Genetic evolution of keratinocytes to cutaneous squamous cell carcinoma

Abstract

Cutaneous squamous cell carcinomas (cSCCs) arise from keratinocytes in the skin, but the molecular changes driving this transformation remain unclear. To better understand this process, we perform multi-omic profiling of keratinocytes, actinic keratoses, and cSCCs. Single-cell mutational analyses reveal that most keratinocytes have remarkably low mutation burdens; however, keratinocytes with TP53 or NOTCH1 mutations exhibit substantially higher burdens. These findings suggest that keratinocytes can withstand high dosages of cumulative ultraviolet radiation, but certain pathogenic mutations break these adaptive mechanisms, inducing a mutator phenotype. Mutational profiling of cSCCs adjacent to actinic keratoses reveals TERT promoter and CDKN2A mutations emerge in actinic keratoses, whereas additional mutations that inactivate ARID2 and activate the mitogen-activated protein kinase pathway delineate the transition to cSCC. Surprisingly, actinic keratoses are often not related to their neighboring cSCC. Spatial analyses reveal gene expression heterogeneity, including checkpoint molecule enrichment at invasive fronts, highlighting tumor and immune cell interactions.

Fig. 1. Keratinocytes have distinct mutational landscapes compared to other cell types.

a Mutation burdens (mutations/megabase; Mut/Mb) of individual keratinocytes (Ker., n = 137) compared to melanocytes (Mel., n = 131) and fibroblasts (Fib., n = 23) derived from 22 independent skin biopsies from 15 unique donors. Each cell represents a single biological unit. b Mutation burden, driver mutations, and mutational signatures for 137 keratinocytes with each column of the three stacked panels representing an individual cell. Top panel: mutation burden of keratinocytes in descending order. Red bars indicate cells harboring one or more pathogenic mutations. Middle panel: tiling plot of pathogenic mutations (rows). Bottom panel: the fractions of different mutational signatures for each cell. White bars indicate keratinocytes with too few mutations to perform signature analysis. c Mutation burdens of keratinocytes with (n = 26) and without (n = 111) pathogenic mutations. d Left panel: fraction of mutations with UV signatures (SBS7a) in keratinocytes (n = 133), melanocytes (n = 123), or fibroblasts (n = 20). Right bar graph: fraction of cells with detectable SBS7a in keratinocytes (69/133), melanocytes (103/123), or fibroblasts (15/20). The center = fraction ± 95% confidence intervals (Poisson exact method). e The data is plotted as in (d) but for SBS1 and SBS5. Fractions of cells with SBS1 and SBS5: keratinocytes (124/133), melanocytes (105/123), and fibroblasts (18/20) ± 95% confidence intervals (Poisson exact method). All comparisons use single cells from separate donors as independent biological units. For all plots, an asterisk (*) or a hash (#) denotes p < 0.05 using the Wilcoxon rank-sum test (two-sided, cell to cell comparisons) or the Poisson test (two-sided, proportion comparisons), respectively. Horizontal bars show the median (panels a and c) or mean (panels d and e). Source data are provided as a Source Data file.

Fig. 2. Clonal architecture of keratinocytes in human skin.

a Clonal structure of keratinocytes from four representative skin biopsies (see Fig. S7 for all biopsies). Each biopsy represents an independent biological unit. The surface area of each biopsy is drawn to scale, as indicated, with dots representing the cells genotyped from each biopsy. The circles group phylogenetically related cells, with pathogenic mutations labeled in red. To the right of each schema, the corresponding phylogenetic trees, rooted in the germline state, are shown for all cells from that biopsy. b The area occupied by individual clones was calculated from the size of each biopsy and the proportion of cells attributed to each clone. Clone areas are shown for keratinocytes and melanocytes with clones harboring pathogenic mutations indicated in red. c Fraction of biopsies with a detectable clone of keratinocytes (Ker., 10 out of 15 biopsies) or melanocytes (Mel., 9 out of 32 biopsies) ± 95% confidence intervals (Poisson exact method). Here, Biopsies are treated as independent biological units. d Fraction of clones with an underlying pathogenic mutation in keratinocytes (5 out of 20 clones from 15 biopsies) or melanocytes (8 out of 14 clones from 32 biopsies) ± 95% confidence intervals (Poisson exact method). An asterisk (*) denotes p < 0.05 (two-sided Poisson exact test). Some donors contributed more than one biopsy from different anatomical sites. Source data are provided as a Source Data file.

