(P 426) MULTI-OMIC ANALYSIS OF DESMOPLASTIC SMALL ROUND CELL TUMORS REVEAL DISTINCT LINEAGES

Objective: Desmoplastic small round cell tumor (DSRCT) is a rare and usually incurable aggressive sarcoma subtype. All malignant cells harbor a pathognomonic EWS-WT1 fusion protein (FP), but FP-targeted agents are nonexistent, and less than 20% of patients survive beyond five years. Lacking the availability of FP-targeted drugs, we seek to better understand how the FP regulates tumor cells to promote the characteristic multilineage phenotypes of epithelial and neuroendocrine (NE) cell fates to discover a therapeutic opportunity.
Our data showed that a subset of DSRCT cells (mostly of epithelial lineage) highly express the androgen receptor (AR), which is the pivotal epigenetic driver of prostate cancer (PCa). We showed that DSRCT depended on AR signaling by targeting AR with enzalutamide, a nonsteroidal antiandrogen, where we observed delayed tumor growth response in a DSRCT xenograft and a patient-derived xenograft. Our team also revealed that the AR cis-elements were enriched for specific TF motifs, including those from forkhead box (FOXA1), response elements for nuclear receptors (ARE, GRE, PGR), and zinc finger proteins (WT1, EGR1). Given the observation of the WT1 (the C-terminal partner of the pathognomonic fusion protein) binding motif, we hypothesized that the FP, FOXA1, and nuclear receptors interact together at the regulatory sites of AR, driving the epithelial and NE lineages in DSRCT.
Herein, we share the results of gene expression and chromatin accessibility analyses to explore fundamental mechanisms that may be driving epithelial and NE biology in DSRCT.

Methods: To assess gene expression and chromatin accessibility, we performed Chromium Single Cell Multiome ATAC + Gene Expression on thirteen fresh-frozen specimens from nine patients. We delineated malignant cells from normal cells using EWS-WT1 gene signatures, copy number variation, and CytoTRACE analysis. CytoTRACE is a computational tool designed to predict differentiation states. We assessed multilineage expression using epithelial and neuroendocrine signatures. Using the ATAC-seq data, we observed significant peaks and motifs associated with different lineages. To assess spatial heterogeneity and proteomic profile, we used the Lunaphore COMET, which enables up multiplex immunofluorescent imaging of slides.

Results: We profiled 85,640 nuclei from nine DSRCT patients using single-nucleus RNA-sequencing and ATAC-sequencing from fresh-frozen tissue. We found a subset of DSRCT patients with low or negative expression of AR that expressed partial NE reprogramming, akin to NE prostate cancer (NEPC, Panel A). One subset (NE: T6, T15, T20, T21, DSRCT-1) exhibited NE markers (SYP, ENO2, CHG, FOXA2, ASCL1, and SOX2), while a separate subset (Epithelial: T14, T19, T3) expressed epithelial markers (MUC1, MUC6, KRT18, KRT23, CDH1). T5a appeared to exhibit both sets of markers – implying a hybrid phenotype or perhaps poorly differentiated phenotype. Additionally, COMET data showed that AR and cytokeratin were anti-correlated with neural specific enolase, an NE marker. This suggests that DSRCT could be classified into three subtypes: 1) Epithelial/AR-positive, 2) NE-positive, and 3) hybrid AR/NE (Panel B).
Given the heterogeneous expression programs in DSRCT, we asked if these states were related to differentiation and plasticity. We used CytoTRACE to predict differentiation states. Expectedly, normal cells (i.e., fibroblasts, and immune cells) were the most differentiated. Neoplastic cells, identified by EWS-WT1 gene targets and copy number variations, had lower differentiation scores, indicative of a dedifferentiated phenotype. The NE subtype had the lowest differentiation score indicative of more plastic capacity, which has previously been implicated in NEPC.
Single-cell ATAC-seq data revealed enriched motifs associated with different DSRCT subtypes. The WT1 motif was enriched in clusters identified as malignant cells. Epithelial/AR-positive subtypes showed enrichment of DNA-response elements for nuclear receptors (ARE, GRE, and PGR) and the forkhead box family of transcription factors (FOXA1, a known pioneer factor for AR). The NE subtype lacked the ARE, GRE, and PGR motifs but was enriched with FOXA1 and neural-specific motifs including NEUROD1, ATOH1, OLIG2, and NANOG (Panel C).

Conclusion: Our results demonstrate significant heterogeneity in DSRCT phenotypes, where we observe both epithelial and NE lineages. Initially, we only detected the Epithelial/AR-positive and NE-positive lineages by gene expression, however, the chromatin accessibility data also supports this result. Next, using CytoTRACE, the NE subtype appeared to exist in a high-plasticity cell state unlike the Epithelial/AR-positive subtype, which directly parallels the relationship between Epithelial/AR and NE that is observed in PCa. The emergence of the NE subtype in PCa is generally in response to therapy. However, it is not known how NE emerges in DSRCT. The CytoTRACE analysis may suggest that NE may have de-differentiated from the more differentiated Epithelial/AR-positive subtype. Under active investigation, we are working to better understand the Epithelial/AR and NE in DSRCT dimorphic states. Given the presence of FOXA1 enriched motifs for both Epithelial/AR and NE, we have also begun to interrogate the FOXA1 cistrome to determine if it acts as a central regulator of DSRCT lineage specification.