Characterisation of the melanocyte lineage in mice and humans to find specific melanocytic biomarkers

Invasion of melanoma cells into the dermis is a key step towards the progression from a rather harmless in situ melanoma to a life-threatening disease. Unfortunately, the cellular and molecular mechanisms that enable this process are largely unknown. To fill out this gap, we engineered a mouse model that permits the activation of a BrafV600E-driven melanomagenic program in the epidermal melanocytes in the tail skin, mimicking the sequence of events and histopathological features observed during the early phases of human melanoma development. Intravital microscopy highlighted that melanoma cells that invade into the dermis change their morphology and thereby use a limited number of focal invading fronts, suggesting multicellular streaming as a mode of invasion. A time-resolved single-cell RNA sequencing analysis revealed two differentiated cell populations (MEL high and low) during radial expansion of newly formed melanoma clones, while the heterogeneity increased upon dermal invasion. This process was accompanied by a progressive loss of melanocytic differentiation markers and the appearance of two intermediate cell states, that both expressed genes involved in cell migration and extracellular matrix remodeling. Nevertheless, the first population (early transition) remained positive for the melanocytic transcription factor Mitf, whereas the other (late transition) downregulated Mitf. The de-differentiation markers of the previously described neural crest-like and mesenchymal-like transcriptional programs were only detected in lesions collected after dermal colonization took place. In fact, immunostaining confirmed that these cells only emerged in the dermis. The appearance of the intermediate states was accompanied by the progressive decrease in Runx3, and a concomitant increase in Runx1 transcriptional activities, as was shown by gene regulatory network analysis. In-silico perturbation experiments in human melanoma cells indicated that RUNX1 may indeed function as a master regulator of these intermediate states. On top, inhibition of Src, a target gene of Runx1 predicted by SCENIC, hampered dermal invasion in mice. Importantly, immunostaining for Runx1 marked an epidermal cell population both in mouse and human melanoma lesions in the invasive radial growth phase. Also, the gene expression signatures of the intermediate states correlated with Breslow thickness in a clinical cohort of primary thin melanomas. In summary, this study provides new insights and tools to better understand, detect and ultimately prevent the earliest steps of primary melanoma invasion.