Many patients with BRAF inhibitor resistance can develop disease at new sites, suggesting that drug-induced selection pressure drives metastasis. Here, we used mass spectrometry–based phosphoproteomic screening to uncover ligand-independent EPHA2 signaling as an adaptation to BRAF inhibitor therapy that led to the adoption of a metastatic phenotype. The EPHA2-mediated invasion was AKT-dependent and readily reversible upon removal of the drug as well as through PI3K and AKT inhibition. In xenograft models, BRAF inhibition led to the development of EPHA2-positive metastases. A retrospective analysis of patients with melanoma on BRAF inhibitor therapy showed that 68% of those failing therapy develop metastases at new disease sites, compared with 35% of patients on dacarbazine. Further IHC staining of melanoma specimens taken from patients on BRAF inhibitor therapy as well as metastatic samples taken from patients failing therapy showed increased EPHA2 staining. We suggest that inhibition of ligand-independent EPHA2 signaling may limit metastases associated with BRAF inhibitor therapy.

Significance: This study provides evidence that BRAF inhibition promotes the adoption of a reversible, therapy-driven metastatic phenotype in melanoma. The cotargeting of ligand-independent EPHA2 signaling and BRAF may be one strategy to prevent the development of therapy-mediated disease at new sites. Cancer Discov; 5(3); 264–73. ©2014 AACR.

See related article by Miao et al., p. 274

This article is highlighted in the In This Issue feature, p. 213

Acquired resistance is the major factor that limits the long-term efficacy of targeted therapy in patients with melanoma (1). The patterns of disease recurrence seen at progression are complex, with 50% of individuals progressing at sites of new metastasis only, 44% at existing sites of metastasis, and 6% at both existing and new sites (2). A number of putative escape mechanisms to BRAF inhibitors have now been described, with the recovery of MAPK signaling known to occur in >70% of lesions that progress on therapy (1). The continued MAPK dependency of BRAF inhibitor–resistant melanomas led to the initiation of clinical trials to evaluate the BRAF–MEK inhibitor combination (3). Despite BRAF–MEK inhibition showing improved progression-free survival compared with BRAF inhibitor alone, resistance was still widespread (3). To date, most of the clinically validated mechanisms of BRAF and BRAF–MEK inhibitor resistance are genetic and include acquired mutations that reactivate the MAPK pathway (BRAF splice mutants, NRAS mutations, MEK1/2 mutations) as well as genetic changes that increase PI3K–AKT signaling (NF1, PTEN, PI3KCA, and AKT1; refs. 4–9). It has also been shown that increased receptor tyrosine kinase (RTK) signaling, secondary to the relief of feedback inhibition in the MAPK pathway, facilitates escape from BRAF and MEK inhibition (10, 11). There is evidence from other tumor histologies that therapeutic intervention drives the adoption of phenotypes such as an epithelial-to-mesenchymal transition (EMT; ref. 12). Although it is known that melanoma cells with intrinsic BRAF inhibitor resistance frequently show increased invasive potential, the role of chronic BRAF inhibitor treatment in mediating the dissemination of melanoma cells has been little explored (13).

Ephrins constitute the largest family of RTKs and play diverse roles in cell migration, development, and maintenance of the stem-cell niche (14, 15). Under physiologic conditions, EPHA2 kinase binds to its membrane-bound ligand ephrin-A1, leading to the inhibition of AKT and reduced cell migration (14). In cancer, EPHA2 is often overexpressed and, following phosphorylation by AKT at S897, can signal in a ligand-independent manner to drive tumorigenic behavior and increased cell migration (16, 17). In the present study, we utilized a label-free phosphoproteomic approach to quantify all of the phosphorylation events associated with long-term inhibition of BRAF signaling. These analyses revealed a previously unsuspected link between BRAF and BRAF–MEK inhibition and the adoption of an aggressive, invasive phenotype driven through ligand-independent EPHA2 signaling.

Phosphoproteomics Identifies BRAF Inhibitor Resistance to Be Associated with an Invasive Signature

We began by asking how chronic vemurafenib treatment rewired the signaling network of BRAF-mutant melanoma at a systems level through the use of a mass spectrometry–based phosphoproteomic platform. The goal of these studies was to uncover phenotypic adaptations to chronic BRAF inhibition. Our approach offers advantages over other proteomic methods such as reverse phase protein array in being comprehensive, unbiased, and not limited by antibody availability. In brief, tyrosine-phosphorylated peptides were retrieved by immunoprecipitation, and the threonine- and serine-phosphorylated peptides were isolated by subjecting the resulting flow-through to strong cation exchange (SCX) and immobilized metal affinity chromatography (IMAC). LC/MS-MS was used to quantify all of the phospho-peptides from isogenic vemurafenib-naïve and vemurafenib-resistant 1205Lu melanoma cells. Levels of peptide phosphorylation were quantified using MaxQuant and the protein–protein interactions characterized using GeneGO (18, 19). Cytoscape mapping of the global signaling changes showed an increase in both the number of nodes and the number of edges in vemurafenib-resistant 1205Lu cells (naïve: 544 nodes, 1,208 edges; resistant: 552 nodes, 1,288 edges). Chronic BRAF inhibition was associated with an enrichment of phospho-proteins involved in adhesion, cytoskeletal remodeling, focal adhesion kinase (FAK), and integrin signaling (Fig. 1A) as well as the emergence of a highly interconnected resistance interactome involving EPHA2, EGFR, EPHB4, FAK1, HDAC1, integrins (ITGA3, ITGA5, ITGAV, ITGB1, and ITGB5), nucleolin, p130CAS, paxillin, SHC1, Tensin-3, and Zyxin (Fig. 1B and C). As this suggested the adoption of a migratory/invasive phenotype, we next characterized a panel of BRAF (designated “R” cell lines) and BRAF–MEK inhibitor–resistant cell lines (designated “RR”), which were generated through chronic drug treatment for >6 months (Supplementary Fig. S1). It was noted that the resistant cell lines had increased motile behavior in both transendothelial migration assays (in which melanoma cells were allowed to migrate through confluent endothelial cell monolayers) and three-dimensional (3-D) spheroid assays (in which melanoma cells migrated into the surrounding collagen matrix; Fig. 1D; Supplementary Fig. S2, not shown). One potential candidate identified from the screen was EPHA2, a cell-surface RTK implicated in development, stem-cell niche maintenance, and cancer progression (14).

Figure 1.

