Machine learning-based integration reveals immunological heterogeneity and the clinical potential of T cell receptor (TCR) gene pattern in hepatocellular carcinoma

Abstract

The T Cell Receptor (TCR) significantly contributes to tumor immunity, whereas the intricate interplay with the Hepatocellular Carcinoma (HCC) microenvironment and clinical significance remains largely unexplored. Here, we aimed to examine the function of TCR signaling in tumor immunity and its clinical significance in HCC. Our objective was to employ TCR signaling genes and a machine learning-based integrative methodology to construct a prognostic prediction system termed the TCR score. Herein, we revealed that the TCR score serves as an independent risk factor for overall survival in HCC patients, demonstrating stable and robust performance. The accuracy of the TCR score significantly exceeds that of traditional clinical variables and published signatures. Additionally, the immune infiltration was abundant in patients with low TCR scores. Single-cell cohort analysis further demonstrates that patients with low TCR scores possess an immune-active tumor microenvironment (TME), with T/NK cells enhancing interactions with myeloid cells through signaling networks such as MIF, MK, and SPP1. In response to these changes in the TME, patients with high TCR scores exhibit poorer outcomes and shorter survival in immunotherapy cohorts. In vitro experiments demonstrated that the key TCR signaling biomarker SOS1 knockdown significantly suppresses the HCC cells’ capability to proliferate, invade, and migrate while enhancing tumor cell apoptosis. The TCR score could function as a robust and potential tool to predict immune activity and improve clinical outcomes for HCC patients.