Ythdf2-mediated m6A mRNA clearance modulates neural development in mice

Background N6-methyladenosine (m6A) modification in mRNAs was recently shown to be dynamically regulated, indicating a pivotal role in multiple developmental processes. Most recently, it was shown that the Mettl3-Mettl14 writer complex of this mark is required for the temporal control of cortical neurogenesis. The m6A reader protein Ythdf2 promotes mRNA degradation by recognizing m6A and recruiting the mRNA decay machinery. Results We show that the conditional depletion of the m6A reader protein Ythdf2 in mice causes lethality at late embryonic developmental stages, with embryos characterized by compromised neural development. We demonstrate that neural stem/progenitor cell (NSPC) self-renewal and spatiotemporal generation of neurons and other cell types are severely impacted by the loss of Ythdf2 in embryonic neocortex. Combining in vivo and in vitro assays, we show that the proliferation and differentiation capabilities of NSPCs decrease significantly in Ythdf2−/− embryos. The Ythdf2−/− neurons are unable to produce normally functioning neurites, leading to failure in recovery upon reactive oxygen species stimulation. Consistently, expression of genes enriched in neural development pathways is significantly disturbed. Detailed analysis of the m6A-methylomes of Ythdf2−/− NSPCs identifies that the JAK-STAT cascade inhibitory genes contribute to neuroprotection and neurite outgrowths show increased expression and m6A enrichment. In agreement with the function of Ythdf2, delayed degradation of neuron differentiation-related m6A-containing mRNAs is seen in Ythdf2−/− NSPCs. Conclusions We show that the m6A reader protein Ythdf2 modulates neural development by promoting m6A-dependent degradation of neural development-related mRNA targets. Electronic supplementary material The online version of this article (10.1186/s13059-018-1436-y) contains supplementary material, which is available to authorized users.


Background
Over the past decade, more than 100 post-transcriptionally modified ribonucleotides have been identified in various types of RNA [1]. Much more recently, epitranscriptomic [2] regulation at the RNA level via reversible RNA methylation has been revealed, beginning from 2011 with the discovery of the reversible potential of N 6 -methyl-adenosine (m 6 A) in mRNA [3]. As a post-transcriptional epitranscriptomic modification, m 6 A is one of the most abundant modifications in mRNA in eukaryotes [4]. It can be written by the methyltransferase complex (Mettl3, Mettl14, Wtap, and Kiaa1429) [5], erased by demethylases (Fto and Alkbh5) [3,6], and read by the binding proteins (Ythdf1-3, Ythdc1-2, and Hnrnp family proteins) [7][8][9][10].
The reversible/dynamic nature of m 6 A in mRNA and the ability to map this modification transcriptome-wide have led to a tremendous increase in the interest and understanding of the multiple biological roles of the dynamic m 6 A modification [10,11]. One of the evolutionarily conserved roles of the m 6 A modification is the regulation of meiosis and fertility. This was shown early for the writers of m 6 A in model organisms [12] and also for the mammalian m 6 A eraser Alkbh5 [6] and the m 6 A reader protein Ythdf2 [13]. The depletion of the m 6 A eraser Fto in mammalian cells causes defects in energy homeostasis and adipocyte differentiation [14]. It is worth mentioning that a loss-of-function mutation in the Fto gene causes growth retardation and multiple malformations in humans [15]. The writer Mettl3 is crucial for maintaining mouse stem cell pluripotency, regulating the reprogramming of somatic cells and the circadian rhythm, and targeting of the gene in mouse causes early embryonic lethality [16][17][18][19][20]. The most recent studies in hematopoietic stem/progenitor cells have uncovered the crucial role of Mettl3 in determining cell fates during vertebrate embryogenesis [21,22]. The m 6 A reader proteins Ythdf1-3 share a set of common mRNA targets and spatiotemporal interplay with each other cooperatively control translation and decay of these common targets in the cytosol [23]. The m 6 A readers Ythdc1 and Hnrnpa2b1 regulate splicing and processing of their mRNA targets [8,24], while Ythdc2 affects translation efficiency as well as stability of target mRNAs [9].
Recently, mutant models of the mammalian m 6 A readers reveal interesting phenotypes, which again include spermatogenesis [9] and oocyte competence [25]. Moreover, Ythdf2-dependent, m 6 A-modified mRNA clearance was shown to impact the highly regulated maternal-to-zygotic transition (MZT) in zebrafish [7,13]. In Drosophila, m 6 A writer (Ime4, dMettl14) and reader (Yt521-b) mutants exhibit flight defects and poor locomotion due to impaired neuronal functions [26]. Most recently, the m 6 A writer Mettl14 was shown to be required for the temporal control of mammalian cortical neurogenesis [27]. These findings strongly suggest the potential role of m 6 A modification during nervous system development, which might be conserved across species. Many histone and DNA encoded epigenetic mechanisms are uncovered to be conserved in this process. Thus, addressing the role of m 6 A methylation in mRNA will be an exciting new field to explore and will shed new light on neural development.
The m 6 A reader Ythdf2 is essential for oocyte competence and mutation of it causes female infertility [25]. Here we describe the early brain developmental failure of mice lacking Ythdf2 due to failure to regulate neural stem/progenitor cell (NSPC) proliferation and differentiation. During embryonic development, apical progenitor cells in the ventricular zone (VZ) serve as primitive neural stem cells that give rise to both the neuronal and glial lineages directly or produce secondary progenitors, termed the basal progenitor, in the subventricular zone (SVZ) in a precisely regulated spatiotemporal order [28]. In this study, Ythdf2 knockout embryos displayed delayed cortical neurogenesis. In vivo and in vitro experiments proved that Ythdf2-deficient NSPCs display decreased proliferation rates. Furthermore, Ythdf2-deficient NSPCs could naturally differentiate to neurons but not glial cells in vitro. However, the properties of differentiated neurons were influenced, seen as less neurite outgrowth and shorter neurites. Removal of Ythdf2 increased the sensitivity of neurons to reactive oxygen species (ROS) stress and decreased their recovery capability. RNA-seq combined with m 6 A-seq uncovered that the m 6 Amodified mRNAs involved in negative regulation of neural development were up-regulated in Ythdf2-deficient NSPCs, in agreement with the function of Ythdf2. The m 6 A-modified mRNA targets, recognized by the Ythdf2 protein in the wild type, were characterized by delayed degradation in Ythdf2 knockout embryos. Taken together, our findings reveal the critical functions of m 6 A modification and its binding protein Ythdf2 in neural development.