Fig. 3. The genetic evolution of a cutaneous squamous cell carcinoma from an actinic keratosis.

a H&E-stained section of a skin biopsy with adjacent areas of squamous cell carcinoma (SCC) and actinic keratosis (AK) dissected, as indicated by the dashed lines. Skin biopsies from 16 different donors (Supplementary Data S4) were analyzed similarly. In five independent cases, squamous cell carcinoma evolved from adjoining actinic keratosis, as shown in this figure (see Fig. S9 for all the cases). Representative images of cases where squamous cell carcinoma did not evolve from adjoining actinic keratosis are shown in Fig. S8. b Scatter plot of mutant allele fractions (MAF) in the squamous cell carcinoma and actinic keratosis reveal three clusters of mutations. c The same scatterplot as shown in panel b with pathogenic mutations annotated. d Copy number alterations were inferred over bins of the genome (columns) for each histologic area (rows) and are shown as a heatmap (red = gain, blue = loss, white = no change). No somatic gains or losses were observed. e Major allele frequency–0.5 (y-axis) for heterozygous SNPs across the genome (x-axis) show loss of heterozygosity over chromosome 9p. f Phylogenetic tree based on somatic mutations (mut) rooted at the germline state. g, h Representative images of immunohistochemistry (IHC) for p53 (g, brown stain) and phospho-MAPK (panel h, purple stain), show keratinocytes overexpressing p53 in both regions with increased phospho-MAPK (Mitogen-activated protein kinase) in the squamous cell carcinoma (n = 5, all the cases where squamous cell carcinoma evolved from adjoining actinic keratosis). Source data are provided as a Source Data file.

Fig. 4. The sequential order of genetic alterations during progression from actinic keratosis to squamous cell carcinoma.

a Phylogenetic trees based on somatic mutations (mut), rooted in the germline states, summarize the evolution of four squamous cell carcinomas (SCC) that evolved from actinic keratoses (AK). See figure S9 for further details on these four cases and Fig. 3 for a summary of the example case. b Eight squamous cell carcinomas that evolved from neighboring precursor lesions were identified as described. The stacked bar plot (top panel) indicates the proportion of mutations, recurrently mutated in these eight cases, in the trunk versus branch of phylogenetic trees. The bar plot (lower panel) indicates the number of cases with a mutation in each pathway. Mutations in the p53, Notch, TERT, and Rb pathways tended to occur early, contributing to the formation of actinic keratoses. Mutations affecting the SWI/SNF chromatin remodeling complex or activating the MAPK/PI3K pathways tended to occur later, driving the transition to squamous cell carcinoma. c The scatterplot shows the frequency of mutations in selected driver genes in normal skin biopsies (total mutations = 234, x-axis) versus squamous cell carcinoma (total mutations = 83, y-axis). Horizontal and vertical error bars for each gene represent 95% confidence intervals for mutation frequencies in normal skin and squamous cell carcinoma, respectively (Poisson exact method). A y = x line is included for orientation. Source data are provided in the Source Data file.

Fig. 5. Spatial heterogeneity in gene expression of immune cells at the interface of squamous cell carcinoma versus actinic keratosis.

Each column of images shows a different view of spatial transcriptomic data from a representative case BB05, including: an H&E overview, annotated spots, and gene expression of immune checkpoints and their ligands. Gene expression intensities represent the combined expression of the checkpoint or ligand genes listed. Zoomed insets show the interface of tumor epithelia and immune cells, illustrating different levels of checkpoint and ligand expression in squamous cell carcinoma versus actinic keratosis. Dotted lines indicate the tumor/immune boundary. This analysis was repeated independently in five cases, and the images shown are representative. See Fig. S14 for an overview of other cases.

Fig. 6. Summary of key events that occur during the evolution of cutaneous squamous cell carcinoma.

After continual exposures to UV radiation, fibroblasts modestly increase their mutation burdens, melanocytes sharply increase their mutation burdens, and keratinocytes have a mixed response. Most keratinocytes accumulate little mutational damage, but a subset with pathogenic mutations builds up mutations more rapidly than other skin cells. UV radiation induces expansion of independent clones of keratinocytes, often in close proximity and admixed, resulting in a complex clonal structure whereby adjacent lesions are not necessarily related. Driver mutations undergo selection in a stereotypical order, linked to histologic and genetic changes that occur during tumor evolution. An immune response builds during progression, but activity is blunted via engagement of immune checkpoints in squamous cell carcinoma.