Genes Menin and SUZ12 may control the timing of human development

To screen for PAX6 expression in live cells, we first integrated an inducible Cas9 gene, which expresses Cas9 following doxycycline treatment, into the AAVS1 locus in the H9 PAX6::H2B-GFP hESC reporter cell line^23,25 (Extended Data Fig. 1a,b). Cas9 protein can be efficiently induced within two days upon doxycycline treatment, and the iCas9 integration does not affect PAX6 expression dynamics (Extended Data Fig. 1c–e). To screen for regulators of differentiation speed at a whole-genome scale, we infected PAX6::H2B-GFP iCas9 hESCs with a genome-scale, human CRISPR KO Brunello lentiviral pooled library²⁶. hESCs were selected for stable gRNA integration with puromycin, and mutations were induced with doxycycline. We then started neuroectoderm differentiation using the dual SMADi method (Fig. 1d). Given that PAX6 expression starts to rapidly increase around day 3 to 4 (Fig. 1b and Extended Data Fig. 1e), and thus the acceleration effect of a mutation, if any, would be most pronounced during this time window, we isolated PAX6-GFP^high and PAX6-GFP^low cells using fluorescence-activated cell sorting (FACS) at the timepoint when we observed the PAX6 level started to increase (Fig. 1e). One replicate was collected at 72 hours (3 days), and two replicates were collected at 84 hours (3.5 days). The gRNA representation in each population was determined by next-generation sequencing. We reasoned that if a gene KO accelerates the differentiation, the gRNA should be repeatedly enriched in the PAX6-GFP^high population and depleted in the PAX6-GFP^low population. By contrast, if a gene KO has no effect on speed, the gRNA should be randomly distributed between the two populations.

The analysis workflow is illustrated in Extended Data Fig. 1f. First, we compared gRNAs enriched in PAX6-GFP^high versus PAX6-GFP^low populations and ranked the genes by Z-score²⁷. Gene-set enrichment analysis (GSEA)²⁸ on the ranked gene list showed an enrichment for multiple chromatin remodelling complexes, such as ATAC, SET1C, npBAF and MLL complexes (Fig. 1f and Extended Data Fig. 1g,h), as well as for pathways associated with mitochondrial metabolism, particularly the tricarboxylic acid (TCA) cycle (Fig. 1f, Extended Data Fig. 1g,i and Supplementary Table 1). This is in line with previous studies highlighting both the chromatin landscape and metabolic activity as key regulators of developmental dynamics^29,30. We then calculated Z-score and P-value using edgeR³¹ for each gene and defined 27 screen hits (Z-score > 1, P < 0.05) (Fig. 1g and Supplementary Table 2). To validate the screen hits and gain insights into the molecular pathways that affect differentiation speed, we adopted both genetic and pharmacological approaches. For the genetic validations, we followed up on the screen hits by infecting PAX6::H2B-GFP iCas9 hESCs with individual lentiviral gRNA targeting each candidate gene, followed by independent differentiation assays. PAX6 level was assessed by GFP live imaging. Among the 27 screen hits, SUZ12 and MEN1 KOs showed the highest fold change of PAX6 expression compared with non-targeting (NT) control (Extended Data Fig. 2a). For the pharmacological approaches, we performed co-essentiality mapping using DepMap database^32,33 and clustered the screen hits based on their essentiality score correlation. The co-essentiality mapping allowed us to infer functional pathways shared by a subset of the screen hits. One prominent cluster emerged from the analysis that contained 5 out of 27 candidates, including members of Polycomb Repressive Complex (PRC) 1 and 2 (BCOR; SUZ12, EED), TCA cycle (DLST), and Menin–MLL complex (MEN1) (Extended Data Fig. 2b). Accordingly, we selected small molecule inhibitors targeting each pathway or interaction and treated hESCs undergoing differentiation (Supplementary Table 3). PAX6 level was assessed at day 4 by flow cytometry. Of all the pathways perturbed, inhibition of PRC1, PRC2 and Menin–MLL interaction showed the highest fold change of PAX6 expression compared to DMSO-treated cells (Extended Data Fig. 2c). Pairwise combinations of these three perturbations led to an additive effect on accelerating neural markers PAX6 and ZBTB16 (Extended Data Fig. 2d).

Taken together, genetic and pharmacological secondary validation experiments underscored MEN1 and SUZ12 as promising candidates for regulating the timing of neuroectoderm differentiation. Intriguingly, Menin and SUZ12 are postulated to have opposing functions on gene expression, namely, Menin/MLL1-mediated histone 3 lysine 4 trimethylation (H3K4me3), an active chromatin mark^34,35, and PRC2-mediated histone 3 lysine 27 trimethylation (H3K27me3), a repressive chromatin mark^36–38, yet both KOs led to accelerated PAX6 expression. Based on these results, we focused on MEN1 and SUZ12 for further investigation.

