Research ArticleCancer

The transcription cofactor c-JUN mediates phenotype switching and BRAF inhibitor resistance in melanoma

See allHide authors and affiliations

Science Signaling  18 Aug 2015:
Vol. 8, Issue 390, pp. ra82
DOI: 10.1126/scisignal.aab1111

Inhibiting two MAPKs is better than one

Mitogen-activated protein kinase (MAPK) pathways, such as the ERK and JNK pathways, mediate critical cellular processes, such as survival, stress responses, and proliferation. These pathways can be hijacked by cancer cells, leading to uncontrolled cell division and metastasis. Many melanoma patients have activating mutations in an upstream kinase, BRAF, in the ERK pathway, but inhibitors of BRAF only produce short-term improvement. Using panels of melanoma cell lines and BRAF inhibitor–treated patient samples, Ramsdale et al. found that increased abundance of the transcription cofactor c-JUN, which is activated by the JNK pathway, mediated both inherent and adaptive resistance to BRAF inhibitors and contributed to metastatic potential. Blocking c-JUN abundance or its activation by the kinase JNK enhanced the efficacy of BRAF inhibitors against melanoma cells. Thus, targeting both MAPK pathways may overcome resistance to treatment with only inhibitors of the ERK pathway.

Abstract

Most patients with BRAF-mutant metastatic melanoma display remarkable but incomplete and short-lived responses to inhibitors of the BRAF kinase or the mitogen-activated protein kinase kinase (MEK), collectively BRAF/MEK inhibitors. We found that inherent resistance to these agents in BRAFV600-mutant melanoma cell lines was associated with high abundance of c-JUN and characteristics of a mesenchymal-like phenotype. Early drug adaptation in drug-sensitive cell lines grown in culture or as xenografts, and in patient samples during therapy, was consistently characterized by down-regulation of SPROUTY4 (a negative feedback regulator of receptor tyrosine kinases and the BRAF-MEK signaling pathway), increased expression of JUN and reduced expression of LEF1. This coincided with a switch in phenotype that resembled an epithelial-mesenchymal transition (EMT). In cultured cells, these BRAF inhibitor-induced changes were reversed upon removal of the drug. Knockdown of SPROUTY4 was sufficient to increase the abundance of c-JUN in the absence of drug treatment. Overexpressing c-JUN in drug-naïve melanoma cells induced similar EMT-like phenotypic changes to BRAF inhibitor treatment, whereas knocking down JUN abrogated the BRAF inhibitor-induced early adaptive changes associated with resistance and enhanced cell death. Combining the BRAF inhibitor with an inhibitor of c-JUN amino-terminal kinase (JNK) reduced c-JUN phosphorylation, decreased cell migration, and increased cell death in melanoma cells. Gene expression data from a panel of melanoma cell lines and a patient cohort showed that JUN expression correlated with a mesenchymal gene signature, implicating c-JUN as a key mediator of the mesenchymal-like phenotype associated with drug resistance.

Introduction

Inhibitors of the BRAF–MAPK (mitogen-activated protein kinase) kinase (BRAF/MEK) pathway are highly effective in achieving clinical responses in most patients with BRAF-mutant metastatic melanoma (15). Unfortunately, most patients display only partial responses to treatment and eventually succumb to progression of the disease due to acquired drug resistance within 6 to 12 months (68). Acquired resistance to BRAF and/or MEK inhibitors in most cases occurs through mechanisms that reactivate the ERK pathway, such as activating NRAS mutations, MEK mutations, BRAF amplification, alternative slicing of BRAF, or activation of receptor tyrosine kinases [reviewed in (9)]. Before development of acquired resistance, early adaptation in melanoma cells may facilitate cell survival during the “responsive” phase of therapy. This adaptive resistance has been reported to involve a rebound in extracellular signal–regulated kinase (ERK) activation through the release of negative feedback regulation mediated by SPROUTY2 (SPRY2) (10). Thus, the current consensus is that restoration of ERK signaling is sufficient for cell survival during drug therapy.

Recently, the phenotypic state of melanoma cells has also been linked to BRAF/MEK inhibitor resistance (1114). Phenotype switching in melanoma from a proliferative (epithelial-like) to an invasive (mesenchymal-like) state has previously been reported to be regulated by specific transcription factors such as microphthalmia-associated transcription factor (MITF), lymphoid enhancer-binding factor 1/T cell–specific transcription factor 4 (LEF1/TCF4), and Zinc Finger E-Box Binding Homeobox (ZEB1/2) (13, 1520), and has also been linked to increased WNT5A signaling in the context of BRAF/MEK inhibitor resistance (21, 22). The critical molecular mechanisms linking alterations in BRAF-MEK-ERK signaling to changes to the inducers of phenotype switching associated with drug resistance have not yet been clearly defined. It is not clear whether phenotype switching drives resistance or is associated with molecular changes within the cell that mediates survival.

We envisaged that investigating the mechanisms of early drug adaptation in cells that survive BRAF/MEK inhibitor treatment could be used to identify targets for achieving more complete and durable responses. Hence, we assessed signaling pathways, molecular features, and phenotypic characteristics that could contribute to the survival of cells during early BRAF/MEK inhibitor treatment and found that c-JUN has a critical role in mediating the changes associated with phenotype switching in melanoma and conferring a survival advantage during early therapeutic responses. These findings delineate critical links between MAPK pathway regulation, phenotypic plasticity, and drug resistance in melanoma.

RESULTS

Inherent resistance to BRAF and MEK inhibitors is associated with reduced SPRY expression, JNK pathway activation, and a mesenchymal-like phenotype

To identify potential mechanisms underlying inherent resistance to mutant BRAF targeting agents in melanoma, we assessed the responses to vemurafenib (PLX4032) (23) in a panel of 22 BRAFV600-mutant melanoma cell lines (24) and identified 4 cell lines that were intrinsically drug-resistant (IC50 > 1 μM) (Fig. 1A). The remaining cell lines displayed varying degrees of drug sensitivity (IC50 < 1 μM) (Fig. 1A). The drug-resistant cell lines also displayed resistance to the BRAF inhibitor dabrafenib (GSK2118436) (4), MEK inhibitors selumetinib (AZD6244) (25) and trametinib (GSK1120212) (5), and the ERK inhibitor SCH772984 (fig. S1A) (26). Thus, inherent resistance in these cell lines was not drug- or target-specific but likely the result of a common mechanism (or mechanisms) of resistance to inhibition of the RAF-MEK-ERK pathway.

Fig. 1 Differential gene expression and protein abundance in melanoma cell lines that are inherently sensitive or resistant to BRAF/MEK inhibitors.