Results and discussion
Ythdf2 −/− targeted mice are embryonic lethal In order to study the biological function of the m 6 A reader Ythdf2, we generated conditional C57BL/6 Ythdf2 targeted mice with LoxP sites flanking the 5′ UTR and exon 1 of the endogenous Ythdf2 locus using CRISPR-Cas9 technology (Fig. 1a). The Ythdf2 +/loxp mice were crossed with mice ubiquitously expressing Cre-recombinase to generate the Ythdf2 +/− mice. Then to get Ythdf2 −/− mice, we intercrossed heterozygous Ythdf2 +/− mice. Interestingly, no viable Ythdf2 −/− newborn mice were identified in this particular knockout strain. The ratio of wild-type, hetero-, and homozygous knockout mice was not consistent with the expected 1:2:1 Mendelian ratio. Noteworthy, the number of postnatal Ythdf2 +/− mice indicated semi-lethality for these mice (Fig. 1b). Furthermore, 34% of Ythdf2 +/− surviving mice have malfunctioning eyes, with eyelids remaining closed (Additional file 1: Figure S1b). Many factors might contribute to this [29,30], such as dysfunction of hypothalamic nerve control, but this was not studied further here.
To assess the stage of developmental failure, we collected embryos at E12.5 and E14.5 from heterozygote intercrosses and genotyped them by both PCR with primers flanking and inside the deleted genomic region (Fig. 1a, c) and western blotting with Ythdf2 antibody (Fig. 1d). PCR and western blot analysis confirmed that the expression of Ythdf2 is completely depleted in Ythdf2 −/− embryos. The Mendelian distribution of wild type, Ythdf2 +/− , and Ythdf2 −/− was 1:2:1 when genotyped at embryonic stages E12.5-14.5 (Fig. 1b), suggesting the stage of embryonic lethality after E14.5. Therefore, we isolated embryos at E18.5 for further analysis. At this stage, 3 out of 41 embryos were genotyped as Ythdf2 −/− (data not shown). Despite the genotype ratio being normal at E12.5 and E14.5, the average number of embryos per litter was significantly less in Ythdf2 +/− intercrosses compared with wild-type intercrosses, especially at the late embryonic stage E18.5 (Fig. 1e). It was reported that removal of Ythdf2 in zebrafish leads to 31.3% cell arrest and lethality at the one-cell stage by Ythdf2 +/− intercross matings [13], consistent with our finding of 30% less embryos at E12.5 and E14.5. According to our data, the major lethality of Ythdf2 −/− embryos occurred between E14.5 and E18.5. Therefore, disruption of the Ythdf2 gene results in embryonic lethality during the late developmental stages of embryogenesis. a b d c e Fig. 1 Ythdf2 −/− mice are embryonic lethal. a The gene-targeting strategy to disrupt the Ythdf2 gene in mouse. Conditional Ythdf2 gene-targeted mouse contains LoxP sites flanking the 5′ UTR and exon 1 of the endogenous Ythdf2 locus. WT_F wild-type forward primer, WT_R wildtype reverse primer, KO_F Ythdf2 −/− forward primer, WT_R Ythdf2 −/− reverse primer, Ex exon. b Numbers of offspring from heterozygous Ythdf2 +/− intercrosses. The number and genotype of embryos at E12.5/E14.5 and postnatal are indicated. c PCR analysis of embryo tail DNA showing a 271-bp wild-type band (WT) and a 550-bp targeted band (KO) with primers displayed in a. d Western blot analysis of the Ythdf2 expression in wild-type and Ythdf2 −/− embryos. Two samples for each genotype. Actin was used as loading control. e Numbers of embryos per litter at E12.5/E14.5 and E18.5 from wild-type or heterozygous intercrosses. Error bars represent mean ± standard deviation, n = 7 litters. *P < 0.05, **P < 0.01, ***P < 0.001, Student's t-test Ythdf2 −/− mice display abnormal cortical development To determine how depletion of Ythdf2 affects embryonic development, we dissected embryos at E12.5, E14.5, and E18.5. Although Ythdf2 −/− embryos at E12. 5 and E14.5 were alive and appeared normal, sagittal sectionings of the whole embryos and H&E staining uncovered dramatically decreased overall cortical thickness of Ythdf2 −/− embryonic fore brains (Fig. 2a). Compared with their wild-type littermates, there was a general 56 μm decrease in the cortical layer at E12.5 and 40 μm decrease in the cortical layer at E14.5, yet the cortexes of both genotypes grew from E12.5 to E14.5 (Fig. 2b).
The Ythdf2 +/− mice are semi-lethal. Thus, we also analyzed a cohort of Ythdf2 +/− mice and found a mean 29 μm decrease in the cortical layer at E12.5 and a mean 24 μm decrease in the cortical layer at E14.5 (Fig. 2a, b). We suspected that the delayed cortical development derived from a defect in the early stages of neurogenesis. In order to determine whether Ythdf2 expression is temporally associated with brain development, we analyzed the expression of Ythdf2 in brain samples by quantitative RT-PCR at E12.5, E13.5, E17.5, and E18.5. Ythdf2 was highly expressed during the early stage of neural development (Additional file 1: Figure S1a To further define the neuronal developmental failure associated with Ythdf2 deficiencies, embryonic brain slices at different developmental stages were stained with the immature neuron marker doublecortin (Dcx). At E12.5, the neuronal layer of Ythdf2 −/− and Ythdf2 +/− embryos was significantly thinner than that of the wild type, here shown as the ratio of the thickness of Dcximmunolabeled neuronal layers over cortical layers (Fig.  2c, d). Taken together, the in vivo evidence indicates a striking phenotype of retarded cortical development, resulting from decreased neurogenesis at the early stages of embryonic brain development.
Basal progenitor cells are decreased in Ythdf2 −/− embryos Neural stem/progenitor cells (NSPCs) and immature neurons are the major cortical components at E12.5 and E14.5 in mice. NSPCs give rise to neurons. Given the profound effects of Ythdf2 targeting on embryonic brain development, we examined the proliferation and differentiation capability of NSPCs during development. The T-box transcription factor Eomes (Tbr2) is specifically expressed in basal progenitor cells, predominantly in the SVZ, which primarily differentiate into superficial layer neurons. In E12.5 and E14.5 Ythdf2 −/− embryos and, to a lesser extent, Ythdf2 +/− embryos, there was a dramatic loss of basal progenitor cells, displayed by the obviously thinner Tbr2 layer, compared to wild type littermate embryos (Fig. 3a). The sex determining region Y-box2 (Sox2) is a marker for apical progenitor cells located in the VZ, which can produce deep layer neurons and basal progenitor cells [31]. The ratio of Tbr2-positive cells to total progenitors (Tbr2 + /Sox2 + ) was decreased markedly at E12.5 and E14.5 in Ythdf2 −/− and Ythdf2 +/− embryos compared with the wild types ( Fig. 3b), suggesting the decrease in neurons (Dcx + ) associates with a reduction in the basal progenitor population in SVZ. However, there was no obvious difference in Sox2-positive apical progenitor cells in VZ layer (Fig. 3a).

