METTL3 counteracts premature aging via m6A-dependent stabilization of MIS12 mRNA.

N6-Methyladenosine (m6A) messenger RNA methylation is a well-known epitranscriptional regulatory mechanism affecting central biological processes, but its function in human cellular senescence remains uninvestigated. Here, we found that levels of both m6A RNA methylation and the methyltransferase METTL3 were reduced in prematurely senescent human mesenchymal stem cell (hMSC) models of progeroid syndromes. Transcriptional profiling of m6A modifications further identified MIS12, for which m6A modifications were reduced in both prematurely senescent hMSCs and METTL3-deficient hMSCs. Knockout of METTL3 accelerated hMSC senescence whereas overexpression of METTL3 rescued the senescent phenotypes. Mechanistically, loss of m6A modifications accelerated the turnover and decreased the expression of MIS12 mRNA while knockout of MIS12 accelerated cellular senescence. Furthermore, m6A reader IGF2BP2 was identified as a key player in recognizing and stabilizing m6A-modified MIS12 mRNA. Taken together, we discovered that METTL3 alleviates hMSC senescence through m6A modification-dependent stabilization of the MIS12 transcript, representing a novel epitranscriptional mechanism in premature stem cell senescence.

INTRODUCTION

Reversible N6-methyladenosine (m6A) mRNA modifications, one of the most abundant epitranscriptomic modifications, regulate transcriptional, post-transcriptional and translational mechanisms (1–5). Accordingly, the m6A modification impinges on central biological processes such as stemness maintenance and differentiation, adipogenesis, and spermatogenesis, and its regulation impacts on pathological processes underlying obesity, infertility, neuronal disorders, aberrant immune signaling and cancers (6–10). Mechanistically, recent studies have revealed that m6A regulates the formation and stability of R-loops, and promotes the phase separation potential of methylated mRNAs (2,11,12), thus providing novel insights into the regulatory role of m6A modifications.

m6A RNA modifications are deposited by a core methyltransferase complex composed of METTL3 and METTL14 and removed by demethylases ALKBH5 and FTO (4,13). As the core catalytic component in the methyltransferase complex (14,15), loss of METTL3 disrupts numerous physiological processes such as progression of spermatogenesis, hematopoiesis and memory formation (16–19). In addition, METTL3 functions as an oncogene in many types of cancers, while its ablation inhibits tumorigenic cell proliferation by affecting m6A levels of target RNAs (20–22). The fate of m6A-modified RNAs is then determined by m6A readers, such as YTH family proteins, which affect their splicing, transport, stability, and translation (10,13). In addition, IGF2BP family members, including IGF2BP1, IGF2BP2, and IGF2BP3, may also function as m6A readers to enhance the stability of target mRNAs, thus affecting gene expression outputs as a consequence of normal and stress conditions (23). Overall, this body of work frames a complex regulatory network role for m6A in fundamental cell biological processes. However, the role of m6A RNA modifications in cellular senescence remains largely unknown.

Premature aging diseases, also named as progeroid syndromes, are rare human disorders characterized by accelerated features of normal human aging, such as hearing loss, alopecia, lipodystrophy, osteoporosis and atherosclerosis (24–28). Of these progerias, Hutchinson-Gilford progeria syndrome (HGPS) and Werner syndrome (WS) are two of the best characterized. HGPS is characterized by progerin accumulation caused by LMNA mutation, leading to dysmorphic nuclei, increased DNA damage, and loss of heterochromatin (29–31). WS is caused by mutations in the WRN gene that encodes a RecQ DNA helicase involved in DNA replication and DNA damage repair. Loss of WRN leads to genomic instability that contributes to accelerated cellular senescence (29,32–34). Dysfunction of transcription, translation, replication, and chromatin organization have been described during the premature aging process (32,35–40). In addition, epigenetic dysregulation has been widely characterized in premature aging, and epigenetic reprogramming has even been suggested as a promising strategy for aging rejuvenation (31,41,42), raising the question whether epitranscriptomic modifications, such as m6A RNA methylations, directly regulate premature aging.

Here, we employed isogenic human mesenchymal stem cells (hMSCs) with LMNA mutation and WRN knockout as HGPS and WS models (29), respectively, to explore the function of m6A RNA modifications in the regulation of premature stem cell aging. We found that decreased m6A modifications were associated with downregulated METTL3 expression, and that hMSCs with METTL3 knockout showed characteristics of accelerated senescence. Conversely, METTL3 overexpression increased m6A modifications and alleviated senescent phenotypes in HGPS and WS hMSCs. To identify specific targets affected by the loss of m6A modifications in premature aging, we conducted RNA sequencing (RNA-seq) and m6A methylated RNA immunoprecipitation sequencing (MeRIP-seq) analyses, and identified MIS12 as a key m6A effector, whose deficiency accelerated hMSC senescence. Mechanistically, we discovered that the m6A reader IGF2BP2 recognized and stabilized m6A-modified MIS12 mRNA to prevent hMSCs from accelerated senescence. Our findings provide novel insights into epitranscriptional mechanisms underpinning stem cell senescence via m6A modifications and identify a potential intervention strategy for the treatment of aging-associated disorders.

