KIT suppresses BRAF^V600E-mutant melanoma by attenuating oncogenic RAS/MAPK signaling

Abstract

The receptor tyrosine kinase KIT promotes survival and migration of melanocytes during development, and excessive KIT activity hyperactivates the RAS/MAPK pathway and can drive formation of melanomas, most notably of rare melanomas that occur on volar and mucosal surfaces of the skin. The much larger fraction of melanomas that occur on sun-exposed skin is driven primarily by BRAF or NRAS activating mutations, but these melanomas exhibit a surprising loss of KIT expression, which raises the question of whether loss of KIT in these tumors facilitates tumorigenesis. To address this question, we introduced a kit(lf) mutation into a strain of Tg(mitfa:BRAF^V600E); p53(lf) melanoma-prone zebrafish. Melanoma onset was accelerated in kit(lf); Tg(mitfa:BRAF^V600E); p53(lf) fish. Tumors from kit(lf) animals were more invasive and had higher RAS/MAPK pathway activation. KIT knockdown also increased RAS/MAPK pathway activation in a BRAF^V600E-mutant human melanoma cell line. We found that pathway stimulation upstream of BRAF^V600E could paradoxically reduce signaling downstream of BRAF^V600E, and wild-type BRAF was necessary for this effect, suggesting that its activation can dampen oncogenic BRAF^V600E signaling. In vivo, expression of wild-type BRAF delayed melanoma onset, but only in a kit-dependent manner. Together, these results suggest that KIT can activate signaling through wild-type RAF proteins, thus interfering with oncogenic BRAF^V600E-driven melanoma formation.

Introduction

Key mutations that drive melanoma progression hyperactivate RAS/MAPK signaling (1,2). In rare cutaneous melanomas that occur on volar surfaces of skin or in mucosa, 15–25% of cases contain mutations or amplifications of the receptor tyrosine kinase KIT that drive tumorigenesis (3–5). Inappropriate activation of KIT leads to ligand-independent, constitutive downstream ERK and AKT signaling (6,7). KIT-mutant melanomas and gastrointestinal stromal tumors (GISTs), which possess similar KIT mutations, have shown clinical responses to imatinib and other tyrosine kinase inhibitors (8,9). Whereas volar and mucosal melanomas represent about 2% of all melanomas, the most common melanomas, representing 90% of all cases, are cutaneous melanomas that occur on intermittently or chronically sun-exposed areas of the skin (10). Paradoxically, KIT expression is either low or undetectable in this latter category of melanomas, which are primarily driven by either BRAF or NRAS activating mutations (11–13). The loss of KIT expression has been attributed to frequent deletion or hypermethylation of the KIT locus in melanoma (14,15). While the loss of KIT in cutaneous melanomas is well documented, it remains unclear whether this loss is a cause or consequence of tumorigenesis. Although previous work in cultured cells has suggested that KIT activity may impede melanoma progression by promoting apoptosis or restricting migration (16–18), the impact that KIT activity has on tumor initiation and oncogenic signaling in melanoma has yet to be explored.

To address these issues, we tested whether the absence of KIT affects melanoma formation by introducing a kit loss-of-function mutation into a zebrafish melanoma model that combines melanocyte lineage-expressed BRAF^V600E with a p53 loss-of-function mutation (19). Loss of kit caused a significant acceleration of tumor initiation. In both zebrafish tumors and human melanoma cell lines, loss of KIT led to an increase in RAS/MAPK pathway signaling. Our mechanistic studies suggest that KIT-mediated activation of wild-type RAF proteins can dampen oncogenic signaling from BRAF^V600E. These data indicate that, in the context of a BRAF^V600E mutation, KIT acts as a tumor suppressor.

Materials and Methods

Analyses of The Cancer Genome Atlas (TCGA)

Data from 384 human melanoma RNA-seq samples were downloaded from the Cancer Genomics Hub (CGHub) (https://cghub.ucsc.edu) using GeneTorrent (v 3.8.5a) (20). The RNAseq TCGA dataset was comprised of 302 metastatic melanoma samples and 82 primary melanomas (2). We compared RNAseq-derived FPKM values of KIT expression in BRAF^WT samples to a BRAF^V600E group, where the BRAF^WT group excluded BRAF^V600E and all other BRAF-mutant melanoma samples. Gene Set Enrichment Analysis (GSEA) was performed using GSEA (v 2.2.0) (21,22). A rank-ordered gene list was generated based on Pearson correlation of the expression level of each gene with KIT expression in TCGA melanoma samples. Only genes with reads in ≥90% of samples were included. The rank-ordered gene list was analyzed for enrichment of a set of genes at least 3-fold upregulated by overexpression of BRAF^V600E in cultured melanocytes (23).

Survival analysis was performed by first sorting patients into groups with high (upper-most 20%) or low (lower-most 20%) BRAF^WT:BRAF^V600E allele ratios, followed by further ranking KIT mRNA expression within each group to define the high and low KIT groups as the upper or lower 50% of these ranked lists.