Comprehensive phosphoproteomics identifies an invasive, motile signaling signature associated with BRAF inhibitor resistance. A, GeneGo enrichment analysis revealed several highly significant pathways (–log P value >2) appearing within the resistance interactome. ECM, extracellular matrix. B, further enrichment analysis of the resistance network showed that several proteins not only were found to be increased (log10 relative peak intensity) following acquired resistance but also recurred across several of the resistance pathways. C, as these proteins would broadly contribute to resistance, we conducted a connectivity analysis that revealed that the redundant nodes formed a highly connected subnetwork. D, acquired resistance to MAPK inhibition leads to increased invasion and migration as seen by modified Boyden chamber assays. Human umbilical vein endothelial cells (HUVEC) were plated onto Transwell inserts and grown to confluency. DiI-labeled (red fluorescence) naïve or resistant melanoma cells were then plated on top of the HUVEC layer and allowed to invade. Nonmigratory cells were removed, and the remaining cells were imaged by fluorescence microscopy. BRAFi, BRAF inhibitor; MEKi, MEK inhibitor.

Figure 1.

Comprehensive phosphoproteomics identifies an invasive, motile signaling signature associated with BRAF inhibitor resistance. A, GeneGo enrichment analysis revealed several highly significant pathways (–log P value >2) appearing within the resistance interactome. ECM, extracellular matrix. B, further enrichment analysis of the resistance network showed that several proteins not only were found to be increased (log10 relative peak intensity) following acquired resistance but also recurred across several of the resistance pathways. C, as these proteins would broadly contribute to resistance, we conducted a connectivity analysis that revealed that the redundant nodes formed a highly connected subnetwork. D, acquired resistance to MAPK inhibition leads to increased invasion and migration as seen by modified Boyden chamber assays. Human umbilical vein endothelial cells (HUVEC) were plated onto Transwell inserts and grown to confluency. DiI-labeled (red fluorescence) naïve or resistant melanoma cells were then plated on top of the HUVEC layer and allowed to invade. Nonmigratory cells were removed, and the remaining cells were imaged by fluorescence microscopy. BRAFi, BRAF inhibitor; MEKi, MEK inhibitor.

Close modal

Ligand-Independent EPHA2 Signaling Drives the Adoption of an Invasive Phenotype

Despite EPHA2 being implicated in the suppression of cell adhesion and migration following stimulation with ephrin-A1, it can also function in a forward signaling, ligand-independent manner following phosphorylation by AKT at S897 (see model in Supplementary Fig. S3; ref. 16). Validation of the phosphoproteomic screen through Western blotting showed increased EPHA2, phospho-EPHA2 (S897), phospho-FAK (Y397), phospho-Paxillin (Y118), and total Paxillin expression in BRAF inhibitor–resistant cell lines (Fig. 2A). Increased S897-EPHA2 expression was also observed in other BRAF inhibitor–resistant (R) cell lines as well as those with acquired BRAF–MEK inhibitor resistance (RR; Fig. 2B). An analysis of drug-naïve 1205Lu and A375 melanoma cells (which expressed low levels of EPHA2 basally) did not reveal vemurafenib to increase EPHA2 phosphorylation at S897, with the addition of the drug instead reducing phosphorylation at this site (Supplementary Fig. S4). This, along with two previous studies showing that vemurafenib and its analogue PLX4720 did not alter EPHA2 kinase activity in in vitro assays, suggested that the induction of ligand-independent EPHA2 signaling was not a direct consequence of BRAF inhibitor treatment (20, 21). In epithelial cancers, unrestricted forward EPHA2 signaling is accompanied by decreased ephrin ligand expression (16). The loss of bidirectional Ephrin-EPH signaling in cell lines with acquired BRAF and BRAF–MEK inhibitor resistance was suggested by the reduction of ephrin-A1 ligand expression (Fig. 2C) and the ability of exogenous ephrin-A1 ligand to suppress the invasion of vemurafenib-resistant 1205LuR cells in a Matrigel assay, in which cells migrated toward serum (Fig. 2D). In drug-resistant melanoma cells, siRNA knockdown of EPHA2 led to a significant (P < 0.005) abrogation of invasiveness in a Matrigel invasion assay in both the presence and absence of the drug, but did not resensitize the cells to vemurafenib-mediated apoptosis (3 μmol/L, 48 hours; Fig. 2E and F; Supplementary Figs. S5 and S6). Conversely, the introduction of EPHA2 into a melanoma cell line that lacked its expression increased cell invasion through Matrigel (Fig. 2G). To rule out that the inhibition of invasion was due to cell death, we also performed Annexin V–DAPI staining and found no differences between nontargeting and EPHA2 siRNA–treated cohorts (data not shown).

Figure 2.

BRAF and BRAF–MEK inhibition induces an invasive phenotype driven through ligand-independent EPHA2 signaling. A, Western blot validation of increased total EPHA2, phospho-EPHA2 (S897), phospho-FAK (Y397), phospho-paxillin (Y118), and paxillin expression in BRAF inhibitor–resistant (R) versus naïve (N) 1205Lu melanoma cells. B, increased EPHA2 and S897-EPHA2 expression is observed in multiple models of BRAF (R) and BRAF–MEK (RR) inhibitor resistance. C, ephrin-A1 ligand expression is decreased in cell lines with BRAF (R) and BRAF–MEK inhibitor (RR) resistance. D, ephrin-A1 ligand prevents invasion of 1205LuR cells through Matrigel. Cells were treated with ephrin-A1 ligand (1 μg/mL, 24 hours). E, siRNA (si) EPHA2 reverses the invasive phenotype of BRAF inhibitor–resistant (R) and BRAF–MEK inhibitor–resistant (RR) cell lines. siRNA knockdown of EPHA2 and representative images of reduced Matrigel invasion. F, knockdown of EPHA2 reduces invasion in the absence and presence of vemurafenib (vemu). Vemurafenib-resistant 1205LuR cells were transfected with 25 nmol/L nontargeting siRNA (NTsi) or 25 nmol/L EPHA2 siRNA in the absence of vemurafenib. After 24 hours, fresh media with 3 μmol/L vemurafenib (black bars) or DMSO (white bars) were added. G, introduction of EPHA2 enhances invasion of SK-MEL-28 cells into collagen. SK-MEL-28 cells were transfected with a plasmid containing EPHA2 and selected for 14 days. After this time, cells were plated on top of Matrigel and allowed to invade. Image shows a Z-stack of phallodin-stained melanoma cells through Matrigel invading toward serum. Bar graph, quantification of invasion relative to control. *, P ≤ 0.05; **, P ≤ 0.01; ***, P ≤ 0.001; ****, P ≤ 0.0001.

