N6-methyladenosine (m⁶A) depletion regulates pluripotency exit by activating signaling pathways in embryonic stem cells

To observe immediate m⁶A depletion effects, we employed short hairpin RNA (shRNA)- or small interfering RNA (siRNA)–mediated knock-down (KD) of Mettl3 and Mettl14 for 10 d. This significantly decreased Mettl3 and Mettl14 protein levels in mESCs (Fig. 1A and SI Appendix, Fig. 1 A, C, E, and I). Depletion of either Mettl3 or Mettl14 reciprocally down-regulates the other complex partner (Fig. 1A and SI Appendix, Fig. 1 A, C, E, and I), indicating mutual coregulation at the protein level (27, 29). Mettl3 or Mettl14 KD increased Wtap protein levels as previously shown (29) (Fig. 1A and SI Appendix, Fig. 1 A, C, E, and I) and reduced m⁶A abundances on mRNA (29) (Fig. 1B). m⁶A depletion resulted in a mixture of both domed, rounded, and flat, loosely formed mESC colonies (Fig. 1C). To determine Nanog protein levels over time in individual cells, we employed protein-fusion reporter mESCs (11). The tethering of fluorescent proteins (FPs) to the carboxy termini of TFs shows functional TF protein levels (11, 12). We confirmed no relation between the degree of m6A depletion and Nanog expression (11) (SI Appendix, Fig. 1 G and H). Immediate m⁶A depletion increases the proportion of Nanog-negative cells (Fig. 1 D and E and SI Appendix, Fig. 1 B, D, and F), although almost all cells remain positive for core pluripotency TFs Oct4 and Sox2 (Fig. 1 F, G, and H and SI Appendix, Fig. 1 J, K, and L). Multispectral protein-level quantification showed that Oct4- and Sox2-positive, Nanog-negative cells were also negative for Klf4 (SI Appendix, Fig. 1M), Esrrb (SI Appendix, Fig. 1N), and Tbx3 (SI Appendix, Fig. 1O). The down-regulation of pluripotency TFs is accompanied by a global increase in pluripotency TF mRNA stability. This is shown by real-time PCR quantification of Nanog and Esrrb transcript abundances after 3 and 6 h of transcription inhibition with Actinomycin D (Fig. 1 I and J) (28). In comparison, no mRNA stability effects were observed for Oct4, which does not retain m⁶A deposition–mediated regulation (27) (SI Appendix, Fig. 1L). These findings indicate that pluripotency TF down-regulation is disconnected from increased pluripotency TF mRNA stability observed in m⁶A depleted cells.

The culture of mESCs in serum/LIF (8, 9) allows cells to undertake a reversible step toward differentiation, marked by the down-regulation of Nanog (10–12). To understand precisely how m⁶A depletion affects this property, we sought to define the dynamics of Nanog-state heterogeneity at single-cell resolution. We created double knock-in, protein-fusion Nanog and Oct4 TF mESCs and measured protein levels continuously over many cell generations. One allele of the endogenous Oct4 gene locus was targeted with a yellow (VENUS) FP (11), and one allele of the Nanog gene locus was targeted with a red (KATUSHKA) FP (Fig. 2A). The FPs are directly fused to the carboxy termini of TF’s and thus present functionally relevant protein levels in mESCs (11, 12) (SI Appendix, Fig. 2 A and B). After m⁶A depletion for 10 d, effects on Nanog^KATUSHKA expression were reproduced for imaging (Fig. 2B and SI Appendix, Fig. 2 C–I). The development of clonal colonies from single cells was monitored by microscopy (11, 12). Cells were marked by constitutive expression of infrared FP (iRFP) targeted to the nuclear membrane. Computer-assisted cell tracking and quantification of time-lapse imaging was performed (11, 37). To compensate for cell cycle–dependent fluctuations in protein intensity, a cell cycle–corrected distribution was calculated. Here, only the median of Nanog^VENUS intensity from the initial first three time points after a cell division was quantified. Using this unbiased approach, we generated protein intensity distributions unaffected by cell-cycle convolution (11) (Fig. 2 C and D).