MATERIALS AND METHODS

Antibodies

Antibodies used for western blotting, dot blotting, immunostaining or RNA immunoprecipitation (RIP): anti-m6A (Synaptic Systems, 202003), anti-METTL3 (Proteintech, 15073-1-AP), anti-METTL14 (Sigma, HPA038002), anti-MIS12 (Abcam, ab70843), anti-IGF2BP2 (Bethyl Laboratories, A303–317A), anti-β-Tubulin (Santa Cruz, sc-5274), anti-GAPDH (Santa Cruz, sc-25778), anti-β-Actin (Santa Cruz, sc-69879), horseradish peroxidase (HRP)-conjugated secondary antibodies (ZSGB-BIO, ZB2305 and ZB2301), Alexa Fluor 488 donkey anti-mouse IgG (Invitrogen, A21202), Alexa Fluor 568 donkey anti-rabbit IgG (Invitrogen, A10042).

Cell culture

hMSC differentiation was performed as previously reported (29). All hMSCs were cultured in MSC culture medium: 90% α-MEM with GlutaMAX (Gibco), 10% fetal bovine serum (FBS, Gemcell, Cat# 100–500, Lot# A77E01F), 0.1 mM non-essential amino acids (NEAA, Gibco), 1% penicillin/streptomycin (Gibco) and 1 ng/ml FGF2 (Joint Protein Central).

Clonal expansion assay

The single-cell clonal expansion assay was carried out as previously described (29). Briefly, 2000 cells were seeded in a gelatin-coated 12-well plate (Corning Incorporated). About 10 days later, the cells were fixed with 4% paraformaldehyde for 30 min and stained with 0.2% crystal violet for 30 min at room temperature. Images were captured with an optical scanner (Hewlett-Packard). The relative cell density was determined using ImageJ.

Senescence-associated-β-galactosidase (SA-β-Gal) staining

SA-β-Gal staining was performed as previously described (29). Briefly, cultured cells were washed twice with PBS and fixed at room temperature for 5 min in 2% formaldehyde and 0.2% glutaraldehyde. Fixed cells were stained with SA-β-Gal staining solution (40 mM citric acid/Na phosphate buffer, 5 mM K4[Fe(CN)6], 5 mM K3[Fe(CN)6], 150 mM NaCl, 2 mM MgCl2, 1 mg/ml X-gal) at 37°C overnight, and images were captured using a microscope digital camera (Olympus). The percentage of SA-β-Gal-positive cells were calculated with ImageJ.

m6A dot blotting assay

Dot blotting assay was performed as previously reported (2,43). Briefly, total RNA was extracted using TRIzol reagent (Invitrogen). mRNA was prepared from total RNA using the Dynabeads mRNA Purification Kit (Life Technologies, 61006). 100–200 ng of mRNA was used for dot blot analysis using an antibody specific for m6A. The results of dot blotting were obtained using ChemiDoc XRS system (Bio-Rad) and quantified by ImageJ. Methylene blue (MB) staining was used as loading control.

Western blotting

For western blotting, as previously reported (29), cells were lysed in RIPA buffer with the protease inhibitor cocktail (Roche). Protein lysates were quantified with a BCA quantification kit (Thermo Fisher Scientific), subjected to SDS-PAGE and electrotransferred to PVDF membranes (Millipore). Membranes were then blocked with 5% non-fat milk in TBST (20 mM Tris–HCl, pH 7.5, 140 mM NaCl, 0.1% Tween-20), incubated with primary antibodies, followed by incubation with HRP-conjugated secondary antibodies. Western blotting results were obtained using ChemiDoc XRS system (Bio-Rad), and quantification was performed with ImageJ.

Immunofluorescence microscopy

Immunostaining was conducted as previously described (29). In brief, cells seeded on microscope coverslips were fixed with 4% paraformaldehyde for 30 min, permeabilized with 0.4% Triton X-100 in PBS for 25 min, and blocked with 10% donkey serum in PBS for 1 h at room temperature. The coverslips were incubated with the primary antibody (diluted with 1% donkey serum in PBS) overnight at 4°C, and then incubated with the fluorescence-labeled secondary antibody (diluted with 1% donkey serum in PBS at 1:500) at room temperature for 1 h. Hoechst 33342 (Invitrogen) was used to stain nuclear DNA. Images were captured with a confocal system (Leica SP5).

Reverse transcription and quantitative PCR (RT-qPCR)

Total RNA was extracted using TRIzol reagent (Invitrogen). 2 μg of RNA was converted to cDNA by using the GoScript Reverse Transcription System (Promega), and 1/100 volume of the cDNA was applied to qPCR. qPCR was performed by using iTaq Universal SYBR Green Supermix (TOYOBO). RT-qPCR primers are listed in Supplementary Table S1.

m6A methylated RNA immunoprecipitation assay (m6A-RIP/MeRIP)

The procedure of m6A immunoprecipitation (m6A-RIP or MeRIP) was modified from previously reported methods (44,45). In brief, purified mRNA was digested by DNase I and then fragmented into ∼100 nt in length by a 45-s incubation at 94°C in RNA fragmentation reagent (Life Technologies, AM8740). The reaction was stopped with a stop buffer (Life Technologies, AM8740), followed by standard ethanol precipitation and collection. 12 μg of anti-m6A antibody was pre-incubated with 50 μl Protein A Dynabeads (Life Technologies, 10013D) in IPP buffer (150 mM NaCl, 0.1% NP-40, 10 mM Tris–HCl, pH 7.4) for 1 h at room temperature. After pre-heating at 75°C for 5 min and chilling on ice immediately, 6 μg of fragmented mRNAs were added to the above prepared antibody-beads mixture and incubated for 4 h at 4°C on a rotator. After extensive washing, immunoprecipitated mixture were digested by proteinase K, and bound RNAs were subjected to phenol-chloroform (Life Technologies, AM9730) extraction and ethanol precipitation and then used for library construction or RT-qPCR analysis. The library constructions of IP-RNA samples were performed using a SMARTer smRNA-Seq Kit (Takara, 635030) and a small fraction of fragmented mRNAs (input) before IP were subjected to RNA-seq using a KAPA Stranded RNA-Seq Library Preparation Kit (KAPA, KK8401) according to the manufacturer's instructions, and sequenced on the Illumina HiSeq X Ten platform.

