Netrin-1 and Its Receptor DCC Are Causally Implicated in Melanoma Progression

Vogelstein and colleagues identified deleted in colorectal cancer (DCC) as a gene frequently silenced in cancer because of LOH on chromosome 18q. DCC expression is markedly reduced in more than 50% of colorectal tumors as well as in many other neoplasms (1). Moreover, loss of DCC is associated with poor prognosis and potentially decreased response to adjuvant chemotherapy in patients with colorectal cancer. Finally, restoration of DCC expression can suppress tumorigenic growth properties in vitro and in nude mice (1). These findings led to the proposal that DCC expression is a constraint for tumor progression, and thus DCC functions as a tumor suppressor gene. However, the localization of the DCC gene close to well-established tumor suppressors such as Smad4 (2) and the absence of increased tumor susceptibility in a mouse model, in which DCC was inactivated (3), created skepticism about its potential role as a tumor suppressor gene (4). However the more recent use of conditional DCC–mutant mouse model have demonstrated that DCC inactivation is associated with increased tumoral progression (5).

DCC, a large 200 kDa transmembrane receptor, belongs to the family of dependence receptors (6). Such receptors induce apoptosis when their trophic ligands are absent, thus conferring a state of cellular dependence on ligand availability for survival (7), and this has been proposed to confer a tumor suppressive activity to these dependence receptors (7, 8). To address this hypothesis, we generated a mouse model in which the DCC locus is point-mutated in Asp1290 to specifically silence the prodeath activity of the receptor without impacting DCC non-cell death signaling (9). These mice spontaneously develop more colorectal cancers and lymphomas than age-matched control animals and are more prone to develop metastatic intestinal adenocarcinoma when backcrossed in an APC-mutant background (9, 10).

Of interest, while loss of DCC has been extensively described in cancer, studies have emerged showing that the ligand of DCC, netrin-1, is also implicated in cancer progression. Netrin-1 was initially discovered as a navigation cue during the development of the nervous system and a key factor for brain wiring (11). Although initially described as a diffusible cue guiding, via long distance processes, different types of axons (12), netrin-1 is currently regarded as a laminin-related sticky protein acting at short distance (13) via its main receptors, DCC and UNC5-Homolog (UNC5H- i.e., UNC5A, UNC5B, UNC5C, and UNC5D; ref. 14). Indeed, netrin-1 appears to be a multifunctional secreted protein that plays key roles in neuronal navigation, angiogenesis, and cell survival (15), and as a consequence, has been associated with numerous diseases including diabetes, cardiovascular diseases, and cancers (15). In the latter case, netrin-1 is often reported to be upregulated, and this upregulation was proposed to act as a selective mechanism blocking apoptosis induced by the dependence receptors DCC and UNC5H (15). Consistently, netrin-1 expression in brain metastases is an independent poor prognostic factor for patient survival (16). Moreover, netrin-1 detection in blood or urine has been proposed to be a predictive marker for cancers (17, 18). Netrin-1 upregulation has been reported not only in cancer cells but also in cancer-associated fibroblasts (19). Efforts to develop drugs that inhibit the interaction of netrin-1 with its receptors are therefore ongoing. Several preclinical proof-of-concept studies have shown that candidate drugs interfering with netrin-1-receptor interactions, either used alone or in combination with conventional chemotherapies or epidrugs, markedly inhibit tumor growth and metastasis development (18, 20, 21). A humanized anti-netrin-1 mAb, also called NP137, has been developed previously (20) and the interim results of a phase I clinical trial have recently been reported (22), showing promising signs of antitumor activity in monotherapy in patients with advanced solid tumors.

While the DCC gene was initially thought to be rarely mutated in cancers, Krauthammer and colleagues identified the DCC gene as the third most frequently mutated gene in sun-exposed melanoma (23). On the basis of this observation and as melanoma still represent a significant clinical challenge (24–26), we investigated the role of the netrin-1/DCC pair in this cancer. We reveal that the netrin-1/DCC pair is causally implicated in disease progression and propose that netrin-1 interference could be an attractive therapeutic perspective for treating melanoma.

A total of 241 patients underwent surgery for skin melanocytic lesions between January 2007 and January 2011 and had enough tissue sample remaining after the diagnosis had been established. According to the French Bioethics law of August 2004, the patients were informed of the research use of the remaining samples and were not opposed to it. The research protocol was approved by the Ethics committee of Saint Louis Hospital. All 241 patients signed a written informed consent (IRB 00006477).

