m6A is required for resolving progenitor identity during planarian stem cell differentiation

Abstract

Regeneration and tissue homeostasis require accurate production of missing cell lineages. Cell production is driven by changes to gene expression, which is shaped by multiple layers of regulation. Here, we find that the ubiquitous mRNA base-modification, m6A, is required for proper cell fate choice and cellular maturation in planarian stem cells (neoblasts). We mapped m6A-enriched regions in 7,600 planarian genes and found that perturbation of the m6A pathway resulted in progressive deterioration of tissues and death. Using single-cell RNA sequencing of >20,000 cells following perturbation of the m6A pathway, we identified an increase in expression of noncanonical histone variants, and that inhibition of the pathway resulted in accumulation of undifferentiated cells throughout the animal in an abnormal transcriptional state. Analysis of >1,000 planarian gene expression datasets revealed that the inhibition of the chromatin modifying complex NuRD had almost indistinguishable consequences, unraveling an unappreciated link between m6A and chromatin modifications. Our findings reveal that m6A is critical for planarian stem cell homeostasis and gene regulation in tissue maintenance and regeneration.

Keywords: differentiation; m6A; planarian; regeneration; stem cells.

Figure 1. Expression of m6A genes is required for planarian regeneration and viability

1. A

   Components of the m6A MTC (left) and readers (right) that were identified in the planarian genome (Fig EV1A and B; Materials and Methods).

2. B

   Effect of RNAi on planarian homeostasis and regeneration. Gene expression was inhibited by feeding the animals with dsRNA against the target gene (Materials and Methods). Phenotypes were observed in kiaa1429 (RNAi) animals following five dsRNA feedings in homeostasis ((i); 10/10 animals), in regeneration ((ii); 10/10 animals; shown 10 days post amputation), and following nine feedings of mettl14 (iii) or ythdc‐1 dsRNA feedings (iv). Matching controls did not show phenotypes (n = 10 per control group). Experiments were performed at least three times with over 10 animals per experiment in each experimental group. Scale = 1 mm.

3. C, D

   The relative abundance of m6A was estimated by using an anti‐m6A‐antibody (Materials and Methods), showing similar m6A levels compared to recent quantification of m6A in mammals (Liu et al, 2020b); (C) Quantification of m6A on RNA isolated from kiaa1429 (RNAi) and control animals following five RNAi feedings (left), or from mettl14 (RNAi) and control animals following nine RNAi feedings (right), revealed a strong reduction in m6A levels following inhibition of gene expression (Student's t‐test P < 0.05). (D) Comparison of m6A levels on polyadenylated and non‐polyadenylated RNA showed that polyadenylated RNA was highly enriched in m6A (Student's t‐test *P < 0.05).

Figure 2. Transcriptome‐wide mapping of m6A in planarians

1. Outline of mapping of m6A on planarian RNA. RNA was extracted from control or kiaa1429 (RNAi) animals in triplicates, with at least 10 animals in each replicate (i). Then, RNA was molecularly tagged with a barcoded RNA oligo (ii). Samples of barcoded RNA were mixed, and methylated RNA was enriched using an anti‐m6A‐antibody; input samples were sequenced without anti‐m6A‐antibody enrichment (iii). RNA was used for RNAseq library preparation, and resultant libraries were sequenced. Sequencing data was mapped to the planarian genome, and m6A‐enriched regions (peaks) were detected using MeTPeak (Cui et al, 2016); read coverage across Smed‐ast8 is shown as normalized RPKM that was scaled to 0–1 across samples (iv).

2. m6A peaks are highly enriched at the 3′ end of the transcript. X‐axis shows the relative position to the transcription start site (TSS); transcript lengths were normalized to 1,000 nt. Y‐axis is the average peak expression (normalized by RPKM) across 740 high‐confidence peaks (≥10 fold‐enriched). Plot produced using the deepTools2 package (Ramírez et al, 2016) (Materials and Methods).

