Characterizing the Dynamics of the Single Cell Tumor Microenvironment and Immune Cell Subpopulations during the Progression of Lung Adenocarcinoma

Abstract

Abstract Background Lung cancer progression typically involves the transition from atypical adenomatous hyperplasia (AAH) to the invasive adenocarcinoma (IA) stage. The immune status in tumor microenvironment (TME) plays a critical role in governing tumor initiation and progression. Nevertheless, the precise variances in the immune microenvironment among these four states remain uncertain. Methods We employed diverse methodologies including single-cell, spatial, and bulk RNA-sequencing datasets to elucidate the intricate dynamics and interplay of immune cells. Leveraging the distinctions observed among the four states, we developed a prediction model utilizing machine learning techniques to assess the potential survival advantages for patients. Results This research involved a cohort of 52 patients representing four distinct states. Through dimension reduction and clustering techniques, we successfully identified and analyzed nine distinct cell types. In-depth investigation of cell-cell communication and spatial transcriptomics indicated variations in the epithelial-cancer-associated fibroblast (CAF) interaction across the four states. Additionally, our analysis revealed the presence of the macrophage migration inhibitory factor (MIF) signaling pathway in all states, which was associated with notable anti-tumor biological processes. Importantly, the machine learning model based on MIF-related genes exhibited a favorable predictive probability of survival time, as evidenced by an area under the curve (AUC) of 0.68. Conclusion Utilizing a multi-dimensional transcriptomics approach, we conducted an in-depth characterization of the temporal evolution of the TME during the progression of LUAD. Our comprehensive analysis elucidated the intricate variances observed across the spectrum from AAH to IA states. Furthermore, we employed an immune-related machine learning model to validate our findings and accurately forecast the potential survival advantages for patients.