The sample series included (i) 36 benign melanocytic lesions: 30 common melanocytic nevi (20 intradermal type, five compound type, and five junctional type) and six dysplastic nevi (four compound type and two junctional type); (ii) 205 malignant melanoma including 25 in situ melanoma, 40 melanoma with a Breslow tumor thickness of less than 1 mm, 40 melanoma with a Breslow tumor thickness between 1 and 2 mm, 50 melanoma with a Breslow tumor thickness between 2 and 4 mm, and 50 melanoma with a Breslow tumor thickness more than to 4 mm.

Patient follow-up over 2–5 years showed that, at the time of the study, in the 36 patients with benign melanocytic lesions there was no metastasis, whereas in the 205 patients with malignant melanoma, 125 had metastases either in the lymph node or in viscera.

An indirect immunoperoxidase method was performed on 5-μm thick tissue sections. A polyclonal goat anti-human netrin-1antibody (AF1109, R&D Systems) was used as primary antibody. A donkey anti-goat biotinylated antibody (ab97108 Abcam) was used as secondary antibody. A DABMap Detection Kit (Roche Diagnostic) was used for detection of the biotinylated antibody. Systematic controls were performed for each slide, including: (i) the absence of primary antibody, and (ii) the use as primary antibody of an irrelevant antibody of the same isotype as the primary antibody.

A Provis AX70 microscope (Olympus) was used. With a wide-field eyepiece number 26.5, it provided a field size of 0.344 mm² at 400 × magnification. Two pathologists (A. Janin and M. Battistella) independently analyzed netrin-1 expression on five different fields at magnification 400 × for each tissue section. To assess the number of positively stained cells, they used a semi-quantitative scoring system ranging from 1 to 4: 1:0–25%; 2: 25–50%; 3: 50–75%; and 4: >75%. Staining intensity was recorded using a similar semi-quantitative approach as follows: 1: weak staining; 2: moderate staining; and 3: strong staining.

The 23 melanoma cell lines and a normal melanocyte cell line (FM516SV) were obtained from the ATTC and cultured in DMEM1X + GlutaMAX + 4.5 g/L d-Glucose + Pyruvate Medium (Gibco by Life Technologies) containing 10% FBS. Routine Mycoplasma testing was performed by MycoAlert Mycoplasma Detection Kit (catalog no. LT07-118). Melanoma cells grown for no more than 20 passages were used in all the experiments.

To analyze the expression of netrin-1, total RNAs of the different melanoma cell lines and of the normal melanocyte cell line (FM516SV) were extracted using the Nucleospin RNAII Kit (Macherey-Nagel). RNA (1 μg) was reverse transcribed using the iScript cDNA Synthesis Kit (Bio-Rad). Expression was assessed by real-time quantitative RT-PCR with specific primers available upon request, on a LightCycler 480 Apparatus (Roche) using the LightCycler TaqMan Master Kit (Roche). Reaction conditions for optimal amplification of each gene, as well as primer sequences are available upon request. The ubiquitously expressed HPRT was used as an internal calibrator. In addition, total RNAs from xenograft tumors were extracted by disrupting tissues with MagNA Lyser Instrument (Roche Applied Science). RT-PCR reactions were performed with PrimeScript RT Reagent Kit (Takara). Five-hundred nanogram total RNA was reverse transcribed using the following conditions: 37°C for 15 minutes and 85°C for 5 seconds. For expression studies, the target transcripts were amplified on a LightCycler2.0 Apparatus (Roche Applied Science), using the Premix Ex Taq Probe qPCR Kit (Takara), according to the manufacturer's instructions. Expression of target genes (DCC) was normalized against glucuronidase beta (GUSB) and TATA-binding protein (TBP), used as housekeeping genes. The amount of target transcripts, normalized against the housekeeping gene, was calculated using the comparative C_t method.

For the generation of inducible stable cell lines, HA-tagged human DCC coding sequence was cloned at NheI/AgeI restriction sites in the pCW57.1-hygro plasmid, a self-inactivating HIV-1–derived vector allowing the doxycycline inducible expression of DCC, as well as the constitutive expression of rtTA protein and hygromycin resistance gene, expressed under the control of the mouse phosphoglycerate kinase (PGK) promoter. Cloning of HA-tagged DCC was performed using Infusion Cloning Kit (Clontech). pCW57.1-hygro plasmid was derived from pCW57.1 (Addgene plasmid #41393) by exchanging puromycin to hygromycin resistance. Stable cell clones were selected for 5 days in hygromycin-containing medium and inducible cell lines were further tested for DCC expression in response to doxycycline by Western blot analysis.