Zebrafish strains and miniCoopR tumor onset assay

The zebrafish mutant alleles used in this study include p53(zdf1) (24), Tg(mitfa:BRAF^V600E) (19), kit(b5) (25), ednrb(b140) (25), kit(e78) (26) and mitfa(w2) (27). Strains were housed and cared for as approved by the Institutional Animal Care and Use Committee (IACUC) at the University of Massachusetts Medical School. miniCoopR constructs were cloned to express either enhanced green fluorescent protein (EGFP) or human BRAF^WT under the mitfa promoter juxtaposed to a mitfa minigene cassette containing the mitfa promoter, open reading frame, and 3’ UTR. These were injected into embryos at the one-cell stage along with Tol2 transposase mRNA. Fish with melanocyte rescue were selected as juveniles approximately two months post-fertilization and monitored weekly by visual inspection for exophytic tumor growth (28).

In-situ hybridization

Tg(mitfa:BRAF^V600E); p53(lf) zebrafish with dorsal anterior melanomas were euthanized, then stored overnight in 4% paraformaldehyde at 4°C. Samples were incubated in 5%, then 30% sucrose, followed by flash freezing in Tissue-tek O.C.T. compound (VWR) on dry ice. 20 µm transverse sections were placed on Superfrost Plus charged slides (Thermo-Fisher) then desiccated and frozen at −80°C. Digoxigenin-labeled riboprobes were synthesized using a zebrafish kita cDNA template transcribed with T3 or T7 polymerases for antisense or sense probes, respectively (Maxi Script Kit; Thermo-Fisher). Tumor samples were rehydrated in a series of decreasing ethanol concentrations and in 2X Saline Sodium Citrate (SSC). They were then incubated in 10 ug/ml Proteinase K for 3–5 minutes and washed with water and triethanolamine (TEA) at pH 8.0. After incubation in an acetic anhydride/TEA solution, samples were dehydrated with increasing ethanol concentrations and dried before beginning an overnight incubation with the riboprobes at 55°C. Washes were performed with 2X SSC, formamide, and then samples were incubated for 30 minutes with RNAse A at 37°C. Samples were incubated in a maleic acid blocking buffer (Roche) for 2 hours followed by an overnight incubation with 1:1000 alkaline phosphatase-conjugated anti-digoxigenin antibody (Roche) at 4°C. Samples were washed with maleate buffer and fluorescence was detected after incubation with Fast Red (HNPP Fluorescent Detection Set; Roche).

Mosaic analysis

Donor blastomeres from kit(lf); Tg(mitfa:BRAF^V600E); p53(lf) embryos were transplanted into Tg(mitfa:BRAF^V600E); p53(lf); mitfa(lf) blastula-stage host embryos or, conversely, donor blastomeres from Tg(mitfa:BRAF^V600E); p53(lf) embryos were transplanted into kit(lf); Tg(mitfa:BRAF^V600E); p53(lf); mitfa(lf) blastula-stage host embryos. Host animals were raised to adulthood and selected for melanocyte positivity, indicating successful donor cell growth. Fish with melanomas that arose on the dorsal anterior region were euthanized, paraffin-embedded, and stained with hematoxylin and eosin to produce transverse sections that were analyzed for tumor cell invasion.

Melanocyte scale density assay

4- to 6-month old fish were treated for 5 minutes with the anesthetic tricaine methanesulfonate and epinephrine, which contracts melanosomes to the central cell body of melanocytes, thereby resolving overlapping cells. Scales were plucked from the dorsal anterior region of fish from the scale rows adjacent to the dorsal midline row. Scales were immediately fixed for ≥30 minutes in 4% paraformaldehyde. After fixation, scales were flat mounted and melanocytes counted. Area was estimated by multiplying maximal antero-posterior and left-right distances on the scale. Five scales from each of three different animals were used to calculate mean melanocyte density.

Cell culture

KIT knockdown experiments were performed in 888MEL cells obtained from the Yale SPORE in Skin Cancer. Cells were grown in F10 (Life Technologies 11550043) media with 5% FBS and 2% penicillin/streptomycin. KIT knockdown was accomplished with the pGIPZ lentiviral shRNA V3LHS_345750 (target sequence 5' - GCATTAAAGCAGCGTATC - 3') or the TRC pLKO.1 shRNA TRCN0000000388 (target sequence 5' - AAACCCAGGGCTGCCTTGGAAAAG - 3'). A non-silencing pGIPZ lentiviral shRNA 22-mer that contains at least 3 or more mismatches against any mammalian gene was used as a negative control (sequence 5’ - ATCTCGCTTGGGCGAGAGTAAG – 3’). Lentiviral particles were produced over the course of 48 hours in HEK293T cells co-transfected with the lentiviral packaging plasmids psPAX and PMD2.G. After harvesting, viral supernatant was filtered with a 0.45µm filter then mixed 1:1 in F10 media with 5% Fetal Bovine Serum (FBS), 2% penicillin/streptomycin, and 2–8 µg/ml Polybrene (Sigma). Cells were selected using 1 µg/ml puromycin for 2–4 days to establish stable cell lines. An identical approach was used to overexpress KIT and EGFP from pLENTI-CMV-KIT or pLENTI-CMV-EGFP constructs in HEK293T, A375, and UACC257 cells, but in these cases the viral supernatant was mixed 1:1 with DMEM plus 10% FBS and 2% penicillin/streptomycin. HEK293T, A375, and UACC257 cells were grown in DMEM plus 10% FBS and 2% penicillin/streptomycin. HEK293T transient transfections were performed with Lipofectamine 2000 (Life Technologies) according to the manufacturer’s protocol.

