Research ArticleCancer

BRAF Inhibitors Induce Metastasis in RAS Mutant or Inhibitor-Resistant Melanoma Cells by Reactivating MEK and ERK Signaling

See allHide authors and affiliations

Science Signaling  25 Mar 2014:
Vol. 7, Issue 318, pp. ra30
DOI: 10.1126/scisignal.2004815


Melanoma is a highly metastatic and lethal form of skin cancer. The protein kinase BRAF is mutated in about 40% of melanomas, and BRAF inhibitors improve progression-free and overall survival in these patients. However, after a relatively short period of disease control, most patients develop resistance because of reactivation of the RAF–ERK (extracellular signal–regulated kinase) pathway, mediated in many cases by mutations in RAS. We found that BRAF inhibition induces invasion and metastasis in RAS mutant melanoma cells through a mechanism mediated by the reactivation of the MEK (mitogen-activated protein kinase kinase)–ERK pathway, increased expression and secretion of interleukin 8, and induction of protease-dependent invasion. These events were accompanied by a cell morphology switch from predominantly rounded to predominantly elongated cells. We also observed similar responses in BRAF inhibitor–resistant melanoma cells. These data show that BRAF inhibitors can induce melanoma cell invasion and metastasis in tumors that develop resistance to these drugs.


The RAS–RAF–MEK [mitogen-activated protein kinase (MAPK) kinase]–ERK (extracellular signal–regulated kinase) pathway is normally activated downstream of receptor tyrosine kinases (RTKs) and regulates cell proliferation, differentiation, apoptosis, and survival (1). This pathway is often constitutively activated in cancer, promoting proliferation, survival, and, consequently, tumor progression. Furthermore, this pathway is particularly important in melanoma, a form of skin cancer that develops from pigment-producing cells called melanocytes. Notably, NRAS and KRAS are mutated in about 20 and 2% of cases, respectively, whereas BRAF is mutated in a further 40% of melanoma cases (2).

Advanced metastatic melanoma is invariably fatal and presents a median survival of 6 to 10 months and a 5-year survival rate of 10 to 15% (3), so treatments for metastatic melanoma are urgently needed. Drugs that inhibit BRAF, such as vemurafenib and dabrafenib, and drugs that inhibit MEK, such as trametinib, increase progression-free and overall survival in patients with BRAF mutant melanoma (46), thus validating BRAF and MEK as therapeutic targets in this disease. However, despite impressive initial responses, most patients develop resistance to BRAF and MEK inhibitors after a relatively short period, and about 20% of patients have primary resistance and do not respond despite the presence of BRAF mutations. Resistance is generally mediated by pathway reactivation, and multiple mechanisms have been described, including increased abundance and activity of RTKs, mutations in NRAS or MAP2K (which encodes MEK), and amplification or alternative splicing of mutant BRAF itself (7).

Unexpectedly, BRAF inhibitors can cause an intriguing phenomenon. Although they inhibit MEK-ERK signaling in BRAF mutant cells, they paradoxically activate this pathway when RAS is mutated or activated by upstream signaling from RTKs. This occurs because in these circumstances, BRAF inhibitors promote BRAF and CRAF homo- or heterodimerization, forming complexes that contain drug-bound and drug-free partners (810). The drug-bound partners are thought to activate the drug-free partners by acting as scaffolds or inducing conformational changes (810). Paradoxical ERK activation can drive tumor growth in some contexts (8, 9). For example, a kinase-inactive BRAF mutant (BRAFD594A) cooperates with oncogenic KRAS (KRASG12D) to induce melanomagenesis in mice (8), and in patients, paradoxical ERK activation can accelerate the growth of preexisting, but previously benign, secondary tumors (11, 12).

We investigated the consequences of BRAF inhibition in melanoma. Beyond driving tumor progression, we now report that genetic or pharmacological inhibition of BRAF can induce RAS mutant melanoma cell invasion and metastasis. We found that invasion was mediated by ERK–interleukin 8 (IL8)–mediated signaling and secretion of extracellular proteases, and that it was accompanied by a tumor cell morphology switch from predominantly rounded to predominantly spindle-shaped cells. Critically, we observed similar responses in BRAF mutant but inhibitor-resistant melanoma cells, suggesting that BRAF inhibitors could induce metastasis not only in RAS mutant tumors but also in BRAF mutant melanoma patients who develop resistance to BRAF inhibitors.


BRAF inhibition induces metastasis in RAS mutant tumors

To investigate the consequences of BRAF inhibition in RAS mutant tumors, we used a mouse model of melanoma mediated by conditional-inducible expression of KRASG12D and kinase-inactive BRAF (BRAFD594A) in mouse melanocytes (8). In accordance with our previous study, BRAFD594A did not induce melanoma even after 24 months, and KRASG12D did not induce tumors in the first 12 months (Fig. 1A). However, KRASG12D did induce tumors in 85% of the mice between 12 and 24 months, and when BRAFD594A and KRASG12D were expressed together, all of the mice developed tumors within 9 months at a median latency of 6 months (Fig. 1A).

Fig. 1 Kinase-inactive BRAF induces metastasis in KRASG12D tumors.

(A) Kaplan-Meier plot of tumor-free survival in 14 tamoxifen-treated Kras+/LSL-G12D;Tyr::CreERT2 (KRASG12D) mice and 25 tamoxifen-treated Braf+/LSL-D594AKras+/LSL-G12D;Tyr::CreERT2 (KRASG12D/BRAFD594A) mice compared with an amalgamated control group consisting of 8 tamoxifen-treated Braf+/LSL-D594A;Tyr::CreERT2 mice, 18 tamoxifen-treated Tyr::CreERT2 mice, 10 ethanol-treated KRASG12D mice, and 6 ethanol-treated KRASG12D/BRAFD594A mice. (B) Photographs of spleens from a KRASG12D/BRAFD594A and an age-matched wild-type mouse, representative of five mice in each group. Arrow indicates the tumor. Scale bars, 1 cm. (C) Photomicrograph of hematoxylin and eosin (H&E) and S100 staining in adjacent sections of a spleen from a KRASG12D/BRAFD594A mouse, representative of five mice. N, normal parenchyma; T, tumor. Scale bar, 0.5 mm (black). (D) Photomicrographs of sections of spleens from KRASG12D and KRASG12D/BRAFD594A mice stained with a pan-melanoma cocktail (HMB45 + MelanA/MART1 + tyrosinase; PanMel), representative of five mice. Scale bar, 100 μm. (E) Collagen invasion assay by mouse melanoma–derived cell lines from KRASG12D and KRASG12D/BRAFD594A mice. Data are means ± SEM from three independent experiments. (F) Collagen invasion by mouse melanoma–derived cells in the presence of 1 μM PLX4720 relative to dimethyl sulfoxide (DMSO)–treated controls (dotted line). Data are means ± SEM from three independent experiments; ***P < 0.001, **P < 0.01, *P < 0.05, Mann-Whitney.