3. The number of m6A peaks, which were determined using MetPeak (Cui et al, 2016), is shown (top) as a function of gene expression in each bin (bottom, box plot).

4. The distribution of the m6A peak length as detected by MeTPeak; peak length was limited to up to 1,000 nt for visualization. Black line represents the median peak length. Processed data for organisms other than planarian were obtained from REPIC (Liu et al, 2020c).

Figure 3. m6A pathway expression is not required for neoblast viability

1. A, B

   The number of neoblasts did not change significantly following inhibition of the MTC‐gene kiaa1429. Shown is FISH of smedwi‐1 (A) and h2b (B) in control and kiaa1429 (RNAi) animals (left). smedwi‐1 ⁺ or h2b ⁺ cells were counted following FISH and showed no significant difference in expression (right; bar represents average normalized cell count; Materials and Methods). Scale = 100 μm.

2. C, D

   Gene expression analysis of smedwi‐1 (C) and h2b (D). The expression of smedwi‐1 and h2b was determined in the different RNAi conditions compared to the expression of these genes in control animals (Materials and Methods). Dashed line marks a twofold change in expression. Measurements performed on biological triplicates. P‐value was calculated by two‐sided Student's t‐test followed by Bonferroni correction for multiple hypotheses. Lack of asterisk indicated a non‐significant change in gene expression. Error bars indicate the 95% confidence interval (***P < 0.001, **P < 0.01, *P < 0.05).

3. E

   Representative FACS plots of cells isolated from control (top) and kiaa1429 (RNAi) animals (bottom) show an increase in cell abundance in the X2 gate (green; Materials and Methods).

4. F

   Quantification of the cells in each FACS gate in control and kiaa1429 (RNAi) animals showed a sixfold increase in the abundance of cells in the X2 gate (Student's t‐test ***P < 0.005). FACS experiments were performed in biological triplicates. Boxes represent the IQR, whiskers represent the 1.5 × IQR, and central band represents the median.

Figure 4. Gene expression changes following inhibition of m6A genes

1. A, B

   Heatmap of genes that were differentially expressed following inhibition of m6A genes. Shown is the expression of the top 25 overexpressed and 25 underexpressed genes (top and bottom, respectively; Dataset EV2) following six mettl14 dsRNA feedings, at different conditions and time points. Rows and columns represent genes and samples, respectively. Blue and red, low to high gene expression (row‐normalized z‐score). Genes that were previously determined to be expressed in the intestine (Fincher et al, 2018) were highlighted in dark blue, or otherwise in dark red.

2. C

   Correlation of gene expression changes between mettl14 (RNAi) and ythdc‐1 (RNAi) compared to their controls. Each colored dot represents a gene (expression >1 transcripts per million, TPM; red and black, significant and nonsignificant change in gene expression compared to controls, respectively).

3. D

   The proportion of genes that were downregulated following RNAi and were found to be expressed in a single cell type. Shown is an analysis based on the top 300 genes that were downregulated following RNAi (six dsRNA feedings for mettl14 or ythdc‐1).

4. E

   FISH for detection of major planarian cell types following inhibition of kiaa1429 by RNAi using previously established cell‐type specific gene expression markers (Fincher et al, 2018). The distribution of most major planarian cell types was comparable between the control (top) and kiaa1429 (RNAi) animals (bottom). A striking exception was the reduction in intestine cells (right; see Fig EV4E–G for quantification), which severely affected the intestine branching morphology in kiaa1429 (RNAi) animals. Scale = 100 μm.

Figure 5. Overexpression of polyadenylated histone‐encoding transcripts

1. A, B

   Heatmap showing eight histone‐encoding genes that were upregulated following inhibition of m6A‐pathway‐encoding genes. Rows and columns represent genes and samples, respectively. Blue and red, low to high gene expression (row‐normalized z‐score).

