, 2, 5–7]. Dysregulation of neurogenesis can result in severe neural system disorders, including microcephaly and neurodegenerative diseases [8, 9]. However, our understanding of the precise mechanisms governing human neural cell specification remains limited, largely due to the lack of suitable human model systems.

Human pluripotent stem cells (hPSCs) provide a valuable in vitro tool to study embryonic development and model diseases, including neurogenesis and its associated disorders, and also represent a potential donor source for cell replacement therapies targeting neurodegenerative conditions [10–12]. Our previous studies revealed distinct requirements for the deposition and removal of the repressive histone modification mark H3 lysine 27 tri-methylation (H3K27me3) during the sequential specification of neural lineages from human embryonic stem cells (hESCs) to NPCs and further to neural subtypes [13–15]. Epigenetic regulators play crucial roles in modulating gene expression, thereby influencing cell fate decisions through mediating chromatin and RNA modifications, for example histone methylation such as H3K27me3 and RNA modifications such as N⁶-methyladenosine (m⁶A) [1, 16–19]. Conventionally, the METTL3/METTL14 complex forms a methyltransferase complex (MTC) responsible for m⁶A on mRNAs and plays critical roles in embryonic development and various diseases such as cancer by regulating RNA stability, decay, splicing, and translation [18, 20–22]. We recently reported that the METTL3/METTL14 complex maintains self-renewal and nucleolar integrity in hESCs [23]. Furthermore, studies in mouse models have indicated that METTL3 functions in neurogenesis and affects some processes such as long-term memory maintenance and newborn neuron maturation [24, 25]. Similarly, METTL14 deficiency in mice prolongs cortical neurogenesis by extending the cell cycle of radial glial cells [26]. Despite these insights, the precise roles and underlying mechanisms of METTL3 as well as METTL14 in directing human neurogenesis from hPSCs to NPCs and their subtypes remain poorly understood.

Here, we employed an inducible knockout system targeting METTL3 or METTL14 in hESCs to systematically investigate their critical functions in human neurogenesis. We demonstrate that METTL3/METTL14 depletion severely impairs the generation of NPCs from hESCs. Moreover, METTL3/METTL14-deficient NPCs exhibit defective long-term proliferation capacity and fail to differentiate into neurons. Mechanistically, METTL3 binds to gene loci essential for neurogenesis as well as chromatin remodeling in human NPCs (hNPCs) and promotes their chromatin accessibility at these sites. Our study defines an essential role for METTL3 and METTL14 in ensuring the fidelity of NPC generation and their neuronal specification potential during human neurogenesis.

Materials and methods

Construction of inducible knockout human embryonic stem cells

We first used a lentiviral system to induce overexpression of METTL3 or METTL14 in H1 cells. Subsequently, the CRISPR/Cas9 [clustered regularly interspaced palindromic repeats (CRISPR)/CRISPR-associated protein 9] system was employed to knock out METTL3 or METTL14, thereby generating inducible METTL3- or METTL14-overexpressing hESC lines in the presence of doxorubicin (DOX) [23].

Neural differentiation

hESCs were seeded at 100% confluence onto Matrigel-coated culture plates. On day 0, neural induction was initiated by culturing the cells in N2B27 medium [50% Dulbecco’s modified Eagle’s medium (DMEM)/F12 (Gibco, C11330500BT), 50% neurobasal medium (Gibco, 21103-049), 0.5× N2 supplement (Gibco, 17502-048), 0.5× B27 supplement (Gibco, 17504-044), 1% Glutamax (Gibco, 35 050 079), 1% non-essential amino acids (NEAAs; Gibco, 11 140 076), 5 μg/ml insulin (Sigma, I9278), and 1 μg/ml heparin (Sigma, H3149)] supplemented with 5 μM SB431542 (Selleck, S1067) and 5 μM dorsomorphin (Selleck, S7840). The medium was changed every other day during this initial induction phase. On day 8, the cells were gently scraped off and replated in N2B27 medium containing 10 μM Y27632 (Selleck, S1049). The following day (day 9), the medium was replaced with fresh N2B27 medium. Subsequent medium changes were performed every other day until day 16 to maintain the culture conditions.

Neural stem cell maintenance

On day 16, neural rosette-like cells were carefully picked and dissociated into single cells using Accutase (Sigma, A6964). These cells were then subjected to suspension culture in low-adhesion plates to generate neural progenitor cells at passage 1 (NPC-P1). The initial culture medium used was N2B27 supplemented with 10 μM Y27632. On the following day, the medium was replaced with N2B27 containing 20 ng/ml basic fibroblast growth factor (bFGF; Peprotech, 100-18C). After 7 days of culture, the cells were collected, dissociated again into single cells using Accutase, and cultured in suspension in low-adhesion plates to generate NPCs at passage 2 (NPC-P2). The medium used for NPC-P2 was again N2B27 supplemented with 10 μM Y27632. On the next day, the medium was changed to N2B27 containing 20 ng/ml bFGF. This protocol was repeated for subsequent passages, maintaining the same medium conditions and culture procedures to ensure consistent NPC maintenance.

Neuronal and glial cell differentiation

NPC-P2 neural stem cells were dissociated with Accutase, counted, and seeded onto Matrigel-coated 24-well plates at a density of 20 000 cells per well. The initial culture medium was N2B27 supplemented with 10 μM Y27632. On the following day, the medium was further supplemented with 20 ng/ml brain-derived neurotrophic factor (BDNF; Peprotech, 450-02), 20 ng/ml glial cell line-derived neurotrophic factor (GDNF; Peprotech, 450-10), and 1 μM cAMP (Sigma, D-0260). Medium changes were performed every other day, and the differentiation process was allowed to proceed for 28 days.

