Spatial transcriptomics reveals evolving tumor microenvironmental archetype along the lung premalignancy-adenocarcinoma continuum

Background: Lung adenocarcinoma (LUAD) is one of the most lethal cancer types worldwide, yet how the tumor microenvironmental (TME) archetypes are remodelled during the transition from premalignant lesions (PMLs) to LUAD remains poorly understood. This study aims to systemically characterize the dynamic evolution of tumor immune archetypes across the PML-LUAD continuum using spatially resolved transcriptomics. Methods: Spatial profiling was conducted on 56 samples from 25 patients with paired PMLs and LUADs using the Visium (spot-level) platform. Among them, 12 paired samples from 6 patients were also profiled with the Xenium 5K (single-cell-resolution) platform. TME cellular organizations and interactions were investigated, and distinct spatial TME archetypes were defined. Subsequently, signature scores representing these microenvironmental architectures were derived from spatial data and validated in bulk RNA-sequencing cohorts. Finally, single-cell RNA sequencing (scRNA-seq) was performed on serial lung samples from a carcinogenesis mouse model with Gprc5a knockout ( Gprc5a -/- ) and exposure to nicotine-derived nitrosamine ketone (NNK), and the dynamic transition of TME archetypes during LUAD development was characterized. Results: Compared with PMLs, LUADs displayed an expansion of cancer-associated fibroblasts, alongside intratumoral enrichment of mature tertiary lymphoid structures and depletion of natural killer cells. Additionally, multiple immune-suppressive signals were observed in LUADs and showed progressive upregulation from distal normal to tumor regions . Altogether, these findings indicate dynamic TME remodelling during malignant progression. Spatial mapping of TME cells revealed four distinct immune archetypes, termed IMmune Phenotype spAtial Contextual Typing in LUAD and PML (IMPACT). PMLs were dominated by IMPACT-1, with enrichment of macrophages in the tumor regions, alongside concurrent intratumoral enrichment of T (subtype 1a) or B (subtype 1b) cells. In contrast, LUADs exhibited mixed lymphocyte infiltration (IMPACT-2) or lymphocyte exclusion (IMPACT-3). These subtypes displayed distinct molecular profiles, showing evidence of tumor-promoting inflammation, immune activation, and different mechanisms of immune evasion. IMPACT signatures applied to bulk RNA sequencing cohorts recapitulated key immune features of different IMPACT subtypes and their associations with disease stages. Finally, scRNA-seq analysis on lung samples from the Gprc5a -/- carcinogenesis mouse model further validated the dynamic transition of IMPACT subtypes overtime. Conclusions: PML-LUAD progression is accompanied by profound TME reprogramming. Patient stratifications by spatial immune archetypes may enable predicting disease developmental trajectories and inform strategies for early cancer interception.