2. C

   Boxplots showing the normalized number of reads that map to a histone gene body and are paired with a read pair containing a polyA (n = 3 biological replicates). Inhibition of kiaa1429 resulted in a dramatic increase in the mapping to polyA in histone‐encoding genes. Boxes represent the IQR, whiskers represent the 1.5 x IQR, and central band represents the median.

3. D

   Quantification of the entire h2b transcript pool including polyA⁺ and polyA⁻ transcripts using qPCR (Materials and Methods). No significant change in the total h2b expression was observed following inhibition of kiaa1429. Error bars represent the 95% confidence interval. Experiment performed in biological triplicates.

4. E

   Comparison of the polyA⁺ and polyA⁻ h2b transcripts showed an overabundance of polyA⁺ h2b in kiaa1429 (RNAi) animals compared with controls.

5. F, G

   Shown is a summary of gene mapping to a representative overexpressed gene neighborhood (contig: dd_Smes_g4_15). Arrows represent gene annotation, which was extracted from PlanMine (Rozanski et al, 2019). Panels show expression in TPM scaled to 0–1.

6. H

   The overexpression of two representative genes from the upregulated genomic neighborhoods was validated by FISH. Cells expressing these genes were found throughout the animal in sub‐epidermal and parenchymal layers (Left, Scale = 100 μm); higher magnification shows perinuclear expression (right, scale = 10 μm).

Figure 6. Cell type‐specific regulation of gene expression by m6A

1. A

   Outline of scRNAseq experiment. Cells were isolated from dissociated animals, purified by FACS and subjected to scRNAseq (top; Materials and Methods). Analysis of cell populations detected in the scRNAseq and visualized by UMAP (dot represents cell, and color represents cluster association). Comparison of scRNAseq of kiaa1429 (RNAi) and control revealed an expansion of a molecularly defined cell cluster (#20, black arrow).

2. B

   Relative representation of clusters between kiaa1429 (RNAi) and control animals. Most clusters were similarly represented between samples. However, depletion of phagocytic intestine cells, epidermal progenitors, and expansion of a kiaa1429 (RNAi) cell cluster was detected (*>90% difference in detectable the relative cell population size between conditions).

3. C, D

   Analysis of cells from the intestine lineage showed depletion in intestine neoblasts (gamma, #6), nkx‐2.2 ⁺ progenitors (#2), and differentiated intestine cells (#0, #1). Arrow indicates transition to differentiated cell state. Normalized cell count of intestine lineage cells are summarized (bottom).

4. E

   FISH using markers for different intestine cell types (Fincher et al, 2018) demonstrated changes to gene expression and morphological defects following inhibition of kiaa1429. Quantification of images found in Appendix Fig S3A. Scale = 100 μm.

5. F

   scRNAseq heatmap produced using Seurat (Stuart et al, 2019) of four representative genes from the kiaa1429 (RNAi)‐overexpressed genomic clusters. The major planarian cell types did not express these four genes. By contrast, a kiaa1429 (RNAi)‐specific cell cluster is characterized by high expression of these genes. Rows and columns represent genes and cells, respectively. Purple to yellow, low to high gene expression. Top row shows assignment of cells to cell types based on gene expression (Dataset EV3).

Figure 7. Inhibition of CHD4 recapitulates effects of the m6A‐pathway inhibition

1. Overexpressed gene neighborhood (contig: dd_Smes_g4_15:3.13–3.17 Mbp). Shown is the normalized and scaled gene expression in three CHD4 (RNAi) samples, 15 days post feedings (top, blue) and in three corresponding control samples (bottom, red).

2. Heatmap showing the eight histone‐encoding genes that were upregulated following inhibition of CHD4, at different time points, in published dataset (Tu et al, 2015). Rows and columns represent genes and samples, respectively. Blue and red, low to high gene expression (row normalized z‐score).

3. Quantification of the entire h2b transcript pool including polyA⁺ and polyA⁻ transcripts using qPCR (Materials and Methods). No significant change in the absolute expression of h2b inhibition of CHD4 (Bonferroni corrected Student's t‐test). Error bars represent the 95% confidence interval (At least two biological replicates, with two technical replicates for each biological replicate).