(A) Heat map of vemurafenib-induced gene expression changes in a panel of melanoma cell lines, ordered by average IC50 to vemurafenib (plotted above from at least two dose-response assays; fig. S1A). An IC50 of 10−6 M was the threshold for drug sensitivity (those above were resistant, labeled in red). Color scale (left) indicates relative gene expression increase (red) or decrease (blue) across the cell lines. Microarray data are means from three independent experiments; *P < 0.05 and **P < 0.005 by Student’s t test. EMT-TFs, EMT transcription factors. (B) GeneGo network analysis for genes with differential expression for vemurafenib sensitivity (table S1) shown as a condensed pathway map. The main hubs for the pathways include the MAPK pathways, the AP-1 complex (JUN), and LEF/TCF that are linked to MITF. The legend indicates the type of protein, interaction, and effect based on the GeneGo network analysis. A white star on an object denotes a complex or multiple proteins. (C) Western blot analysis for a subset of the melanoma cell lines shown in (A), including all four drug-resistant (DrugR) cell lines. MKK, MAPK kinase kinase; ATF2, activating transcription factor 2; GAPDH, glyceraldehyde-3-phosphate dehydrogenase.

Microarray gene expression data for the cell line panel were interrogated to identify genes and pathways associated with differences between the 18 drug-sensitive cell lines and the 4 drug-resistant lines. From a set of differentially expressed genes (table S1), we identified clusters of significantly overrepresented Gene Ontology (GO) terms (table S2). The GO terms in the top-ranked cluster pertained to developmental processes, and cell differentiation and associated genes included FOS, SPRY2, WNT5A, ZEB2, TGFB1, MITF, TYR, and DUSP6. The GO terms in the second-ranked cluster included cell localization and motility, and other similarly ranked terms were “actin filament–based process” and “actin cytoskeleton organization.” Two other high-ranking GO terms (in cluster 3) related to pigmentation and included MITF and TYR. Overall, these GO term clusters point toward differences in the modulators and markers of melanoma cell phenotype and regulators of the MAPK signaling.

GeneGo analysis to link the differentially expressed genes (table S1) into networks revealed key hubs (Fig. 1B) involving MAPK signaling pathways including the ERK and c-JUN N-terminal kinase (JNK) pathways; the activator protein-1 (AP-1) complex, of which c-JUN is a key component in melanoma (27); and the LEF/TCF and MITFs, which have been linked to melanoma phenotype (15, 16, 18). Drug-resistant cell lines had significantly lower expression of ERK pathway target genes such as FOS and negative regulators such as SPRY2, DUSP4, and DUSP6 (Fig. 1A). Conversely, expression of most genes encoding components of the JNK and p38 MAPK pathways was increased in the drug-resistant cell lines, including JUN (Fig. 1A), which encodes c-JUN, a JNK target.

Phenotype switching from a proliferative to invasive state in melanoma cells resembles the epithelial-mesenchymal transition (EMT) (28) associated with some epithelial cancers, often featuring changes in expression of specific markers such as E-cadherin to N-cadherin and activation of the transforming growth factor β (TGFβ) and WNT5a signaling pathways (29). The reduction of MITF (15), a switch from LEF1 to TCF4 (16) and from ZEB2 to ZEB1 (13), is specifically implicated in melanoma phenotype switching (14). Together, the gene expression patterns of the phenotype markers, modulators, signaling pathway activators, and EMT transcription factors (Fig. 1A) revealed that most drug-sensitive cell lines were epithelial-like, and all the drug-resistant cell lines were mesenchymal-like. Consistent with the gene expression, amounts of the negative regulator proteins SPRY2 and SPRY4 were lower in drug-resistant cell lines than in drug-sensitive cell lines (Fig. 1C). Higher abundance of phosphorylated c-JUN in drug-resistant cells relative to drug-sensitive cell lines indicated increased JNK pathway activity (Fig. 1C). In addition, the abundance of LEF1 and MITF proteins was also low or undetectable in the four drug-resistant cell lines (Fig. 1C).

Although expression of genes encoding phenotype markers and EMT transcription factors (Fig. 1A) suggested that A375 cells were more similar to the mesenchymal-like drug-resistant lines than to the epithelial-like drug-sensitive cell lines, analysis by the more sensitive quantitative polymerase chain reaction (qPCR) method revealed that A375 cells had relatively lower expression of JUN, WNT5A, and SNAI2/SLUG compared to the drug-resistant cell line RPMI7951 (fig. S1B). In addition, A375 cells were inherently less migratory than RPMI7951 cells, displaying a difference in phenotype-associated cellular functional characteristics (fig. S1C). Overall, our candidate analyses confirmed that the drug-resistant cell lines had lower gene expression and protein amounts of the ERK pathway negative regulators SPRY2 and SPRY4, displayed activation of the JNK signaling pathway and higher gene expression and protein abundance of c-JUN, and expressed markers and characteristics of a mesenchymal-like phenotype.

Inhibition of JNK or reduction in c-JUN abundance enhanced drug responses in the drug-resistant cell lines

To determine the role of JNK signaling in relation to BRAF inhibitor resistance, drug-resistant cell lines were treated with vemurafenib in combination with either JNK inhibitor, SP600125, or JNK-In-XVI (JNKIn8) (table S4). Compared to vemurafenib alone, combination therapy with JNKIn8 substantially improved treatment responses in the drug-resistant cell lines (Fig. 2, A to C, and fig. S2, A to F). Calculation of excess relative activity over that expected under the Bliss independence assumption (30), as well as the excess over the highest single activity (HSA), indicated that drug synergy occurred at intermediate to high doses of both compounds. Vemurafenib induced an increase in total and phosphorylated c-JUN in the RPMI7951, HS294T, and CO57 cell lines (Fig. 2C). JNKIn8 prevented c-JUN phosphorylation in all cell lines as expected, but in RPMI7951, HS294T, and CO57 cell lines, JNKIn8 also impaired vemurafenib-induced increases in the abundance of JUN, phosphorylated c-JUN, and phosphorylated ERK. Knockdown of JUN using small interfering RNA (siRNA) before vemurafenib treatment (Fig. 2D) enhanced the cytotoxic effect of vemurafenib (Fig. 2E) and increased cell death (fig. S2F), indicating that JNK/c-JUN signaling was contributing to inherent vemurafenib resistance in the drug-resistant cell lines.

Fig. 2 Effects of combining vemurafenib with JNK inhibitors or JUN siRNA.