Postmortem examinations revealed multinodular tumors in the spleens of 5 of 25 (20%) KRASG12D/BRAFD594A mice (Fig. 1B), but not in KRASG12D or BRAFD594A mice. The tumors presented large intraparenchymal nodules of spindle cells that stained positive for the melanoma antigen S100 and the pan-melanoma marker cocktail [which detects human melanoma black 45 (HMB45), melanoma antigen recognized by T cells (MART1), and tyrosinase] with an expansive growth pattern that largely replaced the normal spleen architecture (Fig. 1, C and D). These features are more consistent with the metastatic spread of melanoma from the skin than development of primary tumors within the spleen. Accordingly, we found that KRASG12D/BRAFD594A melanoma cells were 3 to 27 times more invasive in collagen I matrices than KRASG12D melanoma cells (Fig. 1E). This suggested that BRAF inhibition stimulated KRASG12D cell invasion, and we confirmed that the BRAF inhibitor PLX4720 increased invasion of KRASG12D and NRASG12D cells, but not of KRASG12D/BRAFD594A cells (Fig. 1F).

Thus, genetic or pharmacological inhibition of BRAF increased RAS mutant melanoma cell invasion. To examine the in vivo consequences, we treated KRASG12D melanoma cells with PLX4720 or DMSO, labeled the cells with vital dyes of different colors, and injected equal numbers of cells into the tail veins of nude mice. Thirty minutes after injection, equal numbers of DMSO (green)– and PLX4720 (orange)–treated cells occupied the lungs, whereas after 24 hours, the PLX4720-treated cells predominated (Fig. 2A). We also injected KRASG12D melanoma cells into the tail veins of nude mice and treated the mice daily with PLX4720. After 21 days, KRASG12D tumor cells replaced significantly more lung parenchyma in PLX4720-treated than in vehicle-treated control mice (Fig. 2, B and C).

Fig. 2 BRAF inhibitor induces metastasis in RASG12D tumors.

(A) Lung colonization assay of KRASG12D (K-7417) cells in nude mice. Tail veins were injected with a composite cell suspension of DMSO-treated cells labeled with green CMFDA and PLX4720-treated cells labeled with orange CMRA. The graph shows the mean percentage ± SEM of CMFDA- or CMRA-positive cells in 20 fields of cells per lung from three mice each at 30 min and 24 hours after injection; representative images are below; ***P < 0.001, Mann-Whitney. (B) Scatter plot showing the proportion (%) of lung parenchyma replaced by K-7417 cells in nude mice after tail vein injection and treatment with vehicle or PLX4720 (25 mg/kg) daily for 21 days. Each group contained six mice. The photomicrographs below show representative H&E staining. Scale bar, 0.5 cm; *P = 0.0292, Mann-Whitney. (C) Scatter plot showing lung weights from the mice in (B); **P = 0.022, Mann-Whitney. (D) Photomicrographs of H&E-stained lymph nodes from mice in (E). T, metastatic tumor; E, extracapsular spread; N, necrosis. Scale bar, 0.5 mm. (E) Photographs of lungs, kidney, and liver from a representative C57BL/6 mouse injected intradermally with 5 × 105 N-17568 and treated daily from days 10 to 40 with vehicle or PLX4720 (25 mg/kg). The arrows indicate the tumors. Images are representative of eight mice. (F) Photomicrographs of H&E-stained lungs from mice in (E). T, metastatic tumor; L, normal lung parenchyma. Scale bar, 0.5 mm.

We validated this observation using a skin metastasis model. We injected NRASG12D melanoma cells intradermally into immunocompetent mice, and after the tumors were palpable (10 days), we treated the mice daily with PLX4720 for 30 days, after which the primary tumors were removed and PLX4720 treatment halted. After a further 30 days, the mice were examined for metastasis burden. The lymph nodes in the PLX4720-treated mice were larger than those in the controls (table S1), and histology revealed a higher tumor burden in the nodes of the PLX4720-treated mice (Fig. 2D). Furthermore, only 1 of the 15 control mice developed a small tumor visible to the naked eye in the lungs, whereas 8 of the 13 PLX4720-treated mice developed multiple visible tumors in the lungs, livers, and kidneys (Fig. 2E). Histology revealed that tumors replaced no more than 10% of the normal lung parenchyma in 5 of the 15 control animals (Fig. 2F and table S1), whereas tumor tissue replaced 50 to 90% of the lung in 6 of the 13 PLX4720-treated animals and no more than 10% of the lung in 2 of these animals (Fig. 2F and table S1), suggesting that BRAF inhibitors promoted RAS mutant melanoma metastasis. To determine whether this was because of MEK-ERK pathway activation, we treated the mice with PLX4720 and the MEK inhibitor PD184352. Tumor replaced less than 10% of the lung in only one of five mice treated with PLX4720 + PD184352, and the mice in this cohort did not develop tumors in any other organs (table S1). This confirms that BRAF inhibitors induced metastasis in RAS mutant melanoma cells by activating the MEK-ERK pathway.

BRAF inhibitor–induced invasion is protease-dependent

We confirmed that PLX4720 also increased invasion of human NRAS and KRAS mutant melanoma cells in collagen matrices and that this was inhibited by PD184352 (Fig. 3A). Two other BRAF inhibitors, L779450 and SB590885, also increased RAS mutant cell invasion two- to ninefold (Fig. 3B), indicating that pharmacological inhibition of BRAF stimulated the invasive capacity of RAS mutant melanoma cells. Because previous studies implicate proteases in melanoma cell invasion (13, 14), we used quantitative mass spectrometry (MS) to identify proteases whose secretion was affected when BRAF was inhibited in RAS mutant cells.

Fig. 3 BRAF inhibitors induce protease-dependent invasion in RAS mutant melanoma cells.

