H3K36 methylation maintains cell identity by regulating opposing lineage programmes

Abstract

The epigenetic mechanisms that maintain differentiated cell states remain incompletely understood. Here we employed histone mutants to uncover a crucial role for H3K36 methylation in the maintenance of cell identities across diverse developmental contexts. Focusing on the experimental induction of pluripotency, we show that H3K36M-mediated depletion of H3K36 methylation endows fibroblasts with a plastic state poised to acquire pluripotency in nearly all cells. At a cellular level, H3K36M facilitates epithelial plasticity by rendering fibroblasts insensitive to TGFβ signals. At a molecular level, H3K36M enables the decommissioning of mesenchymal enhancers and the parallel activation of epithelial/stem cell enhancers. This enhancer rewiring is Tet dependent and redirects Sox2 from promiscuous somatic to pluripotency targets. Our findings reveal a previously unappreciated dual role for H3K36 methylation in the maintenance of cell identity by integrating a crucial developmental pathway into sustained expression of cell-type-specific programmes, and by opposing the expression of alternative lineage programmes through enhancer methylation.

Extended Data Fig. 1:. Key role of H3K36-methylation in cell identity maintenance.

(a) Colony counts for AP staining of reprogrammable MEFs transduced as indicated (Fig. 1b,c). K36M wells were confluent and could not be counted. P values were determined by two-sided unpaired Student’s t test, error bars indicate mean ± SD (n=3 biologically independent experiments). (b) Mice with dox-inducible alleles of WT H3.3 or K36M in the Col1a1 locus were crossed with mice harboring a dox-inducible OKSM cassette in the same locus, and an EGFP reporter in the 3’UTR of Pou5f1. (c) Immunofluorescence of MEFs derived as in a). Scale bar 50 μm. (d) Mass spectrometry of histone modifications in day 4 reprogramming intermediates (n=2 independent biological experiments). (e) MEFs without endogenous OKSM but with inducible H3.3 WT or K36M were doxycycline-treated for 2 days, then replated and transduced with constitutive OKSM virus. Doxycycline was added to the medium for the indicated intervals, intracellular flow cytometry for Nanog was performed on day 8. (f) Quantification of Nanog positivity by flow cytometry on day 8 in K36M cells treated with doxycycline for the indicated time. Cells were either not pre-treated or pre-treated with doxycycline 2 days prior to initiation of reprogramming (see k). Error bars indicate mean ± SD (n=3 independent biological experiments). (g) Fraction of Oct4-GFP+ cells during reprogramming in FBS/LIF medium supplemented with ascorbic acid (left) and without supplementation (right), error bars indicate mean ± SD (n=3 independent biological experiments). (h) Membrane dye dilution assay for reprogramming cultures. (i) Percentage of viable cells as assessed by Annexin V/PI negativity on day 2 and 4 of reprogramming. Error bars indicate mean ± SD (n=3 independent biological experiments). (j) Fraction of Oct4-GFP+ cells after sorting of positive cells and expansion on gelatin (top), and in picked iPSCs passaged on feeders (bottom). (k) Day 10 K36M reprogramming cultures were sorted by Oct4-GFP reporter positivity. Positive cells were maintained in FBS/LIF, negative cells underwent continued reprogramming in AGi medium. (l) Quantification of Oct4-GFP+ cells by flow cytometry in K36M cells sorted by Oct4-GFP reporter status (see i), error bars indicate mean ± SD (n=3 independent biological experiments).

Extended Data Fig. 2:. K36M enhances the reprogramming of different cell types and generates iPSCs highly similar to control iPSCs.