(A) Drug dose-response curves for vemurafenib in combination with a JNK inhibitor (JNKIn8) in the drug-resistant cell lines. (B) Percentage of excess activity over that expected under the Bliss independence assumption for each of the various dose combinations. Red indicates synergy; blue indicates antagonism. (C) Western blot analysis in lysates from the indicated cells cultured with vemurafenib (Vem), JNK inhibitor (JNKi), or both for 48 hours at the indicated doses. (D and E) Western blot analysis (D) and dimethyl sulfoxide (DMSO) or vemurafenib dose-response curves (E) in drug-resistant cell lines transfected with JUN siRNA (siJUN) 24 hours before drug exposure. Data are means ± SEM from four biological replicates.

Increased c-JUN expression and abundance attenuate the response to RAF-MEK-ERK pathway inhibitors

In most of the drug-sensitive cell lines, high doses of vemurafenib and the other BRAF/MEK/ERK inhibitors did not kill all of the cells (fig. S1A), revealing that regardless of dose escalation, there was a proportion of cells in each of those cell lines that was able to survive drug treatment. Only the dose-response curves for the cell line C089-M1 plateaued at zero, indicating complete response. This variability in drug response is consistent with the partial responses observed in most melanoma patients, with only a small proportion of cases displaying complete responses. C089-M1 was the only cell line with undetectable amount of c-JUN protein (Fig. 1C), raising the hypothesis that c-JUN provides a survival advantage in the other drug-sensitive cell lines that displayed only partial drug responses. To test this, we specifically evaluated the role of c-JUN during early drug adaptation.

Early drug adaptation mediated by SPRY2/4 down-regulation increased JUN expression in melanoma cells and patient samples

Because early drug adaptation to RAF-ERK pathway inhibitors was reported to result from a rebound of ERK activation due to decreased SPRY2 expression (10), we assessed the effects of vemurafenib treatment on SPRY and c-JUN in three drug-sensitive cell lines. In each case, SPRY2 and SPRY4 expression was reduced before an increase in JUN expression, and c-JUN phosphorylation was observed (Fig. 3, A and B). The increase in c-JUN was also detected in A375 cells treated with high doses of vemurafenib in culture (fig. S3A) as well as in xenografts in nonobese diabetic (NOD)–severe combined immunodeficient (SCID)–ILγ (NSG) mice that were fed chow containing the BRAF inhibitor PLX4720 (Fig. 3C and fig. S3B).

Fig. 3 Increased c-JUN is a consequence of SPRY4 down-regulation after inhibition of the RAF-MEK-ERK pathway.

(A and B) Abundance of SPRY2, SPRY4, and c-JUN mRNA (A) and protein (B) assessed in drug-sensitive cell lines over 72 hours in culture with 500 nM vemurafenib. Data are mean fold change ± SEM from three replicates each. Blots are representative of two independent replicates. (C) Expression of c-JUN relative to GAPDH in A375 xenografts from mice fed either control chow (circles) or chow containing PLX4720 (40 mg/kg; triangles). Data are means ± SEM from three mice each; symbols mark means from technical triplicates on each tumor. (D) Representative immunohistochemical staining for total and phosphorylated c-JUN in a melanoma sample after initial treatment with a BRAF/MEK inhibitor (details for this patient, #5, in table S3). Scale bar, 50 μm. (E) Fold change in c-JUN expression in matched patient samples treated with vemurafenib (hatched) or dabrafenib (solid) acquired before treatment (Pre), early during treatment (Early), and upon tumor progression (Prog). Data are means ± SEM from three replicates for each patient sample. (F) Change in c-JUN expression (top; rel?ative to GAPDH) and Western blot analysis (bottom) in A375 cells cultured for 48 hours with vemurafenib (V; 500 nM), MEK inhibitor selumetinib (M; 500 nM), ERK inhibitor SCH772984 (E; 100 nM), or combinations thereof compared to those cultured with DMSO. (G) Western blot analysis in A375 cells 72 hours after transfection with control (siNT) or combined SMARTpools of SPRY2- and SPRY4-targeted siRNA. Blot is representative of >3 experiments.

To determine whether these increases in JUN expression were consistent with drug-induced adaptive changes in patients, we compared melanoma samples from a cohort of patients early during treatment with vemurafenib or dabrafenib to matched pretreatment melanoma samples (table S3) (31, 32). Immunohistochemical analysis confirmed positive JUN and phosphorylated JUN staining in the melanoma cells of the patient samples (Fig. 3D and fig. S3D). There was variability in the positivity within and across the patient samples particularly between patient-matched samples before treatment and those obtained early during the treatment regimen (fig. S3D). Samples obtained early during treatment expressed significantly higher JUN mRNA relative to those obtained before treatment (Fig. 3E and fig. S3C). Thus, residual tumor cells surviving BRAF inhibition expressed higher JUN, consistent with our findings from the cell lines grown in vitro and in vivo. Furthermore, exposing A375 cells in culture to combinations of BRAF, MEK, and ERK inhibitors induced JUN expression and c-JUN activation (Fig. 3F), suggesting that the increase in c-JUN was a consequence of inhibition of the RAF-MEK-ERK pathway.

To determine whether the drug-induced change in c-JUN was linked to negative feedback regulation by SPRY proteins (10), SPRY2 and SPRY4 were knocked down using siRNA in A375 cells sans drug, which resulted in increased abundance of c-JUN and phosphorylated c-JUN (Fig. 3G). Additional experiments in multiple cell lines, including the use of siRNA to different target sequences in SPRY2, showed that the increase in c-JUN was mediated primarily through the reduction in SPRY4, although SPRY2 reduction contributed (fig. S3, E and F). Thus, the increase in c-JUN is a consequence of reduced expression of the ERK target genes SPRY2 and SPRY4 when the ERK pathway is inhibited.

Drug-induced increase in c-JUN is linked to an EMT-like phenotype switch

Consistent with phenotypic variations in the drug-sensitive and drug-resistant cell lines described above (Fig. 1), a change in the cellular morphology of A375 cells was observed during early vemurafenib adaptation (Fig. 4A). This was associated with a decrease in the amount of E-cadherin and change in its localization within 48 hours of treatment (fig. S4A) along with an increase in vimentin and N-cadherin abundance (Fig. 4D and fig. S4A). Similar effects were also observed in SKMEL28 cells but at a later time point. These morphological and expression pattern alterations were consistent with previous descriptions of EMT programs in epithelial cancers (28) and of the phenotypic switch in melanoma (33). In melanoma, loss of E-cadherin was reported to increase c-JUN through Rho-mediated stabilization of c-JUN protein (34, 35). In both A375 and SKMEL28 cells, the time frame of phenotypic marker changes coincided with markedly increased c-JUN abundance (Fig. 4D).

Fig. 4 Vemurafenib induces a phenotype switch resembling EMT.