For immunoblotting analysis of DCC, inducible cell lines were treated with doxycycline for 24 hours in a medium containing 10% serum. Cells were then lysed in SDS buffer (10 mmol/L Tris-HCl, 10% glycerol, 5% SDS, 1% TX100, and 0.1 mol/L DTT) in the presence of protease inhibitor mixture (Roche Applied Science) and phosphatase inhibitor (Sigma). Lysates were then sonicated and incubated for 10 minutes at 95°C, and protein extracts (20 μg per lane) were loaded onto 4%–15% SDS-polyacrylamide gels (Bio-Rad) and blotted onto nitrocellulose membranes using Trans-Blot Turbo Transfer System (Bio-Rad). Filters were blocked with 10% non-fat dried milk in PBS/0.1% Tween 20 (PBS-T) for 2 hours and then incubated over night with a rabbit polyclonal α-HA (1:1,000, Sigma) and a rabbit polyclonal α-GAPDH (1:4,000, Santa Cruz Biotechnology). After three washes with PBS-T, filters were incubated with an α-rabbit HRP-conjugated secondary antibody (1:10,000, Jackson ImmunoResearch) for 1 hour. Detection was performed using West Dura Chemiluminescence System (Pierce). Membranes were imaged on the ChemiDoc Touch Imaging System (Bio-Rad).

Five-week-old (20–22 g body weight) female SCID mice were obtained from Charles River animal facility. The mice were housed in sterilized filter-topped cages and maintained in a pathogen-free animal facility. Melanoma cell lines expressing netrin-1 (M2Ge, Skmel5, WM793, Skmel3, XPC, and M4Be) were implanted by subcutaneous injection of 5 × 10⁶ cells in 100 μL of growth factor–reduced Matrigel (Corning) diluted in 100 μL of PBS into the right flank of the mice. When the tumor reached between 150 and 200 mm³, 100 mg/kg of dacarbazine was injected intraperitoneally once a week for 4 weeks and with the NET1-H-mAb every 2 days at 20 mg/kg during the 4 weeks. In addition, stable melanoma cell lines for DCC (Mewo and WM239A) were implanted by subcutaneous injection of 5 × 10⁶ cells in 100 μL of growth factor–reduced Matrigel (Corning) diluted in 100 μL of PBS into the right flank of the mice. When tumors reached 100 mm³ in approximately 1 week after injection, 1 mg of doxycycline in 200 μL or an equal volume of PPI water for the control group was injected intraperitoneally each day for 10 days. For each xenograft experiment, tumor sizes were measured twice a week with a caliper. The tumor volume was calculated with the formula v = 0.5 x (l x w²), where v is volume, l is length, and w is width. All experiments were performed in accordance with relevant guidelines and regulations of animal ethics committee (Authorization no CLB-2015-004; accreditation of laboratory animal care by CECCAP, ENS Lyon-PBES).

IHC staining was performed on an automated Immunostainer (Ventana Discovery XT) using a DABmap kit and REDmap kit according to the manufacturer's instructions. Tissue samples were fixed in 10% buffered formalin and embedded in paraffin. After antigen unmasking [EDTA buffer (pH 8.4), 98°C for 35 minutes], apoptotic cell staining was performed using a rabbit cleaved caspase-3 antibody (clone 5A1E, 9664L, Cell Signaling Technology) and the rabbit cleaved Parp (clone D64E10, 6525, Cell Signaling Technology). Cell proliferation was determined using a Ki67 (clone SP6, F/RM-9106-5, Neomarkers). The slides were scanned by an Axio-Scan Zeiss microscope. Image quantification was conducted using the software Zen blue, by two blinded analyzers including a pathologist.