Mother-to-daughter inheritance analysis showed that m⁶A-depleted, Nanog^KATUSHKA-negative daughter cells are four times less likely to reacquire Nanog^KATUSHKA expression (Fig. 2 D and E). Nearly all daughter cells inherit similar Oct4^VENUS levels from their mothers (SI Appendix, Fig. 2J). Cell survival or proliferation remain unchanged (38) (SI Appendix, Fig. 2 K and L). Lineage-tracing reveals that m⁶A-depleted cells inherit the Nanog^KATUSHKA-negative, Oct4^VENUS-positive state for many cell generations (Fig. 2 F and G). While on average, over six generations are required before half of the Nanog^KATUSHKA-negative cells re-express Nanog, control cells require less than two generations (Fig. 2 F and G). Likewise, m⁶A-depleted, Nanog^KATUSHKA-negative, Oct4^VENUS-positive cells are supported long-term, with low probability for cell-state departure (Fig. 2E). While either Nanog state accumulates, low-frequency dynamic interchange between these populations (Fig. 2E) suggests differential and opposing regulation downstream of m⁶A depletion. We examined whether these diverse cell-state dynamics stemmed from the activation of signaling pathways in m⁶A-depleted cells.

KEGG (Kyoto Encyclopedia of Genes and Genomes) analysis of m⁶A sequencing (m⁶A-seq) data indicated that mRNAs of FGF (fibroblast growth factor), MAPK (mitogen-activated protein kinase) and PI3K (Phosphoinositide 3-kinase) signaling regulators are m⁶A modified (SI Appendix, Fig. 3A) (27). m⁶A-depleted cells over 10 d show increased mRNA levels of Fgf5 and Fgf2 ligands but not Fgf4 (Fig. 3A and SI Appendix, Fig. 3 B and C). This corresponds to elevated Fgf5 mRNA stability (Fig. 3A). Downstream phosphorylation of the FGFR1 receptor upon m⁶A depletion (Fig. 3B) activates the Mek1/2 and Erk1/2 signaling cascade (39, 40) (Fig. 3 C and D). Multispectral, single-cell, protein-level quantification for Nanog and pErk shows that pErk activation levels are similar in Nanog-positive and -negative, m⁶A-depleted cell populations (Fig. 3 E and F and SI Appendix, Fig. 3D). pErk activation is therefore not a by-product of the increase in the proportion of Nanog-negative cells. Strikingly, inhibition of MAPK-signaling by PD0325901 reverses the accumulation of the m⁶A-depleted, Nanog^VENUS-negative cell fraction and reinstates Nanog^VENUS expression (Fig. 3 G and H). We recapitulated these effects on the Nanog^VENUS distribution by adding PD173074, a specific inhibitor of FGF receptors (20) (SI Appendix, Fig. 3E). PD173074 treatment effectively down-regulated pErk (SI Appendix, Fig. 3F). These findings show that Nanog^VENUS-negative cell-state accumulation is regulated by FGF-dependent Mek-pErk signaling in m⁶A-depleted cells.

We asked whether pErk activation in m⁶A-depleted cells promotes the expression of key TFs involved in Nanog down-regulation. Otx2 and Oct6 (also known as Pou3f1) are associated with Nanog depletion in mESCs (17, 41, 42) and the peri-implantation epiblast (43–46). Multispectral, single-cell, TF protein-level quantification shows that m⁶A depletion for 10 d is accompanied by an increase in Otx2 (Fig. 4A and SI Appendix, Fig. 4 A and C) and Oct6 (Fig. 4B and SI Appendix, Fig. 4 B and C) expressing cells. Mek-pErk appears central to the up-regulation of these TFs as MAPK inhibition by PD0325901 (Fig. 4C) or the blockade of FGF receptors by PD173074 (SI Appendix, Fig. 4D) depletes Otx2 and Oct6 levels. We asked whether Otx2 and Oct6 regulate the Nanog-negative cell state by knocking-down either one or both TFs simultaneously. Otx2 and Oct6 KD significantly reduced the Nanog^VENUS-negative cell fraction (Fig. 4 D–F and SI Appendix, Fig. 4E). The increase in Otx2 and Oct6 mRNA and protein levels also appears to be independent of mRNA stability in m⁶A-depleted cells (Fig. 4 G and H). Instead, pErk activation stimulates the up-regulation of epiblast TFs, promoting Nanog-negative cell-state accumulation.