Based on this observation, we further examined the specificity and extent of transcriptional changes in MEN1 and SUZ12 KO cells. We first assessed dynamic patterns of transcriptional changes during the natural differentiation by performing TCseq⁴⁰ using the temporal data of NT cells. This time-course analysis revealed eight distinct gene expression patterns (clusters) underlying neuroectoderm differentiation. Genes in cluster 1 and 4 showed a monotonic upregulated expression and were strongly enriched for forebrain development (Fig. 2c and Extended Data Fig. 3d). Genes in cluster 7 and 8 showed monotonic downregulated expression and were linked to ribosome and mitochondria-related cellular processes (Fig. 2d and Extended Data Fig. 3e). The remaining clusters showed transient expression patterns, with genes peaking at various timepoints during differentiation (Extended Data Fig. 3f). We then mapped the expression level of genes in each cluster in MEN1 and SUZ12 KO cells side-by-side with NT controls (Fig. 2e,f and Extended Data Fig. 4a). Both KOs accelerated the upregulation of cluster 1 and 4 genes at day 4 (Fig. 2g), whereas only MEN1 KO was also able to promote downregulation of cluster 7 and 8 genes (Fig. 2h). qRT-PCR analysis using a finer time course showed a similar trend to the RNA-seq data, indicating faster upregulation of neural genes in the KOs (Extended Data Fig. 4b). The degree of acceleration was similar between PAX6-bound⁴¹ versus non-PAX6-bound genes (Extended Data Fig. 4c), and PAX6 loss or gain of function did not alter the expression dynamics of other neural markers (Extended Data Fig. 4d–i).

Together, these findings indicate that loss of MEN1 or SUZ12 accelerates the progression of gene regulatory networks driving neuroectoderm differentiation, mostly through accelerating upregulation of forebrain development-associated genes. In the cases of monotonic down or transient expressions, we observed a reduced ability of SUZ12 KO to downregulate gene expression, but the directionality of changes and the overall patterns were preserved (Fig. 2f and Extended Data Fig. 3f).

During DE differentiation, NT cells showed similar expression dynamics of the definitive endoderm markers GATA6 and CXCR4 (refs. ^54,56), compared to the H9 Dendra2 control cells. MEN1, SUZ12 and double KO cells showed a marked increase in GATA6 at day 1 and CXCR4 at day 2 (Fig. 5c and Extended Data Fig. 7c,d). In addition to genetic KOs, we explored the functional relevance of Menin–MLL1 interaction and PRC2 by using two inhibitors validated in the neuroectoderm secondary screen (Extended Data Fig. 2c): VTP50469 (ref. ⁵⁷), a Menin–MLL1/2 inhibitor (MEN1i), and Tazemetostat⁵⁸, an EZH2 inhibitor functionally targeting PRC2 (EZH2i), respectively. In line with the genetic KOs, MEN1i and EZH2i treatment during differentiation accelerated expression of the endoderm markers GATA6 and CXCR4 (Fig. 5d and Extended Data Fig. 7f). EZH2i or SUZ12 KO also promoted SOX17 expression (Extended Data Fig. 7e,g). Notably, the combination of MEN1i and EZH2i resulted in a further increased expression of DE markers. qRT-PCR on DE markers (EOMES, GATA6, SOX17 and FOXA2) at the onset of gene induction recapitulated the trend of elevated expression in Menin- and SUZ12-targeted cells (Extended Data Fig. 7h).

For CM differentiation, qRT-PCR confirmed that cells progressed through the following differentiation stages: cardiac mesoderm (EOMES, MESP1), cardiac progenitor (PDGFRA, TBX5) and CM (TNNT2, MYH6)⁵⁹. Cells treated with MEN1i or EZH2i appeared to move more quickly through those stages with a reduction in cardiac mesoderm markers by day 3 and increased expression of cardiac progenitor and CM markers from day 3 onwards (Extended Data Fig. 8a). To assess the functional property of CM, hESCs were seeded in a multi-well plate and treated with MEN1i and/or EZH2i for the duration of differentiation. CM contraction can typically be detected around days 7–9, so we examined the percentage of wells that exhibit contraction during this time window. While we observed contraction in all control wells around day 9, MEN1i or EZH2i treatment induced premature contraction by day 8, and the combined inhibition by day 7 (Fig. 5e). In line with the contraction phenotype, we detected a higher percentage of cells expressing CM-specific cell-surface marker, SIRPA (ref. ⁶⁰), in KOs and drug-treated cells at day 7 (Fig. 5f). While drug treatment did not further increase the percentage of cells expressing cardiac troponin (cTnT, encoded by TNNT2) at day 7—which was already around 75%—the treatment did further elevate the cTnT expression level in the cTnT+ population (Extended Data Fig. 8b–d). Combined inhibition further enhanced the premature contraction and increased marker expression observed in single drug inhibition. These results indicate that, in addition to neuroectoderm, Menin and SUZ12 regulate the speed of transition from ESCs to endoderm and a mesoderm-derived lineage as well.