The DCC^mut model was described previously (9). For the tgNETRIN-1 mouse line (in C57BL/6 background), Human NETRIN-1 was cloned into the Rosa26-lox-stop-lox plasmid. The mice were generated by SEAT CNRS Gustave Roussy Phenomin. These two models were intercrossed with the inducible model of predisposition to melanoma (TyrCreER^T2:BRAF^CA/+). CreER^T2 expression and local melanomagenesis induction were activated by applying 200 μL of 2 mg/mL solution of tamoxifen (70% Z-isomer, Sigma) diluted in 100% ethanol on the back skin of 6–7 weeks old mice for 4 consecutive days. The mice were monitored twice a week for a maximum of 1 year. The mutant or wild-type animals for DCC (respectively, overexpressing netrin-1 or not) on this background BRAF^V600E were followed to define the impact of the abolition of apoptosis induced by receptor mutation or ligand gain on the incidence and tumor progression during animal aging. Melanoma tumors as well as different organs were recovered to carry out histologic studies and search for the presence of metastasis. All experiments were performed in accordance with relevant guidelines and regulations of animal ethics committee (Authorization no. CLB-2014- 010; accreditation of laboratory animal care by CECCAPP, ENS Lyon-PBES).

As DCC is frequently mutated in melanoma (23) or not expressed by melanoma cancer cell lines (27), we first conditionally reexpressed DCC in the amelanotic MeWo DCC-negative melanoma cell line using a sleeping beauty system (Fig. 1A). On the basis of this approach, doxycycline treatment of MeWo cells induced DCC expression (Fig. 1A). These doxycycline-dependent DCC-expressing MeWo cells were then engrafted in SCID mice, which were treated daily with doxycycline once tumors had reached 100 mm³. As shown in Fig. 1B–G, intratumoral reexpression of DCC (Fig. 1E) was associated with tumor growth inhibition (Fig. 1B,–D), increased tumor cell death (cCaspase-3, cParp), and decreased cell proliferation (Ki67; Fig. 1F and G). Similar effects were observed when DCC was conditionally reexpressed in the amelanotic WM239A melanoma cell line (Supplementary Fig. S1A–S1E). Altogether, these results suggest that DCC expression is sufficient to prevent tumor growth in vivo.

To further demonstrate the implication of the prodeath activity of DCC in the regulation of melanoma progression, we used the recently developed DCC-mutant mouse model, in which the endogenous DCC gene locus is point mutated to convert Asp1290 into Asn1290 to specifically inactivate the prodeath activity of DCC, hereinafter referred to as DCC^mut (9, 10). It has been shown that in this mouse model, the positive signaling induced by netrin-1 binding to DCC is not affected, while DCC is unable to trigger cell death (9). We intercrossed mice harboring two DCC^mut alleles with the classical murine melanoma model TyrCreER^T2:BRAF^CA/+ (28, 29), as almost 60% of human melanomas show somatic mutations that constitutionally activate serine/threonine-protein kinase BRAF (BRAF^V600E). The TyrCreER^T2:BRAF^CA/+ murine model, which conditionally expresses the dominant oncogenic form BRAF^V600E of the BRAF protein, was shown to be prone to develop melanoma neoplasia during ageing (28). TyrCreER^T2:BRAF^CA/+ and TyrCreER^T2:BRAF^CA/+;DCC^mut/mut mice were then treated with tamoxifen and are hereafter referred to as BRAF^V600E and BRAF^V600E;DCC^mut, respectively (Fig. 2A). Melanoma development was monitored up to 12 months after tamoxifen skin treatment. Upon treatment with tamoxifen, around 20% of the BRAF^V600E mice developed cutaneous lesions identified as melanoma by pathologists (Fig. 2B–D; Supplementary Fig. S2). Remarkably and as shown in Fig. 2D, DCC mutation, confirmed at different tumor locations (Fig. 2C), was associated with a significantly increased propensity to develop melanoma, because 71.4% of BRAF^V600E;DCC^mut mice developed melanoma lesions within the first 12 months of tamoxifen application (Fig. 2D). This increase in propensity to develop melanoma was moreover associated in 60% of these mice (Supplementary Table S1), with histologic features associated with increased aggressiveness, including increased cellularity within the tumor together with more frequent mitoses, a denser collagenous stroma (Fig. 2E, c and d), and acanthosis and ulceration of the overlying epidermis (Fig. 2E, a and b). A detailed pathology report is presented in Supplementary Table S2 for each BRAF^V600E; DCC^mut tumors. Collectively these results demonstrate that DCC constitutes a constraint for melanoma development through its prodeath activity.

