(P 129) ELUCIDATING THE EVOLUTIONARY DYNAMICS OF DDLS LIPOSARCOMA USING SINGLE-CELL SEQUENCING AND MATHEMATICAL LINEAGE TRACING

Objective: Well-differentiated and dedifferentiated liposarcomas (WDLS; DDLS) are the most common and clinically aggressive types of genetically complex sarcoma that largely develop and evolve through copy number alterations (CNAs). WDLS and DDLS are characterized by 12q amplification, encompassing the oncogenes CDK4, MDM2, and HMGA2 in > 95% of cases. More than 60% of patients with retroperitoneal WDLS and DDLS eventually die of disease despite complete surgical resection. Moreover, chemotherapy is associated with substantial toxicity and low response rates. The development of new targeted therapies is vital for improving patient outcomes; however, the complexity of CNAs in these liposarcomas has made it difficult to find the true drivers of oncogenesis and disease progression.

Methods: Here we employ single cell genomic analysis to study CNA evolution in a DDLS patient-derived xenograft (PDX) model. Regrettably, existing tools for tracking CNAs at the single-cell level demand extensive computational resources, often taking many hours to analyze a single tumor. To address this limitation, we developed the Maximum Entropy Lineage Tracing (MELT) algorithm, a fast and scalable lineage tracing tool for single-cell copy number data. To validate our approach and uncover copy number signatures of liposarcoma evolution, we performed single-cell copy number profiling on 545 cells across five consecutive passages of this established DDLS PDX model.

Results: In addition to identifying the well-known 12q amplification associated with tumor initiation, MELT unveiled a lineage-specific gain of chromosome 5q21-35, which encompasses several oncogenes, including IL3, PDGFRB, TLX3, FGFR4, and FLT4. The prevalence of this lineage increased from approximately 5% in passage 1 to 35% in passage 6 and was associated with an enriched proliferation transcriptomic signature and positive immunohistochemical staining for PDGFRB and FGFR4. Moreover, MELT accomplishes this with accuracy comparable to leading methods but in seconds rather than hours.

Conclusion: The MELT algorithm was developed to resolve the accumulation of essential molecular drivers of tumor evolution with unprecedented efficiency and granularity utilizing single cell copy number data. MELT identified the evolution of gain on 5q21-35 that was associated with an increase in cellular proliferation in a DDLS PDX. MELT shows promise for comprehensively characterizing CNA’s and tracing tumor evolution in liposarcoma tissue samples, cell lines and PDX models.