Next, we explored whether Menin and SUZ12 can further affect timing at a later developmental stage, independent of lineage specification from ESC stage. We focused on neurogenesis and induced genetic KO only after the differentiating cells have reached neuroectoderm stage. To this end, we infected iCas9 hESCs with gRNAs and did not induce Cas9 until day 10 (Extended Data Fig. 9a,b). For the pharmacological approach, MEN1i or EZH2i was applied starting at the neural progenitor stage (from day 20 onward). We probed neurogenesis speed by assessing the transition from SOX2+ neural progenitor to post-mitotic neuron, marked by HuC/D and MAP2. Both genetic and pharmacological manipulations led to a higher percentage of HuC/D+ or MAP2+ neurons and a further reduction of SOX2+ progenitors at day 25. Double KO or combined MEN1i and EZH2i treatment further enhanced this impact (Fig. 5g–i and Extended Data Fig. 9c–e).

Overall, chemical and genetic inhibition of Menin and SUZ12 produced similar phenotypes in all differentiation paradigms tested, suggesting that Menin–MLL interaction and EZH2–PRC2 play a crucial role in timing regulation across lineages and developmental stages. SUZ12 KO appears to achieve a higher degree of accelerated gene expression than EZH2i in neuroectoderm, DE and neuronal differentiation, which could be explained by incomplete depletion of H3K27me3 by EZH2i (Extended Data Fig. 9f,g) and potential EZH1-PRC2 activity⁶¹. In addition, MEN1i or EZH2i did not induce detectable cell fate changes, as evidenced by similar expression levels of markers at the differentiation endpoint compared to control (Extended Data Fig. 9h).

Finally, we explored whether the functions of Menin and SUZ12 are conserved in another species. We differentiated mouse epiblast stem cells (mEpiSCs) into neuroectoderm with MEN1i or EZH2i treatment. The dynamics of Pax6 and Zbtb16 remained largely unaffected, and the downregulation of E-cadherin was only mildly accelerated (Extended Data Fig. 9i). However, we observed both Pax6-negative non-neural cells and Map2-positive neurons in addition to Pax6-positive neuroectoderm, complicating direct timing comparison using mouse cells (Extended Data Fig. 9j). Furthermore, despite the broad overlap of H3K4me3 peaks between hESCs and mEpiSCs, Menin's binding partner, MLL1, exhibited distinct binding patterns (Extended Data Fig. 9k), raising the possibility that Menin's functional role differs between mouse and human cells during primed pluripotency.

All work using human pluripotent stem cells (hPSCs) was approved by and conducted in compliance with the Tri-Institutional Stem Cell Initiative Embryonic Stem Cell Research Oversight Committee (Tri-SCI PSCRO) at Memorial Sloan Kettering Cancer Center, Rockefeller University and Weill Cornell Medicine. hPSC line WA09 (H9; 46XX) was from WiCell Stemcell Bank. The PAX6::H2B-GFP line was generated previously and derived from WA09 hPSCs²³. The PAX6::H2B-GFP iCas9 line was derived from the PAX6::H2B-GFP line. hPSCs were maintained in Essential 8 media (E8; Life Technologies, #A1517001) in feeder-free conditions on vitronectin-coated plates (VTN-N; Thermo Fisher, #A14700). hPSCs were passaged as clumps every 3–4 days with EDTA (0.5 M EDTA/PBS) and routinely tested for mycoplasma contamination.

PAX6::H2B-GFP iCas9 knock-in hPSCs line was generated using CRISPR/Cas9-mediated homologous recombination by transfecting PAX6::H2B-GFP hPSCs with Hygro-Cas9 donor plasmid and AAVS1-Neo-M2rtTA (Extended Data Fig. 1a). NT, MEN1 KO, SUZ12 KO, and double KO hPSC lines were generated by transfecting PAX6::H2B-GFP iCas9 hPSCs with the following gRNAs: NT, AGCGCAGATAGCGCGTATCA; MEN1, CCAGGCATGATCCTCAGACA; SUZ12, TTCTCTATAACAACAATACA. For RNA-seq, polyclonal populations with stably integrated gRNAs were selected using puromycin and underwent doxycycline-induced mutations. For ATAC-seq, CUT&RUN and lineage differentiation assays, selected clones with gRNA integration were validated by genomic PCR and screened for Karyotype banding. To minimize variability due to long-term culture of edited cells, mutations were induced before each differentiation assay by 2-day treatment of 1 μg ml⁻¹ doxycycline. Protein loss was confirmed by western blot. H9 Dendra2 line was generated by transfecting H9 hPSCs with Dendra2-Neo donor plasmid. Selected clones were validated by flow cytometry.