(a) Reprogramming of GMPs to iPSCs. Alkaline phosphatase staining of iPSC colonies at the indicated timepoints. Quantification of colony counts. P values were determined by two-sided unpaired Student’s t test, n=3 biologically independent experiments. (b) Reprogramming of keratinocytes to iPSCs. Alkaline phosphatase staining on day 13 of cells cultured for the indicated timeframes. Area percentage of well that is AP positive. P values were determined by two-sided unpaired Student’s t test, error bars indicate mean ± SD (n=3 independent biological experiments). (c) Immunofluorescence for Nanog, H3K36me3, H3K36me2, and H3K27me3 of passaged iPSC cultures from WT and K36M backgrounds cultured without doxycycline on irradiated feeders. (d) Relative expression (RNA-seq) of key pluripotency genes in passaged iPSCs of both backgrounds, n=2 biologically independent experiments. (e) DNA methylation at MEF (n=63,696) and ESC enhancers (n=72,638) for MEFs and passaged iPSCs of both backgrounds (RRBS). Box plot center line indicates median; lower/upper hinges indicate 25th/75th percentiles; whiskers extend to 1.5x IQR. (f) Representative gene tracks showing RNA-seq, ATAC-seq, and RRBS data for Cdh1 and Pou5f1 in passaged iPSCs of both backgrounds. (g) Correlation matrices for key histone modifications and chromatin accessibility in passaged WT and K36M iPSCs (CUT&Tag and ATAC-seq). (h) Correlation plot of K36M vs. WT derived iPSCs for H3K36me3 over gene bodies (CUT&Tag). (i) Correlation plot of K36M vs. WT derived iPSCs for H3K36me2 in genome-wide 50kb bins (CUT&Tag). (j) Representative gene tracks showing H3K36me3, H3K36me2, H3K27me3, and H3K4me3 at pluripotency gene Nanog. (k) Quantification of the diameter of embryoid bodies from passaged iPSCs of both backgrounds (without doxycycline). P value determined by two-sided unpaired Student’s t test, n=49 for WT, n=44 for K36M. (l) qPCR for Nanog, Nestin, Sox7, and Gata6 in embryoid bodies derived from passaged iPSCs of both backgrounds (without doxycycline), error bars indicate mean ± SD (n=3 independent biological experiments). (m) H&E sections of teratomas generated with iPSCs of both backgrounds (without doxycycline). 4/4 WT and 6/6 K36M iPSC lines produced well-differentiated teratomas. Images depict tissue-like structures of all germ layers. (n) Coat-color chimeras generated by blastocyst injection of K36M iPSCs.

Extended Data Fig. 3:. Single-cell RNA-seq reveals main trajectories of WT and K36M reprogramming intermediates.

(a) UMAP embedding of scRNA-seq data (Seurat framework) using MEFs, reprogramming intermediates on days 2, 4, 6, 8 for WT and K36M, as well as passaged iPSCs (n=38,743 total number of cells). (b) Expression of pluripotency gene Nanog projected on the same UMAP embedding as in (a). (c) Expression of mesenchymal gene Prrx1 projected on the same UMAP embedding as in (a). (d,e) Diffusion pseudotime mapping of day 2 to day 8 intermediates undergoing reprogramming. WT cells are colored in blue, K36M cells in red. (f) Expression of pluripotency gene Nanog projected on the same pseudotime embedding as in (d). (g) Expression of epithelial gene Cdh1 projected on the same pseudotime embedding as in (d). (h) Expression of mesenchymal gene Zeb1 projected on the same pseudotime embedding as in (d). (i) Relative expression (RNA-seq) of mesenchymal and epithelial genes in MEFs expressing H3.3 WT or K36M, but not OKSM, n=2 biologically independent experiments. (j) Gene ontology terms of genes downregulated in K36M MEFs without OKSM. Analysis and p values from geneontology.org. (k) Gene ontology terms of genes upregulated in K36M MEFs without OKSM. Analysis and p values from geneontology.org.

Extended Data Fig. 4:. K36M confers epithelial plasticity on cells undergoing reprogramming.

(a) Gene expression of Twist1 and Cdh1 on the same UMAP embedding as used in Fig. 2d. Dashed circles encompass day-2 and day-4 samples for WT (blue circle) and K36M (red circle). Solid arrow indicates switch-like MET in K36M samples, dashed arrow indicates heterogeneous maintenance/activation of mesenchymal/epithelial programs in WT cells. (b) Immunofluorescence for Vimentin and Epcam in WT and K36M cells on day 4 of reprogramming. Scale bar = 25 μm. Three independent biological experiments with similar results. (c) Correlation plots of single-cell RNA-seq data comparing transcriptional programs within each cell to MEFs (y-axis) and ESCs (x-axis)91. For each sample, the corresponding cells are colored according to their Epcam expression levels, whereas other cells are greyed out. (d) Correlation plots as in (c), expression data of Twist1 is superimposed. (e) Correlation plots as in (c), expression data of Pou5f1 is superimposed.

