Spatiotemporal profiling defines persistence and resistance dynamics during targeted treatment of BRAF-mutant melanoma

Over half of BRAF-mutant melanoma patients with initial response to targeted therapy will recur with resistant disease. It is thought that resistance can arise from the ability of cells to enter and exit a slow cycling persister state, evade treatment with relative dormancy, and repopulate the tumor when reactivated. However, the expression states of persister and resistant cells, their spatial relationships in the tumor environment, and the evolutionary progression among these populations are not well understood. Using patient derived xenograft (PDX) models, we track melanoma evolution longitudinally from pre-treatment through maximum treatment response and tumor regrowth. We perform spatial transcriptomics (ST) to 1. define expression signatures within treatment-sensitive, persister, and re-emergent populations, 2. uncover fluctuating metabolic priorities as a hallmark of tumor evolution both across treatment time and geographically across tissue sections, and 3. identify specific genes and pathways within persister and resistant lineages as candidate resistance mechanisms, suggesting potential therapeutic targets. Our novel computational pipeline integrates ST data with deep learning imaging features derived from accompanying histopathological slides. We identify phenotypic states that correlate with the persister cell state, exploiting the juxtaposition of transcriptomic and histological features toward the goal of identifying clonal populations using imaging data alone. In summary, we provide insight into the dynamics of shifting expression state and lineage selection in melanoma during multiple stages of treatment with novel spatiotemporal resolution, defining an evolutionary roadmap to acquired resistance.