To induce neuroectodermal fate, hESCs were dissociated into single cells using Accutase (Innovative Cell Technologies, #AT104-500) and plated at 300,000 cells per cm² onto Matrigel-coated wells (Corning, #354234) in E8 medium, supplemented with 10 μM Y-27632 (R&D, #1254) (day −1). On the following day (day 0), culture medium was replaced with Essential 6 medium (E6; Thermo Fisher Scientific, #A1516401) supplemented with 10 μM SB431542 (Tocris, #1614) and 100 nM LDN193189 (Stem Cell Technologies, #72142). Differentiation medium was changed daily from day 1 to 9.

Generation of neurons continues as following. From day 10 to 20, cells were cultured in 1:1 Neurobasal (Thermo Fisher, #21103049) and DMEM/F12 (Thermo Fisher, #11330-032) medium supplemented with 1X N2 (Thermo Fisher, #17502048), 1X B27 without vitamin A (Thermo Fisher, #12587010) and Penicillin Streptomycin (Pen/Strep; Thermo Fisher Scientific, #15140-122) to generate a neurogenic population of cortical NPCs. On day 20, cells were either cryopreserved in STEM-CELLBANKER freezing media (Amsbio #11924) at approximately ten million cells per vial, or induced for neurogenesis as following: NPCs were dissociated using Accutase for 30 min and replated at a density of 120,000 cells per cm² in plates coated with poly-l-ornithine (PO; Sigma-Aldrich, #P3655), laminin (LAM; R&D, #3400-010-1) and fibronectin (FN; Thermo Fisher, #356008). After replating, cells were cultured in low-glucose neuronal culture medium, containing 1:9 Neurobasal and Neurobasal-A medium (Thermo Fisher, #A2477501) supplemented with 1X B27 without vitamin A, 1X GlutaMAX Supplement (Thermo Fisher, #35050061), Pen/Strep, 20 ng ml⁻¹ BDNF (R&D, #248-BD), 20 ng ml⁻¹ GDNF (PeproTech, #450-10), 200 µM cyclic monophosphate sodium salt (cAMP; Sigma-Aldrich, #D0627) and 200 µM ascorbic acid (AA; Sigma-Aldrich, #4034-100). During replating, 10 μM Y-27632 was added to increase cell viability. Half of the medium was replaced every other day.

For experiments in mouse, B6.129_4 mouse epiblast stem cells (mEpiSCs) were cultured on Mouse Embryonic Fibroblast feeders (Thermo, #A34181) in mouse N2B27 (ref. ⁸⁰) supplemented with 12.5 ng ml⁻¹ FGF2, 20 ng ml⁻¹ Activin A and 175 nM NVP-TNKS656. mEpiSCs were differentiated to neuroectodermal fate as follows: on day 0 mEpiSC colonies were lifted from feeders using 0.5 U µl⁻¹ collagenase IV in HBSS++, dissociated to single cells using Accutase, then plated at 400,000 cells per cm² on Matrigel-coated plates in mouse N2B27 supplemented with CEPT cocktail (50 nM chroman I, 5 µM Emricasan, 1:1,000 Poly amines, 0.7 µM Trans-ISRIB), 100 nM LDN193189, 10 µM SB431542, 200 nM LGK-974 and 2 µM cyclopamine. On days 1–3, cells were fed daily with the same media as day 0, excluding the CEPT cocktail. On days 4–6, cells were fed with mouse N2B27 supplemented with 2 µM cyclopamine only.

The Brunello human CRISPR Knockout Pooled Library²⁶ was used for this screen. Three hundred million PAX6::H2B-GFP iCas9 hESCs were infected with the lentiviral library at an MOI of 0.3 in a total of 31 × 100-mm plates for 20 hours. The infected cells were then selected by adding 0.4 μg ml⁻¹ puromycin to the E8 medium for 4 days. After selection, one billion hESCs were dissociated with Accutase, pooled from all cultural plates, and cryopreserved in STEM-CELLBANKER freezing media (Amsbio #11924). To perform each replicate of the screen, two hundred million hESCs were thawed and plated at a density of 120,000–150,000 cells per cm² in E8 medium with 10 μM Y-27632. To induce Cas9 expression, hESCs were treated with 1 μg ml⁻¹ doxycycline for 48 hours. They were maintained in the E8 medium for another 24 hours before harvesting for differentiation. hESCs were differentiated as described in the 'Neural differentiation' subsection. After 72–84 hours of neural induction, cells were dissociated using Accutase and sorted using FACSAria according to GFP expression. PAX6-GFP^high population gate was established based on day 0 ESC, and an equal number of PAX6-GFP^low cells were sorted for comparison. One replicate was collected at 72 hours (3 days) and two replicates were collected at 84 hours (3.5 days). Day 0 ESC, PAX6-GFP^high and PAX6-GFP^low cells were pelleted and snap-frozen. Cell pellets were lysed, and genomic DNA was extracted (Qiagen) and quantified by Qubit (Thermo Scientific). A quantity of gDNA covering 1,000× representation of sgRNAs was PCR amplified to add Illumina adapters and multiplexing barcodes. Amplicons were quantified by Qubit and Bioanalyzer (Agilent) and sequenced on Illumina HiSeq 2500. Sequencing reads were aligned to the screened library and counts were obtained for each gRNA.