Mitotic capability of apical progenitor cells is impaired in
The non-self-renewing basal progenitors only experience one or two mitotic cycles, and the majority of basal progenitors are established by asymmetric division of apical progenitor cells during early cortical development [32,33]. We propose that the decrease in basal progenitors (Tbr2 + ) might be caused by the reduced mitotic capability of the Ythdf2-depleted apical progenitor cells. The E12.5 and E14.5 sagittal sections of wild type, Ythdf2 +/− , and Ythdf2 −/− embryos were co-stained with the mitotic phase marker phospho-histone H3 (Phh3) and Sox2 to quantify the mitotic capability of the apical progenitor cells. The number of Phh3-positive cells decreased more than two-fold in Ythdf2 −/− cortex compared with wild type at E12.5 and E14.5. In Ythdf2 +/− cortex, the number of Phh3-positive cells was significantly reduced compared to the wild type cortex, yet was higher than in Ythdf2 −/− cortex (Fig. 3c, d). Additionally, apical progenitor cells could also maintain the population by several rounds of symmetric division in the VZ layer [34]. As there were no obvious changes in the number of apical progenitor cells (Sox2 + ) in the VZ layer, we concluded that Ythdf2-dependent defective neurogenesis was caused by the decreased generation of basal progenitors from apical progenitors.

Ythdf2 −/− NSPCs exhibit decreased proliferation in vitro
To further understand how Ythdf2 regulates neurogenesis, we cultured neurospheres consisting of NSPCs derived from E14.5 wild-type and Ythdf2 −/− embryonic fore brain. The Ythdf2 −/− neurospheres were smaller than the wild-type spheres (Additional file 1: Figure S2a, b). We first monitored the influence of Ythdf2 on NSPC proliferation. NSPCs dissociated from the primary neurospheres were seeded for proliferation testing and the cell growth was determined at 0, 24, 72, and 120 h. Compared with the wild type, Ythdf2 −/− NSPCs showed a slightly decreased proliferation rate after 24 h and a more pronounced reduction after 72 h culturing (Fig. 4a). This result is in agreement with the decreased mitotic capability of stem/progenitor cells observed in vivo.

