Comprehensive Epithelial Atlas Mapping Cellular Dynamics, Chromosomal Alterations and Molecular Programs in Pre- and Early-Stage Lung Adenocarcinoma Manifested as Ground-Glass Nodule

Abstract RATIONALE: Precancerous lesions and early-stage lung adenocarcinoma (LUAD) manifested as ground-glass nodule (GGN) are features with favorable prognosis and indolent growth, previous studies have described the unique immune microenvironment of GGN, but the characteristics of epithelial cells remain largely unexplored. METHODS: We integrated in-house generated and publicly available single-cell sequencing data and extracted epithelial cells for precancerous and early-stage LUAD sample, along with paired WES/WGS data. Proteomic analysis was also conducted for selected samples, and bulk RNA-seq data were collected for external validation. RESULTS: We included over 1 million cells from 191 samples, classified as normal, precursor lesions, minimally invasive adenocarcinoma (MIA), ground-glass invasive adenocarcinoma (gIAC), and solid invasive adenocarcinoma (sIAC). The precursor lesions, MIA, and gIAC groups were classified as GGN. In total, we identified 113,123 high-quality epithelial cells. Most normal epithelial subtypes (AT2, AT1, club cells) declined as disease advanced. Within GGN, two unique intermediate cell types (SCGB3A2+SFTPB+ AT0 like cells/SCGB3A1+SCGB1A1- Club like cells) were enriched, both showing high expression of MHC-II molecules and enrichment in antigen presentation and leukocyte activation pathways, single-sample analysis indicated that these intermediate cells were associated with increased lymphocyte infiltration, which together suggested that they promote an inflammatory microenvironment in GGN. Additionally, two cell subpopulations lacking known markers showed significant chromosomal copy number variations and were classified as malignant cells. Clustering of malignant cells showed patient-specific grouping, highlighting cellular heterogeneity. Calculation of pairwise transcriptomic similarity indicated lower intrer-tumor heterogeneity in GGN compared to sIAC, and one-way ANOVA on principal components within samples showing reduced intra-sample heterogeneity in GGN. InferCNV combined with uphyloplot2 revealed recurring chromosomal alterations in GGN, which were validated through WES/WGS data. We further examined previously identified meta-programs (MP), finding that GGN was significantly enriched in chromatin-related programs, while MYC, hypoxia and cell-cycle programs were prominent in sIAC. We then generated a new 50-gene meta-program that was more representative of most GGN samples using cNMF, and expression of genes within the new MP were verified through proteomic analysis in 25 GGNs and paired normal tissues. Prognostic relevance of the new GGN MP was confirmed in over 2000 bulk RNA-seq samples from 11 datasets, suggesting its fundamental role in LUAD genesis and GGN progression, with potential clinical utility. CONCLUSIONS: This study presents the largest epithelial cell atlas for precancerous lesions and early-stage LUAD manifested as GGN, elucidating the intra- and inter-tumoral heterogeneity and complexity of GGN epithelial cells.