IGF2BP2 RIP-qPCR

RIP was performed as previously described (46–48) with some modifications. Briefly, cells seeded in a 15-cm dish at 80–90% confluency were harvested in cold PBS by the cell lifter (Corning Incorporated), pelleted by centrifuge for 5 min at 1000 × g (at 4°C) and washed once with cold PBS. The cell pellet was re-suspended with 400 μl lysis-IP buffer (50 mM Tris–HCl, pH 7.4, 150 mM NaCl, 5 mM MgCl2, 0.5% NP-40, 1 mM DTT, 1:100 protease inhibitor cocktail (Roche), 400 U/ml RNase inhibitor (Promega)), and incubated at 4°C for 2 h on a rotator and then centrifuged for 15 min to obtain clear lysates. One-tenth volume of the supernatant was saved as input and total RNA was extracted using TRIzol reagent. The remaining supernatant was incubated with IgG-conjugated protein A Dynabeads (Life Technologies) precoated with or without the anti-IGF2BP2 antibody in 500 μl lysis-IP buffer at 4°C overnight. After washing with lysis-IP buffer for five times, beads were resuspended in 300 μl elution buffer (5 mM Tris–HCl, pH 7.5, 1 mM EDTA, pH 8.0, 0.05% SDS, 4 μl proteinase K (New England Biolabs)) and digested at 55°C for 1 h. Co-immunoprecipitated RNAs were obtained after digestion by phenol–chloroform extraction and ethanol precipitation and then analyzed by RT-qPCR.

RNA stability analysis

Assay for RNA stability analysis was performed as previously described (23,49) with some modifications. hMSCs were cultured in six-well culture plates (Corning Incorporated) to ∼80% confluence. Actinomycin D (ACTD, J&K Scientific) was added at a final concentration of 5 μg/ml. Cells were collected at five time points (0, 2, 4, 6, 8 h) after the addition of ACTD and total RNA was extracted with TRIzol reagent. RT-qPCR was performed as above and GAPDH was used as loading control for normalization. For each RNA transcript of interest, a semi-log graph was plotted and its RNA half-life (t(1/2)) was determined in each condition tested.

Liquid chromatography-tandem mass spectrometry (LC-MS/MS)

LC–MS/MS analysis was conducted as previously reported (49–51). In brief, 100–200 ng of mRNA was digested by 0.1 U Nuclease P1 (Sigma) and 1.0 U calf intestinal phosphatase (New England Biolabs) in a final reaction volume of 50 μl, adjusted with water and incubated at 37°C overnight. The sample was then filtered (MW cutoff: 3 kDa, Pall, Port Washington) and subjected to LC–MS/MS analysis for detection of m6A. The nucleosides were separated by reverse phase ultra-performance liquid chromatography on a C18 column with online mass spectrometry detection using a G6410B triple quadrupole mass spectrometer (Agilent Technologies) in the positive ion mode. The nucleosides were quantified by using the nucleoside-to-base ion mass transitions of m/z 282 to 150 (m6A) and m/z 268 to 136 (A). Quantification was carried out by comparison with a standard curve obtained from pure nucleoside standards run with the same batch of samples. The ratio of m6A to A was calculated based on the calibrated concentrations.

Plasmid construction

For the construction of overexpression (OE) plasmids, cDNA prepared by reverse transcription of RNA from hMSCs was used as template to amplify genes of interest. Sequence information of full-length human METTL3 (Gene ID: NM_019852.5) was obtained from NCBI (https://www.ncbi.nlm.nih.gov/). cDNA of METTL3 was cloned into vector pLE4 (a kind gift from Dr Tomoaki Hishida) and pLE4-Luc (Luciferase) was used as control.

For lentiviral CRISPR/Cas9-mediated knockout, the sgRNA targeting METTL3 or MIS12 was cloned into the lenti-CRISPRv2 (Addgene, #52961) vector with an hSpCas9 expression cassette.

For lentiviral shRNA-mediated knockdown, the shRNA targeting IGF2BP2 was cloned into the pLVTHM vector (Addgene, #12247).

All the primers used for cloning are listed in Supplementary Table S1.

Lentivirus packaging

For lentivirus packaging, HEK293T cells were co-transfected with lentiviral vectors, psPAX2 (Addgene, #12260) and pMD2G (Addgene, #12259) using VigoFect Transfection Reagent (Vigorous Biotechnology). Viral particles were collected by ultracentrifugation at 19 400 × g at 4°C for 2.5 h.

