Ke Qi, Suoqin Jin, Han Ma, Tengfei Wang, Yijun Lou, Qing Nie, Xiufen Zou
The spatial patterns of immune cells within the tumor microenvironment hold profound prognostic implications, yet a mechanistic understanding of their formation and functional impact remains lacking. Current mechanistic models operate in a theoretical vacuum, while spatial transcriptomics (ST) provides static snapshots without dynamic insight. To bridge this gap, we present an integrative framework that couples a reaction-diffusion model of tumor-immune kinetics with multicohort ST data. We theoretically explore the spatiotemporal conditions for Turing spatial patterns. Stability analysis identifies a sufficient condition for immune-dominant patterns and predicts that tumor cell motility disrupts the spot-dominant state. The numerical simulations reveal that the spot-dominant immune pattern provides stronger tumor suppression than the stripe-dominant pattern and the increased movement of persistent and resistant tumor cells disrupts this immune dominance. Furthermore, analyses of ST data from three independent tumor cohorts confirm modeling findings, identifying distinct spot-like and stripe-like immune patterns that correlate with the patient prognosis and revealing enrichment of persister cells in stripe regions. Integrating these approaches, we demonstrate that spot-like, but not stripe-like, immune patterns associate with effective tumor suppression and serve as a robust, prognostic biomarker for favorable outcomes. This pattern-driven prognostic power is linked to the underlying biology: spatial cell-cell communication analysis reveals that spot-patterned niches are characterized by enhanced MHC-I antigen presentation, while stripe-like, immunosuppressive microenvironments exhibit aberrant ECM-mediated signaling. Together, this study bridges theoretical modeling and computational spatial omics, establishing a spatiotemporal analysis framework for dissecting immune spatial organization and therapeutic outcomes in solid tumors.