Serial Spatial Transcriptomics Reveal Divergent Routes to Therapy Resistance in Metastatic Breast Cancer

Metastatic solid tumors persist by evolving therapeutic resistance through complex, heterogeneous adaptive strategies that challenge standard precision medicine approaches. Current clinical decision making relies on bulk biomarkers, failing to resolve the spatial architecture and cellular contexts in which resistance mechanisms emerge. We present a patient-centric spatial framework, profiling 345,207 cells from four metastatic breast cancer patients across ten biopsies spanning personalized treatment courses of up to 3.5 years. By integrating probabilistic topic modeling with spatial deep learning, we observe fundamental principles of metastatic survival: pathway independence, microenvironment remodeling, and compensatory signaling. While these principles are universal, the underlying mechanisms are distinct: pathway independence manifested variously as the extinction of luminal identity, constitutive ESR1 activation, or spatial partitioning into drug-refractory invasive nests. Immune sanctuary was achieved through either genetic evasion mechanisms or physical exclusion via expanded fibroblast barriers. Compensatory transcriptional programs were engaged through rewired ligand-receptor networks and alternative survival pathway activation. These findings establish spatial profiling as a means to identify which mechanisms underlie each resistance principle in individual patients, enabling rational design of multi-axis combination therapies and earlier therapeutic decisions.