Abstract
BACKGROUND: Brain metastases from breast cancer (BCBM) are highly fatal malignancies with limited treatment options. Developing preclinical models that accurately replicate metastatic phenotypes is essential for advancing next-generation therapies. However, existing BCBM models often fail to capture patient heterogeneity or clinically relevant metastatic biology, limiting their translational potential. MATERIAL AND METHODS: To address this gap, we established a cohort of clinically relevant BCBM models that reflect the heterogeneity and complexity of the disease. We collected 24 molecularly and clinically diverse BCBM samples, dissociated them into single cells, and intracranially injected them into CD1 nude mice. Half of these samples exhibited tumorigenicity, and the resulting xenografts were harvested, dissociated into single-cell suspensions, and cultured under conditions optimized for cancer stem cells (CSCs), potentially enriching for metastasis-initiating cells (MICs). RESULTS: MICs displayed key stem-like properties, including the expression of stem and breast cancer markers, self-renewal capacity, and tumorigenic potential in orthotopic models. Upon intracardiac or intranipple injection, MICs successfully colonized the brain and other organs, primarily the bones, recapitulating the metastatic cascade observed in patients. MIC-driven brain metastases closely mirrored patient tumor histology and cytoarchitecture. CONCLUSION: Brain tropism appeared to be mediated by enhanced MIC adhesion to brain endothelium and interactions with the brain microenvironment, which promoted MIC survival and outgrowth. To assess the fidelity of these models, we performed whole genome and RNA sequencing on matched patient samples, MICs, and MIC-driven brain metastases, revealing a strong similarity to the corresponding patient tumors. Notably, MICs adapted their transcriptomic profiles according to their specific metastatic niche—brain, bone, or mammary fat pad. Additionally, MICs demonstrated high chemoresistance, underscoring their potential as a reliable model for identifying novel therapeutic target