Background: Hepatocyte-like cells (iHEPs) generated by transcription factor-mediated direct reprogramming of somatic cells have been studied as potential cell sources for the development of novel therapies targeting liver diseases. The mechanisms involved in direct reprogramming, stability after long-term in vitro expansion, and safety profile of reprogrammed cells in different experimental models, however, still require further investigation.Methods: iHEPs were generated by forced expression of Foxa2/Hnf4a in mouse mesenchymal stromal cells and characterized their phenotype stability by in vitro and in vivo analyses.Results: The iHEPs expressed mixed hepatocyte and liver progenitor cell markers, were highly proliferative, and presented metabolic activities in functional assays. A progressive loss of hepatic phenotype, however, was observed after several passages, leading to an increase in alpha-SMA + fibroblast-like cells, which could be distinguished and sorted from iHEPs by differential mitochondrial content. The resulting purified iHEPs proliferated, maintained liver progenitor cell markers, and, upon stimulation with lineage maturation media, increased expression of either biliary or hepatocyte markers. In vivo functionality was assessed in independent pre-clinical mouse models. Minimal engraftment was observed following transplantation in mice with acute acetaminophen-induced liver injury. In contrast, upon transplantation in a transgenic mouse model presenting host hepatocyte senescence, widespread engraftment and uncontrolled proliferation of iHEPs was observed, forming islands of epithelial-like cells, adipocytelike cells, or cells presenting both morphologies. Conclusion: The results have significant implications for cell reprogramming, suggesting that iHEPs generated by Foxa2/Hnf4a expression have an unstable phenotype and depend on transgene expression for maintenance of hepatocyte-like characteristics, showing a tendency to return to the mesenchymal phenotype of origin and a compromised safety profile.