(P 427) UNDERSTANDING THE MOLECULAR DRIVERS THAT REGULATE EWING SARCOMA CELL FATE

Objective: Like other translocation-positive sarcomas, Ewing sarcoma (ES) harbors a pathognomonic fusion protein (FP) capable of widespread epigenetic reprogramming that can lead cells to acquire a high-grade, undifferentiated phenotypic state. Though EWS-FLI1’s action as an aberrant transcription factor with GGAA-microsatellite and ETS motif binding sites is well known, it remains to be determined precisely how its genome-wide stimulatory and repressive effects coalesce to block differentiation, alter cell plasticity, or promote a quasi-stem-like cell state that better resembles neuroectodermal cells or mesenchymal stem cells.
We first performed an unsupervised comparison of the ES transcriptome to other cancers available in the Cancer Cell Line Encyclopedia (CCLE) and existing tumor expression datasets to investigate this. Though traditionally classified as a ‘bone cancer’, ES cells clustered far from other sarcoma subtypes or malignancies. This unique transcriptional signature implied that the ES FP’s effects are so broad sweeping as to invoke a ES-specific phenotypic state that strongly diverged from other known cell types. To better understand how this might occur, we employed a CRISPR knockout (CRISPR-KO) model of A673, targeting exon 4 of the N-terminus EWSR1 portion of the FP and generated a clonal pool (EWSR1 CRISPR KO). We used single-cell RNA and ATAC sequencing (scMultiome) to identify the transcriptional and chromatin changes after the FP knockout (FP KO).

Methods: To assess gene expression and chromatin accessibility, we performed Chromium Single Cell Multiome ATAC + Gene Expression on A673 WT and A673 EWSR1 CRISPR KO (EWSR1-KO) cells. Briefly, filtered gene barcoded matrices from Cell Ranger (10x Genomics) were processed by Seurat v4. Low-quality nuclei and additional outliers were removed. The data were log-normalized, and the top variable genes were selected for PCA dimensional reduction. ATAC-seq data were analyzed with the Signac package. Single cells were filtered for greater than 1,000 fragments, percent reads in peaks of at least 15%, nucleosome signals less than 10, and TSS enrichment score greater than 2. Data were normalized, followed by dimensional reduction with latent semantic indexing. We integrated the two data modalities using weighted neighbor clustering for integrative analysis of both RNA and ATAC data. chromVar detected motif enrichment per cell in sparse single-cell ATAC-seq data. Peak to gene annotations and transcription factor (TF)-gene regulatory network (GRN) were found by the Functional Inference of Gene Regulation (FigR) framework. We assessed differentiation by CytoTRACE, a computational tool designed to predict differentiation states.

Results: We identified eleven cell clusters of ES cell heterogeneity. Since the EWSR1-KO is a pooled population, we used known FP-regulated genes to delineate cells bearing a FP KO from wild-type (FP WT) cells. We observed robust expression of FP target genes (EWS-FLI1 On), PRKCB, and LIPI in ten clusters, suggesting regular FP function (FP WT). Conversely, FP repressed genes (EWS-FLI1 Off), LOX, and IGFBP3 were elevated in cluster 10, which contained only EWSR1-KO cells. This suggests that the FP was successfully knocked out in cluster 10 (FP KO).
Next, we hypothesized that loss of the FP would induce a mesenchymal fate. To answer this, we used SingleR, which compares the gene expression profile of cancer cell lines to identify the closest cell type. The FP KO cluster did not resemble ES, but all other clusters identified closely to ES and related tumors. A granular identification with cell lines confirmed that all clusters except FP KO were identified as A673 cells. However, the FP KO resembled cell lines with a fibroblastic phenotype.
scMultiome enabled us to study the relationship between the FP target, repressed genes, and chromatin accessibility. We observed no significant difference between the FP KO and FP WT when investigating motifs for the FP or FLI1. However, the FP KO cells were enriched for motifs in the Fos/Jun family and Smad2/3, which have roles in EMT and bone and mesenchymal differentiation. The FP-KO cells significantly expressed the Fos/Jun family of proteins. Next, we identified peak-gene associations within a  500 kb window around the transcriptional start sites. We used HOMER motif discovery algorithm to search for motifs at these peaks. When we analyzed peaks linked to genes expressed in FP-WT, we found enrichment of the Fos/Jun protein family (p < 1e-34), Ets family, and GGAA-microsatellites. However, the FP-KO were strongly enriched for Fos/Jun protein family (p < 1e-540) with no change in Ets family motif.

Conclusion: Our data suggest that knocking out the FP in the A673 cell line significantly shifted the ES lineage. While the FP KO did not lead to cell death in the A673 cell line, our results suggest a novel gene regulatory network through the Fos/Jun family of proteins that may act to sustain an ES lineage that is not dependent on the FP. We observed that peaks linked to genes positively associated or regulated by the FP were enriched for Fos/Jun family, Ets family, and GGAA-microsatellites. This suggests a cooperative interaction between these TFs may drive the genes linked to the FP. Interestingly, upon losing FP, there was a significant enrichment of Fos/Jun family motifs but not Ets family or GGAA-microsatellites. This suggests that FP loss activates FP-repressed genes and enables chromatin remodeling to open previously inaccessible DNA binding sites.
Significant questions remain. It is unknown if opening Fos/Jun repressed sites with an active FP would still lead to a mesenchymal phenotype, and it remains to be determined whether our findings apply to other ES cell lines that poorly tolerate FP loss.