m6A epitranscriptomic modification regulates neural progenitor-to-glial cell transition in the retina

  1. Yanling Xin
  2. Qinghai He
  3. Huilin Liang
  4. Ke Zhang
  5. Jingyi Guo
  6. Qi Zhong
  7. Dan Chen
  8. Jinyan Li
  9. Yizhi Liu
  10. Shuyi Chen  Is a corresponding author
  1. State Key Laboratory of Ophthalmology, Zhongshan Ophthalmic Center, Sun Yat-sen University, Guangdong Provincial Key Laboratory of Ophthalmology and Visual Science, China

Abstract

N 6-methyladenosine (m6A) is the most prevalent mRNA internal modification and has been shown to regulate the development, physiology, and pathology of various tissues. However, the functions of the m6A epitranscriptome in the visual system remain unclear. In this study, using a retina-specific conditional knockout mouse model, we show that retinas deficient in Mettl3, the core component of the m6A methyltransferase complex, exhibit structural and functional abnormalities beginning at the end of retinogenesis. Immunohistological and single-cell RNA sequencing (scRNA-seq) analyses of retinogenesis processes reveal that retinal progenitor cells (RPCs) and Müller glial cells are the two cell types primarily affected by Mettl3 deficiency. Integrative analyses of scRNA-seq and MeRIP-seq data suggest that m6A fine-tunes the transcriptomic transition from RPCs to Müller cells by promoting the degradation of RPC transcripts, the disruption of which leads to abnormalities in late retinogenesis and likely compromises the glial functions of Müller cells. Overexpression of m6A-regulated RPC transcripts in late RPCs partially recapitulates the Mettl3-deficient retinal phenotype. Collectively, our study reveals an epitranscriptomic mechanism governing progenitor-to-glial cell transition during late retinogenesis, which is essential for the homeostasis of the mature retina. The mechanism revealed in this study might also apply to other nervous systems.

Editor's evaluation

The authors' study investigated the role of m6A epitranscriptomic modification in the developing mouse retina. The study clearly demonstrated the defects of Mettl3CKO retina in mice, including cellular disorganization and abnormal physiological responses. Enriched scRNA-seq and MeRIP-seq data provide excellent resources to study the function of m6A modification in retinogenesis.

https://doi.org/10.7554/eLife.79994.sa0

eLife digest

The retina is a layer in the eye that converts light into electrical signals, which allows us to see. It is a part of the central nervous system and is made of brain cells, such as neurons and supporting cells called glia. These supporting cells protect neurons, supply them with nutrients and maintain steady surrounding conditions. The retina shares many characteristics with other neural tissues, so it is useful for biologists to study these structures.

One way for cells to control the activity of genes is by chemically modifying messenger RNA molecules. These alterations can affect various aspects of mRNA and the proteins that are ultimately produced. The most common mRNA modification, referred to as m6A, plays a key role in the development and healthy performance of various tissues. However, it is unclear whether m6A is involved in how glial cells in the retina develop.

To address this question, Xin et al. studied the impact of blocking m6A in the retina of mice. These genetically modified mice displayed abnormalities as the retina developed. Analysis of the mRNA produced in single cells and the pattern of modifications revealed that m6A is involved in the development of glia. In particular, m6A helps to remove the mRNA associated with early-stage proto-glia, allowing the cells to mature and transition to their final form.

The finding by Xin et al. that the m6A RNA modification is an essential part of retina development could help to understand eye diseases. In addition, this discovery may apply to other brain regions, and, in time, such work could lead to new treatments for neurodegenerative diseases.

Introduction

The retina is the neural sensory component of the visual system that is responsible for light perception and preliminary visual information processing. As a lateral derivative of the neural tube, the retina shares basic developmental, structural, and physiological principles with other neural tissues and has become an excellent model for studying neural biology. The retina is composed of six types of neurons, including rod and cone photoreceptors, bipolar cells, horizontal cells, amacrine cells, and retinal ganglion cells, that form complex neural circuits to detect and process visual signals (Masland, 2001). In addition to neurons, the retina contains one type of glial cell, Müller cells. As the sole glial cell inside the retina, Müller cells play pivotal roles in maintaining structural, physiological, and functional homeostasis in the neural retina (Newman and Reichenbach, 1996; Vecino et al., 2016). During development, all retinal neurons and Müller cells are derived from multipotent retinal progenitor cells (RPCs), which go through a series of competent states to give rise to various types of retinal cells in a sequential while overlapping manner. For example, in mice, retinal ganglion cells, cone photoreceptors, and GABAergic amacrine cells are born early before birth, while glycinergic amacrine cells, rod photoreceptors, and bipolar cells are born late during the postnatal period before RPCs finally become Müller glial cells (Agathocleous and Harris, 2009; Bassett and Wallace, 2012; Cepko, 2014; Heavner and Pevny, 2012). Similar to glial cells in other neural tissues, Müller cells maintain significant levels of gene expression signatures of progenitors (Blackshaw et al., 2004; Nelson et al., 2011; Roesch et al., 2008), which has encouraged people to explore the in vivo reprogramming potentials of Müller cells for treating retinal degeneration disease purposes in recent years (Hoang et al., 2020; Jorstad et al., 2017; Yao et al., 2018; Zhou et al., 2020). However, to function as glial cells, Müller cells must develop and maintain their unique transcriptome distinct from that of RPCs (Lin et al., 2019; Nelson et al., 2011). How the transcriptome of Müller cells is established and maintained is currently not well understood.

RNA modifications constitute an important layer of post-transcriptional regulation that orchestrates the metabolism, location, and function of transcripts, which are collectively referred to as the epitranscriptome. Among the various modes of mRNA modification, N6-methyladenosine (m6A) is the most prevalent internal modification (Meyer and Jaffrey, 2014; Meyer and Jaffrey, 2017; Roundtree et al., 2017; Shi et al., 2019; Zaccara et al., 2019). m6A is cotranscriptionally installed on RNAs in the nucleus by a multisubunit methyltransferase ‘writer’ complex, in which METTL3 and METTL14 form the catalytic core; but only METTL3 has methyltransferase catalytic activity, while METTL14 functions as an allosteric activator (Bokar et al., 1994; Bokar et al., 1997; Liu et al., 2014; Sledź and Jinek, 2016; Wang et al., 2016a; Wang et al., 2016b). Two enzymes, FTO and ALKBH5, have been demonstrated to have m6A demethylase ‘eraser’ activity, emphasizing the reversible nature of the modification (Jia et al., 2011; Zheng et al., 2013). The effects of m6A on RNAs are mediated by various m6A readers, such as YTH family and HNRNP family proteins (Lee et al., 2020; Zaccara et al., 2019; Zhao et al., 2017). Gene knockdown and genetic knockout studies on m6A writers, erasers, and readers have shown that m6A plays important roles in regulating various aspects of organism development, physiology, and disease progression (Frye and Blanco, 2016; Liu et al., 2019; Livneh et al., 2020). However, the functions of the m6A epitranscriptome in the visual system remain unclear.

To test whether m6A plays roles during vertebrate retinal development, we used morpholinos to knock down the m6A-writer complex in zebrafish. Analysis of the retinal phenotype of these morphant zebrafish indicated that m6A promotes the timely differentiation and survival of RPCs in zebrafish (Huang et al., 2021). In this study, using a mouse model of retina-specific conditional knockout of Mettl3 and combining scRNA-seq, MeRIP-seq, and gene expression manipulation, we reveal that the m6A epitranscriptome plays critical roles during late retinogenesis by fine-tuning the transcriptomic transition from RPCs to Müller cells.

Results

Mettl3 deficiency leads to structural and physiological abnormalities in the retina

To investigate the function of the m6A epitranscriptome in the retina, we conditionally knocked out Mettl3, the core component of the m6A-writer complex, in RPCs using Six3-Cre (Furuta et al., 2000) and Mettl3floxed (Lin et al., 2017) mice. METTL3 is abundantly and ubiquitously expressed in the retina from early embryonic age to adult stage (Figure 1—figure supplement 1A–D). Immunofluorescence staining (IF) demonstrated that Mettl3 was efficiently knocked out in the central retinas of Six3-Cre+; Mettl3floxed/floxed mice (hereafter referred to as Mettl3-CKO) (Figure 1A and A’), and MeRIP-qPCR examination of several candidate genes showed that m6A levels were efficiently downregulated in Mettl3-CKO retinas (Figure 1—figure supplement 1E). Even though Mettl3 was efficiently deleted from RPCs, the retinas of Mettl3-CKO mice at p0 appeared grossly normal, exhibiting an already separated retinal ganglion cell layer (RGL) and a still-developing retinoblast layer (RBL), comparable to littermate control retinas (Figure 1B and B’). However, at p14, the time point when retinogenesis should have completed and mature retinal organization should have been established, histological examination showed that the Mettl3-CKO retinas exhibited severe structural disorganization (Figure 1C and C’). At this age, the control retinas were well laminated into three cellular layers and two synaptic layers (Figure 1C). In the mutant retinas, although all the cellular and synaptic layers were distinguishable, the outer nuclear layer (ONL) presented many rosettes, and some inner nuclear layer (INL) cells were drawn into the ONL (Figure 1C’, the asterisks indicate the rosette structures). We then examined the visual function of the retina using multifocal electroretinogram (mfERG). The results showed that the visual responses from most regions of Mettl3-CKO retinas exhibited abnormal wave forms and much reduced P1 peak amplitudes compared with control retinas (Figure 1D), demonstrating disrupted visual function of the mutant retinas. To examine whether retinal neurons and glial cells were generated in structurally distorted Mettl3-CKO retinas, we performed IF analyses of characteristic markers of various types of retinal cells on p15 retinal sections. The IF results showed that all types of retinal cells, including RECOVERIN+ photoreceptors, VSX2+ bipolar cells, SOX9+ Müller glia, CALBINDIN+ horizontal cells, HPC-1+ amacrine cells, and BRN3A+ retinal ganglion cells, were generated in Mettl3-CKO retinas, although some cells in the INL, especially bipolar cells, Müller glia, and horizontal cells, were displaced into the ONL (Figure 1—figure supplement 2).

Figure 1 with 2 supplements see all
Mettl3 deficiency leads to structural and physiological abnormalities in the retina.

(A, A’) Confocal images of METTL3 expression in p0 control (Ctrl) and Mettl3-CKO retinas showing that Mettl3 was efficiently deleted in the mutant retinas. (B, B’) H&E staining of p0 control and Mettl3-CKO retinas showing that the mutant retinas had a grossly normal histological structure. (C, C’) H&E staining of p14 control and Mettl3-CKO retinas showing that the mutant retinal structure was severely disrupted. * indicates rosette structures. (D) Multifocal electroretinogram (ERG) response recordings in each hexagonal region of the retinas stimulated with 19 hexagonal light signals showing that the visual function of Mettl3-CKO retinas was severely disrupted. The numbers on the top of each waveform plot indicate the averaged ERG response in the respective hexagonal area in nV/deg2. (E, E’) Confocal images of p14 retinas labeled with GFP to highlight Müller cells (mice on the Rlbp1-GFP background). The white arrows indicate the outer limiting membrane (OLM). * in (E’) indicates mislocalized somata of Müller cells. (F, F’) Confocal images of p14 retinas stained for N-CADHERIN to highlight the OLM. The white arrows in (F) indicate the OLM, and those in (F’) indicate breaks between the OLM. (G, G’) Confocal images of p6 retinas stained for N-CADHERIN showing that breaks in the OLM (white arrowheads) in Mettl3-CKO retinas started to appear at p6. RBL: retinoblast layer; GCL: ganglion cell layer; ONL: outer nuclear layer; OPL: outer plexiform layer; INL: inner nuclear layer; IPL: inner plexiform layer. Scale bars are 50 μm.

In the retina, Müller cells monitor the retina by extending elaborate cellular processes to interact with every retinal neuron and form an outer limiting membrane (OLM) and an inner limiting membrane (ILM) to ensheath the entire retina (Wang et al., 2017). Mechanical and genetic ablation studies have demonstrated essential roles of Müller cells in maintaining the structural and functional homeostasis of the retina (Byrne et al., 2013; MacDonald et al., 2015; Shen et al., 2012). To examine how Müller cells were affected by Mettl3 knockout, we placed Mettl3-CKO mice on the Rlbp1-GFP mouse background, whose Müller glia express GFP (Vázquez-Chona et al., 2009). GFP IF illustrated severely distorted morphology of Müller glia in Mettl3-CKO retinas: somata of many mutant Müller glia translocated from their normal position in the middle of the INL to move upward and mingle with photoreceptors in the ONL (Figure 1E’, the asterisks indicate the soma of Müller glia). Moreover, the apical edges of the mutant Müller glia failed to reach the apical edge of the retina, where they normally organize into the OLM; instead, they formed a discontinuous OLM that was separated into each rosette structure (Figure 1E’, the white arrows indicate the broken OLM). We then further examined the OLM by performing N-cadherin IF, which showed that the OLM in Mettl3-CKO retinas became discontinuous and exhibited many breaks, through which many photoreceptors fled into the subretinal space (Figure 1F’, the white arrows indicate breaks in the OLM). By examining the retinas of pups at different postnatal ages, we found that occasional breaks along the OLM started to appear as early as p6 (Figure 1G’, the white arrowheads indicate breaks in the OLM), coinciding with the time when morphological maturation of Müller glia starts (Wang et al., 2017).

Taken together, these results demonstrate that Mettl3 deficiency leads to structural and functional abnormalities in the mature retina and indicate that dysfunction of Müller glial cells might be the cause of these abnormalities.

Mettl3 deficiency distorts late-stage retinogenesis

We next performed BrdU labeling assays to examine the proliferative activity of RPCs. At E15.5, there were similar numbers of BrdU+ cells in Mettl3-CKO retinas and control retinas (Figure 2—figure supplement 1), suggesting that Mettl3 deficiency does not affect the proliferation of early RPCs. However, at p1, fewer BrdU+ cells were observed in Mettl3-CKO retinas than in control retinas (Figure 2A, A’ , and B), indicating that there were fewer proliferating RPCs in Mettl3-CKO retinas. To examine whether the cell cycle progression of late RPCs was affected by Mettl3 deletion, we harvested p1 retinas at 6 hr after BrdU injection and subjected them to costaining with BrdU and pH3. We calculated the percentage of BrdU+; pH3+ RPCs vs. BrdU+ RPCs, which showed that the percentage was significantly lower in Mettl3-CKO retinas than in control retinas (Figure 2C, C’ , and D), suggesting that S-to-M phase progression was prolonged in Mettl3-CKO late RPCs. The postnatal period occurs at the end of retinogenesis, when an increasing number of RPCs withdraw from the cell cycle and undergo terminal differentiation. We examined the rate of cell cycle withdrawal of RPCs by injecting BrdU at p1 and monitored their KI67 expression 48 hr later at p3. The results showed that there were significantly fewer BrdU+; KI67- RPCs in Mettl3-CKO retinas at p3 (Figure 2E, E’ , and F), meaning that Mettl3-CKO late RPCs withdrew from the cell cycle slower than control cells. Consistent with this slower withdrawal rate, at p5, when proliferating retinogenesis events were mostly terminated in control central retinas (Figure 2G and H), a significant number of proliferating cells were still observable in Mettl3-CKO central retinas (Figure 2G’ and H). To test the role of Mettl3 in late RPCs directly, we performed in vivo electroporation on retinas of p1 pups to specifically introduce shRNAs targeting Mettl3 into late RPCs. We harvested the retinas at p5 after 2 hr of BrdU labeling. Because our in vivo electroporation procedure often targets peripheral region of the retina where retinogenesis continues at this time point, we still observed many BrdU+ cells in the region, but most of these BrdU+ cells were GFP- in the retinas electroporated with control plasmids (Figure 2—figure supplement 2A and B). However, a significantly higher percentage of GFP+ cells were co-labeled with BrdU in the retinas electroporated with plasmids expressing shRNAs targeting Mettl3 (Figure 2—figure supplement 2A’ and B), suggesting that cell cycle withdrawal was delayed in the experimental group. Taken together, these experiments demonstrated that Mettl3 is required in late RPCs for their proper progression through the cell cycle and timely withdrawal from retinogenesis. This end stage of retinogenesis is the peak time point at which rods, bipolar cells, and Müller glia are generated. To examine whether Mettl3 deficiency affects retinal differentiation, we counted the numbers of rods, Müller glia, and bipolar cells in the retina at p7, when the major retinogenesis phase has passed in the central retinas, even in Mettl3-CKO mice. The results showed that the number of SOX9+ Müller glia was increased in Mettl3-CKO retinas (Figure 2I, I’ , and J, SOX9+ cells in the ganglion cell layer and inner plexiform layer were excluded to avoid miscounting astrocytes). On the other hand, the number of VSX2+ bipolar cells remained relatively unchanged (Figure 2K, K’ , and L), while the number of RHODOPSIN+ photoreceptors was reduced (Figure 2M, M’ , and N). We also determined the cell densities of other retinal neurons. The number of BRN3A+ RGCs, CALBINDIN+ amacrine cells, and CAR+ cones was comparable between mutant and control retinas (Figure 2—figure supplement 3A–F), while the number of CALBINDIN+ horizontal cells was significantly reduced in the mutant retinas (Figure 2—figure supplement 3E, E', and G).