Extended Data Fig. 5:. K36M disrupts TGFβ signaling and modulates epithelial plasticity in diverse contexts.

(a) Flow cytometry histograms displaying Epcam expression in day-4 reprogramming intermediates for WT and K36M samples. Untreated control cells are compared to cells treated with 250 nM Repsox (TGFβi) or 2.5 ng/ml recombinant TGFβ-1 or -2 (rTGFβ-1, rTGFβ-2). (b) Fraction of Oct4-GFP+ cells treated with TGFβi or rTGFβ in day 4 reprogramming intermediates. Error bars indicate mean ± SD (n=3). (c) Representative tracks for expression of mesenchymal gene Col1a2 on day 8 of reprogramming, WT or K36M cells were treated as indicated. (d) Representative tracks for expression of epithelial gene Cdh1 on day 8 of reprogramming, WT or K36M cells were treated as indicated. (e) Representative tracks for miR-200a and miR-290. (f) Schematic of K36M’s effect on TGFβ signaling and miRNA expression during reprogramming. (g) De-differentiation of MEFs to induced myogenic progenitor cells (iMPCs). qRT-PCR for myotube marker Myh1 and iMPC marker Pax7, P values were determined by two-sided unpaired Student’s t test, error bars indicate mean ± SD (n=3 biologically independent experiments). Flow cytometry for Pax7-GFP reporter positive cells. (h,i) Immunofluorescence of Nanog and K36M in ES cells of both backgrounds, cultured in S/L (g) or 2iL (h) conditions. Result is representative of three independent biological experiments. (j) Differentiation of ESC aggregates to elongated gastruloids. Representative brightfield images (scale bar = 500 μm) and quantification of long axis diameter (line = mean). P value was determined by two-sided unpaired Student’s t test, n=19 for WT, n=20 for K36M. (k) Differentiation of ESCs to pre-somitic mesoderm. Representative immunofluorescence for Cdh2 (scale bar = 50 μm). qRT-PCR for mesodermal transcription factors Tbx6 and Msgn1. P values were determined by two-sided unpaired Student’s t test, error bars indicate mean ± SD (n=3 independent biological experiments).

Extended Data Fig. 6:. H3K36me2 and H3K36me3 cooperatively safeguard cell identity.

(a) Histone methyltransferases and demethylases implicated in the regulation of H3K36me2 (top) and H3K36me3 (bottom). (b) Fraction of Epcam+ cells on day 4 of reprogramming (top) in WT cells with knockdown of indicated histone methyltransferases. Colony counts after 6 days of doxycycline followed by 4 days of independent growth (bottom), error bars indicate mean ± SD (n=3 biologically independent experiments). (c,d) Fraction of Epcam+ (c) or Oct4-GFP+ (d) cells on day 8 of reprogramming in WT and K36M cells transduced with either empty vector or dox-inducible Nsd2, error bars indicate mean ± SD (n=3 biologically independent experiments). (e) Flow cytometry for Epcam on day 8 of reprogramming in WT cells with knockdown of the indicated histone demethylases. (f,g) Fraction of Epcam+ cells on day 4 (f) and day 8 (g) of reprogramming in WT cells with knockdown of the indicated histone demethylases, error bars indicate mean ± SD (n=3 biologically independent experiments). (h) Fraction of Oct4-GFP+ cells on day 8 of reprogramming in WT cells with knockdown of the indicated histone demethylases. P values were determined by unpaired Student’s t test, error bars indicate mean ± SD (n=3 biologically independent experiments). (i,j) Flow cytometry for Epcam during reprogramming in WT cells with overexpression of the indicated histone demethylases, error bars indicate mean ± SD (n=3 biologically independent experiments). (k) Fraction of Oct4-GFP+ cells on day 4 and day 8 of reprogramming in WT cells with overexpression of the indicated histone demethylases, error bars indicate mean ± SD (n=3 biologically independent experiments). (l) qRT-PCR for mesenchymal genes Vim and Prrx1, epithelial genes Epcam and Cdh1, and pluripotency gene Pou5f1 on day 4 of reprogramming in WT cells overexpressing Kdm2a vs. empty vector control. P values were determined by unpaired Student’s t test, error bars indicate mean ± SD (n=3 independent biological experiments).

