Fig. 2: Single cell analysis of patient HNC uncovered an epithelial-specific opposite signatures for partial EMT and PI3K signalling.

A–C UMAP plot of single cells from 6 primary and metastatic patient tumours assigned to 13 clusters using the “uwot” package. MT: metastatic tumours collected from lymph nodes; PT: primary tumours. D The clusters were assigned to the indicated cell types by differentially expressed genes and presented as (E) heatmaps. Yellow: high expression; purple: low expression. Selected genes are highlighted and labelled in cluster-matching colours. F Dot plot shows gene expression of epithelial cell markers across the clusters containing 12,341 single cells that were sorted by average expression of these genes. Blue: high expression; gray: low expression; the size of the dots indicates the percentage of cells expressing specific genes. G Pie chart shows the percentage of primary (PT) and metastatic (MT) tumour cells matching atypical, classical, basal and metastatic subtype signatures. H Box plots indicating ssGSEA EMT scores were significantly higher in G1 compared to G2M/S phase indicating EMT is high in non-proliferating cells. Conversely, PI3K/AKT/mTOR hallmark gene set scores were significantly higher in S phase indicating cell proliferation. Comparisons between groups (G1 = 587, G2M = 146, S = 191 cells) were performed using Kruskal-Wallis t test and unpaired two samples Wilcoxon test. Median values are shown in each boxplot. All box plots include the median line, the box denotes the interquartile range (IQR), whiskers denote the rest of the data distribution and outliers are denoted by points greater than ±1.5 × IQR. I Heatmap shows the expression of common p-EMT genes in metastatic and primary patient tumours. J Dot plot illustrates a positive Spearman correlation (two-sided p-value < 0.01) between the genes and p-EMT program in cancer cells. K Scatter graphs show a decreasing percentage of cycling malignant cells and increasing p-EMT ssGSEA scores. The malignant cells are divided into ten sliding windows for each tumour. The consistent negative correlation was calculated using Spearman’s correlation test with two-sided p-value. 95% confidence bands of the best-fit line are shown in gray. Source data are provided as a Source Data file.