Quantitative reverse transcription–PCR and RNA sequencing

For RNA extraction, 1 million cells were collected and lysed using Trizol reagent. Subsequently, 2 µg of the extracted RNA was reverse-transcribed into cDNA using the HiScript IV RT SuperMix for qPCR (Vazyme, R423). The resulting cDNA was diluted 50-fold and used for quantitative reverse transcription–PCR (qRT–PCR) analysis with chamQ SYBR qPCR Master Mix (Vazyme, Q311). The qRT–PCR was performed using a CFX96 machine (Bio-Rad). Each sample included three biological replicates. All primer sequences are listed in Supplementary Table S1.

For RNA sequencing (RNA-seq), 1 µg of RNA from each sample was submitted to a sequencing service provider for library preparation and sequencing. Each sample included two biological replicates to ensure data reliability. The libraries of these samples were sequenced by the NextSeq 500 Mid Output Kit (Illumina). The raw sequencing data were mapped with the human mRNA reference sequence (GRCh38/hg38) using RSEM (rsem-1.2.4) and Bowtie2 (v2.2.5). The gene expression abundance was normalized and characterized through TPM (transcripts per million). These data were analyzed using glbase. The correlation analysis among samples was processed using R software. The differentially expressed genes among samples and heatmap analysis were processed by edgeR package and pheatmap, respectively.

Western blot

For protein extraction, 2 million cells were collected and centrifuged to remove the supernatant. The cell pellet was then lysed in 200 µl of RIPA buffer (Beyotime, P0013B) supplemented with 100× protease inhibitor cocktail (Roche, 04 693 132 001) and 1 mM phenylmethylsulfonyl fluoride (PMSF; Sigma, P7626). The lysates were sonicated to ensure complete disruption of cellular components, followed by mixing with 67 µl of 4× SDS loading buffer (Thermo, NP0008). The samples were boiled for 10 min to denature the proteins and then subjected to electrophoresis on a pre-cast 10% sodium dodecylsulfate (SDS)–polyacrylamide gel electrophoresis (PAGE) gel. After electrophoresis, proteins were transferred to a polyvinylidene difluoride (PVDF) membrane. The membrane was blocked to prevent non-specific binding and then incubated with primary antibodies at the dilutions specified by the manufacturers. Following thorough washing, the membrane was incubated with secondary antibodies conjugated to horseradish peroxidase. After additional washing, the proteins were visualized using an ECL chemiluminescent substrate and imaged using a GelView9000Lite machine (BLT). The details of the antibodies in this study are given in Supplementary Table S2.

Immunofluorescence

Cells were seeded onto 14 mm round coverslips coated with 400× Matrigel at an appropriate density. The following day, cells were fixed with 4% paraformaldehyde for 20 min at room temperature. After fixation, cells were washed three times with phosphate-buffered saline (PBS) for 5 min each. They were then incubated with primary antibodies diluted in blocking buffer overnight at 4°C for 16 h. After primary antibody incubation, cells were washed three times with PBS for 5 min each. Subsequently, cells were incubated with secondary antibodies and 4′,6-diamidino-2-phenylindole (DAPI; Sigma, D9542), both diluted in antibody dilution buffer, for 1.5 h at room temperature in the dark. Cells were then washed three times with PBS for 5 min each. Coverslips were mounted onto glass slides using mounting medium, inverted, and stored at 4°C in the dark. For long-term storage, samples were kept at −20°C. Fluorescence imaging was performed using a confocal laser scanning microscope LSM 800 (Zeiss). The details of the antibodies in this study are given in Supplementary Table S2.

Flow cytometry

A total of 500 000 cells were fixed by 200 µl of 1× Fixation Buffer (BD, 554 655). The fixed cells were incubated at room temperature in the dark for 30 min. Following fixation, the cells were washed once with PBS and centrifuged to pellet them. The supernatant was removed. The cells were then permeabilized by adding 1× Permeabilization Buffer (BD, 557 885) and incubating at 4°C for 15 min. After permeabilization, the cells were again washed with PBS. Next, the primary antibody was prepared by diluting it with 1× Permeabilization Buffer as recommended by the manufacturer and applied to the cells. The cells were incubated with the primary antibody at 37°C for 30 min. Afterward, the cells were washed once with PBS. Subsequently, the secondary antibody was diluted with 1× Permeabilization Buffer according to the recommended ratio and added to the cells. This was followed by incubation at 37°C for an additional 30 min. Finally, the cells were washed once more with PBS, resuspended in PBS, and analyzed using a CytoFlex (Beckman) flow cytometer. The details of the antibodies in this study are given in Supplementary Table S2.

5-Ethynyl-2'-deoxyuridine assay

One million cells were seeded into a Matrigel-coated culture plate. The cells were treated with 10 µM 5-ethynyl-2'-deoxyuridine (EdU; from Click-iT EdU Alexa Fluor 647 Flow Cytometry Assay Kit, Invitrogen, C10424) for 20 h. Subsequently, the cells were digested into single-cell suspension using Accutase, collected, and fixed with 200 µl of 1× Fixation Buffer at room temperature in the dark for 30 min. After fixation, the cells were washed once with PBS, centrifuged for supernatant removal, and permeabilized with 1× Permeabilization Buffer at 4°C for 15 min. A second wash with PBS was performed. The EdU reaction solution was prepared according to the manufacturer’s instructions, mixed with the cells, and incubated at room temperature for 1 h. Finally, the cells were washed with PBS, resuspended in PBS, and analyzed using a CytoFlex flow cytometer (Beckman).