Lentiviral CRISPR/Cas9-mediated gene knockout and lentiviral shRNA-mediated gene knockdown

Lentiviral CRISPR/Cas9-mediated gene knockout and lentiviral shRNA-mediated knockdown were conducted as previously described (52,53). Lentiviruses packaged with pLenti-CRISPRv2-METTL3, pLenti-CRISPRv2-MIS12 or pLVTHM-IGF2BP2 were transduced into WT-hMSCs. Lentiviruses containing pLenti-CRISPRv2-NTC (non-targeting control) or pLVTHM-CTRL (control) were used as negative controls. For lentiviral CRISPR/Cas9-mediated gene knockout, 48 h post-transfection, cells were treated with puromycin (Thermo Fisher Scientific) for screening with serial passaging. Phenotypic analyses were conducted after two passages.

RNA-seq

Total RNA was extracted using TRIzol reagents from 1 × 106 cells per duplicate and genomic DNA removed using a DNA-free Kit (Thermo Fisher Scientific). Library construction was conducted using a KAPA Stranded RNA-Seq Library Preparation Kit (KAPA, KK8401) according to the manufacturer's instructions. Quality control and sequencing were performed by Novogene Bioinformatics Technology Co., Ltd.

Bioinformatic analysis

General read preprocessing: two biological replicates for each condition were performed. Adaptor sequences were trimmed off for all raw reads using the FASTX_toolkit (version 0.0.14). Reads that were less than 35 nt in length or containing an ambiguous nucleotide were discarded by Trimmomatic (version 0.36). The remaining reads were aligned to the human reference genome (hg19) using HISAT2 (version 2.1.0). Only uniquely mapped reads with mapping quality score ≥20 were kept for the subsequent analysis of each sample.

For MeRIP-seq: the m6A peaks were called using the MACS2 peak-calling software (version 2.1.2) with the default options except for ‘–nomodel,–keepdup all’. A stringent cutoff threshold for a P-value of 1 × 10−5 and an enrichment score of 3 were used to obtain high-confidence peaks. Each peak was annotated based on the UCSC gene annotation information by applying BEDTools ‘IntersectBed’ (version 2.25.0). Motif enrichment was analyzed by HOMER selecting a motif length of 6 nucleotides. Background regions were generated by shuffling peaks along the transcriptome using the shuffleBed tool from the BEDTools software.

For RNA-seq: the number of uniquely mapped reads was counted using the HTSeq python package (version 0.9.1). The expression of genes was quantified as reads per kilobase of exon model per million mapped reads (RPKM). Differentially expressed genes between WT and HGPS hMSCs, WS hMSCs, or between control and METTL3-knockout hMSCs were determined using the R-package DEseq2 software (fold-change cutoff = 2.0, P-value cutoff = 5 × 10−2).

For functional enrichment analysis: Gene ontology (GO) analysis of genes with m6A modification or with differential expression was performed using metascape (http://metascape.org) or ClusterProfiler, an R package that analyzes and visualizes functional profiles (GO or KEGG) of genes and gene clusters. GO terms with a P-value < 0.05 were set to be statistically significant.

Statistical analysis

Results were presented as mean ± SEM. Two-tailed Student's t-test was conducted by using the Graph-Pad Prism Software. P-values < 0.05, P-values < 0.01 and P-values < 0.001 were considered statistically significant (marked with *, **, ***, respectively).

RESULTS

Decreased mRNA m6A modifications in HGPS and WS hMSCs

Here, we leveraged premature aging cellular models, including HGPS hMSCs (with heterozygous (G608G/+) or homozygous (G608G/G608G) LMNA mutation) and WS hMSCs (with a biallelic WRN knockout) (29,31,32,54–56), to reveal unknown aspects of m6A biology as a consequence of aging. Premature aging phenotypes were first validated (Figure 1A–C). Subsequently, m6A-specific dot blot analysis revealed a significant reduction of m6A levels in HGPS and WS hMSCs (Figure 1D), as confirmed by both immunofluorescence microscopic analysis and liquid chromatography-tandem mass spectrometry (LC-MS/MS) (Figure 1E, F). Consistent with this observation, protein levels of METTL3 were also reduced in HGPS and WS hMSCs (Figure 1G). By comparison, we did not detect any significant alterations in the expression of another m6A methyltransferase, METTL14 (Figure 1G). Taken together, these results suggest that decreased METTL3 expression may contribute to m6A loss in prematurely aged hMSCs.

Figure 1.

Reduced m6A methylation in prematurely senescent hMSCs. (A) Growth curve analysis showing population doubling of WT, HGPS (LMNAG608G/+, LMNAG608G/G608G) and WS (WRN−/−) hMSCs. Data are presented as means ± SEM, n = 3. Statistical significances are shown for passage 11. **P < 0.01; ***P < 0.001. Subsequent biological analyses in prematurely senescent hMSCs were conducted at Passage 9. (B) Clonal formation assay in WT, HGPS and WS hMSCs. Data are presented as means ± SEM, n = 3. ***P < 0.001. (C) Analysis of SA-β-Gal activity in WT, HGPS and WS hMSCs. Scale bar, 200 μm. Data are presented as means ± SEM, n = 3. ***P < 0.001. (D) Dot blot analysis of m6A levels in mRNA extracted from the indicated hMSCs. Methylene blue (MB) staining was used as RNA loading control. Data are presented as means ± SEM. ***P < 0.001. (E) Immunostaining analysis of m6A in WT, HGPS and WS hMSCs. Scale bar, 50 μm. Data are presented as means ± SEM, n = 3. *P < 0.05; **P < 0.01. (F) LC–MS/MS analysis of m6A levels in mRNA extracted from the indicated hMSCs. Data are presented as means ± SEM, n = 3. ***P < 0.001. (G) Western blot analysis of expression of core m6A methyltransferases in WT, HGPS and WS hMSCs. β-Tubulin was used as loading control. Data are presented as means ± SEM, n = 3. *P < 0.05; **P < 0.01; ns, not significant.

