N6-methyladenosine RNA modification regulates embryonic neural stem cell self-renewal through histone modifications

Internal N6-methyladenosine (m6A) modification is widespread in messenger RNAs (mRNAs) and is catalyzed by heterodimers of methyltransferase-like protein 3 (Mettl3) and Mettl14. To understand the role of m6A in development, we deleted Mettl14 in embryonic neural stem cells (NSCs) in a mouse model. Phenotypically, NSCs lacking Mettl14 displayed markedly decreased proliferation and premature differentiation, suggesting that m6A modification enhances NSC self-renewal. Decreases in the NSC pool led to a decreased number of late-born neurons during cortical neurogenesis. Mechanistically, we discovered a genome-wide increase in specific histone modifications in Mettl14 knockout versus control NSCs. These changes correlated with altered gene expression and observed cellular phenotypes, suggesting functional significance of altered histone modifications in knockout cells. Finally, we found that m6A regulates histone modification in part by destabilizing transcripts that encode histone-modifying enzymes. Our results suggest an essential role for m6A in development and reveal m6A-regulated histone modifications as a previously unknown mechanism of gene regulation in mammalian cells. Using a genetic approach, Wang et al. demonstrate an essential function for m6A mRNA modification in promoting neural stem cell proliferation and reveal interactions between m6A and histone modification as a novel gene regulatory mechanism.


Results
Mettl14 knockout decreases NSC proliferation and promotes premature NSC differentiation in vitro. To assess Mettl14 loss of function in vivo, we generated Mettl14-conditional knockout mice (Mettl14 f/f ) by flanking Mettl14 exon 2 with loxP sites. Cre-mediated exon 2 excision results in an out-of-frame mutation that abolishes Mettl14 function ( Supplementary Fig. 1a,b). To assess whether the KO strategy deletes Mettl14 in vivo, we evaluated whether Mettl14 was deleted globally using EIIa-cre transgenic mice, which express Cre at zygotic stages ( Supplementary Fig. 1c,d). Mettl14 +/− heterozygotes were viable and fertile and exhibited no discernible morphological or growth abnormalities, whereas no Mettl14 −/− offspring were observed after crosses of Mettl14 +/− mice (Supplementary Table 1). We then collected embryos resulting from crosses of heterozygotes at embryonic day 7.5 (E7.5), E8.5 and E9.5 for genotyping. Mettl14 −/− embryos were identified at Mendelian ratios when we N 6 -methyladenosine RNA modification regulates embryonic neural stem cell self-renewal through histone modifications combined genotyping results from all three stages (Supplementary  Table 2). But most Mettl14 −/− embryos were dead and many had regressed ( Supplementary Fig. 1e), indicating that Mettl14 activity is required for early embryogenesis, a phenotype similar to that of global Mettl3-KO mice 13 . Of seven Mettl14 −/− embryos identified at either E7.5 or E8.5, four were male and three were female, suggesting that phenotypes were not gender specific ( Supplementary Fig. 1f).
We then assessed the potential effects of Mettl14 deletion in NSCs. To do so, we crossed Mettl14 f/f mice with a Nestin-Cre transgenic line to generate Mettl14 f/f ;Nestin-Cre (Mettl14-cKO) mice and littermate controls, including Mettl14 f/+ ;Nestin-cre (heterozygous) and Mettl14 f/f (non-deleted) mice. Newborn pups were alive and showed no overt morphologic phenotypes ( Supplementary Fig. 1g) and normal body weight ( Supplementary Fig. 1h). However,

Nature NeuroscieNce
all Mettl14-cKO mice were dead within the first neonatal week ( Supplementary Fig. 1i). When we examined the brains of postnatal day 0 (P0) Mettl14-cKO pups, we observed no anomalies in gross anatomy, but we found moderately reduced cortical length (Fig. 1a,b). Hematoxylin and eosin (H&E) staining of coronal sections of P0 mouse brain revealed enlargement of the ventricle and a 23% decrease in cortical thickness in Mettl14-cKO brains relative to littermate Mettl14 f/f controls (Fig. 1c,d). We next examined Mettl14 expression in RGCs by carrying out Mettl14 and Pax6 co-immunostaining on coronal sections of E17.5 brain from nondeleted, cKO and heterozygous mice. Mettl14 was readily detectible in Pax6 + cells in the cortex of nondeleted and heterozygous controls, but not in cKO mice (Fig. 1e). Together, these results suggest that Mettl14 is required for normal function of NSCs that serve as cortical progenitors. Next, we evaluated m 6 A function in isolated embryonic NSCs cultured in vitro. To determine which embryonic stages are appropriate    25 Graphs represent the mean ± s.d. Dots represent data from individual data points. ns, non-significant. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001.

