(P 139) T-CELL PROFILING IN HIGH GRADE SARCOMAS

Objective: Memory T cells play crucial roles in anti-tumor immunity. Resident memory T cells (TRM) have been associated with durable immune responses and improved patient survival across multiple tumor types, in particular interferon gamma secreting TRM (TRM-IFNG). Early clinical trial data has shown promising activity of programmed death 1 (PD-1) checkpoint inhibitor therapy in select histologic subtypes of advanced soft tissue sarcoma (STS). Since the anti-tumor response to anti-PD-1 is dependent on cytotoxic T cells, it is imperative that we understand the T-cell phenotypes in STS. We seek to explore T-cell profiles in high grade sarcomas and examine the ability of T-cell signatures to predict disease-specific survival (DSS) and recurrence-free survival (RFS) in patients.

Methods: Paired single cell gene expression and chromatin accessibility profiles from dedifferentiated liposarcoma (DDLPS) was performed utilizing 10x Genomics multiome platform. Data integration was done using Weighted-Nearest Neighbor (WNN) algorithm implemented in Seurat R package. SingleR was used to perform cell-type annotation in order to define distinct stromal and immune compartments. Moreover, using 10x Genomics Chromium Single Cell Immune Profiling, parallel single cell gene expression and TCR sequencing was done on T-cell populations from undifferentiated pleomorphic sarcoma (UPS). CD8+ and CD4+ T-cell populations were identified with enrichment of tissue residency (CD69, FOS, RGS1), effector (IFNG, TNF, GZMA, PRF1) and exhaustion (CTLA-4, PDCD1, TOX, LAG3) markers. Resident and circulating T cell gene signatures were used to compute sample-specific enrichment scores for TCGA sarcoma samples (DDLPS, UPS, and soft tissue leiomyosarcoma (STLMS)) using the GSVA function in R. Kaplan-Meir plots were generated separately for each histology and T-cell signature, splitting patients with positive versus negative GSVA enrichment scores for DSS and RFS. Multivariate analysis was performed using Cox regression using the different gene signature enrichment scores and other relevant clinical covariates.

Results: UPS had 3 T-cell clusters with features of resident memory (CD69, FOS and RGS1), of which one resembled the TRM-IFNG cluster with expression of effector markers IFNG, GZMA and PRF1G, and low levels of exhaustion markers TOX, but high level of PDCD1. This contrasts with DDLPS which had 2 TRM clusters which both had markers of exhaustion with a strikingly high expression of TOX, even with expression of chemokines CCL4, CCL5, and cytotoxic markers GZMA and PRF1G. Upon defining unique TCR clonotypes in UPS, we found that tumor had higher expansion of certain clonotypes with overall lower diversity compared to the PBMC repertoire. Additionally, we identified matched clonotypes in PBMC that were expanded in tumor. When analyzing the corresponding gene expression profiles of the expanded clonotypes in UPS, we discovered that effector cytotoxic CD8+ T-cells were the most expanded T-cell phenotype. Kaplan-Meir analysis of the TCGA sarcoma dataset showed that the circulating CD8+ T cell signature is significantly associated with longer RFS (p-value, 0.03 and 0.014) in patients with STLMS and UPS, respectively. Additionally, in patients with UPS, the CD8+ stem cell memory T cell and TRM-IFNG signatures were prognostic for improved DSS (p-value, 0.052 and 0.09, respectively).

Conclusion: While TRM populations, including TRM-IFNG, were found in UPS, DDLPS had more exhausted TRM phenotypes. In line with these findings, the TRM-IFNG signature is more predictive of improved DSS in UPS but not in DDLPS. Further investigation needs to be done to identify which T-cell subsets could be better targets for immunotherapy development in DDLPS.