Growth suppression by dual BRAF(V600E) and NRAS(Q61) oncogene expression is mediated by SPRY4 in melanoma

The underlying forces that shape mutational patterns within any type of cancer have been poorly characterized. One of the best preserved exclusionary relationships is that between BRAF(V600E) and NRAS(Q61) in melanomas. To explore possible mechanisms which could explain this phenomenon, we overexpressed NRAS(Q61) in a set of BRAF(V600E) melanoma lines and vice versa. Controlled expression of a second activating oncogene led to growth arrest (“synthetic suppression”) in a subset of cells, which was accompanied by cell cycle arrest and senescence in several melanoma cell lines along with apoptosis. Through differential gene expression analysis, we identified SPRY4 as the potential mediator of this synthetic response to dual oncogene suppression. Ectopic introduction of SPRY4 recapitulated the growth arrest phenotype of dual BRAF(V600E)/NRAS(Q61) expression while SPRY4 depletion led to a partial rescue from oncogenic antagonism. This study thus defined SPRY4 as a potential mediator of synthetic suppression, which is likely to contribute to the observed exclusivity between BRAF(V600E) and NRAS(Q61R) mutations in melanoma. Further leverage of the SPRY4 pathway may also hold therapeutic promise for NRAS(Q61) melanomas.


Introduction
Within tumors, mutational patterns reflect strong evolutionary forces and yet represent only a single snapshot of a complex physiology. Over 45,000 tumor specimens have been subjected to various analyses including whole exome sequencing (www.cbioportal.org). One of the most commonly activated networks is the RAS-MAPK, which impacts both cancer cell proliferation and survival [1,2]. While recurrent oncogenic lesions in BRAF (pV600) and N/ K/HRAS (pG12/13, p.Q61) predominate within the RAS-MAPK pathway, they are rarely identified in conjunction within any single tumor specimen [3,4].
Among the myriad of tumors analyzed to date, cutaneous melanoma bears some of the highest mutational burdens [5]. Thus, it is somewhat surprising that exclusion between BRAF c.1799T>A(V600E) and NRAS c.181C>A (Q61K)/ NRAS c.182A>G (Q61R) mutations is so pronounced in melanoma; there is only a single melanoma tumor specimen out of 366 sequenced which harbored concurrent BRAF c.1799T>A(V600E)/BRAF c.1798G>A (V600M) and NRAS c.37G>C (G13R) mutations (TCGA-ES-A2NC sample; www.bioportal.org). The biological pressures that govern the emergence and patterning of these activating alleles have not been well characterized. A priori, redundancy and antagonism, through growth arrest, apoptosis, senescence or other means, are both possible explanations. Under a redundancy model, the second oncogenic hit would have minimal functional impact and thus exist as a low probability "passenger" oncogene. Alternatively, under an antagonistic framework, an additional activating allele would functionally interfere with tumor growth and thus drop out of the final tumor population. Petti et al. showed that forced expression of NRAS(Q61R) in a single BRAF (V600E) melanoma line led to growth arrest and induction of SA-ß-gal [6], consistent with senescence. These results suggest that the introduction of a rival oncogene impinges on two cancer processes: oncogene-induced senescence (OIS) and synthetic lethality. In the former, expression of a strong activating allele in the context of a noncancerous cell leads to the onset of senescence due to a battery of compensatory mechanisms [7] such as normal telomerase activity. Since melanoma cells have already breached OIS during their initial transformation, it would be more appropriate to describe oncogene exclusion as "secondary OIS". For synthetic lethality, the viability of a cancer cell is compromised when two mutations co-exist whether these changes be activating or loss-of-function [8]. While synthetic lethal interactions may be condition-dependent, there is much enthusiasm about identifying such genetic pairs since the potency of synthetically lethal interactions could offer clues about potentially "druggable" targets. Furthermore, since dual mutant states may be antagonistic but not necessarily lethal, perhaps "synthetic suppression" could be a more encompassing term. Along these lines, we set out to more deeply characterize the mechanism(s) which proscribe the concurrence of BRAF(pV600E) and NRAS(pQ61) mutations in melanoma with an eye towards novel pathways which could countermand constitutive BRAF or NRAS signaling.

