(P 131) HETEROGENEITY OF THE IMMUNE MICROENVIRONMENT AND IMMUNOTHERAPY RESPONSE IN A DEDIFFERENTIATED LIPOSARCOMA MOUSE MODEL

Objective: Dedifferentiated liposarcoma (DDLPS) continues to present a clinical challenge. While surgical resection is the most effective treatment option, patients have high rates of recurrence and distant metastasis. Patients with metastatic disease have a poor overall survival due to the low response rates to systemic chemotherapy. While immunotherapy has shown tremendous efficacy in some solid tumors, it has failed to benefit patients with DDLPS. An early trial with programmed death 1 (PD-1) immune checkpoint blockade (ICB) showed transient responses in about 20% of patients, followed by rapid progression. Taken together, these data suggest that there are tumor elements contributing to ICB response in human DDLPS, but our inability to achieve sustained immune responses in DDLPS underscores a critical need for understanding the mechanisms of treatment response in order to improve overall response rates. Here we use a novel preclinical genetically engineered mouse model of liposarcoma to investigate the immune microenvironment of DDLPS and syngeneic liposarcoma cell lines to characterize the immune tumor microenvironment (TME) following ICB treatment.

Methods: We have established a fully penetrant, genetically engineered mouse model of liposarcoma which combines Pten and Trp53 deletions in targeted adipocytes in C57BL/6 mice, hereby designated ACPP. We performed H&E staining and bulk RNA sequencing on 10 mouse and 9 human DDLPS with matched species-specific normal fat. Histology was verified by a sarcoma pathologist. Gene set enrichment analysis (GSEA) for the Molecular Signatures Database (MSigDb) Hallmark pathways of the differentially expressed genes (DEGs) in both mouse and human DDLPS were compared. The immune TME was analyzed by multicolor immunohistochemistry (mIHC). From the ACPP model, we developed two syngeneic cell lines (N1011 and N1018). Cell lines were injected into the flanks of WT B6 mice and mice were randomized to anti-PD1, anti-LAG3, or isotype control (n=4/group/cell line). Tumor immune populations were analyzed via flow cytometry.

Results: Histology and bulk RNA sequencing confirmed APCC tumors had characteristics of human DDLPS. GSEA revealed upregulated gene sets in mouse tumors are transcriptionally similar to human with statistical significance (NES 2.5, padj 2x10-50). Similar to human DDLPS, we identified both high (n=5) and low (n=5) lymphocyte infiltrating tumors, defined by greater or less than 20% CD3+ cells by mIHC, respectively. CD3hi tumors had a significantly greater % of CD8+ lymphocytes compared to CD3lo (7% v. < 1%; padj < 0.01). However, there was no significant difference in PD1 expression between the two groups. CD3hi tumors also had a significantly greater % of Tregs (FOXP3+) compared to CD3lo (4% v. < 1%; padj < 0.01). Macrophage (F4/80+) and B-cell (CD19+) levels were not significantly different between groups. N1011 and N1018 CLDTs showed different responses to ICB treatment. N1011 showed no growth delay with either anti-PD1 or anti-LAG3. Strikingly, N1018 tumors treated with anti-PD1 and anti-LAG3 showed a delay in the outgrowth of the tumor when compared to controls. Flow cytometry analysis of CLDTs showed similar levels of F4/80+ macrophages between cell lines. However, N1018 anti-PD1 treated tumors showed an increase in the percentage of CD8+ T cells (padj < 0.01) opposed to N1011 which showed no post-treatment differences.

Conclusion: DDLPS generated from the ACPP model mimicked the biological behavior of human DDLPS with similar histology and variable immune cell subsets in the TME. Two syngeneic DDLPS cells lines had different growth rates and immune changes when treated with ICB. Overall, this model allows the unique opportunity to study the TME and mechanisms of ICB efficacy in DDLPS.