Transcriptomic analysis functionally maps the intrinsically disordered domain of EWS/FLI and reveals novel transcriptional dependencies for oncogenesis

Abstract

EWS/FLI is the pathognomic fusion oncoprotein that drives Ewing sarcoma. The amino-terminal EWS portion coordinates transcriptional regulation and the carboxy-terminal FLI portion contains an ETS DNA-binding domain. EWS/FLI acts as an aberrant transcription factor, orchestrating a complex mix of gene activation and repression, from both high affinity ETS motifs and repetitive GGAA-microsatellites. Our overarching hypothesis is that executing multi-faceted transcriptional regulation requires EWS/FLI to use distinct molecular mechanisms at different loci. Many attempts have been made to map distinct functions to specific features of the EWS domain, but described deletion mutants are either fully active or completely "dead" and other approaches have been limited by the repetitive and disordered nature of the EWS domain. Here, we use transcriptomic approaches to show an EWS/FLI mutant, called DAF, previously thought to be nonfunctional, displays context-dependent and partial transcriptional activity but lacks transforming capacity. Using transcriptomic and phenotypic anchorage-independent growth profiles of other EWS/FLI mutants coupled with reported EWS/FLI localization data, we have mapped the critical structure-function requirements of the EWS domain for EWS/FLI-mediated oncogenesis. This approach defined unique classes of EWS/FLI response elements and revealed novel structure-function relationships required for EWS/FLI activation at these response elements.

Keywords: EWS/FLI; Ewing sarcoma; RNA-seq; intrinsically disordered domains; structure-function.

Figure 1. DAF regulation at EWS/FLI target genes is context-dependent

A. Schematic depicting EWS/FLI cDNA constructs introduced to cells. 3F denotes the N-terminal 3XFLAG tag. The EWS domain is blue and the FLI domain is gray. SYGQ subdomains are depicted in lighter shade, and tyrosine (Y) to alanine (A) mutations in DAF are represented in yellow. B,C) qRT-PCR data showing the fold change following rescue with EWS/FLI constructs at B. EWS/FLI-activated genes NKX2-2 and NR0B1 and C. EWS/FLI-repressed genes TGFBR2, IGFBP3, and LOX. Data shown depicts 3 technical replicates and is a representative sample of data acquired from 3 biological replicates. Mean and standard deviation are shown P-values were determined using a Tukey's honest significance test for multiple comparisons. *p < 0.05, **p < 0.01, ***p < 0.001, ns = not significant. D,E) Luciferase reporter assays show EWS/FLI and DAF activation from D. the isolated NR0B1 microsatellite paired with the SV40 or the native promoter and E. artificial microsatellites upstream of the SV40 promoter. Data are plotted as the ratio of firefly to Renilla luminescence units to control for transfection efficiency, and show 3 technical replicates of a representative result from 3 independent experiments.

Figure 2. Partial rescue of EWS/FLI global transcription by DAF is insufficient for transformation

A. Experimental schematic depicting study design for a single replicate. Briefly, cells were transduced with shRNA targeting either luciferase as a control (iLuc) or EWS/FLI (iEF). iEF cells were subsequently rescued with different constructs. After 10 days of selection, these cells were harvested for protein or RNA or seeded into agar. For data analysis purposes, iEF is used as the control to build a transcriptional profile for each EWS/FLI construct. B. Table showing number of genes differentially expressed for each construct as compared to iEF, using a adjusted p < 0.05 cutoff (Benjamini-Hochberg correction). C. Principle component analysis (PCA) plot of the transcriptional profiles of different test conditions. Principle component 2 on the y-axis is plotted against principle component 1 on the x-axis. Different cell conditions are depicted by color and different replicates are represented with different shapes. D-F) Venn diagrams comparing the overlapping genes differentially expressed by D. WT-EF and endogenous EWS/FLI, E. WT-EF and Δ22, and F. WT-EF and DAF. Genes included in these analyses met a cutoff of |fold change| > 2 and adjusted p < 0.05 (Benjamini-Hochberg). P-values were determined using a Chi-square test. G. Volcano plots of differentially expressed genes comparing rescued cells against unrescued cells. The -log(p-value) is plotted against the -log2(FoldChange) for each gene. Genes meeting a cutoff of |log2(FoldChange)| > 1 are shown in yellow. Genes meeting a cutoff of adjusted p < 0.05 are shown in red. Genes meeting both cutoffs are shown in green. H. Colony formation assays of cells used for transcriptional profiling. Representative agars are shown on the left with corresponding colony counts on the right. Counts are depicted as mean and standard deviation of 3 technical replicates, and these are a representative sample of 3 independent experiments. P-values were determined using a Tukey's honest significance test for multiple comparisons. *p < 0.05, **p < 0.01, ***p < 0.001, ns = not significant.