(A) Representative merged images of cells stained with phalloidin (red; to detect actin) and 4′,6-diamidino-2-phenylindole (DAPI) (blue; to stain the nuclei). SKMEL28 cells contain more vesicles and become highly pigmented after vemurafenib treatment. Scale bar, 50 μm. Images are representative of similar morphological changes observed in five or more experiments. (B) Relative expression of each indicated gene in A375 and SKMEL28 cells at various time points after treatment with vemurafenib (500 nM). Blue, relative decrease; red, relative increase (scale, left). The shaded regions are time points for which data were not obtained. (C and D) Western blot analysis within the first 72 hours (C) or longer (D) after vemurafenib (500 nM).

The EMT-like phenotypic changes were preceded by signaling alterations and various gene expression changes (Fig. 4, B and C, and fig. S4, B and C). In particular, LEF1 gene and protein expression was reduced in both A375 and SKMEL28 cell lines (Fig. 4, B and C, and fig. S4, B and C), consistent with the differences observed between the drug-sensitive and drug-resistant cell lines. These changes corresponded with increased c-JUN gene and protein expression (Fig. 4 and fig. S4), suggesting that a drug-induced increase in c-JUN was linked with phenotype switching marked by altered LEF1 expression.

Reversal of resistance and cell phenotype upon drug withdrawal is associated with a reversal in effects on SPRY, c-JUN, and LEF1

To determine which drug-induced effects on resistance, signaling, gene expression, and phenotype described above were maintained by drug treatment, we adapted A375 cells to vemurafenib for 3 weeks. After this period of treatment, drug withdrawal resensitized cells to the drug and reversed most of the drug-induced effects (Fig. 5, A to D). Similar studies in SKMEL28 and WM2664 cells (Fig. 5, E to G) demonstrated most of the same changes. In particular, drug withdrawal consistently reversed the reduced SPRY2/SPRY4, increased c-JUN and the phosphorylated ERK rebound (Fig. 5). Drug withdrawal also caused the cells to revert to a less migratory (Fig. 5B), more proliferative (fig. S5) state. Collectively, our data show that drug-induced effects on MITF and other EMT transcription factors, such as SNAI2/SLUG, can vary between the cell lines (Fig. 5, D and G), whereas the expression and abundance of SPRY2/4, c-JUN, and LEF1 display consistent patterns after drug treatment. All the cell lines were resensitized to the drug after a period of drug withdrawal (Fig. 5D), indicating that the drug-induced changes were maintaining the cells in a drug-resistant state.

Fig. 5 Signaling alterations and phenotype switching associated with drug adaption and resistance are reversed upon withdrawal of vemurafenib.

(A to D) Dose response to vemurafenib (A), cell migration measured by the xCELLigence system using cell invasion and migration (CIM) plates (B), relative gene expression (C), and Western blot analysis (D) in cultures of A375 cells treated in parallel with either DMSO for 4 weeks (black), 500 nM vemurafenib for 4 weeks (red), or 500 nM vemurafenib for 3 weeks, and then DMSO for 1 week (blue). The connecting lines in (C) represent the trend in gene expression changes observed after drug withdrawal for duplicate samples. (E to G) Vemurafenib dose-response curves (E and F) and Western blot analysis (G) in WM2664 and SKMEL28 cells treated with DMSO (black), 500 nM vemurafenib (red), or 500 nM vemurafenib, and then DMSO (blue) for the number of weeks indicated.

Enforced expression of c-JUN mediates a phenotype switch

Because the signaling, transcriptional, and phenotype changes associated with vemurafenib responsiveness corresponded directly with the expression, abundance, and phosphorylation of c-JUN, we next investigated if its enforced expression was sufficient to drive these changes in the absence of the drug. Stable expression of c-JUN in A375 cells was sufficient to alter cell morphology and gene expression, consistent with an EMT-like switch in the absence of drug treatment (Fig. 6, A to C). Increased phosphorylation of c-JUN was also observed in c-JUN–transfected cells (Fig. 6C), which resulted in reduced LEF1 and increased WNT5A, MITF, and SNAI2/SLUG mRNA and protein abundance (Fig. 6, B and C), suggesting that drug-induced c-JUN controls expression of LEF1 and SNAI2/SLUG. However, c-JUN expression and vemurafenib treatment had opposite effects on MITF in A375 cells (Figs. 5, C and D, and 6, B and C), suggesting that c-JUN may cooperate with other drug-induced factors to regulate MITF expression in A375 cells. Overexpression of c-JUN was also sufficient to increase the abundance of phosphorylated ERK (Fig. 6C) and the expression of its downstream target SPRY2 (Fig. 6B).

Fig. 6 c-JUN drives phenotype switching and adaptation to vemurafenib.

(A to C) Immunofluorescence (A), gene expression (B), and Western blotting (C) of c-JUN abundance in A375 cells transfected with an empty or c-JUN expression vector. Data were analyzed at 7 days (A) or 3 and 10 days (B and C) after selection of transfected cells. Scale bar, 100 μm (A). (D to F) Western blot (D), qPCR (E), and fluorescence-activated cell sorting analysis of propidium iodide (PI) staining (F) to assess protein and mRNA abundance and cell death in A375 cultures transfected with control (siNT) or c-JUN–targeted SMARTpool siRNA for 24 hours, and then exposed to solvent (DMSO) or vemurafenib (Vem; 500 nM). (G) Dose-response curves in A375 cells after 3 days of exposure to combinations of vemurafenib and JNK inhibitor (JNKIn8). (H) Percentage of PI-positive (dead) A375 cells pretreated with DMSO or 500 nM vemurafenib and then 2 days of exposure to vemurafenib and JNKIn8. Data are means. ± SEM from triplicate experiments.

c-JUN mediates cell survival during early drug adaptation

We next examined the impact of c-JUN depletion in drug-treated cells. Reduction of drug-induced c-JUN protein abundance using siRNAs invoked a marked and significant increase in vemurafenib-induced cell death in A375 cells (Fig. 6F and fig. S6A). Although A375 cells were responsive to vemurafenib at low doses (Fig. 6G), the effects were mainly cytostatic; higher doses to vemurafenib were required to see a cytotoxic effect (Fig. 6H). However, knocking down c-JUN markedly increased the cytotoxic effects of low-dose vemurafenib treatment (Fig. 6F). Similar effects were also observed in SKMEL28 and WM2664 cells (fig. S6B). Moreover, the extent of cell death in A375 cultures correlated with the extent of reduction in vemurafenib-induced c-JUN abundance (fig. S6, C and D). These data indicate that c-JUN plays a critical role in mediating cell survival during the early drug-adaptive phase. Depleting c-JUN also reduced ERK activation, as measured by abundance of phosphorylated ERK (Fig. 6D). As expected, JUN knockdown did not affect drug-induced changes in SPRY2 and SPRY4 (Fig. 6, D and E), which were previously shown to be upstream of c-JUN (Fig. 3F and fig. S3, D to F). Our c-JUN knockdown data suggest that increased WNT5A signaling, recently implicated in drug-induced phenotype switching associated with resistance during early drug treatment (11), is mediated through c-JUN (Fig. 6E), consistent with data from a previous report (36). Collectively, our findings based on direct manipulation of c-JUN expression in drug-naïve or drug-induced contexts support a central role for c-JUN in mediating the adaptive resistance process in melanoma cells.