Figure 2.

BRAF and BRAF–MEK inhibition induces an invasive phenotype driven through ligand-independent EPHA2 signaling. A, Western blot validation of increased total EPHA2, phospho-EPHA2 (S897), phospho-FAK (Y397), phospho-paxillin (Y118), and paxillin expression in BRAF inhibitor–resistant (R) versus naïve (N) 1205Lu melanoma cells. B, increased EPHA2 and S897-EPHA2 expression is observed in multiple models of BRAF (R) and BRAF–MEK (RR) inhibitor resistance. C, ephrin-A1 ligand expression is decreased in cell lines with BRAF (R) and BRAF–MEK inhibitor (RR) resistance. D, ephrin-A1 ligand prevents invasion of 1205LuR cells through Matrigel. Cells were treated with ephrin-A1 ligand (1 μg/mL, 24 hours). E, siRNA (si) EPHA2 reverses the invasive phenotype of BRAF inhibitor–resistant (R) and BRAF–MEK inhibitor–resistant (RR) cell lines. siRNA knockdown of EPHA2 and representative images of reduced Matrigel invasion. F, knockdown of EPHA2 reduces invasion in the absence and presence of vemurafenib (vemu). Vemurafenib-resistant 1205LuR cells were transfected with 25 nmol/L nontargeting siRNA (NTsi) or 25 nmol/L EPHA2 siRNA in the absence of vemurafenib. After 24 hours, fresh media with 3 μmol/L vemurafenib (black bars) or DMSO (white bars) were added. G, introduction of EPHA2 enhances invasion of SK-MEL-28 cells into collagen. SK-MEL-28 cells were transfected with a plasmid containing EPHA2 and selected for 14 days. After this time, cells were plated on top of Matrigel and allowed to invade. Image shows a Z-stack of phallodin-stained melanoma cells through Matrigel invading toward serum. Bar graph, quantification of invasion relative to control. *, P ≤ 0.05; **, P ≤ 0.01; ***, P ≤ 0.001; ****, P ≤ 0.0001.

Close modal

Ligand-Independent EPHA2 Signaling Is AKT Dependent

In all cases, S897 phosphorylation of EPHA2 was mediated through PI3K–AKT, with Western blotting showing the resistance phenotype to be associated with increased phospho-AKT and phospho-GSK3β levels (Fig. 3A and Supplementary Fig. S7A) and the ability of a PI3K (GDC-0941, 3 μmol/L) and an AKT inhibitor (MK-2206, 3 μmol/L) to reverse receptor phosphorylation (Fig. 3B). Further studies showed PI3K inhibition (PI-103, 1 μmol/L) to prevent melanoma cell invasion in a 3-D spheroid assay (Fig. 3C). Treatment with the ephrin-A1 ligand restored EPHA2 tyrosine phosphorylation (Y772), and thus kinase activity, and inhibited AKT phosphorylation on S473, AKT-dependent EPHA2 phosphorylation at S897, and FAK phosphorylation at Y397 (Fig. 3D and Supplementary Fig. S7B and S7C). The potential role of FAK signaling in the drug-mediated invasive phenotype was suggested by the ability of the FAK inhibitor PF-228 to inhibit invasion in a 3-D spheroid assay (Supplementary Fig. S7C). Introduction of an EPHA2 S897A mutant that is refractory to AKT-mediated phosphorylation prevented its detection by an AKT substrate antibody (Fig. 3E) and reversed the invasion of BRAF inhibitor–resistant cells through Matrigel and their migration in a scratch assay (Fig. 3F). Although increased phosphorylation by AKT explained the ligand-independent EPHA2 signaling, it did not explain the increased EPHA2 expression in BRAF inhibitor–resistant melanoma cells. We therefore asked whether EPHA2 expression in BRAF inhibitor–resistant cell lines was dependent upon continuous drug selection pressure. Removal of BRAF or BRAF–MEK inhibition led to an initial decrease in the fitness of the resistant cells as seen by an increased doubling time, followed by the partial reversal of resistance (Supplementary Figs. S8 and S9)—an effect associated with decreased phosphorylation of both AKT and EPHA2 S897 (Supplementary Fig. S10A). Phenotypically, these effects were paralleled by a marked reduction in the invasive capacity of the cells (Supplementary Fig. S10B and S10C).

Figure 3.

Ligand-independent phospho-EPHA2 (S897) signaling is PI3K–AKT dependent. A, Western blot showing expression of phospho-AKT (pAKT), AKT, PTEN, and GAPDH in melanoma cells that are drug naïve (N), BRAF inhibitor–resistant (R), and BRAF–MEK inhibitor–resistant (RR). B, the PI3K inhibitor GDC-0941 (3 μmol/L, 24 hours) or the AKT inhibitor MK-2206 (3 μmol/L, 1 hour) decreases the phosphorylation of AKT (S473) and EPHA2 (S897). C, PI3K inhibition reduces the invasion of BRAF inhibitor–resistant melanoma cell lines. Top, 1205LuR cells were plated on top of Matrigel and treated with either vehicle or PI-103 (1 μmol/L) for 24 hours before being stained with phalloidin and imaged used confocal microscopy. Bottom, 1205LuR cells were grown as 3-D spheroids, implanted into collagen, and treated with vehicle or 1 μmol/L PI-103 before staining with calcein AM. Invading cells were imaged with an inverted fluorescence microscope, and levels of invasion were calculated with ImageJ software. D, ephrin-A1 ligand reduces EPHA2 (S897) phosphorylation levels in 1205LuR cells. Cells were treated with ephrin-A1 ligand (1 μg/mL, 24 hour) before being analyzed by Western blot. E, transfection of the EPHA2 serine to alanine substitution at position 897 mutant plasmid (S897A) prevents S897 phosphorylation of EPHA2. 1205LuR cells were transfected with either wild-type (control) or mutant S897A EPHA2 plasmid. Lysates from transfected cells were immunoprecipitated with antibodies against S897 and total EPHA2 and immunoblotted with anti–phospho-serine/threonine AKT substrate antibody. Total, nonimmunoprecipitated protein was also probed for EPHA2 and GAPDH as loading controls. F, mutant S897A expression reduces 1205LuR cell invasion and scratch wound closure. 1205LuR cells transfected with wild-type (control) or S897A mutant EPHA2 plasmid were either allowed to invade through Matrigel or plated into 6-well tissue culture plates and grown to confluency before being scratched with a p20 pipet tip, after which the wound was allowed to close over 24 hours.

