Microenvironment-derived factors driving metastatic plasticity in melanoma

Abstract

Cellular plasticity is a state in which cancer cells exist along a reversible phenotypic spectrum, and underlies key traits such as drug resistance and metastasis. Melanoma plasticity is linked to phenotype switching, where the microenvironment induces switches between invasive/MITF^LO versus proliferative/MITF^HI states. Since MITF also induces pigmentation, we hypothesize that macrometastatic success should be favoured by microenvironments that induce a MITF^HI/differentiated/proliferative state. Zebrafish imaging demonstrates that after extravasation, melanoma cells become pigmented and enact a gene expression program of melanocyte differentiation. We screened for microenvironmental factors leading to phenotype switching, and find that EDN3 induces a state that is both proliferative and differentiated. CRISPR-mediated inactivation of EDN3, or its synthetic enzyme ECE2, from the microenvironment abrogates phenotype switching and increases animal survival. These results demonstrate that after metastatic dissemination, the microenvironment provides signals to promote phenotype switching and provide proof that targeting tumour cell plasticity is a viable therapeutic opportunity.

Figure 1. Melanomas become differentiated after dissemination to secondary sites.

(a, top) ZMEL1-GFP zebrafish melanoma cells (mitf-BRAF^V600E; p53^−/−) were orthotopically transplanted into the ventral skin of the transparent casper strain and then imaged using brightfield and GFP over 28 days. At days 1 and 7 post-transplant, the tumours are GFP+ but devoid of pigmentation, a marker of differentiation. By day 14 post transplant, the primary tumour mass in the orthotopic site has become deeply pigmented and the animal has developed small anterior metastases that are GFP+ but unpigmented (red arrowhead). By day 28, these metastases have now enlarged and are clearly pigmented (black arrowhead), consistent with metastatic differentiation. (a, bottom) ZMEL1-GFP cells were transplanted into the vasculature of larval casper recipients to assess direct differentiation capacity at sites of metastatic colonization, bypassing the primary skin site. Similar to what is seen in the orthotopic transplantation, the cells are initially unpigmented at days 1 and 7, but become increasingly pigmented at days 14 and day 28, indicating that cells can directly differentiate after extravasation. (b) Histological analysis of a larval casper recipient transplanted with ZMEL1-GFP cells shows heterogeneous acquisition of pigmentation, with cells near muscle invasive disease showing increased evidence of melanization (N=notochord, T=tumour, M=muscle). (c) Time-lapse imaging of ZMEL1-GFP morphology in a larval casper recipient shows that cells that exit the vasculature and enter the tailfin epithelial layer gradually acquire a dendritic phenotype that is characteristic of differentiated melanocytes. (d) Enlargement of a dendritic cell in the tailfin. Images are representative of n=10–20 fish per group.

Figure 2. Disseminated metastatic cells exhibit a differentiation gene signature.

(a) ZMEL1-GFP cells were transplanted into the vasculature of a larval zebrafish, and then fish grew until 21 days when they had widespread tumour dissemination. These fish were then disaggregated and the post-dissemination ZMEL1-GFP+ cells were isolated by fluorescence-activated cell sorting (FACS) (left). The parental ZMEL1-GFP cells maintained in culture were trypsinized and similarly subject to FACS sorting (right). These two populations were then subject to RNA-seq. (b) Expression of melanocyte/pigmentation genes (that is, PMEl, TYR, SLC24A5) in the disseminated ZMEL1 cells compared with parental, showing a significant upregulation of a differentiation gene program. (c) GSEA shows a significant enrichment between the disseminated ZMEL signature and a signature of human differentiated melanocytes. (d) GSEA shows a significant enrichment between the disseminated ZMEL1 signature and the human melanoma subtypes classified as ‘differentiated/proliferative'. (e,f) Cox proportional Hazard model of TCGA data for melanocyte differentiation genes in either stage I (e) localized disease versus stage III/IV (f) metastatic disease shows that the differentiation signature portends a worse prognosis in metastatic patients. (g,h) Kaplan–Meier survival analysis for the melanosome protein PMEL, showing significantly worse survival for stage III/IV patients compared with stage I. Error bars are s.e.m., with n=2–6 animals per group.

Figure 3. Microenvironmental factors inducing the differentiated/proliferative state.