(A) Collagen invasion of human NRAS, KRAS, and BRAF mutant melanoma cells treated with PLX4720 (1 μM, NRAS and KRAS cells; 0.3 μM, BRAF cells) alone or with 1 μM PD184352 as indicated. Data are means ± SEM fold invasion relative to DMSO-treated controls (dotted line) from three independent experiments; ***P < 0.001, **P < 0.01, *P < 0.05, Mann-Whitney test. ns, not significant. (B) Invasion of WM1791c and WM1366 cells in the presence of L779450 (0.3 μM) or SB590885 (0.1 μM). Data are means ± SEM fold invasion relative to DMSO-treated controls from two independent experiments; **P < 0.01, *P < 0.05, t test. (C) Schematic representation of STRING protein interaction network showing nine differentially secreted extracellular proteases from the conditioned medium (CM) of PLX4720-treated WM1791c cells. Color coding shows where evidence for the interactions between the individual proteases was obtained. (D) PLAU, PLAUR, and MMP1 mRNA abundance by quantitative reverse transcription polymerase chain reaction (qRT-PCR) in WM1791c and WM1366 cells treated with PLX4720 (1 μM, 24 hours). Data are normalized to glyceraldehyde-3-phosphate dehydrogenase (GAPDH) mRNA and presented as means ± SEM relative to DMSO-treated controls from three independent experiments; **P < 0.01, *P < 0.05, Mann-Whitney. (E) Western blot for uPA in the CM and total lysates (TL) from WM1791c and WM1366 cells treated with either DMSO (D) or PLX4720 (PLX; 1 μM, 24 hours). (F) Gel images of MMP1 activity by collagen zymography in CM from WM1791c and WM1366 cells treated as in (E). (G) Collagen invasion of human (WM1791c and WM1361) or mouse (N-790 and N-17568) RAS mutant melanoma cells in the presence of protease inhibitor (PI) cocktail with or without 1 μM PLX4720. Data are means ± SEM fold invasion relative to DMSO controls from three independent experiments; **P < 0.01, *P < 0.05, Mann-Whitney. (H) Collagen invasion of mouse KRASG12D and KRASG12D/BRAFD594A mutant melanoma cell lines treated as in (G). Data are mean fold invasion relative to DMSO-treated KRASG12D cells ± SEM from three independent experiments.

PLX4720 altered the secretion of 153 proteins in WM1791c cells (table S2A). Gene Ontology analysis using the Database for Annotation, Visualization and Integrated Discovery (DAVID; revealed that one of the most enriched groups contained proteins implicated in proteolysis (table S2B). Using the Search Tool for the Retrieval of Interacting Genes/Proteins (STRING;, we generated a connectivity map of the secreted proteases (Fig. 3C). This map revealed that the most increased proteins included components of the metalloproteinase 1 (MMP1) and plasminogen-plasmin pathways (table S2B). In RAS mutant cells, PLX4720 treatment increased the transcript abundance of PLAU [encoding urokinase-type plasminogen activator (uPA)], PLAUR [encoding uPA receptor (uPAR)], and MMP1 [encoding matrix metallopeptidase 1 (MMP1)] (Fig. 3D) and increased the secretion of uPA and MMP1 into the medium (Fig. 3, E and F). These data implicated proteases in RAS mutant melanoma cell invasion. Indeed, cotreatment with a cocktail of protease inhibitors blocked PLX4720-induced invasion by RAS mutant cells (Fig. 3G) and intrinsic invasion by KRASG12D/BRAFD594A cells (Fig. 3H).

BRAF inhibitor–induced invasion is dependent on IL8 signaling

Previous studies show that IL8 increased PLAU expression and uPA protein abundance in human skin (15) and that BRAF inhibitors increase IL8 expression and protein secretion in RAS mutant melanoma cells (16). Therefore, we investigated whether IL8 regulated RAS mutant melanoma cell invasion when BRAF was inhibited. We confirmed that PLX4720 activated ERK and increased IL8 secretion in RAS mutant melanoma cells (Fig. 4A), but inhibited ERK and reduced IL8 secretion in BRAF mutant cells (Fig. 4B). The MEK inhibitor PD184352 blocked the PLX4720-induced increase in IL8 expression in RAS mutant cells (Fig. 4C), thus corroborating that the increase in IL8 expression observed after BRAF inhibitor treatment was ERK-induced. Critically, in RAS mutant cells, exogenous IL8 increased the abundance of PLAU and MMP1 mRNA (Fig. 4D), increased uPA secretion (Fig. 4E), and stimulated invasion (Fig. 4F). However, when IL8 was knocked down by RNA interference, PLX4720 did not induce PLAU or MMP1 expression (Fig. 4G), and when either IL8 or one of its receptors (IL8RA and IL8RB) was knocked down, PLX4720 did not induce invasion (Fig. 4H), confirming a critical role for IL8 in BRAF inhibitor–induced invasion of RAS mutant melanoma.

Fig. 4 BRAF inhibitor–induced invasion is dependent on IL8 signaling.

(A and B) IL8 abundance by enzyme-linked immunosorbent assay (ELISA) in the CM of (A) WM1791c, WM1361, and WM1366 cells or (B) A375 cells treated with DMSO (D) or PLX4720 [P; 1 μM (A), 0.3 μM (B); 24 hours], and a Western blot for phosphorylated (ppERK) and total ERK in these cells. Data are means ± SEM from three independent experiments; *P < 0.05, Mann-Whitney. (C) IL8 and MMP1 mRNA amounts by qRT-PCR in WM1791c and WM1361 cells treated with PLX4720 (1 μM; 24 hours) and PD184352 (1 μM; 24 hours). Data are means ± SEM relative to DMSO-treated controls (dotted line) from three independent experiments. (D) PLAU and MMP1 mRNA abundance determined by qRT-PCR in WM1366 cells treated with IL8 (50 ng/ml; 48 hours). Data are means ± SEM relative to water-treated controls (dotted line) from three independent experiments; **P < 0.01, Mann-Whitney. (E) Western blot for uPA in the CM of WM1791c and WM1366 cells treated with DMSO (D), IL8, or PLX4720 (1 μM; 48 hours). Blot is representative of three independent experiments. (F) Collagen invasion of cells stimulated with IL8 (10 ng/ml). Data are means ± SEM fold invasion relative to that in water-treated controls from two independent experiments; **P < 0.01, *P < 0.05, Mann-Whitney. (G) PLAU and MMP1 mRNA amounts by qRT-PCR in WM1791c and WM1366 after transfection with scrambled control (SC), or IL8-targeted (siIL8) small interfering RNA (siRNA) and treated with PLX4720 (1 μM). Data are means ± SEM relative to SC controls (dotted line) from two independent experiments; **P < 0.01, Mann-Whitney. (H) Collagen invasion of WM1791c cells transfected with SC, siIL8 1 or siIL8 2, or siRNA against the IL8A or IL8B receptors (siIL8RA and siIL8RB), and treated with DMSO or PLX4720 (1 μM). Data are means ± SEM fold invasion relative to that in DMSO-treated SC controls from three independent experiments; **P < 0.01, Mann-Whitney.