BRET Assay

BRET was performed as described by Lavoie et. al. (29). In brief, HEK293T cells were plated at 300,000 cells per well in six-well tissue culture-treated plates. After 48 hours, each well was incubated in 2 ml Opti-MEM, then transfected with 7.5 µl Lipofectamine 2000 plus a specified combination of pLHCX-RLuc-CRAF (100 ng), pLPCS-V-BRAF^V600E, pLPCS-V-BRAF^R188L/V600E, pLPCS-V-BRAF^WT, or pLPCS-V-BRAF^R188L (0, 50, 100, 200, 300, 400, or 600 ng), and pCDNA3.1-NRAS^Q61K (250 ng) or pCDNA3.1-EMPTY (250 ng) in 200 µl Opti-MEM. Either KIT stimulation with SCF or NRAS^Q61K overexpression was used for induction of upstream pathway activity (STIM), as indicated. 48 hours after transfection, cells were resuspended in 500 µl Tyrode’s buffer, then split onto opaque 96-well plates (Perkin Elmer 6005680). An initial reading of total Venus expression was taken on a PE Envision plate reader (excitation 480±30nm, emission at 530±10nm). 10 µl of coelenterazine h (Biotium 10111-1) was then added automatically through the PE Envision robotic pump to a final concentration of 2.5 µM. BRET readings were taken using two filter sets (emission 485±15nm, emission 530±10nm) 5 minutes after adding coelenterazine. The BRET signal was determined by calculating the 530/485 signal ratio for a given sample and subtracting the 530/485 ratio measured when expressing RLuc-CRAF alone. These values were plotted as a function of the total 530/485 ratio (as a measurement of [Acceptor]/[Donor]) for a range of [Acceptor]/[Donor] concentrations, then fit with a one-site binding hyperbolic equation in GraphPad Prism (v 6.05). Each condition was run in biological triplicate, and each of these replicates was run on a separate plate in immediate succession to one another. Each plate was calculated separately for background subtraction, then all data points from the three biological replicates were combined to calculate Kd, referred to here as BRET⁵⁰.

Western Blotting

Protein was harvested from zebrafish melanomas 2–6 weeks post-tumor onset by first euthanizing the fish with tricaine methanesulfonate followed by surgical removal of the tumor. Tumors were triturated and lysed in ice-cold RIPA buffer containing a cOmplete protease inhibitor tablet (Roche). Protein concentration was measured using the Pierce BCA Protein Assay Kit (Life Technologies). Samples were run on 10% polyacrylamide gels, transferred, and developed using fluorophore-conjugated antibodies (LI-COR). Antibodies against the following proteins were used: Mitfa (28); RAF-1 (c-RAF), pMEK1/2 (S217/221), MEK1/2, p44/42 MAPK (ERK1/2), alpha-Tubulin, pAKT S473, AKT, pc-KIT, c-KIT (Cell Signaling 9422, 9154, 8727, 4695, 38735, 4060, 4685, 3391, 3308, respectively); pERK (Sigma m8159); BRAF^V600E (Spring Bioscience E19290); total BRAF (Millipore 10146); IRDye 800CW Donkey anti-Rabbit and IRDye 680RD Goat anti-Mouse (LI-COR 926-32213 and 926-68070, respectively). All quantitative measurements were calculated based on measurements from a LI-COR Odyssey Imaging System and quantified with Image Studio Lite (v 5.0) software. For background subtraction, a median pixel intensity for regions 3 pixels above and below a selected band was calculated and subtracted from the mean pixel intensity of that band.

Growth Curves

Cells were plated at 20,000–50,000 cells per well in 6-well plates and growth was assayed by resuspending and directly counting the cells on a hemocytometer with samples taken in biological triplicate for 6 days. Cells were grown in their respective media specified above, plus 0.5% FBS with either 200 ng/ml Bovine Serum Albumin or Stem Cell Factor administered on days 1 and 4.

Statistical Analysis

Significance calculations were performed on samples collected in a minimum of biological triplicate. P values from two-tailed Student’s t tests were calculated for all comparisons of continuous variables. All further significance tests were performed in Graphpad Prism (v6.05). Fisher’s exact tests were used to calculate P values for 2×2 contingency tables of the association of KIT expression with melanoma patient clinical parameters, whereas a P value for the 2×3 contingency table of zebrafish melanoma invasiveness was calculated by a chi-squared test. F-tests were used to determine significance of changes upon stimulation of BRET titration curves. Log-rank tests were used to calculate significance of changes in zebrafish tumor onset curves as well as the differences in overall survival of select human melanoma cohorts. A P value < 0.05 was considered significant.

Results

As described below, we aimed to test whether the loss of KIT would impact melanoma initiation by using a zebrafish strain that develops fully-penetrant melanomas driven by human BRAF^V600E (19). To first confirm that examining the loss of KIT in a BRAF^V600E-mutant background was appropriate, we investigated whether reduced KIT expression was associated with oncogenic BRAF mutations in human melanoma. Based on normalized transcript amounts from tumor samples in The Cancer Genome Atlas (TCGA) (2), BRAF^V600E-mutant melanomas generally expressed lower levels of KIT than melanomas with wild-type BRAF (BRAF^WT) (Supplementary Fig. S1A). Contributing to this difference were tumors, in the BRAF^WT group that contained KIT gain-of-function mutations. As a further indication of the inverse relationship of KIT expression and BRAF^V600E mutations, gene set enrichment analysis (GSEA) revealed that a set of genes upregulated by BRAF^V600E (23) was associated with low KIT expression in BRAF^V600E-mutant melanomas (Supplementary Fig. S1B). These data indicate that oncogenic BRAF mutations and activity are correlated with low KIT expression in human melanoma.

