m6A mRNA methylation facilitates resolution of naïve pluripotency toward differentiation

See allHide authors and affiliations

Science  27 Feb 2015:
Vol. 347, Issue 6225, pp. 1002-1006
DOI: 10.1126/science.1261417

mRNA modification regulates pluripotency

When stem cells progress from an embryonic pluripotent state toward a particular lineage, molecular switches dismantle the transcription factor network that keeps the cell pluripotent. Geula et al. now show that N6-methyladenosine (m6A), a messenger RNA (mRNA) modification present on transcripts of pluripotency factors, drives this transition. Methylation destabilized mRNA transcripts and limited their translation efficiency, which promoted the timely decay of naïve pluripotency. This m6A methylation was also critical for mammalian development.

Science, this issue p. 1002


Naïve and primed pluripotent states retain distinct molecular properties, yet limited knowledge exists on how their state transitions are regulated. Here, we identify Mettl3, an N6-methyladenosine (m6A) transferase, as a regulator for terminating murine naïve pluripotency. Mettl3 knockout preimplantation epiblasts and naïve embryonic stem cells are depleted for m6A in mRNAs, yet are viable. However, they fail to adequately terminate their naïve state and, subsequently, undergo aberrant and restricted lineage priming at the postimplantation stage, which leads to early embryonic lethality. m6A predominantly and directly reduces mRNA stability, including that of key naïve pluripotency-promoting transcripts. This study highlights a critical role for an mRNA epigenetic modification in vivo and identifies regulatory modules that functionally influence naïve and primed pluripotency in an opposing manner.

Murine pluripotent embryonic stem cells (ESCs) reside in a “naïve” molecular state that largely resembles that of the preimplantation inner cell mass, whereas epiblast stem cells (EpiSCs), derived from the postimplantation epiblast, resemble an advanced developmental stage and are already “primed” for differentiation (1, 2). Limited knowledge exists with regard to the molecular regulators that are critical for transitioning toward or for exclusively maintaining primed pluripotent EpiSCs. Thus, we conducted a small interfering RNA (siRNA) screen against selected transcriptional and epigenetic regulators previously evaluated in the context of naïve pluripotency modulation (1, 2) and tested whether primed EpiSCs , harboring a green fluorescent protein (GFP) reporter knock-in allele under the control of endogenous Oct4 promoter(Oct4-GFP+/+), might selectively rely on some of these factors (Fig. 1A and fig. S1). Regulators that specifically inhibited the stability and viability of Oct4-GFP+ primed cells included the epigenetic repressors Dnmt1, Eed, and Suz12 Polycomb components; Mbd3; and N6-adenosine methyltransferase Mettl3, a component of the N6-methyladenosine (m6A) mRNA methylating complex (3) (Fig. 1A and fig. S1).

Fig. 1 Derivation and characterization of Mettl3 KO ESCs.

(A) siRNA screen for regulators that destabilize Oct4-GFP+ mouse naïve or primed pluripotent cells. Error bars indicate SD (n = 3). Student’s t test *P < 0.05 relative to scrambled control. Gray boxes highlight EpiSC-specific regulators. (B) ESC derivation efficiency after mating of Mettl3 heterozygote mice. (C) Liquid chromatography–tandem mass spectrometry analysis of m6A percentage relative to adenosine in purified mRNA. Error bars indicate SD (n = 3). (D) ESC morphology and immunostaining.

We subsequently focused on the role of m6A in pluripotency transitions, because the biological role of RNA modifications is only starting to be unveiled (4). m6A is an RNA modification catalyzed by Mettl3 (methyl transferase–like 3) and Mettl14, components of a partially characterized multicomponent methyltransferase complex (3). This modification is removed by Fto and Alkbh5 demethylases. A role for m6A was demonstrated in regulation of gene expression through mRNA splicing, localization, and degradation, as well as in modulating the binding capacity of m6A binding “reader” proteins such as Ythdf1-3 (4).

