Single-cell-marker-based subtyping and multi-level analyses uncover the prognostic effects, dysregulations and therapeutic indicative potential of an eight-gene signature in lung adenocarcinoma

BACKGROUND: Lung adenocarcinoma (LUAD), the predominant subtype of non-small cell lung cancer, exhibits substantial molecular heterogeneity, complicating the identification of reliable prognostic and therapeutic markers. Single-cell transcriptome analysis, combined with other omics and multi-level analyses, offers a feasible solution to address this issue. METHODS: We clustered, annotated, and identified landmark genes for single cells in LUAD dataset. Using the expression profiles of these marker genes, we classified TCGA-LUAD samples via non-negative matrix factorization (NMF) clustering. The LUAD subtypes were compared for prognosis, tumor stage, gene expression, aging and stemness scores, and immunotherapy response (assessed by TIDE). LASSO-Cox regression analysis selected key genes for a prognostic model, which was validated in training and external datasets. RT-qPCR was used to validate mRNA expression levels. Protein-level dysregulation was confirmed through proteomics and immunohistochemistry (IHC). Additionally, the functional effects of the key genes and their drug-sensitivity associations were explored in LUAD cell lines. RESULTS: Single cells were annotated into 12 types/subtypes, including three epithelial-associated and two fibroblast-associated subtypes. The top 100 marker genes for each cell type/subtype were identified. Based on these marker gene expressions, TCGA-LUAD samples were classified into C1 and C2 subtypes. Compared with C1 LUADs, C2 LUADs exhibited poorer prognosis, a higher proportion of late-stage tumors, higher MKI67 expression, and higher levels of aging and stemness, but lower MHC II gene expression and TIDE scores. LASSO-Cox regression analysis identified an eight-gene signature that effectively distinguished patient outcomes in both training and validation datasets. Dysregulation of these eight genes was confirmed at the mRNA level by RT-qPCR and at the protein level by proteomic and IHC analyses. Moreover, these eight genes were associated with sensitivity to 3-28 anti-cancer drugs in LUAD cell lines. CONCLUSION: Subtyping based on single-cell derived marker genes and multi-level analyses provide a comprehensive understanding of LUAD biology. The distinct features of C1 and C2 subtypes highlight the need for tailored therapies. The eight-gene signature may serve as a novel prognostic, diagnostic, and therapeutic indicator.