Extended Data Fig. 7:. K36M alters H3K36me2/3 deposition and gene expression.

(a) H3K36me3 deposition at expressed genes (RPKM>0.1, n=10,251 for WT, n=10,496 for K36M) of indicated expression quintiles for WT (left) and K36M (right) samples on day 4 of reprogramming. Center line indicates median; lower/upper hinges indicate 25th/75th percentiles; whiskers extend to 1.5x IQR. (b) H3K36me3 density over gene bodies of differentially expressed genes (n=1,872) on day 4 (upregulated = red, downregulated = blue). Box plots as in (a). P values were determined by two-sided Wilcoxon rank sum test. (c) Fold change of gene expression (y axis) vs. fold change of H3K36me3 (x axis) between K36M and WT samples on day 4 of reprogramming. (d) Representative gene tracks for H3K36me3 and RNA levels at mesenchymal gene Vim, epithelial gene Cdh1, and pluripotency gene Pou5f1. (e) Profile plots of H3K36me2 at promoters, enhancers, and gene bodies. (f) Profile plots of H3K36me2 and H3K27ac at H3K36me2 domains containing down- or upregulated enhancers. (g) Representative tracks for Prrx1, a mesenchymal gene downregulated in K36M samples on day 4, Krt8, and Pou5f1, epithelial/pluripotency genes upregulated in K36M samples on day 4. Putative regulatory elements highlighted in grey. (h) Gene ontology terms of genes closest to H3K36me2 domain-embedded enhancers that are differentially downregulated in K36M cells. Analysis and p values from geneontology.org. (i) Gene ontology terms of genes closest to H3K36me2 domain-embedded enhancers that are differentially upregulated in K36M cells. Analysis and p values from geneontology.org. (j) Dot plot representing enrichment of ENCODE data for differentially active enhancers within H3K36me2 domains. P values were determined by Fisher’s exact test.

Extended Data Fig. 8:. PRC2 contributes to the K36M-dependent silencing of the somatic program.

(a) H3K27me3 deposition within H3K36me2 domains in K36M vs. WT samples in day-4 reprogramming intermediates. Domains gaining H3K27me3 are colored in red, domains losing H3K27me3 are colored in blue. (b) Ontology terms for genes with promoters overlapping H3K36me2 domains and gaining H3K27me3. Analysis and p values from geneontology.org. (c) Heatmaps for H3K27me3 and H3K4me3 at promoters mesenchymal and epithelial genes in WT and K36M samples. (d,e) Fraction of Epcam+ and Oct4-GFP+ cells in WT (blue) and K36M (red) samples with knockdown of indicated PRC2 components (top). Log2(fold change) of fraction normalized to control siRNA (bottom). Error bars indicate mean ± SD (n=3 independent biological experiments). (f) Representative histograms of flow cytometry for Epcam in K36M cells with control siRNA and knockdown of Ezh2 or Suz12. (g) qRT-PCR for mesenchymal (Vim, Prrx1, Zeb1), epithelial (Cdh1, Epcam), and pluripotency (Pou5f1) marker genes, error bars indicate mean ± SD (n=3 biologically independent experiments). (h) Immunofluorescence for H3K27me3 in WT and K36M cells transfected with control siRNA or knockdown of Ezh2 or Suz12. Representative result from three independent biological experiments.

Extended Data Fig. 9:. K36M rewires DNA methylation patterns.

