Volume 59, Issue 4, Pages 664-676 (August 2015) Interactions of Melanoma Cells with Distal Keratinocytes Trigger Metastasis via Notch Signaling Inhibition of MITF  Tamar Golan, Arielle R. Messer, Aya Amitai-Lange, Ze’ev Melamed, Reut Ohana, Rachel E. Bell, Oxana Kapitansky, Galya Lerman, Shoshana Greenberger, Mehdi Khaled, Nira Amar, Jean Albrengues, Cedric Gaggioli, Pinchas Gonen, Yuval Tabach, David Sprinzak, Ruby Shalom-Feuerstein, Carmit Levy  Molecular Cell  Volume 59, Issue 4, Pages 664-676 (August 2015) DOI: 10.1016/j.molcel.2015.06.028 Copyright © 2015 Elsevier Inc. Terms and Conditions

Molecular Cell 2015 59, 664-676DOI: (10.1016/j.molcel.2015.06.028) Copyright © 2015 Elsevier Inc. Terms and Conditions

Figure 1 Interaction of Melanoma Cells with Differentiated Keratinocytes Promotes Invasion (A) WM3526 cells stably transfected with GFP-expressing plasmid were co-cultured with the indicated primary human cell types, and then an invasion assay was conducted. Graph demonstrates proportion (%) of invading cells, and a representative image of invading cells is shown for each primary cell type in the lower panel; blue is DAPI nuclear staining. Error bars represent ± SEM, ∗ indicates p < 0.05 (n = 4). (B) Melanin content of WM3526 cells co-cultured with differentiated keratinocytes and naive melanoma cells (control). (C) Left panel: Experimental design. Right panel: WM3526 cells stably transfected with a luciferase reporter were co-cultured with the indicated primary human cell types. Co-cultured cells were co-grafted into immunodeficient mice and analyzed using bioluminescence live imaging. One representative image of bioluminescence in the lung is shown for each primary cell type. Graph represents the average bioluminescence activity in the lung and is indicative of metastasis. Error bars represent ± SEM, ∗ indicates p < 0.05 (n = 4). See also Figure S1. Molecular Cell 2015 59, 664-676DOI: (10.1016/j.molcel.2015.06.028) Copyright © 2015 Elsevier Inc. Terms and Conditions

Figure 2 Differentiated Keratinocytes Induce Notch Signaling in Melanoma Cells (A) NOTCH1, NOTCH2, and NOTCH3 mRNA levels in highly invasive (WM1716/WM3314) and poorly invasive (WM3526/WM3682) melanoma cells. Results normalized to actin; n = 3, error bars represent ± SEM. (B) Immunofluorescence analysis of normal human skin vertical cross sections (20× magnification) using antibodies directed against Jagged2, Jagged1, DLL1, or MITF (red). DAPI-stained nuclei appear blue. MITF expression reveals melanocytes located at the basal epidermis. (C) Left panel: Experimental design. Right panel: WM3682 cells were transfected with a Notch signaling red fluorescent reporter and a GFP-expressing plasmid. Transfected cells were co-cultured with the indicated primary human cell types and treated with DAPT or vehicle control DMSO. Green signal indicates all melanoma cells; red is indicative of active Notch signaling. (D) WM3526 cells were co-cultured with the indicated primary human cell types and treated with DAPT or vehicle control DMSO before being subjected to an invasion assay. The ratio of invading cells after DAPT treatment to invading cells after vehicle treatment is presented. Error bars represent ± SEM, ∗ indicates p < 0.05 (n = 4). See also Figure S2. Molecular Cell 2015 59, 664-676DOI: (10.1016/j.molcel.2015.06.028) Copyright © 2015 Elsevier Inc. Terms and Conditions

Figure 3 Melanoma Patient Samples Exhibiting DLL1 and Notch Signaling Activation Associated with Melanoma Vertical Transition (A) Immunofluorescence analysis of a vertical cross-section from a single melanoma patient sample (20× magnification) indicates expression of DLL1 (green) and Melan-A (red). DAPI-stained nuclei appear blue. Melan-A indicates normal melanocytes located in the basal epidermis and melanoma cells at various disease stages. Lower panels are higher magnifications of the images above. Each image was generated by merging of serial images (see Experimental Procedures). Arrows indicate the distance between melanoma cells and the DLL1-expressing cells. (B) Immunofluorescence analysis of a vertical cross-section from a single melanoma patient sample indicates expression of NICD (green) and Melan-A (red). DAPI-stained nuclei appear blue. NICD indicates Notch activation and Melan-A indicates normal melanocytes located at the basal epidermis and melanoma cells at various disease stages. See also Figure S3. Molecular Cell 2015 59, 664-676DOI: (10.1016/j.molcel.2015.06.028) Copyright © 2015 Elsevier Inc. Terms and Conditions