4. Comparison of the polyA⁺ and polyA⁻ h2b expression level shows an increase in polyA⁺ h2b in CHD4 (RNAi) animals compared to control animals.

5. Heatmap showing the gene expression levels of intestine‐specific TFs and genes following CHD4 RNAi at different time points. Rows and columns represent genes and samples, respectively. Blue and red, low and high gene expression (row normalized z‐score).

6. Quantification of m6A on RNA isolated from CHD4 (RNAi) or control animals showed no significant change in abundance of m6A methylation (Student's t‐test).

Figure 8. Potential models for interaction of m6A and NuRD

Shown are two models demonstrating the potential regulatory activity of m6A on neoblast differentiation.

1. m6A might regulate neoblast homeostasis by directly regulating NuRD or through a mediator that modulates NuRD function.

2. In the alternative model m6A and NuRD regulate similar processes, in parallel, without an association between the pathways.

Figure EV1. m6A genes are conserved and functional in planarians

1. A

   Conservation of the genes encoding the MTC is shown across diversity of organisms. Annotation for A. thaliana was obtained from a previous analysis (Růžička et al, 2017).

2. B

   Domain analysis of putative YTH‐domain (red block) containing genes.

3. C, D

   The gene expression of planarian MTC and reader‐encoding genes, across different cell types and lineages, was extracted from the planarian scRNAseq resource (Plass et al, 2018) (blue to red, low and high expression levels, respectively). The expression of MTC‐encoding genes is widespread in different cell types, including in the stem cell compartment (black arrow, indicated in the mettl3 panel). The expression of reader‐encoding genes is not widespread, except for ythdc‐1, which is expressed across multiple cell types and conditions.

4. E

   Inhibition of m6A genes resulted in lysis of the animals and eventually death. Shown are representative control and kiaa1429 (RNAi) animals. Scale = 1 mm.

5. F

   Inhibition of m6A genes resulted in defects in food uptake. Shown are animals following feeding with calf liver mixed with a red food color. Control animals show normal food uptake (top) compared with kiaa1429 (RNAi) animals, which stopped eating. Size measurements were performed on separate animals, which were subjected to the same RNAi treatment. Student's t‐test; ****P < 0.0001. Scale = 1 mm.

6. G

   Inhibition of mettl14 and ythdc‐1 by RNAi has resulted in size reduction. Shown are representative images (left) and measurement of animal sizes following nine RNAi feedings. Significance was calculated using one‐way ANOVA followed by Dunnett's multiple comparison test; ***P < 0.001; *P < 0.05. Scale = 1 mm.

7. H

   qPCR analysis showed that the gene expression of mettl3 (top) and wtap (bottom) is not downregulated significantly following RNAi, which could likely explain the lack of penetrant phenotypes in these conditions. Error bars indicate the 95% confidence interval. Samples include two technical replicates and at least two biological replicates.

8. I

   Schematic of the design of two non‐overlapping gene fragments that were used for synthesis of dsRNA targeting the kiaa1429 gene. The sequence used for the experiments presented in the manuscript was labeled “original kiaa1429 clone.”

9. J

   dsRNA that was produced using a second cloned fragment of kiaa1429, labeled “non‐overlapping kiaa1429 (RNAi),” produced phenotypes that were similar to the phenotypes observed in a non‐overlapping clone. This further indicated that the kiaa1429 (RNAi) phenotype resulted from inhibition of kiaa1429 gene expression and not because of an off‐target effect of the RNAi. Student's t‐test ***P < 0.001. Scale = 1 mm.

Figure EV2. m6A is abundant on planarian mRNA and depleted following kiaa1429 RNAi

1. The number of RNAseq reads for each m6A‐seq2 pulldown library per 10 M of the corresponding input reads is shown. The number of reads in each pulldown library in the m6A‐seq2 protocol correlates with the abundance of m6A (Dierks et al, 2021) (red and blue dot, control and kiaa1429 (RNAi) sample, respectively. ***Student's t‐test P < 0.001).

