Single-cell analysis and machine learning-based integration develop an immune-responsive signature of antigen-presenting cancer-associated fibroblasts in lung adenocarcinoma.

Background: Cancer-associated fibroblasts (CAFs) are pivotal regulators of the tumor immune microenvironment, shaping immune responses and influencing therapeutic outcomes. While previous studies have predominantly focused on CAF subpopulations that impair responses to immune checkpoint inhibitors (ICIs), CAF subsets associated with favorable ICIs responses in lung adenocarcinoma (LUAD) remain underexplored. In this study, we integrated bulk RNA and single-cell RNA sequencing data from LUAD samples to identify CAF subpopulations relevant to ICIs efficacy. Methods: Using a machine learning-driven approach, we developed a robust immune response signature based on this antigen-presenting CAFs (apCAFs) subset to predict ICIs responses. Results: We uncovered a novel subset of apCAFs exhibiting macrophage-like features, characterized by the expression of major histocompatibility complex (MHC) class II, CD74, and costimulatory molecules (CD80, CD86, CD83, and CD40). This subset, distinct from classic apCAFs described in other cancer types, is strongly associated with favorable ICIs responses across multiple datasets. Notably, these macrophage-like apCAFs are present in LUAD samples prior to treatment, although their abundance varies among individuals. Patients classified as high-risk using signature calculated by a machine learning-driven approach exhibited lower overall survival rates and diminished immune cell infiltration following ICIs therapy. Conclusions: Collectively, our findings establish a critical link between macrophage-like apCAFs and ICIs efficacy, offering a clinically applicable signature for patient stratification and guiding therapeutic strategies targeting the tumor microenvironment.