Research ArticleCANCER

Functional antagonism of chromatin modulators regulates epithelial-mesenchymal transition

See allHide authors and affiliations

Science Advances  24 Feb 2021:
Vol. 7, no. 9, eabd7974
DOI: 10.1126/sciadv.abd7974
  • Fig. 1 A dominant role for chromatin factors as regulators of EMT homeostasis revealed by phenotypic CRISPRi screens.

    (A) Schematic of EMT reporter lung cancer cell line generation. (B) Diagram of phenotypic CRISPRi screen. (C) Selection of quasi-epithelial H1944 and quasi-mesenchymal A549 cell lines by epithelial and mesenchymal gene expression scoring of human NSCLC cell lines (see Materials and Methods). Lung fibroblasts (mesenchymal) are shown for comparison. CCLE, Cancer Cell Line Encyclopedia. (D) Immunoblotting of samples from intermediate time point of the CRISPRi screen. MGT#1 fluorescence micrograph (above) was taken before lysis. (E) Gating strategy for FACS purification of MGT#1-high and MGT#1-low populations. (F) Scatter plot of differential enrichment analysis for sgRNAs from the indicated libraries in the “GSK126 + dox” arm of the A549-MGT#1 screen. Fold change (FC) and significance were calculated by comparing MGT#1-mVenus–high and MGT#1-mVenus–low populations, upon performing all the relevant quality controls (see Materials and Methods and figs. S2 and S3). Yellow denotes gRNAs significant by FC and P value. Blue dots are significant by P value (i.e., potential hits) above the nonsignificant gray dots. Dot size represents the absolute log2FC value. (G) MA plot of gRNA abundance (x axis) and difference in gRNA abundance (y axis) in the GSK126 + dox arm of the A549-MGT#1 screen. Dot color and labels are consistent with (F), whereas size is fixed.

  • Fig. 2 Genetic loss of potential EMT regulators phenocopies CRISPRi screen.

    (A) Gating strategy for assessing fluctuations in EMT by FACS analysis of MGT#1 expression. SSC-A, side scatter area; FSC-A, forward scatter area. (B) Bubble plot of the FACS analysis of MGT#1 expression in A549-MGT#1 carrying individual KO mutations. (C) FACS analysis of MGT#1 fluorescence intensity in A549-MGT#1 cells with the indicated genotypes and treatments. WT, wild type; DMSO, dimethyl sulfoxide. (D) FACS analysis of MGT#1 fluorescence intensity in A549-MGT#1 cells with the indicated treatments. (E) Immunoblot of E-cadherin expression in A549-MGT#1 carrying individual KO mutations. Quantification of normalized protein band intensity is shown.

  • Fig. 3 Context-dependent cellular phenotypes by genetic loss of individual EMT regulators.

    (A) Violin plot showing the distribution of the dependency for proliferation of all CCLE NSCLC cell lines for individual genes within the indicated families. The plot shows the phenotypic dependency calculated by the DepMap project. CRISPR (Avana) CERES score and RNAi DEMETER2 score are displayed in the left and right panels, respectively. Selected genes are labeled within their families. ANOVA, analysis of variance. (B) Bar plot of control and mutant A549-MGT#1 cell proliferation in parallel longitudinal analysis. Statistics: Significant by t test and Holm-Sidak post hoc test (P < 0.05; n = 4), BRD2, EPC1, ARID1A, and ACVR1 KOs versus control. (C) Bar plot of control and mutant A549-MGT#1 ± GSK126 cell colony formation assay. Statistics: Significant by t test and Holm-Sidak post hoc test (P < 0.05; n = 3) in DMSO group: ARID2, ARID1A, DOT1L, and ACVR1 KOs; GSK126: EPC1, ARID1A, BRD2, DOT1L, KMT2A, and ACVR1 KOs. (D) Left: Line plot of parallel longitudinal high-content wound healing analysis of A549-MGT#1 cells with the indicated genotypes under homeostatic conditions. Each dot represents the mean in each time point. Statistics: Two-way ANOVA and Dunnet post hoc test (n = 4). Asterisks denote significance for the indicated comparison. Antagonistic regulators of EMT and motility in A549 cells are shown to the right. (E) Left: Schematic representation of three-dimensional (3D) invasion assay. Right: Migration depth of DRAQ5-stained nuclei for each time point and clone normalized to time point T = 0 hours from high-content imaging. Statistical analysis for time point 24 hours shows corrected multiple t test (*P < 0.05; ***P < 0.001; n = 4). (F) FACS analysis (left) and quantification (right) of MGT#1 expression in lung and brain tumor cells with the indicated genotypes.

  • Fig. 4 Antagonistic chromatin regulators bind to core set of accessible targets in lung cancer cells.