2. The number of m6A‐rich regions (peaks) was larger in control samples and depleted following inhibition of kiaa1429. Moreover, the fold‐enrichment of the peaks in the control samples over the gene expression observed in the input samples is greater, in comparison to the kiaa1429 (RNAi) samples.

3. A 2d‐density plot showing the correlation between m6A‐enriched regions in control and kiaa1429 (RNAi) animals. The density plot shows that m6a peaks were more highly enriched in the control samples compared to kiaa1429 (RNAi) samples, demonstrating the depletion of m6A following inhibition of kiaa1429.

4. Profile of m6A‐enriched regions across genes in control and kiaa1429 (RNAi) showed enrichment toward the 3′‐end. The length of transcripts with detectable m6A‐enriched regions was normalized to 1,000 nt. Then, the expression across the transcript was computed by generating bins of gene expression (Materials and Methods) and calculating the log‐fold change between the anti‐m6A‐antibody pulldown library and the input sample. K‐means was used to separate three profiles of m6A‐enriched regions.

5. Shown is the RPKM normalized expression of the planarian rRNA 28S (block arrow) in the m6A pulldown library and in control. There was no detectable m6A peak on planarian rRNA.

Figure EV3. Analysis of m6A peak length by species mixing experiment

1. A

   RNA was extracted (i) from control or (ii) kiaa1429 (RNAi) animals in duplicates, with at least 10 animals in each replicate, (iii) from meiosis‐blocked S. cerevisiae strain with high levels of m6A (Ndt80Δ/Δ) (Dierks et al, 2021), and (iv) from S. cerevisiae strain devoid of m6A, due to deletion of the methyltransferase (Ime4Δ/Δ/Ndt80Δ/Δ). RNA was combined into two pools (Materials and Methods), and was processed according to the m6A‐seq2 protocol (Dierks et al, 2021). Following sequencing, reads were associated with the original samples based on their 3′‐end barcode (red and blue, representing mix #1 and mix #2, respectively), and then based on their mapping to either the planarian transcriptome or the yeast genome (Materials and Methods).

2. B

   Violin plot comparing the length of m6A peaks in yeast from either the REPIC database (Liu et al, 2020c), or from the data collected in this experiment; m6A regions larger than 1,000 bp were excluded from the plot, horizontal line indicates the median length.

3. C

   Comparison of planarian m6A peaks in our initial m6A‐seq2 experiment, and in the m6a‐seq2 profiling in the species RNA mixing experiment. Shown are peaks shorter than 1,000 bp. To avoid comparison of lowly enriched peaks, shown are peaks with fold‐enrichments that are larger than the median fold‐enrichment.

4. D, E

   Shown is the most frequent sequence motif detected in our mixed m6A‐seq2 experiment (Materials and Methods), as detected by HOMER (Heinz et al, 2010) for yeast (D) and planarian (E).

5. F, G

   Shown are the fold change and associated P‐values for each k‐mer. DRACH‐like k‐mers are colored as in the figure legend. Data is shown for yeast (F; panels represent two biological replicates) and planarian (G; panels represent two biological replicates).

Figure EV4. Reduction in intestinal cells following inhibition of m6A genes

1. A, B

   Shown is H3P labeling using an anti‐H3P‐antibody in kiaa1429 (RNAi) and in control animals (scale = 100 and 10 μm, top and bottom panels, respectively; top panels show a maximal intensity projection of the H3P and DAPI signal; Materials and Methods). (B) The number of H3P⁺ cells was counted in a region anterior to the pharynx and posterior to the brain, in a z‐stack and was normalized to the area of the counting (Materials and Methods). There was an insignificant difference in the number of H3P⁺ cells between the conditions (Student's t‐test; n = 6 biological replicates). Boxes represent the IQR, whiskers represent the 1.5 × IQR, and central band represents the median.

