Integrative Single-Cell and Machine Learning Analysis Reveals Immune Microenvironment Remodelling in Lymph Node Metastasis of Lung Adenocarcinoma

Abstract

Lymph node metastasis is a pivotal determinant of prognosis in lung adenocarcinoma, yet its impact on tumour microenvironment remodelling remains insufficiently characterised. In this study, we employed single-cell RNA sequencing to compare metastatic and non-metastatic lymph nodes, delineating metastasis-associated immune and stromal alterations. Metastatic nodes exhibited marked reductions in dendritic cell and T cell infiltration alongside increases in monocytes and SPP1⁺ macrophages, indicative of an immunosuppressive milieu. Intercellular communication analysis revealed strengthened interactions among SPP1⁺ macrophages, monocytes, and epithelial cells, suggesting coordinated signalling that may further enforce immune suppression. Integrating differentially expressed genes with multi-omic features, we developed an ensemble machine learning model, LNRScore, which robustly stratified patients into distinct risk groups. A high LNRScore was associated with poorer prognosis and reduced immune infiltration, whereas a low LNRScore correlated with higher immunogenicity and greater predicted responsiveness to immunotherapy based on TCIA assessments. Further analyses identified HMGA1 as a core gene within the model, closely linked to adverse outcomes; functional assays demonstrated that high HMGA1 expression promotes the proliferation and migration of the LLC cell line, supporting its role in metastatic progression. Collectively, this study defines the immune microenvironmental remodelling associated with lymph node metastasis, establishes an effective risk prediction model (LNRScore), and highlights HMGA1 as a potential target for precision diagnosis and therapy in lung adenocarcinoma.