Background Atezolizumab plus bevacizumab (Atez/Bev) has become the first-line therapy for unresectable hepatocellular carcinoma (HCC). However, reliable biomarkers to predict therapeutic response remain lacking. Methods We conducted a high-throughput plasma proteomic screen using the Olink proximity extension assay to identify predictive biomarkers in 78 patients with unresectable HCC treated with Atez/Bev. Validation was performed in an independent cohort of 89 patients. Tumorous expression of the lead candidate, carbonic anhydrase 9 (CA9), was examined using transcriptomic and single-cell analyses. Functional studies were performed in a syngeneic mouse model using Car9-overexpressing hepatoma cells, treated with anti-programmed death-ligand 1/vascular endothelial growth factor antibodies, with or without a CA9 inhibitor. Results High plasma CA9 levels were significantly associated with poor objective response rate, disease control rate, progression-free survival, and overall survival in both discovery and validation cohorts. Multivariable analysis confirmed CA9 as an independent predictor of treatment resistance and poor outcomes. Tumorous CA9 expression correlated strongly with circulating CA9 and was restricted to malignant cells. In vivo, tumors overexpressing Car9 showed resistance to Atez/Bev therapy, characterized by reduced CD8+T cell infiltration, suppressed cytotoxic gene expression, enrichment of M2-like macrophages, and pro-angiogenic signaling. The addition of a CA9 inhibitor reversed resistance and restored antitumor efficacy of Atez/Bev in this model. Conclusions Tumor-derived circulating CA9 is a predictive biomarker and functional driver of resistance to Atez/Bev therapy in HCC. Targeting CA9 may enhance therapeutic response.
Building similarity graph...
Analyzing shared references across papers
Loading...
Yu Sato
Takahiro Kodama
Kazuki Maesaka
Journal for ImmunoTherapy of Cancer
The University of Osaka
Hokkaido University
University of Ulsan
Building similarity graph...
Analyzing shared references across papers
Loading...
Sato et al. (Wed,) studied this question.
www.synapsesocial.com/papers/69d9e67a78050d08c1b76de3 — DOI: https://doi.org/10.1136/jitc-2025-013384