BRAF inhibition induces a metastatic phenotype in RAS mutant tumors

We confirmed that PLX4720 activated ERK in RAS mutant mouse melanoma cells (Fig. 5A) and allografts (Fig. 5B and fig. S1A). In RAS mutant allografts, PLX4720 also increased the abundance of uPA (Fig. 5C) and the expression of Cxcl1 and Cxcl2 (Fig. 5D), which encode chemokine (C-X-C motif) ligand 1 (CXCL1) and CXCL2 [the mouse functional equivalents of IL8 (17)], respectively. In the presence of PLX4720, RAS mutant primary tumor cells at the tumor invasive front switched from a rounded to an elongated morphology (Fig. 5E and table S3) and were more likely to invade the tumor-stroma interface as single cells, dissecting into the collagen bundles (Fig. 5E and fig. S1B). Similarly, cells in the metastatic lung and lymph node deposits in PLX4720-treated animals adopted an elongated morphology suggestive of a highly invasive mesenchymal phenotype (Fig. 5F, fig. S1C, and table S3). Finally, the melanoma cells at the invasive front of the KRASG12D tumors presented a rounded morphology, whereas in the KRASG12D/BRAFD594A tumors, they presented an elongated morphology (Fig. 5G, fig. S1D, and table S3). Furthermore, KRASG12D/BRAFD594A tumors expressed high abundance of uPA (Fig. 5H). Therefore, these data suggest that in vivo, BRAF inhibition promotes the phosphorylation of ERK and the expression of Cxcl1 and Cxcl2. Treatment with the BRAF inhibitor also increased uPA protein abundance, and this was accompanied by switching from a rounded to a spindle shape in tumor cells, characteristic of a protease-driven mesenchymal mode of invasion.

Fig. 5 BRAF inhibition induces a metastatic phenotype in RAS mutant tumors.

(A) Western blot for phosphorylated and total ERK in RAS mutant mouse melanoma cells treated with DMSO (D), or the BRAF inhibitors PLX4720 or SB590885 (both 1 μM, 24 hours). (B) Ratio of ppERK/ERK in NRASG12D melanoma allografts (N-17568 cells) in C57BL/6 mice treated with vehicle or PLX4720 (25 mg/kg per day by oral gavage for 30 days). Data are means ± SEM from 10 tumors assayed in duplicate per group; **P < 0.01, Mann-Whitney. (C) Photomicrographs of 4′,6-diamidino-2-phenylindole (DAPI) and uPA staining in KRASG12D tumor allografts in C57BL/6 mice treated with vehicle or PLX4720. Scale bar, 300 μm. (D) Scatter plots of Cxcl1 and Cxcl2 mRNA abundance by qRT-PCR in tumors from (B). Data are means ± SEM of duplicates from six tumors per group; ***P < 0.001, Mann-Whitney. (E) Photomicrographs of H&E-stained tumor allografts from vehicle- or PLX4720-treated NRASG12D and KRASG12D tumor allografts in (B) and (C), representative of 10 allografts. T, tumor; S, peritumoral collagenous stroma. Black arrows, single tumor cells invading the stroma; white arrows, inflammatory cells. Scale bars, 200 μm. (F) Photomicrograph of H&E staining of NRASG12D tumors in the lymph nodes and lungs of the mice from (B), representative of 10 mice. T, tumor deposits; L, normal lung parenchyma; IF, invasive front. Black arrow, round cells; white arrow, elongated cells. Scale bars, 50 μm (top) and 0.5 mm (bottom). (G) Photomicrograph of H&E-stained sections of the tumor/stromal interface of cutaneous primary melanomas from KRASG12D and KRASG12D/BRAFD594A mice, representative of five mice each. T and IF as in (F); S, pericollagenous stroma. Arrows, single invasive tumor cells. Scale bar, 200 μm. (H) Photomicrographs of DAPI and uPA staining in KRASG12D/BRAFD594A tumors, representative of three tumors. Scale bar, 300 μm.

BRAF inhibitor–resistant melanoma cells display a metastatic phenotype

Our data thus far have shown that BRAF inhibitors induced RAS mutant melanoma cell invasion and metastasis, and we previously reported that drug-resistant BRAFV600E mutant melanoma cells also displayed increased invasion (18). We examined IL8 and PLAU mRNA abundance in these cells and found that in BRAF inhibitor–resistant, BRAF mutant A375 cells (A375/R), the abundance of IL8 and PLAU was increased compared to parental A375 cells (Fig. 6A). We also found that PLX4720 increased IL8 and PLAUR expression in xenografts derived from a patient whose BRAFV600E mutant tumors developed resistance to vemurafenib (Fig. 6, B and C). Additionally, PLX4720 treatment activated ERK and increased the invasive behavior of cells derived from this patient’s inhibitor-resistant BRAFV600E tumor (RM4 cells; Fig. 6D). Furthermore, we observed increased uPA abundance in tumors from BRAF mutant patients with intrinsic or acquired resistance after vemurafenib treatment (Fig. 6E and tables S4 and S5). Finally, we also observed a morphology switch from rounded to elongated cells at the invasive front of the tumors and an increased number of spindle cells invading the peritumoral stroma in the drug-resistant tumors after vemurafenib treatment (Fig. 6F, fig. S2, A and B, and table S3). Thus, the data above indicate that BRAF inhibition induced an invasive phenotype in BRAF inhibitor–resistant, BRAF mutant cells.

Fig. 6 BRAF inhibitor–resistant melanoma cells display a metastatic phenotype.

(A) IL8 and PLAU mRNA abundance by qRT-PCR in BRAFV600E mutant, PLX4720-resistant A375 cells, relative to that in A375 parental cells (dotted line) and are means ± SEM from two independent experiments each in triplicate; **P < 0.01, *P < 0.05, Mann-Whitney. (B and C) Scatter plot of IL8 (B) and PLAU (C) mRNA abundance in a vemurafenib-resistant BRAFV600E patient tumor–derived xenografts in nonobese diabetic–severe combined immunodeficient (NOD-SCID) mice. Data are means ± SEM from triplicates from 11 tumors per group; *P < 0.05, Mann-Whitney. (D) Matrigel invasion of a patient-derived vemurafenib-resistant BRAFV600E cell line (RM4) in the absence or presence of PLX4720 (1 μM). Data are means ± SEM fold invasion relative to that in DMSO-treated controls from three independent experiments; *P < 0.05, Mann-Whitney. Bottom: Western blot for phosphorylated ERK in these cells. (E) Photomicrographs of uPA-stained BRAFV600E melanoma sections before and after vemurafenib treatment in patients. T, tumor; S, stroma; dotted line, invasive front. Scale bar, 100 μm. Images are representative of whole-face sections from each tumor. (F) Photomicrographs of H&E- and HMB45-stained samples from a vemurafenib-resistant BRAFE600 melanoma patient. Left, a photomicrograph representative of three subcutaneous metastases that occurred before vemurafenib treatment; right, a photomicrograph of progressive melanoma metastases after vemurafenib treatment. Black arrows, HMB45-positive single invasive tumor cells. Scale bars, 100 μm (top) and 40 μm (bottom).