(a) Number of colonies following alkaline phosphatase staining of WT and K36M cell cultures transduced with non-selectable, dox-inducible lentiviruses for the expression of SKM, OSM, or OKM. Cultures were induced for 12 days and stained on day 15, n=3 independent biological experiments. (b) Correlation plot of log2(fold-change) differences (K36M vs. WT) at Sox2 peaks called in WT and K36M samples. Differences of Sox2 enrichment are correlated with differences in H3K27ac abundance. Pearson correlation with corresponding two-sided t test, R = 0.56, p<2.2e-16. (c) Profile plots showing H3K36me2 abundance at ectopic and ESC-specific Sox2 sides in WT and K36M cells on day 4 of reprogramming. (d) Sox2 enrichment at Sox2 binding sites as defined in iPSCs, log2(RPKM). (e) Correlation plot of log2(fold-change) differences (K36M vs. WT) at Sox2 peaks called in WT and K36M samples. Differences of Sox2 enrichment are correlated with differences in chromatin accessibility (as measured by ATAC-seq). Pearson correlation with corresponding two-sided t test, R = 0.68, p<2.2e-16. (f) Chromatin accessibility (as measured by ATAC-seq) at ectopic and ESC-specific Sox2 binding sites in MEFs and passaged iPSCs, log2(RPKM+1). (g) DNA methylation at differentially methylated regions losing (left, n=30,294) or gaining (right, n=28,060) methylation in iPSCs vs. MEFs. Box plot center line indicates median; lower/upper hinges indicate 25th/75th percentiles; whiskers extend to 1.5x IQR. (h-k) Representative gene tracks of Cdh1, Krt8, the miR-290 cluster, and Pou5f1. Putative regulatory elements affected by DNA demethylation are highlighted in grey. (l) CpG density at differentially active enhancers in H3K36me2 domains (n=4,939). P value determined by two-sided Wilcoxon rank sum test. Box plots as in (g).

Extended Data Fig. 10:. DNA demethylation is limiting for K36M-dependent enhancer rewiring.

(a) DNA methylation (WGBS) at differentially active enhancers in H3K36me2 domains (n=4,939) in day-4 reprogramming intermediates of WT, K36M and K36M+DMOG samples. Center line indicates median; lower/upper hinges indicate 25th/75th percentiles; whiskers extend to 1.5x IQR. (b) DNA methylation (WGBS) at ectopic (n=45,095) and ESC-exclusive (n=27,708) Sox2 binding sites in day-4 reprogramming intermediates of WT, K36M and K36M+DMOG samples. Box plots as in (a). (c) Representative histogram plots from flow cytometric analysis for Epcam of K36M cells with knockdown of the indicated Tet demethylases. (d) Fraction of Epcam+ cells in K36M cells with Tet knockdown on day 4 of reprogramming, error bars indicate mean ± SD (n=3 biologically independent experiments). (e) qPCR of miRNAs miR-200b-3p, miR-205-5p, and miR-290a-5p in untreated (K36M Ctrl) and DMOG-treated K36M cells (K36M DMOG). P values were determined by unpaired Student’s t test, error bars indicate mean ± SD (n=3 independent biological experiments). (f) Fraction of Epcam+ cells in K36M cultures transduced with either an empty vector or dox-inducible overexpression vectors for Dnmt3a and Dnmt3b, error bars indicate mean ± SD (n=3 independent biological experiments). (g) Bisulfite-seq of a Cdh1 enhancer in K36M cells transduced with either empty vector (left), or overexpression of Dnmt3a (middle) or Dnmt3b (right). (h) Quantification of H3K36me2 levels within H3K36me2 domains (n=7,610) on day 4 of reprogramming in WT, untreated K36M cells (K36M Ctrl), and DMOG-treated K36M cells (K36M DMOG). Center line indicates median; lower/upper hinges indicate 25th/75th percentiles; whiskers extend to 1.5x IQR. P values were determined by two-sided Wilcoxon rank sum test. (i-k) Representative gene tracks of Krt8, Pou5f1, and the miR-290 cluster. Putative regulatory elements highlighted in grey.

Fig. 1:. Histone mutant analysis reveals crucial role of H3K36-methylation in cell identity maintenance.