JNK inhibition overcomes the drug adaptive changes and abrogates c-JUN–mediated adaptive resistance

As a potential therapeutic strategy for overcoming adaptive resistance, we assessed the impact of inhibiting the JNK pathway, which we previously observed to enhance drug response in the resistant cell lines (Fig. 2, A and B). Combining vemurafenib with the JNK inhibitor JNKIn8 enhanced the efficacy of vemurafenib treatment in A375 cells (Fig. 6G). These effects were additive to synergistic (fig. S6E), and the dose-response curves suggested that the residual cells surviving the higher doses of vemurafenib were susceptible to combination treatment with JNKIn8, confirmed by increased cell death (Fig. 6H). The amount of cell death after combination treatment with vemurafenib and JNKIn8 was higher and more synergistic after a short-term “drug adaptation” in vemurafenib (Fig. 6H and fig. S6, H and I), suggesting that early drug adaption resulting in high JUN and phosphorylated c-JUN increases the responsiveness to combination treatment of vemurafenib with JNK inhibitors. Combination treatment with JNKIn8 at lower doses that did not result in cell death was used to demonstrate reduction in vemurafenib-induced cell migration (fig. S6F) by preventing c-JUN phosphorylation (fig. S6G). Together, these data support a therapeutically exploitable role for c-JUN in mediating cell survival and EMT-like phenotype switching during the early drug adaptive process.

Immunohistochemical staining of some of the patient tumor samples also confirmed expression of c-JUN and activation of c-JUN (phosphorylated c-JUN) specifically within the melanoma cells of the tumor samples (Fig. 3D and fig. S3D). The sample set available for immunohistochemical analysis was too small to conclude about specific trends, because the results revealed heterogeneity within the individual tumor samples and variability across the patients between the pretreatment and early on-treatment samples. Variation in the amount of phosphorylated c-JUN could be due to differences in time of the early during-treatment samples between the patients, because our data revealed variation in the extent of phosphorylated c-JUN at different times after treatment in the various cell lines (Figs. 3B and 4, C and D). Similar variation in phosphorylated c-JUN culminating with a peak in high phosphorylated c-JUN has also recently been shown in a study using proteomic analysis to assess patterns of BRAF inhibitor resistance in melanoma (37). We were interested in determining which signaling pathways may be implicated in vemurafenib-induced activation of c-JUN through the reduction of SPRY regulation. Previous studies show that SPRY4, the down-regulation of which is sufficient to increase c-JUN, suppresses RAS activation (38), which also activates multiple signaling pathways impinging on c-JUN, including the JNK (39) and the Rho–ROCK (Rho-associated protein kinase) pathways (40, 41). We found that ROCK inhibition counteracted the vemurafenib-induced increase in c-JUN abundance and activity (Fig. 7A) as well as the siSPRY-mediated increase and activation of c-JUN (fig. S7, B and C). ROCK inhibition enhances cell death after treatment with BRAF inhibitors (42); on the basis of our data, we propose that this effect is due to a reduction in c-JUN abundance and activity.

Fig. 7 Signaling and phenotypic changes associated with cell survival and resistance during early drug adaptation.

(A) Western blot analysis in A375 cells cultured with DMSO, vemurafenib (500 nM), or a combination of 500 nM vemurafenib and ROCKi Y-27632 (10 μM) for 48 hours. (B) Schematic representation of the BRAF/MEK inhibitor–induced changes in BRAF-mutant melanomas that are initially responsive to treatment based on our data and that previously reported. GTP, guanosine triphosphate; GDP, guanosine diphosphate. (C) Heat maps of relative gene expression in a cohort of melanoma patient samples [National Center for Biotechnology Information (NCBI) Gene Expression Omnibus (GEO) archive GSE22153] and across a separate panel of melanoma cell lines (Array Express repository E-MTAB-1496). Cell line data were transformed using a variance-stabilizing transformation and then quantile-normalized. Samples for both sets have been arranged in ascending order of JUN expression.

Analysis of patients’ tumor material before and after BRAF inhibitor treatment revealed that treatment resulted in early selection for surviving cells expressing lower SPRY4, higher JUN, and lower LEF1 (fig. S7A) compared to tumors analyzed before therapy. Expression trends for several other genes, including MITF, varied between patient-matched tumor samples taken before treatment and early during treatment (fig. S7A); this was similar to our observations across multiple cell lines during early drug treatment (Fig. 5, D to G). These clinical data strongly support the molecular insights gained from our studies in melanoma cell lines, informing our model of the role of c-JUN in drug response (Fig. 7B).

To further confirm the link between c-JUN and a mesenchymal-like phenotype in all melanomas rather than just in the context of BRAF/MEK inhibition, we interrogated publicly available data sets of a separate 55–cell line melanoma panel (43) and a cohort of 57 late-stage melanoma patient samples (44). We observed a strong association between the expression of JUN and phenotype markers in both sets (Fig. 7C). All cell lines with high JUN expression were previously reported to be mesenchymal-like (45). Strongly supporting our findings in BRAF-mutant melanomas described above, we observed a clear association between high JUN expression with low expression of the epithelial genes and higher expression of the mesenchymal genes. These data suggest that high JUN expression is inherently linked to a mesenchymal-like migratory phenotype in melanomas.

DISCUSSION

Although BRAF and MEK inhibitors have revolutionized therapy for patients with BRAFV600-mutant melanomas, because of progression of disease in patients after incomplete therapeutic responses, the progression-free survival rates in patients have only marginally improved. A range of studies have described various mechanisms of acquired drug resistance to BRAF/MEK inhibitors, and a few have described early adaptive resistance (10, 46, 47). A link between cell phenotype and drug resistance mediated through changes in MITF and WNT5A transcript and protein abundance has also been proposed (11, 12, 21). Our study reveals a clear link between alterations in the MAPK signaling pathways, AP-1 complex activity, and cellular phenotype in the context of inherent and early adaptive resistance to BRAF/MEK inhibitors in melanomas.