Annexin V/propidium iodide apoptosis detection

A total of 50 000 cells were used for the Annexin V/propidium iodide (PI) analysis according to the manufacturer’s instructions from the Annexin V-FITC-PI Apoptosis Detection Kit (Vazyme, A211). The reaction solution must be prepared fresh and used immediately. These cells were resuspended with the Annexin V/PI solution and incubated at room temperature for 10 min, then 400 µl of 1× Binding Buffer was added. These samples were immediately analyzed using a CytoFlex (Beckman) Flow Cytometer.

CUT&TAG-seq

For library construction, 100 000 cells were used according to the procedures of the Hyperactive Universal CUT&Tag Assay Kit for Illumina Pro (Vazyme, TD904-01). The primary and secondary antibodies were used at the ratios according to the instruction manual. These sample libraries were conducted on a NextSeq 500 plateform.

These adaptors of raw files were removed using Cutadapt (v1.13) with parameters -m 35 -e 0.1. Then, the reference human genome (hg38) was used to align these processed sequences using bowtie2 (v2.4.1) with the recommended parameters in the official protocol (https://yezhengstat.github.io/CUTTag_tutorial/). The low-quality mapped reads and tagging duplicated reads were filtered out by Samtools (v1.3.1) and Picard Tools (v1.90). We carried out the peak calling of METTL3 using MACS2 (v2.1.0) with -log(q-value) > 4 for downstream analysis. The bigwig files were generated for visualization by DeepTools (v3.4.3) with bamCoverage, computeMatrix, plotHeatmap, and plotProfile modules. ChIPseeker (v1.26.0) in the R package was used for annotation of peaks with the parameters c (-3000, 3000) of tssRegion. Gene Ontology (GO) analysis and motif analysis were conducted by the R package clusterProfiler (v3.18.0) and the Motifs Genome module in Homer, respectively.The details of the antibodies in this study are given in Supplementary Table S2.

ATAC-seq

A total of 100 000 single cells were used for construction of the sequencing library following the protocol of the Hyperactive ATAC-Seq Library Prep Kit for Illumina (Vazyme, TD711). The fully constructed ATAC libraries were sequenced on a NextSeq 500 plateform. Adaptors were removed from the raw data using cutadapt (v1.13). The reads were aligned to the reference sequence of the human genome (UCSC hg38) using bowtie2 (v2.2.5). The duplicates were removed by Samtools (v1.3.1) and Picard Tools (v1.90). The tools, including MACS2 callpeak, bdgcmp, and deeptools (v2.4.2), were used to pile up signals and plot signal density heatmaps and profiles.

Results

METTL3 deficiency impairs the generation of neural progenitor cells

To investigate the precise roles of METTL3 in human neurogenesis, we performed default neural differentiation in hESCs under defined conditions with dual SMAD inhibitors [15] (Fig. 1A). Since METTL3 deletion in hESCs causes nucleolar stress and subsequent loss of self-renewal capacity, we established inducible knockout of METTL3 in hESCs (METTL3-OE/KO hESCs) through introducing an inducible exogenous METTL3 expression dependent on DOX treatment in wild-type (WT) hESCs and subsequently targeting endogenous METTL3 in these cells [23]. We then triggered the neural differentiation process in METTL3-OE/KO hESCs upon DOX treatment (DOX+) or withdrawal (DOX−). Upon differentiation at day 16, both WT cells and METTL3-OE/KO cells in the presence of DOX efficiently gave rise to NPCs, as evidenced by forming typical rosette-like morphology and expressing the NPC marker PAX6 (Fig. 1B, C). In contrast, in the absence of DOX, METTL3-OE/KO cells did not exhibit a rosette-like NPC phenotype and express the NPC marker PAX6 (Fig. 1B, C). Furthermore, the pluripotent markers OCT4 and NANOG in three cells were fully silenced, confirming complete differentiation (Fig. 1D). However, expression of the NPC marker genes SOX2, PAX6, and FOXG1 was higher in WT cells and METTL3-OE/KO cells plus DOX than those of METTL3-OE/KO cells without DOX at day 16 of neural differentiation at both mRNA and protein levels (Fig. 1D, E). Indeed, METTL3 was highly expressed in METTL3-OE/KO cells plus DOX, while METTL3 was not detected in METTL3-OE/KO cells without DOX (Fig. 1D, E). Moreover, immunostaining data confirmed the presence of rosette-like structures and high expression of NPC markers SOX2, NESTIN, and PAX6 in WT cells and METTL3-OE/KO cells with DOX treatment, but not in METTL3-OE/KO cells upon DOX withdrawal (Fig. 1F). Lastly, we also performed whole-genome transcriptome analysis from WT and METTL3-OE/KO cells with or without DOX at day 16 upon differentiation (Fig. 1G). Consistent with the above phenotype, down-regulated genes in METTL3-deficient (DOX−) cells compared with WT cells at day 16 were enriched in functions related to nervous system development, brain development, sensory system development, and so on (Fig. 1G), while the up-regulated genes in METTL3-deficient (DOX−) cells were associated with placenta, epidermis, and mesoderm development, including limb, circulation system, and kidney development (Supplementary Fig. S1A). Furthermore, these expression defects were rescued by DOX-induced METTL3 restoration (Fig. 1G). Together, these data demonstrate that METTL3 deficiency impairs the generation of NPCs from hESCs.

Figure 1.