Nature NeuroscieNce
to select Mettl14-deficient NSCs, we examined Mettl14 protein expression in coronal sections prepared from E13.5, E15.5, E17.5 and P0 brains from cKO, heterozygous and nondeleted control mice. Immunostaining revealed residual Mettl14 staining in the cerebral cortex at E13.5 in Mettl14-cKO brain, whereas Mettl14 signals in cortex were absent from E15.5 onward (Fig. 1f). Heterozygous mice showed Mettl14 signals comparable to those of nondeleted controls. Thus, for further analysis we chose E14.5 and E17.5 cortical NSCs and cultured them as neurospheres for 7 d before harvesting them for analysis. We observed comparable phenotypes in subsequent in vitro analysis of E14.5 and E17.5 NSCs. Unless stated otherwise, our results are those of experiments conducted in E14.5 NSCs. Following confirmation of Mettl14 loss in KO NSCs by western blotting (Supplementary Fig. 2a), we assessed m 6 A levels from E14.5 neurospheres. Thin-layer chromatography (TLC) analysis revealed an almost total loss of m 6 A in polyA RNA isolated from Mettl14-KO versus nondeleted NSCs, whereas heterozygous cells displayed m 6 A levels comparable to those seen in nondeleted controls (Fig. 2a), suggesting that the KO system that we generated is ideal for studying m 6 A function in NSCs.
To characterize KO versus control NSCs, we used a Celigo image cytometer and software to image neurospheres and assess their number and size. Although Mettl14-KO, heterozygous and nondeleted control NSCs derived from E14.5 embryos formed a similar number of neurospheres, neurosphere size, as reflected by neurosphere area in this system, decreased by ~55% in KO versus nondeleted control cells, whereas neurosphere size from heterozygous cells was comparable to that seen in nondeleted controls (Fig. 2b,c). Consistently, those same Mettl14-KO NSCs exhibited significantly decreased proliferation, as determined by cell-counting analysis (Fig. 2d). Similar proliferation defects were detected in NSCs taken from E17.5 Mettl14-cKO mice ( Supplementary Fig. 2b). Annexin V flow cytometry ( Supplementary Fig. 2c,d) and TUNEL (terminal deoxynucleotidyl transferase (TdT) dUTP nick-end labeling) analysis ( Supplementary Fig. 2e) of E14.5 NSCs confirmed that the effects were not a result of increased apoptosis. To ensure that proliferation defects were a result of Mettl14 deletion, we performed  Fig. 2f) and the restoration of m 6 A ( Supplementary Fig. 2g). Notably, Mettl14 overexpression did not increase the proliferation of nondeleted NSC controls (Fig. 2e), but increased the proliferation of Mettl14-KO NSCs to rates comparable to that of controls (Fig. 2e). These results suggest that Mettl14 and concomitant m 6 A RNA modification regulate NSC proliferation, at least in vitro.
It is well established that decreased NSC proliferation is coupled with premature NSC differentiation 14 . Thus, we checked for the presence of cells expressing the neuronal marker Tuj1 in E14.5 NSCs cultured for 7 d as neurospheres. Immunostaining analysis revealed a 6.2-fold increase in the number of Tuj1 + cells in KO versus control NSCs (Fig. 2f,g), whereas the number of Tuj1 + cells was comparable between heterozygous and nondeleted controls, suggesting that Mettl14 loss leads to premature neuronal differentiation. Together, these results suggest that Mettl14 regulates NSC self-renewal.