Figure 3.

Ligand-independent phospho-EPHA2 (S897) signaling is PI3K–AKT dependent. A, Western blot showing expression of phospho-AKT (pAKT), AKT, PTEN, and GAPDH in melanoma cells that are drug naïve (N), BRAF inhibitor–resistant (R), and BRAF–MEK inhibitor–resistant (RR). B, the PI3K inhibitor GDC-0941 (3 μmol/L, 24 hours) or the AKT inhibitor MK-2206 (3 μmol/L, 1 hour) decreases the phosphorylation of AKT (S473) and EPHA2 (S897). C, PI3K inhibition reduces the invasion of BRAF inhibitor–resistant melanoma cell lines. Top, 1205LuR cells were plated on top of Matrigel and treated with either vehicle or PI-103 (1 μmol/L) for 24 hours before being stained with phalloidin and imaged used confocal microscopy. Bottom, 1205LuR cells were grown as 3-D spheroids, implanted into collagen, and treated with vehicle or 1 μmol/L PI-103 before staining with calcein AM. Invading cells were imaged with an inverted fluorescence microscope, and levels of invasion were calculated with ImageJ software. D, ephrin-A1 ligand reduces EPHA2 (S897) phosphorylation levels in 1205LuR cells. Cells were treated with ephrin-A1 ligand (1 μg/mL, 24 hour) before being analyzed by Western blot. E, transfection of the EPHA2 serine to alanine substitution at position 897 mutant plasmid (S897A) prevents S897 phosphorylation of EPHA2. 1205LuR cells were transfected with either wild-type (control) or mutant S897A EPHA2 plasmid. Lysates from transfected cells were immunoprecipitated with antibodies against S897 and total EPHA2 and immunoblotted with anti–phospho-serine/threonine AKT substrate antibody. Total, nonimmunoprecipitated protein was also probed for EPHA2 and GAPDH as loading controls. F, mutant S897A expression reduces 1205LuR cell invasion and scratch wound closure. 1205LuR cells transfected with wild-type (control) or S897A mutant EPHA2 plasmid were either allowed to invade through Matrigel or plated into 6-well tissue culture plates and grown to confluency before being scratched with a p20 pipet tip, after which the wound was allowed to close over 24 hours.

Close modal

Ligand-Independent EPHA2 Signaling Is Induced In Vivo Following BRAF Inhibition and Is Associated with Metastatic Dissemination

The in vivo relevance of ligand-independent EPHA2 signaling and its link to metastasis was explored in specimens of patients with melanoma. It was found that whereas 0% of primary (stage I/II) melanoma lesions (n = 12) stained strongly (+2/3) for EPHA2, 21% of metastatic lesions (stage III/IV; n = 19) showed strong staining (Fig. 4A and B; Supplementary Table S1), with 57.9% of metastatic lesions showing some strong focal S897-EPHA2 staining. We next established an in vivo resistance model in which fragments of treatment-naïve human melanoma specimens were implanted subcutaneously into nude mice [patient-derived xenografts (PDX); ref. 22]. Vemurafenib dosing was commenced over 100 to 200 days, until resistance emerged (22). Local or distant metastases were frequently observed in vemurafenib-resistant PDXs, but not in the vehicle-treated animals (as these were sacrificed at much earlier time points). Staining of matched pairs of subcutaneous and metastatic specimens from the same drug-treated animals at necropsy showed high levels of immunohistochemical staining for EPHA2 in the metastatic lesions that were absent from the matched primary subcutaneous lesions (Fig. 4C). In common with other studies, a heterogeneous pattern of EPHA2 staining was noted, with both membranous and cytoplasmic/nuclear staining observed (17).

Figure 4.

Increased total and phosphorylated (S897) EPHA2 in clinical specimens is associated with metastatic dissemination. A, representative images of sequentially sectioned primary (S897-EPHA2: +1–2, total EPHA2: +2) and metastatic (met; S897-EPHA2: +2, total EPHA2: +3) IHC-stained patient tumor specimens. B, metastasis is associated with higher tumor-wide–positive staining (+2/3) for phospho-EPHA2, increased focal S897-EPHA2 staining, and greater tumor-wide total EPHA2 staining. Representative images of phospho-EPHA2 staining of the advancing edge of two separate metastatic tumor specimens. Arrows, focal S897-EPHA2 staining at the leading edge. C, images from three matched pairs of primary and metastatic vemurafenib-treated patient-derived melanoma mouse xenografts (xeno) stained for total EPHA2, showing greatly increased levels of total EPHA2 expression in the matched metastatic versus primary lesion. Images show paired samples from three independent mice. D, vemurafenib treatment increases EPHA2 expression in specimens of patients with melanoma. Representative images of staining for EPHA2 and S897-EPHA2 in matched melanoma lesions (from patient E) before treatment and collected on vemurafenib therapy. See Supplementary Table S2 for details. E, vemurafenib resistance is associated with increased EPHA2 expression in specimens derived from a patient with matched primary melanoma and metastatic lesions that emerged on therapy. Images show IHC staining for EPHA2, S897-EPHA2, and pAKT on a matched pair of primary (before vemurafenib treatment) and subcutaneous metastatic (after vemurafenib treatment) lesions.

Figure 4.

Increased total and phosphorylated (S897) EPHA2 in clinical specimens is associated with metastatic dissemination. A, representative images of sequentially sectioned primary (S897-EPHA2: +1–2, total EPHA2: +2) and metastatic (met; S897-EPHA2: +2, total EPHA2: +3) IHC-stained patient tumor specimens. B, metastasis is associated with higher tumor-wide–positive staining (+2/3) for phospho-EPHA2, increased focal S897-EPHA2 staining, and greater tumor-wide total EPHA2 staining. Representative images of phospho-EPHA2 staining of the advancing edge of two separate metastatic tumor specimens. Arrows, focal S897-EPHA2 staining at the leading edge. C, images from three matched pairs of primary and metastatic vemurafenib-treated patient-derived melanoma mouse xenografts (xeno) stained for total EPHA2, showing greatly increased levels of total EPHA2 expression in the matched metastatic versus primary lesion. Images show paired samples from three independent mice. D, vemurafenib treatment increases EPHA2 expression in specimens of patients with melanoma. Representative images of staining for EPHA2 and S897-EPHA2 in matched melanoma lesions (from patient E) before treatment and collected on vemurafenib therapy. See Supplementary Table S2 for details. E, vemurafenib resistance is associated with increased EPHA2 expression in specimens derived from a patient with matched primary melanoma and metastatic lesions that emerged on therapy. Images show IHC staining for EPHA2, S897-EPHA2, and pAKT on a matched pair of primary (before vemurafenib treatment) and subcutaneous metastatic (after vemurafenib treatment) lesions.