Specifically, we uncovered a central role for c-JUN in mediating phenotype switching and cell survival that is associated with adaptive BRAF-MEK-ERK pathway inhibitor resistance in BRAF-mutant melanomas. We demonstrated that this is a direct consequence of down-regulation of SPRY4 after ERK pathway inhibition. In addition, down-regulation of WNT5A signaling and E-cadherin has also been previously linked to JNK/c-JUN activation (35, 48). On the basis of our findings described here together with previously published reports, we propose a model for early drug adaption highlighting the central role of c-JUN. In this model (Fig. 7B), Rho/ROCK is a key signaling pathway that is repressed by SPRY4 and can initiate and maintain c-JUN–mediated adaptive resistance during the early response phase of treatment. This network mediated by c-JUN promotes an EMT-like phenotype switch, giving rise to melanoma cells with a more migratory phenotype that can survive drug treatment.

A key implication of our study is that targeting the RAF-MEK-ERK pathway will induce rapid adaptation of initially drug-responsive melanomas by causing selection for cells that have reduced proliferative potential and a survival advantage with a potentially greater capacity to metastasize. A report showing that drug withdrawal restored responsiveness and intermittent dosing resulted in longer-term drug responses (49) supports our findings that drug-induced effects provide a selective pressure for maintaining cells that are able to survive treatment. We thus believe that effective combination therapy to prevent early adaptation and residual disease is a priority in drug development. Our data do suggest that dose and timing of combination therapy may influence efficacy and should be considered an important aspect of clinical evaluation of combination therapies.

c-JUN has been previously suggested to play a critical role in melanoma and contribute to tumor progression (27). Our data suggest that c-JUN is also a prime target for overcoming early adaptive processes after BRAF/MEK/ERK inhibition. However, because of the difficulty of directly targeting transcription factors, we assessed the effects of inhibition of the upstream JNK signaling pathway that is known to activate and stabilize c-JUN (27). Our data shown here demonstrate that JNK inhibition provides a clinically translatable therapeutic strategy for preventing early drug adaptive resistance mediated through c-JUN. Similar findings have also been recently reported by Fallahi-Sichani et al. (50), confirming that targeting the JNK-c-JUN pathway offers a potential therapeutic combination that could be used for achieving more complete, durable responses to BRAF/MEK inhibitor treatment in patients with BRAF-mutant melanoma. We propose that the best window of opportunity for treatment with combination of BRAF and JNK inhibitors would be in the early stages of adaptation to BRAF/MEK inhibitor therapy when the amount of total and phosphorylated c-JUN is increased. Although JNK inhibitors have been in clinical development for inflammatory conditions, effective clinical candidates are necessary for conclusive preclinical evaluation using in vivo melanoma models (51).

Beyond melanoma, sustained JNK signaling promotes EMT, invasion, and survival of breast cancer cells by regulating ERK activity (52), and JNK activation has been reported to predict for poor response to sorafenib in hepatocellular carcinoma (53). Hence, we propose that combination treatment with the JNK pathway inhibitor may potentially be more widely applicable for overcoming drug resistance in other cancers that undergo EMT. Phenotypic plasticity in epithelial cancers, driven primarily through EMT, is the main contributing factor to cancer progression through metastases, and EMT is a feature of drug resistance in many cancers (54). “Drugging” the EMT/phenotype switching process has the potential to greatly enhance cancer therapy, and the identification of new targets to overcome cell survival associated with phenotypic changes has wide-ranging clinical applications for cancer treatment (55). We propose that the JNK–c-JUN pathway offers a useful therapeutic target that could be beneficial in preventing phenotypic plasticity and overcoming therapy resistance.

MATERIALS AND METHODS

Cell lines

Melanoma cell lines were obtained from the American Type Culture Collection or from the Australasian Biospecimen Network–Oncology Cell Line Bank at the Queensland Institute of Medical Research (QIMR). PCR-based short tandem repeat (STR) analysis using six STR loci was performed to confirm the individuality of each cell line and the identity of the cell lines through the study. BRAF mutations in exon 15 were detected by Sanger sequencing using the conditions previously described. The sequencing products were run on a 3700 Genetic Analyzer (Applied Biosystems), and the sequencing data were then analyzed using Sequencer 4.6 (Gene Codes Corporation).

Drug dose-response SRB assays

The cell lines were seeded into 96-well microtiter plates for 48 hours before the addition of drug ranging in concentration from 0.01 nM to 30 μM, incubated for 72 hours, and analyzed for total cell number using the sulforhodamine B (SRB) assay. Cells were fixed in situ by the addition of cold trichloroacetic acid (Merck) and stained using 0.4% (w/v) SRB/1% (v/v) acetic acid. The bound dye was solubilized in 10 mM tris buffer, and the absorbance was measured at 515 nM using a spectrophotometer. The IC50 value for each cell line was calculated as the drug dose resulting in 50% reduction of SRB absorbance relative to solvent-treated cells obtained upon fitting a dose-response curve to the SRB absorbance readings. The drugs used in this study are listed in table S4.

Gene expression analysis on cell lines by microarrays

Cell lines were harvested when they reached 50 to 80% confluency. RNA was extracted using Qiagen miRNeasy kit. The labeling for Affymetrix DNA microarray analysis was performed according to the manufacturer’s instructions. Biotin-labeled complementary RNA (cRNA), produced by in vitro transcription, was fragmented and then hybridized to Affymetrix 1.0 ST expression array. For each cell line, three independent expression arrays were performed, robust microarray–normalized, and background-corrected. Expression data from two or more experimental replicates for each cell line were averaged and analyzed. Gene probes with interquartile range less than 0.25 expression units or that did not display an intensity of at least log2 (100) units in at least 25% of the samples were also omitted from further analysis. Genes that were differentially expressed between the drug-resistant versus the drug-sensitive cell lines were identified using the “limma” algorithm (56, 57), having adjusted P values less than 0.05 and absolute log fold change greater than 1. Functional category enrichment analysis using the NCBI Database for Annotation, Visualization and Integrated Discovery (DAVID) functional annotation tool was used to identify GO terms (58) for the significant differentially expressed gene set. The three highest levels were selected for each of the Molecular Function, Biological Processes, and Cellular Compartment categories of GO terms. GeneGo network analyses were performed using the shortest path, one-step method (METACORE).

Retroviral transduction and establishment of stable cell lines

Stable cell lines expressing c-JUN were generated via retroviral transduction. Human embryonic kidney (HEK) 293T cells were cotransfected with 3 μg of pBABEpuro(hu c-JUN) construct or empty pBABEpuro vector control and 3 μg each of the Moloney murine leukemia virus gag-pol expression plasmid pEQ-PAM3(-E) and the RD114 env expression plasmid pRDF by complexing with 40.5 μl of polyethylenimine (1 mg/ml). The HEK293T viral supernatant was collected, filtered, and supplemented with protamine sulfate (10 μg/ml) before being used to infect melanoma cell lines. Infections were repeated four times over a 48-hour period. Virus-infected cells were selected in puromycin (0.4 μg/ml) for 3 days until mock-infected cultures were completely killed. Transduced cell lines were maintained in puromycin (0.2 μg/ml).