Figure 3. DAF-rescued genes have a distinct function from those rescued only by WT-EF

A. Heatmap showing unsupervised hierarchical clustering of the 100 most variable genes across samples from RNA-seq analysis. Red depicts activation and blue depicts repression. This analysis identified a cluster of genes regulated by both DAF and WT-EF. B. The top five categories from ToppGene functional enrichment analysis are shown for the Gene Ontology: Molecular Function category. Functional enrichment was performed for genes activated either only by WT-EF or by WT-EF and DAF. Genes were included in analysis if they met a cutoff of 2 fold-change and Benjamini-Hochberg adjusted p-value < 0.05. C,D) Top five enriched motifs for genes activated either only by C. WT-EF or by D. WT-EF and DAF using HOMER de novo motif enrichment analysis of target gene promoters. The letters are color coded by base: green for adenine, red for thymine, blue for cytosine, and yellow for guanine. The size of the letter reflects the prevalence of that base in that particular position in the detected DNA motif. The larger the letter, the more conserved that particular base is in that position across instances of the motif, Genes were included in analysis if they met a cutoff of 2 fold-change and Benjamini-Hochberg adjusted p-value < 0.05.

Figure 4. DAF most robustly rescues EWS/FLI function at activated microsatellites

A. Schematic depicting workflow used to analyze gene rescue at direct microsatellite targets. B. Scatterplots depicting construct-specific rescue of activated microsatellites. Data is plotted as log2(FoldChange) of each rescue construct on the y-axis against the log2(FoldChange) of regulation by endogenous EWS/FLI on the x-axis. Dotted lines depict x = 0 and y = 0. Data points in red represent genes whose change in expression was significant in both KD and rescue conditions. Data points in gray indicate genes whose change in expression was significant in only the KD condition. Data points in yellow indicate genes whose change in expression was significant in only the rescue condition. Data points in black indicate genes with no detectable change in expression in these experiments. Significance was defined as Benjamini-Hochberg adjusted p < 0.05 with no fold-change cutoff. Lines of best fit were derived from the linear model only for genes with significant changes in both KD and rescue conditions to represent the “volume” of rescued activity. Pearson correlation coefficients and their p-values are depicted with the slope from the linear model in the boxed inset. The pie chart inset for each panel shows the proportion of genes belonging to each functional group for that construct. C,D) Compilation depicting the rescue of direct EWS/FLI targets by different constructs at C. activated and D. repressed target genes. Genes included in this analysis were only those genes with detectable differential expression by endogenous EWS/FLI, and thus represent the “red” and “gray” scatterplot data for activated and “blue” and “gray” scatterplot data for repressed genes. In each pie chart, gray depicts the proportion of genes which are not rescued. Red depicts the proportion of genes that are differentially activated and blue depicts the proportion of genes that are differentially repressed by a given construct.

Figure 5. mut9 shows transcriptional activity intermediate to DAF and WT-EF, and transforms cells

A. Schematic depicting EWS/FLI cDNA constructs introduced to cells. 3F denotes the N-terminal 3XFLAG tag. The EWS domain is blue and the FLI domain is gray. SYGQ subdomains are depicted in lighter blue, and tyrosine to alanine mutations in DAF are represented in yellow. B. Colony formation assays of cells used for transcriptional profiling. Counts are depicted as mean and standard deviation of 3 technical replicates, and these are a representative sample of 3 independent experiments. P-values were determined using a Tukey's honest significance test for multiple comparisons. *p < 0.05, **p < 0.01, ***p < 0.001, ns = not significant. C. Volcano plots of differentially expressed genes as comparing rescued cells against unrescued cells. The -log(p-value) is plotted against the -log2(FoldChange) for each gene. Genes meeting a cutoff of |log2(FoldChange)| > 1 are shown in yellow. Genes meeting a cutoff of adjusted p < 0.05 are shown in red. Genes meeting both cutoffs are shown in green. D. Table showing number of genes differentially expressed for each construct as compared to iEF, using a cutoff as adjusted p-value less than 0.05 using a Benjamini-Hochberg correction. E. Unsupervised hierarchical clustering of constructs by their respective differential expression of 66 core transcriptional regulator genes identified by GO:0043565 (Figure 3B).

Figure 6. Distinct features of EWS contribute to regulation at different direct EWS/FLI targets

A, B. Compilation depicting the rescue of direct EWS/FLI targets by different constructs at A. activated and B. repressed target genes. Genes included in this analysis were only those genes with detectable differential expression by endogenous EWS/FLI, and thus represent the “red” and “gray” scatterplot data for activated and “blue” and “gray” scatterplot data for repressed genes. In each pie chart, gray depicts the proportion of genes which are not rescued. Red depicts the proportion of genes that are differentially activated and blue depicts the proportion of genes that are differentially repressed by a given construct. C. Global model of the involvement of different features of EWS in different types of direct EWS/FLI target genes.