(a) Lysine-to-methionine (K-to-M) mutants of histone H3.3 dominantly block histone methylation at the respective residue across the genome. (b) WT or mutant histones (K4M, K9M, K27M, and K36M) were co-expressed with OKSM in fibroblasts during reprogramming. (c) Alkaline phosphatase (AP) staining of transgene-independent iPSC colonies. Transduced MEFs were treated with dox for 12 days, followed by 3 days of withdrawal. (d) OKSM transgene dependency assay. MEFs were treated with dox, ascorbic acid, and CHIR99021 as indicated, and iPSCs were scored by AP staining on day 15. (e) Marker dynamics during MEF to iPSC reprogramming. (f) Percentage of Thy1⁻ and SSEA1⁺ intermediates on day 6 of reprogramming. P values were determined by two-sided unpaired Student’s t test, error bars indicate mean ± SD (n=3 biologically independent experiments). (g) Flow cytometry analysis of Thy1 and SSEA1 at early timepoints of reprogramming in WT and K36M cells. (h) Abundance of intermediate populations that reprogram with high efficiency (SSEA1⁺/Sca1⁻/Epcam⁺) on days 4 and 6 of reprogramming. P values were determined by two-sided unpaired Student’s t test, error bars indicate mean ± SD (n=3 biologically independent experiments). (i) Flow cytometry using the Oct4-GFP reporter as a readout for successful activation of the endogenous pluripotency network on day 4 and day 8 of reprogramming. (j) Detection of Oct4-GFP⁺ cells at different timepoints of reprogramming in WT and K36M samples (ascorbic acid and CHIR99021 condition). Error bars indicate mean ± SD (n=3 biologically independent experiments). (k) AP staining of human reprogramming cultures at the indicated timepoints. Human fibroblasts were transduced with constitutive vectors expressing H3.3 WT or K36M, and OKSM. (l) Quantification of AP colony counts on days 9 and 12. P values were determined by two-sided unpaired Student’s t test, n=3 biologically independent experiments. (m) qRT-PCR for epithelial and pluripotency-associated genes in human reprogramming cultures on days 9 and 12 of reprogramming. P values were determined by two-sided unpaired Student’s t test, error bars indicate mean ± SD (n=3 biologically independent experiments).

Fig. 2:. K36M endows iPSC intermediates with transcriptional homogeneity and epithelial plasticity.

(a) Multidimensional scaling (MDS) plot of RNA-seq data based on genes with most variable expression among all timepoints and conditions. Filled circles, MEFs expressing both histone genes and OKSM for indicated number of days; open circles, MEFs expressing histone genes alone for 4 days. (b) Heatmaps showing gene expression dynamics in WT and K36M reprogramming intermediates for genes differentially expressed between MEFs and iPSCs (n=2 biologically independent replicates). (c) Scatter plot showing gene expression differences (RPKM at log2 scale) between WT and K36M reprogramming intermediates on day 4. Genes that are upregulated in iPSCs vs. MEFs are highlighted in green, genes that are downregulated in iPSCs vs. MEFs are highlighted in brown (n=2 biologically independent replicates were integrated for this analysis). (d) UMAP embedding of single-cell RNA-seq data using MEFs, reprogramming intermediates on days 2, 4, 6, 8 for WT and K36M, as well as passaged iPSCs. For each indicated group, one sample was encapsulated leading to n=38,743 cells total. (e) Expression of Nanog projected on the same UMAP embedding as shown in (d). Dashed circles highlight scattered expression of Nanog in WT cells (blue) on day 8 compared to homogeneous expression in K36M cells on days 6 and 8 (red). (f) Quantification of transcriptional heterogeneity of indicated samples as measured by average distance between cells within each sample in the same UMAP embedding as shown in (d). Median values for all cells within each sample are plotted. (g) Selected differentially expressed genes between WT and K36M samples that distinguish day 2 and day 4 intermediates. (h) Expression of the mesenchymal regulator Zeb1 projected on the same UMAP embedding as shown in (d). Dashed circles highlight d2 and d4 samples for WT (blue) and K36M (red). (i) Expression of epithelial gene Epcam projected on the same UMAP embedding as used in (d). Dashed circles highlight d2 and d4 samples for WT (blue) and K36M (red). (j) Flow cytometric quantification of Epcam expression in WT vs. K36M reprogramming intermediates.

Fig. 3:. K36M acts downstream of TGFβ and Smad2 but upstream of Zeb1.