Sequencing reads were aligned to the screened library and the CRISPR screen was analysed using the edgeR pipeline as previously described³¹. Z-score was calculated for each gene as previously described⁸¹. gRNA enrichment was calculated by comparing PAX6-GFP^high to PAX6-GFP^low populations. To identify candidate hits, we removed essential genes and genes with only one gRNA recovered. Essential genes were defined as genes significantly (P-value < 0.05) depleted in both PAX6-GFP^high versus day 0 and PAX6-GFP^low versus day 0 comparisons (Supplementary Table 2). Genes with a Z-score>1 and P-value < 0.05 were considered hits and followed up with secondary validation experiments.

We analysed the DepMap 22Q2 Public+Score, Chronos dataset that contains the genetic perturbation scores of 18,018 genes in 957 cancer cell lines⁸². We curated a list of top 27 screen hits and included top five correlates of each gene using Pearson correlation coefficients from DepMap gene essentiality scores above a minimum threshold (r > 0.25), resulting in a list of 96 genes. Correlation matrix heat map and network diagram were generated as previously described⁸³. Gene cluster that includes the most candidate hits was shown.

Top scoring gRNAs of the top 27 candidate genes from the pooled screen were selected for the genetic approach in the secondary validation. gRNAs were cloned into the lentiGuide-Puro plasmid (Addgene #52963) as described by the Zhang lab⁸⁴. PAX6::H2B-GFP iCas9 hESCs were infected with the gRNAs individually, selected using puromycin for 4 days, and Cas9 expression was induced by doxycycline for 2 days. Cells were differentiated as described in the 'Neural differentiation' subsection in array in 96-well plates and continuously imaged every 6 hours for 10 days by an imaging reader (Agilent BioTek Cytation 7). In order to image GFP signal, E6 medium was replaced by a customized DMEM/F12 (without riboflavin, pyridoxine hydrochloride, and phenol red) made by the MSKCC Media Preparation Core, supplemented 10 mg ml⁻¹ transferrin (Sigma-Aldrich, #T0665), 5 mg ml⁻¹ insulin (Sigma-Aldrich, #91077 C), 500 μM sodium selenite (Sigma-Aldrich, #S5261), 221 mM l-Ascorbic acid 2-phosphate sesquimagnesium salt hydrate (Sigma-Aldrich, #A8960), and 54.3 mg ml⁻¹ sodium bicarbonate (Sigma-Aldrich, #S5761). For pharmacological approach, each small molecule inhibitor was tested across a logarithmic dose range, and the highest concentration that did not induce toxicity was selected for use in the differentiation assays. Small-molecule inhibitors were added to culture medium starting from day 0 and maintained throughout the neuroectoderm differentiation. A list of small molecule inhibitors, their molecular targets and concentrations used is reported in Supplementary Table 3.

Phase separation in cells lysed in TRIzol was induced with 200 µl chloroform. RNA was extracted from the aqueous phase using the MagMAX mirVana Total RNA Isolation (Thermo Fisher, #A27828↗) on the KingFisher Flex Magnetic Particle Processor (Thermo Scientific, #5400630) according to the manufacturer's protocol with 350 µl input. After RiboGreen quantification and quality control by Agilent BioAnalyzer, 500 ng of total RNA with RIN values of 9.3–10 underwent polyA selection and TruSeq library preparation according to instructions provided by Illumina (TruSeq Stranded mRNA LT Kit, #RS-122-2102), with eight cycles of PCR. Samples were barcoded and run on a NovaSeq 6000 in a PE100 run, using the NovaSeq 6000 S4 Reagent Kit (200 cycles) (Illumina). An average of 34 million paired reads was generated per sample. Ribosomal reads represented 0.17–0.52% of the total reads generated and the percentage of mRNA bases averaged 86%. The R package Rsubread⁸⁵ was used to map reads to the human genome (GRCh38) using the align() function. Rsubread was also used to compute the expression count matrix from the mapped reads (BAM files) using the featureCounts() function. The R package DESeq2⁸⁶ used to normalize the raw counts (median of ratios method) and perform differential gene expression analysis. Data are presented on a gene-by-gene basis as normalized counts or after carrying out a variance-stabilizing transformation followed by PCA analysis using the top 500 variable features.