We targeted the endogenous Mettl3 locus in mouse ESCs and generated a truncated out-of-frame allele (fig. S2). Mettl3+/– mice were obtained and used to derive embryonic day (E)3.5 blastocysts (Fig. 1B). Mettl3–/– knockout (KO) blastocysts retained normal morphology and expression of pluripotency markers (fig. S3, A and B) and yielded ESCs at the expected ratio (Fig. 1B and fig. S3, C and D). Quantitative mass spectrometry analysis (MS) of m6A levels in purified mRNA from wild-type Mettl3+/+ (WT) and KO cells showed that Mettl3 ablation leads to near-complete depletion of m6A on mRNA (Fig. 1C).

Mettl3–/– ESCs preserve their naïve pluripotent identity, as evidenced by domed-shape colony morphology and expression of pluripotency markers (Fig. 1D and fig. S3, F to I). To test their differentiation ability, WT and KO cells were transferred to differentiation media for embryoid bodies (EBs) for 8 to 21 days. KO cells generated dense EB spheres but failed to undergo the characteristic cavitation observed in WT EBs (fig. S4, A and B). KO EBs failed to robustly up-regulate early developmental markers and to adequately repress pluripotent genes (Fig. 2A and fig. S4C). Twenty-one–day–old EBs were disaggregated and replated in ESC growth conditions, and only KO EBs efficiently regenerated stable ESCs (fig. S4, D and E). In vitro protocols did not efficiently differentiate Mettl3–/– ESCs into mature neurons (fig. S4F). Consistently, Mettl3 KO ESCs did not contribute to embryo chimera formation after blastocyst microinjection (fig. S5).

Fig. 2 Mettl3 KO ESCs resist termination of naïve pluripotency.

(A) qPCR analysis for pluripotency (top) and differentiation markers (bottom) expression after 8 days of EB induction. Error bars indicate SD (n = 3). (B) Immunostaining of teratoma sections. DAPI, 4′,6-diamidino-2-phenylindole. (Arrows highlight regions with pluripotency marker expression.) (C) Morphology and immunostaining images of Mettl3 WT and KO ESCs transferred to primed conditions.

WT and KO ESCs were injected into immunodeficient mice to generate mature teratomas. Mettl3–/– ESCs generated larger tumors, and histological analysis showed that KO teratomas were poorly differentiated. However, mature structures were abundant in WT teratomas (fig. S6, A to C). Immunostaining showed that KO teratomas poorly expressed differentiation markers, such as Gata4/6 and troponin, and diffusely expressed pluripotent markers, such as Oct4 and Nanog, even 6 weeks after subcutaneous in vivo growth (Fig. 2B and fig. S6, D and E). Disaggregation of dissected KO tumors demonstrated that >75% of their cells still expressed Oct4-GFP pluripotency reporter (fig. S6F). Only KO teratomas contained pluripotent cells that could rapidly recover in culture and give rise to ESC colonies within 6 days and could form stable Oct4-GFP+ ESC lines (fig. S6G). The resistance to differentiation could be rescued by reconstitution with a WT Mettl3 transgenic allele (fig. S7). Deletion in ESCs of the other methyltransferase complex component, Mettl14 (5), showed an equivalent reduction in m6A in mRNA and recapitulated in vitro resistance to differentiation as seen with Mettl3 KO ESCs (fig. S8).

We next wanted to test the ability of female Mettl3–/– ESCs to convert from naïve pluripotent state into primed epiblastlike state in vitro, by applying primed fibroblast growth factor 2 (FGF2)–activin conditions (2). WT colonies obtained a typical flat morphology (Fig. 2C); reduced expression of naïve markers (Nanog, Rex1, and Klf4); and induced primed pluripotency markers (Xist, Foxa2, Brachyury, and Fgf5) (fig. S9, A to D). Mettl3–/– colonies, however, kept their naïve-domed shape; failed to down-regulate naïve markers, such as Esrrb and Nanog; and up-regulated only Fgf5 but not Xist, Foxa2, or Brachyury (Fig. 2C and fig. S9, A to D). Other indicators for pluripotency priming, such as enriched cytoplasmic localization of Tfe3 upon pluripotency priming and X-chromosome inactivation (evident by formation of foci for histone H3 trimethylated at lysine 27), were observed only in WT cells (fig. S9, B and E). KO ESCs were also more resilient to differentiation in the absence of LIF (fig. S9F). In conclusion, depletion of m6A in mRNA of Mettl3–/– ESCs hampers their priming and differentiation competence, which leads to a “hyper”-naïve pluripotency phenotype.