(a) Effects of TGFβ signaling on reprogramming. (b) Flow cytometry for Epcam in day 8 reprogramming intermediates for WT and K36M samples. Untreated controls vs. cells treated with 250 nM Repsox (TGFβi) or 2.5 ng/ml recombinant TGFβ-1 or -2 (rTGFβ-1, rTGFβ-2). (c) Fraction of Oct4-GFP⁺ cells in reprogramming cultures treated with TGFβi or rTGFβ-1/rTGFβ-2 on day 8. P values were determined by two-sided unpaired Student’s t test, error bars indicate mean ± SD (n=3 independent biological experiments). (d) Heatmap showing gene expression (z-score) of mesenchymal and epithelial genes in WT and K36M reprogramming intermediates treated with either TGFβi or rTGFβ-1 (n=2 biologically independent replicates). (e) Immunoblot analysis for phospho-Smad2, Smad2 and GAPDH in WT and K36M reprogramming intermediates treated with either TGFβi or rTGFβ-1. Blot is representative of three independent biological experiments. (f) Epcam expression (fold-change) of WT and K36M reprogramming cultures (day 4) transfected with the indicated siRNAs relative to control. Error bars indicate mean ± SD (n=3 independent biological experiments). (g) Fraction of Oct4-GFP⁺ cells in day 8 WT samples treated with control siRNA or siRNA targeting Zeb1. P values were determined by two-sided unpaired Student’s t test, error bars indicate mean ± SD (n=3 independent biological experiments). (h) Flow cytometry for Epcam on day 4 in WT and K36M reprogramming intermediates transduced with empty vector or dox-inducible Zeb1. (i) small RNA-seq of day-4 reprogramming intermediates. Heatmap of key pro-epithelial and pro-pluripotency miRNAs (n=2 biologically independent replicates). (j) MEF-to-neuron conversion. Representative immunofluorescence images (scale bar = 50 μm) and quantification of Tubb3⁺ neurons (line = mean). P value determined by two-sided unpaired Student’s t test, n=3 biologically independent experiments. (k) MEF-to-myotube conversion. Representative immunofluorescence images (scale bar = 50 μm) and quantification of Myh1⁺ myotubes (line = mean). P values determined by two-sided unpaired Student’s t test, n=3 biologically independent experiments. (l) Epidermal stem cells exposed to recombinant TGFβ-1 (10 ng/ml). Representative images for phase and p63 immunofluorescence. Fraction of p63⁺ cells per field. P value determined by unpaired Student’s t test, error bars indicate mean ± SD (n=3 independent biological experiments).

Fig. 4:. K36M decommissions MEF enhancers and activates ESC enhancers.

(a) Profile plot of mean H3K36me3 density over gene bodies ± 10kb. TSS, transcription start site, TES, transcription end site (n=2 biologically independent replicates were integrated for this analysis). (b) H3K36me3 deposition at differentially expressed genes (n=2,068) for WT (blue) and K36M (red) samples on day 4 of reprogramming. Center line indicates median (n=2 biologically independent replicates were integrated for this analysis). (c) Profile plot of median H3K36me2 density over H3K36me2 domains (n=2 biologically independent replicates were integrated for this analysis). (d) Scatter plot showing expression of genes proximal to or overlapping with H3K36me2 domains in WT and K36M samples. Red = upregulated, blue = downregulated genes in OKSM/K36M vs. OKSM/WT cells on day 4 (n=2 biologically independent replicates were integrated for this analysis). (e) Heatmaps showing signal intensities for H3K36me2 and H3K27me3 at H3K36me2 domains including a 2kb flanking region (n=2 biologically independent replicates were integrated for this analysis). (f) Heatmaps showing signal intensities for H3K36me2, H3K27ac and chromatin accessibility (ATAC-Seq) around differentially regulated (> 2-fold) enhancers within H3K36me2 domains. Enhancers were called by measuring H3K27ac abundance in regions proximal to H3K36me2 domains excluding promoters (n=2 biologically independent replicates were integrated for this analysis). (g) Scatter plot showing H3K27ac abundance at enhancers embedded within H3K36me2 domains in WT and K36M reprogramming intermediates (n=2 biologically independent replicates were integrated for this analysis). Colors show overlap with MEF (beige) or ESC enhancers (green). (h) Fraction of differentially regulated (> 2-fold) enhancers embedded within H3K36me2 domains and overlapping with MEF (beige) or ESC enhancers (green) in WT and K36M reprogramming intermediates. (i) H3K4me3 enrichment at enhancer-proximal promoters (n=3,687) in WT and K36M cells on day 4, and of corresponding RNA expression. Center line indicates median (n=2 biologically independent replicates were integrated for this analysis). (j, k) Representative tracks of mesenchymal gene Vim, and of epithelial gene Cdh1 in WT and K36M reprogramming intermediates on day 4 (n=2 biologically independent replicates). Putative regulatory elements are highlighted in grey.

Fig. 5:. K36M-dependent chromatin rewiring alters transcription factor binding preferences.