Molecular response to dual oncogenesis
Examination of signaling cascades which might be activated in response to dual oncogene expression was not initially revealing. There were some modest and inconsistent increases in both pMEK and pERK, which did not appear to correlate with the observed response (Fig. S4). Even when broader signal profiles were obtained with phosphokinase arrays, no recurrent phosphorylation events were noted (data not shown). These findings indicate that signaling differences may not underpin the antagonistic phenotype, but rather, the molecular circuitry may itself be reprogrammed. Thus, we set out to map the molecular events which are downstream of dual oncogenesis.
While it is likely that many concurrent pathways have been activated to bring about growth arrest in the SK-MEL-119 NRAS* + iBRAF* and GMEL BRAF* + iNRAS*, SPRY4 transcripts were among the most upregulated ones in both antagonized SK-MEL-119 NRAS* + iBRAF* and GMEL-BRAF* + iNRAS* lines, but not in the neutral A375 BRAF* + iNRAS* cell line ( Fig. 3d Table). As SPRY4 has been implicated as a tumor suppressor, we set out to explore the possibility that SPRY4 is mediating growth suppression selectively in the antagonistic lines.
To more directly prove that SPRY4 is mediating the synthetic suppression, we set out to determine if SPRY4 itself is growth suppressive, both in vitro and in vivo, and if the depletion of SPRY4 can rescue cells from the observed antagonism. As shown in Fig. 5a, constitutive expression of SPRY4 dramatically inhibited proliferation of SK-MEL-119 NRAS* and GMEL BRAF* cells (upper panel) but not in neutral A375 BRAF* and SK-MEL-63 NRAS* cells (lower panel). In vitro analysis of SK-MEL-119 NRAS* confirmed the increase in apoptosis and SA-ß-galactosidase enzyme activity with SPRY4 induction (Fig. S5a). Figure S5b shows the reduction in density along with morphologic changes associated with SPRY4 overexpression (day 5).
To further investigate whether SPRY4 overexpression can reduce tumor growth in vivo, SK-MEL-119 NRAS* cells were transfected either with empty CD516B-2-Vector or CD516B-2-SPRY4 and subjected to xenograft experiments. As shown in Fig. 5b, SPRY4 profoundly inhibited tumorigenesis in NSG mice (p < 0.0001) with no change in body weight (Fig. S5c). Histologically, tumors that overexpressed SPRY4 had reduced overall cellularity and increased fibrous stroma (Fig. 5c). Compared to control tumors, SPRY4 tumors also harbored reduced Ki67, enhanced tumor cell apoptosis as shown by TUNEL staining and increased SAß-gal expression (Fig. 5c). Taken together, SPRY4 appears to suppress in vivo tumor xenograft growth, which can be partially explained by an increase in senescence similar to the observations in vitro. We next sought evidence of rescue from oncogene antagonism by depleting SPRY4 in SK-MEL-119 NRAS* + iBRAF* cells using siRNA's against SPRY4. Effective SPRY4 suppression by siRNA in SK-MEL-119 NRAS* + iBRAF* (2.47-fold decrease) and rival oncogene induction by doxcycyline lines were confirmed by western blotting (Fig. 6a). Phenotypically, in the SK-MEL-119 NRAS* + iVector i.e. Tet-On-vector control ( Fig. 6b; left panel) cells, SPRY4 depletion alone had minimal effects on proliferation as growth in the siNTC and siSPRY4 lines were similar in the SK-MEL-119 NRAS* + iVector control. However, upon induction of iBRAF* in SK-MEL-119 NRAS* , there was a significant arrest, as expected, with Fig. 2 Ectopic induction rival oncogene inhibit melanoma cell growth and enhance associated phenotypes. On fifth day following rival oncogene induction with doxycycline (50-100 ng/ml), three antagonistic NRAS* + iBRAF* lines (SK-MEL-119 NRAS* and WM1361 NRAS* , GMEL BRAF* ) and two neutral BRAF* + iNRAS* (SK-MEL-63 NRAS* and A375 BRAF* ) lines were assayed for their phenotypic dependence. a Senescence was detected by senescenceassociated expression of β-galactosidase (SA-β-gal) staining, (b) different phases of cell cycle were detected with PI staining and c apoptosis was detected by Annexin-V staining and d colony formation was detected by 0.1% of crystal violet in 24 well plates. Student's t test, doxycycline vs. no-doxycycline, ◇ p ≤ 0.05. Bar, 40 µm. The data presented are representative of three independent experiments the control siNTC (Fig. 6b, black circle line to gray box line). When SPRY4 was additionally depleted with siSPRY4, there was a significant rescue (Fig. 6b, gray box line to yellow inverted triangle line) from iBRAF*-mediated suppression. These results support the idea that SPRY4 can, at least in part, mediate oncogene antagonism.

