SummaryOnce melanomas have progressed with acquired resistance to mitogen-activated protein kinase (MAPK)-targeted therapy, mutational heterogeneity presents a major challenge. We therefore examined the therapy phase before acquired resistance had developed and discovered the melanoma survival oncogene MITF as a driver of an early non-mutational and reversible drug-tolerance state, which is induced by PAX3-mediated upregulation of MITF. A drug-repositioning screen identified the HIV1-protease inhibitor nelfinavir as potent suppressor of PAX3 and MITF expression. Nelfinavir profoundly sensitizes BRAF and NRAS mutant melanoma cells to MAPK-pathway inhibitors. Moreover, nelfinavir is effective in BRAF and NRAS mutant melanoma cells isolated from patients progressed on MAPK inhibitor (MAPKi) therapy and in BRAF/NRAS/PTEN mutant tumors. We demonstrate that inhibiting a driver of MAPKi-induced drug tolerance could improve current approaches of targeted melanoma therapy.

Despite the general focus on an invasive and de-differentiated phenotype as main driver of cancer metastasis, in melanoma patients many metastatic lesions display a high degree of pigmentation, indicative for a differentiated phenotype. Indeed, studies in mice and fish show that melanoma cells switch to a differentiated phenotype at secondary sites, possibly because in melanoma differentiation is closely linked to proliferation through the lineage-specific transcriptional master regulator MITF. Importantly, while a lot of effort has gone into identifying factors that induce the de-differentiated/invasive phenotype, it is not well understood how the switch to the differentiated/proliferative phenotype is controlled. We identify collagen as a contributor to this switch. We demonstrate that collagen stiffness induces melanoma differentiation through a YAP/PAX3/MITF axis and show that in melanoma patients increased collagen abundance correlates with nuclear YAP localization. However, the interrogation of large patient datasets revealed that in the context of the tumour microenvironment, YAP function is more complex. In the absence of fibroblasts, YAP/PAX3-mediated transcription prevails, but in the presence of fibroblasts tumour growth factor-β suppresses YAP/PAX3-mediated MITF expression and induces YAP/TEAD/SMAD-driven transcription and a de-differentiated phenotype. Intriguingly, while high collagen expression is correlated with poorer patient survival, the worst prognosis is seen in patients with high collagen expression, who also express MITF target genes such as the differentiation markers TRPM1, TYR and TYRP1, as well as CDK4. In summary, we reveal a distinct lineage-specific route of YAP signalling that contributes to the regulation of melanoma pigmentation and uncovers a set of potential biomarkers predictive for poor survival.

Approaches to prolong responses to BRAF targeting drugs in melanoma patients are challenged by phenotype heterogeneity. Melanomas of a “MITF‐high” phenotype usually respond well to BRAF inhibitor therapy, but these melanomas also contain subpopulations of the de novo resistance “AXL‐high” phenotype. > 50% of melanomas progress with enriched “AXL‐high” populations, and because AXL is linked to de‐differentiation and invasiveness avoiding an “AXL‐high relapse” is desirable. We discovered that phenotype heterogeneity is supported during the response phase of BRAF inhibitor therapy due to MITF‐induced expression of endothelin 1 (EDN1). EDN1 expression is enhanced in tumours of patients on treatment and confers drug resistance through ERK re‐activation in a paracrine manner. Most importantly, EDN1 not only supports MITF‐high populations through the endothelin receptor B (EDNRB), but also AXL‐high populations through EDNRA, making it a master regulator of phenotype heterogeneity. Endothelin receptor antagonists suppress AXL‐high‐expressing cells and sensitize to BRAF inhibition, suggesting that targeting EDN1 signalling could improve BRAF inhibitor responses without selecting for AXL‐high cells.

A major challenge for managing melanoma is its tumour heterogeneity based on individual co‐existing melanoma cell phenotypes. These phenotypes display variable responses to standard therapies, and they drive individual steps of melanoma progression; hence, understanding their behaviour is imperative. Melanoma phenotypes are defined by distinct transcriptional states, which relate to different melanocyte lineage development phases, ranging from a mesenchymal, neural crest‐like to a proliferative, melanocytic phenotype. It is thought that adaptive phenotype plasticity based on transcriptional reprogramming drives melanoma progression, but at which stage individual phenotypes dominate and moreover, how they interact is poorly understood. We monitored melanocytic and mesenchymal phenotypes throughout melanoma progression and detected transcriptional reprogramming at different stages, with a gain in mesenchymal traits in circulating melanoma cells (CTCs) and proliferative features in metastatic tumours. Intriguingly, we found that distinct phenotype populations interact in a cooperative manner, which generates tumours of greater “fitness,” supports CTCs and expands organotropic cues in metastases. Fibronectin, expressed in mesenchymal cells, acts as key player in cooperativity and promotes survival of melanocytic cells. Our data reveal an important role for inter‐phenotype communications at various stages of disease progression, suggesting these communications could act as therapeutic target.