To determine whether the loss of an m 6 A demethylase would have the opposite effect on NSC proliferation, we knocked down two reported m 6 A demethylases, Fto and Alkbh5 9,15 , separately in wildtype NSCs. Reverse-transcription quantitative PCR (RT-qPCR) analysis showed high knockdown efficiency in each case, and western blots revealed a marked decrease in Fto and Alkbh5 protein levels in the respective knockdown cells (Supplementary Fig. 2h,i,k,l). Nonetheless, NSC proliferation in vitro was not altered by the loss of either protein ( Supplementary Fig. 2j,m). Some reports suggest that changes in expression of either Fto or Alkbh5 have only moderate effects on m 6 A levels and that both factors likely regulate m 6 A modification of a subset of transcripts [15][16][17] . Our results suggest that mRNAs that function in NSC proliferation are not regulated by either Fto or Alkbh5.
The size of the cortical RGC pool is reduced in Mettl14-cKO mouse brain. We next examined the effect of Mettl14 on the proliferation of primary cortical stem cells or RGCs in vivo. To do so, we determined the number of S-phase cells in the cortex of E13.5, E15.5 or E17.5 Mettl14-cKO, heterozygous and nondeleted control embryos by injecting pregnant females with bromodeoxyuridine (BrdU) and harvesting embryos 0.5 h later, which detected only cells undergoing DNA replication at that time point. Immunostaining showed that the number of BrdU + cells decreased by 19% in E15.5 Mettl14-cKO compared with nondeleted control brain ( Supplementary Fig. 3a,d), and that number was 40% when analysis was conducted at E17.5 (Fig. 3a,d). Similarly, we also observed a 44% and 45% decrease in the number of cells expressing the mitotic marker phospho-histone H3 (PH3) at the apical membrane of the cortical VZ in cKO versus nondeleted control brain at E15.5 and E17.5, respectively ( Supplementary Fig. 3b,d and Fig. 3b,d). To determine the number of RGCs, we assessed brain coronal sections at all three stages with the RGC marker Pax6. Consistently, we detected a 45% decrease in the number of Pax6 + RGCs in the VZ of E17.5 Mettl14-cKO brain versus controls (Fig. 3c,d) and a 14% decrease at E15.5 ( Supplementary Fig. 3c,d). All experiments showed highly comparable results between heterozygous and nondeleted control RGCs (Fig. 3a-d and Supplementary Fig. 3a-d). We did not detect differences in BrdU, PH3 or Pax6 staining relative to that in nondeleted controls in the cortex of E13.5 Mettl14-cKO brains ( Supplementary  Fig. 3e-g), consistent with the finding that residual Mettl14 is present in cortex at E13.5 (Fig. 1f). Immunostaining with the apoptosis marker cleaved caspase-3 revealed no change in the number of apoptotic cells in the cortex of E17.5 and E15.5 Mettl14-cKO brains relative to nondeleted controls ( Supplementary Fig. 3h,i). To understand how Mettl14 loss might affect RGC proliferation, we examined cell cycle progression and cell cycle exit of RGCs from the brains of E15.5 and E17.5 Mettl14-cKO versus control mice. We carried out sequential 5-iodo-2ʹ-deoxyuridine (IdU) and BrdU injection to evaluate cell cycle progression, followed by IdU and BrdU double-staining of cortical sections 18 . We then determined the percentage of IdU + BrdUcells, which represented the cells that had progressed past S-phase, versus all IdU + cells, a group that included both proliferating cells and cells that had left S-phase. We detected a 38% and 43% decrease in E15.5 and E17.5 Mettl14-cKO embryos, respectively, compared with the nondeleted control, suggesting that Mettl14 loss disrupts normal RGC cell cycle progression (Fig. 3e-g). Heterozygous and nondeleted control RGCs yielded comparable results (Fig. 3e-g). To determine whether Mettl14 loss alters cell cycle exit, we performed BrdU-Ki67 double-staining of cortical sections from the brains of mice pulsed with BrdU and analyzed 24 h later. Mettl14 loss resulted in a 50% and 39% decrease in cells exiting the cell cycle from E15.5 and E17.5 Mettl14-cKO embryos, respectively, versus nondeleted controls, suggesting that Mettl14 is required for normal RGC cell cycle exit (Fig. 3h-j). Heterozygous and nondeleted control RGCs yielded comparable results (Fig. 3h-j). Together, these data strongly suggest that Mettl14 regulates the RGC cell cycle and that the RGC pool in cortex is substantially reduced in Mettl14-cKO mice.
