Soft tissue sarcoma (STS) are rare mesenchymal-originated tumors with more than 50 different histologies identified. Not every histology in STS responds to immunotherapy and immunologic predictive markers are lacking. The purpose of this study is to establish an immune classification of STS by analysis of the transcriptome. This was performed by using a deconvolution method that allowed us to quantify 8 immune populations and endothelial cells. As a secondary objective, we searched for a surrogate biomarker that could be assessable in the clinic. We analyzed transcriptomic data of four publicly available datasets, accounting for 608 complex genomic STS, including leiomyosarcoma (LMS, 35.4%), dedifferentiated liposarcoma (DDLPS, 33.9%) and undifferentiated pleomorphic sarcoma (UPS, 30.8%). By using the MCP-Counter deconvolution method, we characterized the tumor microenvironment (TME) of these tumors and established a robust immune classification that is consistent through various cohorts. We classified the patients into 5 Sarcoma Immune Classes (SIC), labeled as A1, A2, B, C1 and C2. The A1 and A2 groups are associated with very low to low immune infiltrates. Conversely, SIC C1 and C2 tumors are characterized by strong to very strong expression of signatures associated to all immune cells. SIC B tumors are characterized by a high expression of endothelial cell signature, an intermediate presence of neutrophils, and a rather low infiltration by other immune cell types. Regarding functional orientation of the TME, gene signatures associated with immune cells chemotaxis activation and survival, expression of major histocompatibility complex class I, and regulatory T-cells are highly expressed in SIC C1 and C2, modestly expressed in B and A2, and very lowly expressed in A1. Interestingly, immune checkpoint genes exhibited strong expression differences between SICs. SIC C2 had a strong expression of PD-1, PD-L2, CTLA-4 and TIM-3 genes. We also found that the lymphoid structure-associated B cell chemoattractant chemokine CXCL13 is remarkably highly expressed in C2 class. CXCL13 is associated with the presence of tertiary lymphoid structures (TLS). Although all histologies are distributed in each SIC group, LMS are more commonly found in the immune low SIC A1 and A2 groups, and we also extended our analysis to other histologies such as synovial sarcoma or gastrointestinal stromal tumors. Our classification is associated with clinical outcome, and SIC group C (C1/C2) has the longest overall survival, as compared to SIC A group (A1/A2) (p = 0.015). We then validated SIC classification using STS FFPE samples (n=32). SIC classification by RNA expression was correlated with quantitative immunohistochemistry (IHC) of CD3 (T-cells), DC-Lamp (activated dendritic cells), CD20 (B cells), CD8 (CD8+ T-cells), and CD34 (endothelial cells). Densities of CD3 (p=0.0033), CD8 (p=0.004) and CD20 (p=0.00043) were significantly higher in SIC C tumors. Tumor SIC B groups have a higher, albeit nonsignificant (p=0.6) expression of CD34+ endothelial cells. In the validation cohort, SIC group C is associated with a better prognosis (p=0.053). We further investigated whether TLS were a marker of the immune-high SIC C group. Not surprisingly, TLS-positive tumors were found only in SIC C2 (100%, 9/9) and C1 (50%, 3/6) groups of the validation cohort. We expanded the validation cohort (n = 93) to analyze the correlation between TLS and CD3+/CD8+/CD20+ tumor-infiltrating lymphocytes. We found that the presences of each type of TILs all are significantly associated with the presence of TLS (CD3, CD8, CD20, all p values < 0.0001). This validation finding suggests that TLS may be a good surrogate marker for identification of SIC C groups and could be validated in the future to correlate with the efficacy of immunotherapy in STS.

Citation Format: Wei-Wu Tom Chen, Florent Petitprez, Cheng-Ming Sun, Laetitia Lacroix, Aurélien de Reyniès, Antoine Italiano, Maud Toulmonde, Carlo Lucchesi, Yec'han Laizet, Catherine Sautès-Fridman, Wolf Herve Fridman. Immune-based classification of soft-tissue sarcoma is associated with clinical outcome and unveils tertiary lymphoid structures as surrogate biomarker for the clinic [abstract]. In: Proceedings of the Fourth CRI-CIMT-EATI-AACR International Cancer Immunotherapy Conference: Translating Science into Survival; Sept 30-Oct 3, 2018; New York, NY. Philadelphia (PA): AACR; Cancer Immunol Res 2019;7(2 Suppl):Abstract nr PR03.