Longitudinal per-lesion in vivo imaging reveals allele-dependent resistance evolution in EGFR-mutant lung cancer

Acquired resistance to targeted therapies remains an inevitable outcome in EGFR-mutant non-small cell lung cancer, yet the spatiotemporal dynamics through which resistant clones emerge and evolve in vivo remain incompletely understood. In particular, how distinct oncogenic EGFR alleles shape evolutionary trajectories under therapeutic pressure within native tumor microenvironments remains unclear. Here, we establish a longitudinal in vivo imaging framework to resolve tumor evolution at single-lesion resolution in genetically engineered mouse models (GEMMs) harboring three clinically relevant EGFR mutations: exon 19 deletion (EGFRD19), L858R (EGFRLR), and L858R/T790M (EGFRLT). Using high-resolution micro-computed tomography, three-dimensional reconstruction, and per-lesion volumetric tracking, we quantitatively map tumor growth dynamics, therapeutic response, and resistance emergence over time in individual lesions within the same animal. We find that EGFR alleles impose distinct evolutionary trajectories. EGFRLT -driven tumors exhibit shorter latency and early emergence of lesions with intrinsic resistance to osimertinib. In contrast, EGFRD19 and EGFRLR tumors show slower growth kinetics, more homogeneous initial responses, and delayed acquisition of resistance during prolonged treatment. Importantly, longitudinal per-lesion imaging reveals marked spatial heterogeneity across all genotypes. Within the same lung microenvironment, individual lesions undergo complete regression, sustained response, or progressive growth, reflecting parallel and spatially distinct evolutionary trajectories. These divergent behaviors emerge despite shared systemic therapy and identical host environment, underscoring lesion-intrinsic and genotype-dependent constraints on evolution. Together, these findings identify oncogenic EGFR genotype as a key determinant of the temporal and spatial architecture of resistance evolution under targeted therapy. More broadly, we provide a quantitative framework to resolve tumor evolution in vivo at lesion-level resolution, applicable to dissecting spatiotemporal dynamics of tumor growth and therapeutic response across oncogene-driven cancers.