Close modal

Analysis of specimens from a cohort of 6 patients with melanoma receiving vemurafenib therapy (14–370 days) revealed increased EPHA2 and S897-EPHA2 expression in the majority of on-treatment and postrelapse samples that were lacking in the pretreatment lesions (5/6; Supplementary Table S2; Fig. 4D shows pre- and posttreatment specimens from patient E). Chronic treatment of A375 cells with vemurafenib (1 μmol/L, 1–4 weeks) further supported the rapid induction of EPHA2 and S897-EPHA2 expression following BRAF inhibition, with increased EPHA2 protein expression observed by 3 weeks and increased EPHA2 and S897-EPHA2 expression seen after 4 weeks of drug treatment (Supplementary Fig. S11). It thus seemed that induction of EPHA2 expression was a relatively rapid adaptation to BRAF inhibition, which may precede the development of acquired drug resistance.

A potential link between acquired resistance and the role of BRAF inhibition in the emergence of new metastases was suggested by the observation that patients failing vemurafenib therapy (n = 28) had a significantly higher incidence of tumor growth at new, distant disease sites (68%) rather than the regrowth of existing lesions, compared with patients receiving dacarbazine (n = 20; 35%; Supplementary Table S3). These differences were seen despite the time on therapy being similar (147.5 days vs. 101.5 days for vemurafenib and dacarbazine, respectively; P = 0.1379; Supplementary Table S3). We next determined whether EPHA2 expression increased in new metastatic lesions that arose on BRAF inhibitor therapy. The IHC staining of a pair of matched pretreatment and therapy-derived metastases from a patient failing vemurafenib therapy showed high levels of EPHA2, phospho-EPHA2, and phospho-AKT expression in the treatment-derived metastatic lesions (Fig. 4E), with much less EPHA2 expression observed in the pretreatment primary lesion.

Hostile microenvironments, such as hypoxia, metabolic changes, and nutrient deprivation, favor the detachment of cancer cells and their migration to more favorable niches (23, 24). We here demonstrate that chronic BRAF inhibition induces a migratory, invasive phenotype in melanoma cells. In epithelial tumors, therapeutic adaptation leads to an EMT, which is sometimes associated with increased tumor invasiveness and metastatic spread (12, 25). Phosphoproteomic screening identified ligand-independent EPHA2 signaling as a key driver of the metastatic phenotype in melanoma cells with acquired BRAF and BRAF–MEK inhibitor resistance. In agreement with this, increased S897 phosphorylated EPHA2 was identified in on-treatment melanoma metastases. Our findings mirror those in astrocytoma, where tumor aggressiveness was correlated with high expression of S897 phosphorylated EPHA2 as well as phospho-AKT (16). In prostate cancer and glioma cell lines, ligand-independent EPHA2 signaling is known to promote invasion in an AKT-dependent manner and was reversed following treatment with ephrin-A1 ligand (16).

Previous modeling studies have demonstrated BRAF inhibitor resistance to be dependent upon continuous drug administration, with treatment withdrawal leading to tumor regression (22). It was further observed that ligand-independent EPHA2-mediated invasion also required constant BRAF and BRAF–MEK inhibitor–mediated selection pressure. The reversibility of the invasive phenotype upon drug removal suggested that ligand-independent EPHA2 signaling could be abrogated through alternate dosing schedules. There is currently some debate about whether intermittent BRAF–MEK inhibitor dosing can forestall resistance better than continuous dosing, with evidence being provided for each scenario (22, 26). There is also clinical evidence that continuation of the drug beyond the time of progression can prolong clinical benefit (22, 26). Our data support the notion that continuous drug dosing can increase the fitness and the metastatic potential of the melanoma cells. It is likely that this could be overcome through intermittent drug dosing, and this may limit the development of new metastases.

Therapeutically, tyrosine kinase inhibitors directed against EPHA2 are unlikely to reverse the migration driven through the EPHA2–AKT axis. Instead, direct targeting of the EPHA2 receptor with agonists should inhibit both ligand-independent EPHA2 signaling and AKT signaling, limiting metastatic dissemination. Targeting of EPHA2 could be an excellent strategy to increase response duration by limiting the development of new disease in patients on BRAF and BRAF–MEK inhibitor therapies.

Cell Culture and Generation of BRAF Inhibitor Resistance

Cells were cultured in 5% FBS, RPMI 1640 media. Parental 1205Lu, SK-MEL-28, and WM164 melanoma cell lines were a gift from Dr. Meenhard Herlyn (The Wistar Institute). Parental A375 cell line was purchased from the American Type Culture Collection on April 18th, 2012. Identities of all cell lines were confirmed by Biosynthesis Inc., through short tandem repeat validation analysis at 6-month intervals. The date of last validation for these studies was December 20, 2013. Dual BRAF and MEK inhibitor–resistant (RR) lines were established by chronically treating 1205Lu, SK-MEL-28, and WM164 for >6 months with 1 μmol/L each vemurafenib and selumetinib. Unless otherwise noted, single-agent vemurafenib (R) cell lines were maintained in 5% FBS media with the addition of vemurafenib at the following concentrations: 1 μmol/L for A375R, 2 μmol/L for WM164R, and 3 μmol/L for 1205LuR. Dual-agent RR inhibitor–resistant lines were maintained in 5% FBS, RPMI 1640 with 1 μmol/L vemurafenib and 1 μmol/L selumetinib.

Proliferation Assay

Assays were performed as described in ref. 27. Briefly, 4,000 cells were seeded into each well of a 96-well plate before drug treatment and allowed to attach overnight. Media containing inhibitor solubilized in DMSO, or an equivalent volume of DMSO alone, were added, and cells were incubated for 3 days before the addition of Alamar blue reagent (Invitrogen).

