Skip to main content
. 2021 May 26;10:e61461. doi: 10.7554/eLife.61461

Figure 5. scRNAseq analysis of EpCAMlo cells reveals specific markers of partial EMT cells.

(A) tSNE of HCT116 and SW480 cells based on the variable expressed genes across EpCAMlo and EpCAMhi populations. (B) tSNE of HCT116 and SW480 cells based on genes from the epithelial to mesenchymal transition (EMT) signature (N = 107). (C) Heatmap of differentially expressed genes between EpCAMlo and EpCAMhi populations. Cells were ranked according to their EMT score (EMT = M – E). (D) Gene expression trends projected over the EMT axis. Expression values were imputed with MAGIC, scaled by their z-score, and smoothened by general additive models to visualize the gene expression trend. (E) RNA velocity analysis of the HCT116 scRNAseq data. Cells from both populations were moving in their respective state (left panel) with a minority population of transitioning cells (right panel). (F) Projection of the pEMT score on the UMAP embedding of the HCT116 cell line. (G) The expression of SPARC on the UMAP embedding of the HCT116 cell line. (H) Left panels: qPCR analysis of SPARC overexpression in the subpopulations of HCT116 and SW480. Gene expression values are relative to GAPDH and normalized to the EpCAMhi subpopulation. Right panel: quantification of the transwell migration assay upon overexpression of SPARC.

Figure 5—source data 1. EMT and CMS4 classifiers.

Figure 5.

Figure 5—figure supplement 1. Further scRNAseq analysis of EpCAMlo cells: EMT/MET transcriptional trajectory.

Figure 5—figure supplement 1.

(A) tSNE of the SW480 cell line indicating an additional subpopulation in the EpCAMhi population (left panel). Using a signature derived from the bulk RNAseq, this population was identified as the ‘sphere’ population (middle panel) and annotated to be excluded for further analysis (right panel). (B) Left panel: the SW480 cell line, after exclusion of the ‘sphere’ population, contains slightly higher variability compared to the HCT116 cell line, as evidenced by the variance of the top 50 principal components. Right panel: while in HCT116 most of the variable expressed genes are differentially expressed between the EpCAMhi and EpCAMlo population, this is not the case in SW480, where most of the highly variable genes do not differ between the two populations. (C) Top panels: expression values of VIM, ZEB1, and CD44 on the UMAP embedding of the HCT116 cell line. Lower panels: projection of the RNA velocity direction of the same genes. (D) HCT116 UMAP embedding annotated with the eight unsupervised clusters. (E) Heatmap of HCT116 with expression values of the epithelial to mesenchymal transition (EMT) signature averaged by the eight clusters. Clusters were ranked according to their EMT score, and genes were clustered in four distinct gene sets using k-means clustering. (F) Schematic diagram showing a transcriptional trajectory with distinct gene arrays through which pEMT cells arise.