Nanog repression is also a feature of the transient EpiLC state acquisition (19). However, in Serum/LIF conditions, epiblast identity is suppressed by the continued presence of LIF (16, 47). In FGF2-supplemented, serum-free medium used to support mEpiSCs, m⁶A depletion led to spontaneous differentiation (29). We asked whether decreasing m⁶A abundances also rendered the recently established FS epiblast cells more amenable to lineage commitment. FS cells are maintained by endogenous FGF activity and low nodal/activin supplementation. These cells are poised to enter mesendodermal differentiation upon increased levels of either signal or canonical Wnt (20). The FS state was induced from 2i/LIF pluripotency conditions for 10 d, followed by m⁶A depletion for a further 10 d. Marker analysis showed that Oct4 TF protein expression was unaffected, and levels of epiblast TFs Otx2 and Oct6 also remained unaltered (Fig. 4I). However, we observed pErk- and pAkt-signaling activation (Fig. 4J), while Wnt activity remained unchanged in m⁶A-depleted FS cells (SI Appendix, Fig. 4F). FS cultures could be expanded long-term and profiling of mesendodermal markers showed no significant up-regulation (Fig. 4K and SI Appendix, Fig. 4G). However, upon lineage induction of m⁶A-depleted cells, a significant enrichment for mesendodermal markers was found as indicated by Mesp1, Mixl1, and brachyury T at the mRNA (Fig. 4K) and protein levels (SI Appendix, Fig. 4H). Together, these observations indicate that m⁶A depletion activates pErk-signaling, promoting epiblast cell identity in Serum/LIF. Upon lineage induction in m⁶A-depleted FS cells, pErk and pAkt coactivation appears to increase the propensity for mesendodermal differentiation.

We asked whether m⁶A depletion also coactivates pAkt, promoting Nanog expression and pluripotency maintenance in Serum/LIF (13, 48). m⁶A-depleted cells showed a significant increase in pAkt activation (Fig. 5A) spanning the Nanog protein intensity distribution (Fig. 5B and SI Appendix, Fig. 5A). Nanog^VENUS-positive cells also displayed a higher pAkt activation, consistent with its role in supporting Nanog expression. PI3K inhibition by LY294002 administration resulted in a transient Akt dephosphorylation (SI Appendix, Fig. 5B) and a significant increase in the proportion of Nanog^VENUS-negative cells (Fig. 5 C and D). Furthermore, inhibition of pAkt increased pErk-mediated expression of Otx2 and Oct6 (SI Appendix, Fig. 5C). pAkt can also promote pluripotency by facilitating the nuclear accumulation of β-catenin (49), resulting in Wnt pathway activation (15, 50). This is achieved by pAkt-dependent phosphorylation and degradation of the Wnt repressor Gsk3β, as also shown herein (SI Appendix, Fig. 5D). Profiling of other signaling pathways indicates that m⁶A depletion does not affect signaling activation more broadly (SI Appendix, Fig. 5E). PI3K-pAkt and Mek-pErk regulators (51, 52) implicated in tumorigenicity (53) also remained unchanged (SI Appendix, Fig. 5F). Extended propagation in Serum/LIF reveals that the proportion of m⁶A-depleted Nanog^VENUS-negative cells becomes progressively reduced over time, likely due to the increased propensity for pluripotency exit (Fig. 5E). Meanwhile, Nanog^VENUS-positive cells gradually expand in culture. These results suggest that alongside increased pluripotency TF mRNA stability (Fig. 1 I and J), pAkt activation reinforces Nanog expression in m⁶A-depleted cells.