TCseq analysis was performed using the TCseq R package v1.28.0. RNA-seq normalized count table of NT day 0 to 10 data was used. Clustering was performed using c-means clustering by increasing cluster number until redundant clusters arose. To identify the effect of genotype in average time-variant signature, the averaged Z scores of selected timepoints were used as the aggregate gene expression signal as a function of time, then underwent nested, multivariate linear regression, contrasting the models with and without considering the effect of genetic KOs. The KO effects of MEN1 and SUZ12 were compared separately with the NT control. P-values were calculated from ANOVA of F tests.

GSEA was performed using GSEA software v4.3.3. For pathway analysis in the pooled screen, a ranked list of 18,897 genes recovered from the screen, ranked by their Z-scores, was used as input for GSEAPreranked. C5.go.v2024.1.Hs.symbols gene set database was used. Top 100 positive pathways were visualized using ggplot2. For enrichment analysis in the NT versus KO comparisons, normalized count table was used as the expression dataset. Gene clusters identified from TCseq were used as gene sets database. Comparisons were done on SUZ12 KO day 4 versus NT day 4 and MEN1 KO day 4 versus NT day 4.

Two clones were used for each KO line and two biological replicates were collected. Profiling of chromatin was performed by ATAC-Seq as previously described⁴⁶. Briefly, 50,000–100,000 frozen hESC were thawed, washed in cold PBS and lysed. The transposition reaction containing TDE1 Tagment DNA Enzyme (Illumina, #20034198) was incubated at 37 °C for 30 minutes. The DNA was cleaned with the MinElute PCR Purification Kit (QIAGEN, #28004) and material was amplified for five cycles using NEBNext High-Fidelity 2X PCR Master Mix (New England Biolabs, #M0541L). After evaluation by real-time PCR, 1–3 additional PCR cycles were done. The final product was cleaned by aMPure XP beads (Beckman Coulter, #A63882) at a 1.0× ratio, and size selection was performed at a 1.5× ratio. Libraries were sequenced on a NovaSeq 6000 X in a PE100 run, using the NovaSeq 6000 S1 and S4 Reagent Kits (200 Cycles) (Illumina). An average of 52 million paired reads were generated per sample. Sequencing data were aligned to the hg38 reference genome using bowtie2 (v2.5.0)⁸⁷. Ends of the aligned reads were shifted to remove Tn5 transposase artefacts as previously described^46,88. Macs2 (v2.2.7.1)⁸⁹ was used to remove duplicate reads and to call peaks, with the options extending the reads to both sides, and using a permissive P-value threshold (-p 1e-2–nomodel–shift −100–extsize 200–scale-to large–keep-dup all). If replicates were available, peaks were called at the combined level and the replicate level. They were then filtered using irreproducible discovery rate (IDR) (v2.0.4.2)⁹⁰ using the replicates with the threshold of 0.01. If a sample was not prepared in replicates, peaks at the combined level were filtered where the RPKM quantification is above 1.85 in at least one sample. The signal intensity of the samples was first written in raw counts and normalized by library size factor defined by DESeq2's median normalization. Differentially accessible peaks were assigned using DESeq2 (v1.39)⁸⁶ using the P_adj threshold of 0.05.

One clone was used for each KO line and two biological replicates were collected. CUT&RUN was performed from 100,000 cells per condition using the following antibodies at 1:100 dilution: rabbit anti-H3K4me3 (Abcam, #ab8580); rabbit anti-H3K27me3 (Cell Signaling Technologies, #9733); rabbit anti-Menin (Cell Signaling Technologies, #6891); rabbit anti-MLL1 (Cell Signaling Technologies, #14689); rabbit anti-SUZ12 (Cell Signaling Technologies, #3737); normal rabbit IgG (Cell Signaling Technologies, #2729). In brief, cells were permeabilized with digitonin and the different antibodies were incubated overnight at 4 °C on a rotator. Cells were washed and incubated with pA-MN. Beads Ca²⁺-induced digestion occurred on ice for 30 min and stopped by chelation. DNA was isolated using an extraction method with phenol and chloroform as previously described⁴⁷. Immunoprecipitated DNA was quantified by PicoGreen and the size was evaluated by Agilent BioAnalyzer. When possible, fragments between 100 and 600 bp were size selected using aMPure XP beads (Beckman Coulter, #A63882) and Illumina sequencing libraries were prepared using the KAPA HTP Library Preparation Kit (Kapa Biosystems KK8234) according to the manufacturer's instructions with 0.3–5 ng input DNA and 14 cycles of PCR. Barcoded libraries were run on the NovaSeq 6000 in a PE100 run, using the NovaSeq 6000 S4 Reagent Kit (200 Cycles) (Illumina). An average of 15 million paired reads were generated per sample. Sequencing data were aligned to the hg38 reference genome using bowtie2 (v2.5.1). Macs2 (v2.2.7.1) was executed to remove duplicate reads and to call peaks with the respective input control and using a permissive P-value threshold. Peaks were further filtered by IDR, with the threshold chosen as 0.01 for H3K4me3 and 0.05 for H3K27me3. The signal intensity of the samples was written during the peak calling (-B -p 1e-2–scale-to large–keep-dup all). It was then normalized by library size factor defined by DESeq2's median normalization and converted to bigWig format. Deeptools⁹¹ was used for heatmaps and gene-profile plots. Tracks of CUT&RUN peaks were visualized using Integrative Genomics Viewer v2.16.1 (IGV, Broad Institute) or trackplot⁹² R script. ChIP-seq data in mEpiSCs were downloaded from the ChIP-Atlas database⁹³ with the threshold chosen as 100 or 500.

