Abstract
Immune checkpoint inhibitors (ICIs) are utilized to treat hepatocellular carcinomas (HCC) without validated biomarkers. We sought to identify genomic signatures and immune patterns predicting treatment efficacy in 1,306 HCC patients to evaluate for potential predictive biomarkers associated with immunotherapy outcome. Programmed death ligand 1 (PD-L1) expression was high in 6.1% and low in 11.6%; deficient mismatch repair/microsatellite instability-high (dMMR/MSI-H) was detected in only 0.2%, and high tumor mutational burden (TMB-H) in 5.1% of the patients. Immuno-oncologic (IO) markers (CD80, HAVCR2, PDCD1, IFNG, CTLA4, IDO1, CD86, LAG3, PDCD1LG2) correlated with PD-L1 expression (q < 0.05). Infiltration of B cells, M1 macrophages, CD8 + T cells, and T(regs) also increased with PD-L1 expression (q < 0.05). CTNNB1 mutations occurred more in PD-L1 negative (35%) than in PD-L1 high (21%) patients (p = 0.03). PD-L1 expression did not affect ICI treatment duration, but CTNNB1-high levels were associated with a shorter duration of treatment in PD-L1 negative tumors (p = 0.04). However, in IO non-exposed tumors, CTNNB1-high expression was associated with improved overall survival (HR = 0.97, 95% CI 0.96-0.98, p < 0.00001). In conclusion, our analysis indicates that expression of PD-L1, CTNNB1, and other molecular markers most associated with IO response in other gastrointestinal malignancies are not directly indicative of such responses in HCC.