Inhibitors

Vemurafenib (PLX4032), selumetinib (AZD6244), GDC-0941, LBH589, PI-103, and MK-2206 were purchased from Selleck Chemicals. Dabrafenib and trametinib were from Chemie Tek.

Phosphoproteomic Analysis: Sample Processing

Naïve and vemurafenib-resistant 1205Lu cell lines were each grown in 10 15-cm tissue culture dishes. Cells were grown to approximately 70% confluency before each dish was washed with 10 mL of ice-cold PBS + 1 mmol/L orthovanadate (Sigma Aldrich). Cells were then lysed according to the manufacturer's instructions for the Phospho-Tyrosine Mouse mAb (P-Tyr-100; Cell Signaling Technology). Lysed proteins were reduced and alkylated before proteolytic digestion, and phosphorylated tyrosine–containing peptides generated from tryptic digestion were enriched with antibody-based (P-Tyr-100) immunoprecipitation. Flow-through from the immunoprecipitation was then further enriched for phosphorylated serines and threonines by SCX peptide fractionation and IMAC. The enriched phospho-tyrosine, serine, and threonine fractions were then subjected to LC/MS-MS, and the resultant tandem mass spectra were searched against SEQUEST and MASCOT databases to identify phosphoproteins. To calculate relative phospho-signal intensities, label-free protein quantification of the mass spectrometry data was analyzed using MaxQuant version 1.2.2.5 (18).

EPHA2 and S897A-Mutant Plasmid Transfection

The human EPHA2 plasmid was generated as described in ref. 28. The mutant S897A plasmid was a gift from Dr. Elena Pasquale. Naïve SK-MEL-28 cells were transfected with 4 μg of EPHA2 or control plasmid, whereas 1205LuR cells were transfected with 4 μg of S897A or wild-type EPHA2 control plasmid with Lipofectamine (Invitrogen) according to the manufacturer's instructions. Stable transfectants were harvested at 14 days, and protein expression levels were confirmed by Western blotting.

Western Blotting

Proteins were extracted and blotted as described in ref. 27. Antibodies against S897-EPHA2, Y772-EPHA2, EPHA2, Y397-FAK, FAK, Y118-PXN, PXN, S9-GSK3, GSK3β, S473-AKT, AKT, phospho-S/T AKT substrate, and PTEN were purchased from Cell Signaling Technology. Anti–ephrin-A1 was from Santa Cruz Biotechnology, and GAPDH was purchased from Sigma Aldrich.

RNA Interference

Cells were transfected as described in ref. 27 with 25 nmol/L of EPHA2 and scrambled siRNA sequences (LifeTechnologies). Cells were transfected for 24 hours in the absence of inhibitor and an additional 48 hours in the presence of inhibitor before further experimentation.

Quantitative Real-Time PCR

QRT-PCR was performed as described in ref. 27. TaqMan Gene Expression Assay Hs00171656_m1 primers/probe were used to quantify EPHA2. The 18S (P/N 4319413E) and GAPDH (Hs99999905_m1) were used to normalize EPHA2. All standards and samples were tested in triplicate, and data were analyzed using SDS software version 2.3.

3-D Spheroid Assays

Collagen-implanted spheroids were prepared using the liquid overlay method as described in ref. 27 and were treated with 100 nmol/L LBH589 or 1 μmol/L PI-103 for 120 hours before being analyzed by fluorescence microscopy. Calculations of the areas of spheroid invasion into the collagen matrix were performed with ImageJ analysis software.

Transendothelial Cell Migration Assays

Migration assays were performed as described in ref. 29. Human umbilical vein endothelial cells (HUVEC) were plated into Transwell inserts and grown to confluence. DiI-labeled naïve or resistant melanoma cell lines were plated on top of the HUVEC layer and allowed to invade for 1 to 4 hours. Nonmigratory cells were removed before imaging with an inverted Nikon Eclipse TS100 microscope. Calculations of the percent spheroid invasion into the collagen matrix were performed with ImageJ analysis software.

Scratch Wound Assays

S897A and control plasmid–transfected 1205LuR melanoma cells were grown to confluency before scratching with a p20 pipette tip. Wounds were imaged at 0 and 24 hours, and percentage wound closure was calculated using ImageJ software 1.46r.

Matrigel Invasion Assays

Cells were overlaid onto Transwell inserts coated with Matrigel (BD) and allowed to invade for 24 to 48 hours. For ephrin-A1 ligand experiments, cells were pretreated for 72 hours with 1 μg/mL ephrin-A1–Fc or IgG-Fc (R&D Systems). Cells were fixed and stained with phalloidin-AF594 and noninvasive cells removed before fluorescence imaging with an inverted Nikon Eclipse TS100 microscope. To quantify levels of invasion, fixed and stained cells were imaged with a Zeiss confocal microscope (20×) at 0 μm with 0.5-μm image slices taken throughout the distance of invasion.

Rates of New Metastasis for Vemurafenib- and Dabrafenib-Treated Patients

Patients managed at the Moffitt Cancer Center were selected from Moffitt medical records and archived melanoma specimens under the Total Cancer Care/Health & Research Informatics and Moffitt pathology systems with written informed consent being approved by the Institutional Review Board of the University of South Florida under the Declaration of Helsinki Protocols. For the vemurafenib- and dabrafenib-treated patients, the target population consisted of subjects with unresectable stage III or stage IV BRAFV600E-mutant cutaneous melanoma treated with single-agent vemurafenib or dabrafenib as the first line of therapy. A similar patient cohort was also identified who received dacarbazine as their first line of therapy. Patients who had similar numbers of restaging scans were selected for each treatment regimen to eliminate sample bias. Deidentified information pertaining to patient demographics and clinical outcome during therapy was collected on subjects with the exclusion of patients who completed less than 2 months of therapy or for whom follow-up information was not available.

Xenograft Studies

Xenograft implantation of tumor pieces from the subcutaneous tissue of a vemurafenib treatment–naïve 44-year-old male with recurrent BRAFV600E-mutant melanoma (HMEX2613) was performed as described in ref. 22 under an approved Novartis Institutional Animal Care and Use Committee protocol. Established HMEX2613 tumors were dosed continuously with 45 mg/kg of vemurafenib. After 100 to 200 days, tumors became resistant to vemurafenib, with 30% of mice being euthanized due to clinical signs. Upon necropsy, widespread metastasis was found. IHC was carried out on subcutaneous resistant tumors and matched with corresponding metastatic lesions.