SK-MEL-119 NRAS* GMEL BRAF* A375 BRAF*
Lastly, we set out to determine if SPRY4 could be linked to a marker of senescence, p21 Waf1/Cip1. Dual BRAF* + NRAS* expression upregulated p21 in three out of the four suppressed lines but not in either of the neutral lines (Fig. 6c). We next generated a tetracycline-inducible SPRY4 cassette in SK-MEL-119 NRAS* and A375 BRAF* lines (Fig. 6d) and showed that dox-mediated expression of SPRY4 led to increased p21 in the susceptible SK-MEL-119 NRAS* line but not the neutral A375 BRAF* line. Moreover, suppression of SPRY4 with siSPRY4 in the SK-MEL-119 NRAS* line partially abrogated the p21 increase which accompanied iBRAF* (Fig. 6e). These findings are consistent with p21 as one of the factors which is downstream of SPRY4 and which may contribute to secondary OIS.

Discussion
Deep tumor sequencing has uncovered a myriad of mutational signatures and patterns [10,11]. Although BRAF and NRAS are frequently mutated in human melanoma, coexistence of BRAF c.1799T>A(V600E) and NRAS c.181C>A (Q61K)/182A>G (Q61R) changes within the same melanoma tumor is essentially nonexistent [2,12] except in situations of acquired resistance to BRAF inhibitors [3]. Despite the widespread recognition of this exclusive relationship, the evolution of these innate patterns as dictated by tumor, host and environmental forces remains largely unknown. The overall objective of our studies is to better understand the underlying pathophysiology which could explain the well-established exclusivity between BRAF (V600E) and NRAS(Q61) mutations in melanoma. We primarily focused on "oncogene antagonism" since mechanistically, a trimming of double mutants would be most compatible with observed mutational pattern in vivo.
Through our analysis, we identified SPRY4 as one possible mediator of oncogene antagonism. This is supported by several lines of evidence including the (i) upregulation of SPRY4 by the rival oncogene only in cell lines exhibiting suppression and in none of the nonsuppressed lines, (ii) stronger growth inhibitory effects of SPRY4 in the antagonistic compared to neutral lines, (iii) direct tumor suppression in vivo by SPRY4 and (iv) partial rescue from oncogene antagonism with depletion of SPRY4. SPRYs and SPREDs comprise a family of proteins which are engaged in a negative regulatory loop in that both are activated by, and serve to repress, MAPK signaling [13]. SPRY4 has been shown to directly bind to and inhibit RAF1 and BRAF (WT), but not BRAF*, through the carboxy-terminal cysteine-rich domain [14,15]. Since our cell lines all harbor BRAF*, either as an innate mutation or as the rival oncogene, MAPK signaling in our experimental conditions may be functionally resistant to SPRY4-mediated suppression. Therefore, the precise molecular mechanism by which SPRY4 mediates synthetic suppression remains to be clarified but may involve upregulation of p21.
The contrasting effects of NRAS* on two distinct BRAF* lines are also worth noting. In the synthetically suppressed GMEL BRAF* + iNRAS* cells, there is a dramatic suppression of MITF and its targets (Fig. S6). Melanoma cells have been shown to be "addicted" to MITF [16] and thus the NRAS*-mediated downregulation of MITF could cooperate with the upregulation of SPRY4 to suppress growth. In oncogene-resistant A375 BRAF* cells, sterol biosynthesis appears to be activated in the context of dual oncogenesis. Both glycolytic and lipid metabolic reprogramming is now well established in cancer and is thought to allow malignant cells to adapt to a hostile microenvironment [17]. The secondary acquisition of an NRAS* mutation in a BRAF* melanoma cell has also been described in the context of therapeutic resistance [18]. In both the A375 BRAF* + iNRAS* and MGH-CH-1 BRAF* + iNRAS* lines, we did observe resistance to BRAF inhibition (Fig.  S7), suggesting that resistance can be engendered de novo even in the absence of drug.