Stable m⁶A depletion by Mettl3 KO hampers the priming and differentiation competence of mESCs, leading to a “hyper-naïve” pluripotency phenotype. This is attributed to increased pluripotency TF mRNA stability of m⁶A-decorated transcripts (28, 29) (SI Appendix, Fig. 6 A and B). We profiled Nanog and Oct4 in two KO ESC lines by multispectral, single-cell, protein-level quantification. KO1 and KO2 cell lines differ in the proportions of Oct4-positive, Nanog-negative cells in Serum/LIF (Fig. 6A). KO1 with a greater Oct4-positive, Nanog-negative cell fraction has higher pErk activation compared to the other cell lines. This activation threshold appears to counteract elevated pluripotency TF mRNA stability and pAkt (Fig. 6 A and B). A comparable increase in pluripotency TF mRNA stability and lower pErk activation in KO2 cells results in a reduced proportion of Nanog-negative cells in Serum/LIF (Fig. 6 A and B). Pharmacological inhibition of pErk in 2i/LIF conditions reinstates homogeneous Nanog expression (Fig. 6C) and increases pAkt levels in m⁶A-depleted cells (Fig. 6D and SI Appendix, Fig. 6C). Induction of differentiation by retinoic acid treatment of wild-type (WT) and KO cells shows that KO1 with higher pErk activation propensity differentiates more efficiently than KO2 (Fig. 6E). This is characterized by a lower proportion of Oct4 and Nanog double-positive cells remaining after 48 h of retinoic acid treatment in KO1 relative to KO2. These findings show that activation of pErk-signaling in stably m⁶A-depleted cells counteracts the pluripotency exit delay imposed by stable m⁶A depletion.

mESCs from the R1 parental line (129 substrain) were electroporated with linearized Oct4-Venus-IRES-neo targeting vector and selected with G418 (200 µg/mL, catalog no. G8168-10ML, Invitrogen). Resistant ES clones were confirmed by Southern blotting. Oct4^VENUS heterozygous fusion mESCs underwent a second round of electroporation with the Nanog-Katushka-2A-BSD targeting vector and were selected with Blasticidin (5 µg/mL, catalog no. A11139-02, Invitrogen). The Nanog^VENUS line, the parental R1 ESCs, and the E14TG2a ESCs were kindly obtained from K.A.’s laboratory (BIOTEC, Technische Universitaet Dresden). Professor Howard Y. Chang kindly provided J1 (129 substrain) Mettl3 KO cell lines (KO1 and KO2) and wild-type control cells for this study grown in Serum/LIF as previously published (28).

Electroporation was conducted as previously described (56). Antibiotic-resistant colonies were picked, expanded in 96-well plates, and DNA was extracted by Proteinase K (0315801001, Roche) digestion and NaCl–ethanol precipitation. Genomic DNA was digested overnight with the indicated enzyme(s), separated on 0.8% agarose gels, and blotted to nylon membranes (BNAZF810S, PALL). Probes were labeled with ³²P by random priming (11585592001, High Prime Kit, Roche). Hybridization and washes were conducted at 65 °C.

mESC lines were maintained in Fetal Bovine Serum (FBS)–containing medium and feeder-free conditions as described previously (11). Retinoic acid–induced differentiation was initiated by removal of LIF and addition of 1 μM retinoic acid (Sigma-Aldrich, R2625) 24 h after cell seeding for 2 d in FBS-containing ESC medium. For 2i/LIF ground-state conditions, mESCs were cultured on Fibronectin-coated plates or flasks (Nunc) in N2B27-based, FBS-free medium supplemented with MEK inhibitor PD0325901 (1 μM) and GSK3β inhibitor CHIR99021 (3.3 μM). In 2i/LIF, mESCs were passaged by dissociating with TrypLE (Gibco, 12605010). mESCs originally cultured in FBS-containing medium were adapted to 2i/LIF conditions for at least five passages before further experimentation. For inhibitor assays in FBS/LIF (Figs. 3–5), the following concentrations of inhibitors PD0325901 (1 μM), PD173074 (0.8 μM), or LY294002 (5 μM) were used. Thereafter, cells were collected for protein extraction followed by immunoblot analyses for designated protein targets.

Mouse FS cells were obtained from R1 ESCs in N2B27-based 2i/LIF conditions by culturing in A_loXR medium (20) for at least five passages. Briefly, A_loXR media contains activin A (3 ng/mL, R&D Systems), XAV939 (2 μM, Selleck), and BMS493 (1 μM, Selleck) in N2B27 basal medium. FS cell clumps were dissociated with Accutase (A1110501, Gibco) and passaged on Fibronectin (Merck Millipore, FC010)–coated plates every 2 to 3 d. Medium was replaced daily. For mesendodermal differentiation, FS cells were dissociated, washed, and seeded on Fibronectin-coated plates in N2B27 basal medium containing 20 ng/mL activin A and 3 μM CHIR99021 (Selleck) for 24 h (20).