Ythdf2-deficient NSPCs show impaired neural differentiation
In differentiation assays, NSPCs dissociated from neurospheres produce both neurons and glial cells after 5 days culturing. We first assessed the mRNA expression profile of Ythdf2 in wild-type neurospheres during differentiation by RT-qPCR. The expression of Ythdf2 was upregulated from Day 0 (D0) to D3 during differentiation and remained high till D5, suggesting the involvement of Ythdf2 in regulating differentiation (Fig. 4b). Neuronal and glial cell lineages can be identified by staining with antibody against microtubule associated protein 2 (Map2) or glial fibrillary acidic protein (Gfap), respectively. We quantified the percentages of Gfap-positive cells for Ythdf2 −/− and wild-type at D5 and D7. Dramatic reduction of glial cells, with abnormal branches (Gfap + ), was observed in differentiated Ythdf2 −/− neurospheres (Fig. 4c, d). However, we did not observe a significant different ratio of Ythdf2 −/− neurons (Map2 + ) at D5, while the ratio of Ythdf2 −/− neurons declined significantly more than the wild type at D7. These results were further substantiated by neuron progenitor antibody neuron-specific class III beta-tubulin (Tuj1) and glial progenitor antibody S100 calcium-binding protein B (S100-β) staining at D3 and D5 (Additional file 1: Figure S2c, d). At D3, the number of glial lineage progenitors had already declined in Ythdf2 −/− cells, while no difference was observed for neuronal lineage progenitor cells (Additional file 1: Figure S2c, d). The TUNEL assay showed significantly more dead Ythdf2 −/− cells, which might result from impaired differentiation (Additional file 1: Figure S2e, f).