We then examined RGC premature differentiation in the cortical VZ of E15.5 and E17.5 Mettl14-cKO brains using Eomes (Tbr2), a marker of intermediate progenitor cells located at the sub-ventricular zone (SVZ), and the proneural marker Neurod2 (ND2). Notably, patches of ND2 + Tbr2 + cells were seen consistently at E15.5 and E17.5 in areas close to the apical surface of the cortical VZ of Mettl14-cKO brain, but were absent in comparably staged littermate controls ( Supplementary Fig. 3j,k and Fig. 3k,l). Together, these data suggest that Mettl14 is required to prevent NSC premature differentiation and maintain the NSC pool in vivo.

Mettl14 deletion results in reduced numbers of late-born neurons.
We next evaluated the effects of Mettl14 loss on cortical neurogenesis. In P0 mice, neurons differentiated from RGCs are found in six distinct cortical layers containing neuronal subtypes identifiable by specific markers. Thus, we stained coronal sections from cKO and comparably staged littermate controls at P0 for the following markers: Cux1, which is expressed in late-born neurons and is a marker of upper neuronal layers II-IV; Sox5, which is expressed in early-born neurons and is a marker of layer V; and Tbr1, which is expressed in postmitotic neurons and is a marker of layer VI to the subplate. Overall, layer organization was comparable in cKO and control mice. When we assessed layer thickness, the thickness of layers marked by Sox5 and Tbr1 did not differ significantly between genotypes (Fig. 4a,b). However, we observed a 70% decrease in the thickness of Cux1 + layers (II-IV) (Fig. 4a,b). To confirm the loss of neurons from these layers, we stained sections from P0 embryos for a different layer II-IV marker, Satb2, and observed an ~34% decrease in the number of Satb2 + e neurons in cKO mice versus littermate controls (Fig. 4c,d). When we examined cortical Cux1 staining at E17.5, we detected a 22% reduction in the thickness of Cux1 + layers and a 50% reduction in the number of newly generated Cux1 + cells residing in a region between the VZ and layer IV in Mettl14-cKO mice versus controls (Fig. 4e,f). These results suggest that Mettl14 loss may deplete the progenitor pool in a way that is reflected by loss of late-born neurons.
Mettl14 knockout leads to genome-wide changes in histone modification that perturb gene expression. To assess molecular mechanisms underlying m 6 A-regulated NSC activity, we cultured NSCs from E14.5 Mettl14-cKO, heterozygous and nondeleted control embryos for 7 d and performed RNA sequencing (RNA-seq). Mettl14-KO NSCs exhibited distinct gene expression profiles relative to nondeleted and heterozygous controls (which showed comparable profiles; Fig. 5a and Supplementary Table 3).
Gene Ontology analysis (GO) suggested that the most significantly upregulated genes function in NSC differentiation, whereas downregulated genes are associated with cell proliferation (Fig. 5b,c and Supplementary Fig. 4a), changes that are reflective of observed phenotypes. We then evaluated potential mechanisms underlying these changes in gene expression. It is well established that m 6 A destabilizes transcripts 4,13,19,20 . However, we detected only a weak correlation between m 6 A loss and increase in transcript abundance ( Supplementary Fig. 4b, Supplementary Tables 3 and 4), suggesting that different m 6 A-related mechanisms modulate mRNA levels. Given that modification of histone tails is a critical mechanism of gene regulation in mammalian cells 21 , we asked whether m 6 A RNA modification may also change histone modifications by performing western blotting on acid-extracted histones from KO versus control NSCs isolated at both E14.5 and E17.5. We evaluated a panel of wellstudied histone modifications that have been reported to regulate stem cell activities, including histone H3 phosphorylation, histone H2A and H2B ubiquitination, three types of histone acetylation, and four types of histone methylation [22][23][24][25][26][27][28][29][30][31][32][33] . These histone marks are associated with either gene activation or repression. Representative western blots of E14.5 NSCs are shown in Fig. 5d. We quantified western blots from E14.5 and E17.5 by calculating the ratios of respective histone modifications to total H3 protein in KO, heterozygous and nondeleted control NSCs. Although we observed no significant change in any of the histone modifications that we tested between heterozygous and nondeleted control samples (Fig. 5d,e), we detected a significant increase in acetylation of histone H3 at lysine 27 (H3K27ac; 111% increase), trimethylation of histone H3 at lysine 4 (H3K4me3; 43% increase) and trimethylation of histone H3 at lysine 27 (H3K27me3; 71% increase) in Mettl14-KO NSCs versus nondeleted controls (Fig. 5d,e). These results suggest that m 6 A regulates specific histone modifications.