We began investigating the relationship between KIT and BRAF^V600E in zebrafish by crossing a kit(b5) loss-of-function mutant strain (25), referred to hereafter as kit(lf), with a Tg(mitfa:BRAF^V600E) strain, that expressed BRAF^V600E in the melanocyte lineage (19). This was important because it was unclear whether melanoma could be investigated in a kit-mutant background - melanomas in zebrafish predominantly arise from dorsal regions containing scale-associated melanocytes, and kit(lf) zebrafish lack these melanocytes. In the resulting kit(lf); Tg(mitfa:BRAF^V600E) strain, dorsal scale-associated melanocytes were present in a typical net-like pattern, although at a density somewhat lower than in Tg(mitfa:BRAF^V600E) animals (Supplementary Fig. S2A). To determine whether BRAF^V600E rescued development of dorsal, scale-associated melanocytes that were missing in kit(lf) fish or whether the oncogene had merely induced mis-migration of otherwise extant stripe-associated melanocytes, we further crossed these strains into an ednrb(lf) background (25). Whereas, kit(lf); ednrb(lf) double mutants lacked melanocytes entirely, kit(lf); ednrb(lf); Tg(mitfa:BRAF^V600E) fish still developed a population of dorsal scale-associated melanocytes (Supplementary Fig. S2B). These results indicate that BRAF can act downstream of, or in parallel to, kit to promote melanocyte development, which is consistent with KIT receptor signaling through BRAF that is thought to occur in this cell type (30,31).

Rescue of melanocyte development by BRAF^V600E enabled us to test whether the loss of kit would have an effect on BRAF^V600E-driven melanoma formation. Before doing so, we assessed whether kit was expressed in melanomas from Tg(mitfa:BRAF^V600E); p53(lf) fish. In contrast to most human BRAF^V600E-mutant melanomas, kit was expressed in Tg(mitfa:BRAF^V600E); p53(lf) zebrafish melanomas (Fig. 1A), making this model useful for interrogating the effects of the loss of kit in melanoma formation. We established a kit(lf); Tg(mitfa:BRAF^V600E); p53(lf) strain and monitored tumor onset in these animals as compared to Tg(mitfa:BRAF^V600E); p53(lf) controls. The kit(lf); Tg(mitfa:BRAF^V600E); p53(lf) fish had markedly accelerated melanoma onset relative to Tg(mitfa:BRAF^V600E); p53(lf) animals; the median melanoma onset in kit(lf); Tg(mitfa:BRAF^V600E); p53(lf) fish was 21 weeks, whereas it was 35 weeks in a Tg(mitfa:BRAF^V600E); p53(lf) background (Fig. 1B). To confirm that accelerated onset was due to the loss of kit rather than background effects, we asked whether a second, independently-derived kit loss-of-function allele, kit(e78) (26), could also affect onset. Indeed, the kit(e78);Tg(mitfa:BRAF^V600E); p53(lf) strain experienced accelerated tumor onset similar to kit(b5); Tg(mitfa:BRAF^V600E); p53(lf) fish (Supplementary Fig. S3A), indicating that the loss of kit is responsible for the accelerated tumor onset. Melanomas from kit(lf); Tg(mitfa:BRAF^V600E); p53(lf) (all further use of “kit(lf)” refers to the kit(b5) strain) were more frequently invasive, with 65% of tumors having penetrated through the musculature and into the spinal column at four weeks post onset, while only 9% of tumors from control fish displayed such invasion (Fig. 1C and Table 1). The invasiveness was determined to be a cell-autonomous property by mosaic analysis experiments in which melanomas derived from kit(lf); Tg(mitfa:BRAF^V600E); p53(lf) cells were invasive in a Tg(mitfa:BRAF^V600E); p53(lf); mitfa(lf) background, whereas melanomas derived from Tg(mitfa:BRAF^V600E); p53(lf) cells in a kit(lf); Tg(mitfa:BRAF^V600E); p53(lf); mitfa(lf) background were not invasive (Supplementary Fig. S3B and S3C). These results indicate that the loss of kit promotes tumor onset and invasiveness, demonstrating that kit acts as a tumor suppressor in BRAF^V600E-mutant melanomas.

Figure 1. kit suppresses BRAF^V600E-driven melanoma formation.

(A) In-situ hybridization of a representative Tg(mitfa:BRAF^V600E); p53(lf) zebrafish melanoma with kit antisense and sense probes. Dashed yellow lines indicate the border between normal muscle tissue and tumor. Yellow squares indicate the locations of the inset magnified for the antisense images in the center. Scale bars = 150 µm (B) Tumor onset curves for Tg(mitfa:BRAF^V600E); p53(lf) and kit(lf); Tg(mitfa:BRAF^V600E); p53(lf) fish. P value = 8.46×10^-12; log-rank test. (C) Representative adult zebrafish with dorsal melanomas. Lower panels show transverse sections from non-invasive Tg(mitfa:BRAF^V600E); p53(lf) and invasive kit(lf); Tg(mitfa:BRAF^V600E); p53(lf) tumors. Quantification of invasiveness is shown in Table 1. Scale bars = 0.5 mm.