We next analyzed the effect of Mettl3 depletion on normally established WT EpiSC viability and stability, in comparison with WT naïve ESCs. Quantitative polymerase chain reaction (qPCR) analysis showed that, in naïve cells, Mettl3 depletion amplifies the already highly expressed naïve pluripotency transcripts, which further boosts naïve circuitry stability (fig. S10, A and B). The increase in minimally expressed lineage factors like Fgf5 is marginal, as their basal transcript levels are very low (fig. S10B). As naïve WT cells progress toward EpiSC stage, pluripotency genes are down-regulated, and lineage commitment markers become abundantly expressed (2) (fig. S10, C and D). At this stage, Mettl3 depletion leads to minimal amplification of pluripotency genes and further boosts the highly expressed lineage commitment markers, which leads to tipping the balance toward differentiation and compromises the stability of the primed state (fig. S11, A and B, and supplementary text). Consistently, we dissected a complex effect for Mettl3 depletion in different stages of cellular reprogramming toward naïve pluripotency (figs. S11 and S12).

To resolve the roles of m6A in the onset of the “hyper”-naïve pluripotent phenotype, we applied m6A sequencing (m6Aseq) (4) on RNA purified from mouse naïve ESCs, 11-day-old EBs, and mouse embryonic fibroblasts (MEFs) (fig. S13). We identified 10,431, 8356, and 11,948 m6A peaks within 6412, 5504, and 6427 expressed genes of ESCs, EBs, and MEFs, respectively, which constitute 44 to 52% of expressed genes (Fig. 3A, figs. S13 and S14, and tables S1 and S2). Functional enrichment analysis of methylated genes revealed statistically significant enrichment of genes involved in various basic cellular processes, as well as for targets of pluripotency regulators (e.g., Nanog) (false discovery rate <1%) (fig. S13G). Note that 28 out of 35 (80%) naïve pluripotency-promoting genes (6) were methylated for m6A, including Nanog, Klf2, and Esrrb but not Oct4, which is expressed in both pluripotent states (Fig. 3A, fig. S15, and table S3) (1). Lineage priming transcripts expressed in WT EBs, like Foxa2 and Sox17, were also positive for m6A (fig. S16, A to C).

Fig. 3 Molecular characterization of Mettl3 KO cells.

(A) The m6A methylation and transcriptional landscape. Normalized read density (reads-per-million) levels are shown as green shades: m6A-IP in WT ESC, EB, and MEF; gray shades: m6A input in WT ESC, EB, and MEF; blue shades: RNAseq in WT ESC, EB, and MEF; and pink shades: RNAseq in KO ESC and EB. In m6A samples, all three biological replicates are shown (replicate number indicated on the left). The genome browser, range is shown at the right side of each track. Significant peaks are indicated in horizontal black rectangles and green highlight. Levels are normalized by the number of reads in each sample. (B) Transcript level changes in KO ESCs compared with WT ESCs, as a function of the number of m6A peaks in each transcript. Box plots describe the distribution of fold-change; medians are indicated in the center of the boxes. (C) Distribution of transcripts’ half-life (in hours) in ESCs (left) and EBs (right), for WT (blue shades) and KO (pink/red). Distributions are shown for genes without m6A (12,461), genes with at least one m6A peak in either condition (n = 7181), and genes with at least three peaks (427). Paired Wilcoxon *P < 0.05, **P < 0.00005. (D) Translation efficiency in WT and KO ESCs. Paired Wilcoxon **P < 10−8.

Global transcriptional profiles of KO EB samples exhibit high correlation with those of WT and KO ESCs (fig. S17A), which validates our functional phenotypic observations (Fig. 2). Pluripotency regulators’ transcript levels are increased in KO EBs compared with WT, and in several cases, this increase is evident already in ESCs (e.g., Klf2, Klf4, and Esrrb) (figs. S15 and S17, F and G). Notably, m6A marked transcripts are significantly higher in KO than in WT ESCs (Fig. 3B and fig. S17, B and E), as well as known targets of Ythdf2 (7) and of Mettl3 in mouse ESCs (8) (fig. S17C). The change in expression level of a given gene in KO compared with WT is positively correlated with the number of m6A peaks of that transcript in ESCs and EBs (Fig. 3B and fig. S17, B and D). Overall, these results suggest that m6A depletion from mRNA by Mettl3 KO increases mRNA levels of methylated transcripts.

