Small nucleotide variants in non-coding regions of the genome can alter transcriptional regulation, leading to changes in gene expression which can activate oncogenic gene regulatory networks. Melanoma is heavily burdened by non-coding variants, representing over 99% of total genetic variation, including the well-characterized TERT promoter mutation. However, the compendium of regulatory non-coding variants is likely still functionally under-characterized. We developed a pipeline to identify hotspots, i.e. recurrently mutated regions, in melanoma containing putatively functional non-coding somatic variants that are located within predicted melanoma-specific regulatory regions. We identified hundreds of statistically significant hotspots, including the hotspot containing the TERT promoter variants, and focused in on a hotspot in the promoter of CDC20. We found that variants in the promoter of CDC20, which putatively disrupt an ETS motif, lead to lower transcriptional activity in reporter assays. Using CRISPR/Cas9, we generated an indel in the CDC20 promoter in a human A375 melanoma cell line and observed decreased expression of CDC20, changes in migration capabilities, and an altered transcriptional state previously associated with neural crest transcriptional programs and melanoma initiation. Overall, our analysis prioritized several recurrent functional non-coding variants that, through downregulation of CDC20, led to perturbation of key melanoma phenotypes.

The role of a neural crest developmental transcriptional program, which critically involves Sox10 upregulation, is a key conserved aspect of melanoma initiation in both humans and zebrafish, yet transcriptional regulation of sox10 expression is incompletely understood. Here we used ATAC-Seq analysis of multiple zebrafish melanoma tumors to identify recurrently open chromatin domains as putative melanoma-specific sox10 enhancers. Screening in vivo with EGFP reporter constructs revealed 9 of 11 putative sox10 enhancers with embryonic activity in zebrafish. Focusing on the most active enhancer region in melanoma, we identified a region 23 kilobases upstream of sox10, termed peak5, that drives EGFP reporter expression in a subset of neural crest cells, Kolmer-Agduhr neurons, and early melanoma patches and tumors with high specificity. A ~200 base pair region, conserved in Cyprinidae, within peak5 is required for transgenic reporter activity in neural crest and melanoma. This region contains dimeric SoxE/Sox10 dimeric binding sites essential for peak5 neural crest and melanoma activity. We show that deletion of the endogenous peak5 conserved genomic locus decreases embryonic sox10 expression and disrupts adult stripe patterning in our melanoma model background. Our work demonstrates the power of linking developmental and cancer models to better understand neural crest identity in melanoma.

The manner in which zebrafish are fed may have important impacts on the behavior of disease models. We examined the effect of different feeding regimens on the rate of overt melanoma tumor onset in a p53/BRAF-dependent model, a commonly used read-out in this and many other cancer models. We demonstrate that increased feeding leads to more rapid melanoma onset. The ability to modulate overt tumor onset rates with this regimen indicates additional flexibility to ‘tune’ the system to more quickly generate tumors for study and to begin to address questions related to cancer metabolism using the zebrafish model.