2. C, D

   Shown are confocal images of kiaa1429 (RNAi) and control animals, which were soaked in F‐ara‐EdU for 16 h (scale = 100 and 10 μm, top and bottom panels; Materials and Methods). (D) The number of F‐ara‐EdU⁺ cells was counted in a region anterior of the pharynx and posterior to the brain and was normalized to the rectangle size (Materials and Methods). There was an insignificant difference in the number of F‐ara‐EdU⁺ cells between the tested conditions (Student's t‐test; n = 5 biological replicates). Boxes represent the IQR, whiskers represent the 1.5 × IQR, and central band represents the median.

3. E–G

   The effect of inhibition of m6A pathway‐encoding genes on the intestine integrity and cells was examined by FISH. (E, F) Inhibition of mettl14 (RNAi) or ythdc‐1 (RNAi) by 9 dsRNA feedings did not result in significant differences in the morphology of the intestine, as estimated by counting the number of intestine branches in confocal images (Materials and Methods). A significant difference in the number of intestine branches was observed following kiaa1429 (RNAi) following five RNAi feedings compared to its control. Scale = 100 μm. (E, G) A significant reduction in the number of intestine cells labeled by dd_3194 was found following either mettl14 (RNAi) or ythdc‐1 (RNAi) (two‐sided t‐test controlled by Bonferroni's correction). The dd_3194⁺ cells were counted in confocal images and the counting was normalized to the length of the intestine branches (Materials and Methods). *P < 0.05.

Figure EV5. Analysis of kiaa1429 (RNAi) phenotypes by scRNAseq

1. A–D

   Quality measurements of scRNAseq libraries prepared from control and kiaa1429 (RNAi) cells. The quality of the libraries produced from both conditions was assessed using the Seurat package (Stuart et al, 2019). The number of expressed genes (A); unique molecular tags (B); non‐coding gene expression (C); and the correlation between the number of expressed genes and the unique molecular tags were highly similar between the libraries (D). This indicated that the quality of libraries was comparable.

2. E

   Expression levels of canonical neoblast markers were highly similar in control and kiaa1429 (RNAi).

3. F

   The polyadenylated transcript expression of h2b and h3 was much higher in kiaa1429 (RNAi) animals compared to controls. The expression was detectable in neoblasts (bottom), and to a much lesser degree in the entire cell population (top), which includes neoblasts as well.

4. G, H

   UMAP representation of neoblasts (color dots, left panel), and their identity (right panel). Neoblast identity was assigned based on expression of previously published gene expression markers (Fincher et al, 2018). Most neoblast clusters were not affected by the RNAi, yet several lineages (right, asterisk) were differentially represented following kiaa1429 (RNAi).

5. I

   Expression of neoblast (smedwi‐1) and specialized intestine neoblast gene expression markers was overlaid on UMAP plots of the intestine lineage (cells represented by dots; gray and purple, low to high ranked expression).

6. J–L

   Comparison of gene expression between several cell clusters. Importantly, pharynx cells are post‐mitotic, and therefore show minimal smedwi‐1 expression (J). Post‐mitotic cells express xbp‐1, which was previously shown to be expressed in differentiating and differentiated cells (Raz et al, 2021) (K). SMAD6/7–2 is expressed in the kiaa1429 (RNAi) specific‐cluster and in neural cells (L).

7. M

   Highly expressed genes in the kiaa1429 (RNAi)‐specific cluster compared to control are shown in UMAP plots. Contig dd_1620 is highly expressed in the kiaa1429 (RNAi)‐specific cluster. The genes SMAD6/7–2 and protocadherin‐like (dd_15376, gene model SMESG000067388) are expressed in the kiaa1429 (RNAi)‐specific cluster (black arrowhead); both genes are also expressed in cells of the neural clusters (e.g., cluster 2).