Abstract PR009: Using omics technologies to understand hepatoblastoma heterogeneity and identify tumor cell types and their microenvironment for precision medicine
Elise Lelou, Abhishek Murti, Saphia Nguyen, Matthew Choi, Cindy Ament, Nikita Sajai, Phoebe N. Miller, Amar Nijagal, Soo-Jin Cho, E. Alejandro Sweet-Cordero, Bruce WangAbstract
Introduction:
Hepatoblastoma (HB) is the most common primary pediatric liver cancer. Among childhood tumors, it has one of the highest mortality rates with 20% cases resulting in death or liver transplantation. Identifying high-risk tumors has been challenged by tumor heterogeneity. Current clinical high-risk stratification relies on imaging and histology, with serum α-fetoprotein (AFP) as the only molecular biomarker. Therefore, it is necessary to improve molecular characterization to enable more accurate patient risk stratification. We previously used single-cell RNA sequencing (scRNA-seq) to identify 5 tumor cell populations shared across patients. We now expand this analysis to include additional freshly resected and frozen tumors, using both single-cell and single-nuclei RNA sequencing (snRNA-seq). In parallel, we characterized 23 hepatoblastomas using spatial transcriptomics to decipher the microenvironment associated with each tumor cell population.
Methods:
Samples from 21 patients were included among three platforms. scRNA-seq was performed using microwell capture on freshly dissociated tumors and snRNA-seq was performed using droplet capture on banked frozen tumors. Data was analyzed using Scanpy. Tissue microarrays of FFPE-Hepatoblastoma tissues from 17 patients were generated for Xenium spatial transcriptomics using a custom 480-gene panel based on our sc/snRNA-seq dataset and analyzed with Scanpy and Squidpy.
Results:
We confirmed the reproducibility of tumor cell populations across techniques using matched scRNA- and snRNA-seq patient samples. It highlights the utility of snRNA-seq for characterizing banked HB samples. Our expanded sc/snRNA-seq datasets validated the previously identified 5 tumor cell populations, and identified 4 new populations (Hepatoblast-III, Wnt-I, Wnt-II, and Neuroendocrine-II (CHGB+)). Surprisingly, neuroendocrine tumor cell types, rarely reported in the literature for HB, were identified in 5 of 21 patients, with Neuroendocrine-I (CHGA/B+) being specific to the high-risk HB patients 8 and 18. Spatial transcriptomics revealed that distinct tumor cell populations occupied discrete spatial regions within the tumor. Using neighborhood analysis, we found that each tumor cell type had a distinct microenvironment. Hepatoblast tumor microenvironment was enriched for vascular endothelial cells and tumor-associated fibroblasts at the tumor-stroma interface, with ligand-receptor analysis identifying HGF-MET and EGF-family signaling. In contrast, Neuroendocrine-I cell nests, which may be associated with high risk, lack other cell types within their microenvironment and exhibit somatostatin autocrine signaling.
Conclusion:
By combining complementary omics techniques, we established a molecular and spatial single-cell resolution atlas of hepatoblastoma. We identified clinically relevant shared tumor cell types, each with distinct microenvironments. These results could improve risk stratification and identify tumor cell type-specific microenvironmental interactions that represent therapeutic target candidates.
Citation Format:
Elise Lelou, Abhishek Murti, Saphia Nguyen, Matthew Choi, Cindy Ament, Nikita Sajai, Phoebe N. Miller, Amar Nijagal, Soo-Jin Cho, E. Alejandro Sweet-Cordero, Bruce Wang. Using omics technologies to understand hepatoblastoma heterogeneity and identify tumor cell types and their microenvironment for precision medicine [abstract]. In: Proceedings of the AACR Special Conference in Cancer Research: Bridging Discovery and Clinical Impact in Pediatric Cancer; 2026 Sep 22-25; Philadelphia, PA. Philadelphia (PA): AACR; Cancer Res 2026;86(18_Suppl_1):Abstract nr PR009.