Viral transduction methodology and short hairpin RNA (shRNA) sequences obtained from the Broad Institute’s The RNA interference (RNAi) Consortium are described in the SI Appendix.

siRNA transfection methodology and siRNA sequences used in this study are described in the SI Appendix.

Total proteins were extracted by radioimmunoprecipitation assay (RIPA) lysis buffer containing the protease inhibitor mixture (Sigma-Aldrich, P8340) and phosphatase inhibitor mixtures (Sigma-Aldrich, P5726; P0044). Protein concentration of the lysate was quantified with Bio-Rad Bradford (Bio-Rad, 5000006) assays. Protein samples were denatured and resolved by Bolt Bis-Tris Plus Gels (Invitrogen). Separated proteins were transferred onto nitrocellulose membranes then detected with primary antibodies against Mettl3 (Abcam ab195352), Mettl14 (Sigma-Aldrich, HPA038002), Wtap (Proteintech, 60188–1-Ig), Oct4 (Santa Cruz, sc-8628), Nanog (Abcam ab80892), Oct6 (Merck Millipore, MABN738), Otx2 (R&D systems, AF1979), GSK-3β (Cell Signaling, 9315S), pGSK3β (Ser9) (Cell Signaling, 9323S), β-catenin (Santa Cruz, sc-7199), pβ-catenin(Ser33/37/Thr41) (Cell Signaling, 9561S), Phlpp2 (Abcam, ab71973), Smad1/5/9 (Abcam, ab66737), pSmad1(Ser463/465)/Smad5(Ser463/465)/Smad9 (Ser465/467) (Cell Signaling, 13820S), Ras (Abcam, ab52939), Smad2/3 (Cell Signaling, 8685S), pSmad2(Ser465/467)/Smad3(Ser423/425) (Cell Signaling, 8828S), mTOR (Cell Signaling, 2983S), pmTOR(S2481) (Abcam, ab137133), Eras (Abcam, ab192868), Pten (Cell Signaling, 9559S), PP2A C subunit (Cell Signaling, 2038S), Akt (Cell Signaling, 4691S), pAkt(Ser473) (Cell Signaling, 4060S), pAkt(Thr308) (Cell Signaling, 13038S), c-Raf (Cell Signaling, 9422S), pc-Raf(Ser338) (Cell Signaling, 9427S), Erk1/2 (Cell Signaling, 4348S), pErk1/2 (Thr202/Tyr204) (Cell Signaling, 9101S), Mek1(Abcam, ab32091), Mek2 (Abcam, ab32517), pMek1/2 (Ser217/221) (Cell Signaling, 9154S), Fgfr1 (Abcam, ab10646), pFgfr1(Tyr653/Tyr654) (Merck Millipore, 06–1433), Histone H3 (Abcam, ab85869), and GAPDH (Abcam, ab125247). The following secondary antibodies were used: Donkey anti-mouse horseradish peroxidase (HRP) (Abcam, ab6820), donkey anti-rabbit (HRP) (Abcam, ab6802), and donkey anti-Goat (HRP) (Thermo Fisher Scientific, PA1-28664). Blots were developed by enhanced chemiluminescence (ECL) (Thermo Fisher Scientific 32209 and 34095) and scanned by a Bio-Rad ChemiDoc XRS + system.

The integrated optical density (IOD) of immunoblot bands was measured by Gel-Pro analyzer software (Rockville). Generally, the IOD of target bands was firstly divided by the IOD of corresponding loading-control (for example, GAPDH) bands to normalize the sample loading-dependent variation between different lanes. Then, normalized quantification of target bands from all sample lanes were divided by the normalized value of target bands from the corresponding control sample lane. The resulting ratios were used as relative abundances of target proteins across different samples. The IOD of phosphorylated protein bands were divided by the IOD of their corresponding total-protein bands. The resulting values were used to calculate relative abundance of phosphorylated protein as previously mentioned. Each analysis for relative protein abundance was performed independently for at least three experimental repeats.

