Updated exploratory biomarker analysis of cadonilimab plus regorafenib as second-line or later therapy in advanced hepatocellular carcinoma: Identification of Bola3 as a potential predictive biomarker

Background: Cadonilimab, a first-in-class bispecific antibody targeting PD-1 and CTLA-4, in combination with regorafenib, a multikinase inhibitor approved for second-line treatment of hepatocellular carcinoma (HCC), has demonstrated promising antitumor activity and manageable safety in heavily pretreated patients with advanced HCC (aHCC). We previously reported the updated clinical outcomes at the 2025 AACR Annual Meeting (Abstract #CT172, NCT05644379). Here, we present updated findings from an exploratory biomarker analysis focusing on the potential predictive role of BOLA3 in treatment response. Methods: Paired plasma samples were collected before and during treatment for cell-free RNA sequencing (cfRNA-seq) to evaluate transcriptional dynamics. Differential expression analysis was performed using DESeq2 (FDR ≤ 0.10), and Hallmark GSEA was applied to assess mitochondrial and immune-related pathway changes. ELISA quantified soluble PD-1 and CTLA-4 at baseline and early on-treatment, calculating individual values. In a prespecified patient subset, baseline tumor tissues underwent semi-quantitative BOLA3 immunohistochemistry (IHC). Multiplex immunofluorescence (mIF) was performed for BOLA3, TOMM20, PD-L1, CD8alpha, and DAPI, enabling cell-level quantification and comparison of tumor versus peritumoral spatial distribution. Results: ELISA detected stable levels of soluble PD-1 and CTLA-4 pre- and on-treatment, with early individual dynamics observed; however, these levels did not correlate with treatment response. cfRNA-seq revealed decreased BOLA3 in the PR patient and increased BOLA3 in the PD patient after treatment. Pathway analysis indicated PD-associated enrichment of mitochondrial signaling and weaker inflammatory response, whereas PR showed the opposite trend. Baseline BOLA3 IHC aligned with transcriptomic findings: higher BOLA3 expression was observed in patients with disease progression versus disease control, maintaining directional association when analyzed as a continuous variable. Preliminary mIF suggested that responders exhibited lower BOLA3/TOMM20 and higher PD-L1/CD8, while progressors showed the reverse phenotype. Conclusions: Integrated exploratory analyses nominate BOLA3 as a candidate outcome-associated biomarker and support a model in which responders exhibit a mitochondria-low, inflamed-high phenotype. These findings provide early, complementary evidence that BOLA3-related features may inform patient stratification, with further validation underway. Research Sponsor: The Joint Funds of the Natural Science Foundation of Tianjin (No.25JCLZJC00120).