Abstract

Background: Lung adenocarcinoma (LUAD) develops through a stepwise pathological status from atypical adenomatous hyperplasia (AAH), adenocarcinoma in situ (AIS), and minimally invasive adenocarcinoma (MIA) to invasive adenocarcinoma (IA). Although malignant epithelial evolution during this process has been increasingly characterized, the dynamic remodeling of cancer-associated fibroblasts (CAFs) and their contribution to the immunosuppressive tumor microenvironment (TME) remain incompletely explored. Methods: Single-cell RNA sequencing data from treatment-naïve LUAD lesions, including AAH, AIS, MIA, and IA, were analyzed together with external bulk transcriptomic cohorts. CAF subsets were characterized according to their transcriptional features, inferred developmental states, transcription factor activity, functional programs, and predicted cell–cell interactions. Ligand–receptor analysis was used to examine MIF-related communication between epithelial cells and CAFs. MIF-related genes were then used to develop a machine learning-based prognostic signature in the TCGA-LUAD cohort, followed by validation in independent GEO cohorts. Results: Single-cell transcriptomic analysis of 131,639 cells from 25 treatment-naïve LUAD lesions identified six CAF subtypes, including alveolar CAFs, antigen-presenting CAFs, extracellular matrix CAFs, EndMT-like CAFs, inflammatory CAFs, and myofibroblastic CAFs. CAF composition differed across pathological stages, with MIA lesions showing a distinct enrichment of eCAFs and reduced proportions of inflammatory and myofibroblastic CAF populations. Compared with pre-invasive lesions, IA lesions exhibited increased proportions of exhausted CD4+ and CD8+ T cells together with reduced cytotoxic features. Cell–cell communication analysis identified enhanced epithelial–CAF interactions in IA, including enrichment of MIF-CD74/CD44 signaling. Based on MIF-related genes, a machine learning prognostic signature was developed and validated in independent cohorts, consistently stratifying patients into distinct risk groups with significantly different survival outcomes. Conclusions: These findings suggest that CAF-related stromal remodeling is associated with immune suppression during LUAD progression. MIF-mediated epithelial–CAF communication may be involved in the formation of an immunosuppressive microenvironment and is associated with poor prognosis. The MIF-related signature may provide a useful approach for prognostic stratification in LUAD.

IPC Classification

Keywords