While there were detectable effects on the cell cycle and on the apoptotic response, the dual NRAS*/BRAF* mutant state is associated with an increase in cells with SA-ß-gal and a decrease in colony formation, both of which correlate with heightened cellular senescence. We interrogated our microarray data using the GSEA Fridman senescence gene set (Fig. S8) and found that only GMEL BRAF* + iNRAS* exhibited a significantly higher level of induction of the senescence gene set relative to all other transcripts (mean log2-fold +0.068 ± 0.025 vs. −0.012 ± 0.003, p < 0.0001). (http://software.broadinstitute.org/gsea/msigdb/cards/ FRIDMAN_SENESCENCE_UP.html) These results echo a prior study in which Petti and colleagues overexpressed NRAS(Q61R) into a single BRAF(V600E) melanoma line and similarly observed the eventual onset of a senescence phenotype [6]. While oncogenes have been shown to induce senescence in primary cells [19,20], NRAS* and BRAF*mutated melanoma cells have theoretically negotiated this Fig. 3 Rival oncogene upregulates SPRY4 expression in growth suppressive melanoma cells. a The overall workflow and heat map results of microarray analysis of mRNA isolated from three melanoma cell lines as indicated. log2-fold differences as obtained by [log2 (+Dox/oncogene) − log2(no Dox/oncogene)] − [log2(+Dox/vector) − log2(no Dox/vector)]. b DAVID enrichment scores (ES) for set of all genes that were significantly increased or decreased by at least twofold among antagonistic; SK-MEL-119 NRAS* + iBRAF*, GMEL-BRAF* + iNRAS* and neutral; A375 BRAF* + iNRAS* cell lines. c DAVID enrichment of growth suppressive overlapping significantly up-and downregulated (=2 folds) genes in SK-MEL-119 NRAS* + iBRAF*, GMEL BRAF* + iNRAS* cell lines but not A375. d SPRY4 transcripts were among the most upregulated ones in both SK-MEL-119 NRAS* and Gmel BRAF* , but not A375 BRAF* cell lines checkpoint to attain a malignant and immortal state. Thus, oncogene antagonism appears distinct from primary OIS and may be more appropriately termed "secondary" OIS.
One senescence-associated gene worth mentioning is CDKN1A (p21). This cell cycle inhibitor is upregulated in senescent cells [21] and we were able to show that the  suppressor p21 [22]. These data suggest a possible mechanism whereby NRAS* + BRAF* increases SPRY4, which in turn stimulates p21 and secondary OIS (Fig. S10).
There are inherent limitations to our analyses. First, the relationship between our in vitro results and in vivo human findings remain mostly inferential since there is a lack of sufficient cohorts of dual mutant tumors to verify our results. Second, our experiments span from days to weeks but are not considered long term. Thus, slow but persistent selection against double mutants may not be detected in some of the synthetic neutral lines. Third, while we provide evidence that SPRY4 may mediate some of the synthetic suppression, it is unlikely to be the only inhibitory mediator. Other upregulated genes (e.g. ID1, DUSP5) have also been implicated in growth arresting physiologies such as senescence [25,26]. Lastly, given our biologic interest in oncogene exclusion, we have not fully examined the genetic and/ or molecular differences along with immune surveillance which dictate sensitivity or resistance to the acquisition of a rival oncogene. In conclusion, we have uncovered a range of proliferative responses to dual BRAF* and NRAS* expression in melanoma lines. Synthetic suppression could be observed in a subset of lines and could explain the clinical phenomenon of mutual exclusivity. Through comparative molecular profiling, we identified SPRY4 as a potential mediator of the arrested physiology, though diverse biologic pathways appear to be engaged in the process. As many cells exhibit markers of senescence with oncogene antagonism despite the absence of traditional gatekeepers of OIS, "secondary" OIS may in fact play a role in the evolution of mutation patterns.