Table 1.

Melanoma invasiveness in kit(lf) mutant zebrafish

	Tg(mitfa:BRAFV600E); p53(lf) (n=11)	kit(lf); Tg(mitfa:BRAFV600E); p53(lf) (n=17)
Skin	100%	100%
Muscle	36%	100%
Muscle + spinal column	9%	65%

Percentages of fish that were positive for melanoma cells at the indicated anatomical depth relating to the representative images in Fig. 1C. Tg(mitfa:BRAF^V600E); p53(lf) versus kit(lf); Tg(mitfa:BRAF^V600E); p53(lf), P value = 0.0004; chi-squared test.

To explore whether signaling downstream from kit is altered in kit(lf); Tg(mitfa:BRAF^V600E); p53(lf) tumors, we analyzed levels of phosphorylated Erk (pErk), phosphorylated Akt (pAkt), and Mitfa by western blotting. There was no significant difference in Akt activation or Mitfa expression between kit(lf); Tg(mitfa:BRAF^V600E); p53(lf) and Tg(mitfa:BRAF^V600E); p53(lf) tumors (Supplementary Fig. S4A-C). However, kit(lf); Tg(mitfa:BRAF^V600E); p53(lf) tumors did possess 2.6-fold higher levels of pErk (Fig. 2A and B). Phospho-specific MEK (pMEK) antibodies did not cross react with zebrafish MEK protein, so we could not quantify pMEK/MEK ratios to further measure pathway activity. ERK activation is the major conduit for BRAF^V600E signaling, therefore KIT inhibition of melanoma formation could be mediated by inhibition of oncogenic BRAF^V600E signaling.

Figure 2. kit suppresses BRAF^V600E oncogenic signaling in zebrafish melanomas.

(A) Western blots and (B) quantification of pErk/Erk for Tg(mitfa:BRAF^V600E); p53(lf) and kit(lf); Tg(mitfa:BRAF^V600E); p53(lf) zebrafish melanomas. pERK/ERK signals were normalized to the mean pERK/ERK signal from Tg(mitfa:BRAF^V600E); p53(lf) tumor samples. Box plot edges represent minimum, 1^st quartile, median, 3^rd quartile, and maximum normalized values. P value < 0.05; Student’s t test.

To determine whether KIT also suppressed RAS/MAPK pathway signaling in human melanoma cells, we knocked down KIT in 888MEL cells, a rare melanoma cell line that has a BRAF^V600E mutation but still expresses KIT (16). 888MEL cells might sustain KIT expression through the transcription factor AP-2, which directly binds and upregulates the KIT promoter (32). Although AP-2 expression is frequently absent in melanoma cell lines, it has been previously shown that 888MEL cells have high levels of this factor (32,33). We observed BRAF^V600E activity by measuring pMEK. MEK is phosphorylated by BRAF and is the most direct measure of pathway activity downstream of BRAF activation. In cells where KIT was knocked down with either of two independent shRNAs, levels of pMEK increased (Fig 3A). Also, KIT-knockdown cells more readily formed colonies when plated at low density in an anchorage-dependent colony formation assay (Fig. 3B). Furthermore, stimulation of KIT-expressing 888MEL cells with Stem Cell Factor (SCF), the KIT ligand, reduced both MEK activation and growth of these cells without affecting expression of either KIT or BRAF^V600E (Fig. 3C and D). These experiments demonstrate that KIT can function in 888MEL cells to suppress RAS/MAPK pathway activity and reduce their growth potential.

Figure 3. KIT suppresses BRAF^V600E oncogenic signaling in a human melanoma cell line.

(A) Western blots of pMEK and MEK for human melanoma 888MEL cells expressing a control non-silencing shRNA or either of two KIT-targeting shRNAs. (B) Colony formation assay for 888MEL cells expressing a control non-silencing shRNA or either of two KIT-targeting shRNAs. (C) Western blot and quantification of 888MEL cells treated with 200 ng/ml Bovine Serum Albumin (BSA) or Stem Cell Factor (SCF) 30 minutes prior to protein harvest. (D) 888MEL cells were counted during adherent growth with 200 ng/ml BSA or SCF treatment administered on days 1 and 4. Data in B, C, and D are represented as mean ±SEM for experiments done in biological triplicate. *P value < 0.05, **P value < 0.01, NS, not significant; Student’s t test.

We considered different possibilities for how KIT could inhibit BRAF^V600E-driven signaling. Since KIT activity normally promotes downstream MEK/ERK signaling, it is paradoxical that the loss of KIT would lead to an increase in downstream pathway activity. Excess ERK activation invokes negative feedback loops, so perhaps KIT upregulates signaling to a level where negative feedback is needed to dampen pathway activity. We consider such mechanisms unlikely because A) a major mode of negative feedback is through upregulation of dual-specificity phosphatases, which remove activating phosphoryl groups from ERK (34) – upon loss of KIT in 888MEL cells, we found increased pMEK (Fig. 3A), arguing that the effect of KIT loss is not centered on ERK, but instead occurs more upstream in the pathway, and B) negative feedback is also accomplished by inhibitory phosphorylation on BRAF by ERK – BRAF^V600E is insensitive to such inhibition and thus not subject to this type of negative feedback (35).