Figure 2 with 4 supplements see all
Mettl3 deficiency distorts late-stage retinogenesis.

(A, A’) Confocal images of p1 retinas stained for BrdU. The pups were injected with BrdU, and their retinas were collected 2 hr later. (B) Quantification of the BrdU+ cells in (A) and (A’). (C, C’) Confocal images of p1 retinas stained for BrdU and pH3. The pups were injected with BrdU, and their retinas were collected 6 hr later. (D) Quantification of the BrdU+ and pH3+ cells in (C) and (C’). (E, E’) Confocal images of p3 retinas stained for BrdU and KI67. The pups were injected with BrdU at p1, and their retinas were collected 48 hr later. (F) Quantification of the BrdU+ and Ki67- cells in (E) and (E’). (G, G’) Confocal images of p5 retinas stained for BrdU. The pups were injected with BrdU, and their retinas were collected 2 hr later. (H) Quantification of the BrdU+ cells in (G) and (G’). (I, I’) Confocal images of p7 retinas stained for SOX9 to illustrate Müller cells. (J) Quantification of the SOX9+ cells in (I) and (I’). SOX9+ cells in the retinal ganglion cell layer and inner plexiform layer were excluded to avoid miscounting of SOX9+ astrocytes. (K, K’) Confocal images of p7 retinas stained for VSX2 to illustrate bipolar cells. (L) Quantification of the VSX2+ cells in (K) and (K’). (M, M’) Confocal images of p7 retinas stained for RHODOPSIN to illustrate rods. (N) Quantification of the RHODOPSIN+ cells in (M) and (M’). The data in (B, D, F, H, J, L, N) are presented as the means ± standard deviations, corresponding to at least three independent biological replicates. * p<0.05, **p<0.01. The scale bars in (G’) and (M’) are 50 μm and apply to (A–G’) and (I–M’), respectively.

To examine whether Mettl3 deficiency affects the survival of retinal cells, we performed TUNEL assay, which showed that apoptosis was significantly elevated in Mettl3-CKO retinas (Figure 2—figure supplement 4A–D). Costaining with cell-type markers showed that both RPCs and their differentiated progenies underwent apoptosis (Figure 2—figure supplement 4E–G). The elevation in cell death events resulted in severe degeneration of all cellular layers in Mettl3-CKO retinas at older ages (Figure 2—figure supplement 4H and H') and might have contributed to the reduced population of photoreceptors at p7 and the reduced population of BrdU+ cells at p0.

Single-cell transcriptome analyses demonstrate distorted and extended late-stage retinogenesis in Mettl3-CKO retinas

To examine how Mettl3 deficiency affects gene expression in the retina, we performed RNA-seq analyses of control and Mettl3-CKO retinas at p7, the time point at which the structural defects became obvious. In principal component analysis (PCA), mutant retinas were clearly separated from control retinas (Figure 3—figure supplement 1A). Differentially expressed gene (DEG) analysis showed that 580 genes were upregulated and 532 genes were downregulated in Mettl3-CKO retinas (Figure 3—figure supplement 1B). Pearson analysis showed only subtle differences between control and mutant retinal transcriptomes (Figure 3—figure supplement 1C), indicating that Mettl3 delicately regulates retinal gene expression.

To examine the transcriptomic and cell status changes in Mettl3-CKO retinal cells in a more comprehensive way, we performed single-cell RNA sequencing (scRNA-seq) on control and Mettl3-CKO retinal cells at p7. A total of 5995 control and 6992 Mettl3-CKO retinal cells passed quality control and were used for subsequent bioinformatic analysis. By applying unsupervised clustering and dimensional reduction and based on the expression of key marker genes, sequenced cells were clustered into nine clusters, representing the typical cellular composition of the retina at p7 (Figure 3A and B, Figure 3—figure supplement 2A). Rods, the largest cell population in the retina, occupied the largest cluster, and other retinal neurons, including cones, amacrine cells, bipolar cells, and Müller glial cells, each formed distinct clusters. Retinal ganglion cells and horizontal cells were not represented, probably due to their small population size. A group of cells along the trajectory toward bipolar cells and rods were judged as precursors for the two cell types based on their high expression of Otx2 and Neurod4. A group of cells at one edge extending from Müller glia highly expressed cell cycle drivers, such as Mki67, Pcna, and Ccnb1, suggesting that they were RPCs and were likely RPCs in the final phase of differentiation that reside on the retinal periphery at this time point; these cells were more abundantly observed in Mettl3-CKO retinas whose retinogenesis period was extended (Figure 3—figure supplement 3).

Figure 3 with 4 supplements see all
Single-cell RNA sequencing (scRNA-seq) analyses reveal distorted transcriptomes of late retinal progenitor cells (RPCs) and Müller glial cells in Mettl3-CKO retinas.

(A) UMAP presenting the cell clusters of all sequenced cells. (B) Heatmap illustrating the expression patterns of key markers of each retinal cell type. (C) UMAP presenting the distribution of control and Mettl3-CKO retinal cells. (D) Bar column presenting the cell compositions in control and Mettl3-CKO retinas. (E) Left: pseudotime presentation illustrating the developmental progression from RPCs to Müller glia; right: cell density distributions of RPCs and Müller glia over the pseudotime period. (F) Left: separate pseudotime presentations of control and Mettl3-CKO RPCs and Müller glia; right: cell density distributions of control and Mettl3-CKO cells over the pseudotime period. (G) UMAP of reclustered RPCs. The red dotted line encircles a distinct cell cluster composed solely of Mettl3-CKO RPCs. (H) Volcano plot showing gene expression differences between mutant RPCs (encircled by the red dotted line in A) and the remaining RPCs. (I) Bubble plot showing the biological processes enriched in genes that were upregulated in the Mettl3-mutant RPC cluster. (J) Bubble plot showing the biological processes enriched in genes that were downregulated in the Mettl3-mutant RPC cluster. (K) UMAP of reclustered Müller glia. The red dotted line encircles a distinct cell cluster composed solely of Mettl3-CKO Müller glia. (L) Volcano plot showing gene expression differences between mutant Müller glia (encircled by the red dotted line in E) and the remaining Müller glia. (M) Bubble plot showing the biological processes enriched in genes that were upregulated in the Mettl3-mutant Müller glial cluster. (N) Bubble plot showing the biological processes enriched in genes that were downregulated in the Mettl3-mutant Müller glial cluster.

Cell number counting showed that the proportion of rods was slightly reduced in Mettl3-CKO retinas; the proportion of RPCs in Mettl3-CKO retinas was nearly twofold higher than that in control retinas (Figure 3D, Figure 3—figure supplement 2B), consistent with the extended retinogenesis observed for Mettl3-CKO RPCs. Our manual counting showed that more Müller glial cells were generated in Mettl3-CKO retinas; however, the scRNA-seq results showed that the proportions of Müller cells in control and Mettl3-CKO retinas were comparable (Figure 3D, Figure 3—figure supplement 2B). These seemingly conflicting results were due to the different sampling strategies adopted in the two experiments. For manual counting, we selected the central region of the retina, while for scRNA-seq, we collected the whole retina. At p7, the remaining KI67+ retinogenesis region was larger in Mettl3-CKO retinas than in control retinas, thus, the region of the retina that had finished Müller gliogenesis was smaller in Mettl3-CKO retinas than in control retinas (Figure 3—figure supplement 3). As the consequence, the elevated Müller gliogenesis phenotype was masked by the smaller region that had finished the process in the scRNA-seq data analysis. To observe the progression of this end-stage retinogenesis, we selected RPCs and Müller glia for pseudotime analysis. The results demonstrated a continuous cell status transition from RPCs to Müller glia (Figure 3E), consistent with the gliogenic status of RPCs at this time point. Analysis of the pseudotime distributions of control and Mettl3-CKO cells revealed that the gliogenesis progression of Mettl3-CKO RPCs lagged behind that of control RPCs (Figure 3F), suggesting delayed Müller gliogenesis in Mettl3-CKO retinas, which is consistent with the extended retinogenesis in Mettl3-CKO retinas revealed by BrdU labeling (Figure 2). Thus, the scRNA-seq examination of p7 control and Mettl3-CKO retinal cells reveals distorted and extended late-stage retinogenesis in the mutant retina at the transcriptomic level.

Mettl3 deficiency distorts the transcriptomes of late RPCs and Müller glial cells

Clustering analysis of all the cells sequenced failed to reveal distinct cell clusters for the cells derived from the mutant retinas (Figure 3C). We suspected that subtle transcriptomic changes in some types of retinal cells were masked in the whole dataset. We thus performed reclustering of individual clusters of cells. The reclustering revealed distinct mutant clusters in the Müller, RPC, bipolar, and rod populations (Figure 3G and K, Figure 3—figure supplement 4A), especially for the Müller glia population, in which the mutant cell cluster was distantly separated from the control cell cluster (Figure 3K). Cell cycle phase analysis of RPCs showed that G2/M and S phase cells were distinctively distributed along two sides of UMAP, but the cell cycle phases were not responsible for the formation of the mutant cell cluster (Figure 3—figure supplement 4B). Since phenotypic analyses indicated that RPCs and Müller glia were the two cell types most significantly affected by Mettl3 deletion and that the mutant cell clusters were more pronounced in the RPC and Müller populations, we next focused on RPCs and Müller glia. DEG analysis of the mutant RPC cell cluster and the remaining RPC cells showed that 286 genes were upregulated and 212 genes were downregulated in mutant RPCs (Figure 3H and Source data 1). The upregulated genes were enriched for biological processes related to gene expression regulation, while downregulated genes were enriched for biological processes related to organelle, metabolism, and cell cycle regulation (Figure 3I and J). Importantly, key cell cycle machinery components, such as Ki67 and Brca2, were upregulated, but many cell cycle regulators, such as Ccna2 and Cdca8, were downregulated (Figure 3H–J and Source data 1), which might be a part of the molecular mechanism underlying the extended cell cycle progression observed in Mettl3-CKO RPCs (Figure 2). DEG analysis of the mutant Müller cell cluster and the remaining Müller cells showed that 239 genes were upregulated and 181 genes were downregulated in mutant Müller cells (Figure 3L and Source data 2). The upregulated genes in the mutant Müller cell cluster were enriched for genes regulating transcription that have been linked with RPC development, such as Six3 and Sox11 (Figure 3L and M and Source data 2). The downregulated genes in the mutant Müller cell cluster were enriched for regulators participating in ‘cell shape’ and ‘actin filament organization,’ including important actin cytoskeleton regulators, such as RhoA and Ezr (Figure 3L and N and Source data 2), indicating compromised physical properties of mutant Müller cells, which explained the distorted morphology of Müller cells and severely disorganized tissue structure in Mettl3-CKO retinas (Figure 1). In addition, many genes that are essential for the supporting functions of Müller cells, such as those involved in the ‘glycolytic process,’ including Aldoa, Eno1, and Rlbp1, were significantly downregulated in the mutant Müller cell cluster (Figure 3L and N and Source data 2), indicating that the physiological functions of Müller glia in Mettl3-CKO retinas were compromised, which likely contributes to the functional abnormality of Mettl3-CKO retinas.

Taken together, reclustering and DEG analyses of scRNA-seq data reveal that Mettl3 deficiency distorts the transcriptomes of late RPCs and Müller cells, which is likely the molecular basis for the distorted cellular behaviors of RPCs and Müller glia in Mettl3-CKO retinas.

The m6A epitranscriptome landscape of the mouse retina

To explore how m6A modification might be involved in retinal development and physiology, we next characterized the mouse retinal m6A epitranscriptome using MeRIP-seq. With a fold enrichment cutoff of 3, we detected 15,471 peaks on 6735 genes, 11,601 peaks on 6356 genes, and 14,041 peaks on 5940 genes in p0, p7, and adult retinas, respectively (Figure 4A and Source data 3). The retinal m6A epitranscriptomes at different ages were similar to each other that of the 8772 m6A-modified genes detected in the retina, 46% were detected in all ages, and 71% were detected in at least two ages (Figure 4A). The retinal m6A peaks were largely localized in the CDS region, with pronounced enrichment around the stop codon (Figure 4B, Figure 4—figure supplement 1A), and enriched for the ‘GGACU’ sequence (Figure 4C), consistent with the topological features observed in the m6A epitranscriptomes of other tissues (Dominissini et al., 2012; Meyer et al., 2012; Zaccara et al., 2019; Zhao et al., 2017). The mouse retinal m6A epitranscriptome was enriched for genes participating in biological processes related to retinal development, structural organization, and visual function, such as ‘synapse organization,’ ‘axonogenesis,’ ‘dendrite development,’ ‘cell junction assembly,’ and ‘retina development,’ including many genes that have been demonstrated to play essential roles in retinal development, physiology, or pathology (Figure 4D and E, Figure 4—figure supplement 1B). The retinal m6A epitranscriptome at p0 was further enriched for genes involved in cell cycle progression, consistent with the proliferative status of the tissue at the time (Figure 4—figure supplement 1B and Source data 4). Thus, the MeRIP-seq analyses revealed that the retinal m6A epitranscriptome is enriched for genes that play key roles in many essential activities of the retina and might participate in regulating these biological processes.

Figure 4 with 1 supplement see all
Mouse retinal m6A epitranscriptome.