Having shown that reexpression of DCC is associated with tumor growth inhibition and that mutation of the death activity of DCC is associated with melanoma progression, we hypothesized that in the human pathology, the progressing melanoma should have negatively selected DCC-induced cell death. One mechanism to silence DCC-induced tumor cell death is the upregulation of netrin-1 (10). We thus analyzed netrin-1 expression by IHC on a panel of melanoma, benign nevus, or of lymph node metastasis. As shown in Fig. 3A–C; Supplementary Table S3, a netrin-1–specific staining is clearly visible in melanoma and metastatic lesions, and was seen in microvascular endothelial cells (black arrows) and in tumor cells. When defining a score for the fraction of cancer cells expressing netrin-1, a high score (75%–100%) was mostly detected in malignant melanoma (Fig. 3B) compared with benign melanocytic lesions, and this trend toward an increase in netrin-1 expression associated with progression was further observed when melanoma with metastasis at time of study was compared with melanoma without metastasis at time of study (Supplementary Fig. S3A). Similar observations were made when the intensity of the netrin-1 signal in cancer cells was analyzed (Fig. 3C; Supplementary Fig. S3B; Supplementary Table S4). These data thus support the view that netrin-1 is upregulated in aggressive melanoma in the human pathology.

Next, we investigated whether forced expression of netrin-1 in the skin of mice was associated with increased melanoma progression. To address this, we generated a mouse model (tgNETRIN-1) in which a loxP-stop-loxP-human netrin-1 cassette was inserted in the endogenous Rosa26 locus. Ectopic netrin-1 expression could then be triggered by Cre recombination (Fig. 4A, top). We intercrossed tgNETRIN-1 mice with TyrCreER^T2:BRAF^CA/+ mice (hereinafter referred to as TyrCreER^T2:BRAF^CA/⁺;tgNETRIN-1), and cutaneously treated TyrCreER^T2:BRAF^CA/+ and TyrCreER^T2:BRAF^CA/+;tgNETRIN-1 mice with tamoxifen, hereafter referred to as BRAF^V600E and BRAF^V600E;tgNETRIN-1, respectively (Fig. 4A and B). As shown in Fig. 4C, forced expression of netrin-1 in BRAF^V600E;tgNETRIN-1 mice was associated with a drastic increase in the frequency of primary melanoma compared with BRAF^V600E mice. No metastases were observed at necropsy. This increase propensity to develop melanoma was associated in 43.75% of these mice (Supplementary Table S5), with increase in aggressiveness markers, such as cellularity and a higher number of mitoses per high power field within the tumor (Fig. 4D, c and d). In addition, the skin of 50% of these mice was thickened (acanthosis) and ulcerated (Fig. 4D, a and b). A detailed pathology report is presented in Supplementary Table S6 for each BRAF^V600E; tgNETRIN-1 tumors. In-line with the proposed mechanisms, when analyzing cell death in the corresponding tumors using active caspase-3 staining, we observed that tumors from the BRAF/tg-netrin-1 background displayed fewer apoptotic cells compared with the BRAF background (Supplementary Fig. S4A and S4B). Taken together, these data demonstrate that the netrin-1 overexpression seen in human melanoma is sufficient per se to trigger melanoma progression in mice.

Because the above data were showing that netrin-1 is causally implicated in melanoma progression and because a therapeutic anti-netrin-1 mAb is currently tested in early clinical trial, we investigated whether in melanoma preclinical models we could detect a tumor growth inhibiting effect of the anti-netrin-1 mAb. We first screened for netrin-1 expression in a panel of melanoma cell lines (general information on the genotype of these cell lines is well detailed in Supplementary Table S7). In agreement with the netrin-1 staining observed in human melanoma, a sizeable fraction of melanoma cell lines expressed netrin-1 both at the mRNA (Fig. 5A) and protein levels (Fig. 5B). Skmel5 and M2Ge were further studied as they displayed the strongest netrin-1 expression. These cells were engrafted into immune-deficient mice, which were systemically treated with either the netrin-1 mAb alone (monotherapy) or in combination with dacarbazine once tumors reached 150 mm³, and the overall survival of mice was followed (Fig. 5C). The netrin-1 mAb was only modestly affecting tumor growth when used alone in agreement with what was previously reported in highly proliferative models (Supplementary Fig. S5A; refs. 20, 30). However, when combined with dacarbazine, the netrin-1 mAb significantly increased overall mouse survival in several models including Skml5-engrafted and M2Ge-engrafted mice (Fig. 5D and E; Supplementary Fig. S5B and S5C). Overall, these data support the hypothesis that targeting netrin-1 in netrin-1–expressing melanoma could constitute a complementary therapeutic strategy to conventional treatments.