Here, we found that BRAF inhibition induced RAS mutant melanoma cell invasion and metastasis. BRAFD594A/KRASG12D cells were more invasive and metastatic than KRASG12D cells, and BRAF inhibitors induced RAS mutant melanoma cell invasion in vitro and metastasis in vivo. By MS, we showed that paradoxical activation of ERK by BRAF inhibition in RAS mutant cells increased the abundance of various extracellular proteases that are associated with melanoma metastasis (13). We confirmed that BRAF inhibitors increased uPA, uPAR, and MMP1 on a transcriptional level, subsequently increasing uPA and MMP1 secretion.

The uPA-uPAR signaling complex promotes cell migration through the degradation of extracellular matrix (19, 20), and increased expression of PLAU and PLAUR is a poor prognostic marker in cancer (2123). We showed that protease inhibitors blocked BRAF inhibitor–induced invasion by RAS mutant melanoma cells, and demonstrated that IL8 was increased at both the transcriptional and secretory levels by BRAF inhibitor–induced paradoxical activation of ERK. Further, exogenously added IL8 was sufficient to increase the expression of MMP1 and PLAU and the secretion of uPA to increase the invasive behavior of RAS mutant melanoma cells in culture. Critically, IL8 knockdown blocked BRAF inhibitor–induced increase in PLAU and MMP1 expression, and knockdown of IL8 or either of its receptors blocked BRAF inhibitor–induced invasion by RAS mutant cells. It was previously reported that IL8 mediates the invasion of melanoma cells through MMP2 (24), but our data show that in the context of paradoxical activation of the MEK-ERK pathway, IL8 also induced invasion through uPA and MMP1.

Melanoma cells invade biological matrices by “amoeboid” or “mesenchymal” migration, but can switch between these forms of movement (2527). Amoeboid migration is mediated by the forces generated by actin-myosin contractility and is associated with a characteristic round cell shape (2527). Mesenchymal movement is mediated by the secretion of proteases, the degradation of extracellular matrix, and the assembly of actin, and is associated with a characteristic spindle cell shape (28, 29). We found that BRAF inhibition induced a morphology switch in RAS mutant melanoma cells in vivo from predominantly rounded to predominantly elongated cells, an observation that is consistent with previous reports linking uPA-uPAR signaling to mesenchymal movement (30, 31).

Critically, we observed similar events in BRAF inhibitor–resistant BRAFV600E mutant melanoma cells. PLX4720 induced IL8 and PLAU expression in BRAF inhibitor–resistant BRAF mutant melanoma cell lines and induced IL8 and PLAUR expression in drug-resistant xenografts from patient samples. PLX4720 also increased invasion of cell lines derived from BRAF inhibitor–resistant patient tumors. Finally, vemurafenib increased the abundance of uPA in tumors from patients with intrinsic or acquired resistance to BRAF inhibitors, and this was accompanied by a switching from predominantly rounded to predominantly elongated cell morphology at the invasive front of the tumors.

IL8 was critical to these responses, and previous studies show that IL8 is regulated by RAS through the RAF-ERK and RAC–JNK (c-Jun N-terminal kinase) pathways signaling to the AP-1 (activating protein 1) transcription factor, as well as through phosphatidylinositol 3-kinase (PI3K)–AKT pathway signaling to nuclear factor κB (NFκB) (Fig. 7) (32). IL8 expression can also be regulated by CCAAT/enhancer binding protein α (C/EBPα), which is phosphorylated by ERK, resulting in inhibition of granulopoiesis in myeloid cells. Similarly, IL8 expression is regulated by the glucocorticoid receptor, which is phosphorylated by ERK and regulates leptin expression in breast cancer cells (33, 34), or by HNF4α, which is phosphorylated by MAPKs in human hepatoma cells (35). Thus, MAPKs appear to regulate IL8 expression in many cell types to control distinct processes, and we posit that BRAF inhibitors activate a signaling network in RAS mutant and BRAF inhibitor–resistant melanoma cells to activate ERK and increase expression of IL8, thereby stimulating protease-dependent mesenchymal invasion and metastasis (Fig. 7). This behavior mirrors the response of tumor cells to conventional DNA-damaging agents and microtubule poisons, in which after an initial response, resistance develops and is accompanied by increased invasion and a switch to an elongated morphology (36).

Fig. 7 Model of BRAF inhibitor–induced invasion.

RAS stimulates IL8 expression through the activation of NFκB by the PI3K-AKT pathway and the activation of AP-1 by the RAF-ERK and RAC-JNK pathways. When BRAF is inhibited in the presence of oncogenic RAS, the pathway is hyperactivated, leading to increased expression of IL8, PLAU, and MMP1, which encode IL8, uPA, and MMP1, respectively. This is accompanied by a switch from a rounded to elongated cell morphology and increased invasion.

Our data have clear clinical implications because reactivation of the RAF-ERK pathway underlies resistance in most melanomas, and 23% of resistant tumors carry NRAS mutations (37). Further, BRAF inhibitors appear to accelerate the growth of previously benign RAS mutant lesions in some patients (38). Thus, our data indicate that once BRAF mutant tumors become resistant to BRAF inhibitors, not only do the cells escape the growth-inhibitory effects of the drug, but their ability to metastasize is enhanced by the continued presence of the drug. This caused rapid disease progression in mice in this study and could explain the rapid deterioration observed in some patients when resistance emerges (39). Critically, BRAF inhibitor–induced metastasis was blocked by MEK inhibition, providing further support for the use of this drug combination in the clinic.