Immunohistochemistry: Xenograft and Patient Samples

Slides were stained using the Ventana Discovery XT automated system (Ventana Medical Systems) per the manufacturer's instructions with proprietary reagents. Slides were deparaffinized on the automated system with EZ Prep solution (Ventana). The retrieval method used was citrate buffer at pH 6.0. Rabbit anti-mouse polyclonal IgG primary antibody for EPHA2 (Santa Cruz Biotechnology; #sc-924) that reacts to human EPHA2 was used at a 1:250 concentration in PSS antibody diluent (Ventana) and incubated for 12 minutes. Ventana anti-mouse secondary antibody was used for 12 minutes, and a Ventana OmniMap kit was used for detection. Slides were counterstained with hematoxylin. Deidentified formalin-fixed paraffin-embedded tissue samples were obtained from Moffitt Pathology, the University of Pittsburgh, and the University of Tuebingen archives under a written informed consent protocol approved by the Institutional Review Board of the University of South Florida under the Declaration of Helsinki Protocols and stained as above for the xenograft samples. Staining was visualized using the Ventana Chromomap Red Kit. Slides were analyzed by two independent observers and consensus scored on a scale from 0 to +3.

Statistical Analysis

Data show the mean of at least three independent experiments ± SEM, unless stated otherwise. GraphPad Prism 6 statistical software was used to perform the two-tailed Student t test and for contingency analyses of patient data (two-tailed Fisher exact test). For all statistical analyses, asterisks indicate *P ≤ 0.05; **P ≤ 0.01; ***P ≤ 0.001; ****P ≤ 0.0001.

J.M. Kirkwood is a consultant/advisory board member for BMS, GSK, and Merck. G.T. Gibney is a consultant/advisory board member for Genentech/Roche. D. Stuart is a Senior Investigator II and has ownership interest (including patents) in Novartis. No potential conflicts of interest were disclosed by the other authors.

Conception and design: K.H.T. Paraiso, F. Meier, D. Stuart, K.S.M. Smalley

Development of methodology: K.H.T. Paraiso, M.D. Thakur, H. Tsao, J.L. Messina, U.N. Rao, G.T. Gibney, D. Stuart, K.S.M. Smalley

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): K.H.T. Paraiso, M.D. Thakur, B. Fang, J.M. Koomen, I.V. Fedorenko, J.K. John, V.K. Sondak, J.L. Messina, J.M. Kirkwood, F. Meier, S. Sloot, G.T. Gibney, H. Tawbi

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): K.H.T. Paraiso, M.D. Thakur, B. Fang, J.M. Koomen, I.V. Fedorenko, K.T. Flaherty, V.K. Sondak, J.M. Kirkwood, F. Meier, S. Sloot, G.T. Gibney, D. Stuart, H. Tawbi, K.S.M. Smalley

Writing, review, and/or revision of the manuscript: K.H.T. Paraiso, M.D. Thakur, B. Fang, J.M. Koomen, H. Tsao, K.T. Flaherty, V.K. Sondak, J.L. Messina, E.B. Pasquale, A. Villagra, J.M. Kirkwood, F. Meier, G.T. Gibney, D. Stuart, H. Tawbi, K.S.M. Smalley

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): K.H.T. Paraiso, B. Fang, V.K. Sondak, F. Meier

Study supervision: K.H.T. Paraiso, V.K. Sondak, D. Stuart, K.S.M. Smalley

Other (helped with the design of some experiments and provided a cDNA construct): E.B. Pasquale

Other (performed diagnosis and review, interpretation, and analysis of the immunohistochemical stains on tumor tissue): U.N. Rao

Work in the Smalley laboratory is supported by R01 CA161107-01 and SPORE grant P50 CA168536-01A1 from the NIH. S. Sloot received a grant from the Melanoma and Sarcoma Groningen Foundation. I.V. Fedorenko was supported by the Joanna M. Nicolay Melanoma Foundation.