(A) Pie graph illustrating the number of genes carrying m6A modification detected by MeRIP-seq in the mouse retina at different ages. The numbers represent the number of genes in the respective pie region. (B) Plots illustrating the distribution of m6A peaks along mouse retinal transcripts. (C) The most enriched motif among m6A peak sequences in the mouse retina transcriptome. (D) IGV views illustrating the m6A peak distribution along transcripts of important retinal regulatory genes revealed by MeRIP-seq. (E) Bubble plot showing the biological processes enriched in genes carrying m6A modification detected in the retinas of mice at all ages.

m6A modification promotes the degradation of RPC transcripts during RPC-to-Müller transition

To investigate how m6A modification regulates the retinal transcriptome, we performed integrative analyses of the scRNA-seq data and the MeRIP-seq data. First, we analyzed the expression levels of m6A-tagged retinal transcripts in control and Mettl3-mutant cell clusters. Single-sample gene set enrichment analysis (ssGSEA) showed that m6A-tagged transcripts tended to be upregulated in Mettl3-mutant RPC and Müller glia clusters, and in Mettl3-mutant rod and bipolar cell clusters, but to a lesser extent (Figure 5A). Grouping genes based on the number of m6A peaks showed that as the number of m6A sites increased, the genes were upregulated to greater levels in Mettl3-mutant RPC and Müller glia clusters (Figure 5B). These results suggest that m6A modification promotes the degradation of modified transcripts in RPCs and Müller cells. Examining the m6A modification status of the DEGs in Mettl3-mutant RPC and Müller glia clusters showed that most of the upregulated genes in Mettl3-mutant clusters were subject to m6A modification, while, in contrast, a much smaller portion of the downregulated genes did (Figure 5C), suggesting that one major direct cause for gene upregulation in Mettl3-mutant RPC and Müller glia clusters was due to lack of m6A-mediated degradation.

Figure 5 with 1 supplement see all
m6A modification promotes the degradation of retinal progenitor cell (RPC) transcripts.

(A) Top: violin plot of single-sample gene set enrichment analysis (ssGSEA) scores illustrating the expression levels of m6A-modified genes in control and Mettl3-CKO retinal cells. Bottom: violin graph of ssGSEA scores of randomly selected genes that do not carry m6A to serve as controls. (B) Violin plots of ssGSEA scores illustrating the expression levels of different groups of m6A-modified genes in control and Mettl3-CKO Müller glia (top) and RPCs (bottom). The genes are grouped based on the number of m6A peaks along their transcript. (C) Pie graphs showing the number of m6A-modified genes that were upregulated (green pies) or downregulated (yellow pie) in Mettl3-mutant RPC or Müller glia clusters. The numbers inside the parentheses indicate the number of genes, and the numbers inside the figure indicate the number of overlapping genes between the two pies. (D) UMAP of reclustered RPCs and Müller glial cells. (E) UMAP of reclustered RPCs and Müller glial cells with colors indicating the sample origin of the cells. (F) Violin plots of area under the curve (AUC) scores of the collective gene expression levels of RPC feature gene set (left) and Müller feature gene set (right) in different clusters of reclustered RPCs and Müller glial cells. Clusters were ordered based on the scores. (G) UMAP graph illustrating the distribution of cell clusters of RPCs, RPCs committed to Müller glial fate (RPCMüller), Müller glial cells just differentiated (MüllerEarly), and Müller glial cells, as well as their corresponding mutant cell clusters (_mut). (H) Violin plot of AUC scores illustrating the activity of m6A-modified RPC genes in the control and mutant cell clusters along the course of RPC-to-Müller transition.

As in most neural tissues, gliogenesis occurs toward the end of retinogenesis. By constitutively expressing a set of progenitor genes, Müller cells maintain a certain level of retinal regeneration potential (Blackshaw et al., 2004; Goldman, 2014; Jadhav et al., 2009). However, to function as a glial cell, Müller cells need to establish a unique transcriptome different from that of progenitors (Lin et al., 2019; Nelson et al., 2011). How this transcriptomic transition from progenitors to glial cells is achieved remains unexplored. Taking advantage of our scRNA-seq data that likely contain transcriptomes of the cells in the transition states between late RPCs and Müller glial cells, we tested whether further analyses could identify such cell clusters. For this purpose, we pulled out late RPCs and Müller glial cells and reclustered them. Reclustering analyses generated nine clusters (Figure 5D). Cells originally annotated as RPCs and Müller cells were clearly separated on the two sides of the UMAP graph (Figure 5—figure supplement 1A), and cells in the inner circle of the graph were composed purely of cells from Mettl3-CKO retinas (Figure 5E), consistent with original clustering analyses (Figure 3A, G and K). In previous work, we compared the transcriptomes of Müller glia and early RPCs, and identified 3717 and 2450 genes that are significantly more abundant in early RPCs and Müller glia, respectively (Lin et al., 2019). Using these two groups of genes as the RPC signature gene set and Müller glia signature gene set, we used AUCell to score RPC and Müller glia features in each cell. Putting the two pure Mettl3-CKO cell cluster (clusters 1 and 6) aside, area under the curve (AUC) scoring showed that RPC clusters 2 and 7 exhibited significantly lower levels of the RPC feature but higher levels of the Müller feature than other RPC clusters, while Müller cluster 4 exhibited significantly lower level of the Müller feature but higher level of the RPC feature than other Müller clusters (Figure 5F), indicating that these three clusters reside in intermediate states between late RPCs and Müller glial cells. We thus classified cells in clusters 2 and 7 as late RPCs committed to Müller glial fate (named RPCMüller), and cells in cluster 4 as Müller cells that had just exited the cell cycle and become Müller cells (named MüllerEarly). The reclustering did not identify distinct mutant clusters corresponding to clusters 2, 4, and 7. Based on the expression patterns of markers for clusters 4, and 7, we classified subsets of cells of clusters 1 and 6 as the corresponding mutant clusters for clusters 4 and 7 (named MüllerEarly_mut and RPCMüller_mut, respectively); based on the pure Mettl3-CKO sample origin, we classified a subset of cells at one edge of cluster 2 as the corresponding mutant cluster for cluster 2 (also named RPCMüller_mut). Thus, our reclusting and RPC/Müller feature analyses identified cell clusters in RPC-to-Müller transition states, as well as their corresponding mutant clusters (Figure 5G). We then scored the RPC and Müller features in this set of cell clusters. AUC scores showed that the RPC feature gradually decreased, and the Müller feature gradually increased from RPC clusters to Müller clusters in the control groups (Figure 5—figure supplement 1E and F). When examining the mutant groups, although the general trend was for the RPC feature to be elevated and for the Müller feature to be repressed compared with the corresponding control clusters, the differences were moderate (Figure 5—figure supplement 1E and F). We then determined the activities of subsets of RPC and Müller signature genes that carry m6A modification. Among the 3717 RPC signature genes, 1642 were subjected to m6A modification, and among the 2450 Müller signature genes, 894 were subjected to m6A modification. AUCell soring showed that, for the m6A subset of RPC signature genes, the activity in mutant clusters was significantly elevated to a level close to that in the control cluster of an earlier state (Figure 5H), while for the m6A subset of Müller signature genes, the activity in the mutant clusters remained relatively unchanged (Figure 5—figure supplement 1G). Thus, these analyses demonstrated that m6A-RPC transcriptome failed to be promptly downregulated during the RPC-to-Müller transition.

Taken together, these analyses on the expression of m6A-tagged genes in control and Mettl3-mutant retinal cells demonstrate that m6A modification promotes the degradation of RPC transcripts during the RPC-to-Müller transition. This process is likely essential for the timely termination of retinogenesis and the proper establishment of the transcriptome of Müller glial cells.

Overexpression of m6A-tagged RPC genes in late RPCs disturbs late-stage retinogenesis

To experimentally test whether m6A modification promotes the degradation of RPC transcripts, we selected eight m6A-tagged genes (Zfp292, Ckap4, Nxt1, Apex1, Bcl7a, Traf4, Rap2b, and Sulf2) that were significantly upregulated in the Mettl3-mutant RPC and Müller clusters (Figure 6—figure supplement 1A). All eight genes carry multiple m6A tags around the stop codon (Figure 6—figure supplement 1B) and are involved in various biological processes, such as transcription, cytoskeleton function, and nuclear export. Using MeRIP-qPCR, we confirmed that the m6A modification levels of the transcripts of these genes were reduced in Mettl3-CKO retinas (Figure 6A, Figure 6—figure supplement 1C). We next measured the half-lives of the transcripts of these genes in the control and Mettl3-CKO retinas at p1. The results showed that the half-lives of the transcripts of five of these genes were extended significantly in Mettl3-CKO retinas compared with those in the retinas of their littermate controls (Figure 6B, Figure 6—figure supplement 1D), demonstrating that m6A modification promotes the degradation of these RPC transcripts in the retina. To test how upregulation of m6A-regulated genes would affect late-stage retinogenesis, we performed in vivo electroporation to overexpress four genes that showed extended transcript half-lives in Mettl3-CKO retinas in p1 RPCs. At p3, 2 days after electroporation, we examined the cell cycle status of GFP-marked electroporated RPCs through KI67 or PCNA staining. The results showed that there were significantly fewer KI67- or PCNA-; GFP+ cells in the retinas electroporated with Zfp292-, Ckap4-, Traf4-, and Bcl7a- overexpression plasmids (Figure 6C and D, Figure 6—figure supplement 2A), suggesting that overexpression of these four genes prohibited RPCs from exiting the cell cycle. After exiting the cell cycle, many differentiating retinal cells migrate away from the central region of the retinoblast layer to their designated position in the mature retina. In the control retinas, we observed many GFP+ cells located near the vitreous edge of the retinoblast layer, away from KI67-positive zones, indicating that they were differentiating/maturing amacrine cells (arrowheads in Figure 6C, Ctrl). In contrast, much fewer such cells were observed in retinas overexpressing Zfp292, Ckap4, Traf4, or Bcl7a (Figure 6C), further supporting the conclusion that overexpression of these four genes prohibited RPCs from exiting the cell cycle. We then traced the cell fates of GFP-marked electroporated RPCs at p14. Overexpression of Zfp292, Ckap4, Traf4, or Bcl7a led to an increase in the percentage of Müller cells at the expense of amacrine cells or rod photoreceptors (Figure 6E and F), suggesting that overexpression of these four genes biased the late RPCs to adopt Müller glial fate versus neuronal fates. Overexpression of Rap2b increased the population of bipolar cells and decreased the population of amacrine cells (Figure 6—figure supplement 2B). We then coelectroporated Zfp292, Ckap4, Traf4, and Bcl7a in the retina. Co-overexpressing of all four genes did not exhibit an additive effect (Figure 6—figure supplement 2C), suggesting that co-overexpressing all four genes might elicit different gene-regulatory network changes that are different from those when expressed alone. Phalloidin staining of the electroporated retinas at p14 showed that while the OLMs of these retinas were grossly intact, occasionally at a few regions (three out of six retinas), the OLMs are broken, and a few rods escaped to the subretinal space (Figure 6—figure supplement 2D, arrows in the inset image indicate the escaped rods). We then tested the effects of knocking down these m6A-regulated genes in late RPCs. We electroporated plasmids expressing shRNAs targeting Zfp292, Ckap4, Traf4, and Bcl7a into p1 retinas. Late RPCs deficient for Bcl7a generated fewer Müller glial cells, while knocking down other three genes did not significantly affect retinogenesis (Figure 6—figure supplement 2E and F). Taken together, in vivo electroporation experiments demonstrated that expression levels of these m6A-regulated genes should normally be kept at relatively low levels and carefully controlled in late RPCs.

Figure 6 with 2 supplements see all
Overexpression of m6A-regulated genes disturbs late-stage retinogenesis.

(A) MeRIP-qPCR results revealed downregulation of m6A modification of the transcripts in Mettl3-CKO retinas. (B) Transcript half-life measurements revealed prolonged half-lives of the transcripts in Mettl3-CKO retinas. (C) Confocal images of p3 retinas stained for GFP and KI67. The retinas were transfected with plasmids overexpressing cDNAs of candidate genes via in vivo electroporation at p1. (D) Quantification of the percentage of GFP-labeled electroporated RPCs that were KI67- 48 hr after electroporation. (E) Confocal images of p14 retinas stained for GFP and SOX9. The retinas were transfected with plasmids overexpressing cDNAs of candidate genes via in vivo electroporation at p1. (F) Quantification of the cell compositions of GFP-labeled electroporated cells. The data in (A, B, D, F) are presented as the means ± standard deviation, corresponding to at least three independent biological replicates. *p<0.05, **p<0.01. The scale bars in (C) and (E) are 50 μm.

Discussion

In this study, using Mettl3-CKO mice as a model, we revealed that Mettl3 is essential for late-stage retinogenesis progression and structural and physiological homeostasis in the mature retina. Mechanistically, we showed that m6A promotes the degradation of RPC transcripts, thereby promoting the termination of retinogenesis and fine-tuning the transcriptomic transition from RPCs to Müller glial cells, the disruption of which leads to abnormalities in late-stage retinogenesis and compromises the glial functions of Müller cells and thus the structural and physiological homeostasis of the mature retina.

One interesting finding of our study is that the regulatory function of Mettl3 during retinal development is restricted to the late-stage retinogenesis period. Six3-Cre mice express CRE recombinase in RPCs beginning from embryonic day 9.5 (Furuta et al., 2000); however, the retinas of Mettl3-CKO mice remained normal until the postnatal period. Our in vivo electroporation experiments to directly knock down Mettl3 in late RPCs also demonstrated that Mettl3 is required in late RPCs for the proper progression of late retinogenesis. Even though all are referred to as RPCs, RPCs continue to adjust their behavior during retinogenesis. For example, RPCs gradually go through a series of competent states to produce different types of retinal cells at different retinogenesis periods so that early RPCs produce retinal ganglion, cone, horizontal, and some amacrine cells, while late RPCs produce rod, some amacrine, and bipolar cells before finally becoming Müller glial cells (Agathocleous and Harris, 2009; Cepko, 2014). Accompanying the gradual changes in the retinogenic behaviors, RPCs likely continue to adjust the gene-regulatory networks to accommodate retinogenic tasks at different developmental stages. Indeed, scRNA-seq analyses of retinal development showed that the transcriptomes of early and late RPCs are distinct from each other (Clark et al., 2019). The postnatal retinogenesis period is the critical time when late RPCs quickly slow the cell cycle and finally exit the cell cycle to terminate retinogenesis in a timely fashion. Our study suggests that the m6A epitranscriptome delicately coordinates the transcriptomic adjustment needed for the timely termination of retinogenesis. In Mettl3-mutant RPCs, key cell cycle machinery components, such as Ki67, were upregulated, but many other cell cycle regulators were downregulated, which is likely the molecular basis for the distorted cell cycle. Furthermore, many m6A-tagged RPC transcripts were abnormally upregulated in Mettl3-mutant RPCs, which likely disrupts the balance of the gene-regulatory networks in late RPCs. Indeed, overexpression of m6A-tagged genes in late RPCs distorts the retinogenesis behavior of the cells. Thus, our study suggests that m6A delicately modulates the transcriptome of late RPCs to prepare them for the timely termination of retinogenesis.

Numerous studies have demonstrated the essential roles of Müller glial cells in maintaining the structural and physiological homeostasis of the mature retina (Byrne et al., 2013; MacDonald et al., 2017; Shen et al., 2012; Wohl et al., 2017). A couple of lines of evidence indicate that dysfunction of mutant Müller glial cells is the cellular basis for the retinal structural abnormality of Mettl3-CKO retinas. First, the structural abnormality of Mettl3-CKO retinas did not appear until p6, coinciding with the time point when maturing Müller glial cells are transitioning from the simple bipolar radial shape inherited from RPCs to a much more sophisticated morphology to form a glial network to intimately interact with every cell in the mature retina (Wang et al., 2017). Second, our scRNA-seq analyses demonstrated that Müller glia are one of the two cell types whose transcriptomes were most severely affected by Mettl3 knockout, and many genes involved in cell–cell interaction and the cytoskeleton were downregulated in Mettl3-CKO Müller glial cells, suggesting that the structural supporting function of Müller glial cells was compromised. On the other hand, late RPCs might indirectly contribute to the structural defect by failing to downregulate m6A-tagged transcripts, leaving a distorted transcriptome to their Müller progenies and compromised the glial supporting function of Müller cells, leading to the structural abnormality of the retina.