Cell culture and reagents

Human cell lines (table S6) were cultured in Dulbecco’s modified Eagle’s medium (DMEM) (Invitrogen) supplemented with 10% fetal bovine serum (FBS), 1% penicillin/streptomycin. Mouse- and human tumor–derived cell lines obtained (table S6) were cultured in DMEM/10% FBS/1% penicillin/streptomycin and primocin (0.1 mg/ml) (InvivoGen). PD184352 and PLX4720 were synthesized in-house, SB590885 and L779450 were purchased from Symansis, GM6001 was from Calbiochem, calpeptin was from Tocris Bioscience, and IL8 was from Sigma.

Cell lysis and Western blotting

Cell lysates were prepared in NP-40 buffer [150 mM NaCl, 50 mM tris (pH 7.5), 2 mM EDTA (pH 8), 25 mM, 1 mM Na3VO4, 10 mM NaF, and 1% NP-40] containing protease inhibitors (Complete, Roche). Conditioned medium was concentrated using centrifugal filter units (Amicon Ultra, Millipore). Protein bands were visualized using fluorescent-labeled secondary antibodies (LI-COR Biosciences) and analyzed on an Odyssey Infrared Scanner (LI-COR Biosciences). The antibodies used were the following: rabbit anti-uPA (H14, Santa Cruz Biotechnology), rabbit anti-ERK2 (C-14, Santa Cruz Biotechnology), mouse anti–α-tubulin (Sigma), and mouse anti–phospho-ERK2 (M8159, Sigma).

Invasion assays

Collagen invasion assays were performed as previously described (27). Briefly, 1 × 104 Superindex cells in 100 μl of serum-free collagen I at 2.3 mg/ml (PureCol, Advance BioMatrix) were dispensed into 96-well ViewPlate (PerkinElmer) coated with bovine serum albumin (BSA). The cells were sedimented at 300g and incubated at 37°C/10% CO2 for 1 hour to coagulate the collagen, then overlaid with DMEM/10% FBS (chemoattractant). After 24 hours, cells were fixed in 4% formaldehyde and stained with Hoechst 33258 (10 mg/ml) (Invitrogen). Confocal z sections were collected at the bottom of the well and at 45 μm using an INCELL 3000 high-content microscope. Nuclear staining was quantified with INCELL 3000 software with the object intensity module. Invasion indices are [number of cells at 45 μm]/[number of cells at 1 μm]. Means of triplicate samples are presented as invasion fold (invasion index of treated cells/invasion index of its control in each case). The Matrigel invasion assay was performed with the BD BioCoat Tumor Invasion System, 96-Multiwell insert plate, 8.0 μm, from BD Biosciences (354168) according to the manufacturer’s guidelines. Where protease inhibitors were used, the cocktail consisted of 20 μM GM6001, 10 μM calpeptin, aprotinin (10 μg/ml), and leupeptin (10 μg/ml).

In vivo studies

All experiments were conducted with local ethical committee approval in accordance with the United Kingdom Home Office regulations and United Kingdom Coordinating Committee on Cancer Research Guidelines (40). For transgenic mice, CreERT2 was activated by tamoxifen (Sigma), and genotyping and expression analysis were as described (41).

Lung colonization assays

For short-term lung colonization assays, 5 × 105 K-7417 cells pretreated with DMSO or 1 μM PLX for 24 hours were stained with CellTracker Green CMFDA or orange CMRA (each) (Molecular Probes, Invitrogen) according to the manufacturer’s instructions. Cells were mixed in 100 μl of phosphate-buffered saline (PBS) and injected into the tail veins of nude mice. Mice were sacrificed after 30 min, or 6 or 24 hours, and the surface of the lungs was examined for CMFDA- or CMRA-stained cells. Cell numbers (average of 20 measurements per lung, three mice per experiment) are expressed as percentage of total number of cells counted. For long-term lung colonization assays, 1.5 × 104 K-7417 cells in 100 μl of PBS were injected into the tail veins of nude mice. Mice were treated daily with vehicle (5% DMSO) or PLX4720 (25 mg/kg per day by oral gavage), and the lungs were weighed after 21 days.


Cells (5 × 105) in 0.1 ml of PBS were inoculated intradermally into the flanks of female C56BL/6 mice (Charles River). The tumors were allowed to establish and grow for 10 days, at which point mice were dosed daily by oral gavage with vehicle (5% DMSO), PLX4720 (25 mg/kg per day), without or with PD184352 (50 mg/kg per day). The tumors were removed after further 30 days. Tumor volumes were determined using volume = length × width × depth (mm) × 0.5236.

Patient-derived xenografts

Fresh tissue from patients was collected immediately after surgery. The tissue was transferred into a sterile petri dish containing DMEM and washed. Necrotic parts of the tumor were removed, and 5 mm3 pieces were implanted subcutaneously in the flanks of a NOD-SCID mouse. When the tumors reached 1.5 cm in size, they were excised, and viable tissue was dissected into 5 mm3 cubes for propagation in secondary host. One day after transplantation, mice were divided into two groups and treated daily with vehicle or PLX4720 (45 mg/kg) by oral gavage.

Real-time qRT-PCR

RNA was extracted from cell lines or tumors (tumor tissue was previously homogenized using the Precellys 24 tissue homogenizer, Bertin Technologies) by RNeasy Kits (Qiagen). RNA was reverse-transcribed to generate complementary DNA using M-MLV Reverse Transcriptase (Sigma). Real-time qRT-PCR was performed with Precision MasterMix (Primerdesign) and TaqMan Gene Expression Assay probes on an Applied Biosystems 7900HT Fast Real Time machine. Relative expression was calculated using the ΔΔCt method and GAPDH as an internal control.

siRNA transfections

WM1791c or WM1366 (2 × 105) (Superindex) cells were seeded in a 35-mm-diameter well the day before transfection. The cells were transfected with 10 nM of the following siRNA: scramble control: 5′-AAGUCCAUGGUGACAGGAGAC-3′; siRNA IL8.1: 5′-GAAGAGGGCUGAGAAUUCA-3′; siRNA IL8.2: 5′-GGACCACACUGCGCCAACA-3′; siRNA IL8RA or ILRB: SMARTpool (Dharmacon, Thermo Scientific) using INTERFERin (Polyplus Transfection). Briefly, 1 μl of siRNA (20 μM) and 6 μl of INTERFERin were diluted in 200 μl of Opti-MEM. The mix was incubated for 10 min before adding to cells on 1.8 ml of medium (DMEM–10% FBS). Forty-eight hours later, cells were used for invasion assays as indicated.


