Single-cell profiling of tumor heterogeneity and the microenvironment in advanced non-small cell lung cancer

Abstract

Lung cancer is a highly heterogeneous disease. Cancer cells and cells within the tumor microenvironment together determine disease progression, as well as response to or escape from treatment. To map the cell type-specific transcriptome landscape of cancer cells and their tumor microenvironment in advanced non-small cell lung cancer (NSCLC), we analyze 42 tissue biopsy samples from stage III/IV NSCLC patients by single cell RNA sequencing and present the large scale, single cell resolution profiles of advanced NSCLCs. In addition to cell types described in previous single cell studies of early stage lung cancer, we are able to identify rare cell types in tumors such as follicular dendritic cells and T helper 17 cells. Tumors from different patients display large heterogeneity in cellular composition, chromosomal structure, developmental trajectory, intercellular signaling network and phenotype dominance. Our study also reveals a correlation of tumor heterogeneity with tumor associated neutrophils, which might help to shed light on their function in NSCLC.

Fig. 1. Advanced NSCLC single-cell atlas.

a Graph illustration of the baseline information of the 42 patients, including subtypes, stages, mutation status, and smoking history. b UMAP plot of 90,406 cells from 42 patients, colored by their 11 major cell types. c Heatmap of canonical cell-type markers of 11 major cell types. d UMAP plot of all cells, colored by patients. e Major cell-type composition of each patient. Biopsies were all taken from the primary lung tumors. Source data are provided as a Source Data file.

Fig. 2. Inter- and intratumor heterogeneity of cancer cells.

a Heatmap of CNA profiles inferred from scRNA-seq of tumor cells of patients. Red indicated genomic amplifications and blue indicated genomic deletions. The x-axis showed all chromosomes in the numerical order. The y-axis was marked by both patient subgroups. b Heatmap displaying proportions of cancer cells of each patient in cancer clusters. The clustering results of cancer cells were generated using resolution 0.4 in Seurat. The arrangement of the patients on the y-axis were based on their similarities using hierarchical clustering. c UMAP visualization of cancer cell clusters. The cluster IDs corresponded to cluster IDs shown in b. d Correlation between ITH_CNA and ITH_GEX for 42 patients. Shaded areas corresponded to the 0.95 confidence interval analyzed by two-sided t-test. e Statistical tests of ITH_CNA and ITH_GEX between patients in different groups, LUSCn, LUADm and LUADn (LUSCn: n = 16, LUADn: n = 6, LUADm: n = 12 biological independent samples. *p ≤ 0.05; ns: p > 0.05). Two-sided unpaired Wilcoxon test was performed to compare between groups. The lower hinge, middle line, and upper hinger of boxplots represented the first, second, and third quartiles of the distributions. The upper and lower whiskers corresponded to the largest and smallest data points within the 1.5 interquartile range. All actual data values were also plotted as dots alongside the boxplots.

Fig. 3. Phenotypes of lung epithelial cells and their evolutionary trajectory into cancer cells.

a UMAP projection of alveolar cells. Alveolar cells could be further divided into two clusters, both of which are AT2 cells. They were denoted as AT2-1 and AT2-2. b Volcano plot of differentially expressed genes between AT2-1 and AT2-2 cells. Difference between percentage of cells expressed in two clusters was plotted against log fold change of average expressions. c UMAP visualization of epithelial cell subtypes. Epithelial cells could be further annotated into basal cells, club cells, and ciliated cells. d Heatmap of canonical marker genes of epithelial lung subtypes. e Developmental trajectories of AT2 cells, club cells, and LUAD tumor cells. Normal cells were shown as a whole for each type, and cancer cells were shown separately for each patient. f Developmental trajectories of basal cells, club cells, and LUSC tumor cells.

Fig. 4. Subtypes and developmental trajectory of T cells.

a UMAP visualization of 6 T cell subtypes and 2 NK cell subtypes (left) and predicted T cell subtypes by singleR (right). b Heatmap of selected markers for 2 NK clusters. c Heatmap of T subtype markers and selected functional genes. d Transitional relationship among CD4 T cells predicted by Slingshot. Rainbow coloring from red to blue represented the begin to end of the trajectory. e Illustration of CD4 T cell differentiation pathways inferred by Monocle and the relative locations of each CD4 T subtypes along the development pathways. The red and blue arrows indicated the two pseudotime directions of cell development. The grey section represented the beginning of the trajectory before the branching point. f Heatmap showing relative expressions of canonical markers of CD4 T cells along inferred trajectories. The red and blue branches correspond to the two developmental directions in e.

Fig. 5. Correlation analysis of cellular composition, tumor subtypes, and ITH.

Cellular composition analysis of cell type between patient groups for a neutrophils and b macrophage subtypes. Two-sided unpaired Wilcoxon test was performed to compare between groups for tests in a and b (LUSCn: n = 16, LUADn: n = 6, LUADm: n = 12 for both a and b. **p ≤ 0.01; *p ≤ 0.05; ns: p > 0.05). The lower hinge, middle line, and upper hinger of boxplots represented the first, second, and third quartiles of the distributions. The upper and lower whiskers corresponded to the largest and smallest data points within the 1.5 interquartile range. All actual data values were also plotted as dots alongside the boxplots. c Survival analysis for tissue-resident macrophage markers (MARCO and CD207) of LUAD and LUSC. d Correlation analysis between ITH_GEX and the cellular composition of patients. Only significantly associated cell types were shown. The tumor subtypes (LUAD, LUSC, and NSCLC) were shown in different colors and p-values were obtained by two-side t-tests (LUAD: n = 18, LUSC: n = 22, and NSCLC: n = 2).

Fig. 6. Cell and gene interaction networks.

a The cellular interaction network among cell types of NSCLC patients. The line width and color were proportional to numbers of interactions between cell types. b Interacting molecular networks. Within each connected network, node (gene) sizes were proportional to the number of neighbors (interacting genes) of each node. Heatmaps shown selected interacting pairs for selected cell types in LUADm, LUADn, and LUSCn groups. Z-scores of expression levels were represented by color, and dot size displayed the proportion of patients who have significant interaction for the given ligand-receptor pair. c chemokine and chemokine receptors between cancer cells, T cells and DCs. d selected growth factors between cancer cells and stromal cells. e selected checkpoints between cancer cells, macrophages, and T cells.