As the last cell type generated by RPCs at the end of retinogenesis and as the major glial cell type in the retina, the close relationship between Müller glia and RPCs has long been noted. Many factors that play important roles in RPC proliferation and differentiation are also highly expressed in Müller glia, and several transcriptomic comparison studies have revealed similarities between Müller glia and RPCs at the transcriptome level (Blackshaw et al., 2004; Nelson et al., 2011; Roesch et al., 2008). Encouraged by these cues, the regenerative potential of Müller glia has been intensively explored in recent years (Goldman, 2014; Hoang et al., 2020; Jorstad et al., 2017; Lahne et al., 2020; Yao et al., 2018; Zhou et al., 2020). However, to function as glial cells, Müller cells must develop a unique transcriptome that differs from that of RPCs to accommodate their structural and physiological supporting roles in the adult neural retina (Lin et al., 2019; Nelson et al., 2011), but how this delicate transcriptome transition is achieved is unclear. By carefully analyzing the single-cell transcriptomes of late RPCs and Müller glial cells of the p7 retina, we were able to capture cell clusters in RPC-to-Müller transition states. The RPC feature gradually decreases, and the Müller feature gradually increases, from RPCs to Müller glial cells through these clusters. However, in Mettl3-mutant cell clusters, the activities of m6A-tagged RPC genes were significantly elevated to a level close to those in the cluster of an earlier state. These analyses demonstrated that m6A promotes the transcriptomic transition from RPCs to Müller glial cells by promoting the degradation of modified RPC transcripts.

In our effort to test the roles of m6A-regulated genes in late retinogenesis, we used in vivo electroporation on retinas of newborn pups to either overexpress or knock down candidate genes. Knocking down candidate genes mostly did not affect late retinogenesis, while overexpressing some candidate genes disrupted the cell cycle progression of late RPCs and promoted Müller gliogenesis at the expense of rod or amacrine cell generation, suggesting that the expression levels of these genes in late RPCs should normally be kept at a relatively low level and carefully controlled. Overexpression of m6A-regulated genes did not faithfully recapitulate the cell population change phenotype observed in Mettl3-CKO retinas. For example, in Mettl3-CKO retinas, the population of rods decreased, while in retinas overexpressing Zfp292 or Ckap4, the population of amacrine cells decreased. Two reasons might contribute to these discrepancies. First, in Mettl3-CKO RPCs, over 200 m6A-tagged genes were upregulated. The retinal cell-type composition change in Mettl3-CKO retinas is likely a collective result of the expression changes of many of these m6A-tagged genes; thus, overexpressing one (or a few) of them may not give rise to exactly the same phenotype as Mettl3-CKO retinas. Second, forced overexpression of Zfp292/Ckap4/Traf4/Bcl7a from plasmids electroporated into RPCs probably increases the expression levels of these genes by several to tens of times, while in Mettl3-CKO RPCs, their expression levels were only moderately elevated (Source data 1 and 2). Different levels of overexpression may elicit different gene-regulatory network changes, and thus different end results.

While this article was under revision, a manuscript was published showing that retinas deficient for all three m6A-readers, Ythdf1, Ythdf2, and Ythdf3, exhibit a phenotype similar to that of Mettl3-CKO retinas, suggesting that the three m6A readers work redundantly to ‘read’ m6A in the retina (Niu et al., 2022). As a derivative of the neural tube, the neural retina shares similar regulatory principles with other central neural tissues. Two groups independently studied the functions of m6A during cortical development using very similar strategies. Although the two studies presented controversial claims regarding the maintenance of neural stem cell (NSC) pools and the differentiation activities of NSCs, both studies showed that the cell cycle of NSCs deficient in Mettl14 (another essential component of the m6A writer complex) was prolonged (Wang et al., 2018; Yoon et al., 2017), a phenotype shared by Mettl3-CKO RPCs, suggesting that some regulatory mechanisms of m6A are conserved between NSCs and RPCs. Similar to the case with Müller cells in the retina, the supporting functions and regeneration potentials of glial cells in the central nervous system are hotly debated neural biological questions (Allen and Lyons, 2018; Guo et al., 2014; Qian et al., 2020; Wang et al., 2021; Zuchero and Barres, 2015). Yoon et al. showed that there are fewer astrocytes in p5 Mettl14-deficient cortices than in control cortices; however, since neurogenesis is still progressing in Mettl14-deficient cortices at this time point, it is not clear whether the final population of astrocytes is reduced or increased in the adult Mettl14-CKO cortex. Furthermore, it is also not clear whether the generated Mettl14-deficient astrocytes are transcriptomically and functionally normal (Yoon et al., 2017). Our study hints at an interesting possibility: that m6A may regulate the transcriptomic transition from NSCs to glial cells in other nervous systems, which warrants further investigation. Future studies are also needed to determine whether m6A is involved in cell fate reprogramming between glial cells and neurons.

Materials and methods

Key resources table
Reagent type (species) or resourceDesignationSource or referenceIdentifiersAdditional information
Gene (Mus musculus)Mettl3GenBankID: 56335
Genetic reagent (M. musculus)Mettl3floxedLin et al., 2017
Genetic reagent (M. musculus)Six3-CreFuruta et al., 2000
Genetic reagent (M. musculus)Rlbp1-GFPVázquez-Chona et al., 2009
AntibodyAnti-METTL3 (rabbit monoclonal)AbcamCat# ab195352;
RRID:AB_2721254
1:500
AntibodyAnti-RECOVERIN (rabbit polyclonal)MilliporeCat# AB5585;
RRID:AB_2253622
1:1000
AntibodyAnti-VSX2 (sheep polyclonal)MilliporeCat# AB9014;
RRID:AB_262173
1:500
AntibodyAnti-SOX9 (rabbit polyclonal)MilliporeCat# ABS571; RRID:AB_27838761:500
AntibodyAnti-CALBINDIN (rabbit polyclonal)SwantCat# CB38;
RRID:AB_10000340
1:500
AntibodyAnti-SYNTAXIN (mouse monoclonal)SigmaCat# S0664;
RRID:AB_477483
1:200
AntibodyAnti-BRN3A (mouse monoclonal)MilliporeCat# MAB1585;
RRID:AB_94166
1:100
AntibodyAnti-GFP (chicken polyclonal)AbcamCat# ab13970;
RRID:AB_300798
1:2000
AntibodyAnti-N-CADHERIN (rabbit polyclonal)Santa CruzCat# sc-7939;
RRID:AB_647794
1:1000
AntibodyAnti-GS (mouse monoclonal)BD BiosciencesCat# 610518;
RRID:AB_397880
1:2000
AntibodyAnti-GFAP (mouse monoclonal)SigmaCat# G3893;
RRID:AB_477010
1:2000
AntibodyAnti-BrdU (mouse monoclonal)GE HealthcareCat# RPN2021:50
AntibodyAnti-pH3 (mouse monoclonal)Cell SignalingCat# 9706;
RRID:AB_331748
1:2000
AntibodyAnti-KI67 (rabbit polyclonal)AbcamCat# ab15580;
RRID:AB_443209
1:500
AntibodyAlexa Fluor 488 donkey anti-mouse (donkey polyclonal)Thermo FisherCat# A212021:1000
AntibodyAlexa Fluor 568 donkey anti-mouse (donkey polyclonal)Thermo FisherCat# A100371:1000
AntibodyAlexa Fluor 488 donkey anti-rabbit (donkey polyclonal)Thermo FisherCat# A212061:1000
AntibodyAlexa Fluor 568 donkey anti-rabbit (donkey polyclonal)Thermo FisherCat# A100421:1000
AntibodyAlexa Fluor 488 donkey anti-chicken (donkey polyclonal)Jackson ImmunoResearchCat# 703-296-1551:1000
Recombinant DNA reagentpCAGIG (plasmid)AddgeneCat# 11159CDS overexpression
Recombinant DNA reagentpSicoR (plasmid)AddgeneCat# 11579shRNA expression
Sequence-based reagentRsp14 forwardThis paperMeRIP-qPCR primersacctggagcccagtcagccc
Sequence-based reagentRsp14 reverseThis paperMeRIP-qPCR primerscacagacggcgaccacgacg
Sequence-based reagentPax6 forwardThis paperMeRIP-qPCR primersactctgccaatgactatgtg
Sequence-based reagentPax6 reverseThis paperMeRIP-qPCR primersctccagttcaggacagttac
Sequence-based reagentSox2 forwardThis paperMeRIP-qPCR primersctggactgcgaactggagaa
Sequence-based reagentSox2 reverseThis paperMeRIP-qPCR primersactctcctctttttgcaccc
Sequence-based reagentRax forwardThis paperMeRIP-qPCR primersggaaattcagcctcgctgtc
Sequence-based reagentRax reverseThis paperMeRIP-qPCR primersccaggtcaagatccttggtc
Sequence-based reagentZfp292 forwardThis paperMeRIP-qPCR primersgggaaataacgaatttcagg
Sequence-based reagentZfp292 reverseThis paperMeRIP-qPCR primersctctcttcaaattaccaggc
Sequence-based reagentCkap4 forwardThis paperMeRIP-qPCR primersggaacgacctggataggttg
Sequence-based reagentCkap4 reverseThis paperMeRIP-qPCR primerscgtaagaaactgtgcccacac
Sequence-based reagentTraf4 forwardThis paperMeRIP-qPCR primersggatgatgcggttttcatccg
Sequence-based reagentTraf4 reverseThis paperMeRIP-qPCR primersccagtttcagatccagtcccg
Sequence-based reagentBcl7a forwardThis paperMeRIP-qPCR primersccaagaagaacctagagcgg
Sequence-based reagentBcl7a reverseThis paperMeRIP-qPCR primersggcagtcacttgaaggttcg
Sequence-based reagentRap2b forwardThis paperMeRIP-qPCR primersggacttggagggtgaacgtga
Sequence-based reagentRap2b reverseThis paperMeRIP-qPCR primersgatgtctccatgaaggggcag
Sequence-based reagentNxt1 forwardThis paperMeRIP-qPCR primersccagtaacacggtgtggaag
Sequence-based reagentNxt1 reverseThis paperMeRIP-qPCR primersgagactggcatttctctgcag
Sequence-based reagentApex1 forwardThis paperMeRIP-qPCR primersgttgggatgaagccttcc
Sequence-based reagentApex1 reverseThis paperMeRIP-qPCR primerscatgagccacattgagatcc
Sequence-based reagentSulf2 forwardThis paperMeRIP-qPCR primersctttggagaaagcacggac
Sequence-based reagentSulf2 reverseThis paperMeRIP-qPCR primerscttctgagccagccaggtc
OtherMettl3GenBankCGTCAGTATCTTGGGCAAATTshRNA targeting site
OtherZfp292GenBankTGTGGCAGTAAGCCATATATAshRNA targeting site
OtherCkap4GenBankACGACCTGGATAGGTTGTTTCshRNA targeting site
OtherTraf4GenBankGCGTATAGTTCCCACTAATTTshRNA targeting site
OtherBcl7aGenBankCGGAGCCAAAGGTTGATGATAshRNA targeting site

Animals

All animal studies were performed in accordance with the protocol approved by the Institutional Animal Care and Use Committee of Zhongshan Ophthalmic Center (protocol number: Z2022011), and all animals were housed in the animal care facility of Zhongshan Ophthalmic Center Six3-Cre mice were kindly provided by Dr. Furuta from the University of Texas (Furuta et al., 2000). Mettl3floxed mice were kindly provided by Dr. Tong Minghan from Shanghai Institute of Biochemistry and Cell Biology, Chinese Academy of Sciences (Lin et al., 2017). Rlbp1-GFP mice were kindly provided by Edward M. Levine from the University of Utah (Vázquez-Chona et al., 2009).

Histology, immunostaining, and images

Request a detailed protocol

For cryosectioning, eyes were fixed overnight in 4% formaldehyde, cryopreserved with 15% and then 30% sucrose, frozen in Tissue-Tek OCT freezing medium, and sectioned using a cryostat microtome (Leica CM1950). For paraffin sectioning, eyes were fixed with Davidson’s Fixative overnight, processed, embedded in paraffin, and sectioned using a microtome (Leica RM2235). Before subsequent histological staining, paraffin sections underwent a standard procedure for deparaffinization, and cryosections were air-dried for 5 min and then soaked in PBS for 5 min to wash off OCT. H&E staining was performed using a standard procedure. For fluorescent immunostaining, section slides were first heated in citrate buffer at 95°C for 30 min for antigen retrieval. After cooling to room temperature, slides were incubated with primary antibodies prepared in PBST with 5% fetal bovine serum at 4°C overnight, washed with PBST, incubated with fluorophore-conjugated secondary antibodies, washed, counterstained with 4′,6-diamidino-2-phenylindole (DAPI), and finally mounted with cover glasses using VECTASHIELD mounting medium (Vector Labs). The following antibodies were used: rabbit anti-METTL3 (Abcam), rabbit anti-RECOVERIN (Millipore), sheep anti-VSX2 (Millipore), rabbit anti-SOX9 (Millipore), rabbit anti-CALBINDIN (Swant), HPC-1 (mouse anti-SYNTAXIN) (Sigma), mouse anti-BRN3A (Millipore), chicken anti-GFP (Abcam), rabbit anti-N-CADHERIN (Santa Cruz), mouse anti-GS (BD Transduction Laboratories), mouse anti-GFAP (Sigma), mouse anti-BrdU (Amersham), rabbit anti-KI67 (Abcam), mouse anti-pH3 (Ser10) (Cell Signaling Technology), Alexa Fluor 488 (568) donkey anti-mouse (Thermo Fisher), Alexa Fluor 488 (568) donkey anti-rabbit (Thermo Fisher), Alexa Fluor 488 donkey anti-chicken (Jackson ImmunoResearch), and Alexa Fluor 488 (568) donkey anti-sheep (Thermo Fisher). Bright-field H&E-stained images were taken under an inverted microscope (Zeiss Axio Observer). Immunofluorescence images were taken under a confocal microscope (Zeiss LSM 880 or LSM 980).

BrdU labeling

Request a detailed protocol

Pups were injected subcutaneously with BrdU (0.32 μmol/g body weight). Then 2 hr, or 6 hr, or 48 hr later, pups were sacrificed, and eyes were collected, fixed with 4% formaldehyde overnight, and subjected to cryosectioning and immunofluorescent staining as described above. After confocal images were taken, the number of BrdU+ cells was counted for each selected 200 μm retinal region.

TUNEL assay

Request a detailed protocol

Eye sections were prepared as described above for immunostaining. TUNEL assays were performed using an In Situ Cell Death Detection Kit (Roche) following the manufacturer’s instructions.

mfERG

Request a detailed protocol

The experiments were performed on p14 control (n = 4) and Mettl3-CKO (n = 2) mice. The mice were anesthetized using sodium pentobarbital and positioned on a warming table to maintain body temperature. For each animal, only the right eye was examined. Before recordings started, the eyes were dilated with 0.5% tropicamide (Santen Pharmaceutical, Japan) and local narcotized with 0.5% proparacaine hydrochloride (Ruinian Best, China). The eyes were positioned 1–2 mm in front of the device (Roland RETImap SLO, Roland, Germany). The animal Goldring recording electrode was placed at the corneal limbus. Subcutaneous silver needle electrodes served as reference and ground electrodes. An optical correction lens was positioned in front of the recording electrode. Viscous 2% methocel gel was applied between the cornea and contact lens. The stimuli were generated by a projector with a refresh rate of 60 Hz. The number of recording hexagons was set at 19, and the optic nerve head (ONH) was adjusted near the edge of the recording area. The stimulation parameter of light was set at mfERG Max. Twelve cycles were averaged for a final result.