    (A) Heatmap of signal intensity for ZMYND8, BRD2, ARID1A, DOT1L, acetyl- and trimethyl-H3K27, and immunoglobulin G (IgG) occupancy at the indicated genomic loci. k-means clustering was used to partition chromatin occupancy into three clusters according to ZMYND8. The number of genes included in each of the clusters and of chromatin regions is indicated to the left and below the heatmap, respectively. (B) Top five GO terms per core set of Cluster I, II, and III genes (color, adjusted P < 0.05; size, gene ratio). (C) Pie charts showing the genomic distribution of the indicated ChIP-seq peaks. Note that the ZMYND8, BRD2, DOT1L, and ARID1A binding mode mirrors the enhancer-decorating mark H3K27ac. (D) IGV view of the indicated ChIP-seq tracks for known epithelial and mesenchymal markers. For each track, scale values are indicated to the left. (E) IGV view of ZMYND8, BRD2, ARID1A, DOT1L, acetyl- and trimethyl-H3K27, and IgG occupancy at the MGT#1 reporter loci. (F) Dendrogram showing hierarchical clustering of the indicated ChIP-seq tracks for loci from (A). Note the dominant effect of TGF-β1 on the clustering. (G) Density plot (above) and heatmap (below) of the indicated ChIP-seq tracks for TGF-β–regulated loci significant by DESeq2 (padj < 0.05). (H) Bubble plot showing the expression data for the selected genes in the indicated conditions. Bubble size and color indicate FC compared to control and normalized expression per sample, respectively.

  • Fig. 5 Antagonistic chromatin regulators targets include a pancancer meta-EMT signature influencing molecular tumor classification.

    (A) Top: Heatmap showing 1081 epithelial and mesenchymal LUAD biopsies defined by the dual scoring analysis is clustered into five clusters by NMF consensus on the basis of cluster I + II + III genes (see Materials and Methods). Selected EPI and MES metagenes are highlighted to the right. Bottom: Composition of the tumor immune infiltrate. For each NMF cluster, boxplots show the proportion of six major classes of immune cells (from CIBERSORT). Lymphocytes represent the reference, and the other five classes are compared to lymphocytes. Significance is indicated by asterisks (Wilcoxon rank sum test, adjusted P < 0.05). (B) Relative frequencies of each NMF cluster in the MES and EPI patients of 32 cancer types, respectively; significance of enrichment was calculated by Fisher’s exact test and highlighted with asterisks (see Materials and Methods): *P < 0.05; **FDR (false discovery rate) < 0.05. (C) Kyoto Encyclopedia of Genes and Genomes (KEGG) pathways displayed were significantly associated with NMF clusters C1 to C5. Two panels referred to the log10(FDR) and gene count ratio. (D) Kaplan-Meier survival analysis for five NMF clusters patients of LUAD, KIRC, and STAD. Time is indicated in days. Different patient groups are compared using the log-rank test, for which the P value is shown. (E) Uniform manifold approximation and projection (UMAP) clustering of patients with LUAD based on either all genes (left) or signature genes from NMF cluster C1 (right). The latter represents the meta-mesenchymal pan-cancer signature genes.

  • Fig. 6 Antagonistic chromatin regulators control meta-EMT activity via enhancers and promoters regulated by RNAPII elongation and converging on the AP-1 transcriptional network.

    (A) Pie charts showing the genomic distribution of the RNAPII in naïve and GSK126-treated A549. The NMF signatures genes from Fig. 5A are highlighted as a subset of promoters (orange). (B) Cumulative distribution plot of the genome-wide RNAPII loading at 5′ and 3′ of each transcript, as defined above. (C) Heatmap (above) and violin plot (below) of RNAPII traveling ratio at 5′ and 3′ of each transcript for naïve and GSK126-treated cells. Clustering was performed using k-means, with k = 3, and the “ns” denotes no significance by one-way ANOVA and Sidak post hoc test. (D) Relative RNAPII traveling ratio between naïve and GSK126-treated cells on selected genes, as defined above. A ±0.5 threshold was decided on the MGT#1-mVenus reporter (more stringent), and selected examples are shown. (E) IGV view of FOS. The asterisks denote changes in elongation as determined in (D). (F) Upstream regulator analysis by IPA on the genes passing the FC in (D). The AP-1 transcription factor components are highlighted in red. (G) Giraph plot showing the distance between JUN ChIP-seq peak lists. Colors denote the known state of the cell line in which ChIP-seq was performed. (H) Heatmap of signal intensity for the indicated ChIP-seq profiles for all JUN peaks in (G). Direct overlap with genomic loci in Fig. 4A was used to partition chromatin occupancy into the indicated clusters. (I) Pie charts showing the genomic distribution of all JUN peaks in (G), above. NMF C1 genes were annotated when a JUN peak was close to the gene (−2.5 and +0.5 kb). The below charts are referred to the two clusters in (H). (J) Upstream regulator analysis by IPA on the genes annotated in (I) as direct JUN/ZMYND8-BRD2 targets.

Supplementary Materials

Stay Connected to Science Advances

Navigate This Article