Fig 2: NGFRhi melanomas are resistant to multiple therapies.a Colony formation assay on IFN? treatment on D10 and D10-TR cells. Tumor cells were treated for 7 days with IFN? (which was refreshed on day 4) and stained with crystal violet. Quantification in b. b Quantification of IFN? and TNF treatment for 7 days on D10 and D10-TR cells (medium was refreshed on day 4). Error bars represent S.D. of three independent replicates. Statistical analysis by unpaired t-test; *p < 0.05, ***p < 0.001. c Cytotoxicity assay in D10 parental versus TR cells for BRAF + MEKi. Titrations were performed with a 10:1 ratio of BRAFi:MEKi concentration. Error bars represent S.D. of three biological replicates; the experiment was performed in four independent replicates (as quantified in d). d Quantification of three cell lines for BRAF/MEKi sensitivity in parental versus TR lines. BRAF inhibitor dabrafenib was given at 10 nM and MEK inhibitor trametinib at 1 nM. Error bars represent S.D. of four independent replicates. Statistical analysis by Mann–Whitney test; *p < 0.05. e Cell count in parental and TR cells of the D10 cell line, treated continuously with 1000 nM of dabrafenib. Error bars represent S.D. of two independent replicates. Statistical analysis by Mann–Whitney test; *p < 0.05. f Average NGFR expression in PDX samples from melanoma. Error bars represent S.D. Statistical analysis by unpaired t-test. *p < 0.05. g NGFR immunohistochemistry score for patient samples prior to and on BRAF + MEK inhibition, divided in responders (R) and nonresponders (NR). Statistical analysis by Mann–Whitney, **p < 0.01. Source data are provided as a Source Data file.

Fig 3: Genetic perturbation of NGFR modulates T cell sensitivity.a Western blot analysis of indicated cell lines transduced with a control hairpin or one of two short hairpins targeting NGFR. Vinculin was used as a loading control. One blot of two independent replicates is shown (the other can be found in Source data). b Quantification of MART-1 T cell cytotoxicity in TR cell lines relative to Ctrl T cells (1:1 ratio tumor cell: T cell) in the presence of absence of shNGFR. Cells were loaded with 100nM MART-1 peptide prior to the assay, and cytotoxicity was normalized to shCtrl cell survival after MART-1 T cell attack. Error bars represent S.D. of four independent experiments with two technical replicates. Statistical analysis by Kruskal–Wallis test; *p < 0.05, ****p < 0.001. c Western blot analysis on D10 and M009R.X1.CL cells transduced with a control ORF or an NGFR-RFP ORF. GAPDH was used as a loading control. One blot of two independent replicates is shown (the other can be found in Source data). d Quantification of cell survival by flow cytometry after T cell attack in vitro in control or NGFR-ORF cells. Error bars represent S.D. of four independent experiments. Statistical analysis by Mann–Whitney, *p < 0.05. e mRNA expression of BDNF in parental and TR cells. Error bars represent S.D. of pooled analysis of three independent replicates. Analysis by unpaired t-test; *p < 0.05. f Quantification of cytotoxicity in TR cell lines relative to shCtrl cells after T cell attack (1:1 ratio tumor : T cell) in the presence or absence of shBDNF. Error bars represent S.D. of six independent experiments. Statistical analysis by Kruskal–Wallis test; *p < 0.05, **p < 0.01. Source data are provided as a Source Data file.

Fig 4: NGFRhi melanoma fractions show immune exclusion in patients.a Examples of two melanomas from patients showing either high or low NGFR expression, and the corresponding CD3 and CD8 stainings for the same regions. Scale bars indicate 50 µm. Quantification in b, c. b, c CD3 and CD8 scores for NGFRneg vs. NGFRpos samples. Only tumor-infiltrating lymphocytes were scored and compared with the relative abundance of tumor cells. Error bars represent S.D. Statistical analysis by Mann–Whitney test; *p < 0.05. d Examples of two melanomas from patients that are heterogeneous for NGFR expression within each sample, and the corresponding CD3 stainings for the same regions (indicated by dotted lines). Scale bars indicate 100 µm. e MCP counter analysis on melanoma cohort of TCGA divided in top and bottom 25% NGFR-expressing tumors. Gene sets for “T cells” and “cytotoxic lymphocytes” are plotted. Error bars represent S.D. Statistical analysis by unpaired t-test. Source data are provided as a Source Data file.

Fig 5: Single-cell analysis uncovers heterogeneities in melanoma differentiation, proliferation, and MAPK signaling states across a wide range of Braf/Mek inhibitor sensitivity.a, b Single-cell protein levels of three melanoma differentiation state markers, Mitf, Ngfr, and Axl, measured by multiplexed immunofluorescence microscopy and overlaid on t-SNE plots. Cells were exposed to either vehicle (DMSO), Braf inhibitor (vemurafenib at 100 nM), or the combination of Braf and Mek inhibitors (vemurafenib at 100 nM and trametinib at 10 nM) for 3 and 5 days. Cells for each experimental condition were randomly selected from the pool of two biologically independent replicates. Single-cell t-SNE maps for all cell lines combined (a) and projections of variation within each individual cell line (b) are shown. c, d Single-cell protein levels of p-ErkT202/Y204, p-S6S235/S236, and Ki67, measured by multiplexed immunofluorescence microscopy and overlaid on t-SNE plots. Treatment conditions are the same as in (a, b). Single-cell t-SNE maps for all cell lines combined (c) and projections of variation within each individual cell line (d) are shown. e Average net growth rates calculated from measurements of live cell count (across at least two biologically independent replicates) at three timepoints (including 0, 3, and 5 days) following exposure to DMSO, vemurafenib (at 100 nM) or vemurafenib (at 100 nM) plus trametinib (at 10 nM). f Drug-induced normalized growth rates (a.k.a. DIP rates) calculated by dividing the average net growth rate for drug-treated cells to that for DMSO-treated cells in each cell line. Normalized growth rates <0 indicate a net cell loss (i.e., drug-induced cytotoxicity), a value of 0 represent no change in viable cell number (i.e., cytostasis), a value >0 indicates a net cell gain, and a value of 1 represents no drug effect as cells grow at the same rate as in the DMSO condition. g Two-sided Pearson’s correlation analysis between the fraction of p-ErkLow/p-S6Low/Ki67Low cells (referred to as fully inhibited cells) and drug-induced normalized growth rate (top left), the fraction of undifferentiated (NgfrLow/AxlHigh) cells (top right), the fraction of neural crest-like (NgfrHigh/AxlLow) cells (bottom left), and the fraction of melanocytic (MitfHigh) cells (bottom right) across 16 melanoma cell lines treated with Braf/Mek inhibitors for 3 and 5 days. Source data are provided as a Source data file.