NPC differentiation of inducible METTL3-KO hESCs. (A) Strategy overview of the neural differentiation from hESCs (see the Materials and methods for details). (B) Morphology of the WT, METTL3-OE/KO hESC lines without (DOX−) or with DOX (DOX+) treatment under NPC differentiation conditions at day 0, day 9, or day 16. Scale bar, 100 μm. METTL3-OE/KO hESCs [23], inducible knockout of METTL3 in hESCs through introducing an inducible exogenous METTL3 expression dependent on DOX treatment in WT hESCs and subsequently targeting endogenous METTL3 in these cells. (C) Fluorescence-activated cell sorting (FACS) results for NPC marker PAX6⁺ cells at day 8 and day 16 during neural differentiation in the indicated cells. (D) The expression level of the pluripotent genes OCT4/NANOG, the NPC genes SOX2/PAX6/FOXG1, and METTL3 using qRT–PCR at day 0, day 8, and day 16 during neural differentiation. WT cells served as controls. (E) The expression of PAX6 and METTL3 proteins using western blot in the indicated cells at day 8 and day 16 during neural differentiation, respectively. (F). Immunostaining analysis for the NPC markers SOX2/NESTIN/PAX6 in WT and METTL3-OE/KO cells in the presence or absence of DOX. Scale bar, 50 μm. (G) Left panel, heatmap of up- or down-regulated genes in WT and METTL3-OE/KO cells in the absence of DOX at day 16 of neural differentiation. Right panel: top GO terms for down-regulated genes in METTL3-OE/KO cells without DOX. Significance was determined by unpaired two-tailed Student’s t-test. ***P < 0.001. **P < 0.01. All error bars represent the SD (standard deviation) from three independent replicates (n = 3).

METTL3 is required for the long-term maintenance of hNPCs

To further investigate the roles of METTL3 in the long-term maintenance and neural subtype differentiation of hNPCs, we picked typical rosette-like cells derived from METTL3-OE/KO cells with DOX (METTL3-OE/KO DOX+) cells at day 16 upon neural differentiation (Fig. 2A). Then, these cells were expanded or differentiated under defined medium in the absence or presence of DOX (Fig. 2A). Firstly, WT NPCs formed typical neural spheres from passage 1 (NPC-P1) through passage 4 (NPC-P4) (Fig. 2B). The neural spheres derived from METTL3-OE/KO cells without DOX (METTL3-OE/KO DOX−) were similar to WT cells at early passages (NPC-P1 and NPC-P2), but exhibited severe defects compared with those of WT cells at later passages (NPC-P3 and NPC-P4) evidenced by reduced size and number (Fig. 2B, C). Moreover, DOX treatment rescued the defective phenotype of neural spheres from METTL3-OE/KO cells without DOX at later passages (Fig. 2B, C). These data demonstrate that METTL3 loss compromises the long-term proliferation of hNPCs.

Figure 2.

METTL3 loss results in a poor long-term proliferation of NPCs. (A) Strategy overview of NPC fidelity validation including the proliferation and subtype differentiation. (B) Morphology (left panel) and proliferation curve (right panel) of WT hNPCs and METTL3-OE/KO hNPCs without (DOX−) or with DOX (DOX+) treatment under defined NPC maintenance conditions at different passages (P). Scale bar, 200 μm. (C) Western blot analysis for PAX6, METTL3, and P53 proteins and quantification on P53 protein in the indicated NPCs at P2 or P4. (D) FACS results for NPC marker PAX6⁺ cells in the indicated NPCs at passage 4 (P4). (E) The expression level of the NPC genes SOX2/PAX6/SOX1 using qRT–PCR in the indicated NPCs at passage 2 (P2) or passage 4 (P4). (F and G) EdU incorporation assay (F) and apoptosis assay (G) of the indicated NPCs at P2 or P4. Early apoptosis, PI negative- and Annexin V-positive cells. Late apoptosis, PI positive- and Annexin V-positive cells. (H) Principal component analysis of the transcriptome profiles in the indicated NPCs at P2 or P4. (I) Left panel: heatmap of up- or down-regulated genes in WT NPCs and METTL3-OE/KO NPCs in the absence of DOX at P4. Right panel: top GO terms and GO terms related to proliferation in down-regulated genes in METTL3-OE/KO NPCs without DOX at P4. Significance was determined by unpaired two-tailed Student’s t-test. **P < 0.01. *P < 0.05. All error bars represent the SD from three independent replicates (n = 3).

To study how METTL3 deficiency results in a defect in NPC proliferation, we analyzed NPC marker expression. Notably, these defective neural spheres from METTL3-OE/KO DOX− cells at NPC-P4 remained in an undifferentiated state evidenced by high expression of NPC markers including PAX6, SOX1, and SOX2 at mRNA and protein levels, similar to METTL3-OE/KO DOX+ and WT cells (Fig. 2C–E). Furthermore, we showed that the percentage of proliferating cells (EdU⁺) was significantly decreased while the apoptosis level (Annexin V⁺) and P53 protein level were greatly increased in METTL3-deficient NPCs (METTL3-OE/KO DOX−) compared with WT NPCs at later passage (P4) (Fig. 2C–G). These defects in EdU incorporation, apoptosis, and P53 protein level were effectively rescued by DOX-induced METTL3 expression (Fig. 2C–G). We then performed RNA-seq analysis for WT, METTL3-OE/KO DOX−, and METTL3-OE/KODOX+ NPCs at P2 and P4 (Fig. 2H). Consistently, the transcriptome profiles were similar among WT, METTL3-OE/KO DOX−, and METTL3-OE/KO DOX+ NPCs at P2 (Fig. 2H), while METTL3-OE/KO DOX− NPCs were distinct from WT and METTL3-OE/KO DOX+ NPCs at P4 (Fig. 2H). Furthermore, down-regulated genes in METTL3-deficient P4 NPCs were related to NPC proliferation and neuronal development (Fig. 2I). DOX treatment restored the transcriptome similar to that of WT cells (Fig. 2H, I). Together, these data demonstrate that METTL3 is required for the long-term self-renewal and proliferation in hESC-derived NPCs.