RNA-seq

Request a detailed protocol

RNA-seq was performed on four control and four CKO retinal samples. The retinas of p6 control and CKO pups were collected directly into TRIzol (Thermo Fisher), and total RNA was extracted following the manufacturer’s instructions. Sequencing libraries were generated using the NEBNext Ultra RNA Library Prep Kit for Illumina (NEB) following the manufacturer’s instructions. The libraries were sequenced on an Illumina HiSeq platform, and 150 bp paired-end reads were generated. Raw data were filtered to remove low-quality reads, and the clean reads were mapped to the mouse genome (GRCm38, mm10) using Hisat2 v2.0.5. featureCounts v1.5.0-p3 was used to count the read numbers mapped to each gene, and then the FPKM of each gene was calculated. Differential expression analysis was performed using the DESeq2 R package. The resulting p-values were adjusted using Benjamini and Hochberg’s approach. Genes with an adjusted p-value<0.05 were considered differentially expressed.

scRNA-seq

Request a detailed protocol

Two retinas from two control pups and two retinas from two Mettl3-CKO pups of the same litter were used for the scRNA-seq experiment. Retinas were digested with Papain (Worthington) to single cells and resuspended in PBS containing 0.04% BSA, and two retinal samples for each experimental group were mixed together as one sample for subsequent scRNA-seq. The scRNA-seq was performed by NovelBio Co., Ltd. Libraries were generated using the 10X Genomics Chromium Controller Instrument and Chromium Single Cell 3′ V3.1 Reagent Kits (10X Genomics). Libraries were sequenced using an Illumina sequencer (Illumina) on a 150 bp paired-end run.

We applied fastp with default parameter filtering of the adaptor sequence and removed the low-quality reads to obtain clean data. Then, the feature-barcode matrices were obtained by aligning reads to the mouse genome (GRCm38, mm10) using CellRanger v3.1.0. Cells containing over 200 expressed genes and a mitochondrial UMI rate below 20% passed the quality filtering, and mitochondrial genes were removed from the expression table. The Seurat package v3.1 was used for cell normalization and regression to obtain the scaled data. PCA was constructed based on the scaled data with the top 2000 highly variable genes, and the top 10 principal components were used for UMAP construction. Utilizing the graph-based cluster method, we acquired unsupervised cell clusters. Judged by the key markers expressed by each cluster, we merged some clusters and annotated the final clusters.

For pseudotime analysis, we applied single-cell trajectory analysis using Monocle2 using DDR-Tree and default parameters. To identify DEGs among samples, the function FindMarkers with the Wilcox rank-sum test algorithm was used under the following criteria: lnFC > 0.25, p-value<0.05, min.pct > 0.1. To characterize the relative activity of gene sets with different m6A features, we performed QuSAGE (2.16.1) analysis. To characterize the activities of RPC or Müller glial feature gene sets, R package AUCell was used. Fisher’s exact test was applied to identify the significant GO terms and FDR was used to correct the p-values.

MeRIP-seq

Request a detailed protocol

The retinas of C57BL/6J mice were collected directly into TRIzol (Thermo Fisher). Total RNA was extracted according to the manufacturer’s instructions, and RNAs from 14 retinas were pooled together for each MeRIP-seq. MeRIP-seq was performed according to a published protocol (Dominissini et al., 2013). Briefly, PolyA-mRNA was purified using a Dynabeads mRNA DIRECT Purification kit (Thermo Fisher). mRNA was fragmented in fragmentation buffer heated at 94℃ for 5 min. Fragmented mRNA was immunoprecipitated against an m6A antibody (Synaptic Systems) bound to Protein G Dynabeads (Thermo Fisher) at 4°C for 2 hr, washed, and eluted by competition with free m6A (Sigma). After recovery of precipitated m6A-mRNA, sequencing libraries were constructed using the VAHTS Universal V6 RNA-seq Library Prep Kit (Vazyme). The libraries were sequenced on an Illumina HiSeq platform on a 150 bp paired-end run. Raw data were filtered to remove low-quality reads to obtain clean reads. Clean reads were mapped to the mouse genome (GRCm38, mm10) using BWA mem v0.7.12. m6A peaks were determined using MACS2 v2.1.0 with the q-value threshold of enrichment set at 0.05. Motif analysis was done using MEME. GO term analysis was performed using clusterProfiler package (Yu et al., 2012).

MeRIP-qPCR

Request a detailed protocol

Total RNA from retinas (5–6 retinas per sample) was extracted using TRIzol. MeRIP for qPCR was performed as described above for MeRIP-seq except that the mRNA fragmentation step was skipped. m6A pull-down and 1% reserved input RNAs were reversed transcribed into cDNAs using the SuperScript II First-strand Synthesis System for RT-PCR (Thermo Fisher). qPCR was performed using the SYBR green-based method on a LightCycler 480 (Roche). The primers are listed in the key resources table. Enrichment of genes in m6A-tagged genes in pull-down samples was calculated based on ΔCt between the pull-down sample and the input sample, and was normalized to an internal negative control, Rsp14, which does not carry m6A modification.

Molecular cloning

Request a detailed protocol

Full-length Zfp292, Ckap4, Traf4, Bcl7a, Rap2b, Nxt1, Apex, and Sulf2 cDNAs were PCR amplified from a cDNA pool derived from embryonic mouse retina total RNA and cloned into the pCAGIG vector (gift from Dr. Connie Cepko and obtained through Addgene). shRNAs targeting Zfp292, Ckap4, Traf4, Bcl7a, and Mettl3 were cloned into pSicoR vector (gift from Dr. Tyler Jacks and obtained through Addgene).

Mouse retina in vivo electroporation

Request a detailed protocol

Newborn mouse pups were anesthetized by chilling on ice. A small incision was made in the eyelid and sclera near the edge of the cornea with a 30-gauge needle (Hamilton). Purified plasmid solutions (2–5 μg/μl) in PBS were injected into the subretinal space through the previous incision using a Hamilton syringe with a 32-gauge blunt-ended needle under a dissecting microscope. After DNA injection, a tweezer-type electrode was placed to softly hold the pup heads, and five 80 V pulses of 50 ms duration and 950 ms intervals were applied using a Gemini X2 pulse generator (BTX). After electroporation, pups were allowed to recover on a warming pad until they regained consciousness and were returned to their mother. The pups were sacrificed at p14, and the retinas were collected and subjected to cryosection as described above. Eye sections were immunostained for GFP (to recognize transduced cells and their progenies) and SOX9 (to recognize Müller cells). For cell composition analysis of the electroporated cells, GFP+ cells in the ONL were judged as rods, GFP+; SOX9+ cells were judged as Müller cells, the GFP+ cells in the INL above the SOX9+ Müller cells (close to the OPL layer) were judged as bipolar cells, while the GFP+ cells in the INL below the SOX9+ Müller cells (close to the IPL layer) were judged as amacrine cells.

RNA half-life measurement

Request a detailed protocol

Mettl3-CKO and control mice were sacrificed at p1, and the retinas were collected and soaked in DMEM supplemented with actinomycin D (5 μm, Sigma) to inhibit transcription and cultured in a CO2 incubator at 37℃. The retinas were collected directly into TRIzoL at 0 hr, 4 hr, 8 hr, and 16 hr after drug treatment. Total RNA was extracted according to the manufacturer’s instructions. Two micrograms of total RNA from each sample was reverse transcribed into cDNA using SuperScript II (Thermo Fisher) and subjected to qPCR measurements using SYBR Green mix (Roche) and LightCycler LC480 (Roche). The primers are listed in the key resources table. The relative mRNA concentrations at each time point against time point 0 were calculated, the lnmRNA concentration at time points 0, 4, 8, and 16 hr were plotted to perform a linear regression analysis as a function of time, and the slope of the line was identified as the decay rate (k). The half-life was calculated with the following formula: t (1/2) = ln2/k (Chen et al., 2008).

Statistical analysis

Request a detailed protocol

All experiments subjected to statistical analysis were performed with at least three biological replicates. Quantification data are presented as the mean ± standard deviation, and the significance of differences between samples was assessed using Student’s t-test.

Data availability

Sequencing data have been deposited in GEO under accession codes GSE180815, GSE181095, and GSE188607.

The following data sets were generated
    1. Chen S
    (2021) NCBI Gene Expression Omnibus
    ID GSE180815. Mettl3 regulates late-stage retinogenesis and mouse retina m6A epitranscriptome.
    1. Chen S
    (2021) NCBI Gene Expression Omnibus
    ID GSE188607. Neonatal and adult mouse retina m6A epitranscriptome [meRIP-seq].

References

    1. Bokar JA
    2. Shambaugh ME
    3. Polayes D
    4. Matera AG
    5. Rottman FM
    (1997)
    Purification and cdna cloning of the adomet-binding subunit of the human mrna (N6-adenosine) -methyltransferase
    RNA 3:1233–1247.

Decision letter

  1. Fadel Tissir
    Reviewing Editor; Université Catholique de Louvain, Belgium
  2. Marianne E Bronner
    Senior Editor; California Institute of Technology, United States

Our editorial process produces two outputs: (i) public reviews designed to be posted alongside the preprint for the benefit of readers; (ii) feedback on the manuscript for the authors, including requests for revisions, shown below. We also include an acceptance summary that explains what the editors found interesting or important about the work.

Decision letter after peer review:

Thank you for submitting your article "m6A epitranscriptomic modification regulates neural progenitor-to-glial cell transition in the retina" for consideration by eLife. Your article has been reviewed by 2 peer reviewers, and the evaluation has been overseen by a Reviewing Editor and Marianne Bronner as the Senior Editor. The reviewers have opted to remain anonymous.

The reviewers have discussed their reviews with one another, and the Reviewing Editor has drafted this to help you prepare a revised submission.

Essential revisions:

1) The authors showed disrupted retina structure in Mettl3cko and claimed that dysfunction of Muller glial cells was the underlying cause. However, based on the data presented in the manuscript, it is not clear whether retinal structure abnormalities are direct effects of Muller glial defects or secondary effects induced by defects in RPCs or other cell types.

2) The authors claimed that m6A is important for RPC to Muller glial transition but didn't clearly state whether m6A regulates gliogenesis, which can be easily assessed by quantifying the number of Muller glial cells. The authors showed that the number of Muller glial cells was increased in mettl3cko at P7 based on Sox9 staining, but the number of Muller glial cells in single cell RNA-seq data stayed unchanged between Mettl3cko and controls. In addition, the number of Muller glial cells in adult retinas (for example, at P14) was not examined. Based on images in Figure 1E, the number of Muller glial cells seemed to be slightly decreased compared to controls. Moreover, in line 130, the authors mentioned that Mettl3cko retinas resembled those with loss of Muller glia in the literature. These discrepancies need to be addressed.

3) The authors showed that there was an increased number of RPCs in Mettl3cko at late stages of retinal development compared to controls. However, some of the data do not seem to be consistent with published literature and their own supplementary data. Specifically, at P7, based on their single cell RNA-seq data, 6% of retinal cells were RPCs in control retinas, and 11% of retinal cells were RPCs in Mettl3cko (Figure 3—figure supplement 2). However, the number seems to be very high for control P7 retinas. In published single cell datasets (Clark et al. 2019), the percentage of retinal cells identified as RPCs was close to 0 at around P7. In addition, in Figure 3—figure supplement 3, the authors stained P7 Mettl3cko and control retinas with Ki67. There was a very limited number of Ki67+ cells in Mettl3cko, which doesn't quite match the 11% found in single cell RNA-seq data.

4) The stage of the retina was not consistent across experiments. The single cell RNA-seq experiment was performed using P7 retinas, whereas MeRIP-seq used P6 retinas. Twenty-four hours could make a big difference in terms of transcription at this stage.

5) Many claims in the manuscript are not fully supported by the data. For example, the authors over-expressed candidate m6A-regulated genes in RPCs and showed that they prevented the RPCs from exiting the cell cycle. However, the data presented in Figure 6 cannot fully support this conclusion. PCNA is not a pan cell cycle marker. Reduction of GFP+PCNA- cells in Mettl3cko doesn't necessarily mean that mutant cells failed to exit the cell cycle.

6) Individual data points and N numbers were not shown in the bar graphs.

7) The authors used six3:cre (expressed from E9.5) to knock out Mettl3 throughout the entire course of retinal development. It would be better if the authors could test the role of m6A in late RPCs directly.

8) In Figure 2, the authors identified müller glia in Mettl3CKO mice using the Sox9 antibody. Due to severe structural abnormality of the retina in Mettl3CKO mice, the identification of cell fate needs additional caution. It is better to verify müller glia using other antibodies (such as glutamine synthetase, GS; GFAP) independently.

9) Some results seem inconsistent. First, in Mettl3CKO mice, the author found an increased number of Müller glia, a reduced number of rods, and no change in other cell types (Figure 2I-N). However, in the retinae with the overexpression of Zfp292, Ckap4, Traf4, or Bcl7a, they found an increase in the percentage of Müller cells at the expense of amacrine cells or rod photoreceptors (Figure 6C-D). Second, in scRNAseq data analysis, only the proportion of RPCs increased, but not Müller glial cells (Figure 3D). This result is not consistent with the IF phenotype (Figure 2I-J). The authors need to discuss these results at least to clarify the inconsistency.

10) Overall, the working model that the authors proposed lacks direct support. The authors may want to identify late RPC clusters that were committed to Müller glia fate as well as were subject to Mttl3CKO, and further verify the expression specificity of m6A-modified transcripts, such as Zfp292, Ckap4, Traf4, or Bcl7a, in such RPC clusters. Also, the authors should experimentally examine the essential roles of the loss of these m6A-modified transcripts in Müller glia development. These analyses will much strengthen the main conclusions of the study.

Reviewer #1 (Recommendations for the authors):

I encourage the authors to include the following info and additional explanations in the manuscript.

1. Line 120, the stage of IF staining should be included.

2. Line 146, please provide references showing that P6 is the peak time point for MG generation.

3. Line 162, the number of BrdU+;Ph3+ RPCs over BrdU+ cells should be quantified. The data presented by the authors cannot support the conclusion.

4. Line 180, please provide evidence showing that the major retinogenesis phase was completed in the central retinas in Mettl3CKO mice.

7. Line 183, how did the authors quantify Rho+ rods based on rhodopsin staining?

8. Line 226, "Considering that RPCs at this age are near the end of retinogenesis, most would become Müller glia. As a consequence, the final proportion of Müller glia in Mettl3CKO retinas should be higher than that in control retinas".

Muller glial proportion was not increased based on single cell RNA-seq data. This claim is confusing.

9. Line 258, "The upregulated genes were enriched for biological processes related to gene expression regulation, while downregulated genes were enriched for biological processes such as 'organelle organization' and 'precursor metabolites and energy'".

These GO terms do not seem to be the top terms based on the figures. Why did the author highlight them?

10. Line 259, "Importantly, key cell cycle machinery components, such as Ki67 and Brca2, was upregulated, but many cell cycle regulators, such as Ccna2 and Cdca8, were downregulated (Figure 3H-3J and Supplementary 261 Table 1), which explained why the cell cycle progression of Mettl3CKO RPCs was distorted".

This claim is very confusing. Does downregulation of Ccna2 and Cdca8 delay cell cycle?

11. Line 386, "p14 showed that while the OLMs of these retinas were grossly intact, at a few regions, the OLMs are broken, and a few rods escaped to the subretinal space".

This may be due to damage from retina sectioning. Please state whether this is a significant phenomenon. In addition, the authors did not quantify the number of Muller glial cells when the four plasmid mix was electroporated into P1 RPCs.

