(P 65) EPIGENETIC HETEROGENEITY AND STATE SPECIFIC TRANSCRIPTIONAL PROGRAMS DRIVE DRUG RESPONSE IN OSTEOSARCOMA

Objective: To define epigenetic subtypes in osteosarcoma and the relevance of these subtypes to drug response

Methods: We used a panel of PDX-derived cell lines derived from primary, resection and metastasis to define epigenetic subtypes. Using ATACseq and H3K27ac as well as RNAseq, we analyzed variability in chromatin accessibility and how this correlates to gene expression. We then analyzed a large panel of human primary OS RNAseq data to determine the relevance of the epigenetic analysis for human OS.

Results: Osteosarcoma exhibits a range of genetic changes, contributing to its heterogeneity. We aim to delve into chromatin accessibility, a higher level of regulation, to clarify how it influences gene expression and determines cellular states. By evaluating chromatin accessibility using ATAC-seq, we identified two distinct cellular states in OS, Epigenetic cluster 1 (EC1) and Epigenetic cluster 2 (EC2). These epigenetic clusters are driven by distinct sets of transcription factors (TF) and correlate with bone development. EC1 represents an early differentiation state regulated by homeobox TFs, exhibiting a profile of accessible chromatin more akin to human skeletal stem cells. Conversely, EC2-accessible regions are regulated by AP1 members. We evaluated whether these TFs are dependencies using the pediatric Depmap. RUNX2, SP7 and SATB2 are specific dependencies for EC1 and FOLS1 and JUN are specific dependencies for EC2. To interrogate whether these ECs are driven by enhancer regulation, H3K27ac CUT&RUN assays were performed. The ECs show a distinct set of unique acetylated regions. To understand the essential TF interactions driving the osteosarcoma transcriptional program in each EC, we constructed models of CRCs. We utilized H3K27ac domains to define superenhancers (SE) and ATAC-seq to demarcate discrete cis-regulatory elements within these domains and find TF motifs. EC-specific SEs were associated with differentially expressed genes in each EC. MYC was identified as a CRC member for all the cell lines in EC1. To define a comprehensive set of genes per EC and evaluate the presence of these cellular states in patient samples, ATAC-seq and RNA-seq data was integrated from a panel of matched cell lines. Patient samples contain the EC-gene signatures and can be classified accordingly. EC2 shows better prognosis than EC1. To identify possible therapeutic implications of these epigenetic-defined clusters, we evaluated a drug response dataset generated across 40 different targeted agents. ECs display differential drug responses, underscoring the potential for tailored therapeutic approaches in osteosarcoma management.

Conclusion: In summary, we have identified two epigenetically cell states regulated by a state-specific set of transcription factors defining a gene signature that could help to define osteosarcoma subclasses and predict drug response.