To determine whether these changes alter NSC proliferation, we searched for chemical inhibitors that antagonize activities associated with upregulated histone modifications to determine whether inhibitor treatment of E14.5 KO NSCs would rescue cell proliferation defects. Three inhibitors were commercially available: MM102, which inhibits mixed-lineage leukemia (MLL) function and H3K4me3 formation; C646, which inhibits the H3K27 acetyltransferase Crebbp (CBP)/p300 activity; and GSK343, which inhibits Ezh2-dependent H3K27me3 formation. We then seeded comparable numbers of NSCs of all three genotypes, added inhibitor or DMSO vehicle at day 0, and determined cell number via MTT assays 4 d later. After DMSO treatment, the number of KO NSCs was ~50% that of nondeleted controls, reflecting slower proliferation, as anticipated (Fig. 5f). GSK343 treatment at 1.25, 2.5 and 5 μ M increased the ratios of KO to nondeleted control NSCs to 64%, 71% and 80%, respectively (Fig. 5f), whereas the percentages of heterozygous to nondeleted control NSCs were unchanged by GSK343 treatment (Supplementary Fig. 4c). These observations suggest that blocking the formation of H3K27me3 rescues the growth defects of KO NSCs. Increased ratios of KO versus nondeleted control NSCs were also seen after C646 treatment ( Fig. 5f and Supplementary  Fig. 4c), suggesting that blocking of H3K27ac also blocks the proliferation defects of KO NSCs. By contrast, treatment of E14.5 NSCs with MM102 had no effect ( Fig. 5f and Supplementary Fig. 4c). These results suggest that m 6 A regulates NSC proliferation, at least in part, through H3K27me3 and H3K27ac modifications.
H3K27me3 marks gene promoters and is associated with silencing 34,35 , whereas H3K27ac, which is enriched at promoters and enhancers, is associated with gene activation 36,37 . Thus, we asked whether increased promoter H3K27me3 was associated with gene downregulation, whereas increased promoter/ enhancer H3K27ac was associated with gene upregulation in E14.5 Mettl14-KO versus control NSCs. To do so, we performed H3K27me3 and H3K27ac ChIP-seq analysis on E14.5 KO versus nondeleted NSCs (Supplementary Tables 5 and 6) and correlated changes in gene expression with altered histone modification. In total, the intensity of 1,610 promoter/enhancer H3K27ac peaks, defined as peaks within a 10-kb region up-or downstream of a transcriptional start site (TSS), were significantly altered in KO versus control cells. Pearson correlation analysis showed a positive correlation between changes in peak intensity and changes in gene expression (r = 0.06195, P = 0.01292) in KO versus control NSCs, suggesting that H3K27ac functions in m 6 A-regulated gene activation. We also detected 434 altered H3K27me3 promoter peaks, defined as peaks within 2 kb upstream of a TSS, in KO versus control NSCs. Although in this case we did not detect a significant correlation between changes in peak intensity and gene expression (P = 0.05784) using all 434 genes, we detected a strongly negative Pearson correlation (r = -0.38804, P < 0.02) when we analyzed only downregulated genes (log 2 fold change ≤ -0.6) in KO versus control NSCs, suggesting that H3K27me3 levels are positively correlated to the repression of genes showing decreased expression.