METTL3 is indispensable for preventing hMSCs from accelerated senescence

Similarly to broadly related studies using a loss of function approaches to probe functional roles of METTL3 (5,6,14,15), we knocked out METTL3 in WT hMSCs using a CRISPR/Cas9-based method (Figure 2A and Supplementary Figure S1A). Knockout of METTL3 (sgMETTL3) led to decreased m6A modification levels as compared to those in control hMSCs (sgNTC) (Figure 2B and Supplementary Figure S1B). Furthermore, METTL3-deficient hMSCs acquired premature aging phenotypes, as evidenced by decreased proliferative capacity and increased percentage of SA-β-Gal-positive cells (Figure 2C, D). These results suggest that METTL3 deficiency accelerates hMSC senescence.

Figure 2.

Loss of METTL3 induces senescence in hMSCs while METTL3 overexpression reverses senescence in prematurely senescent hMSCs. (A) Immunostaining analysis of METTL3 in control (sgNTC) and METTL3-knockout (sgMETTL3) hMSCs. Scale bar, 20 μm. Data are presented as means ± SEM, n = 3. ***P < 0.001. (B) Dot blot analysis of m6A mRNA modification in control and METTL3-knockout hMSCs. Methylene blue staining was used as RNA loading control. Data are presented as means ± SEM. *P < 0.05. (C) Clonal formation assay of control and METTL3-knockout hMSCs. Data are presented as means ± SEM, n = 3. ***P < 0.001. (D) SA-β-Gal staining of control and METTL3-knockout hMSCs. Scale bar, 200 μm. Data are presented as means ± SEM, n = 3. ***P < 0.001. (E) Clonal formation assay of hMSCs overexpressing Luc or METTL3. Data are presented as means ± SEM, n = 3. *P < 0.05; **P < 0.01. (F) SA-β-Gal staining of hMSCs overexpressing Luc or METTL3. Scale bar, 200 μm. Data are presented as means ± SEM, n = 3. **P < 0.01; ***P < 0.001.

To investigate whether increased m6A might delay the onset of premature senescence in progeria hMSCs, we overexpressed METTL3 in HGPS and WS hMSCs. As expected, overexpression of METTL3 increased global m6A modification levels (Supplementary Figure S1C, D). Moreover, enhanced cell proliferative capacity and reduced SA-β-Gal-positive cells were observed in HGPS and WS hMSCs upon METTL3 overexpression (Figure 2E, F). Collectively, these results suggest that intact METTL3 function is important for staving off senescence in hMSCs.

Transcriptome-wide profiling of m6A modifications in WT and prematurely aged hMSCs

We next performed RNA-seq and MeRIP-seq in HGPS and WS hMSCs to depict a transcriptome-wide m6A landscape in premature aging (Supplementary Figure S2A-B). High reproducibility was confirmed by examining Pearson's correlation coefficients between replicates of each condition. And >10 000 m6A peaks in overlapping regions for each pair of independent replicates were detected, demonstrating the high quality of MeRIP-seq data (Supplementary Figure S2A–C). Further analysis revealed that m6A modifications were mainly found in the protein-coding mRNAs (Supplementary Figure S2D), typically located in a consensus ‘DRACH’ motif (D = A/G/U; R = A/G; and H = U/A/C) (Figure 3A) and also highly enriched in the coding sequence, 3′-untranslated regions (3′ UTR) and near the stop codons (Figure 3B and Supplementary Figure S2E) as previously reported (50,57,58).

Figure 3.

m6A profiling in prematurely senescent hMSCs. (A) m6A motif identified in WT, HGPS and WS hMSCs. (B) Distribution of m6A peaks along the 5′ UTR, CDS, and 3′ UTR regions of total mRNA from each cell line normalized for length. (C) Heatmap showing shared and specific m6A peaks in the indicated hMSCs. The colour key (up) from light to dark represents m6A enrichment from low to high. GO analysis is shown on the right. The colour key (down) from yellow to red indicates -log10 (P value) from low to high. (D) IGV plots showing examples of shared and specific m6A peaks. Peaks are represented as subtracted read densities (IP minus input). (E) Heatmap showing clustering of differentially expressed genes in prematurely senescent hMSCs relative to WT-hMSCs. The colour key (up) from blue to red represents scaled expression values from low to high. GO analysis is shown on the right. The colour key (down) from yellow to red indicates -log10 (P value) from low to high.

We integrated the RNA-seq analysis with the MeRIP-seq analysis and revealed that transcripts with m6A modifications were overall more highly expressed than transcripts without m6A modifications (Supplementary Figure S3A), pointing to a potential regulatory role of m6A in gene expression homeostasis. We then subdivided all the transcripts according to their expression levels and found m6A modification levels were positively associated with mRNA expression levels (Supplementary Figure S3B). Furthermore, we noticed that the expression of mRNAs with reduced m6A modifications as a consequence of premature senescence was downregulated, while upregulated mRNAs were correlated with the gain of m6A modifications (Supplementary Figure S3C), and that this correlation occurred across different regions of the transcripts (Supplementary Figure S3D). Altogether, our data suggest a positive correlation between m6A modification levels and mRNA expression levels, consistent with previous study (59).