mESCs were dissociated with 0.05% trypsin (Gibco, 25300054) or TrypLE. Single-cell suspensions were obtained by resuspension of the cell pallet with phosphate-buffered saline (PBS) containing 0.5% bovine serum albumin (BSA) and filtration with a 40-μm cell strainer. All samples were run on a BD LSRFortessa cell analyzer. The blue 488-nm laser and GFP detector was used to analyze the VENUS fluorescence signal. The yellow-green 561-nm laser and PE-Cy5 detector was used to analyze KATUSHKA fluorescence signal. The red 640-nm laser and Alexa Fluro 700 detector was used to analyze iRFP713 fluorescence signal. Background signal from R1 unmodified wild-type mESCs on each channel was used to align each negative gate at 10³ signal intensities.

For intracellular flow cytometry analysis, cells were dissociated and then fixed using 1.5% formaldehyde (Santa Cruz, sc-281692) for 10 min at room temperature. Fixed cells were permeabilized by resuspension in 100% methanol overnight at −20 °C. Permeabilized cells were washed twice by staining buffer (PBS with 0.5% BSA and 0.02% NaN₃) and then incubated with 10 µL of Alexa Fluor 647 conjugated mouse anti-pAkt (S473) (BD Bioscience, 560343) or 5 µL of mouse IgG1 κ Isotype Control (BD Bioscience, 557714) in 50 µL staining buffer for 60 min at 4 °C. Stained cells were washed five times with staining buffer before running on a BD LSRFortessa cell analyzer.

Isolation of Nanog^VENUS fluorescent cell subpopulations in Scr or shM3 conditions was performed on BD FACS (Fluorescence-activated cell sorting) Aria II cell sorter with a BD FacsDiva software package. In order to obtain enough cells with high purity for immunoblot analysis, we applied two rounds of sorting. The first sorting was carried out on “Yield” mode to get a preliminary enrichment of the target cell population. Preliminarily sorted cells were resorted with “Purity” Mode to attain high purity of the target cell population. The purity (more than 98%) of each sorted population was confirmed immediately after sorting. FACS analyses were performed on at least three independent experiments.

Total RNA was isolated using TRIZOL (Invitrogen, 15596026) according to the manufacturer’s instructions. Polyadenylated mRNA was purified by Dynabeads mRNA Purification Kit (Invitrogen, 61006). Purified mRNA was used for m⁶A dot blot with a Bio-Dot Apparatus. In brief, mRNA was denatured and applied to rehydrated Hybond-N⁺ membrane (GE, RPN203B) within Bio-Dot Apparatus by gentle suction vacuum. The membrane was then cross-linked in an ultraviolet (UV) light crosslinker BLX-E254nm (Vilber, France) for 5 min, followed by baking at 80 °C for 30 min. The blotted membrane was blocked with 5% nonfat milk and incubated with primary rabbit anti-m⁶ A antibody (Synaptic Systems, 202003) and followed by secondary donkey anti-rabbit (HRP) (Abcam, ab6802). Blots were developed by ECL (Thermo Fisher Scientific 34095) and scanned by Bio-Rad ChemiDoc XRS+ system. The signal density of the dot blot experiment is quantified by Gel-Pro analyzer software (Media Cybernetics) in all experiments. The relative m⁶A abundance was calculated as the ratio of dot intensities between shM3 and Scr. Methylene blue staining was used as RNA loading control.

qRT-PCR was used to assess the relative abundance of mRNA. Methodology and primers used for qRT-PCR are described in the SI Appendix.

Control, Mettl3 KD, or Mettl3 KO cells were seeded in 6 cm dished at 50% confluency for 24 h. Actinomycin D (Sigma-Aldrich, A1410) was added to culture medium at 5 μg/mL for 6 h, 3 h, and 0 h before collection. The total RNA was purified and applied for RT-qPCR analysis to examine mRNA abundance. The degradation rate of RNA was estimated as previously shown (53).