To further assess the relevance of altered H2K27ac and H3K27me3 modifications to NSC gene expression, we asked whether altered transcript abundance seen in KO versus control NSCs could be rescued by treating cells with inhibitors of H2K27ac (C646) or of H3K27me3 (GSK343). Using overlaying ChIP-seq and RNA-seq data and coupling that to Ingenuity pathway analysis (IPA), we picked five differentiation-related genes showing increased H3K27ac and increased expression and five cell-proliferation related genes showing increased H3K27me3 but decreased expression for rescue experiment. Indeed, C646 treatment resulted in significantly decreased expression of the neuritogenesis regulators Kif26a 38 , Gas7 39 and Pdgfrb 40 , in KO NSCs when compared with that in nondeleted NSCs (Fig. 6a-c), whereas GSK343 treatment increased expression of the transcription factors Egr2 and Egr3, which are known to promote proliferation 41 (Fig. 6d-f). These results suggest that m 6 A-regulated histone modification functions in NSC gene expression. m 6 A regulates the stability of CBP and p300 transcripts. We then asked how m 6 A might regulate histone modifications. To do so, we first evaluated the presence of m 6 A on transcripts encoding the H3K27 acetyltransferases CBP and p300 and the polycomb repressive complex (PRC2) subunits Ezh2, Suz12 and Eed, which catalyze H3K27me3, by methylated RNA immunoprecipitation (meRIP). We detected a 20-30% enrichment of m 6 A over input in CBP (Crebbp) and p300 (Ep300) mRNAs, which was lost in E14.5 Mettl14-KO NSCs (Fig. 7a). In contrast, only a 0.4-0.6% enrichment of m 6 A was observed in Ezh2, Eed and Suz12 mRNAs, and the extremely low levels that we observed for Ezh2 and Eed persisted in KO cells ( Supplementary Fig. 5), suggesting that the signals that we detected are a result of the immunoprecipitation background.
We then evaluated potential changes in the stability of CBP and p300 mRNAs. We observed a significant increase in both CBP and p300 mRNA levels in E14.5 Mettl14-KO versus control NSCs (Fig. 7b). We then assayed mRNA stability by treating E14.5 cultured KO and control NSCs with actinomycin D (ActD) to block transcription and harvesting cells 3 and 6 h later. Both CBP and p300 showed significantly increased mRNA stability in Mettl14-KO NSCs compared with nondeleted control NSCs (Fig. 7c), suggesting that m 6 A may regulate histone modification by destabilizing transcripts that encode histone modifiers.

Discussion
By conditionally inactivating Mettl14 in embryonic NSCs, we discovered that Mettl14 is required for NSC proliferation and maintains NSCs in an undifferentiated state (Figs. 1-3). Thus, our findings reveal a previously unknown, but essential, function of m 6 A RNA methylation in the regulation of NSC self-renewal. We also observed decreased numbers of late-born neurons, which are generated from RGCs after E15.5, in the cortex of Mettl14-cKO animals at E17.5 and P0 (Fig. 4), consistent with the loss of Mettl14 expression and a decrease in size of the RGC pool. Notably, although Mettl14 loss promoted premature NSC differentiation, the identity of neuronal subtypes in each neuronal layer was not obviously affected in cortex of Mettl14-cKO brain. Thus, we conclude that RGCs lacking Mettl14 remain capable of differentiation and migration, and propose that cortical defects seen in Mettl14-KO mice are primarily a result of perturbed NSC self-renewal. Overall, our results provide a benchmark to further explore mechanisms underlying perturbed m 6 A RNA methylation in neurodevelopmental disorders. We also detected Mettl14 expression in postmitotic cortical neurons (Fig. 1e); thus, the possibility of a Mettl14 function in these cells cannot be excluded. However, given that the Mettl14-cKO mice in our study died shortly after birth, examination of potential Mettl14 function in neurons at later postnatal or adult stages was not possible. We also note that, although we found a specific reduction in the number of upper layer projection neurons in the cortex of Mettl14-cKO versus control mice, we do not exclude the possibility that Mettl14 regulates production of neurons in other cortical layers. Analysis of that effect is likely to require the use of a Cre driver to delete Mettl14 at an earlier time point than E13.5.
