Mitochondrial niches of residual disease in EGFR-mutant NSCLC: immune-constrained persistence and therapeutic interception.

Therapeutic failure in EGFR-mutant non-small cell lung cancer (NSCLC) is commonly viewed as a consequence of acquired resistance, but this perspective may overlook an earlier and potentially more tractable phase marked by the survival of residual cells after effective EGFR inhibition. In this Review, we define the mitochondrial niche as a temporally bounded, hypothesis-generating model of early residual disease in which tumor-cell mitochondrial stress adaptation is considered alongside fibroblast-matrix support and incomplete immune clearance. The conceptual contribution of this framework is to place independently studied tumor-intrinsic, stromal, and immune processes on a common temporal axis and to generate testable questions about their potential interaction; whether these processes coexist and function as a coordinated niche in EGFR-mutant residual disease has not yet been directly demonstrated. We distinguish DTPs and cycling persisters as experimentally defined cell states, MRD as a clinical detection state, and acquired resistance as overt regrowth under continued therapy. In selected preclinical models, EGFR-targeted therapy is associated with survival states involving mitochondrial fuel use, redox control, organelle-quality surveillance, or apoptotic restraint. In parallel, selected cancer-associated fibroblast states may buffer therapy-induced stress, whereas treatment-associated inflammatory remodeling often fails to generate productive antitumor immunity. By treating residual disease as a distinct, low-burden and testable state rather than as direct evidence of a single persister ecology, this framework prioritizes candidate biomarker classes and mechanism-based interventions for prospective validation before stable resistance is established. Candidate readouts include driver-ctDNA kinetics, residual-lesion status, stromal-state markers, and selected immune features, all requiring prospective validation.