Reduced mRNA m6A methylation in cell cycle-related genes during hMSC senescence

To understand the biological relevance of m6A modifications in premature senescence, we investigated shared and cell type-specific m6A peaks in WT, HGPS and WS hMSCs (Figure 3C, D). Transcripts with shared m6A peaks (such as USP1) were mainly enriched for mRNA processing, RNA splicing, and histone modification pathways (Figure 3C, D), most of which are essential for cell survival (60–64). By contrast, transcripts with m6A peaks specific for premature aging models were enriched for cell morphogenesis, cellular protein catabolic process and actin cytoskeleton organization in HGPS hMSCs, such as PKD2 and EGF (Figure 3C, D), and translation-related process and RNA metabolism in WS hMSCs, such as RPL7A (Figure 3C, D). Consistent with our data showing a global loss of mRNA methylation in premature aging cell models (Figure 1D–F), m6A modifications were decreased in all three hMSC senescence models relative to those in WT hMSCs (Figure 3C). Moreover, the premature senescence-associated hypomethylated transcripts that harbored WT-specific m6A peaks were involved in functional categories related to cell division, microtubule-based process, regulation of cell cycle, DNA repair and chromosome organization (Figure 3C), such as ASPM (Figure 3D). Most of these terms are cell cycle-related pathways, which are dysregulated during aging (65–67).

Next, we investigated gene expression profiles of WT, HGPS and WS hMSCs and found distinct patterns in each cell line (Figure 3E and Supplementary Figure S4A-D). The gene expression patterns of heterozygous and homozygous HGPS hMSCs were similar to each other, but were largely different from those of WT and WS hMSCs. In HGPS hMSCs, highly expressed genes were related to exocytosis, endoplasmic reticulum stress, immune responses, metabolic processes and autophagy (Figure 3E). In WS hMSCs, pathways regulating actin filament-based processes, cell morphogenesis, cell projection organization, kinase activities, as well as protein catabolic processes, were significantly enriched with highly expressed genes (Figure 3E). In addition, we found the genes upregulated in both HGPS and WS hMSCs were enriched in developmental pathways (Supplementary Figure S4A, B). By comparison, highly expressed genes in WT hMSCs (also representing downregulated genes in premature senescence) were enriched in chromosome segregation, nuclear division, histone modification, DNA replication, cell cycle phase transition and spindle organization pathways (Figure 3E and Supplementary Figure S4C-D), most of which were cell cycle-related terms as aforementioned (Figure 3C). These data suggest that cell cycle-related genes showing m6A loss and downregulated expression may play predominant roles in regulating hMSC senescence.

We next compared the m6A profiles in METTL3-deficient and control hMSCs (Supplementary Figure S5A–E) and observed similar patterns to those of WT and prematurely senescent hMSCs. Shared m6A-modified transcripts between METTL3-deficient and control hMSCs were mainly enriched in histone modification and RNA processing pathways (Supplementary Figure S5F, G). We also noticed the presence of a small fraction of transcripts with increased m6A modifications after METTL3 deletion (Supplementary Figure S5F, G), likely due to the existence of other m6A methyltransferases. Notably, regulation of cell cycle progression was also enriched for transcripts showing m6A loss after METTL3 knockout (Supplementary Figure S5F, G), in further support of altered m6A modifications in cell cycle-related genes potentially contributing to hMSC senescence.

MIS12 functions as a key m6A target in premature senescence

In search for key regulators mediating m6A-associated effects in hMSC senescence, we discovered a common loss of mRNA methylations near the stop codons of transcripts both in prematurely senescent and METTL3-deficient hMSCs (Figure 4A, B and Supplementary Figure S6A–D). GO analysis of these transcripts revealed enrichment of cell cycle-related terms (Figure 4C) as described above. Among these transcripts, we noticed that m6A modifications in MIS12, a key regulator of cell proliferation (68), were markedly reduced in both prematurely senescent and METTL3-deficient hMSCs (Figure 4D and Supplementary Figure S6E), as validated by MeRIP-qPCR analysis (Figure 4E, F). We also detected a significant decrease in MIS12 expression at both the mRNA and protein levels in HGPS and WS hMSCs as well as in METTL3-deficient hMSCs (Figure 5A–D).

Figure 4.

MIS12 is a downstream target of METTL3 that regulates hMSC senescence. (A) m6A frequency near stop codon regions in transcripts from WT-, HGPS- and WS-hMSCs. (B) m6A frequency near stop codon regions in transcripts from control and METTL3-knockout hMSCs. (C) Pie chart showing genes with m6A loss at stop codon region that overlap between prematurely senescent hMSCs and METTL3-knockout hMSCs. (D) IGV plots showing m6A modification on MIS12 mRNA in different cell lines. Peaks are represented as subtracted read densities (IP minus input). White arrows indicate the direction of gene transcription and yellow triangle indicates the location of the m6A peak. (E) MeRIP-qPCR detection for m6A enrichment on MIS12 mRNA of WT-, HGPS- and WS-hMSCs. Data are presented as means ± SEM, n = 6. *P < 0.05; **P < 0.01. (F) MeRIP-qPCR detection for m6A enrichment on MIS12 mRNA of control and METTL3-knockout hMSCs. Data are presented as means ± SEM, n = 6. ***P < 0.001.