Ythdf2-deficient neurons display abnormal neurite outgrowth and increased sensitivity to arsenite
Whereas neuronal lineage differentiation (Map2 + or Tuj1 + ) was not affected at D5 in Ythdf2 −/− cells, the morphological analysis of Map2-positive cells showed that Ythdf2 −/− differentiated neurons had less and shorter primary neurites (axons and dendrites). The mean number of branching neurites per neuron in Ythdf2 −/− differentiated cells is less than in the wild type (Fig. 4e), and the mean length of the longest neurite in Ythdf2 −/− differentiated cells is shorter than in the wild type (Fig. 4f ). The neurite outgrowth is pivotal in neuronal development and maturation, synaptic formation, neuronal function, and functional recovery in diseases [35]. The severe effect on neurite branching and extension of a b c d Fig. 3 The number of basal progenitors and mitotic capability of apical progenitors depends on Ythdf2. a Immunostaining of E12.5 and E14.5 sagittal sections with Tbr2 (green) and Sox2 (red) antibodies in wild type, Ythdf2 +/− , and Ythdf2 −/− embryos. VZ ventricular zone, SVZ subventricular zone, IZ intermediate zone, CP cortical plate. Nuclei were counterstained with DAPI. b Percentage of Tbr2 + cells over Tbr2 + /Sox2 + at E12.5 and E14.5. Error bars represent mean ± standard deviation, n = 3 biological and 3 technical replicates. Scale bars, 20 μm. c Immunostaining of E12.5 and E14.5 sagittal sections with Phh3 (green) and Sox2 (red) antibodies in wild type, Ythdf2 +/− , and Ythdf2 −/− embryos. Nuclei were counterstained with DAPI. d Number of Phh3 + cells per 400 μm of the cortical wall at E12.5/E14.5 from c. Error bars represent mean ± standard deviation, n = 3 biological and 3 technical replicates. *P < 0.05, **P < 0.01, ***P < 0.001, Student's t-test. Scale bars, 20 μm Ythdf2 −/− neurons might also contribute to the defective neurogenesis during neural development. Besides, we found that differentiated neurons in vitro were more sensitive to arsenite treatment. Arsenite was demonstrated to induce oxidative stress by generating ROS and depleting antioxidants in cell lines and mammalian brain [36,37]. It is reported that after arsenite treatment, Ythdf2 can co-localize with P body to regulate mRNA decay [7]. We treated differentiated neurons with 5 μM arsenite for 24 h in vitro, followed by recovery in fresh medium for 24 h (Additional file 1: Figure S3a). For wild-type neurons, the mean length of neurites was shortened and the number of neurites reduced after 24-h arsenite exposure (Additional file 1: Figure S3b, c). However, after 24-h culture in fresh medium, the remaining neurites recovered to the original length and the neurite number partially increased as the growth of new neurites needs longer time (Additional file 1: Figure S3b, c). In contrast, Ythdf2 −/− neurons showed increased sensitivity to arsenite exposure compared to wild-type neurons. After 24-h recovery, Ythdf2 −/− neurites could not outgrow to the original length and no new neurites projected.
Negative regulation of neural development pathways enriched in Ythdf2 −/− neurospheres To address the molecular mechanism of modulating NSPC proliferation and differentiation, we performed mRNA sequencing in the wild-type and Ythdf2 −/− neurospheres with three biological replicates. We identified 2144 up-regulated differentially expressed genes (DEGs) and 1756 down-regulated DEGs (Additional file 1: Figure S4a). With more stringent criteria (fold change > 1.5, P < 0.05 in three replicates), 151 significantly up-regulated and 316 significantly down-regulated genes were identified in Ythdf2 −/− neurospheres. Interestingly, the up-regulated genes were significantly associated with axon guidance, synapse assembly, neuron differentiation, and apoptosis. All these biological processes are subordinate to nerve development (Additional file 1: Figure S4b). The JAK-STAT signaling pathway is up-regulated in neurons and glial cells, which contributes to the neuroprotection and neurite outgrowth [38,39]. The genes, highly enriched for Gene Ontology (GO) term "negative regulation of JAK-STAT cascade", inhibit this cascade, such as Flrt2, Flrt3, Ptprd, and Lrrtm1 and 4. On the contrary, clustered terms, such as "positive regulation of cell differentiation", "positive regulation of transcription", "positive regulation of GTPase activity", and "negative regulation of neuron apoptotic process", were dominant in down-regulated genes. Initially, we quantified the m 6 A/A ratio of the total mRNAs purified from the wild-type and Ythdf2 −/− neurospheres by LC-MS/MS. In Ythdf2 −/− neurospheres, the m 6 A abundance was increased by around 10% on average compared with the wild type (Fig. 5a). This is consistent with the m 6 A-dependent RNA decay function of Ythdf2 [7] and correlates very well with a study in zebrafish on the role of Ythdf2 in the maternal-tozygotic transition [23]. We identified 16,626 common m 6 A sites from 8201 genes and 17,734 common m 6 A sites from 8585 genes in three biological replicates of wild-type and Ythdf2 −/− neurospheres, respectively (Additional file 1: Figure S5a). The highly overrepresented m 6 A RRACH (R = G/A, H=U/A) motif identified using the HOMER algorithm in both wild-type (P = 1e-471) and Ythdf2 −/− (P = 1e-475) neurospheres proved the successful enrichment of m 6 A-modified mRNA ( Fig. 5b and Additional file 1: Figure S5b). The m 6 A sites were significantly enriched at start codons, stop codons, and 3′ UTRs. The m 6 A profile is thus in very good agreement with those reported previously ( Fig. 5c and Additional file 1: Figure S6a). Based on the statistics from three biological replicates, 3095 m 6 A sites from 2464 genes and 4109 m 6 A sites from 2619 genes were identified to have lower or higher m 6 A levels in three biological replicates of Ythdf2 −/− neurospheres ( Fig. 5d and Additional file 1: Figure S6b). m 6 A sites with significantly higher enrichment (fold change > 1.5) in all three Ythdf2 −/− replicates were analyzed further. Based on this stringent criterion, 78 m 6 A sites from 69 genes were markedly up-regulated. These genes were enriched for functional clusters like transcription regulation, phosphorylation, and neuron projection development (Additional file 1: Figure S7a). On the other hand, 102 m 6 A sites from 99 genes were down-regulated. These genes were enriched for functional clusters like transcription regulation, transport, rhythmic process, and apoptosis (Additional file 1: Figure S7b). Among these 168 genes, 115 genes had conserved m 6 A sites across samples, while 65 and 54 genes had newly occurring or absent m 6 A sites, respectively, in all three Ythdf2 −/− neurospheres (Additional file 1: Figure S7c).

Ythdf2 is required for degradation of genes related to neuron differentiation
It is well established that Ythdf2 specifically binds mRNAs containing m 6 A and promotes mRNA decay [1,7]. In Ythdf2-depleted zebrafish embryos, Ythdf2-targeted mRNAs had extended lifetimes as seen by increased mRNA levels [13]. Hence, we focused on verifying candidate genes with increased mRNA transcripts and enrichment of m 6 A sites. Among these genes, Nrp2 and Nrxn3 were involved in nerve development and cell differentiation; Flrt2 and Ptprd were enriched in negative regulation of JAK-STAT cascade, regulation of synapse assembly and axon guidance, and neuron differentiation; Ddr2 was related to fibroblast proliferation; Hlf was involved in rhythmic process; and Nrp2 and other genes showed enrichment of representative m 6 A peaks in Ythdf2 −/− neurospheres ( Fig. 5e and Additional file 1: Figure S8). To further substantiate these findings, we performed m 6 A immunoprecipitation (IP) combined with RT-qPCR. Consistent with our initial findings, m 6 A IP showed that m 6 A levels increased significantly, while non-methylated actin was used as negative control (Fig. 6a). RT-qPCR showed that target genes were markedly enriched in Ythdf2 −/− neurospheres compared with the wild type (Fig. 6b). Next, we analyzed whether these m 6 A-enriched genes are real Ythdf2 targets by RNA IP (RIP) analysis. We confirmed that the Ythdf2 antibody was applicable to IP (Additional file 1: Figure S9). Compared with Ythdf2 −/− , Nrp2 mRNA and other candidates were enriched by Ythdf2 protein in the wild type, which was verified by qPCR (Fig. 6c). To examine whether increased gene expression was due to loss of Ythdf2-mediated RNA decay, we measured the mRNA life time of these candidate genes by inhibition of transcription with actinomycin D in wild-type and Ythdf2 −/− neurospheres. After actinomycin D treatment, mRNA levels of Nrp2 and the other candidate genes in wild type declined more rapidly than in Ythdf2 −/− neurospheres ( Fig. 6d and Additional file 1: Figure S10). Thus, the increased levels of m 6 Amodified mRNA transcripts in the absence of Ythdf2 were caused by delayed mRNA clearance, which might contribute to the defects in neurogenesis.