Sections (3 mm) of formalin-fixed paraffin-embedded material were used. Slides were dewaxed, and antigen retrieval was performed with citrate buffer (pH 6) followed by blocking in PBS–Tween 0.1% + 1% BSA for 15 min and overnight incubation with uPA antibody (1:100 in PBS + 1% BSA). Antibody detection was performed with Alexa Fluor–conjugated secondary antibodies (Invitrogen). Slides were counterstained with DAPI. Samples were analyzed, and pictures were taken using a Leica SP2 confocal scanning microscope (Leica Microsystems).

Enzyme-linked immunosorbent assay

Abundance of IL8 protein was quantified using human CXCL8/IL8 DuoSet ELISA Development System (R&D Systems). For treatments, 1.5 ml of fresh medium containing the drug was added, and conditioned medium was collected after 24 hours.


Mouse tumors were formalin-fixed and stained with H&E, S100 (MenaPath) as described (41), or a pan-melanoma cocktail (HMB45 + MelanA/MART1 + tyrosinase) (CM165, Biocare Medical) diluted 1:80 for 48 min. The staining was performed on a Ventana BenchMark XT machine. The antigen was retrieved using Ventana cell conditioning solution for 30 min. The mouse antibody was blocked for 32 min using mouse immunoglobulin G blocking reagent MKB-2213 (Vector Laboratories). Human samples were stained with H&E or an antibody against uPA (AF1310, R&D Systems) at a concentration of 10 μg/ml. Before incubation, the tissue was subjected to heat-induced epitope retrieval using pH 9 retrieval solution (S2367, Dako). The antibody was then detected using a biotinylated secondary antibody followed by an avidin/biotin complex alkaline phosphatase–based system. Vector Red was used as the substrate.

Histological analysis

The invasive front of the primary tumors was defined as the group of melanoma cells with at least 50% cell surface in contact with the peritumoral collagenous stroma. Shape of the cells was determined quantitatively using photomicrographs of the samples taken with a 20× lens and averaging 15 random cells at the invasive front of the primary tumors, or in contact with the peritumoral stroma in lung and lymph nodes metastasis. Shape was expressed as the ratio between the short and long diameter of the nuclei as previously described (42). For metastasis studies, in all animals in vehicle- and PLX4720-treated groups, the primary tumor, local recurrences, macroscopically enlarged lymph nodes, lungs, heart, kidneys, liver, and pancreas were examined. In PLX4720 + PD184352–treated animals, only enlarged lymph nodes, lungs, and heart were collected. We examined microscopically all available organs and assessed the presence or absence of lymph node and visceral metastases, as well as area of organ replaced by tumor cells. We also determined lymph node size by measuring the longest diameter, and quantified extent of necrosis in the tumors of the lymph nodes.

Secretome study

WM1791c cells were seeded at 80% confluence in 150-mm plates, grown for 24 hours, washed three times with PBS, and incubated in serum-free medium for an additional 24 hours in the presence of vehicle (DMSO) or PLX4720 (1 μM). Conditioned media were collected and cleared from cell and debris by centrifugation. Resulting supernatants were supplemented with 0.01 μM aprotinin and 1 μM leupeptin (Sigma), and concentrated 60-fold in Centriprep 3 (Millipore). Proteins were then precipitated with 10% trichloroacetic acid/acetone, and the resulting pellet was solubilized in standard cell lysis buffer (8 M urea). Protein concentration was determined using the bicinchoninic acid (BCA) assay (Sigma).

Solution digestion

Protein samples diluted in 8 M urea were subsequently reduced to 2 M using 100 mM ammonium acetate (pH 8) before tryptic digestion. Proteins were reduced with 10 mM dithiothreitol for 30 min at 56°C, then alkylated with 55 mM iodoacetamide for 1 hour at room temperature in the dark. Cell lysates were diluted to a final urea concentration of 1.6 M with 50 mM ammonium bicarbonate, and digested with trypsin (substrate/enzyme = 50) at 37°C overnight with end-over-end rotation. The resulting peptide solutions were acidified with 10% trifluoroacetic acid and desalted on a Waters C18 solid-phase extraction plate. Eluted peptides were divided into 100-μg aliquots, lyophilized to complete dryness, and stored at −80°C until needed.

iTRAQ labeling

For DMSO- and PLX4720-treated samples, duplicates were performed such that peptides were independently labeled and analyzed by liquid chromatography–tandem MS (LC-MS/MS) twice. Desalted peptides were labeled with iTRAQ reagents (43) according to the manufacturer’s instructions. Briefly, 100-μg aliquots of dried peptides were reconstituted in 30 μl of 0.5 M triethylammonium bicarbonate. One tube of iTRAQ reagent (AB SCIEX) was reconstituted in 70 μl of ethanol and added to each peptide solution. The reaction was allowed to proceed for 1 hour at room temperature. Derivatized peptides were combined, dried by vacuum centrifugation, and desalted on a Waters C18 solid-phase extraction plate. iTRAQ-labeled peptides were lyophilized to complete dryness and stored at –80°C until needed. Secretome samples were labeled with iTRAQ 4plex reagent (114-vehicle 1, 115-PLX4720 1, 116-1PLX47201, or 117-vehicle).

Liquid chromatography–MS/MS

Reversed-phase chromatography was performed with an HP 1200 platform (Agilent). Forty percent of samples were analyzed as 4 μl of injection. Peptides were resolved on a 75-mm internal diameter (ID) C18 PepMap column (3 μm of particle size; LC Packings/Dionex). A range of linear gradients of 96:4 to 50:50 buffer A/buffer B (buffer A: 2% acetonitrile/0.1% formic acid; buffer B: 80% acetonitrile/0.1% formic acid) at 300 nl/min were used for specific sample analyses. Peptides were ionized by electrospray ionization using 1.5 kV applied directly to the postcolumn LC eluent via a microtee built into the nanospray source. Sample was infused into an LTQ Velos Orbitrap mass spectrometer (Thermo Fisher Scientific) using a noncoated SilicaTip emitter (20-μm ID, 10-μm tapered tip; New Objective). The ion transfer tube was heated to 200°C, and the S-lens set to 60%. MS/MS was acquired using data-dependent acquisition to sequence the top 10 most intense ions using enhanced ion trap (IT) scans. Automatic gain control was set to 1,000,000 for Fourier transform MS (FT-MS) and 30,000 for IT-MS/MS, full FT-MS maximum inject time was 500 ms, and normalized collision energy was set to 35% with an activation time of 10 ms. Wideband activation was used to cofragment precursor ions undergoing neutral loss of up to –20 mass/charge ratio (m/z) from the parent ion, including loss of water/ammonia. MS/MS was acquired with a 10-ppm mass window for either 15s based on a maximum exclusion list of 500 entries.