Figure 5.

m6A promotes IGF2BP2-mediated MIS12 mRNA stability to prevent accelerated senescence in hMSC. (A) qPCR analysis of MIS12 mRNA levels in WT, HGPS and WS hMSCs. Data are presented as means ± SEM, n = 6. ***P < 0.001. (B) Western blot analysis of MIS12 protein levels in WT, HGPS and WS hMSCs. β-Tubulin was used as loading control. Data are presented as means ± SEM, n = 3. *P < 0.05; **P < 0.01. (C) qPCR analysis of MIS12 mRNA levels in control and METTL3-knockout hMSCs. Data are presented as means ± SEM, n = 6. ***P < 0.001. (D) Western blot analysis of MIS12 protein levels in control and METTL3-knockout hMSCs. β-Tubulin was used as loading control. Data are presented as means ± SEM, n = 3. *P < 0.05. (E) RIP-qPCR analysis showing enrichment of IGF2BP2 on MIS12 mRNA in WT hMSCs. IgG was used as isotype control. GAPDH was used as a negative control. Data are presented as means ± SEM, n = 3. ***P < 0.001; ns, not significant. (F) RIP-qPCR analysis showing enrichment of IGF2BP2 on MIS12 mRNA in control and METTL3-knockout hMSCs. IgG was used as isotype control. Data are presented as means ± SEM, n = 3. ***P < 0.001. (G) Western blot analysis of IGF2BP2 protein levels in control (shCTRL) and IGF2BP2-silenced (shIGF2BP2) hMSCs. β-Actin was used as loading control. Data are presented as means ± SEM, n = 3. **P < 0.01. (H) qPCR analysis of the MIS12 mRNA levels in control and IGF2BP2-silenced hMSCs. Data are presented as means ± SEM, n = 3. *P < 0.05. (I) Western blot analysis of the MIS12 protein levels in control and IGF2BP2-silenced hMSCs. β-Actin was used as loading control. Data are presented as means ± SEM, n = 3. ***P < 0.001. (J) Western blot analysis of MIS12 protein levels in control (sgNTC) and MIS12-knockout (sgMIS12) hMSCs. β-Tubulin was used as loading control. Data are presented as means ± SEM, n = 3. **P < 0.01. (K) Clonal formation assay of control and MIS12-knockout hMSCs. Data are presented as means ± SEM, n = 3. *P < 0.05. (L) SA-β-Gal staining of control and MIS12-knockout hMSCs. Scale bar, 200 μm. Data are presented as means ± SEM, n = 3. ***P < 0.001. (M) Clonal formation assay of control and IGF2BP2-silenced hMSCs. Data are presented as means ± SEM, n = 3. *P < 0.05. (N) SA-β-Gal staining of control and IGF2BP2-silenced hMSCs. Scale bar, 200 μm. Data are presented as means ± SEM, n = 3. **P < 0.01.

When we examined the stability of MIS12 mRNA, we observed a shortened mRNA half-life of MIS12 in HGPS and WS hMSCs as well as in METTL3-deficient hMSCs (Supplementary Figure S7A, B), indicating that m6A modifications may promote MIS12 mRNA stability. Interestingly, a recent study has reported that IGF2BPs recognize m6A modifications in thousands of mRNA transcripts and promote the stability and translation of these mRNAs, including MIS12 identified as one of the targets of IGF2BP2 by high-throughput sequencing analyses (23). Accordingly, we asked whether IGF2BP2 recognition and stabilization of m6A-modified MIS12 mRNA might regulate hMSC senescence.

To test this hypothesis, we conducted RIP-qPCR analysis with an anti-IGF2BP2 antibody to detect a potential interaction between IGF2BP2 and MIS12 mRNA in hMSCs. Indeed, we observed a significantly higher enrichment of IGF2BP2 on MIS12 mRNA relative to IgG control (Figure 5E). Importantly, METTL3 deficiency disrupted this interaction (Figure 5F), implying that IGF2BP2 binds MIS12 mRNA in an METTL3/m6A-dependent manner. To confirm that IGF2BP2 promotes MIS12 expression, we knocked down IGF2BP2 with shRNA in WT hMSCs and detected a significant decrease in MIS12 expression at both the mRNA and protein levels (Figure 5G–I). We also observed a shortened MIS12 mRNA half-life after IGF2BP2 knockdown (Supplementary Figure S7C), suggesting that IGF2BP2 could recognize and stabilize m6A-tagged MIS12 mRNA.

Finally, we used a CRISPR/Cas9-based method to ablate MIS12 in WT hMSCs (Figure 5J). As expected, MIS12 deficiency accelerated hMSC senescence, as evidenced by decreased proliferative capacity and increased SA-β-Gal staining (Figure 5K, L). Likewise, IGF2BP2 knockdown also led to accelerated cellular senescence in WT hMSCs (Figure 5M, N). Collectively, these data suggest that IGF2BP2 and its target mRNA MIS12 contribute to m6A-associated regulation of hMSC senescence.