Determination of the transcript copy number of BRAF and NRAS and its comparison to endogenous levels
Total mRNA was collected from doxycycline-induced and noninduced melanoma cell lines and 2 µg RNA was converted to cDNA (High Capacity RNA-to-cDNA Kit). The Wagatsuma method [29] was used to determine the exact transcript copy numbers of BRAF and NRAS, both of which was compared with two standard curves plotted with the threshold cycles (Cq values) from recombinant plasmid (rDNA) and sample cDNA diluted 1/10 five serial dilutions. The standards of sample cDNA are from Tet-On BRAF(V600E) SK-MEL-119 NRAS* designated as "SK-MEL-119 NRAS* + iBRAF*", or Tet-On NRAS(Q61R) GMEL BRAF* designated as "GMEL BRAF* + iNRAS*" cell lines induced with doxycycline 100 ng/ml for 3 days. We first obtained the standard curves of BRAF and NRAS using Fig. 6 Rival oncogene-induced growth suppression is rescued by SPRY4 silencing in vitro. a Western blot analysis and quantitative densitometry of the protein expression in SK-MEL-119 NRAS* + iBRAF*cells that ectopically express rival oncogenes and siRNA SPRY4 constructs. Total cell lysate extracts at 48 h were probed with antibodies for BRAF(V600E), BRAF(WT), NRAS(Q61R), NRAS (WT), SPRY4 and internal loading control GAPDH. b Silencing of SPRY4 abrogates the growth inhibition effect of rival oncogene expression in SK-MEL-119 NRAS* cells as compared to siRNA nontarget control (siNTC) vector. c Differential regulation of SPRY4 and p21 proteins upon second oncogene induction. d SPRY4 overexpression and (e) SPRY4 silencing effect on p21 protein expression in the presence of rival oncogene, among suppressive (Red) and nonsuppressive (Green) cell lines was confirmed by western blotting. Student's t test, doxycycline vs. no-doxycycline, ◇ p ≤ 0.05. The data presented are representative of three independent experiments recombinant plasmids (pcDNA-NRASQ61R or pcDNA-BRAFV600E) solutions. The copy numbers of recombinant plasmid (rDNA) was within the range of 101−105 copies per reaction, and the standard curve from SK-MEL-119 NRAS* + iBRAF* or from GMEL BRAF* + iNRAS* cDNA was parallel (arbitrary units) to their corresponding recombinant plasmid standard curve (Fig. S1a, b). The standard curve slopes/E-amp (amplification efficiency) for the two target genes were (BRAF: −3.2377/2.036 for the recombinant plasmid (rDNA) standard and −3.3810/1.976 for the sample cDNA solutions and NRAS: −2.9053/2.209 for the recombinant plasmid (rDNA) standard and −3.4705/ 1.941 for the sample cDNA solutions). These results confirm that samples amplify target BRAF and NRAS nearly at same efficiency as rDNA plasmid solutions, respectively.
To get transcript copy numbers, we took the efficiencies of both the standards into an account and calculated the copy number of mRNA [29,30]. Next, we determined the transcript copy number of BRAF and NRAS mRNA in a broader panel of cell lines and doxycycline induced or noninduced SK-MEL-119 NRAS* + iBRAF* and GMEL BRAF* + iNRAS* melanoma cells (Fig. S1c, d). The mRNA copies per micro liter of the sample solution at 0 −1000 ng/ml doxycycline at third day was determined to range from 1.54 × 10 6 to 99.97 × 10 6 for BRAF mRNA in SK-MEL-119 NRAS* + iBRAF* and 81.047× 10 6 to 4847.11 × 10 6 for NRAS mRNA in GMEL BRAF* + iNRAS* and RNA expression found to be well sustained in all the melanoma lines. Hence, by calculating the average value of triplicate samples at 50−100 ng/ml doxycycline in respective cell lines, the transcript copy number after induction for BRAF was from 27.66 × 10 6 to 42.48 × 10 6 in SK-MEL-119 NRAS* + iBRAF* which is nearly parallel to A373-C6 and K2 cells and for NRAS was from 1373.25 × 10 6 to 2092.42 × 10 6 which is almost parallel to A373-C6 and SK-MEL-28 cell lines. Thus, the mRNA copy number obtained with 50−100 ng/ml doxycycline were close to naïve cell lines. Therefore, the induction was in physiological range and comparable to endogenous level of many melanoma cell line (Fig. S1c, d).