METTL3-deficient NPCs fail to differentiate into neurons

To further characterize the role of METTL3 in neuron generation, we performed spontaneous differentiation from NPCs at P2 by supplementing the culture with BDNF, GDNF, and cAMP (Fig. 3A). We utilized the picked METTL3-OE/KO DOX+ NPCs at P2 as donor NPCs for random differentiation. Indeed, WT NPCs effectively differentiated into neurons, displaying the typical neuronal morphology and highly expressing neuronal marker microtubule-associated protein 2 (MAP2) at day 28 (Fig. 3B–D). In contrast, in the absence of DOX, METTL3-OE/KO NPCs failed to differentiate into neurons, as indicated by morphology and MAP2 expression at day 28 (Fig. 3B–D). In addition, the astrocytes were normally differentiated from these cell lines at day 28 (Fig. 3D). Interestingly, upon DOX withdrawal, these METTL3-OE/KO cells maintained a colony-like morphology and continued to express high levels of two NPC markers SOX2 and NESTIN at day 28 of differentiation (Fig. 3B, E). Indeed, neuronal markers TUJ1, MAP2, and NEUROD1 were suppressed, while the NPC genes SOX1 and PAX6 remained highly expressed in METTL3-OE/KO cells without DOX at day 28 of spontaneous differentiation (Fig. 3F). Furthermore, addition of DOX effectively restored the neuron generation in METTL3-OE/KO cells at day 28 (Fig. 3B–F). These data indicate that METTL3 loss blocks the neuronal generation from NPCs.

Figure 3.

METTL3-deficient NPCs fail to differentiate into neurons. (A) Strategic diagram of the neural subtype differentiation from human NPCs at passage 2 (P2) under defined medium (see the Materials and methods for details). (B) Morphology of the indicated NPCs in their undifferentiated state and their differentiated cells at day 28. Scale bar, 50 μm. (C) Western blot analysis for the neuronal marker MAP2 and METTL3 in the indicated undifferentiated NPCs and their differentiated cells at day 28. (D) Immunostaining and quantification results on MAP2 and glial fibrillary acidic protein (GFAP; a glial marker) in the indicated undifferentiated NPCs or differentiated cells. Scale bar, 50 μm. (E) Immunostaining and quantification results on the NPC markers SOX2 and NESTIN in the indicated undifferentiated NPCs or differentiated cells. Scale bar, 50 μm. (F) qRT–PCR analysis of the neuronal markers TUJ1/MAP2/NEUROD1, the NPC markers SOX1/PAX6, and METTL3 in the indicated differentiated cells at day 28. (G) Spearman’s rank correlation analysis of the transcriptome profiles of the indicated differentiated NPCs at day 28. (H) Heatmap (left panel) and GO analysis (right panel) of up- or down-regulated genes in WT differentiated cells and METTL3-OE/KO differentiated cells in the absence of DOX at day 28. Significance was determined by unpaired two-tailed Student’s t-test. ***P < 0.001. ns, no significance. All error bars represent the SD from three independent replicates (n = 3).

We then analyzed RNA-seq data from differentiated WT, METTL3-OE/KO DOX+, and METTL3-OE/KO DOX− cells at day 28 (Fig. 3F). Correlation analysis revealed a clear differentiation defect in METTL3-deficient cells compared with WT cells at day 28, while the transcriptome of METTL3-OE/KO cells with DOX was similar to that of WT controls (Fig. 3G). Heatmap analysis further showed that down-regulated genes in METTL3-deficient cells were related to neurogenesis, synapse organization, etc, whereas up-regulated genes were enriched in extracellular structure organization and cell adhesion, etc (Fig. 3H). Together, these data demonstrate that METTL3 is essential for the fate transition from NPCs into neurons.

METTL14 is also essential for neural fate specifications from hESCs

Conventionally, METTL3 functions within the METTL3/METTL14 methyltransferase complex (MTC) to deposit m⁶A. To validate whether the essential roles of METTL3 depend on the intact MTC, we conducted the neural differentiation process in METTL14-deficient hESCs. Similar to the METTL3 model, we employed an inducible METTL14 knockout hESC line (METTL14-OE/KO) for neural induction [23] (Fig. 4A). Notably, in the absence of DOX, METTL14-OE/KO cells failed to undergo normal neural differentiation, as detected by morphology and NPC marker expression upon differentiation at day 16 (Fig. 4A–C; Supplementary Fig. S1B). In contrast, METTL14-OE/KO cells with DOX treatment (METTL14-OE/KO DOX+) could restore the NPC generation, similar to WT cells (Fig. 4A–C). We further picked typical rosette-like cells derived from METTL14-OE/KO DOX+ cells at day 16 of neural differentiation for subsequent NPC proliferation and random differentiation experiments (Fig. 4D). Consistently, at later passages, METTL14-deficient NPCs exhibited severe proliferation defects in terms of reduced size and cell number, decreased EdU incorporation, and increased apoptosis of neural spheres compared with WT cells (Fig. 4E–G). Moreover, we employed the picked METTL14-OE/KO DOX+ NPCs for random differentiation and showed that METTL14-deficient (DOX−) NPCs fail to generate neurons evidenced by morphology and the silencing of the neuronal marker MAP2 at day 28, while astrocyte differentiation remained unaffected (Fig. 4H, I). WT NPCs underwent normal neuronal differentiation including typical neuronal morphology and robust MAP2 expression at day 28 (Fig. 4H, I). In the presence of DOX, METTL14-OE/KO NPCs exhibited similar proliferation capability and neuronal differentiation to WT controls (Fig. 4D–I). Taken together, these data demonstrate that METTL3 and METTL14 determine neural fate specifications from hESCs (Fig. 4J).

Figure 4.