While debates on the role of DCC as a tumor suppressor somehow dampened the interest of scientists for studying DCC in the field of oncology (3, 31), the demonstration that the conditional deletion of DCC in mice was associated with increased metastasis in p53-deficient mammary tumors (5) and that the silencing of DCC death activity was associated with intestinal cancer progression (9) and increased lymphoma development (10), has strengthened the view that DCC is constraining tumor progression by engaging cell death. Consistently, Halaban and colleagues elegantly established that DCC is the third most frequently mutated gene in sun-exposed melanoma, suggesting that DCC may also be a tumor suppressor in melanoma (23). Furthermore, DCC was also found as one of the genes identified as mutated in both cutaneous malignant melanoma and Parkinson disease (32). We provide here evidence that DCC expression is a constraint for melanoma cell growth in engrafted mice and we demonstrate that the silencing of the death activity of DCC in murine skin cells is associated with increased melanoma development in BRAF-mutant settings. Thus, even though it remains to be shown whether the mutations of DCC detected in human melanoma are loss-of-proapoptotic function mutations, the data obtained here and the frequency of DCC mutation strongly supports the view that DCC is a tumor suppressor in melanoma.

According to the dependence receptor hypothesis, cancer cells have to silence DCC-induced cell death to survive (7, 8) and this can be achieved through at least two mechanisms, the first being the silencing of the receptor or of the receptor prodeath activity by for example mutation as extensively described for DCC in colorectal cancer (9). The second known mechanism is the autocrine/paracrine expression of the ligand, in our case, netrin-1. Along this line it has been shown that netrin-1 is upregulated in a fraction of various cancers (10, 17, 18, 33) and that interference with netrin-1/netrin-1 receptor interaction is associated with tumor growth inhibition in several preclinical models (10, 18, 33, 34). We show that a similar mechanism occurs in melanoma, in which netrin-1 expression is detected both in a fraction of human melanoma cell lines and in a fraction of human melanoma samples. We also show that this netrin-1 expression promotes melanoma development in mice in a BRAF-mutant background. Of interest, an anti-netrin-1 mAb is currently in phase I of a clinical trial assessing all solid tumors (https://clinicaltrials.gov/ct2/show/NCT02977195) with reported signs of clinical efficacy (22), and we present evidence that in preclinical models, combining the netrin-1 mAb and dacarbazine enhances animal survival. Future work should also investigate whether combining netrin-1 interference with novel drugs, such as immune checkpoint inhibitors, recently developed and/or approved in the melanoma indication are also biologically relevant and deserve to be assessed in clinical combination trials.

M. Battistella has received speakers bureau honoraria from Bristol Myers Squibb. C. Lebbé reports receiving a commercial research grant from BMS and Roche, has received speakers bureau honoraria from BMS, MSD, Novartis, Amgen, and Roche, and has provided expert testimony for BMS, MSD, Novartis, Amgen, Roche, Avantis, Pierre-Fabre, Pfizer, and Incyte. P. Mehlen is an interim CEO (paid consultant) for Netris Pharma and has ownership interest (including patents) in Netris Pharma. No potential conflicts of interest were disclosed by the other authors.

Conception and design: A. Boussouar, L. Larue, P. Mehlen

Development of methodology: A. Boussouar, N. Gadot, L. Larue

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): A. Boussouar, A. Manceau, A. Paradisi, N. Gadot, J. Vial, L. Larue, M. Battistella, C. Leboeuf, C. Lebbé, A. Janin

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): A. Boussouar, A. Tortereau, A. Paradisi, M. Battistella, C. Lebbé, A. Janin

Writing, review, and/or revision of the manuscript: A. Boussouar, M. Battistella, C. Lebbé, A. Janin, P. Mehlen

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): A. Boussouar, D. Neves, P. Mehlen

Study supervision: A. Boussouar, P. Mehlen

We wish to thank Julie Caramel, Stéphane Dalle, and Thibault Voeltzel for materials and scientific advises. This work was supported by institutional grants from CNRS, University of Lyon, Centre Léon Bérard, and from the Ligue Contre le Cancer, INCA, Pair Melanoma, ANR, and ERC.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.