Cellular colony formation assay
Melanoma cells were seeded in 12-well plates at a density of 50−200 cells per well. The media were changed every other day, and the colonies were counted at day 10-14th after staining with 0.1% of crystal violet for at least an hour as described previously [27]. The number of colonies was determined by counting entire field of view from triplicate wells for each cell line under an Olympus SZ-PT Stereoscope using a ×10 eyepiece. The data were expressed as means ± SD of at least three independent experiments.

Cell cycle analysis
Cell cycle analyses were performed to evaluate the distribution of cells in various cell cycle phases (subG1, G1, S, and G2/M) by measuring the DNA content of nuclei labeled with propidium iodide (PI) (Life Technologies). Briefly, Tet-On BRAF(V600E) cells (SK-MEL-119 NRAS* + iBRAF*, WM1361 NRAS* + iBRAF*, and SK-MEL-63 NRAS* + iBRAF*) and Tet-On NRAS(Q61R) (GMEL-BRAF* + iNRAS* and A375 BRAF* + iNRAS*) viable cells were plated in triplicates at 2.0 × 10 5 cells per well in six-well plates and incubated with and without doxycycline (0.1 µg/ ml) for 4-5 days at 37°C in humidified 5% CO 2 incubator. After trypsinization cells were harvested and prepared singlecell suspension in 1 × PBS buffer. Cells were washed twice and centrifuged at 300 × g for 5 min, resuspended~6 × 10 6 cells/ml, 500 µl cell suspension were aliquoted into two polypropylene tubes one for cell cycle and another for apoptosis. For cell cycle, cells were fixed with ice-cold 70% (v/v) ethanol drop wise while gently vortexing and kept at −20°C overnight prior to propidium iodide (PI) staining and flow cytometric analysis. Cells were centrifuged and washed twice with cold 1× PBS and added 0.5 ml of propidium iodide staining solution to cell pellet and mix well (100 µg/ml of propidium iodide and 100 µg/ml of RNase A in 0.1%Triton X-100) in 1× PBS (Sigma-Aldrich, St. Louis, MO) and incubated at room temperature in dark for 30 min [27]. Samples were analyzed by flow cytometry (BD FACS Calibur flow cytometer, BD Biosciences-US, Sparks Glencoe, MD). FlowJo, version 7.6.5 software (Ashland, OR) was used to calculate the percentages of cells in various cell cycle phases. All experiments were performed at least three times in triplicate.

Apoptosis assay
Cells were washed with PBS, incubated with Annexin-V Alexa Fluor 647 and 4′,6-diamidino-2-phenylindole (DAPI) for 15 minutes at room temperature in the dark as per the manufacturer's protocol (Life Technologies). The percentage of Annexin-V-positive cells was determined by flow cytometry BD FACSAria, (BD Biosciences-US) and results were analyzed using FlowJo, version 7.6.5 software (Ashland, OR). As stated before, the cell cycle and apoptosis assays were performed in parallel and in triplicate for each condition.