METTL14 is required for neural fate specifications from hESCs. (A) Morphology and immunostaining analysis for the NPC markers NESTIN/PAX6 in the WT, METTL14-OE/KO hESC lines without or with DOX treatment under NPC differentiation conditions at day 16. Scale bar, 100 μm or 50 μm. (B) FACS results for NPC marker PAX6⁺ cells at day 16 during neural differentiation in the indicated cells. (C) The expression level of the NPC genes PAX6/FOXG1, and METTL14 using qRT–PCR at day 0, day 8, and day 16 during neural differentiation. (D) Strategy overview of NPC fidelity validation including the proliferation and random differentiation. (E) Morphology and proliferation curve of the WT hNPCs and METTL14-OE/KO hNPCs without (DOX−) or with DOX (DOX+) treatment at different passages (P). Scale bar, 200 μm. (F and G) EdU incorporation assay (F) and apoptosis assay (G) of the indicated NPCs at P2 or P4. Early apoptosis, PI negative- and Annexin V-positive cells. Late apoptosis, PI positive- and Annexin V-positive cells. (H) Morphology and immunostaining results for neuronal markers MAP2/NEUN, glial marker GFAP, and NPC markers SOX2/NESTIN for the indicated differentiated cells at day 28 of random differentiation. Scale bar, 50 μm. (I) Quantification results for immunostaining analysis(H) on MAP2/NEUN, GFAP, and SOX2/NESTIN in the indicated differentiated cells. Scale bar, 50 μm. (J) Schematic model of METTL3 and METTL14 regulating neural fate determinations in hESCs. Significance was determined by unpaired two-tailed Student’s t-test. ***P < 0.001. *P < 0.01. All error bars represent the SD from three independent replicates (n = 3).

METTL3 occupies neurogenesis genes in hNPCs

To investigate the molecular mechanisms underlying the neurogenesis defects induced by METTL3 deficiency, we performed CUT&TAG-seq for the genome-wide mapping and enrichment of METTL3 in METTL3-OE/KO NPCs at P2 with or without DOX treatment (Fig. 5A). The overall METTL3 intensity was enriched in METTL3-OE/KO NPCs plus DOX but not in the absence of DOX (Fig. 5A). Furthermore, the enriched peaks of METTL3 were localized at promoter regions of genes which were related to transcription, cell cycle, nervous development, neuron differentiation, chromatin remodeling, etc in METTL3-OE/KO NPCs upon DOX treatment at P2 (Fig. 5B, C), indicating that METTL3 might regulate the transcription of neurogenesis genes. Furthermore, we analyzed the binding motifs of known TFs enriched in these METTL3-associated promoter regions (Fig. 5D). All the top five motifs have been reported to be critical neural regulators, including NFY, SP1, SP5, and SP2 [27–30]. In addition, METTL3-binding genes exhibited high expression levels and open chromatin accessibility in both WT NPCs at P2 and neurons (Fig. 5E, F). Indeed, METTL3 bound many known critical NPC regulators (PAX6, SOX1, and SOX2), NPC proliferation markers (NES, FZD9, and LHX2), and neuronal regulators (TUBB3, NEUROD1, and NEUROG2) (Fig. 5G), indicating that this might provide a potential explanation for the neurogenesis defects, including impaired NPC generation, poor NPC proliferation, and failure in neuron differentiation, observed in METTL3-deficient hESCs.

Figure 5.

METTL3 binds neurogenesis genes in human NPCs. (A) Signal densities and heatmap of CUT&TAG-seq data from METTL3-associated peaks in METTL3-OE/KO hNPCs without (DOX−) or with DOX (DOX+) treatment at P2. (B) Pie chart of METTL3-associated peaks in METTL3-OE/KO hNPCs with DOX treatment at P2. (C) Top GO terms of METTL3-associated genes mapped from peaks in promoter regions. (D) Top five motifs of METTL3-associated genes mapped from peaks in promoter regions. (E) mRNA expression level of METTL3-associated genes or other genes in WT NPCs and differentiated cells at day 28. (F) Chromatin accessibility of METTL3-associated genes or other genes in WT NPCs and differentiated cells at day 28. (G) Genomic views of the METTL3 binding of genes related to neural development. (H and I) qRT-PCR and western blot analysis of BRM, a component of the BAF complex in the indicated cells at day 16 of neural differentiation, NPCs at P4, and differentiated cells at day 28. Significance was determined by unpaired two-tailed Student’s t-test. **P < 0.01. All error bars represent the SD from three independent replicates (n = 3).

Interestingly, METTL3 also occupied the promoters of chromatin-remodeling regulators, such as components of the BAF (Brg/Brm-associated factor) complex which has been reported to be critical in neural development [31, 32] (Fig. 5C, G). We then examined the expression level of the core BAF subunit BRM in METTL3-OE/KO and METTL14-OE/KO cells without DOX (Fig. 5H, I; Supplementary Fig. S1C). Consistent with the defective phenotypes, after DOX withdrawal, the mRNA and protein levels of BRM were substantially reduced in METTL3-OE/KO or METTL14-OE/KO cells at the NPC generation stage (NPC-Day 16), the later P4 of NPCs (NPC-P4), and the neuron differentiation stage (Neuron-Day 28) (Fig. 5H, I; Supplementary Fig. S1C), indicating that decreased BRM may contribute to the neurogenesis failure in METTL3- or METTL14-deficient models. Together, these data demonstrate that METTL3 directly binds and regulates the expression of neurogenesis genes.