For these reasons, we considered mechanisms whereby KIT could more directly interfere with BRAF^V600E-driven signaling. RTK activation of RAS leads to membrane-localized BRAF:CRAF dimer formation and activation (29,36). The V600E mutation shifts BRAF into a constitutively active conformation capable of phosphorylating MEK in a RAS-independent manner with vastly increased kinase activity (37–39). Previously it was reported that BRAF^V600E kinase activity was reduced when it heterodimerized with CRAF (40), so we hypothesized that KIT could drive formation of BRAF^V600E:CRAF complexes, which would have less activity than uncomplexed BRAF^V600E. To test this, we used a bioluminescence resonance energy transfer (BRET) assay to measure the interaction between Renilla Luciferase-tagged CRAF (RLuc-CRAF) and Venus-tagged BRAF^V600E (V-BRAF^V600E). We determined transfection conditions in HEK293T cells where formation of V-BRAF^V600E:RLuc-CRAF dimers was induced by co-expression with NRAS^Q61K as a constitutively active surrogate for pathway stimulation (STIM). Under stimulating conditions, an increase in dimerization was measured between V-BRAF^V600E and RLuc-CRAF as a decrease in BRET⁵⁰, the acceptor/donor ratio that yields 50 percent of maximal signal of a BRET titration curve (Supplementary Fig. S5A and Supplementary Table S1). We used site directed mutagenesis of the V-BRAF^V600E construct to introduce an R188L mutation, which disrupts the RAS-binding domain of BRAF (41). V-BRAF^R188L/V600E was less sensitive to upstream stimulation-induced RLuc-CRAF binding, as indicated by a less dramatic shift in the BRET⁵⁰ for its titration curve (Supplementary Fig. S5A and Supplementary Table S1). Levels of pMEK increased when cells expressing V-BRAF^V600E were stimulated by upstream pathway activity, but did not increase as much for V-BRAF^R188L/V600E stimulation (Supplementary Fig. S5B and S5C). These data suggest that, in the context of upstream stimulation, formation of BRAF^V600E:CRAF complexes does not diminish the total signaling output of BRAF^V600E-mutant cells.

We therefore examined another hypothesis. Since the BRAF^WT allele can continue to be expressed in BRAF^V600E-mutant cancer cells (42–44), we reasoned that upstream activation of this pathway could recruit BRAF^WT proteins to form dimers, thus engaging pathway components which possess relatively low kinase activity. To determine if BRAF^WT:CRAF dimers could dampen BRAF^V600E activity, we again used BRET assays to determine transfection conditions in HEK293T cells that yielded stimulation-dependent increases in V-BRAF^WT:RLuc-CRAF heterodimers. Upstream stimulation induced a decrease in BRET⁵⁰, indicating a robust interaction of V-BRAF^WT and RLuc-CRAF (Fig. 4A and Table 2). Upstream stimulation only modestly induced an interaction between a V-BRAF^R188L mutant and RLuc-CRAF. Using transfection conditions that yielded low, medium, and near saturating levels of V-BRAF^WT:RLuc-CRAF interaction, we measured MEK phosphorylation. With upstream stimulation, pMEK levels increased when any amount of BRAF^WT was expressed in cells (Fig. 4B). A less robust increase was observed when BRAF^R188L was expressed. We next asked whether expression of BRAF^WT could impact BRAF^V600E-driven MEK activation. In the absence of pathway stimulation, neither the addition of BRAF^WT nor BRAF^R188L affected pMEK levels in BRAF^V600E-expressing cells (Fig. 4C). However, with upstream stimulation, BRAF^WT expression reduced pMEK levels in BRAF^V600E-expressing cells, whereas BRAF^R188L expression had no effect (Fig. 4D). Quantification of relative pMEK/MEK was calculated in biological triplicate samples of all BRAF titrations for each experimental condition (Fig. 4E-F).

Figure 4. Upstream stimulation of BRAF^WT can dampen BRAF^V600E oncogenic signaling.

(A) BRET titration curves measuring association between RLuc-CRAF and varying levels of V-BRAF^WT or V-BRAF^R188L in the presence or absence of stimulation from NRAS^Q61K (STIM). Data points represent the mean ±SEM for either the [Acceptor]/[Donor] ratio on the x-axis, or the BRET signal on the y-axis for three biological replicates. Quantification of BRET⁵⁰ values is shown in Table 2. (B-D) Western blots of pMEK and MEK from HEK293T cells expressing BRAF^WT or BRAF^R188L with upstream stimulation alone (B), with BRAF^V600E alone (C), or with both stimulation and BRAF^V600E (D). The highest bands on the BRAF blots are Venus-tagged BRAF^V600E, whereas the lower bands are untagged BRAF^WT or BRAF^R188L, with BRAF^WT running slightly higher in lane 2 of panels B and D when activated by upstream stimulation. (E-G) Quantification of pMEK/MEK levels for cells expressing varying levels of BRAF^WT or BRAF^R188L with upstream stimulation alone (E), with BRAF^V600E alone (F), or with both stimulation and BRAF^V600E (G). Fold changes were calculated in comparison to control cells in which EGFP was expressed instead of BRAF^V600E or BRAF^R188L. Quantification indicates the mean ±SEM of three biological replicate experiments. P values were calculated comparing each condition of BRAF^WT to BRAF^R188L expression. *P value < 0.05, **P value < 0.01, ***P value < 0.001, ****P value < 0.0001; Student’s t test.

