Fig. 3

Deregulation of pro-LNM cluster L4 genes associated with p53-DREAM pathway-mediated repression: A Dot plot illustrating the overlap of prognostic gene signatures (survival and LNM-associated genes) with gene sets related to cell cycle and p53. Cell cycle gene sets include genes reported to be periodically expressed during G1 or S phase (G1/S markers, n = 15 genes) and G2 or M phase (G2/M-markers, n = 48 genes) across four cell lines [62]. P53-regulated gene sets include genes that are transcriptionally repressed by the p53-DREAM pathway (p53-DREAM targets, n = 202 genes) [60] and genes that are transcriptionally activated by p53 (P53-activated, n = 343 genes) [89]. Y-axes show − log10 p-values (hypergeometric tests) indicating the significance of overlaps between the prognostic gene signatures and the cell cycle/p53-related gene sets. Horizontal dashed red lines indicate significance thresholds (equivalent to p = 0.05). Point colors indicate the number of overlapping genes. Point sizes indicate the percentage of genes within the prognostic signature that overlap with the cell cycle/p53-related gene set. B Spaghetti plots showing expression changes of pro-LNM cluster L4 genes in Hela cells that were serially collected at 14 timepoints over the course of two cell cycles following cell synchronization (data derived from Dominguez et al. [62]). Y-axes represent expression (normalized FPKM) of genes (points). Lines connect each gene across timepoints. X-axis labels indicate the timepoint (number of hours since cell synchronization) as well as the phase at which cells were harvested. The top panel shows all genes within cluster L4, with point and line colors indicating the phase within which each gene was specifically expressed as reported by Dominguez et al. [62]. Genes labeled “None” represent non-periodic genes, i.e., genes that were stably expressed across phases. The middle and bottom panels show cluster L4 genes that were periodically expressed in G1/S and G2/M phases, respectively. These represent the genes shown in A (top panels). C Differential expression of cluster L4 genes between p53 inactivated and p53 proficient primary HNCs based on bulk gene expression analysis. i The box plots show mean expression (normalized counts) of cluster L4 genes in two HNC bulk gene expression studies including the i TCGA [16] and ii Wichmann [91] studies. Primary HNCs are stratified into groups based on their TP53 mutation and human papillomavirus (HPV) status to illustrate differences of cluster L4 gene expression between p53 proficient and deficient HNCs. P53-proficient HNCs represent those that are HPV − ve and TP53^wt, while p53-inactivated HNCs represent those with TP53 mutations (TP53mut) or HPV positivity (HPV + ve). Expression of cluster L4 genes is also shown for tumor-adjacent normal tissue in the TCGA dataset. ii Expression of cluster L4 genes in HPV − ve primary HNCs of the TCGA study, comparing levels in TP53^wt HNCs with those in HNCs with TP53 mutations of four major functional categories. TP53^mut HNCs were restricted to those with TP53 mutations of only one functional category, to exclude ambiguity in cases with multiple mutations of different categories. Asterisks indicate Wilcoxon rank sum test p-values: *p < 0.05, **p < 0.01, ***p < 0.001. D Upregulation of cluster L4 genes in proliferating malignant cells of TP53^mut HNCs. i Box plots showing mean expression (normalized counts) of cluster L4 genes in cells (points) of TP53^wt and TP53^mut HNCs, with cells stratified by cell type (malignant or non-malignant) as well as cell cycle phase. ii Density plots of malignant cells shown in i, illustrating the distribution of cluster L4 gene expression within each unsupervised cell cluster. Cells are stratified into those derived from TP53^mut and TP53^wt HNCs. Horizontal dashed lines indicate mean cluster L4 gene expression within each cell cluster. iii Forest plot of linear regression coefficients (estimates) that indicate the association of cluster L4 gene expression with TP53 mutation status and cell cycle phase within malignant cells (based on analysis of data shown in i). Coefficients were derived from a multiple linear regression model estimating the association of mean expression of cluster L4 genes (dependent variable) with two independent variables including TP53 mutation status (the mutation status of the overall tumor) and cell cycle phase. Blue points and lines indicate coefficients and 95% confidence intervals, respectively. E Smoothed scatter plots illustrating the correlation of pro-LNM cluster L4 genes with CDKN1A within malignant cells. Scaled mean expression (normalized counts) of cluster L4 genes (X-axes) is plotted against CDKN1A expression (Y-axes) in malignant cells (points) of primary HNCs, including i TP53^mut HNCs of the Puram dataset, ii TP53^wt HNCs of the Puram dataset, and iii all HNCs of the Stanford dataset (TP53 mutation status unknown). Pearson correlation coefficients (R), linear regression p-values (P), and regression lines (red dashed lines) indicate associations between cluster L4 genes expression and CDKN1A