Among the 4 scientific objectives of THRIVE, we first aimed to identify molecular features linked to cancer risk and early development (1st objective). In this regard, we identified that 17% of children with HB show mosaic genetic alterations in the 11p15.5 locus in their liver. These alterations occurred before common cancer-related mutations (e.g. CTNNB1 mutations) and may represent a pre-cancerous stage. Spatial transcriptomics and single-nucleus RNA sequencing revealed that these regions exhibit a distinct tumour microenvironment. Overall, our findings suggest early 11p15.5 alterations may help identify high-risk patients.
To build a detailed cellular map of liver cancer (2nd objective), we used single-cell RNA sequencing and defined immune cell signatures, to dissect the immune cells types in HCC, uncovering intratumoral heterogeneity. In paediatric HB, we identified liver progenitor and immune cold tumour subtypes linked to poorer outcomes. Using single-cell multiomics, we showed that certain HB cell clones are resistant to chemotherapy, linking clonal evolution to response.
To identify molecular markers of treatment response (3rd objective), by integrating single-cell and bulk RNA sequencing data, we uncovered two mechanisms driving response patterns to atezolizumab + bevacizumab, the current standard of care treatment in advanced HCC: one driven by immune activity and the other by angiogenesis. We also uncovered two resistance mechanisms, involving either immunosuppressive myeloid cells or TGF-/Notch activation. Further, we developed an AI method to predict HCC molecular subtypes from standard H&E tumour slides, laying the foundation for future models predicting response to immunotherapies. For HB, we identified molecular features linked to chemotherapy resistance.
To discover or repurpose affordable therapies (4th objective), we developed innovative pre-clinical models: (a) a mouse model mimicking Metabolic and Alcohol-Associated Liver Disease (Met-ALD); (b) HCC and HB patient-derived organoids (PDOs); and (c) a PDO integrating bacteria+immune cells to better simulate the tumour environment. Also, in collaboration with CRUK, we helped develop 27 genetically engineered, immunocompetent mouse models recapitulating 4 HCC molecular classes. Drug screening in these models revealed cladribine as a promising candidate enhancing therapy efficacy. Finally, in a MASH-HCC mouse model, we showed NRP1 blockade enhances immunotherapy efficacy.