Conclusions
Ythdf2 is essential for oocyte maturation and early zygotic development in zebrafish and mouse [13,25]. Ythdf2 was recently reported to be required for oocyte competence through the post-transcriptional regulation of the maternal transcriptome and homozygous Ythdf2 −/− mice were reported to be partially permissive at weaning, with approximately 80% loss of homozygous Ythdf2 −/− mice in inbred C57BL/6 mice [25]. The targeting of the homozygous Ythdf2 −/− mice in our inbred C57BL/6 background and the majority of Ythdf2 −/− embryos died at late embryonic stages (Fig. 1). Of note, intercrossing Ythdf2 +/− mice results in constantly smaller litter size than from wild type matings (Fig. 1e), indicating an essential role of Ythdf2 in early embryo development. Here we reveal a crucial role of m 6 A in mRNA and its binding protein Ythdf2 in neural development at embryonic developmental stages. The mammalian nervous system arises from the ectoderm, with both neurons and glial cells (astrocytes and oligodendrocytes) generated from NSCs in a precisely regulated spatiotemporal order [34]. We propose that erroneous recognition and degradation of m 6 A-containing mRNA at this stage leads to the dysregulation of neural development. The m 6 A level in mRNAs is higher in brain than in other studied mouse organs, indicating a crucial role during normal brain development [11]. Recent studies found m 6 A-modifying enzymes Mettl3, Alkbh5, and Fto to be involved in regulating progression of glioblastoma, indicating that m 6 A epitranscriptomic regulation plays roles in the nervous system. Very recently, Yoon et al. [27] used a methyltransferase Mettl3-Mettl14 complex knockout to demonstrate that m 6 A depletion extends cortical neurogenesis by protracting cell cycle progression of NPCs. In this study, we demonstrate the severe impact of Ythdf2 deletion on corticogenesis, neurogenesis, and gliogenesis. During early neural development, the decreased thickness of cortex is attributed to the dramatically thinner CP and SVZ layers, composed of neurons (Dcx + ) and basal progenitor cells (Tbr2 + ), respectively (Figs. 2c and 3a). Multiple factors are supposed to contribute to this. First, consistent with the documented function of m 6 A in proliferation of NPCs [27], our in vivo and in vitro evidence reveals that the proliferation capability of the NSPCs is severely compromised in Ythdf2 −/− embryonic cortex NSPCs. Further, apical progenitors symmetrically divide into more apical progenitors to expand the stem cell/ progenitor VZ pool [34]. No significant change in the thickness of the VZ layer in Ythdf2 −/− embryos suggests that the symmetric division of apical progenitors is not disturbed. However, apical mitosis is significantly decreased in Ythdf2 −/− embryos. Apical progenitor cells can give rise to basal progenitor cells and neurons by asymmetric division [40,41]. The switch between symmetric and asymmetric cell division of Ythdf2 −/− neural progenitor cells, which determines self-renewal or differentiation, may be disturbed. It is worth mentioning that it is not confirmed that the neural defects observed contribute to embryonic lethality. So it will be interesting to address this relationship by generating neural-specific Ythdf2 knockout mice. Second, the NSPCs derived from Ythdf2 −/− embryo brains generated similar numbers of neurons as wild type in vitro (Fig. 4c and Additional file 1: Figure S2c). However, morphological analysis demonstrated abnormal neurite outgrowth of Ythdf2 −/− neurons that are more vulnerable to stress and fail to recover from neurite degeneration (Fig. 4e, f and Additional file 1: Figure S3). Proper neurite outgrowth and branching is pivotal for establishing neuronal circuits which facilitate nervous system function [42]. Interestingly, RNA-seq analysis shows that differentially expressed genes (DEGs) relate to functions such as axon regulation, synapse assembly, and neuron differentiation (Additional file 1: Figure S4). Among them, genes such as Ddr2, Rnf135, Flrt2, Hlf, Nrp2, Nrxn3, and Ptprd have both up-regulated mRNA and m 6 A levels (Fig. 6a, b). Ythdf2-mediated mRNA decay affects the translation efficiency and lifetime of m 6 A-modified mRNA targets [7]. By recruiting the Ccr4-not deadenylase complex, Ythdf2 initiates the degradation of its mRNA targets at specialized decay sites [43]. The RIP combined RT-qPCR and mRNA life-time assays display that mRNA levels of these genes are stabilized due to the complete absence of Ythdf2 in Ythdf2 −/− NSPCs (Fig. 6c, d and Additional file 1: Figure S10). Delayed mRNA degradation causes the retention of m 6 Amodified transcripts in Ythdf2 −/− neurospheres, leading to increased m 6 A enrichment.
Last but not least, while homozygous Ythdf2 −/− is embryonic lethal, heterozygous Ythdf2 +/− is unexpectedly only partially lethal, with 30% of the surviving Ythdf2 +/− mice having eye defects (Additional file 1: Figure S1b), which may reflect haploid insufficiency of Ythdf2 and a malfunctioning nervous system. Furthermore, m 6 A is highly enriched in mouse brain, and the level is dramatically increased with postnatal aging [11]. Taken together, we propose that m 6 A and Ythdf2 have a pivotal function in brain not only during embryonic neural development but also in postnatal life. Thus, functions of m 6 A and Ythdf2 on postnatal nervous system development merits further investigations.
During revision of this manuscript, two studies relating to the role of m 6 A in the adult mammalian nervous system were reported. One study found that either the m 6 A methyltransferase Mettl14 or the m 6 A-binding protein Ythdf1 regulate functional axon regeneration in the peripheral nervous system in vivo by modulating injury-induced protein translation [44]. In another study it was discovered that m 6 A in mRNA regulates histone modification in part by destabilizing transcripts that encode histone-modifying enzymes, which might be a previously unknown mechanism of gene regulation in mammalian cells [45].
In summary, our study demonstrates a pivotal function of Ythdf2-mediated m 6 A epitranscriptomic regulation in cortical neurogenesis during embryonic neural development, via regulating RNA degradation of m 6 A-tagged genes associated with neural development and differentiation.