Figure 4 Increased miR-222/221 Expression in Melanoma Cells upon Interaction with Differentiated Keratinocytes Promotes Invasion (A) miRNA small-RNA-seq analysis: Venn diagram showing the overlap (n = 2) between miRNAs differentially expressed in melanoma upon culturing on DLL1 coated plate, miRNAs previously shown (Ozsolak et al., 2007) to be bound at their promotor by MITF, and miRNAs shown to promote melanoma invasion (Levy et al., 2010). ∗p = 3.26e-6. (B) Left panel: Experimental design. Right panel: WM3682 cells were transfected with a luciferase reporter driven by the miR-222/221 promoter. Transfected cells were co-cultured with the indicated primary human cell types and luciferase activity was measured. Error bars represent ± SEM, ∗ indicates p < 0.05 (n = 4). (C) WM3682 cells were stably transfected with a vector expressing miR-222 or scrambled miRNA (control) and injected subcutaneously into immunodeficient mice. One representative image of H&E staining of lung tissue is shown; arrow indicates metastases. (D) WM3314 and WM1716 cells were transfected with miR-222 antagomiR (anti222), miR-221 antagomiR (anti221), both antagomiRs (anti222/221), or antagomiR control; WM3682 and WM3526 cells were transfected with miR-222 mimic, miR-221 mimic, both miR-222/221 mimics, or scrambled control, and an invasion assay was performed. Error bars represent ± SEM, ∗ indicates p < 0.05 (n = 4). (E) Upper panels: miR-222 expression levels in normal human skin, benign nevi, superficial-spreading melanomas, and invasive melanoma were detected by in situ hybridization (green). DAPI-stained nuclei appear blue. Lower panels: H&E staining of same magnification and location. White dashed lines represent epidermal-dermal junction. (F) WM3526 cells were transfected with anti-miR-222/221, siMITF, anti-miR-222/221 plus siMITF, or scrambled siRNA before being subjected to an invasion assay. Error bars represent ± SEM, ∗ indicates p < 0.05 (n = 4). See also Figure S4. Molecular Cell 2015 59, 664-676DOI: (10.1016/j.molcel.2015.06.028) Copyright © 2015 Elsevier Inc. Terms and Conditions

Figure 5 MITF Blocks Melanoma Invasion by Repressing miR-222/221 Expression (A) EMSA was conducted using a probe corresponding to the E-Box region of the human miR-222/221 promoter. WM3682 nuclear extracts were used as the source of MITF. Biotinylated E-Box probe (WT) or E-Box point-mutated probe (Mut probe) were used in competition analyses. Bands corresponding to MITF binding and free probe are marked with arrows. (B) WM3682 cells were transfected with a luciferase reporter driven by a WT or mutated E-Box miR-222/221 promoter (mutations marked in red) or an empty promoter (control). Luciferase activity was normalized to Renilla luciferase. Fold changes relative to control are shown; error bars represent ± SEM (n = 5). (C) ChIP analysis was performed on extracts from WM3682 cells transfected with siMITF or a scrambled siRNA. Protein-chromatin crosslinked complexes were precipitated with the indicated antibodies or with control antibody (cAb). PCR primers spanning the region encoding the miR-222/221 promoter (lower panel) were used. Promoter occupancy is represented as relative to control (cAb). Error bars represent ± SEM, ∗ indicates p < 0.05 (n = 4). (D) WM3682 and 451LU cells were transfected with siMITF or a scrambled siRNA (control); WM1716 and WM3314 cells were transfected with MITF cDNA or empty plasmid (control). Levels of MITF mRNA were normalized to levels of actin; miR-222/221 and miR-211 levels were normalized to RNU48. Expression is represented as relative to levels in control treated cells. Error bars represent ± SEM, ∗ indicates p < 0.05 (n = 4). (E) Pearson’s correlation for the expression levels of each miRNA relative to MITF in 11 melanoma cell lines. miR-211 and miR-222/221 are highlighted (correlations of p < 0.05). (F) MITF protein levels and mature miR-221/222 expression levels in five melanomas; β-tubulin was used as a loading control. PCR data were normalized to RNU48. Error bars represent ± SEM (n = 5). See also Figure S5. Molecular Cell 2015 59, 664-676DOI: (10.1016/j.molcel.2015.06.028) Copyright © 2015 Elsevier Inc. Terms and Conditions