Changes in mRNA levels reflect the difference between transcription and degradation rates. A recent study showed that certain Ythdf proteins, abundantly expressed in ESCs and EBs (fig. S18A), mediate degradation of methylated mRNA (7). To measure mRNA degradation rates, we monitored mRNA levels after transcription inhibition with actinomycin-D (fig. S18B), followed by calculation of degradation and half-life rates. The half-life of methylated transcripts is significantly shorter than that of unmethylated transcripts in WT ESCs (P < 3 × 10−6) and WT EBs (P < 0.005) (Fig. 3C). Subsequently, in KO cells, the half-life of methylated transcripts increases significantly in ESCs and EBs (P < 2 × 10−16) (fig. S18). Previously identified mRNA bound targets of Ythdf2 and pluripotent gene transcripts also showed a significantly increased half-life that was prominent in KO samples (figs. S18D and S19). Transcript half-life was assessed by reverse transcription with qPCR (RT-qPCR) with and without actinomycin-D treatment in ESCs, and this procedure validated increased half-life rates (>twofold change) of Klf4, Nanog, Sox2, and Zfp42 but not Oct4 and Mta2, as those transcripts are unmethylated (fig. S20). Combined knockdown of Nanog, Klf4, and Esrrb in Mettl3 KO ESCs improved their ability to up-regulate lineage commitment markers upon in vitro priming (fig. S21). Overall, these results indicate that increased stability of methylated pluripotent mRNA transcripts, which resulted in matching alterations of protein levels (fig. S22), contributed to the “hyperpluripotent” phenotype induced by Mettl3 ablation.

Analysis of translation topology in WT and KO ESCs and EBs was conducted by using ribosome profiling [Riboseq (9)] (figs. S22C and S23). To identify whether differences in translation existed and were correlated with m6A abundance, we calculated translation efficiency (TE; ribosome footprint reads per million reads (RPM)/mRNA RPM) (9). When we analyzed the TE in cells of KO compared with WT cells, we found a modest yet significantly increased TE in KO compared to WT ESCs (Fig. 3D and fig. S22D), including of pluripotency-promoting transcripts, such as Utf1, Lin28, and Gbx2 (fig. S22E). This effect was observed on methylated and unmethylated transcripts with significantly higher GC content, yet independent of the length of the 3′ untranslated region or m6A deposition (Fig. 3D and fig. S22, F to I). Collectively, absence of m6A leads to an indirect increase in translation efficiency of GC-rich transcripts and a direct increase in mRNA stability of m6A-decorated transcripts. This includes transcripts of prominent naïve pluripotency regulators that stabilize this state and shield its responsiveness to lineage priming cues. Upon depletion of Mettl3, we detected significant changes in splicing patterns that were directly dependent on m6A and significant changes in adenosine-to-inosine RNA editing that were indirectly dependent on m6A (figs. S24 and S25) (10). However, how and whether the latter changes alter Mettl3-KO pluripotency regulation remain to be defined (supplementary text).