Transfection of the siRNA

Cell lines were transfected with 50 nM siRNA complexed with 0.1% DharmaFECT1 1 day before drug treatment. Lipid and siRNA were each separately diluted in nonsupplemented medium for 5 min and complexed for 15 min before addition to cells. Cells were assessed in SRB or death assays and by Western blot analysis 2 to 4 days after transfection and/or drug treatment. siRNAs used include siNT ON-TARGETplus Nontargeting siRNA #1 (Thermo-Fisher Scientific; cat. no. D-001810-01-05), siJUN GENOME SMARTpool (Thermo-Fisher Scientific; cat. no. M-003268-03-0005), siJUN ON-TARGETplus SMARTpool (Thermo-Fisher Scientific; cat. no. L-003268-00-0005), siSPRY2 ON-TARGETplus SMARTpool (Thermo-Fisher Scientific; cat. no. L-005206-00-0005), siSPRY4 ON-TARGETplus SMARTpool (Thermo-Fisher Scientific; cat. no. L-015457-01-0005), and siSPRY2 (Santa Cruz Biotechnology; cat. no. SC-41037).

Western blot analysis

Whole-cell lysates were prepared in lysis buffer containing 50 mM Hepes, 150 mM NaCl, 1 mM EGTA, 1.5 mM MgCl2, 1% (v/v) Triton X-100, 10% (v/v) glycerol, 50 mM NaF, 10 mM Na3VO4 with complete mini protease inhibitor cocktail (Roche), and protein concentrations were quantitated using the DC Protein Assay reagents (Bio-Rad). Protein aliquots (20 μg) were separated on 10 to 12% SDS–polyacrylamide gels and transferred onto polyvinylidene difluoride membranes (Immobilon-P, Millipore). Membranes were blocked with tris-buffered saline/0.1% (v/v) Tween-20 (TBST) containing 5% (w/v) skim milk powder (Diploma) and probed with antibodies listed in table S6 detected by chemiluminescence and autoradiography.

Immunofluorescence and microscopy analysis

Cells were fixed in 4% paraformaldehyde in phosphate-buffered saline (PBS) for 15 to 20 min and washed before permeabilization with 0.1% Triton X-100 in PBS for 1 to 5 min. After incubation with antibody blocking (Dako) for 30 min to 1 hour, they were stained with phalloidin-TRITC (tetramethylrhodamine isothiocyanate; 50 μg/ml) (Sigma-Aldrich) for 1 hour, washed, and mounted in ProLong Gold Antifade reagent containing DAPI (Invitrogen). For immunofluorescence detection of E-cadherin and vimentin, cells were stained with antibodies (BD Transduction) to E-cadherin conjugated to FITC (fluorescein isothiocyanate) and to vimentin conjugated to TRITC for 1 hour, washed, and mounted in ProLong Gold Antifade reagent containing DAPI (Invitrogen). Cells were analyzed using the BX51 or BX61 microscope (Olympus). Microscopic images were captured using the SPOT Advanced software.

Death and apoptosis assays

All suspension and adherent cells were harvested and pooled after drug treatment, and stained with PI in PBS at room temperature for 5 min to detect dead cells or with tetramethylrhodamine ethyl ester (TMRE) to assess the mitochondrial membrane potential and detect apoptotic cells. Stained cells were analyzed on the Canto II (Becton Dickinson), and the percentage of PI- or TMRE-positive cells were assessed using the FCS Express (BD Biosciences) as described previously (59).

Gene expression analysis by reverse transcription PCR

For analysis of cell lines and xenograft tumors, total RNA was extracted at the indicated time points using the PureLink RNA Mini Kit (Life Technologies). Total mRNA was used for reverse transcription using the High-Capacity cDNA Reverse Transcription Kit (Life Technologies). Reverse-transcribed complementary DNA (cDNA) (30 ng) was used for qPCR using Fast SYBR Green PCR Master Mix and gene-specific primers (shown in table S5) in triplicate reactions using an Applied Biosystems StepOnePlus Real-Time PCR System. Relative expression was determined using the comparative CT method (Applied Biosystems) followed by normalization to GAPDH.

Migration assays

Cell migration assays were performed using the RTCA DP Instrument (ACEA Bioscience Inc.). Cells to be assayed were passaged 1 day before the experiment and reached 60 to 80% confluence. The lower chamber wells of the CIM plates were filled with 160 μl of 10% fetal bovine serum/RPMI/Hepes medium, and the upper chambers were filled with 50 μl of serum-free medium to cover the membrane, both containing the drug treatment at the appropriate concentration. A serum gradient was established before the baseline measurement. Cells (15,000) diluted in serum-free medium were then seeded to each upper chamber. The CIM plates were loaded to the RTCA DP Analyzer inside a TC incubator. Measurements were taken every 15 min for 300 repetitions (>72 hours).

Xenograft tumor growth and treatment of mice

A375 cells (4× 106) mixed in a 1:1 ratio with high-concentration Matrigel (Matrigel Matrix, BD Biosciences) were injected subcutaneously into the right flank of each female NSG mouse. Tumors were monitored every 2 to 3 days, and caliper measurements were taken of width and length of each tumor. The volume was calculated as (length × width × width)/2. Treatment commenced 10 days after injection of A375 cells, and average volume of tumors was 183.4 mm3 (range, 39.4 to 401 mm3). Three mice were fed a diet of control chow, and three mice were fed a diet of PLX4720 in chow (approximate dosing of 40 mg/kg). One week after treatment commenced, the mice were sacrificed, the tumors were resected, and the tumor slurry was prepared using the McIlwain tissue chopper, snap-frozen in liquid nitrogen, and stored at −80°C. RNA was extracted from cell pellets and analyzed for gene expression by reverse transcription PCR (RT-PCR) described above.