DISCUSSION

Our results reveal that m6A RNA modifications play a central role in regulating premature senescence in hMSCs. Firstly, we found that m6A modification levels were reduced concomitant with the downregulation of METTL3 in prematurely senescent hMSCs. Secondly, the depletion of METTL3 led to reduced m6A modifications and MIS12 downregulation, accelerating hMSC senescence. Thirdly, overexpression of METTL3 restored m6A modifications and prevented premature senescence in hMSCs. Finally, IGF2BP2 recognized and stabilized m6A-modified MIS12 mRNA to prevent accelerated hMSC senescence. Together, our findings demonstrate that METTL3 and m6A modifications could alleviate hMSC senescence through IGF2BP2-mediated enhancement of MIS12 mRNA stability (Figure 6), adding an epitranscriptional layer to the regulation of premature senescence in stem cells.

Figure 6.

A model illustrating the protective role of METTL3/m6A in alleviating hMSC senescence. In young cells, METTL3 maintains m6A levels to stabilize MIS12 mRNA; facilitating IGF2BP2 binding, MIS12 expression and hMSC self-renewal. In aged cells, METTL3 downregulation results in reduced m6A levels; preventing IGF2BP2 binding and accelerating MIS12 mRNA turnover, causing accelerated cellular senescence in hMSCs.

By targeting different mRNAs in different cell types and organs (13), m6A regulation by METTL3 can shape biological processes in very different ways in both normal and diseased conditions. Aberrant m6A modifications are known to lead to impaired embryonic development, tumorigenesis and the occurrence of diabetes by affecting the fate of specifically targeted mRNAs (21,69–71). Of significance for our findings, abnormal m6A modifications are recently implicated in Parkinson's disease (PD) and Alzheimer's disease (AD) models, well-known age-related neurodegenerative disorders (72,73). Moreover, dysregulation of m6A has also been implicated in age-related infertility and heart failure in mice (74,75). These findings all point to a regulatory role of m6A in aging.

Consistent with our findings, Min et al. has reported that decreased m6A modifications in human blood cells collected from aged individuals lead to cellular senescence in human fibroblasts, partially attributed to the instability of AGO2 mRNA that may be recapitulated by METTL3 depletion (76). However, genetic differences between individuals were not investigated in this study, and a causal link between m6A and aging has not yet been established. In our study, we employed isogenic wild-type and prematurely senescent hMSC models. We observed m6A loss during premature aging as a consequence of METTL3 downregulation. We also showed that knockout of METTL3 accelerated senescence in young hMSCs, reflected by increased SA-β-Gal staining and impaired proliferation (consistent with the findings in cancer cells (59)). Combined with the functional validations showing that METTL3 overexpression alleviated cellular senescence in prematurely senescent hMSCs, we reached a conclusion that METTL3 as well as m6A play important roles in stem cell senescence. Consistent with our conclusion, a recent study has reported that METTL3 deficiency shortens lifespan in Drosophila (77).

In cellular models of human premature aging, m6A profiling revealed declined methylation and expression levels of cell cycle-related pathways. Cell cycle arrest is a common feature in many aging contexts (65–67). As a cell cycle-related factor, MIS12 is known to interact with DSN1, NSL1, and PMF1, forming the MIS12 complex (MIS12C) that promotes chromosome segregation and kinetochore formation during mitosis (78) and thus plays critical regulatory roles in cell proliferation (68). Consistently, knockdown of MIS12 in human and chicken cells results in chromosome mis-segregation and chromosome biorientation defects (79). However, little is known about the function of MIS12 in cellular senescence. In our study, we observed significant reduction of both m6A modification and MIS12 mRNA stability in prematurely senescent hMSCs. METTL3 deficiency decreased stability of MIS12 transcripts in hMSCs, likely due to reduced m6A modifications around their stop codons preferentially recognized and bound by the m6A reader IGF2BP2. Indeed, it has been reported that IGF2BP2 selectively recognizes m6A-modified MIS12 mRNA and promotes mRNA stability and translation efficiency (23). These data offer insights into precise molecular mechanisms that underlie the regulatory roles of m6A during stem cell senescence.

Epigenetic changes have a powerful influence on the aging process (41,80–84) and genetic reprogramming targeting such changes may lead to cellular rejuvenation (41). For example, partial reprogramming using the Yamanaka factors Oct4, Sox2, Klf4 and c-Myc (OSKM) restores histone modifications, including H3K9me3 and H4K20me3, and extends lifespan in a progeria mouse model (42). Rejuvenation from an epitranscriptomic aspect has not yet been reported, whereas our findings, for the first time, demonstrate that m6A loss is a senescence biomarker in hMSCs and that the restoration of m6A through METTL3 overexpression may rescue premature senescence phenotypes in hMSCs.

Taken together, our study puts forth a regulatory model in which IGF2BP2-mediated enhancement of MIS12 mRNA stability is increased by METTL3-mediated m6A modifications to reverse the senescent phenotypes of hMSCs. The existence of a direct molecular connection between MIS12 expression and its m6A status suggests the manipulation of m6A modification levels as a potential strategy for cellular rejuvenation. Our work sheds light on the poorly understood molecular mechanism of m6A in aging and identifies METLL3 and MIS12 as novel biomarkers for diagnosis and candidate targets for the treatment of age-associated disorders.

DATA AVAILABILITY

RNA-seq and MeRIP-seq data generated in this study have been deposited in the Genome Sequence Archive (85) in National Genomics Data Center (86), Beijing Institute of Genomics (China National Center for Bioinformation), Chinese Academy of Sciences, under accession number HRA000206 that are publicly accessible at http://bigd.big.ac.cn/gsa-human.

Click here for additional data file.