Q-PCR analysis
Cells were seeded in 10 cm plates at a density of 1.0×10 6 cells/well and allowed to adhere overnight in Tet-free growth medium. Next day, medium was replaced with or without doxycycline 50-100 ng/ml for a defined time of 4-5 days before total RNA was isolated using RNeasy isolation kit (Qiagen, Germany) according to the manufacturer's instructions. The first strand cDNA was reverse transcribed from 2 µg of RNA using high-capacity RNA to cDNA kit according to the manufacturer's instructions (Applied Biosystems, Foster City, CA, USA). Microarray data were verified by qPCR using the gene-specific primers (Invitro

Microarray preparation and analysis
Cells were seeded 24 h before treatment with or without doxycycline 50-100 ng/ml for a defined time of 4-5 days. RNA was extracted from 2 × 10 6 cells from each condition (biologic duplicates) using the RNeasy Mini Kit (Qiagen). The RNA specimens were submitted to the Massachusetts Institute of Technology (MIT) Core Facility and profiled with Affymetrix Primeview GeneChip arrays as per the standard operating procedures of the Core Facility. Probe set intensity values were converted into log2 space after adding a pseudo-count of 1. Expressed genes were those with log2(expression) > log2(100 units), which is approximately log2 ∼6.64. Only transcript probes which were expressed in at least three or more samples were used for the comparative analysis. The effect of BRAF* or NRAS* was quantified as log2-fold differences using the formula: 2 Oncogene effectlog 2-fold = [Oncogene(+Dox) log 2-expression − Oncogene(−Dox) log 2-expression ] − [Vector(+Dox) log 2expression − Vector(-Dox) log 2-expression ]. The entire expression set in log2 expression levels is shown in Table S1.

Functional enrichment analysis using DAVID
Probes that exhibited >2-fold change (+1.0 or −1.0 log2fold) for each line (Tables S2-S4) and for probes that were shared in SK-MEL-119 NRAS* + iBRAF* and GMEL BRAF* + iNRAS* but NOT A375 BRAF* + iNRAS* (Table S5) were subjected to DAVIDV6.8 (https://david.ncifcrf.gov/) functional clustering annotation under the default mode. Per DAVID's website for functional clustering, "The geometric mean (in -log scale) of member's p values in a corresponding annotation cluster, is used to rank their biological significance. Thus, the top ranked annotation groups most likely have consistent lower p values for their annotation members." Annotation clustering Enrichment Scores are presented in Tables S2-S5. The annotation cluster is designated by the first GO term within the cluster, which by convention, has the lowest p value in the cluster.

In vivo tumor growth assay
Sixteen Nod-SCID-gamma male mice of 6-week-old (Jackson Laboratory, Bar Harbor, ME) were divided into two groups, eight mice per group were injected subcutaneously (s.c.) with 0.2 million SK-MEL-119 NRASQ61R cells constitutively expressing either control CD516B2vector or CD516B2-SPRY4 in 0.1 ml 10% DMEM growth medium per flank of mice. Animals were monitored twice weekly for 7 weeks. Body weights and tumor size were measured every week as described previously [27]. Data were expressed as mean ± SD, N = 8. Tumor histology was confirmed by hematoxylin/eosin staining of formalin-fixed and paraffin-embedded tissue. Animals were maintained in well-ventilated animal facility and tested in accordance with the MGH Animal Care and Use Committee guidelines. independent public data (i.e. results not from our laboratory), these lines were labeled "CONSISTENT" with the designated line. Lastly, cell lines that were either newly developed or lacked STR information and public domain information were experimentally assayed for MITF levels by RNAseq. Those that showed MITF expression were considered "COMPA-TIBLE" with melanoma. The following represents the level of confirmation for cell lines in this study: A375 BRAF* (CON-FIRMED), GMEL BRAF* (CONFIRMED), WM1361 NRAS* (CONFIRMED), SK-MEL-63 NRAS* (CONSISTENT) and MGH-CH-1 BRAF* (COMPATIBLE). The cell lines were thawed and collected between third or fourth passages for original naive and stably expressing transductant cells. After experiment, the cell lines were either used up or discarded at 15th to 25th day. All used cell lines were mycoplasma tested as instructed by the company (Invitrogen).

Statistical analysis
Data from different experiments were represented as means ± SD from at least three independent experiments. To analyze cell viability, linear regression analysis was performed using GraphPad Prism 7.0 (GraphPad Software, La Jolla, CA). Significance was established at p = 0.05, as usual.

Animal material
The mice experiments were performed in accordance with a protocol approved by the Institutional Animal Care and Use Committee (IACUC) of MGH.