单细胞转录组分析揭示肺腺癌的瘤内异质性

The TCGA-LUAD (n = 503), GSE68465 (n = 442), GSE72094 (n = 398), and GSE26939 (n = 115) datasets were retrieved for prognostic assessment. Subgroup analysis was performed for the epithelial cells, endothelial cells, immune cells, and fibroblasts, and the transcription factors and tumor-related pathways enriched in each subgroup were analyzed using PROGENy and DoRothEA package. The InferCNV software was used to calculate the copy number variations (CNVs) in tumor cell subgroups with normal epithelial cells as the reference. The association between the annotated cell types and survival was analyzed using the Scissor software.

We identified eight major cell types in LUAD, namely epithelial cells, NK cells, T and B cells, endothelial cells, mast cells, myeloid cells, and fibroblasts, of which the epithelial cells and B cells showed a marked increase in the tumor samples. In addition, we also detected an intense signal transduction network from the cancer-associated fibroblasts (CAFs) to malignant cells, mainly involving the DCN/MET, COLA1/DDR1, COL1A1/SDC1, and COL1A2/SDC1 pathways. The tumor differentiation trajectory consisted of state 1 and state 2, which were enriched in HIF1A, and state 4. Furthermore, only a few B cells originated from the normal tissue, suggesting significant recruitment and infiltration of B cells in LUAD. Based on differentially upregulated genes in the cells positively and negatively associated with survival, we established a prognostic model that showed satisfactory predictive performance in three different cohorts. States 3 and 2 of epithelial cells included the majority of cells with KRAS mutation, whereas state 2 showed high frequency of EGFR mutations.

We analyzed intra-tumor heterogeneity of LUAD at the single-cell level and developed a prognostic index that was highly effective across multiple cohorts.