Our data suggest, for the first time, to our knowledge, the existence of cross-talk between RNA and histone modification. Although m 6 A reportedly regulates gene expression through diverse mechanisms such as mRNA stability 4,13,19,20 , splicing 15,42-44 and translation 45,46 , an interaction between m 6 A modification and epigenetic mechanisms has not been explored. Thus, our finding that m 6 A RNA methylation regulates specific histone modifications, including H3K27me3, H3K27ac and H3K4me3, represents a previously unknown mechanism of gene regulation. Among these modifications, H3K27ac and H3K4me3 are associated with gene activation, and H3K27me3e with repression, consistent with our observations that there is no marked bias toward gene activation or repression in Mettl14-KO NSCs compared with controls (Fig. 5a). Importantly, our data show that, in Mettl14-KO cells, different m 6 A-regulated histone marks regulate expression of genes of different function (Fig. 6), suggesting that m 6 A-regulated active and repressive histone modifications work synergistically to ensure an NSC ground state. Synergy is also evident in our observation that treatment of cells with either H3K27me3 or H3K27ac inhibitors can, at least partially, restore cell proliferation in Mettl14-KO NSCs (Fig. 5). We did not observe any effects of the H3K4me3 inhibitor MM-102 on NSC proliferation. However, unlike H3K27ac and H3K27me3, which are regulated by distinct enzymes, H3K4me3e modification is catalyzed by a panel of SET domain proteins 47 . Although MM-102 is widely used to test H3K4me3 function, it inhibits the formation of a MLL1-WDR5 complex and therefore targets only the subset of H3K4me3 modifications that are dependent on MLL1 48 . As yet, there are no inhibitors available that inhibit activities of most H3K4me3 methyltransferases and therefore sufficiently abolish H3K4me3. Thus, we cannot exclude the possibility that m 6 A-regulated H3K4me3 regulates NSC proliferation. Future studies may reveal that m 6 A regulates levels of histone modifications other than H3K27ac, H3K27me3 and H3K4me3.
We further showed that m 6 A regulates histone modifications directly by destabilizing transcripts that encode histone modifiers. It is evident that such a mechanism is applicable to the H3K27 acetyltransferase CBP/p300, but not to subunits in the PRC2 complex, suggesting that m 6 A regulates histone modifications through distinct mechanisms. Identification of those mechanisms warrants future investigation.
In summary, we propose a model (Fig. 7d) in which m 6 A loss alters histone modifications partly by regulating mRNA stability of histone modifiers, and altered histone modifications aberrantly repress proliferation-related genes and activate differentiationrelated genes, resulting in a loss of NSC ground state. Our results provide an in-depth analysis of m 6 A in brain neural stem cells and suggest that there is an interaction between m 6 A and histone modification as a mechanism of gene regulation.
Mettl14 ES targeting vector was linearized by NotI and transfected into G4 male ES cells as described 52,53 . 24 h after transfection, 150 μ g/ml Hygromycin was added to the ES medium. Hygromycin-resistant ES colonies were picked 10 d after transfection. Targeting of Mettl14 was confirmed by genomic PCR analysis using primer pairs: M14cF and SA-R for 5′ -end; UptpA-F and M14-cR for 3′ -end.
Generation of Mettl14 conventional and conditional knockout mice. Positive ES clones were used for injection into c57 blastocysts and generation of chimerical mice.
To produce Mettl14 f/+ mice, the chimeras were crossed with wild-type c57 for germ line transmission and then crossed with Atcb-Flpe transgenic mice (The Jackson Laboratory, # 003800) to remove FRT flanked selection cassette. Male Mettl14 f/+ mice were crossed to female EIIa-Cre transgenic mice (The Jackson Laboratory, # 003724) to obtain Mettl14 +/− mice, and Mettl14 +/− mice were intercrossed to obtain Mettl14-conventional knockout mice. Sex of embryos was determined.
To conditionally knock out Mettl14 in brain, floxed Mettl14 mice were bred with Nestin-Cre transgenic mice (The Jackson Laboratory, # 003771) to generate Mettl14 f/f ;Nestin-cre. Sex of embryos was not determined. Mice were maintained at the Sanford Burnham Prebys Medical Discovery Institute Institutional animal facility, and experiments were performed in accordance with experimental protocols approved by local Institutional Animal Care and Use Committees.
Genotyping. Genotyping was performed using the MyTaq Extract-PCR kit (Bioline) with primer sets corresponding to the primer list table (Supplementary Table 7).