Generation of conditionally Ythdf2 gene-targeted mice
The Ythdf2 conditional knockout mouse model (mYthdf2-CKO) was generated as described in Additional file 1: Figure S1a by Applied Systemcell Inc. (CA, USA) using CRISPR-Cas9 technology. A cocktail of active guide RNA molecules (gRNAs), two single-stranded oligo donor nucleotides (ssODNs) and qualified Cas-9 mRNA was microinjected into the cytoplasm of C57BL/6 embryos. Two LoxP sites were inserted, flanking the upstream of 5′ UTR and intron 1 regions, resulting in loss and changes in size of PCR products. Ythdf2 fl/fl mice were genotyped and further sequenced for the LoxP cassettes at the designated locations. Potential Ythdf2-CKO mice were generated by crossing Ythdf2 fl/fl mice with Cre_Del_GT_07 mice from the Norwegian Transgenic Center (NTS, Oslo, Norway).

RT-qPCR analysis
The total RNA was isolated using TRIzol LS Reagent (Life Technologies, 10,296-010). Normally, 1 μg total RNA was used for reverse transcription using High-Capacity cDNA Reverse Transcription Kit (ThermoFisher, 4,368,814). The quantitative PCR reactions were carried out with Power SYBR Green PCR Master Mix (Life Technologies, 4,368,708) on a StepOnePlus™ Real-Time PCR System instrument (Applied Biosystems). Primers used in this study were as follows.

Western blotting
Total protein lysate was extracted with RIPA buffer (20 mM Tris-HCl, pH 7.4, 20% glycerol, 0.5% NP40, 1 mM MgCl 2 , 150 mM NaCl, 1 mM EDTA, 1 mM EGTA). Protein concentrations were measured using the Bradford Assay, and 50-100 μg protein extracts were subjected to SDS-PAGE. Then proteins were transferred to a nitrocellulose membrane, blocked with 5% non-fat milk and incubated with first antibodies for 1 h at room temperature. After incubation with secondary antibody against mouse (1:10,000) or rabbit (1: 10,000) for 1 h at room temperature, the membrane was visualized with an ECL Western Blotting Detection Kit (32,106, Thermo).

Immunohistochemistry and immunofluorescence
For immunohistochemistry, embryonic brain tissues were dissected in cold PBS and fixed in 4% PFA at 4°C for 48 h. Slides (4 μm thick) were sectioned by microtome (HM355s, Thermo Scientific) and deparaffinized and cleared in Clear-Rite™ 3 (6901TS, Thermo) followed by rehydration in an EtOH gradient. After antigen retrieval in citrate buffer (pH 6.4), the slides were blocked with blocking buffer (5% goat gut, 5% BSA, 0.1% tween-20, 0.5% Triton X-100) for 1 h, and incubated with primary antibodies overnight at 4°C. Secondary antibodies were applied at room temperature for 1 h. For immunofluorescence, cultured cells were fixed with 4% paraformaldehyde (PFA), permeabilized with 0.1% Triton X-100, and stained with primary antibodies and secondary antibodies. Nuclei were visualized with mounting medium with DAPI (BioNordika, H-1200). Images were taken with a Leica SP8 confocal microscope equipped with a ×40 oil immersion lens.

H&E staining
Tissue slides were stained in haematoxylin (Richard-Allen Scientific, 12,687,756) and eosin (Nerliens Meszansky, 161,170) after dehydration and rehydration, followed by differentiation in acetic acid in 100% ethanol at 1:50,000 dilution for 5 s. Then, the sections were dehydrated in ascending series of ethanol, treated with xylene, and coverslipped using Cytoseal XYL xylene-based mounting medium (8312-4, Thermo). Images were taken with a Zeiss AxioPlan 2 microscope system.