Table 2.

BRET⁵⁰ values for V-BRAF^WT and V-BRAF^R188L

	V-BRAFWT BRET50	V-BRAFR188L BRET50	Fold difference
+ STIM	1.037	4.050	3.905****
+ EMPTY	4.473	4.711	1.053 (NS)
Fold change	4.313****	1.163***

BRET⁵⁰ and BRET⁵⁰ fold changes for each of two Venus-BRAF constructs expressed with RLuc-CRAF in HEK293T cells in the presence or absence of stimulation with NRAS^Q61K (STIM). Associated with the BRET titration curve in Fig. 4A. Data obtained from biological triplicate samples.

^***P value < 0.001,

^****P value < 0.0001,

NS, not significant; F-test.

We repeated these tests using HEK293T cells stably expressing KIT or enhanced green fluorescent protein (EGFP). Upon stimulation with SCF, we detected KIT phosphorylation (Supplementary Fig. S6A). In addition, SCF treatment increased levels of pMEK and promoted the interaction of V-BRAF^WT with RLuc-CRAF (Supplementary Fig. S6A and S6B and Supplementary Table S2), although both were more modest than the increases observed when cells were stimulated by NRAS^Q61K. SCF stimulated, KIT or EGFP-expressing cells were transfected with BRAF constructs and, as before, BRAF^WT reduced BRAF^V600E-driven signaling, but only in the KIT expressing cells (Supplementary Fig. S6C and S6D). These data indicate that BRAF^WT can attenuate BRAF^V600E-driven MEK activation, and this attenuation is most pronounced in the context of robust upstream stimulation.

Our studies in HEK293T cell lines suggested that signaling through BRAF^WT was important for KIT-mediated tumor suppression. To further explore the relationship between KIT and BRAF^WT, we examined signaling in melanoma cells. A375 cells are BRAF^V600E-mutant and have no detectable BRAF^WT expression, whereas UACC257 cells are BRAF^V600E-mutant but retain expression of BRAF^WT (45). Neither cell line expressed KIT endogenously (Supplementary Fig. S7A). To reconstitute upstream signaling in these cells, we introduced exogenous KIT and grew cells in the presence of SCF. KIT expression promoted growth of A375 cells and increased pMEK levels. The opposite was observed in UACC257 cells, in which KIT expression slowed growth and decreased pMEK levels (Supplementary Fig. S7B-S7D). These results further suggest that the tumor suppressive activity of KIT is related to the expression of BRAF^WT. Additionally, it is notable that 888MEL cells, which displayed KIT-mediated growth and signaling inhibition, also expressed BRAF^WT (Supplementary Fig. S7E and S7F). Lastly, we examined whether the link between KIT and BRAF^WT extended to melanoma clinical data. TCGA melanoma samples were segregated into quintiles based on BRAF^WT-to-BRAF^V600E ratios, and the upper-most (high BRAF^WT:BRAF^V600E) and lower-most (low BRAF^WT:BRAF^V600E) quintiles were examined. Within each group, cohorts with high and low KIT expression were defined. In the high BRAF^WT:BRAF^V600E group, patients whose tumors expressed high levels of KIT experienced a survival benefit as compared to patients whose tumors had low KIT expression (Supplementary Fig. S8A). In the low BRAF^WT:BRAF^V600E group, there was no correlation between KIT expression and patient survival (Supplementary Fig. S8B). We further observed that high KIT expression was associated with a decreased likelihood for melanomas to present with regional or distant metastases (tumor stage III/IV) amongst the group of patients with high BRAF^WT:BRAF^V600E ratios (Supplementary Table S3). KIT expression had no association with the likelihood of metastases in the group of patients with low BRAF^WT:BRAF^V600E ratios (Supplementary Table S3). We further analyzed potential relationships between KIT expression and other common melanoma tumor suppressors. This analysis revealed no significant link between KIT and the loss of PTEN, TP53, or CDKN2A, suggesting that there is no interdependence or redundancy of KIT with these suppressors (Supplementary Table S4).

To test the link between KIT and BRAF^WT in vivo, we employed the established ‘miniCoopR’ method for generating transgenic melanomas in zebrafish (28). The miniCoopR vector juxtaposes a transgene of interest to a copy of the mitfa melanocyte specification gene. When this construct is injected into melanocyte-deficient mitfa(lf) zebrafish embryos, melanocytes are rescued cell-autonomously by the mitfa gene, and each rescued melanocyte expresses the neighboring transgene of interest (Fig. 5A). miniCoopR vectors containing either BRAF^WT or EGFP open reading frames under control of the mitfa promoter were injected into Tg(mitfa:BRAF^V600E); p53(lf); mitfa(lf) embryos, and animals with rescued melanocytes were monitored weekly for melanoma onset. Expression of BRAF^WT delayed median tumor onset by seven weeks as compared to expression of EGFP (Fig. 5B). This change was not due to altered expression of BRAF^V600E, as levels of this oncoprotein were similar in BRAF^WT and EGFP-expressing cohorts (Supplementary Fig. S9). To determine if this delayed onset was dependent on kit, the same experiment was performed, except miniCoopR constructs were injected into a kit(lf); Tg(mitfa:BRAF^V600E); p53(lf); mitfa(lf) background. In this background, melanocyte rescue was poor and consequently tumor onset of the miniCoopR-EGFP control cohort was slower as compared to miniCoopR-EGFP in a Tg(mitfa:BRAF^V600E); p53(lf); mitfa(lf) background. Within the kit(lf); Tg(mitfa:BRAF^V600E); p53(lf); mitfa(lf) background, the miniCoopR-EGFP control cohort and miniCoopR-BRAF^WT animals exhibited no difference in melanoma onset (Fig. 5C). These data indicate that BRAF^WT can inhibit BRAF^V600E-driven tumor onset, but this inhibition is dependent upon expression of KIT.