12. Figure 4 and Figure 5 may be combined.

Reviewer #2 (Recommendations for the authors):

1. The authors used six3:cre (expressed from E9.5) to knock out Mettl3 throughout the entire course of retinal development. It would be better if the authors could test the role of m6A in late RPCs directly.

2. In Figure 2, the authors identified müller glia in Mettl3CKO mice using the Sox9 antibody. Due to severe structural abnormality of the retina in Mettl3CKO mice, the identification of cell fate needs additional caution. It is better to verify müller glia using other antibodies (such as glutamine synthetase, GS; GFAP) independently.

3. Some results seem inconsistent. First, in Mettl3CKO mice, the author found an increased number of Müller glia, a reduced number of rods, and no change in other cell types (Figure 2I-N). However, in the retinae with the overexpression of Zfp292, Ckap4, Traf4, or Bcl7a, they found an increase in the percentage of Müller cells at the expense of amacrine cells or rod photoreceptors (Figure 6C-D). Second, in scRNAseq data analysis, only the proportion of RPCs increased, but not Müller glial cells (Figure 3D). This result is not consistent with the IF phenotype (Figure 2I-J). The authors need to discuss these results at least to clarify the inconsistency.

4. Overall, the working model that the authors proposed lacks direct support. The authors may want to identify late RPC clusters that were committed to Müller glia fate as well as were subject to Mttl3CKO, and further verify the expression specificity of m6A-modified transcripts, such as Zfp292, Ckap4, Traf4, or Bcl7a, in such RPC clusters. Also, the authors should experimentally examine the essential roles of the loss of these m6A-modified transcripts in Müller glia development. These analyses will much strengthen the main conclusions of the study.

[Editors’ note: further revisions were suggested prior to acceptance, as described below.]

Thank you for resubmitting your work entitled "m6A epitranscriptomic modification regulates neural progenitor-to-glial cell transition in the retina" for further consideration by eLife. Your revised article has been evaluated by Marianne Bronner (Senior Editor) and a Reviewing Editor.

The manuscript has been improved but there are some remaining issues that need to be addressed, as outlined below:

While the reviewers appreciate the efforts that have been invested in this revised manuscript, they feel that an additional revision is needed to put the study into the context of previous report, and clarify the novelty of the findings.

Reviewer #1 (Recommendations for the authors):

The authors improved their manuscript. However, to make the study convincing, additional investigation and explanation still need to be conducted.

1. It is still not clear whether m6A depletion specifically affects RPC to muller glia (MG) transition, or just elongates the RPC cell cycle as shown in previously published papers, which investigated the role of m6A in brain development. Elongation of RPC cell cycle can also lead to increased production of MG.

2. The authors showed that there are significant cell death in the cko retinas at P0, P6 and P14, and this may be responsible for the not so significant increase of MG in cko. But they didn't show which types of cells are dying. Many TUNEL positive cells are in ONL (Figure 2-supplement 4, A'-C'). This cell death was also not carefully integrated into their conclusions when they explained the phenotypes or results.

3. The authors claimed that cko RPCs withdrew from the cell cycle shower than control RPCs (Figure 2E and F). If this is the case, there should be more proliferating RPCs in cko. However, the author showed that there are significantly less BrdU+ cells at P1 (2 hours after BrdU injection) in cko (Figure 2B). These results do not seem to support each other.

4. The authors didn't show that Mettl3 is depleted and m6A levels are lowered at embryonic stages.

5. For all the shRNA-based experiments, there aren't any control experiments to show their efficiency and specificity.

6. Lin 254-258, the argument is weak and hard to understand. The Ki67+ region is only at the very tip of the retina in cko. The authors need to provide stronger evidence to clarify whether the number of muller glial cells is increased or unchanged in the cko at different stages.

7. The authors claim that Mettl3cko "…compromises the function of Muller cells" (abstract line 32). There is no direct evidence to support this point.

8. The role of m6A on neural development has been extensively studied in the brain. The authors did not discuss these published papers and did not explain how their work improved the field.

Other concerns:

1. In Figure 1A and 1A', the background color looked very different. It looks like different exposure duration were used. A western blot may be better to show the deletion efficiency.

2. Penetrance (50%) of the OLM break phenotype was not included in the manuscript.

3. Line 161, 2 hour Brdu pulse chase experiment cannot support that the cells proliferate slower. It only suggests that there are less proliferating cells.

4. Figure 2D, label on Y axis was not corrected.

5. Line 201-206, the number of cells cannot be quantified based on markers expressed in the cytosol. These markers may be upregulated in individual cells.

6. The m6A IP was not shown for cko retinas in Figure 6-Supplement 1.

Reviewer #2 (Recommendations for the authors):

The authors have enlisted new experiments and analyses that substantially strengthen the paper in this revised manuscript. All these efforts should be applauded. In particular, the new data of shRNA targeting Mettl3 expressed in late RPCs directly addressed the cell-type specificity of Mettl3's role. Also, the detailed analysis of scRNA data of late RPCs further clarified the transition from late RPCs to the muller glial cells. Furthermore, the authors examined the influences of m6A-associated transcripts on muller glia development by electroporating shRNAs targeting these transcripts into P1 retinae. Besides, additional discussions in the revised manuscript help to clarify the concerns the reviewers raised. In all, all these improvements are satisfying. However, I am surprised that the authors did not provide the background of previous studies of m6A in retinal development. Adding these research backgrounds will further improve the clarity of this study by putting it into context.

https://doi.org/10.7554/eLife.79994.sa1

Author response

Essential revisions:

1) The authors showed disrupted retina structure in Mettl3cko and claimed that dysfunction of Muller glial cells was the underlying cause. However, based on the data presented in the manuscript, it is not clear whether retinal structure abnormalities are direct effects of Muller glial defects or secondary effects induced by defects in RPCs or other cell types.

We acknowledge that our study did not directly investigate what happens to the retina if Mettl3 is specifically knocked out from Müller glial cells and their RPC precursors; we lack the necessary tools. However, several lines of evidence indicate that dysfunctional Müller glial cells are the key cells responsible for the structural abnormality of Mettl3CKO retinas:

1. As the sole neural glial cells inside the retina, Müller cells play key roles in maintaining the structural and physiological homeostasis of the mature retina, as demonstrated in numerous studies (MacDonald R et al., Curr Opin Neurobiol. 2017, 47:31-37; Shen W, et al., J Neurosci. 2012, 32(45):15715-27; Byrne L et al., PLoS One. 2013, 8(9):e76075; Wohl S et al., Nat Commun. 2017, 8(1): 1603).

2. The structural abnormality in Mettl3CKO retinas did not appear until p6, coinciding with the time point at which maturing Müller glial cells are transitioning from the simple bipolar radial shape inherited from RPCs to a much more sophisticated morphology; this morphology allows them to form a glial network and intimately interact with every cell in the mature retina (Wang J et al., J Comp Neurol. 2017, 525(8):1759-1777).

3. Our scRNA-seq analyses demonstrated that Müller glia were one of the two cell types whose transcriptomes were most severely affected by Mettl3 knockout (the other being RPCs), and that many genes involved in cell-cell interactions and the cytoskeleton were downregulated in Mettl3CKO Müller glial cells, suggesting that the structural supporting function of Müller glial cells was compromised.

On the other hand, we did not exclude the involvement of other cell types, especially RPCs, in the structural defects. Indeed, we think that RPCs indirectly contributed to the structural abnormality: the transcriptome abnormality of Mettl3CKO Müller cells was rooted in the failure of RPC transcripts, especially m6A-tagged RPC transcripts, to be promptly downregulated in Mettl3CKO RPCs during RPC-to-Müller transition. The failure likely disrupted the balance of the gene-regulatory-network of the developing Müller glia and compromised the glial supporting function of Müller cells, leading to the structural abnormality of the retina.

We have included a discussion of the above in the Discussion section of the revised manuscript.

2) The authors claimed that m6A is important for RPC to Muller glial transition but didn't clearly state whether m6A regulates gliogenesis, which can be easily assessed by quantifying the number of Muller glial cells. The authors showed that the number of Muller glial cells was increased in mettl3cko at P7 based on Sox9 staining, but the number of Muller glial cells in single cell RNA-seq data stayed unchanged between Mettl3cko and controls. In addition, the number of Muller glial cells in adult retinas (for example, at P14) was not examined. Based on images in Figure 1E, the number of Muller glial cells seemed to be slightly decreased compared to controls. Moreover, in line 130, the authors mentioned that Mettl3cko retinas resembled those with loss of Muller glia in the literature. These discrepancies need to be addressed.

Regarding the number of Müller glial cells, at p7, more Müller glial cells were generated in the central retinas of Mettl3CKO mice than in those of control mice based on our manual counting of SOX9-stained retinal sections (We recounted these images, but the conclusion remains the same. Please see our response to comment #8). Then, why did the scRNA-seq analysis show that the percentage of Müller glial cells in Mettl3CKO retinas was comparable to that in control retinas at p7? This result is due to the different sampling strategies adopted in the manual image counting and scRNA-seq experiments: for manual counting, we selected the central region of the retina, while for scRNA-seq, we collected the whole retina. At p7, the remaining KI67+ retinogenesis region was larger in Mettl3CKO retinas than in control retinas (revised Figure 3—figure supplement 3). In other words, the region of the retina that had finished Müller gliogenesis was smaller in Mettl3CKO retinas than in control retinas. Thus, the elevated Müller gliogenesis phenotype was masked by the smaller region that had finished the process in the scRNA-seq data.

Regarding the population of Müller glial cells at p14, the reviewer is correct; the increased population of Müller cells in Mettl3CKO retinas observed at p7 was no longer obvious at p14. The reason for the declining population of Müller glial cells is the elevated cell death in Mettl3CKO retinas. Cell death was significantly elevated in Mettl3CKO retinas, including Müller glial cells (revised Figure 2—figure supplement 4), which led to the degeneration of all cell types in the retina, including Müller glial cells.

Regarding the term ‘gliogenesis’, our data clearly showed that more Müller glial cells were generated by Mettl3CKO RPCs than by control RPCs. However, we did not use the term ‘gliogenesis’ for two reasons: first, even though more Müller glial cells were generated by Mettl3CKO RPCs, the onset of Müller gliogenesis was delayed for Mettl3CKO RPCs; second, even though more Müller glial cells were generated by Mettl3CKO RPCs, the transcriptome of these cells was distorted with a certain level of RPC feature (revised Figure 5H), which likely compromised their glial function. For these reasons, we hesitate to say ‘m6A promotes gliogenesis’.

Regarding line 130, we have referenced the literature to emphasize that Müller glial cells play key roles in maintaining the structural homeostasis of the retina. In Mettl3CKO retinas, even though more Müller glial cells were generated than the number observed in control retinas, these cells are dysfunctional; thus, the retinas exhibited a phenotype similar to that of retinas in which Müller glial cells are depleted. We have revised the sentence for clarity.

3) The authors showed that there was an increased number of RPCs in Mettl3cko at late stages of retinal development compared to controls. However, some of the data do not seem to be consistent with published literature and their own supplementary data. Specifically, at P7, based on their single cell RNA-seq data, 6% of retinal cells were RPCs in control retinas, and 11% of retinal cells were RPCs in Mettl3cko (Figure 3—figure supplement 2). However, the number seems to be very high for control P7 retinas. In published single cell datasets (Clark et al. 2019), the percentage of retinal cells identified as RPCs was close to 0 at around P7. In addition, in Figure 3—figure supplement 3, the authors stained P7 Mettl3cko and control retinas with Ki67. There was a very limited number of Ki67+ cells in Mettl3cko, which doesn't quite match the 11% found in single cell RNA-seq data.

To address the reviewer’s concern, we applied the annotation criteria of Clark et al.’s 2019 Neuron paper to annotate our cells to determine how many cells would be annotated as late RPC. We used the R package SingleR and used p5 and p8 cells from the Neuron paper as reference cells to annotate our cells from control retinas. Please see the Author response image 1: 2.74% of our control cells were annotated as late RPCs.

Author response image 1

We then pulled out data from Supplementary Table S5 of the Neuron paper and calculated the retinal cell type compositions in their p5 and p8 retinas. Please see Author response image 2: in the Neuron paper, at p5, 12.04% of retinal cells were annotated as late RPCs; at p8, 0.1% of retinal cells were annotated as late RPCs. They did not sequence p7 retinas.

Author response image 2

Thus, for our p7 retinal cells, the percentage of cells annotated as RPCs in our paper (5.37%) (rounded to 6% by Excel, as shown in Supplementary Figure 8 in our original manuscript) is smaller than that annotated by SingleR based on the annotation standard in the Neuron paper (2.74%). However, after carefully inspecting the data, we feel that, at least for our p7 retinal cells, our annotation more closely reflects the real situation:1. For the other types of cells in p5, p8 and p14 retinas described in the Neuron paper, a few obviously abnormal numbers are evident (please see the Author response image 2): At p5, 20.04% of cells were annotated as Müller glial cells; the percentage fell to 0.48% at p8 but then increased to 17.86% at p14; the percentage of amacrine cells was 11.37% at p5, 0.26% at p8, and 0.14% at p14; and the percentage of bipolar cells was 1.18% at p5, 19.55% at p8, and 20.34% at p14. None of these numbers accurately reflect the real populations of cell types at the various time points, suggesting that caution should be taken when using these numbers.

2. We checked the cell cycle status of the cells annotated as “Müller Glia” in Clark et al.’s p5 retinas. Of 1087 cells, 68 were in G2/M phase, and 138 were in S phase; thus, approximately 19% of these cells were actually cycling RPCs, making the RPC percentage 15.9%.

3. Similarly, we checked the cell cycle status of our p7 retinal cells annotated as “Photoreceptor Precursor” by SingleR; 24% of them were in S phase, indicating that they were actually cycling RPCs.

4. The cell cycle status of the cells annotated as RPCs in our paper is either S phase or G2/M phase, demonstrating that these cells are proliferating RPCs, while most cells annotated as Müller cells were in G1 phase (Figure 5—figure supplement 1B).

5. Regarding the Mki67+ cells, as shown in Figure 5—figure supplement 1B-1D, only portion of proliferating RPC cells (mostly G1/M-phase cells) are Mki67+ cells. We calculated the percentages of Mki67+ cells in our scRNA-Seq data, which were 3.99% in control retinas and 7.93% in Mettl3CKO retinas.

6. To further address the reviewer’s concern about the discrepancy between the scRNA-seq data and the KI67 staining in our original Supplementary Figure 3, we performed KI67 staining on a few more p7 retinal samples and replaced the original images with images showing the whole retina. The abundances of KI67+ cells in these images roughly match the percentages calculated from the scRNA-seq data mentioned above.

Based on the above evidence, we feel that, at least for our p7 retinal cells, our annotation is close to the real situation.

4) The stage of the retina was not consistent across experiments. The single cell RNA-seq experiment was performed using P7 retinas, whereas MeRIP-seq used P6 retinas. Twenty-four hours could make a big difference in terms of transcription at this stage.

We performed p6 retina MeRIP-seq before scRNA-seq. Because the m6A epitranscriptome of the mouse retina is relatively stable during postnatal development (Figure 4A, and Figure 4—figure supplementary file 1 and file 2), we used p6 retina MeRIP-seq data in the scRNA-seq and MeRIP-seq integrative analyses in original manuscript. To address the reviewer’s concern, in the course of revising this manuscript, we collected p7 retinas and performed MeRIP-seq on them. Comparing the p7 MeRIP-seq results and p6 MeRIP-seq results showed that the two data sets were quite similar to each other (please see Supplementary Table 3 in the original manuscript and Figure 4—figure supplementary file 1 in the revised manuscript). We have replaced the p6 MeRIP-seq data with the p7 MeRIP-seq data in the revised manuscript.