Definitive endoderm differentiation was performed as previously described⁹⁴. hESCs were dissociated into single cells using Accutase and plated at 125,000 cells per cm² onto vitronectin-coated plates in E8 medium with 10 μM ROCK inhibitor Y-27632 (day −1). The next day (day 0), cells were washed with 1xDPBS once and culture medium was changed to Advanced RPMI (Thermo Fisher Scientific, #12633012) with penicillin streptomycin and 1X GlutaMAX, supplemented with 50 ng ml⁻¹ Activin A (R&D systems, #338-AC-MTO) for 3 days (day 0–2) and 5 μM CHIR99021 (R&D systems, #4423/50) for the first day (day 0). Marker expression was assessed using flow cytometry and gating was established using hESCs as negative control.

Cardiomyocyte differentiation was performed as previously described⁵⁵ with some modifications. hESCs were dissociated into single cells using Accutase and plated at 80,000 cells per cm² onto matrigel-coated plates in E8 medium with 10 μM ROCK inhibitor Y-27632 (day −2). Cells were maintained in E8 medium for another day (day −1) before changing to base medium containing RPMI 1640 medium (Fisher Scientific, #MT10041CV) and 1X B27 supplement minus insulin (Fisher Scientific, #A1895601) or Advanced RPMI (Thermo Fisher Scientific, #12633012) with Penicillin Streptomycin and 1X GlutaMAX, supplemented with 5 μM CHIR99021 (day 0). On day 1, CHIR99021 was removed, and culture medium was changed into base medium. From day 2 to 3, cells were treated with IWP-2 (Stemgent, #04-0034) in the base medium for 48 hours. From day 4 onwards, IWP-2 was removed, and base medium was changed every other day. Marker expression was assessed using flow cytometry and gating was established using hESCs as negative control. Contraction was assessed in 96-well plates using an imaging reader (Agilent BioTek Cytation 7). A 5-second video with 20 fps acquisition speed and 1,390 × 1,390 μm field-of-view was recorded for each well and contraction phenotype was observed by eye and counted manually. The position of imaged area was kept constant across all wells and at least six wells were assessed for each condition.

The following small molecules targeting EZH2 and Menin–MLL interaction were used in the study: Tazemetostat (EPZ-6438) (Selleckchem, #S7128); VTP50469 (Selleckchem, #S8934). The following doses were used: for definitive endoderm differentiation and neurogenesis, 0.025% DMSO, 2.5 μM VTP50469, and 2.5 μM Tazemetostat; for cardiomyocyte differentiation and neuroectoderm differentiation of mEpiSCs, 0.0125% or 0.025% DMSO, 1.25 μM VTP50469, and 1.25 μM tazemetostat. For human neuroectoderm, definitive endoderm, cardiomyocyte and mouse neuroectoderm differentiation, small-molecule inhibitors were added to the culture medium starting from day 0 and maintained throughout the differentiation. For neurogenesis, small-molecule inhibitors were added to the culture media starting from day 20 and maintained throughout the differentiation.

Cells were collected and lysed in RIPA buffer (Sigma) with 1:100 Halt Protease and Phosphatase Inhibitor Cocktail (Thermo Fisher Scientific) and then sonicated for 3 × 30 sec at 4 °C. Protein lysates were centrifugated for 15 min at more than 15,000g at 4 °C and supernatant was collected and quantified by Precision Red Advanced Protein Assay (Cytoskeleton). Protein (5–10 μg) was boiled in NuPAGE LDS sample buffer (Invitrogen) at 95 °C for 5 min and separated using NuPAGE 4–12% Bis-Tris Protein Gel (Invitrogen) in NuPAGE MES SDS Running Buffer (Invitrogen). Proteins were electrophoretically transferred to nitrocellulose membranes (Thermo Fisher Scientific) with NuPAGE Transfer Buffer (Invitrogen). Blots were blocked for 60 min at room temperature in TBS-T + 5% nonfat milk (Cell Signaling Technologies) and incubated overnight in the same solution with the respective primary antibodies at 4 °C. The following primary antibodies were used at 1:500 dilution: mouse anti-Cas9 (Cell Signaling Technologies, #14697S); rabbit anti-Menin (EpiCypher, #13-2021); rabbit anti-SUZ12 (Cell Signaling Technologies, #3737S); HRP-conjugated rabbit anti-GAPDH (Cell Signaling Technologies, #3683S). The following secondary antibodies were incubated for 1 h at room temperature at 1:1,000 dilution: anti-mouse IgG HRP-linked (Cell Signaling Technologies, #7076); anti-rabbit IgG HRP-linked (Cell Signaling Technologies, #7074). Blots were revealed using SuperSignalTM West Femto Chemiluminescent Substrate (Thermo Fisher Scientific) at ChemiDoc XRS+ system (Bio-Rad). Chemiluminescence was imaged and analysed using Image lab software version 6.1.0 (Bio-Rad).