BRM rescues neurogenesis defects in METTL3-deficient hESCs

To investigate whether BRM could rescue the neurogenesis defects caused by METTL3 loss, including impaired NPC generation, maintenance, and neuron differentiation, we introduced a lentivirus expressing BRM into METTL3-OE/KO hESCs and performed neural differentiation in vitro (Fig. 6A). Upon DOX withdrawal, METTL3-OE/KO hESCs expressing exogenous BRM, comparable with WT cells, underwent normal neural differentiation, as evidenced by forming a typical rosette-like NPC morphology, expressing marker genes, and whole-genomic gene expression profiles at day 16 (Fig. 6B–D; Supplementary Fig. S1D, E). In contrast, METTL3 loss blocked NPC generation in METTL3-OE/KO cells without DOX (Fig. 6B–D). We then utilized the picked METTL3-OE/KO DOX+ NPCs as control NPCs and withdrew DOX treatment for further proliferation analysis and random differentiation experiments. Indeed, in the absence of DOX, exogenous BRM restored the normal proliferation capability, evidenced by the size and cell number of neural spheres, EdU insertion, apoptosis analysis, and whole-genomic gene expression profiles, which were distinct from those of METTL3-OE/KO NPCs at later passage (NPC-P4) (Fig. 6E–G; Supplementary Fig. S1F, G). In addition, the forced expression of BRM also rescued the neuronal differentiation defects induced by METTL3 deficiency, evidenced by morphology changes and expression levels of marker genes, and whole-genomic gene expression profiles (Fig. 6H, I; Supplementary Fig. S1H, I). Interestingly, based on ATAC-seq results, the closed chromatin regions induced by METTL3 deficiency were re-opened in BRM-rescued cells at day 16 of neural differentiation (NPC-Day 16) and passage 2 of NPCs (NPC-P2) (Supplementary Fig. S1J, K). Moreover, the open chromatin regions by METTL3 deficiency were decreased in BRM-rescued cells (Supplementary Fig. S1J, K). Collectively, these data demonstrate that reduced BRM is responsible for neurogenesis defects in METTL3-deficient hESCs, and BRM could drive neural fate transition independently of METTL3.

Figure 6.

BRM restores normal neurogenesis in METTL3-deficient hESCs. (A) Strategic diagram of rescue experiments of BRM in METTL3-OE/KO hESCs without DOX during neurogenesis. (B) Morphology and FACS analysis of the WT, METTL3-OE/KO hESC lines without DOX (−DOX) or with BRM overexpression (+BRM) under NPC differentiation conditions at day 16. Scale bar, 200 μm. (C) qRT-PCR analysis of the NPC markers PAX6/SOX1, BRM, and METTL3 in the indicated cell lines at day 0 and day 16 of neural differentiation. (D) Immunostaining analysis for the NPC markers SOX2/NESTIN/PAX6 in METTL3-OE/KO NPCs with BRM overexpression (+BRM). Scale bar, 50 μm. (E) Morphology and proliferation curve of the WT, METTL3-OE/KO hESC lines without DOX (−DOX) or with BRM overexpression (+BRM) under defined NPC maintenance conditions at passage 1 (P1) or P4. Scale bar, 200 μm. (F and G) EdU incorporation assay (F) and apoptosis assay (G) of the indicated NPCs at P4. (H) Morphology, immunostaining, and quantification results on MAP2, SOX2, and NESTIN of the indicated randomly differentiated cells at day 28. Scale bar, 50 μm. (I) qRT-PCR analysis of the NPC markers PAX6/SOX1, the neuronal marker MAP2/TUJ1, BRM, and METTL3 of the indicated randomly differentiated cells at day 28. Significance was determined by unpaired two-tailed Student’s t-test. **P < 0.01. All error bars represent the SD from three independent replicates (n = 3).

Discussion

In this study, we uncover the essential roles of METTL3, the core catalytic subunit of canonical m⁶A methyltransferase, and METTL14 in orchestrating the sequential specification of neural cell fates during human neurogenesis. We demonstrate that METTL3/METTL14 is indispensable for the generation and maintenance of NPCs derived from hESCs, as well as for their subsequent differentiation into neurons. Furthermore, METTL3 binds directly to key neurogenesis-associated genomic loci, including those encoding components of the chromatin-remodeling BAF complex, and promotes chromatin accessibility in human NPCs. Collectively, our findings systematically define METTL3/METTL14 as a critical guardian ensuring the fidelity of generation of NPCs and their subsequent neuronal specification potential during human neurogenesis.

Neurogenesis entails a highly orchestrated process and demands exquisite coordination among diverse regulators, such as neural lineage-specific TFs and epigenetic regulators. Our data highlighted the crucial roles of METTL3 and METTL14 in the sequential neurogenesis process, which would be a study reference for other epigenetic regulators in neural development. Our prior work revealed the coordination of the neural TF SOX2 and the H3K27me3 methyltransferase EZH2 in human neural fate decision [13]. The BAF chromatin-remodeling complex gene, also implicated as a vital regulator for neural development by altering chromatin structure dependent on its ATPase activity and interaction with other epigenetic regulators [32, 33], emerges here as a key downstream target of METTL3. Our data show that BRM, a catalyzed subunit of the BAF complex, is regulated by METTL3 throughout human neurogenesis progression from hESCs to hNPCs and then to neurons. METTL3 or METTL14 depletion not only recapitulated neurogenic defects (impaired NPC generation, compromised long-term NPC proliferation, and failed neuronal differentiation) but also led to BRM down-regulation. Significantly, forced expression of BRM completely rescued these neurogenic defects and reset the permissive chromatin accessibility and gene expression profiles in METTL3-deficient cells, functionally linking BRM as a critical effector downstream of METTL3. These data suggest that the permissive accessibility of neurogenesis genes regulated by METTL3 or BRM might be essential for neural differentiation. Intriguingly, deficiency in the H3K27me3 demethylase KDM6 similarly disrupted NPC proliferation and neuronal generation, accompanied by BAF complex dysregulation [15]. This convergence suggests a potential mechanistic link where METTL3 may cooperate with KDM6 to orchestrate the dynamic transitions of the BAF complex during human neural development. Unraveling the precise interplay between METTL3/METTL14, KDM6, and the BAF complex represents a compelling avenue for future investigation. To investigate the potential mechanisms or identify key regulators involving in the neural differentiation of pluripotent stem cells, green fluorescent protein (GFP)-tagged reporter systems under the control of NPC marker genes as well as new models such as haploid stem cells have been widely employed [34–39]. These systems and models represent valuable tools warranting further investigation.