Figure 5. BRAF^WT delays BRAF^V600E-driven melanoma formation in a kit-dependent manner.

(A) Overview of a miniCoopR tumor onset experiment as described in Materials and Methods. Tumor onset was monitored weekly for fish expressing BRAF^WT or EGFP. (B) Tumor onset of Tg(mitfa:BRAF^V600E); p53(lf); mitfa(lf) zebrafish with melanocytes rescued by miniCoopR-BRAF^WT or miniCoopR-EGFP, P = 8.05×10⁻⁷; log-rank test. (C) Tumor onset of kit(lf); Tg(mitfa:BRAF^V600E); p53(lf); mitfa(lf) zebrafish with melanocytes rescued by miniCoopR-BRAF^WT or miniCoopR-EGFP, P value = 0.664; log-rank test.

Discussion

Our results show that KIT can suppress BRAF^V600E-driven melanoma formation. The loss of KIT resulted in increased BRAF^V600E-driven oncogenic signaling in a zebrafish melanoma model and human melanoma cells. Our mechanistic studies suggest that BRAF^V600E activity can be reduced by BRAF^WT, but only under conditions where BRAF^WT receives upstream pathway stimulation. In vivo data support this notion, as expression of BRAF^WT suppressed BRAF^V600E-driven melanoma initiation in a KIT-dependent manner.

There are several possible ways in which KIT could inhibit BRAF^V600E-driven oncogenic signaling. Important negative regulators of oncogenic signaling include the dual-specificity phosphatases (DUSPs), which dephosphorylate and inactivate ERK (34,46). Potentially, KIT could stimulate DUSP activity either by signaling through ERK or by signaling independently, leading to pathway inhibition. However, upon knockdown of KIT in BRAF^V600E-mutant melanoma cells, we observed increased pMEK levels. As negative feedback from DUSPs is not expected to have an effect on MEK phosphorylation status (34,47), this led us to further investigate potential mechanisms upstream of ERK. We also considered a mechanism whereby KIT signaling would drive dimerization of BRAF^V600E with CRAF, which has been reported to lower the kinase activity of BRAF^V600E (40). While this is an attractive model, additional observations suggest that the situation is more complex. In our assay, we did not find reduction of downstream signaling when BRAF^V600E was driven into dimers with CRAF. Our assay differs from the one used by Karreth et al. in that we measured downstream signaling (i.e. MEK phosphorylation) under conditions of upstream pathway stimulation. Stimulation not only facilitates membrane recruitment and dimerization of RAF species, but also promotes their activation through phosphorylation at a series of sites in the negatively-charged region (39). It is possible that BRAF^V600E:CRAF dimers, under conditions of upstream stimulation, receive additional activating cues, enhancing the activity of these species.

An alternative model to explain our findings would have the ratio of high activity to low activity BRAF species determine flux through the signaling pathway. In this model, under conditions of no upstream stimulation, only high activity BRAF^V600E would be functional. By contrast, upon upstream stimulation a mixture of low activity BRAF^WT:CRAF and high activity BRAF^V600E:CRAF dimers form. Although upstream stimulation could increase the collective number of active BRAF species, overall pathway signaling would be diminished if low activity BRAF^WT:CRAF species have a prominent role in determining signaling flux. Such a role could manifest in a variety of ways. For example, BRAF^WT:CRAF dimers could compete with BRAF^V600E:CRAF dimers for interaction with their shared MEK substrate. Although not considered in our analysis, BRAF^WT and BRAF^V600E could also potentially compete for interaction with KSR scaffolding proteins (48). In each of these examples, any interaction of BRAF^WT with limiting downstream components could attenuate signaling.

These proposed mechanisms highlight the interplay between wild-type and oncogenic signaling in tumors. During tumorigenesis, oncogenic mutations in genes typically affect one allele, leaving the other unaffected. In tumors driven by oncogenic RAS genes, loss of the corresponding wild-type gene is frequently observed (49). A rationale for this loss comes from elegant genetic studies in the mouse which showed that the wild-type RAS gene has a tumor suppressive effect in certain tumor types (50). By contrast, loss of the BRAF^WT allele is rare in tumors driven by oncogenic BRAF, and co-expression of wild-type and oncogenic variants is evident in the majority of tumors (43). In such tumors we propose that signaling through BRAF^WT has a suppressive effect, but that the loss of KIT effectively abrogates this effect. More generally, our results reveal that, under certain circumstances, an unexpected increase of oncogenic RAS/MAPK pathway activity could occur upon loss or inhibition of upstream signaling components.