Injection of S-phase tracer. IdU/BrdU double-labeling was performed as previously described 18 . Briefly, pregnant females were injected intraperitoneally with Iododeoxyuridine (IdU, Sigma) (100 mg/kg body weight) and 1.5 h later with the same dose of BrdU (Sigma) and killed after 0.5 h.
For BrdU single-labeling, pregnant females were injected intraperitoneally with BrdU (100 mg/kg body weight) and killed after 0.5 or 24 h.
For BrdU staining, sections were treated with 2 N HCl at 37 °C for 30 min and 0.1 M borate buffer, pH 8.5, for 10 min at room temperature.
Pax6-Tbr2 and Pax6-Mettl14 co-staining were performed according to the protocol described in a previous paper 54 . In brief, sections were incubated with highly diluted (1:30,000) primary antibody overnight at 4 °C and biotinylated goat anti-rabbit secondary antibody for 1 h at room temperature. The signal was amplified with the horseradish-peroxidase-based Vectastain ABC Kit (Vector Laboratories, Cat. # PK-6101) and Cyanine 3 Tyramide System (Perkin Elmer, Cat. # NEL704A001KT).
Fluorescence images were acquired by Zeiss LSM 710 confocal microscope and analyzed in ImageJ software.
NPC isolation and culture. Cortex region was dissected out from embryonic brains and triturated by pipetting. Dissociated cells were cultured as neurospheres with NeuroCult Proleferation Medium (Stemcell Tech.) according to the manufacturer's protocols. Lentiviral constructs harboring shRNAs against Alkbh5 or Fto were purchased from Sigma-Aldrich (see below for details). Stable knockdown lines were generated using standard viral infection and puromycin selection (2 μ g/ml).

Replication
Describe whether the experimental findings were reliably reproduced.
Phenotypes observed are robust and were reliably reproduced. All experiments on mice embryos and newborns were repeated using at least 3 independent litters that consist all reported genotypes.

Randomization
Describe how samples/organisms/participants were allocated into experimental groups.
Mice were allocated to different groups based on the genotypes. No randomization was used in this study.

Blinding
Describe whether the investigators were blinded to group allocation during data collection and/or analysis.
Mice were allocated to different groups based on the genotypes. No blinding was used in this study.
Note: all studies involving animals and/or human research participants must disclose whether blinding and randomization were used.

Statistical parameters
For all figures and tables that use statistical methods, confirm that the following items are present in relevant figure legends (or in the Methods section if additional space is needed).

n/a Confirmed
The exact sample size (n) for each experimental group/condition, given as a discrete number and unit of measurement (animals, litters, cultures, etc.) A description of how samples were collected, noting whether measurements were taken from distinct samples or whether the same sample was measured repeatedly A statement indicating how many times each experiment was replicated The statistical test(s) used and whether they are one-or two-sided (note: only common tests should be described solely by name; more complex techniques should be described in the Methods section) A description of any assumptions or corrections, such as an adjustment for multiple comparisons The test results (e.g. P values) given as exact values whenever possible and with confidence intervals noted A clear description of statistics including central tendency (e.g. median, mean) and variation (e.g. standard deviation, interquartile range)

Clearly defined error bars
See the web collection on statistics for biologists for further resources and guidance. 10. Eukaryotic cell lines a. State the source of each eukaryotic cell line used. The only cell lines used in the study are primary cultured E14.5/E17.5/P0 mouse NSCs.
b. Describe the method of cell line authentication used. NSC genotypes were determined by genomic PCR.
c. Report whether the cell lines were tested for mycoplasma contamination.
All NSC lines were tested negative for mycoplasma contamination.
d. If any of the cell lines used are listed in the database of commonly misidentified cell lines maintained by ICLAC, provide a scientific rationale for their use.
No commonly misidentified cell lines were used in this study.

Animals and human research participants
Policy information about studies involving animals; when reporting animal research, follow the ARRIVE guidelines 11. Description of research animals Provide details on animals and/or animal-derived materials used in the study.
Policy information about studies involving human research participants

Description of human research participants
Describe the covariate-relevant population characteristics of the human research participants.
This study did not involve human research participants.
the ChIP-seq data. the data. Detailed methods are described in the manuscript. Scripts are available upon request.