Hypoxia-induced modulation of immunotherapy efficacy in hepatocellular carcinoma

Abstract

Background Hepatocellular carcinoma (HCC) is a leading cause of cancer-related mortality, with most cases diagnosed at an advanced stage, making immunotherapy a key treatment strategy. However, the response rate to PD-L1 inhibitors remains low, necessitating further exploration of resistance mechanisms and predictive biomarkers. Hypoxia is a major contributor to immunotherapy resistance, as HIF-1α upregulates PD-L1 expression and activates genes that help tumor cells adapt to hypoxia, ultimately reducing immunotherapy efficacy. This study integrated bioinformatics, machine learning, and deep learning to identify key hypoxia-associated genes and pathways contributing to PD-L1 expression. A hypoxia risk score model was developed to stratify cases by risk, and a Kolmogorov-Arnold Network (KAN) deep learning model was constructed to predict immunotherapy response. Additionally, an in vitro hypoxia-induced drug-resistant HepG2 cell model was established, and the role of NOXA in apoptosis regulation was examined through flow cytometry and AI-based image analysis. Results and Conclusion 52 HCC-Hypoxia Overlap genes (HHOs) were identified, with 14 PD-L1 regulatory genes and 10 hub genes influencing immunotherapy response. PMAIP1 (NOXA) was significantly associated with immunotherapy response (p < 0.001). A hypoxia risk score model integrating PMAIP1 and 9 hypoxia risk-associated genes demonstrated high predictive accuracy (AUC = 0.815, 0.774, 0.771 for 1-, 2-, and 3-year survival, respectively). The KAN deep learning model incorporating 11 key genes achieved high predictive accuracy (AUC = 0.936 training, 0.7 test). SVM-based integration of hypoxia risk score and KAN model improved prediction performance (AUC = 0.725 test set). Experimental validation demonstrated that hypoxia enhances drug resistance in HepG2 cells, while NOXA knockdown alters apoptosis patterns, potentially modulating treatment response. These findings highlight NOXA as a potential therapeutic target and establish a robust model for predicting immunotherapy response, advancing precision medicine in HCC treatment.