5) Many claims in the manuscript are not fully supported by the data. For example, the authors over-expressed candidate m6A-regulated genes in RPCs and showed that they prevented the RPCs from exiting the cell cycle. However, the data presented in Figure 6 cannot fully support this conclusion. PCNA is not a pan cell cycle marker. Reduction of GFP+PCNA- cells in Mettl3cko doesn't necessarily mean that mutant cells failed to exit the cell cycle.

We agree with the reviewer that PCNA is not a pan cell cycle marker. To address the reviewer’s concern, we repeated these experiments with KI67. The results showed that, similar to the PCNA results, there were fewer GFP+; KI67- cells among RPCs overexpressing Zfp292/Ckap4/Traf4/Bcl7a than in control RPCs. Although KI67 is not a pan cell cycle marker either (please see our response to comment #3), since both the PCNA- (S phase) and Ki67- (G2/M phase) cell populations were reduced in RPCs overexpressing Zfp292/Ckap4/Traf4/Bcl7a, the data strongly suggest that overexpressing Zfp292/Ckap4/Traf4/Bcl7a prevented RPCs from exiting the cell cycle.

Furthermore, another phenomenon indicates that fewer RPCs overexpressing Zfp292/Ckap4/Traf4/Bcl7a exited the cell cycle: after exiting the cell cycle, many differentiating retinal cells migrate away from the central region of the retinoblast layer to their designated position in the mature retina. For example, differentiating/maturing amacrine cells move toward the vitreous side of the retinoblast layer. In the control samples, we observed many GFP+ cells located near the vitreous edge of the retinoblast layer, away from Ki67- or PCNA- positive zones, indicating that they were differentiating/maturing amacrine cells (arrows in revised Figure 6C-Ctrl). In contrast, much fewer such cells were observed in retinas overexpressing Zfp292/Ckap4/Traf4/Bcl7a, suggesting that withdrawal from the cell cycle was delayed in these retina.

We have replaced the PCNA data in Figure 6 with the KI67 data and presented the PCNA data in Figure 6—figure supplement 2. We have also added a description of the tissue distribution patterns of GFP+ cells to the revised manuscript.

6) Individual data points and N numbers were not shown in the bar graphs.

We have added individual data points to the bar graphs. Since the N value is clearly reflected by the data points, we have not added N values to the bar graphs.

7) The authors used six3:cre (expressed from E9.5) to knock out Mettl3 throughout the entire course of retinal development. It would be better if the authors could test the role of m6A in late RPCs directly.

We thank the reviewer for the suggestion.

To test the role of m6A in late RPCs directly, we designed and synthesize plasmids expressing shRNAs targeting Mettl3 and introduced these shRNAs into late RPCs through in vivo electroporation on retinas of p1 pups. We examined the cell proliferation status of the electroporated cells at p5. The results showed that, in the retinas electroporated with plasmids expressing shRNAs targeting Mettl3, significantly more GFP marked electroporated cells were labeled with 2-hour BrdU treatment than the retinas electroporated with the empty control vector (Figure 2—figure supplement 2), suggesting that cell cycle exit of the late RPCs in the experimental group was delayed. In addition, in the retinas electroporated with plasmids expressing shRNAs targeting Mettl3, fewer GFP marked electroporated cells had moved to the vitreous side of the retinoblast layer compared with the control retinas, indicating that the differentiation of Mettl3 deficient late RPCs was delayed. Thus, these experiments demonstrated that Mettl3 is required in late RPCs for their proper retinogenesis activities.

8) In Figure 2, the authors identified müller glia in Mettl3CKO mice using the Sox9 antibody. Due to severe structural abnormality of the retina in Mettl3CKO mice, the identification of cell fate needs additional caution. It is better to verify müller glia using other antibodies (such as glutamine synthetase, GS; GFAP) independently.

We agree with the reviewer that SOX9 is not an ideal marker for Müller glial cells, especially for structurally distorted Mettl3CKO retinas. However, other Müller glial cell markers, such as GS and GFAP, have their own problems. GFAP is normally expressed in astrocyte but not Müller glial cells unless they are stimulated by injuries to the retina. GS is a cytoplasmic protein and presents challenges for counting Müller cells, which have complex cellular processes; moreover, at p7, GS is just starting to be expressed in Müller glial cells; the weak expression of GS in Müller glial cells (especially in Mettl3CKO retinas) at this time makes it unsuitable as a counting marker for Müller glial cells. To address the reviewer’s concern, we recounted SOX9+ cells in p7 retinas but only considered those located in the INL+OPL+ONL region (i.e., excluding those in the IPL+GCL region, which might be astrocytes) as Müller glial cells. The results showed that there were still more SOX9+ Müller glial cells in Mettl3CKO retinas than in control retinas (Revised Figure 2J). Thus, our original conclusion that more Müller glial cells were generated by Mettl3CKO RPCs than by control RPCs still stands.

9) Some results seem inconsistent. First, in Mettl3CKO mice, the author found an increased number of Müller glia, a reduced number of rods, and no change in other cell types (Figure 2I-N). However, in the retinae with the overexpression of Zfp292, Ckap4, Traf4, or Bcl7a, they found an increase in the percentage of Müller cells at the expense of amacrine cells or rod photoreceptors (Figure 6C-D). Second, in scRNAseq data analysis, only the proportion of RPCs increased, but not Müller glial cells (Figure 3D). This result is not consistent with the IF phenotype (Figure 2I-J). The authors need to discuss these results at least to clarify the inconsistency.

Regarding the reviewer’s concern about the discrepancy in Müller percentage numbers obtained from scRNA-seq analysis and IF image counting, please see our response to comment #2.

The following points address the reviewer’s concern that retinas overexpressing Zfp292/Ckap4/Traf4/Bcl7a did not exhibit exact the same cell composition changes observed for Mettl3CKO retinas: First, in Mettl3CKO RPCs, over 200 m6A-tagged genes were upregulated. It is likely that the change in retinal cell type composition in Mettl3CKO retinas was the collective result of expression changes of many of these m6A-tagged genes, and the overexpression of one (or a few) of them may not have given rise to the exact same phenotype as that of Mettl3CKO retinas. Second, forced overexpression of Zfp292/Ckap4/Traf4/Bcl7a from plasmids electroporated into RPCs (with CAG used as the promoter) probably increased the expression levels of these genes by several to tens of times, while in Mettl3CKO RPCs, the expression levels of these genes were only moderately elevated compared with those of control RPCs (Figure 3—figure supplementary file 1). Different levels of overexpression may elicit different changes in gene-regulatory-networks and thus yield different end results.

We have added the above information to the Discussion section of the revised manuscript.

10) Overall, the working model that the authors proposed lacks direct support. The authors may want to identify late RPC clusters that were committed to Müller glia fate as well as were subject to Mttl3CKO, and further verify the expression specificity of m6A-modified transcripts, such as Zfp292, Ckap4, Traf4, or Bcl7a, in such RPC clusters. Also, the authors should experimentally examine the essential roles of the loss of these m6A-modified transcripts in Müller glia development. These analyses will much strengthen the main conclusions of the study.

We thank the reviewer for the valuable suggestion.

To identify a putative RPC cluster that was committed to the Müller glia fate, we pulled out the data on RPC cells and Müller glial cells from the scRNA-Seq data (green and blue cells in Figure 3A) and reperformed clustering analyses, which separated the cells into 9 clusters. Clusters 0, 1, 3, and 4 were composed of cells originally annotated as Müller glial cells, while clusters 2, 5, 6, 7, and 8 were composed of cells originally annotated as RPCs (revised Figure 5D and Figure 5—figure supplement 1A). We then used AUCell and early RPC and Müller cell signature genes (Lin et al. Invest Ophthalmol Vis Sci. 2019, 60 (13): 4436-4450) to score the RPC and Müller features in each cell. Both the RPC feature scoring and the Müller feature scoring showed that RPC clusters 2 and 7 and Müller cell cluster 4 laid at the junction between RPCs and Müller cells (revised Figure 5F). We thus considered RPC clusters 2 and 7 as putative RPC clusters committed to the Müller glia fate and called them ‘RPCMüller’, and we considered Müller cell cluster 4 as the cluster that had just exited the cell cycle and become Müller cells and called them ‘MüllerEarly’. Our reclustering did not identify distinct mutant clusters corresponding to clusters 2, 4, and 7. Based on the expression patterns of markers for clusters 4 and 7, we classified subsets of cells of clusters 1 and 6 as corresponding mutant clusters for clusters 4 and 7 (‘MüllerEarly-mut’ and ‘RPCMüller-mut’), respectively. Based on the pure CKO sample origin, a subset of cluster 2 was considered the corresponding mutant cluster for cluster 2 (‘RPCMüller-mut’) (revised Figure 5G). Thus, by further analyses of RPCs and Müller cells, we identified an RPC cluster committed to the Müller glial fate and a Müller cluster at an early stage of differentiation, as well as their corresponding mutant clusters. We then used AUCell to score the expression levels of m6A-tagged RPC genes in the newly annotated cell clusters. The results showed that the expression levels of m6A-tagged RPC genes in mutant clusters were significantly elevated to a level close to that in the control cluster of an earlier state (Figure 5H). Thus, these analyses are consistent with our previous analyses (original Figure 5E), and strengthen our hypothesis that m6A promotes the transcriptomic transition from RPCs to Müller glial cells by promoting the degradation of modified RPC transcripts. We have added the above new analyses to the revised manuscript.

To examine the effects of the loss of these m6A-modified transcripts on Müller glia development, we designed and synthesized plasmids expressing shRNAs targeting Zfp292/Ckap4/Traf4/Bcl7a and electroporated these plasmids into p1 retinas. Late RPCs deficient for Bcl7a generated fewer Müller glial cells (rods and amacrine cells were slightly increased but did not show statistical significance), suggesting that Bcl7a promotes Müller glial cell development; while knocking down the other three genes did not significantly affect retinogenesis. Taken together, the gene knockdown and overexpression experiments suggest that the expression of these genes in late RPCs should normally be kept at low levels and carefully controlled. We have added the above new data to the revised manuscript.

Regarding the expression specificity of Zfp292, Ckap4, Traf4, and Bcl7a, AUC scoring of the expression levels of these genes in the above set of cell clusters showed that the expression levels of these genes were significantly elevated in all mutant clusters compared with the corresponding control clusters; however, within the control clusters, most of these genes did not show obvious RPC-specific expression (Figure 6—figure supplement 1A). We selected these m6A-regulated genes as RPC-specific genes based on our RNA-Seq comparison of sorted mature Müller glial cells and E12.5 retinas (Lin et al. Invest Ophthalmol Vis Sci. 2019, 60 (13): 4436-4450). Consistent with our RNA-seq comparison, analyzing Early RPC, Late RPC and Müller glia cells of Clark’s Neuron paper showed that these genes were more abundantly expressed in early RPCs, and often also so in late RPCs (please see Author response image 3). The lack of specificity of these genes in late RPCs at p7 might be due to the fact that, at this time point, the expression levels of these genes in this end-stage late RPCs were reduced to levels close to those in Müller glial cells. Because of the above new information, we feel that it is not accurate to call these genes late RPC-specific genes, but the results of overexpression and knockdown experiments targeting these genes were still in line with the conclusion that ‘m6A plays essential roles during late retinogenesis by promoting the degradation of m6A-regulated RPC transcripts’. We have revised the paragraph accordingly.

Author response image 3

Reviewer #1 (Recommendations for the authors):

I encourage the authors to include the following info and additional explanations in the manuscript.

1. Line 120, the stage of IF staining should be included.

These experiments were performed on the retinas of p15 pups. We have added the information to the revised manuscript.

2. Line 146, please provide references showing that P6 is the peak time point for MG generation.

We intended to say ‘the time when morphological maturation of Müller glia starts’. The sentence has been revised and the corresponding reference has been added.

3. Line 162, the number of BrdU+;Ph3+ RPCs over BrdU+ cells should be quantified. The data presented by the authors cannot support the conclusion.

We have replaced Figure 2D with the quantification of BrdU+; pH3+ over BrdU+ percentages.

4. Line 180, please provide evidence showing that the major retinogenesis phase was completed in the central retinas in Mettl3CKO mice.

Please see Figure 3—figure supplement 3 in the revised manuscript, which showed that there were only a few KI67+ cells in the central retina of Mettl3CKO mice, suggesting that the major retinogenesis phase has passed in this region. We have replaced ‘completed’ with ‘passed’ in the revised manuscript.

7. Line 183, how did the authors quantify Rho+ rods based on rhodopsin staining?

We acknowledge that a cytoplasmic marker like RHODOPSIN is not ideal for counting cells, especially cells in the compacted ONL of the retina, and cones might be miss-counted as rods. We thus have revised the sentence to generally call RHODDPSIN+ cells as photoreceptors.

8. Line 226, "Considering that RPCs at this age are near the end of retinogenesis, most would become Müller glia. As a consequence, the final proportion of Müller glia in Mettl3CKO retinas should be higher than that in control retinas".

Muller glial proportion was not increased based on single cell RNA-seq data. This claim is confusing.

We have revised the paragraph. Please see our response to comment #2.

9. Line 258, "The upregulated genes were enriched for biological processes related to gene expression regulation, while downregulated genes were enriched for biological processes such as 'organelle organization' and 'precursor metabolites and energy'".

These GO terms do not seem to be the top terms based on the figures. Why did the author highlight them?

Among the 10 GO terms, 3 were related to organelles (organelle fission, organelle localization, organelle organization), 4 were related to molecule metabolism (precursor metabolites and energy, ATP metabolic process, ribonucleotide metabolic process, pyruvate metabolic process), and the remainder were related to the cell cycle. Our intent was to select one ‘organelle’ term and one ‘metabolism’ term as examples. We have revised the sentence.

10. Line 259, "Importantly, key cell cycle machinery components, such as Ki67 and Brca2, was upregulated, but many cell cycle regulators, such as Ccna2 and Cdca8, were downregulated (Figure 3H-3J and Supplementary 261 Table 1), which explained why the cell cycle progression of Mettl3CKO RPCs was distorted".

This claim is very confusing. Does downregulation of Ccna2 and Cdca8 delay cell cycle?

Ccna2 plays an important role in regulating G1/S and G2/M transitions during the cell cycle; Cdca8 plays an important role in regulating mitosis. Thus, downregulation of these genes might be part of the molecular mechanism underlying the extended S-to-M phase progression observed for Mettl3CKO RPCs. We have revised the sentence.

11. Line 386, "p14 showed that while the OLMs of these retinas were grossly intact, at a few regions, the OLMs are broken, and a few rods escaped to the subretinal space".

This may be due to damage from retina sectioning. Please state whether this is a significant phenomenon. In addition, the authors did not quantify the number of Muller glial cells when the four plasmid mix was electroporated into P1 RPCs.

We do not think the broken OLM in Figure 6E of the original manuscript was caused by tissue sectioning. Tissue sectioning could cause broken OLM, but in such situations, we do not see photoreceptors leak out as those shown in the DAPI channel insert image. However, we acknowledge that the penetrance of this phenotype was not high (3 out 6 retinas). We thus have moved these images to the supplementary figure and revised the text.

We have added the cell type composition quantification results of the retinas electroporated with the four plasmids mix in the revised manuscript (Figure 6—figure supplement 2C).

12. Figure 4 and Figure 5 may be combined.

In the revised manuscript, we have added new analyses to Figure 5 and moved some of the panels of Figure 5 to Figure 6. We feel it is better to keep Figure 4 and Figure 5 separate.

[Editors’ note: further revisions were suggested prior to acceptance, as described below.]

