Abstract Introduction: Liver cancer, predominantly hepatocellular carcinoma (HCC) in adults, has become the fastest-growing cause of cancer death in the United States. Most cases of HCC are diagnosed at the metastatic stage and have a 5-year survival rate below 3%. Liver cancer in children, most commonly hepatoblastoma (HB), has a better prognosis than in adults, but the 5-year survival falls from 70% to 30% if the tumor has disseminated. Without basic research into the biology of liver cancer metastasis, cure rates for this disease are unlikely to improve. Preclinical mouse models have played a central role in linking basic and translational cancer research. However, current mouse models of liver cancer have mostly addressed the early stages of the disease and have rarely focused on its metastatic forms. Accurate model systems of liver cancer metastasis are urgently needed to elucidate the associated cellular and molecular mechanisms. Methods: Metastasis-initiating cells often hijack normal stem cell pathways to gain survival and growth advantages. Recent advances in 3D stem cell spheroid culture have provided a means to rapidly select and propagate cancer cells with stemness potential for in-depth biologic analysis. We have previously generated two liver cancer genetic mouse models (GMMs) of HCC and HB, both of which developed frequent primary tumors in the liver but only rare metastases. Our laboratory recently developed a robust protocol for growing liver cancer spheroids (LCSs) from primary tumors in the two GMM using 3D stem cell culture. To evaluate the in vivo tumorigenicity of the LCSs, we developed a novel ultrasound-guided, noninvasive intrahepatic injection method for transplanting cells into the liver with high accuracy and minimal animal manipulation. Results: We found that a subset of LCSs from each model exhibited potent tumorigenicity, driving widespread metastases in vivo upon orthotopic transplant. Gene set enrichment analysis demonstrated substantial enrichment of a number of metastatic and recurrent human HCC gene signatures in the LCS-driven HCC tumors and the enrichment of a human HB gene signature in the LCS-driven HB tumors. A large RNA-seq transcriptomic analysis revealed a significant enrichment of inflammatory and extracellular processes in the metastatic LCSs and tumors they generated compared to their nonmetastatic counterparts in both models, suggesting a critical role of tumor microenvironment (TME) in liver cancer metastasis. Conclusion: Our LCS-based approach to model liver cancer metastasis represents a unique strategy to model liver cancer metastasis by using primary tumors from non- or low-metastatic GMMs, an approach that can potentially be extended to other cancer types, considering the rapid advances in cancer spheroid studies across a wide range of solid malignant tumors. It also permits rapid mechanistic studies of liver cancer metastasis in the future by allowing gene manipulation in vitro within a uniform population of tumor-derived, metastasis-initiating LCSs, with subsequent tracking of metastasis development in transplantation models to be built in disease-relevant TMEs. Citation Format: Liyuan Li, Maoxiang Qian, David Finkelstein, Melissa Johnson, Dolores H. López-Terrada, Liqin Zhu. Model liver cancer metastasis using 3D spheroids derived from primary tumors in liver cancer genetic mouse models [abstract]. In: Proceedings of the AACR Special Conference: Advances in Modeling Cancer in Mice: Technology, Biology, and Beyond; 2017 Sep 24-27; Orlando, Florida. Philadelphia (PA): AACR; Cancer Res 2018;78(10 Suppl):Abstract nr B40.