(a) Ingenuity Pathway Analysis of ZMEL1-GFP cells after metastatic dissemination suggests six pathways (left) that could mediate microenvironmental-mediated differentiation/proliferation. P values indicate estimated likelihood that the indicated pathway is altered in the RNA-seq data set. To test these pathways, ZMEL-GFP cells were treated in vitro with agonists of these six pathways at the indicated doses (middle) and differentiation and proliferation was assessed using the InCell high-throughput/high content scanning system (right). (b,c) Heatmaps showing the effect of the various agonists on proliferation tested in either low- or high-serum conditions in ZMEL1 melanoma cells (b) or A375 melanoma cells (c). Red indicates increased proliferation, whereas blue indicates decreased proliferation (compared with dimethylsulphoxide control). (d,e) Heatmaps showing the effect of the agonists on cell elongation, which is a reflection of melanoma differentiation, in ZMEL1 (d) or A375 (e) melanoma cells. Yellow indicates increased elongation, while purple indicates decreased proliferation. (f) To more directly test whether elongation is associated with differentiation, ZMEL1-GFP cells were treated with Endothelin-3 and L-DOPA, or the combination and melanin content assessed. Both agonists caused a significant increase in melanin content (*=P<0.05). (g) Photograph of the cell pellets (equal numbers of cells) treated with either dimethylsulphoxide, Endothelin-3 or L-DOPA. (h) Proliferation curves showing opposing effects of Endothelin-3 as compared with L-DOPA in ZMEL1 cells, suggesting that EDN3 is uniquely capable of inducing both proliferation and differentiation across multiple cell types. Error bars are s.e.m., with n=4–6 biological replicates per group.

Figure 4. Microenvironmental CRISPR against EDN3b abrogates phenotype switching.

(a) Control wild-type fish of the AB strain that was injected with Cas9 protein alone. Higher magnification views of the melanocytes from this fish is shown below, indicating these are fully mature, pigmented melanocytes. (b) An AB fish in the F1 generation that had been injected with a CRISPR gRNA against EDN3b, showing a severe loss of melanocytes over the entire body of the zebrafish. Higher magnification views (below) from this fish show a decreased number of pale, misshapen melanocytes. (c) An AB fish in the F1 generation that had been injected with a CRISPR gRNA against ECE2b, showing a melanocyte defect that is highly similar to that seen in the EDN3b mutant. (d) Schema for testing the effects of microenvironmental EDN3b on melanoma growth and differentiation. Equal numbers of ZMEL1-GFP cells were transplanted subcutaneously either in a WT recipient, EDN3b mutant recipient or the ECE2b mutant recipient, who differ only in the loss of function EDN3b in the microenvironment. (e) Tumour growth in the WT recipient shows large pigmented tumours in multiple subcutaneous sites, consistent with phenotype switching to a more differentiated/proliferative state. (f,g) Tumour growth in the EDN3b or ECE2b recipients is impaired, with smaller tumours that are markedly less pigmented, consistent with reduced phenotypic switching due to loss of endothelin-3 signalling from the microenvironment. (h) Quantification of tumour area in WT versus EDN3b versus ECE2b backgrounds at 14 days post transplant demonstrates a significant decrease in tumour size in the CRISPR mutant (*=P<0.05, WT versus EDN3b and WT versus ECE2b, ANOVA, n=15 WT, n=24 EDN3b, n=19 ECE2b). (i) Measurement of mitf-GFP pixel intensity in WT versus EDN3b versus ECE2b recipients shows a significant decrease in overall GFP+ intensity in the ECE2b mutant (*=P<0.05, WT versus ECE2b). (j) Kaplan–Meier survival curves of WT versus EDN3b−/− recipient fish, showing a significantly longer survival time in the EDN3b and ECE2b recipients (+, P=0.0035, logrank test). Error bars are s.e.m. with the numbers of animals indicated above.

Figure 5. Histologic analysis of melanomas that develop in WT background versus ECE2b or EDN3b-deficient backgrounds.

(a–c) H&E staining demonstrates areas of central necrosis (red arrows) in the ECE2b and EDN3b-deficient background when compared with WT (WT=0/4, ECE2b=4/4, EDN3b=4/4 with central necrosis). (d–f) A significant decrease in proliferation, as assessed by phospho-H3 staining, is seen in the EDN3b-deficient tumours, which is quantified in the (s) panel below (*P<0.05, ANOVA). (g–i) A significant increase in apoptosis as measured by cleaved caspase expression is seen in both ECE2b and EDN3b-deficient backgrounds, quantified in panel (t) below (*,+, P<0.05, ANOVA). (j–l) A significant decrease in melanin content, as measured by Fontana-Masson staining, is seen in the ECE2b and EDN3b backgrounds as compared with WT and quantified in panel (u) below (*,+, P<0.05, ANOVA). (m–o) Staining for human BRAF^V600E was uniform across all three genotypes, as was GFP expression (p–r), indicating that expression of those transgenes from the mitfa promoter was not affected. Error bars are s.e.m.

Figure 6. An overview model of phenotype switching factors in melanoma.

The major driver of both proliferation and differentiation is EDN3 (synthesized by ECE2b), whereas proliferation alone can be augmented by pro-proliferation signals such as SCF and IGF1 and differentiation augmented by L-DOPA. These convergent signals yield cells that transition from the invasive gene signature to an MITF-dominated gene signature with expression of proliferation genes (MYC) and differentiation genes (PMEL, TYRP1, TYR).