Integrated single cell and bulk sequencing analysis identifies tumor reactive CXCR6(+) CD8 T cells as a predictor of immune infiltration and immunotherapy outcomes in hepatocellular carcinoma

BACKGROUND: Various immune cell types in the tumor microenvironment (TME) of hepatocellular carcinoma (HCC) have been identified as important parameters associated with prognosis and responsiveness to immunotherapy. However, how various factors influence immune cell infiltration remains incompletely understood. Hence, we investigated the single cell multi-omics landscape of immune infiltration in HCC, particularly key gene and cell subsets that influence immune infiltration, thus potentially linking the immunotherapy response and immune cell infiltration. METHODS: We grouped patients with HCC according to immune cell infiltration scores calculated by single sample gene set enrichment analysis (ssGSEA). Differential expression analysis, functional enrichment, clinical trait association, gene mutation analysis, tumor immune dysfunction and exclusion (TIDE) and prognostic model construction were used to investigate the immune infiltration landscape through multi-omics. Stepwise regression was further used to identify key genes regulating immune infiltration. Single cell analysis was performed to explore expression patterns of candidate genes and investigate associated cellular populations. Correlation analysis, ROC analysis, Immunotherapy cohorts were used to explore and confirm the role of key gene and cellular population in predicting immune infiltration state and immunotherapy response. Immunohistochemistry and multiplexed fluorescence staining were used to further validated our results. RESULTS: Patients with HCC were clustered into high and low immune infiltration groups. Mutations of CTNNB1 and TTN were significantly associated with immune infiltration and altered enrichment of cell populations in the TME. TIDE analysis demonstrated that T cell dysfunction and the T cell exclusion score were elevated in the high and low infiltration groups, respectively. Six risk genes and five risk immune cell types were identified and used to construct risk scores and a nomogram model. CXCR6 and LTA, identified by stepwise regression, were highly associated with immune infiltration. Single cell analysis revealed that LTA was expressed primarily in tumor infiltrating T lymphocytes and partial B lymphocytes, whereas CXCR6 was enriched predominantly in T and NK cells. Notably, CXCR6(+) CD8 T cells were characterized as tumor enriched cells that may be potential predictors of high immune infiltration and the immune-checkpoint blockade response, and may serve as therapeutic targets. CONCLUSION: We constructed a comprehensive single cell and multi-omics landscape of immune infiltration in HCC, and delineated key genes and cellular populations regulating immune infiltration and immunotherapy response, thus providing insights into the mechanisms of immune infiltration and future therapeutic control.