To attain feeder-free monolayer mESC culture, cells were grown on circle glass coverslips coated with Recombinant Human E-Cadherin (25 µg/ml ⁻¹, Primorigen, S2071-500UG) according to the manufacturer’s instructions in 24-well plates for 2 d. ESCs were then rinsed briefly and fixed with 4% paraformaldehyde (Santa Cruz, sc-281692) in PBS at room temperature for 5 min. After permeabilization with PBS containing 0.25% Triton X-100 (For pErk1/2 and pStat3 staining, cells were permeabilized with ice-cold 100% methanol for 10 min at −20 °C, followed by rinse in 1× PBS for 5 min) and blocking with 5% donkey serum for 30 min, the coverslips were incubated with a primary antibody combination at optimized dilutions overnight at 4 °C. The next day, coverslips were washed with PBS four times and incubated with the secondary antibody combination at room temperature for 1 h. For triple fluorescence staining, three primary antibodies raised in different species were used, followed by compatible secondary antibodies, conjugated with fluorophores: Alexa Fluor 488, 555, or 647. Negative control-staining was done using secondary antibodies only. After secondary antibody incubation, coverslips were washed five times and counterstained with antifade Mounting medium containing DAPI (Vector Laboratories, H1200). Images of stained cells were then acquired by wide-field microscopy.

Cells seeded in 8-well µ-Slides (Ibidi, 80826) were used to estimate the correlation between FP-fusion and total-protein levels. Imaging of Nanog^KATUSHKA and Oct4^VENUS FP-fusions in live cells was first done followed by in situ immunostaining of fixed cells. Exposure time used for immune-staining yielded no fluorescence protein detection, ensuring no interference from FP-fusion signal during the imaging of immuno-stained samples. Spatial shifts in image x, y between DAPI images from staining and live-cell imaging of corresponding Nanog^KATUSHKA or Oct4^VENUS signals were realigned by manual correction using Photoshop software.

The following antibodies were used: goat anti-Oct-3/4 (Santa Cruz, sc-8628), rabbit anti-Oct3/4 (Abcam, ab19857), rabbit anti-Nanog (Abcam ab80892), mouse anti-Nanog (BD Pharmingen, 560259), mouse anti-Sox2 (Abcam, ab79351), mouse anti-Esrrb (R&D systems, PPH6705-00), mouse anti-Klf4 (Abcam, ab75486), rabbit anti-Tbx3 (gift from Prof. Hitoshi Niwa), mouse anti-Oct6 (Merck Millipore, MABN738), goat anti-Otx2 (R&D systems, AF1979), rabbit anti-pErk1/2 (Thr202/Tyr204) (Cell Signaling, 9101s), Stat3 (Cell Signaling, 12640S), pStat3(Tyr705) (Cell Signaling, 9145S), and goat anti-T (1/200, R&D, AF2085). Alexa Fluor 488–conjugated donkey anti-mouse (Invitrogen, A-21202), donkey anti-rabbit (Invitrogen, A21206), Alexa Fluor 555–conjugated donkey anti-mouse (Invitrogen, A31570), donkey anti-rabbit (Invitrogen, A31572); Alexa Fluor 647–conjugated donkey anti-mouse (Invitrogen, A31571), donkey anti-mouse (Invitrogen, A-21447) were used as secondary antibodies.

The following filters sets were used: Filter Set 38 HE (Zeiss, 489038–9901-000) for the 488-nm channel, mCherry hardcoated (HC) Filterset (AHF, F36-508) for the 555-nm channel, Cy5 ET Filterset (AHF, F46-006) for the 647-nm channel, and Filter Set 49 (Zeiss, 488049–9901-000) for the DAPI channel.

Staining images were firstly background corrected with the BaSiC (57) package for quantitative analyses. Nuclei areas were determined and segmented according to DAPI-staining by a custom-made Fiji script. For pErk1/2-staining analysis, manual segmentation was used to determine cytoplasmic areas for quantification (SI Appendix, Fig. 4L). The intensity of fluorescence signals within segmented areas for individual cells were quantified using the program Fiji. Data processing and visualization of staining quantifications were done with RStudio (R version 3.6) and the ggplot2 package. Cells within a range of 40 to 650 px in nuclear area were analyzed to reduce segmentation artifacts. The original quantified intensities were transformed by Log10 function for final plotting.

