TY - JOUR AU - Gao, Xingxing AU - Huang, Hechen AU - Wang, Yubo AU - Pan, Caixu AU - Yin, Shengyong AU - Zhou, Lin AU - Zheng, Shusen PY - 2021 M3 - Original Research TI - Tumor Immune Microenvironment Characterization in Hepatocellular Carcinoma Identifies Four Prognostic and Immunotherapeutically Relevant Subclasses JO - Frontiers in Oncology UR - https://www.frontiersin.org/articles/10.3389/fonc.2020.610513 VL - 10 SN - 2234-943X N2 - PurposeThe tumor microenvironment (TME) plays a critical role in the pathogenesis of hepatocellular carcinoma (HCC). However, underlying compositions and functions that drive the establishment and maintenance of the TME classifications are less-well understood.MethodsA total of 766 HCC patients from three public cohorts were clustered into four immune-related subclasses based on 13 TME signatures (11 immune-related cells and 2 immune-related pathways) calculated by MCP-counter. After analyzing the landscapes of functional annotation, methylation, somatic mutation, and clinical characteristics, we built a TME-based Support Vector Machine of 365 patients (discovery phase) and 401 patients (validation phase). We applied this SVM model on another two independent cohorts of patients who received sorafenib/pembrolizumab treatment.ResultsAbout 33% of patients displayed an immune desert pattern. The other subclasses were different in abundance of tumor infiltrating cells. The Immunogenic subclass (17%) associated with the best prognosis presented a massive T cell infiltration and an activation of immune checkpoint pathway. The 13 TME signatures showed a good potential to predict the TME classification (average AUC = 88%). Molecular characteristics of immunohistochemistry from Zhejiang cohort supported our SVM classification. The optimum response to pembrolizumab (78%) and sorafenib (81%) was observed in patients belonging to the Immunogenic subclass.ConclusionsThe HCC patients from distinct immune subclass showed significant differences in clinical prognosis and response to personalized treatment. Based on tumor transcriptome data, our workflow can help to predict the clinical outcomes and to find appropriate treatment strategies for HCC patients. ER -