(a) Scatter plots showing Sox2 occupancy at sites specific to early reprogramming intermediates (“ectopic”, left panel) and ESCs (“ESC-specific”, right panel) between WT and K36M reprogramming intermediates (n=2 biologically independent replicates were integrated for this analysis). (b) Representative tracks showing ectopic Sox2 binding to the Acta2 promoter (n=2 biologically independent replicates). (c) Representative tracks showing ESC-specific Sox2 binding at the proximal enhancer of Pou5f1 (n=2 biologically independent replicates). (d) DNA methylation at ectopic and ESC-specific Sox2 binding sites using published Methyl-seq data of MEFs (n=2) and RRBS data of ESCs (n=2)^,. (e) Scatter plots of differentially methylated regions (DMRs) between WT and K36M reprogramming intermediates on day 4 and day 8 (n=2 biologically independent replicates were integrated for this analysis). DMRs overlapping ESC-specific Sox2 binding sites are highlighted in purple. (f) Heatmaps showing percent DNA methylation at ectopic and ESC-specific Sox2 binding sites in MEFs, reprogramming intermediates on d4 and d8, and passaged iPSCs (n=2 biologically independent replicates for d4 and d8 samples, one sample for each genotype in uninduced MEFs and iPSCs). (g) Profile plots showing enrichment of Sox2 (left panel) and H3K36me2 (right panel) over differentially demethylated regions in WT (blue) and K36M (red) reprogramming intermediates on day 4 (n=2 biologically independent replicates were integrated for this analysis). (h) DNA methylation at ectopic and ESC-specific Sox2 binding sites in WT, Dnmt3a knockout (KO), and Dnmt3b KO MEFs, as well as in WT and Tet triple KO (TKO) ESCs (n=2). (i) Representative tracks showing DNA methylation at the Pou5f1 locus in reprogramming intermediates (RRBS) and WT/Dnmt3a KO/Dnmt3b KO MEFs (Methyl-Seq) (n=2 biologically independent replicates).

Fig. 6:. DNA demethylation is limiting for K36M-dependent enhancer activation and reprogramming.

(a) Dot blot assay to quantify 5hmC levels in MEFs as well as day 4 WT and K36M reprogramming intermediates in the presence and absence of DMOG (1 mM). (b) Flow cytometric quantification of Epcam levels on day 4 of reprogramming in untreated or DMOG-treated WT and K36M intermediates. (c) Quantification of Oct4-GFP⁺ cells detected in day-4 K36M reprogramming intermediates in the presence or absence of DMOG. P value determined by two-sided unpaired Student’s t test, error bars indicate mean ± SD (n=3 independent biological experiments). (d) Relative expression of the fibroblast gene Vim, the epithelial gene Epcam and the pluripotency gene Pou5f1 in untreated and DMOG-treated K36M intermediates. Error bars indicate mean ± SD (n=3). P values were determined by two-sided unpaired Student’s t test. (e) Fraction of Oct4-GFP⁺ K36M cultures transduced with either an empty vector or dox-inducible overexpression vectors for Dnmt3a and Dnmt3b, error bars indicate mean ± SD (n=3 independent biological experiments). (f) Quantification of Sox2, H3K27ac, and H3K27me3 levels at sites that are gained or lost in K36M vs. WT samples and the effect of DMOG on these enrichment patterns (K36M only). Center line indicates median (n=2 biologically independent replicates were integrated for this analysis). (g) Representative gene tracks (K36me2, H3K27ac, H3K27me3, Sox2, WGBS) showing the mesenchymal gene Prrx1 and the epithelial gene Cdh1 for WT, K36M, K36M+DMOG conditions on day 4 of reprogramming (n=2 biologically independent replicates). (h) Schematic of DMOG washout experiment (top), and quantification of Oct4-GFP⁺ cells on day 8 of reprogramming (bottom). K36M cells were either left untreated (Ctrl) or treated with DMOG for 8 days (DMOG d0-d8) or 4 days, (DMOG d0-d4). P values were determined by two-sided unpaired Student’s t test, error bars indicate mean ± SD (n=3 biologically independent experiments). (i) Summary highlighting the dual role of K36M-mediated H3K36me2 depletion on active mesenchymal vs. repressed epithelial and pluripotency genes.