Cells were collected in Trizol and processed using Direct-zol RNA Miniprep kit (Zymo Research, # R2050). A total of 1 μg of RNA was used to generate cDNA using iScript (Bio-Rad). Real-time PCR was performed using SsoFAST EvaGreen Mix (Bio-Rad) in a Bio-Rad CFX96 Thermal Cycler. The manufacturers protocol was used for all steps. Samples in which the housekeeping gene had a cycle threshold (Ct) value above 30 were repeated. Primer sequences used in the study are listed in the Supplementary Table. 5

Cultured cells were fixed with 4% PFA in PBS for 20 min at room temperature, washed three times with PBS, permeabilized for 30 min in 0.5% Triton X-100 in PBS and then blocked in a solution containing 5% Normal donkey serum, 2% BSA and 0.25% Triton X-100 for 1 hour at room temperature. Primary antibodies were incubated overnight at 4 °C in the same blocking solution. Secondary antibodies conjugated to either Alexa 488, Alexa 555 or Alexa 647 (Thermo) were incubated for 45 min at 1:400 dilution in blocking solution. Cell nuclei were stained with 5 μM 4′-6-diamidino-2-phenylindole (DAPI) in PBS. The following primary antibodies and dilutions were used: mouse anti-HuC/HuD 1:500 (Thermo Scientific, #A-21271); chicken anti-MAP2 1:2,000 (Abcam, # ab5392); rat anti-SOX2 1:500 (Thermo Scientific, #14-9811-82). To quantify the percentage of positive cells, images were converted to 8-bit and a threshold was set for DAPI to exclude bright cells. 'Analyse Particles' was used to quantify the total number of cells. A threshold was set for each marker, and cells with over 50% of DAPI area that met the threshold requirement were counted as positive. Thresholds were kept constant across all wells per replicate. 6–10 images of were taken from 3 individual wells for each condition and averaged.

Cells were dissociated to single cells using Accutase. For live cell staining, cells were incubated with conjugated antibodies in 3% fetal bovine serum (FBS) in PBS for 45 min at 4 °C. Cell nuclei were stained with DAPI in PBS, and live cells were gated based on negative DAPI signal. Fixed cell staining was performed using BD Cytofix/Cytoperm kit (BD Biosciences, #554714) according to manufacturer's instructions. The following antibodies were used at 1:100 dilution: Alexa 488 mouse anti-PAX6 (BD Biosciences, #561664); PE-CF594 mouse anti-PLZF (ZBTB16) (BD Biosciences, #565738); Alexa 647 mouse anti-CXCR4 (Biotechne, # FAB172R-100UG); PE rabbit anti-GATA6 (Cell Signaling Technologies, #26452S); APC goat anti-SOX17 (Biotechne, #IC1924A); PE mouse anti-SIRP alpha (R&D systems, #FAB4546P); APC rat anti-CD324 (E-Cadherin) (Biolegend, #147311); PE Mouse Anti-Cardiac Troponin T (BD Biosciences, #564767). Results were analysed using FlowJo. Mean fluorescence intensity (MFI) of markers assessed by flow cytometry is reported in the Supplementary Table. 6

No statistical method was used to predetermine sample sizes, but our sample sizes are similar to those reported in the previous publications^95,96. Data collection and analysis were not performed blind to the conditions of the experiments. For secondary validation experiments, samples were de-identified respect to the molecular or genetic target and a number code was assigned for each condition. No randomization was involved in the differentiation assays. No data were excluded from the analyses unless the differentiation experiment itself failed. Statistical analysis was carried out using GraphPad Prism 10. The details of statistical analyses and significance values are specified in the figures and figure legends. Data were reported as the mean ± s.d. or s.e.m, or independent replicates shown as individual data points. Data distribution was assumed to be normal, but this was not formally tested. For statistical analyses conducted using one-way ANOVA test with post-hoc Dunnett's multiple comparisons test, results from one-way ANOVA test with post-hoc Tukey's multiple comparisons test are provided in the Supplementary Table 7.

Further information on research design is available in thelinked to this article. Nature Portfolio Reporting Summary