In total, 2,000 ESCs from Scr or shM3 conditions were seeded in one well of 96-well µ-plates (Catalog no. 89626, Ibidi) coated with Recombinant Human E-Cadherin (50 µg/mL ⁻¹, Primorigen, S2071-500UG). After seeding, plates were leveled on the microscope stage with 5% CO₂ using a self-developed continuous gas delivery lid system and cultured at 37 °C. Imaging was initiated as soon as position grids were established in each well. Imaging was conducted on previously mentioned wide-field microscope (Zeiss), with a pE-4000 CoolLED light engine, Micro-Manager Version 2.0 beta control software, Hamamatsu camera, and a 10× FLUAR objective (Zeiss). Images were acquired on live cells at 10 ms in bright-field and 250 ms for the VENUS channel using a yellow fluorescence protein (YFP) Filter Set (AHF, F46-003), 250 ms for the KATUSHKA channel using an mCherry HC Filter Set (AHF, F36-508), and 50 ms for the nuclear marker iRFP channel using a customized iRFP filter (AHF). Continuous time-lapse imaging of mESCs was conducted for a minimum of 4 d. Channels were imaged continuously at 30-min intervals per position. All experiments were performed over three independent experimental repeats.

Single-cell tracking was performed as previously described (11). Briefly, tracking was done by using “the tracking tool” (TTT) program (37) on a desktop computer with 32 GB of random access memory (RAM), Quad-core central processing unit (CPU), and Windows 64-bit operating systems. Individual cells from a selected colony were identified and tracked through manually evaluating every time point of iRFP channel in a time-lapse movie. All relevant properties and behavior (division, death, and migration) of cells of interest were marked, stored, and displayed in pedigrees. All cell tracking was done manually; only cells with unequivocal identity that could clearly be identified when evaluating the video were used for analysis and all cells with questionable identity were excluded from analyses. “the tracking tool”

Fluorescence image normalization was done as previously described (57). Briefly, after subtracting the estimated time-dependent background and correcting uneven illumination, normalized images were used for fluorescence-intensity quantification of tracked single cells. The software package QTFy (37) was used to semiautomatically quantify the nuclear signal of tracked cells, resulting in normalized intensity time traces. Nuclei were first automatically segmented based on the nucmem-iRFP signal and further corrected manually by referring to multichannel images. Oct4^VENUS or Nanog^KATUSHKA signals were then quantified electronically from nuclear segmentation.

To compensate for cell cycle–dependent fluctuations in signal intensity, cell cycle–corrected values were used for distributions and cell-state transition analyses. To this end, only the initial Nanog^KATUSHKA or Oct4^VENUS intensities were considered (that is, median of first three time points) for each single cell. For heat-tree visualization in Fig. 2H, the absolute Oct4^VENUS or Nanog^KATUSHKA intensities in single-cell trajectories were cell cycle corrected by nuclear volume, as previously shown (11). The resulting intensity concentrations over the entire cell lifetime were shown in heat trees.

For Nanog expression, an intensity of Log10 1.4 a.u. (arbitrary units) nicely separated the bimodal Nanog^KATUSHKA intensity distribution in all three Mettl3 KD experiments. For Oct4 expression, the negative gate was set to Log10 1.7 with ∼5% of Oct4^VENUS-negative cells being present across all three independent experiments, reflecting FACS analyses. For heat-tree visualization in (Fig. 2H), cells visually negative in time lapse throughout their lifetimes were quantified to determine the electronic negative gate threshold for Nanog^KATUSHKA and/or Oct4^VNEUS intensity concentrations.

Mettl3 and Mettl14 target gene list from the SI Appendix, Table 1 in Wang’s publication (27) were used for KEGG pathway enrichment analysis. The KEGG enrichment analysis was done by RStudio (R version 3.6) and the cluster-Profiler package (58). Top 30 enriched pathways were shown in a dot plot in SI Appendix, Fig. 3A.

At least three independent experiments were done for each assay. Data are presented as the mean ± SD. Two-tailed Student’s t tests assuming unequal variances were performed to assess the statistical significance of differences between groups. Pearson correlation coefficients (r) were calculated to assess correlation. Statistical significance (P) was assessed by a two-tailed t test of r = 0. Immunoblots shown are the representative images of at least three independent experiments. For box plots, the center line represents the median, the box limits show the upper and lower quartiles, whiskers represent 1.5 times the interquartile range, and outliers are represented as individual data points.