Data processing

The iTRAQ-labeled peptides were fragmented under collision-induced dissociation conditions to display reporter ions at m/z 114.1 and 117.1. The ratios of the peak area of the iTRAQ reporter ions represent the relative abundance of the peptides in the samples, which were automatically quantified with Proteome Discoverer v1.4. For this purpose, raw MS/MS data were submitted for database searching using Proteome Discoverer v1.4 and Mascot v2.3. MS/MS spectra from each acquisition were extracted and converted into .mgf files, and then searched against Swiss-Prot database 57.14 (514,789 sequences) using Mascot. Search parameters included trypsin specificity with up to two missed cleavages, fixed carbamidomethylation on cysteine, fixed iTRAQ modification on N terminus and lysine, variable deamidation on asparagine and glutamine, variable oxidation on methionine, and variable phosphorylation on serine, threonine, and tyrosine. MS/MS-based peptide and protein identifications were grouped and validated using Scaffold v3.0 (Proteome Software Inc.). Protein identifications were automatically accepted if they contained at least two unique peptides assigned with at least 95% confidence by PeptideProphet. Raw MS/MS files can be found at

Data analysis

For iTRAQ quantification, the ratios of iTRAQ reporter ion intensities in MS/MS spectra (m/z 114.11 to 117.11) from raw data sets were used to calculate fold changes between samples. Ratios were derived by Proteome Discoverer v1.2. False discovery rate (FDR) for peptide search is a statistical value that determines the number of false identifications among all identifications. Proteome Discoverer calculates the percentage of false identifications using a separate decoy database (reverse database) that contains the reversed sequences of the protein entries. An FDR threshold of 1% was used in this study. Precursor and reporter ion window tolerance were fixed at 5 ppm and 0.05 dalton, respectively. The criteria specified for generation of peak lists include signal-to-noise ratio of 1.5 and inclusion of precursor mass range of 600 to 8000 daltons. Proteome Discoverer software performs automated statistical analysis of the results and uses unique peptides to calculate accurate relative protein quantification. The relative protein abundance ratios were then determined with respect to the control sample (vehicle) (tag 114 or 117). The ratios of all peptides corresponding to the same proteins were averaged. The resulting protein ratios were again normalized by their population median. Two technical replicates were included in each experiment (114 = control, 115 = PLX, 116 = control, and 117 = PLX), and two biological replicates were analyzed. The ratios 115:114 and 117:116 were averaged. A fold change cutoff of 1.5 was set to identify molecules whose expression was differentially regulated.

Collagen zymography

MMP1 activity was assessed in the serum-free condition medium of vehicle (DMSO)– or PLX4720 (1 μM)–treated WM1791c or WM1366 cells for 24 hours by collagen zymography as previously described (44). Electrophoresis was carried out using a Mini-PROTEAN 3 Cell system (Bio-Rad). Conditioned medium (30 μl) was mixed in nonreducing sample buffer containing SDS, glycerol, and bromophenol blue (without 2-mercaptoethanol) and subjected to electrophoresis on 10% polyacrylamide SDS gels (1.5 mm thick) containing bovine skin collagen type I (0.5 mg/ml). After electrophoresis, the polyacrylamide gels were washed three times (20 min each) in 2.5% Triton X-100 for 1 hour to remove all traces of SDS and incubated at 37°C in developing buffer containing 100 mM tris-HCl, 5 mM CaCl2, 0.005% Triton X-100, and 0.001% NaN3 (pH 8.0) for 24 hours. Gels were stained with 0.25% Coomassie brilliant blue G-250 in 50% methanol and 10% acetic acid solution and destained with 40% methanol and 10% acetic acid solution. The MMP1 activity was identified as clear zones of lysis against a blue background.


Mann-Whitney test or t test was performed for mRNA expression, fold invasive index, lung colonization assays, scatter plots, and cell shape differences and to assess whether the difference in metastatic lymph node size was significant among the treated versus vehicle animals. χ2 test was used to assess local recurrences, and Fisher’s exact test to calculate number of visceral organs with metastatic disease in treated versus vehicle-treated control groups.


Fig. S1. BRAF inhibition induces a metastatic phenotype in RAS mutant tumors.

Fig. S2. BRAFV600E inhibitor–resistant melanoma cells display a metastatic phenotype.

Table S1. BRAF inhibition induces metastasis in RASG12D mutant tumors.

Table S2. BRAF inhibitors induce protease-dependent invasion in RAS mutant melanoma cells.

Table S3. Cell shape quantification.

Table S4. Responses to vemurafenib in BRAF mutant melanoma patients.

Table S5. Abundance of uPA in patient melanoma samples before and after vemurafenib treatment.

Table S6. Cell lines and mutational status.


Acknowledgments: We thank G. Ashton and the Histology Unit for assistance with immunohistochemistry. Human research samples were obtained from the Manchester Cancer Research Centre (MCRC) Biobank. Although the MCRC Biobank provides the samples, it cannot endorse studies performed with, or interpretation of results from the same. We thank H. Paterson and D. Robertson for assistance with microscopy, and A. Thomson and A. Paul from the Institute of Cancer Research (ICR) proteomics core facility for help with the MS experiments and analysis. Funding: This work was supported by a Federation of European Biochemical Societies Long-Term Fellowship (B.S.-L.), Cancer Research UK (refs: C5759/A12738, C107/A10433, and C309/A2187), the European Organization for Research and Treatment of Cancer (EORTC) Melanoma Group, the Tegger Foundation, and Wenner-Gren Foundations, Stockholm (M.P.), International Association of Cancer Research (ref: 09-0773), the Harry J. Lloyd Charitable Trust, National Health Service (NHS) funding to the National Institute for Health Research Biomedical Research Centre at the Royal Marsden Hospital (J.L. and M.G.), ICR, and the Cancer Research UK Manchester Institute. Author contributions: B.S.-L. designed and performed the experiments, analyzed the data, and wrote the paper. R.M. supervised the project, analyzed the data, and wrote the paper. A.V. performed the histopathological analysis of the tumors and scorings. M.R.G. performed the proteomics experiments and analyzed the data. G.S. performed some cell-based experiments. M.P. provided tumor-derived cell lines. A.Z., D.N.-D., and C.S. synthesized and provided the PLX4720 and PD184352 inhibitors. S.T., A.H., M.G., J.L., and P.L. provided clinical samples. M.C. analyzed the data. Competing interests: All ICR authors are part of a “Rewards to Inventors Scheme” that could provide financial benefit to programs that are subsequently commercialized. Data and materials availability: MS data are available at
View Abstract

Stay Connected to Science Signaling

Navigate This Article