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

1.
Chapman
PB
,
Hauschild
A
,
Robert
C
,
Haanen
JB
,
Ascierto
P
,
Larkin
J
, et al
Improved survival with vemurafenib in melanoma with BRAF V600E mutation
.
N Engl J Med
2011
;
364
:
2507
16
.
2.
Menzies
AM
,
Haydu
LE
,
Carlino
MS
,
Azer
MW
,
Carr
PJ
,
Kefford
RF
, et al
Inter- and intra-patient heterogeneity of response and progression to targeted therapy in metastatic melanoma
.
PLoS ONE
2014
;
9
:
e85004
.
3.
Flaherty
KT
,
Infante
JR
,
Daud
A
,
Gonzalez
R
,
Kefford
RF
,
Sosman
J
, et al
Combined BRAF and MEK inhibition in melanoma with BRAF V600 mutations
.
N Engl J Med
2012
;
367
:
1694
703
.
4.
Shi
H
,
Hong
A
,
Kong
X
,
Koya
RC
,
Song
C
,
Moriceau
G
, et al
A novel AKT1 mutant amplifies an adaptive melanoma response to BRAF inhibition
.
Cancer Discov
2014
;
4
:
69
79
.
5.
Shi
H
,
Hugo
W
,
Kong
X
,
Hong
A
,
Koya
RC
,
Moriceau
G
, et al
Acquired resistance and clonal evolution in melanoma during BRAF inhibitor therapy
.
Cancer Discov
2014
;
4
:
80
93
6.
Poulikakos
PI
,
Persaud
Y
,
Janakiraman
M
,
Kong
X
,
Ng
C
,
Moriceau
G
, et al
RAF inhibitor resistance is mediated by dimerization of aberrantly spliced BRAF(V600E)
.
Nature
2011
;
480
:
387
90
.
7.
Nazarian
R
,
Shi
H
,
Wang
Q
,
Kong
X
,
Koya
RC
,
Lee
H
, et al
Melanomas acquire resistance to B-RAF(V600E) inhibition by RTK or N-RAS upregulation
.
Nature
2010
;
468
:
973
7
.
8.
Paraiso
KH
,
Xiang
Y
,
Rebecca
VW
,
Abel
EV
,
Chen
YA
,
Munko
AC
, et al
PTEN loss confers BRAF inhibitor resistance to melanoma cells through the suppression of BIM expression
.
Cancer Res
2011
;
71
:
2750
60
.
9.
Wagle
N
,
Van Allen
EM
,
Treacy
DJ
,
Frederick
DT
,
Cooper
ZA
,
Taylor-Weiner
A
, et al
MAP kinase pathway alterations in BRAF-mutant melanoma patients with acquired resistance to combined RAF/MEK inhibition
.
Cancer Discov
2014
;
4
:
61
8
.
10.
Abel
EV
,
Basile
KJ
,
Kugel
CH
 3rd
,
Witkiewicz
AK
,
Le
K
,
Amaravadi
RK
, et al
Melanoma adapts to RAF/MEK inhibitors through FOXD3-mediated upregulation of ERBB3
.
Eur J Clin Invest
2013
;
123
:
2155
68
.
11.
Lito
P
,
Pratilas
CA
,
Joseph
EW
,
Tadi
M
,
Halilovic
E
,
Zubrowski
M
, et al
Relief of profound feedback inhibition of mitogenic signaling by RAF inhibitors attenuates their activity in BRAFV600E melanomas
.
Cancer Cell
2012
;
22
:
668
82
.
12.
Yu
M
,
Bardia
A
,
Wittner
BS
,
Stott
SL
,
Smas
ME
,
Ting
DT
, et al
Circulating breast tumor cells exhibit dynamic changes in epithelial and mesenchymal composition
.
Science
2013
;
339
:
580
4
.
13.
O'Connell
MP
,
Marchbank
K
,
Webster
MR
,
Valiga
AA
,
Kaur
A
,
Vultur
A
, et al
Hypoxia induces phenotypic plasticity and therapy resistance in melanoma via the tyrosine kinase receptors ROR1 and ROR2
.
Cancer Discov
2013
;
3
:
1378
93
.
14.
Pasquale
EB
. 
Eph receptors and ephrins in cancer: bidirectional signalling and beyond
.
Nat Rev Cancer
2010
;
10
:
165
80
.
15.
Murai
KK
,
Pasquale
EB
. 
Restraining stem cell niche plasticity: a new talent of Eph receptors
.
Cell Stem Cell
2010
;
7
:
647
8
.
16.
Miao
H
,
Li
DQ
,
Mukherjee
A
,
Guo
H
,
Petty
A
,
Cutter
J
, et al
EphA2 mediates ligand-dependent inhibition and ligand-independent promotion of cell migration and invasion via a reciprocal regulatory loop with akt
.
Cancer Cell
2009
;
16
:
9
20
.
17.
Binda
E
,
Visioli
A
,
Giani
F
,
Lamorte
G
,
Copetti
M
,
Pitter
KL
, et al
The EphA2 receptor drives self-renewal and tumorigenicity in stem-like tumor-propagating cells from human glioblastomas
.
Cancer Cell
2012
;
22
:
765
80
.
18.
Cox
J
,
Hein
MY
,
Luber
CA
,
Paron
I
,
Nagaraj
N
,
Mann
M
. 
Accurate proteome-wide label-free quantification by delayed normalization and maximal peptide ratio extraction, termed MaxLFQ
.
Mol Cell Proteomics
2014
;
13
:
2513
26
.
19.
Fedorenko
IV
,
Fang
B
,
Munko
AC
,
Gibney
GT
,
Koomen
JM
,
Smalley
KS
. 
Phosphoproteomic analysis of basal and therapy-induced adaptive signaling networks in BRAF and NRAS mutant melanoma
.
Proteomics
2014 Oct 22
.
[Epub ahead of print]
.
20.
Bollag
G
,
Hirth
P
,
Tsai
J
,
Zhang
J
,
Ibrahim
PN
,
Cho
H
, et al
Clinical efficacy of a RAF inhibitor needs broad target blockade in BRAF-mutant melanoma
.
Nature
2010
;
467
:
596
9
.
21.
Davis
MI
,
Hunt
JP
,
Herrgard
S
,
Ciceri
P
,
Wodicka
LM
,
Pallares
G
, et al
Comprehensive analysis of kinase inhibitor selectivity
.
Nat Biotechnol
2011
;
29
:
1046
51
.
22.
Das Thakur
M
,
Salangsang
F
,
Landman
AS
,
Sellers
WR
,
Pryer
NK
,
Levesque
MP
, et al
Modelling vemurafenib resistance in melanoma reveals a strategy to forestall drug resistance
.
Nature
2013
;
494
:
251
5
.
23.
Friedl
P
,
Alexander
S
. 
Cancer invasion and the microenvironment: plasticity and reciprocity
.
Cell
2011
;
147
:
992
1009
.
24.
Lee
SL
,
Rouhi
P
,
Dahl Jensen
L
,
Zhang
D
,
Ji
H
,
Hauptmann
G
, et al
Hypoxia-induced pathological angiogenesis mediates tumor cell dissemination, invasion, and metastasis in a zebrafish tumor model
.
Proc Natl Acad Sci U S A
2009
;
106
:
19485
90
.
25.
Sharma
SV
,
Lee
DY
,
Li
B
,
Quinlan
MP
,
Takahashi
F
,
Maheswaran
S
, et al
A chromatin-mediated reversible drug-tolerant state in cancer cell subpopulations
.
Cell
2010
;
141
:
69
80
.
26.
Carlino
MS
,
Gowrishankar
K
,
Saunders
CA
,
Pupo
GM
,
Snoyman
S
,
Zhang
XD
, et al
Antiproliferative effects of continued mitogen-activated protein kinase pathway inhibition following acquired resistance to BRAF and/or MEK inhibition in melanoma
.
Mol Cancer Ther
2013
;
12
:
1332
42
.
27.
Haarberg
HE
,
Paraiso
KH
,
Wood
E
,
Rebecca
VW
,
Sondak
VK
,
Koomen
JM
, et al
Inhibition of Wee1, AKT, and CDK4 underlies the efficacy of the HSP90 inhibitor XL888 in an in vivo model of NRAS-mutant melanoma
.
Mol Cancer Ther
2013
;
12
:
901
12
.
28.
Udayakumar
D
,
Zhang
G
,
Ji
Z
,
Njauw
CN
,
Mroz
P
,
Tsao
H
. 
EphA2 is a critical oncogene in melanoma
.
Oncogene
2011
;
30
:
4921
9
.
29.
John
JK
,
Paraiso
KH
,
Rebecca
VW
,
Cantini
LP
,
Abel
EV
,
Pagano
N
, et al
GSK3beta inhibition blocks melanoma cell/host interactions by downregulating N-Cadherin expression and decreasing FAK phosphorylation
.
J Invest Dermatol
2012
;
132
:
2818
27
.