We then determined the extent to which the in vitro observed phenotypes correlate with in vivo development and crossed heterozygote mice to obtain Mettl3–/– litters. KO of Mettl3 is embryonic lethal (fig. S26). Inspecting E12.5, E10.5, and E8.5 embryos, we saw that all KO embryos were already absorbed (fig. S26, B to G). Whereas E3.5 KO blastocysts retain normal characteristics (Fig. 4A and fig. S3, A and B), postimplantation E5.5 to E7.5 KO embryos were deformed and relatively deficient in adopting the typical postimplantation epiblast egg cylinder shape (fig. S26, E to G). Note that Oct4+ cells were readily detected at E5.5 to E7.5 in KO postimplantation epiblasts, which suggested that pluripotent cells existed in vivo and excluded precocious differentiation as the cause for embryonic lethality (Fig. 4B and fig. S27). However, the typical down-regulation and retraction in Nanog expression seen in WT embryos at E5 to E5.5 (11) was not observed in KO embryos (fig. S28A). Moreover, at E6.0 to E7.5, Nanog expression in WT embryos was reinitiated and restricted to the proximal posterior epiblast (Fig. 4C). On the contrary, Nanog was diffusely expressed throughout the entire Oct4+ epiblast in KO E6.0 to E7.5 embryos, (Fig. 4C and fig. S28, B and C). Notably, X-chromosome inactivation and Esrrb down-regulation were observed in both WT and KO postimplantation epiblasts (fig. S29), which indicated that some level of priming does occur and that the latter relative resistance to priming is less severe in some characteristics in vivo relative to that observed in vitro (Fig. 2). Nevertheless, reduced competence to undergo priming was also evident in vivo by the fact that early differentiation markers such as Brachyury+, Foxa2+ cells or early Oct4+/Blimp1+ primordial germ cells were not induced in KO postimplantation embryos (Fig. 4D and figs. S30 to S32). Further, E6.5 KO Oct4+ epiblasts expanded and maintained in primed FGF2-activin conditions yielded Oct4+/Esrrb+/Nanog+ naïve-like pluripotent lines (fig. S33). Collectively, the retention of widespread Nanog expression and maintenance of Oct4 expression without up-regulating lineage commitment genes in vivo are largely consistent with the in vitro observed phenotypes. Further, the latter reiterate relative resistance to terminate aspects of naïve pluripotency and the formation of an inadequately primed pluripotent epiblast in Mettl3 KO postimplantation embryos (fig. S34).

Fig. 4 Mettl3 regulates pluripotency priming and differentiation in vivo.

(A to D) Representative immunostaining images of WT and KO histological sections at the indicated developmental stages. (Epiblasts are marked by dashed outline; arrows highlight Brachyury+ cells.)

In summary, we identify m6A mRNA methylation as a regulator acting at molecular switches, during resolution of murine naïve pluripotency, to safeguard an authentic and timely down-regulation of pluripotency factors, which is needed for proper lineage priming and differentiation (fig. S34). These findings set the stage for dissecting the role of m6A in other developmental transitions (12, 13) and for exploring other potential regulatory roles for m6A and its reader proteins.

Supplementary Materials

Materials and Methods

Supplementary Text

Figures S1 to S34

Tables S1 to S6

References (1460)

References and Notes

  1. Acknowledgments: G.R. is supported by grants from the Flight Attendant Medical Research Institute (FAMRI), the Israel Science Foundation (grant no. 1667/12), the Israeli Centers of Excellence (I-CORE) Program and The Israel Science Foundation (grants no. 41/11 and no. 1796/12), and the Ernest and Bonnie Beutler Research Program. G.R. is a member of the Sagol Neuroscience Network and holds the Djerassi Chair for Oncology (Sackler Faculty of Medicine, Tel-Aviv University, Israel). D.D. is supported by a Human Frontier Science Program long-term fellowship. J.H.H. is supported by a generous gift from Ilana and Pascal Mantoux; the New York Stem Cell Foundation; FAMRI; the Kimmel Innovator Research Award; the European Research Council starting grant (StG-2011-281906); the Leona M. and Harry B. Helmsley Charitable Trust; Britain Israel Research and Academic Exchange Partnership Regenerative Medicine Initiative; The Sir Charles Clore Research Prize; the Israel Science Foundation (Bikura, Morasha, ICORE, and Regular research programs); the Israel Cancer Research Fund; the Benoziyo endowment fund; the Helen and Martin Kimmel Institute for Stem Cell Research; Fritz Thyssen Stiftung; and Erica A. and Robert Drake. J.H.H. is a New York Stem Cell Foundation–Robertson Investigator. We thank Weizmann Institute management for providing critical financial and infrastructural support. DNA sequencing data have been deposited under National Center for Biotechnology Information, Gene Expression Omnibus submission GSE61998. All authors declare lack of conflict of interest.
View Abstract

Navigate This Article