(P 137) TARGETING PRIMARY AND SECONDARY ONCOGENIC DRIVERS IN RETROPERITONEAL DEDIFFERENTIATED LIPOSARCOMAS

Objective: Dedifferentiated liposarcoma (DDLPS) is characterized by the amplification of MDM2, CDK4, and HMGA2. Targeted therapies directed at these alterations have shown limited antitumor activity when administered as monotherapy. This study aimed to determine whether well-differentiated liposarcoma (WDLPS) undergoes progressive evolution to DDLPS by analyzing the transcriptomic profiles of retroperitoneal liposarcoma. To achieve this, we leveraged the well-differentiated (WD) component of DDLPS to investigate the evolutionary trajectory between WDLPS and DDLPS. Additionally, the study sought to identify and validate therapeutic targets for DDLPS, including primary oncogenic drivers (MDM2 and CDK4) and secondary alterations identified through transcriptomic analysis, using in-house generated patient-derived DDLPS tumor models.

Methods: In a contemporary retrospective series of primary retroperitoneal liposarcoma (N=107), we sampled paired tumor (DD and/or WD) and normal fat (NF) components of DDLPS and WDLPS subtypes. Transcriptomic profile were measured with RNA-Seq data, normalized using the trimmed mean of M-value (TMM) algorithm. A pre-ranked Gene set enrichment analysis (GSEA) was evaluated using t-statistic to assess enrichment in biological hallmarks. A statistical FDR threshold of 0.05 was considered for significance. We generated a library of 13 patient-derived xenografts (PDX) from untreated primary DDLPS and 3 2D cell lines, which are marked by amplification of MDM2, CDK4, and HMGA2 genes and characterized by a different array of histological dedifferentiation lineages, as well as levels of amplification of the liposarcoma driver genes. Two of these PDXs (LS-NV and LS-MP) and all cell lines were exploited to functionally validate therapeutic targets identified by the RNA-Seq analysis and comparatively evaluate the effectiveness of chemotherapeutics with an indication in soft tissue sarcoma.

Results: Marked transcriptional changes exist among paired components of DDLPS (DD, WD, NF) or WDLPS (WD, NF). GSEA highlighted deregulation in hallmarks. An upregulation of G2M checkpoint and mitotic spindle gene sets was observed in the WD component of DDLPS compared to WDLPS and progressively increases in the DD component. Some targetable genes from the leading edge of these pathways (AURKA, CHK1, and EZH2) as well as the primary oncogenic drivers (MDM2, CDK4, and HMGA2) of liposarcoma were functionally validated with targeted agents in cell lines of DDLPS, showing a variable extent of anti-cancer activity with some differences among the cell models. As a further validation, these druggable targets were tested in vivo and compared with chemotherapeutics administered as single agent. PDX models LS-NV (characterized by myogenic non-rhabdomyoblastic differentiation) and LS-MP (characterized by homologous pleomorphic liposarcoma-like differentiation) were MDM2-amplified and consistently recapitulate the morphology and transcriptomic profile of their paired clinical tumors. Cabazitaxel and ifosfamide resulted in the highest maximum tumor volume inhibition (mTVI%), which was 94-100% and 83-98%, respectively. MDM2 inhibition with milademetan resulted in different effects in the two analyzed models, which were higher in LS-MP (mTVI% 83%) compared to LS-NV (mTVI%35%). Conversely, the AURKA inhibitor alisertib resulted in a higher mTVI% in the LS-NV (93%) compared to the LS-MP (50%) model. Analysis of tumor transcriptome profiles at the end of treatment is ongoing to investigate modifiers of tumor response.

Conclusion: We identified and validated therapeutic targets for DDLPS beyond the oncogenic drivers of this disease. These findings have significant implications for developing combination treatments, which will be tested in new in vivo experiments to prioritize innovative treatment strategies for clinical trials.