METTL3 functions canonically within the METTL3/METTL14 MTC to deposit m⁶A and regulate various RNA metabolism processes [20]. The non-conventional functions of the MTC complex have also been reported to regulate nascent RNA synthesis, heterochromatin, and repression of retroviral elements through binding chromatin [40–42]. In our data, chromatin profiling in human NPCs confirms widespread genomic occurrence of METTL3 (Fig. 4). Notably, METTL3-bound genes critical for neurogenesis exhibited elevated expression and enhanced chromatin accessibility, strongly suggesting a direct role for chromatin-associated METTL3 in transcriptional regulation during human neural development. METTL3/14 may recruit chromatin remodelers or other epigenetic regulators to modulate chromatin accessibility and thereby promote the transcription of target genes. For instance, METTL3 physically interacts with the H3K9 tri-methyltransferase SETDB1 to direct H3K9me3 distribution [41], or recruits the H3K9me2 demethylase KDM3B to facilitate H3K9me2 demethylation [38]. In our previous study, we also found that METTL3/METTL14 mediate the degradation of the H3K9me3 methyltransferases SUV39H1/H2 [23]. Thus, future studies will focus on identifying METTL3-interacting proteins during human neurogenesis to elucidate the precise mechanisms through which the METTL3/14 complex regulates target gene expression. Recent studies reported that METTL14 regulates bivalent domains or H3K27me3 deposition in mouse ESCs independent of METTL3 or m⁶A modification [43, 44]. In addition, we demonstrated that the catalytic activity of METTL3 is dispensable for neural differentiation of hESCs based on rescue experiments with METTL3’s catalytic mutant (Supplementary Fig. S2A, B). However, another report suggests that the association of METTL3 with chromatin depends on its catalytic activity [41]. The precise relationship among the catalytic activity and chromatin binding of METTL3, its interacting proteins, and target gene expression during neurogenesis warrants further investigation. However, a recent report showed that METTL3 primarily regulates NPC proliferation but not generation or differentiation via m⁶A-dependent stabilization of SLIT2 mRNA in small-molecule-assisted shut-off (SMASh)-tagged hESCs because of reduced m⁶A in SLIT2 mRNA [45]. It is possible that the SMASh-tagged system results in low residual levels of METTL3 protein, which may limit the interpretation of METTL3’s role across the full neurogenesis process. In contrast, our study employed an inducible knockout system to investigate the roles of METTL3.

Furthermore, we deciphered the essential contribution of METTL14 similar to that of METTL3 in human neurogenesis, indicating that both METTL3 and METTL14 determine neural fate specifications from hESCs dependent on their MTC complex. Although the METTL3/METTL14 complex is known to maintain nucleolar integrity in hESCs [23], we showed that METTL3 deficiency does not affect the number of nucleoli and H3K9me3 levels (Supplementary Fig. S2C, D), suggesting that the role of METTL3 in neural differentiation is not associated with its effects on nucleoli and H3K9me3. Furthermore, BRM overexpression could not rescue the impaired self-renewal and nucleolar integrity of hESCs induced by METTL3 loss (Supplementary Fig. S2E, F). These data indicate that the roles of METTL3 and BRM in human ESC maintenance and neural differentiation depend on distinct mechanisms. Another study revealed that METTL3 loss impairs the hPSC proliferation and differentiation towards embryonic and extraembryonic lineages due to up-regulation of transposable elements and naïve genes [46]. Similarly, both that study and our previous work using a similar inducible knockout system observed reduced proliferation in METTL3-deficient hPSCs. However, that study emphasized dysregulation of transposable elements and the defective differentiation induced by METTL3 loss due to failure to exit pluripotency [46]. Our previous work revealed that METTL3/METTL14 regulate self-renewal of hESCs through maintaining nucleolar integrity [23], while in this study we showed the impaired neural differentiation induced by METTL3 loss due to dysregulation of chromatin remodelers and de-repression of non-neural lineage genes rather than the defect in the exit of pluripotency. These divergent mechanisms highlight context-dependent functions of METTL3 and warrant further investigation in future studies. Together, our findings highlight the essential requirement for both METTL3 and METTL14 in human neurogenesis.

Supplementary data

Supplementary data is available at NAR online.

Conflict of interest

The authors declare no competing interests.

Funding

The National Key Research and Development Program of China, Stem Cell and Translational Research [2022YFA1105001]; the Innovation Technology Commission of the Hong Kong SAR Health@InnoHK Program; the National Natural Science Foundation of China [32270624 and 31971374]; the Youth Innovation Promotion Association of the Chinese Academy of Sciences [2022360]; Guangzhou Key Research and Development Program [202206010041]; Guangdong Provincial Key Laboratory of Stem Cell and Regenerative Medicine [2020B1212060052]; Science and Technology Planning Project of Guangdong Province, China [2023B1212060050 and 2023B1212120009]. Funding to pay the Open Access publication charges for this article was provided by the National Key Research and Development Program of China, Stem Cell and Translational Research.

Data availability

The RNA-Seq, ATAC-seq, and CUT&TAG-seq data have been deposited in the Genome Sequence Archive for Human database under the accession codes HRA012368 and HRA013853 at https://ngdc.cncb.ac.cn/gsa-human. The authors declare that all data supporting the current study’s findings are available from the corresponding authors on reasonable request.