Reviewer #1 (Recommendations for the authors):

The authors improved their manuscript. However, to make the study convincing, additional investigation and explanation still need to be conducted.

Major concerns:

1. It is still not clear whether m6A depletion specifically affects RPC to muller glia (MG) transition, or just elongates the RPC cell cycle as shown in previously published papers, which investigated the role of m6A in brain development. Elongation of RPC cell cycle can also lead to increased production of MG.

In Figure 5, we showed that Mettl3-mutant Müller cells retained certain features of RPCs, and the m6A-tagged RPC transcripts failed to be promptly downregulated during the RPC-to-Müller transition, demonstrating that m6A depletion directly regulates the RPC-to-Müller transition.

2. The authors showed that there are significant cell death in the cko retinas at P0, P6 and P14, and this may be responsible for the not so significant increase of MG in cko. But they didn't show which types of cells are dying. Many TUNEL positive cells are in ONL (Figure 2-supplement 4, A'-C'). This cell death was also not carefully integrated into their conclusions when they explained the phenotypes or results.

We performed costaining of retinal cell type markers with TUNEL, which showed that both RPCs and their differentiated progenies were dying. This led to the degeneration of the retina at older ages and might have contributed to the reduced population of photoreceptors at p7. We have added the above information to the revised version of the manuscript (Figure 2—figure supplement 4E-4G).

The elevations in cell death events would affect the cell composition of the retina, which is exactly why we chose p7 to count cells in the central retina, when the region had just finished retinogenesis, to keep the influence of the cell death events minimal. Why retinal cells die in Mettl3-CKO retinas is an interesting question that is worth future study. But, we believe further analyses on cell death will not affect the conclusions of this manuscript.

3. The authors claimed that cko RPCs withdrew from the cell cycle shower than control RPCs (Figure 2E and F). If this is the case, there should be more proliferating RPCs in cko. However, the author showed that there are significantly less BrdU+ cells at P1 (2 hours after BrdU injection) in cko (Figure 2B). These results do not seem to support each other.

Elevations in cell death activity (Figure 2—figure supplementary 4A’, 4D, and 4E) might reduce the population of RPCs in CKO retinas.

4. The authors didn't show that Mettl3 is depleted and m6A levels are lowered at embryonic stages.

We stained E15.5 retinas with a METTL3 antibody, which showed that Mettl3 was efficiently deleted from the central retina (revised Figure2—figure supplement 1). Since no obvious phenotype was observed at this stage, we did not measure m6A level changes (which would require large amounts of tissues, meaning that large amounts of CKO and control embryos would need to be collected).

5. For all the shRNA-based experiments, there aren't any control experiments to show their efficiency and specificity.

We assessed the gene knockdown efficiencies of the shRNAs in mouse embryonic fibroblasts (MEFs): we generated shRNA-expressing lentiviruses (the cloning vector pSicoR is a lentivirus vector); used these viruses to infect MEFs; harvested MEFs two days after infection; and performed RT-qPCR (empty vector virus was used as control). We have added this information to the revised version of the manuscript (Figure6—figure supplement 2E).

6. Lin 254-258, the argument is weak and hard to understand. The Ki67+ region is only at the very tip of the retina in cko. The authors need to provide stronger evidence to clarify whether the number of muller glial cells is increased or unchanged in the cko at different stages.

Our manual counting of p7 retinas clearly demonstrated that more Müller cells were generated by CKO-RPCs. We have explained in detail why we chose p7 central retinas (see response to comment #2 above), how we counted (see response to comment #8 in the previous round of revision), why the increase in the population of Müller cells was not obvious at p14 (see response to comment #2 in the previous round of revision), and why the manual counting and scRNA-Seq results were seemingly inconsistent (see line 254-258, and the response to comment #2 in the previous round of revision). As we demonstrated in our response to comment #3 in our previous round of revision, bioinformatic analysis of scRNA-Seq data has its caveats and may not faithfully annotate every cell, but the general trends are usually consistent with the real situation in the tissue, which is the case here. We feel we have clearly demonstrated the point that more Müller cells were generated by CKO-RPCs.

7. The authors claim that Mettl3cko "…compromises the function of Muller cells" (abstract line 32). There is no direct evidence to support this point.

As we explained in our response to comment #1 in our previous round of revision, we acknowledge that our study did not directly test the role of Mettl3 specifically in Müller cells, but several lines of our data indicate that CKO-Müller cells were dysfunctional. We have revised the sentence to “…likely compromises the glial functions of Müller cells.”.

8. The role of m6A on neural development has been extensively studied in the brain. The authors did not discuss these published papers and did not explain how their work improved the field.

Thank you for the suggestion.

In the Discussion section of the revised manuscript, we compare our study with two published studies investigating the roles of m6A during cortical development. Please see the last paragraph of the Discussion section of the revised manuscript.

Other concerns:

1. In Figure 1A and 1A', the background color looked very different. It looks like different exposure duration were used. A western blot may be better to show the deletion efficiency.

We feel that our METTL3 immunostaining results clearly demonstrated that Mettl3 was efficiently knocked out in the central region of the CKO retina at spatial and single-cell resolution (Figure 1A’), which is a better and more meaningful way than western blotting to demonstrate the deletion effect.

2. Penetrance (50%) of the OLM break phenotype was not included in the manuscript.

The information has been added to the revised manuscript.

3. Line 161, 2 hour Brdu pulse chase experiment cannot support that the cells proliferate slower. It only suggests that there are less proliferating cells.

We have revised the sentence to “…indicating that there were fewer proliferating RPCs in Mettl3-CKO retinas.”.

4. Figure 2D, label on Y axis was not corrected.

The error has been corrected.

5. Line 201-206, the number of cells cannot be quantified based on markers expressed in the cytosol. These markers may be upregulated in individual cells.

We agree with the reviewer that nuclear markers are better than cytosolic markers in terms of cell counting, but we think both can be used to count cells. We currently do not have suitable nuclear markers on hand for these experiments. The results may not be absolutely precise, but we believe they very closely approximate the real situation.

6. The m6A IP was not shown for cko retinas in Figure 6-Supplement 1.

We are not sure which panel this comment is referring to. If the reviewer is referring to panel B, we have the following response: Because MeRIP-Seq is not a good way to quantify differential methylation levels between samples (Mclntyre A, et al., Scientific Reports (2020), 10:590), we did not perform MeRIP-Seq for CKO retinas.

Reviewer #2 (Recommendations for the authors):

The authors have enlisted new experiments and analyses that substantially strengthen the paper in this revised manuscript. All these efforts should be applauded. In particular, the new data of shRNA targeting Mettl3 expressed in late RPCs directly addressed the cell-type specificity of Mettl3's role. Also, the detailed analysis of scRNA data of late RPCs further clarified the transition from late RPCs to the muller glial cells. Furthermore, the authors examined the influences of m6A-associated transcripts on muller glia development by electroporating shRNAs targeting these transcripts into P1 retinae. Besides, additional discussions in the revised manuscript help to clarify the concerns the reviewers raised. In all, all these improvements are satisfying. However, I am surprised that the authors did not provide the background of previous studies of m6A in retinal development. Adding these research backgrounds will further improve the clarity of this study by putting it into context.

We are glad that the reviewer is satisfied with our revision efforts and thank you for the suggestion.

Before we started our Mettl3-CKO mouse project, we tested the function of m6A during vertebrate retinal development by knocking down m6A-writer components in zebrafish using morpholinos. The study was published last year in BBRC. In addition, during the time when this manuscript was under revision, A paper studying the role of m6A-readers during retinal development was published. We have discussed these works in the introduction or the Discussion sections in the revised manuscript.

https://doi.org/10.7554/eLife.79994.sa2

Article and author information

Author details

  1. Yanling Xin

    State Key Laboratory of Ophthalmology, Zhongshan Ophthalmic Center, Sun Yat-sen University, Guangdong Provincial Key Laboratory of Ophthalmology and Visual Science, Guangzhou, China
    Contribution
    Formal analysis, Investigation
    Competing interests
    No competing interests declared
  2. Qinghai He

    State Key Laboratory of Ophthalmology, Zhongshan Ophthalmic Center, Sun Yat-sen University, Guangdong Provincial Key Laboratory of Ophthalmology and Visual Science, Guangzhou, China
    Contribution
    Investigation
    Competing interests
    No competing interests declared
  3. Huilin Liang

    State Key Laboratory of Ophthalmology, Zhongshan Ophthalmic Center, Sun Yat-sen University, Guangdong Provincial Key Laboratory of Ophthalmology and Visual Science, Guangzhou, China
    Contribution
    Investigation
    Competing interests
    No competing interests declared
  4. Ke Zhang

    State Key Laboratory of Ophthalmology, Zhongshan Ophthalmic Center, Sun Yat-sen University, Guangdong Provincial Key Laboratory of Ophthalmology and Visual Science, Guangzhou, China
    Contribution
    Investigation
    Competing interests
    No competing interests declared
  5. Jingyi Guo

    State Key Laboratory of Ophthalmology, Zhongshan Ophthalmic Center, Sun Yat-sen University, Guangdong Provincial Key Laboratory of Ophthalmology and Visual Science, Guangzhou, China
    Contribution
    Investigation
    Competing interests
    No competing interests declared
  6. Qi Zhong

    State Key Laboratory of Ophthalmology, Zhongshan Ophthalmic Center, Sun Yat-sen University, Guangdong Provincial Key Laboratory of Ophthalmology and Visual Science, Guangzhou, China
    Contribution
    Investigation
    Competing interests
    No competing interests declared
  7. Dan Chen

    State Key Laboratory of Ophthalmology, Zhongshan Ophthalmic Center, Sun Yat-sen University, Guangdong Provincial Key Laboratory of Ophthalmology and Visual Science, Guangzhou, China
    Contribution
    Investigation
    Competing interests
    No competing interests declared
  8. Jinyan Li

    State Key Laboratory of Ophthalmology, Zhongshan Ophthalmic Center, Sun Yat-sen University, Guangdong Provincial Key Laboratory of Ophthalmology and Visual Science, Guangzhou, China
    Contribution
    Investigation
    Competing interests
    No competing interests declared
  9. Yizhi Liu

    State Key Laboratory of Ophthalmology, Zhongshan Ophthalmic Center, Sun Yat-sen University, Guangdong Provincial Key Laboratory of Ophthalmology and Visual Science, Guangzhou, China
    Contribution
    Conceptualization
    Competing interests
    No competing interests declared
  10. Shuyi Chen

    State Key Laboratory of Ophthalmology, Zhongshan Ophthalmic Center, Sun Yat-sen University, Guangdong Provincial Key Laboratory of Ophthalmology and Visual Science, Guangzhou, China
    Contribution
    Conceptualization, Formal analysis, Supervision, Funding acquisition, Writing - original draft, Project administration
    For correspondence
    chenshy23@mail.sysu.edu.cn
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0003-3037-6853

Funding

National Natural Science Foundation of China (81870659)

  • Shuyi Chen

National Natural Science Foundation of China (81721003)

  • Yizhi Liu

The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.

Acknowledgements

We thank Dr. Minghan Tong, from Shanghai Institute of Biochemistry and Cell Biology, Chinese Academy of Sciences, for providing the Mettl3floxed mice, Dr. Edward M Levine, from the University of Utah, for providing the Rlbp1-GFP mice, Dr. Yasuhide Furata, from the University of Texas, for providing the Six3-Cre mice. We thank the staff of Laboratory Animal Center at State Key Laboratory of Ophthalmology, Zhongshan Ophthalmic Center for technical support. This study was supported by the National Natural Science Foundation of China (81870659 and 81721003).

Ethics

All animal studies were performed in accordance with the protocol approved by the Institutional Animal Care and Use Committee of Zhongshan Ophthalmic Center (protocol number: Z2022011) , and all animals were housed in the animal care facility of Zhongshan Ophthalmic Center.

Senior Editor

  1. Marianne E Bronner, California Institute of Technology, United States

Reviewing Editor

  1. Fadel Tissir, Université Catholique de Louvain, Belgium

Publication history

  1. Received: May 4, 2022
  2. Preprint posted: May 8, 2022 (view preprint)
  3. Accepted: November 13, 2022
  4. Version of Record published: December 2, 2022 (version 1)

Copyright

© 2022, Xin et al.

This article is distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use and redistribution provided that the original author and source are credited.

Metrics

  • 344
    Page views
  • 60
    Downloads
  • 1
    Citations

Article citation count generated by polling the highest count across the following sources: Crossref, PubMed Central, Scopus.

Download links

A two-part list of links to download the article, or parts of the article, in various formats.

Downloads (link to download the article as PDF)

Open citations (links to open the citations from this article in various online reference manager services)

Cite this article (links to download the citations from this article in formats compatible with various reference manager tools)

  1. Yanling Xin
  2. Qinghai He
  3. Huilin Liang
  4. Ke Zhang
  5. Jingyi Guo
  6. Qi Zhong
  7. Dan Chen
  8. Jinyan Li
  9. Yizhi Liu
  10. Shuyi Chen
(2022)
m6A epitranscriptomic modification regulates neural progenitor-to-glial cell transition in the retina
eLife 11:e79994.
https://doi.org/10.7554/eLife.79994

Further reading

    1. Developmental Biology
    Marianne E Emmert, Parul Aggarwal ... Roger Cornwall
    Research Article Updated

    Neonatal brachial plexus injury (NBPI) causes disabling and incurable muscle contractures that result from impaired longitudinal growth of denervated muscles. This deficit in muscle growth is driven by increased proteasome-mediated protein degradation, suggesting a dysregulation of muscle proteostasis. The myostatin (MSTN) pathway, a prominent muscle-specific regulator of proteostasis, is a putative signaling mechanism by which neonatal denervation could impair longitudinal muscle growth, and thus a potential target to prevent NBPI-induced contractures. Through a mouse model of NBPI, our present study revealed that pharmacologic inhibition of MSTN signaling induces hypertrophy, restores longitudinal growth, and prevents contractures in denervated muscles of female but not male mice, despite inducing hypertrophy of normally innervated muscles in both sexes. Additionally, the MSTN-dependent impairment of longitudinal muscle growth after NBPI in female mice is associated with perturbation of 20S proteasome activity, but not through alterations in canonical MSTN signaling pathways. These findings reveal a sex dimorphism in the regulation of neonatal longitudinal muscle growth and contractures, thereby providing insights into contracture pathophysiology, identifying a potential muscle-specific therapeutic target for contracture prevention, and underscoring the importance of sex as a biological variable in the pathophysiology of neuromuscular disorders.

    1. Developmental Biology
    2. Genetics and Genomics
    Ankit Sabharwal, Mark D Wishman ... Stephen C Ekker
    Research Advance Updated

    The clinical and largely unpredictable heterogeneity of phenotypes in patients with mitochondrial disorders demonstrates the ongoing challenges in the understanding of this semi-autonomous organelle in biology and disease. Previously, we used the gene-breaking transposon to create 1200 transgenic zebrafish strains tagging protein-coding genes (Ichino et al., 2020), including the lrpprc locus. Here, we present and characterize a new genetic revertible animal model that recapitulates components of Leigh Syndrome French Canadian Type (LSFC), a mitochondrial disorder that includes diagnostic liver dysfunction. LSFC is caused by allelic variations in the LRPPRC gene, involved in mitochondrial mRNA polyadenylation and translation. lrpprc zebrafish homozygous mutants displayed biochemical and mitochondrial phenotypes similar to clinical manifestations observed in patients, including dysfunction in lipid homeostasis. We were able to rescue these phenotypes in the disease model using a liver-specific genetic model therapy, functionally demonstrating a previously under-recognized critical role for the liver in the pathophysiology of this disease.