PURPOSE: Immunotherapy has transformed the neoadjuvant treatment landscape for patients with resectable locally advanced non-small cell lung cancer (NSCLC). However, a population of patients cannot obtain major pathologic response (MPR) and thus benefit less from neoadjuvant immunotherapy, highlighting the need to uncover the underlying mechanisms driving resistance to immunotherapy. METHODS: Two published single-cell RNA sequencing (scRNA-seq) datasets were used to analyze the subsets of cancer-associated fibroblasts (CAFs) and T cells and functional alterations after neoadjuvant immunotherapy. The stromal signature predicting ICI response was identified and validated using our local cohort with stage III NSCLC receiving neoadjuvant immunotherapy and other 4 public ICI transcriptomic cohorts. RESULTS: Non-MPR tumors showed higher enrichment of CAFs and increased extracellular matrix deposition than MPR tumors, as suggested by bioinformatic analysis. Further, CAF-mediated immune suppression may involve reciprocal interactions with T cells in addition to a physical barrier mechanism. In contrast, MPR tumors demonstrated therapy-induced activation of memory CD8 T cells into an effector phenotype. Additionally, neoadjuvant immunotherapy resulted in expansion of precursor exhausted T (Texp) cells, which were remodeled into an anti-tumor phenotype. Notably, we identified metabolic heterogeneity within distinct T cell clusters during immunotherapy. Methionine recycling emerged as a predictive factor for T-cell differentiation and a favorable pathological response. The stromal signature was associated with ICI response, and this association was validated in five independent ICI transcriptomic cohorts. CONCLUSION: These discoveries underscore the distinct tumor microenvironments in MPR and non-MPR patients and may elucidate resistance mechanisms to immunotherapy in NSCLC.