Patient samples analyses

Fresh-frozen tumor specimens were obtained before, during, and on progression from patients participating in clinical trial protocols involving either vemurafenib or dabrafenib at the Melanoma Institute Australia and Westmead Hospital (New South Wales, Australia). Studies had local institutional review board approval, and all patients provided written informed consent. Banked fresh-frozen melanoma samples were reviewed and scored for tumor content. Only tumors with >80% tumor content were selected for this study. Total RNA was extracted from 20 to 30 μg of fresh-frozen tissue. Tissue samples were homogenized using a high-speed agitation Polytron blender (Kinematica) in the presence of TRIzol. The RNA was isolated and purified with an RNeasy purification kit (Qiagen) with deoxyribonuclease I digestion on the column. The quality of the RNA preparations was assessed using the Agilent 2100 Bioanalyzer (Agilent Technologies). RNA integrity scores were > 8 for all the samples analyzed. cRNA amplification and labeling were performed using the Illumina TotalPrep RNA Amplification Kit according to the manufacturer’s directions (Ambion) with 250 ng of total RNA as input material. Illumina microarrays were used to profile gene expression as described previously (31, 60).

cDNA was generated using an oligo(dT) primer, and qPCR was conducted using the comparative CT method (Applied Biosystems) as described above. Relative expression was determined by normalization to GAPDH. Four-micrometer whole sections were freshly cut from formalin-fixed paraffin-embedded tumor blocks. Slides were dewaxed through histolene and graded alcohols to water before antigen retrieval in Target Retrieval Solution (pH 9) (Dako) at 1250°C for 3 min in a pressure cooker. After cooling, slides were loaded onto a Dako autostainer for the following incubations: 1 hour in primary antibody (c-JUN or phosphorylated c-JUN Ser73; table S6), 1 hour in EnVision+ Rabbit (Dako), and 10-min color development with 3,3′-diaminobenzidine (DAB) (Dako). Sections were washed in TBST buffer between each incubation. After DAB staining, slides were counterstained with hematoxylin, mounted with coverslips, and stored at room temperature. Stained sections were analyzed using the BX51 microscope (Olympus).

Data presentation and statistical analysis

Data were graphed and statistically analyzed using the computer programs Microsoft Excel (Microsoft), GraphPad Prism (GraphPad Software), and R (61) programs.

Combination drug treatment analyses

To analyze the dose-response data from combinations of multiple vemurafenib and JNKIn8 doses on various melanoma cell lines, the activities were converted into relative activities f, such that f = 0 at no drug and f = 1 for maximum response. At each dose combination, we calculated the Bliss excess, which is the difference between the measured relative activity and the relative activity predicted from the assumption of Bliss independence, fbliss = f1 + f2f1× f2 (62). We also calculated the HSA excess (30), which is the difference between the observed fractional activity and the highest of the fractional activities from single doses of vemurafenib or JNKIn8.

SUPPLEMENTARY MATERIALS

www.sciencesignaling.org/cgi/content/full/8/390/ra82/DC1

Fig. S1. BRAF, MEK, and ERK inhibitor drug dose responses, gene expression changes and cell migration in sensitive and resistant melanoma cell lines.

Fig. S2. Effects of JNK inhibitor combination or siJUN in the drug-resistant cell lines.

Fig. S3. BRAF inhibition induces JUN expression and c-JUN abundance and activation by down-regulating SPRY.

Fig. S4. Vemurafenib-induced gene expression and protein abundance changes in A375 cells and xenografts.

Fig. S5. Effects of drug treatment and drug withdrawal on cell proliferation.

Fig. S6. c-JUN knockdown or JNK inhibition promotes apoptosis and decreases cell migration in response to vemurafenib.

Fig. S7. Gene expression in patient samples early during drug treatment and the role of the ROCK pathway during early drug adaptive responses.

Table S1. List of genes and exon array probes from “limma” differential gene expression calculation.

Table S2. Selected GO annotations from NCBI DAVID analysis.

Table S3. Melanoma patient information.

Table S4. Small molecule inhibitors used in this study.

Table S5. Primer sequences used for RT-PCR.

Table S6. Antibodies used in this study.

REFERENCES AND NOTES

Acknowledgments: We are grateful to C. Schmidt (QIMR, Australia) for provision of the melanoma cell lines from the Australasian Biospecimen Network–Oncology Cell Line Bank at the QIMR, Plexxikon Inc. for PLX4032 compound and PLX4720 chow, and P. Darcy [Peter MacCallum Cancer Centre (PMCC), Australia] for the PEQ and RD114 retroviral packaging constructs. We thank R. Salemi, A. Rynska, H. Do, and A. Dobrovic (PMCC) for conducting the BRAF high-resolution melt PCR and BRAF sequencing, and A. Foo (PMCC) for conducting the melanoma cell line microarrays. We thank U. Putz, C. P. Goh, and S.-S. Tan (Howard Florey Institute, Melbourne, Australia) for advice in relation to SPROUTY2 reagents and R. Anderson with members of the Metastasis Laboratory (PMCC) for advice and assistance with cell phenotype analysis and migration assays. We thank the staff from the Flow Cytometry, Microscopy, Bioinformatics and Functional Genomics Core facilities (PMCC) including R. Rossi, J. Danne, C. Johnson, J. Ellul, J. Luu, and P. Madhamshettiwar. We acknowledge the contribution made by melanoma patients through their consent for the analysis of tumor samples. “In memory of Eilis.” Funding: This project was supported by funding to P.T.F. from the Peter MacCallum Cancer Foundation, the CASS Foundation, and the National Health and Medical Research Council of Australia (NHMRC 1042980). F.Z.L. receives scholarships from the China Scholarships Council and Cancer Therapeutics CRC. M.S. was supported by the Australian National Health and Medical Research Council, Pfizer Australia, and the Victorian Endowment for Science, Knowledge and Innovation. Author contributions: P.T.F. conceptualized and designed the study, supervised the experimental work, and developed the methodology. R.R., S.A.-O., T.W., P.T.F., F.Z.L., P.E.B., and R.J.Y. conducted the experimental work and analyzed the data. R.N.J. conducted the differential gene expression and pathway analyses and contributed to data presentation, statistical analysis, and concepts on pathway interactions. K.E.S. and R.B.P. provided the melanoma cell line microarray data. H.R. and G.V.L. provided cDNA, sections, and microarray data from the patient melanoma samples. S.E.B. and M.S. provided xenograft tumor samples. G.A.M. provided expertise and discussion on clinical responses to BRAF/MEK inhibitors. G.B. provided PLX4032 and PLX4720 chow and expertise and discussion on MAPK signaling pathways. E.T. conducted some of the Western blots and provided expertise on EMT transcription factors. A.S.D. provided expertise and discussion on EMT concepts. P.T.F. wrote and revised the manuscript and A.S.D., R.B.P., R.N.J., R.R., G.A.M., M.S., G.B., H.R., and E.T. edited or commented on the manuscript. Competing interests: G.B. is an employee of Plexxikon. G.A.M. receives research grant support from Novartis, Pfizer, Ventana, and Celgene; is a consultant for Provectus; and is on the advisory board for GSK, Roche-Genentech, Novartis, BMS, Millennium, Merck, and Amgen. G.V.L. is a consultant advisor for Amgen, BMS, GSK, Merck, Novartis, Provectus, and Roche. Data and materials availability: The microarray data have been deposited to NCBI GEO and are accessible through GEO Series accession number GSE45558.
View Abstract

Navigate This Article