Figure 6 MITF Cooperates with RBPJK to Repress Transcription of miR-222/221 in Melanoma Cells (A) WM3682 and WM3314 cells were transfected with siRBPJK or RBPJK cDNA and the appropriate control (scrambled siRNA or empty plasmid, respectively). qRT-PCR was performed to determine the levels of mature miR-221 and miR-222. Data were normalized to RNU48. Error bars represent ± SEM, ∗ indicates p < 0.05 (n = 4). (B) ChIP analysis was performed on extracts from WM3682 and WM3314 cells transfected with siRBPJK or siControl. Protein-chromatin crosslinked complexes were precipitated with H3K27me antibody or H3K4me antibody. PCR primers spanning the miR-222/221 promoter were used. Error bars represent ± SEM, ∗ indicates p < 0.05 (n = 4). (C) Protein-chromatin crosslinked complexes from the samples described in (B) were precipitated with Pol II antibody. Several PCR primers along the miR-222/221 locus were used (primer position correlates with schematic of locus illustrated below graph). The actin promoter and gene served as a control (lower panels). Error bars represent ± SEM, ∗ indicates p < 0.05 (n = 4). (D) Upper panel: Schematic of miR-222/221 promoter reporter plasmids: WT, mutated in E-Box, or mutated in R-Box. Mutations are indicated in red; numbers indicate the nucleotide position. Lower panels: WM3682 cells were transfected with a luciferase reporter driven by either a WT or mutated E-Box miR-222/221 promoter and with RBPJK cDNA (green bars) or empty plasmid (control; black bars). WM3314 cells were transfected with a luciferase reporter driven by either a WT or mutated R-Box miR-222/221 promoter and with either MITF cDNA (blue bars) or empty plasmid (control; black bars). Luciferase activity was normalized to Renilla luciferase activity. Fold changes relative to control are shown. Error bars represent ± SEM, ∗ indicates p < 0.05 (t test, n = 5). In parallel, western blot analysis was conducted on harvested cells to monitor the indicated proteins. β-tubulin served as a loading control. (E) ChIP analysis was performed on extracts from WM3682 cells transfected with siRBPJK or siControl. Protein-chromatin crosslinked complexes were precipitated with H3K27me3, KDM5A, or RBPJK antibodies. PCR primers spanning the miR-222/221 promoter were used. Error bars represent ± SEM (n = 3). (F) WM3314 cells were transfected with a luciferase reporter driven by the miR-222/221 promoter and with either MITF cDNA or empty plasmid (control), and in addition either KDM5A cDNA or KDM5A (H483A) cDNA. Luciferase activity was normalized to Renilla luciferase activity. Fold changes relative to control are shown. Error bars represent ± SEM; the ∗ indicates p < 0.05 (n = 5). In parallel, western blot analysis was conducted on harvested cells to monitor MITF and KDM5A protein levels. β-tubulin served as a loading control. See also Figure S6. Molecular Cell 2015 59, 664-676DOI: (10.1016/j.molcel.2015.06.028) Copyright © 2015 Elsevier Inc. Terms and Conditions

Figure 7 Notch Signaling Activation Reduces MITF Occupancy of miR-222/221 Promoter and Promotes Melanoma Invasion (A) WM3314 cells were transfected with a luciferase reporter driven by the miR-222/221 promoter and with either MITF or NICD cDNA or empty plasmid. Luciferase activity was normalized to Renilla luciferase activity. Fold changes relative to control are shown. Error bars represent ± SEM, ∗ indicates p < 0.05 (n = 4). (B) WM3314 cells were transfected with MITF or NICD cDNA or empty plasmid and subjected to EMSA analysis. Nuclear extracts were incubated with a probe containing the MITF binding site sequence found in the miR-222/221 promoter. (C) Left panel: Experimental design. Right panel: WM3526 and WM3682 cells were seeded on DLL1-coated plates and treated with DAPT or vehicle control DMSO. qRT-PCR was performed to determine the levels of pre- and mature miR-221/222. Data were normalized to actin or RNU48, respectively. Error bars represent ± SEM, ∗ indicates p < 0.05 (n = 4). (D) Pearson’s correlation for the mRNA expression levels of each gene relative to MITF. mRNA levels were calculated using the microarray profiles of 88 melanoma cell lines (Lin et al., 2008). MITF and MITF-related genes are marked in green; Notch and Notch-related genes are marked in red. (E) WM3526 and WM3682 cells were transfected with anti222/221 or antagomiR control, treated with DAPT or DMSO, and seeded on DLL1-coated plates. An invasion assay was then performed. Error bars represent ± SEM, ∗ indicates p < 0.05 (n = 4). (F) A Venn diagram describing the overlap between miR-222/221 potential targets, genes with expression levels that significantly positively correlate (p < 0.05) with MITF levels, and genes identified as involved in invasion. (G) WM3314 cells were transfected with a luciferase reporter possessing either WT or mutant 3′ UTRs of GRB10 (upper panel) or ESR1 (lower panel) and with either MITF cDNA, NICD cDNA, or empty plasmid. Error bars represent ± SEM, the ∗ indicates p < 0.05 (t test, n = 5). Schematic shows predicted miR-222/221 target sites identified in GRB10 and ESR1 3′ UTRs. Numbers indicate positions of miR-222 binding sites. (H) Average gene expression of melanoma cells exhibiting radial versus vertical growth in clinical samples was calculated for the indicated genes. The fold change between vertical and radial growth is presented in the graph. Notch-related genes are marked in black and miR-222/221 target genes are indicated by gray. See also Figure S7. Molecular Cell 2015 59, 664-676DOI: (10.1016/j.molcel.2015.06.028) Copyright © 2015 Elsevier Inc. Terms and Conditions