Fig 2: Loss of H3K36me2 domains and enhancer H3K27ac affects expression of target genes(A) Log fold-change (LFC) of various subsets of DEGs. The lower and upper hinge in the box plots correspond to the first and third quartile; the upper whiskers extend to the largest value = 1.5 * IQR and vice versa for the lower whiskers.(B) LFC of putative target genes for various differential binding (DB) site subsets. For the box plots, the lower and upper hinge correspond to the first and third quartile; the upper whiskers extend to the largest value = 1.5 * IQR and vice versa for the lower whiskers.(C) Logistic regression model outputs for expression downregulation status based on its distance to and/or whether it shares a TAD with a cluster B bin.(D) Permutation test on downregulated genes’ tendency to share a TAD with cluster B bin, controlling for distance.(E) Example loci illustrating genome-wide phenomenon using differential signal tracks in which enrichment values of the respective parental line were subtracted from the corresponding KOs, after which the average across lines was taken.(F) Schematic model of epigenetic dysregulation resulting from the absence of NSD1 (created with https://biorender.com). Note that, in the absence of NSD1, PRC2 deposits H3K27me3 in the same intergenic regions where H3K36me2 was depleted. In addition, H3K27ac decreases around distal enhancers located in these H3K36me2-depleted regions.

Fig 3: Changes in transcriptome and pathways resulted from loss of NSD1 and reduced H3K36me2 levels(A) GSEA enrichment plot of hallmark gene sets significantly associated with the aggregated ranking of differentially expressed genes and genes targeted by differentially acetylated enhancers using their test statistics.(B) Motifs exhibiting differential activity between up- versus downregulated peaks, with dots representing the strength of association in each direction and triangles their difference.(C) Aggregated motif density plots around differential H3K27ac sites for the most significant differentially associated motifs in each direction.(D) Stratified rank-rank hypergeometric overlap plot of gene expression differences between NSD1-WT versus MT and WT versus KO.(E) Distribution of expression changes for leading edge genes of hallmark gene sets significantly associated with the aggregated ranking of differential gene expression for both NSD1-WT versus MT and WT versus KO. The error bars indicate the interquartile range (IQR).

Fig 4: Epigenomic characterization of NSD1-WT and KO HNSCC cell lines(A) Genome-wide prevalence of modifications based on mass spectrometry; diamonds represent within-condition averages; p values obtained using Welch’s t test: H3K36me2 p = 0.04; H3K27me3 p = 0.16. In this plot and all the following ones, Cal27-KO corresponds to replicate 1, Det562-KO to replicate 2, and FaDu-KO to replicate 1, unless otherwise specified.(B) Genome-browser tracks displaying individual cell-line differences (KO-PA) as lines and condition averages as area plots in a lighter shade; ChIP-seq signals shown are MS-normalized logCPM and beta values are used for WGBS; regions of noticeable difference in Figure 1B are highlighted.(C) Heatmaps showing H3K36me2 enrichment patterns centered on IGRs; n = 10,630. Numbers displayed at the bottom of aggregate plots correspond to the intergenic/genic ratio where TSS/TES and outer edges are excluded.(D) Distribution of differential beta values within actively transcribed genes, all genes, intergenic regions, and regions depleted of H3K36me2 (corresponding to regions defined in Figure 3A as “cluster B”); median values are shown at the top. For the boxplots, the lower and upper hinge correspond to the first and third quartile; the upper whiskers extend to the largest value = 1.5 * IQR and vice versa for the lower whiskers.(E) Spearman correlation of differential enrichment between NSD1-WT versus KO and WT versus MT.

Fig 5: Loss of NSD1 preferentially impacts distal intergenic cis-regulatory elements(A) Scatterplots of H3K36me2 enrichment (10-kb resolution) comparing a representative WT parental sample (Cal27) against its NSD1-KO counterpart (replicate 1; see also Figure S3A for other cell lines).(B) Overlap enrichment result of Ensembl annotations with bins consistently labeled as cluster B (i.e., identified as B in all three paired WT versus KO comparisons). Stratification is applied to only focus on intergenic regions to avoid spurious associations to annotations confounded by their predominantly intergenic localization. The size of the dots corresponds to number of bins overlapping the corresponding annotation. *** represents p-value < 1e-5 based on Fisher’s exact test of bins overlapping a specific class of annotated regions versus a background of all non-quiescent bins, meaning >10 reads in at least one mark in one sample.(C) Aggregate plots of differential signal enrichment centered around CREs overlapping (n = 5,193) consistent cluster B bins. Values are averaged across all three WT versus KO comparisons.(D) Log fold change of H3K27ac normalized enrichment values comparing all differentially bound sites to those overlapping consistent cluster B bins. For the box plots, the lower and upper hinge correspond to the first and third quartile; the upper whiskers extend to the largest value = 1.5 * IQR and vice versa for the lower whiskers.(E) Distribution of genomic compartments overlapping various subsets of H3K27ac peaks categorized by differential binding status.