TUNEL assay
Cells were grown on coverslips, fixed on ice with 4% PFA for 10 min, and permeabilized with 0.2% Triton X-100 in PBS-Tween for 30 min on ice. After incubating in 3% H 2 O 2 in PBS for 10 min, slides were rinsed twice with PBST. Slides were incubated with 50 μl TUNEL reaction mixture for 60 min at 37°C. Nuclei were visualized with mounting medium with DAPI (BioNordika, H-1200). Images were taken with a Leica SP8 confocal microscope equipped with a × 40 oil immersion lens.

Proliferation assay
Dissociated single NSPCs were seeded at a density of 1.0 × 10 4 per well in 96-well plates. The proliferation rates were measured at 24, 72, and 120 h with Presto-Blue Cell Viability reagent (A13262, ThermoFisher Scientific) as instructed.

Ythdf2 RIP
Ythdf2 RIP was carried out with a modified procedure [46]. Briefly, 1 × 10 7 collected NSPCs were lysed in NETN buffer (20 mM  and DNase I digestion following the manufacturer's instructions. We applied 1 mg total RNA for further mRNA purification with a Dynabeads mRNA DIRECT™ purification kit (Thermo, 61,011) for two rounds. The mRNA quality was checked using a 2100 Bioanalyzer instrument with an Agilent RNA 6000 Nano kit (5067-1511). RNA fragmentation (1 μg) was performed by sonication at 10 ng/μl in 100 μl RNase-free water with Bioruptor Pico (Diagenode) with 30 cycles of 30 s on followed by 30 s off; 5% of the fragmented RNA was saved as input. m 6 A IP was performed with an EpiMark® N 6 -Methyladenosine Enrichment Kit (NEB, E1610S) following the kit manual adapted for the KingFisher™ Duo Prime Purification System. In detail, 1 μl N6methyladenosine antibody from the kit and 25 μl Protein G beads (NEB #S1430) were used for each affinity pull down. After incubating with RNA, the beads were washed with 200 μl low salt reaction buffer twice, and then 200 μl high salt reaction buffer twice. RNA that was pulled down (IP) was eluted with 50 μl RLT buffer twice, and recovered by RNA Clean and Concentrator-5 (Zymo). Both input and IP were subjected to RNA library preparation with Truseq Stranded mRNA Library Prep Kit with the RFP incubation step shorten from 8 min to 20 s. Sequencing was carried out on Illumina HiSeq 4000 according to the manufacturer's instructions.

Sequencing data analysis
The sequencing data were mapped to mouse genome version mm10 downloaded from UCSC. Data analysis was carried out as previously described. Briefly, reads were aligned to mm10 using TopHat v2.0.142. For input analysis (RNA-seq), RPKM were calculated by Cuffnorm3. For m 6 A peak calling, the longest isoform was used if multiple isoforms were detected. Aligned reads were extended to 100 nucleotides (average fragment size) and converted from genome-based coordinates to isoformbased coordinates to eliminate interference from introns in peak calling. The longest isoform of each mouse gene was scanned using a 100-nucleotide sliding window with 10-nucleotide steps. To reduce bias from potentially inaccurate gene structure annotation and the arbitrary use of the longest isoform, windows with read counts less than 1/20 of the top window in both m 6 A IP and input samples were excluded. For each gene, the read count in each window was normalized by the median count of all windows of that gene. The window was called positive if FDR < 1% and log2(enrichment score) ≥ 1. Overlapping positive windows were merged. The following four numbers were calculated to obtain the enrichment score of each peak (or window): read count of the IP sample in the current peak/window (a); median read count of the IP sample in all 100-nucleotide windows on the current mRNA (b); read count of the input sample in the current peak/window (c); and median read count of the input sample in all 100-nucleotident windows on the current mRNA (d). The enrichment score of each window was calculated as (a × d)/(b × c). Common peaks shared in the triplicates of a sample were kept with the peak annotation from replicate 1.
For motif analysis, consensus motif was determined by using HOMER4. For GO analysis, differentially expressed genes or m 6 A-modified genes were uploaded to DAVID (http://david.abcc.ncifcrf.gov/). The GO terms were ranked and presented according to −log2(P value).

mRNA life-time assay
Wild-type and Ythdf2 −/− neurospheres were trypsinized with TrypLE™ Express Enzyme (Thermo, 12,605,010). The dissociated cells were seeded into plates coated with PDL (Millipore, A-003-E) and laminin (R&D Systems, 3446-005-01). After 12-h culturing, cells were treated with 5 μg/ml actinomycin D (Sigma, A9415) for 2 and 4 h, while cells without treatment were used as 0 h. Cells were collected at designated time points and total RNA was extracted for reverse transcription and qPCR.

Statistical analysis
All statistical analyses were performed with GraphPad Prism 5. Student's t-test was adapted and data are shown as mean ± standard deviation. P value is used for significance.

Additional file
Additional file 1: Figures S1-S10. This document contains additional supporting evidence for this study presented in the form of supplemental figures. (PDF 1160 kb)