C-Jun drives melanoma progression in PTEN wild type melanoma cells

Due to the critical impact of active AP-1 transcription factors in melanoma, it is important to define their target genes and to identify and ultimately inhibit oncogenic signals. Here we mapped the genome-wide occupancy of the AP-1 family member c-Jun in different melanoma cells and correlated AP-1 binding with transcriptome data to detect genes in melanoma regulated by c-Jun. Our analysis shows that c-Jun supports the malignant phenotype by deregulating genes in cancer-relevant signaling pathways, such as mitogen-activated protein kinase (MAPK) and phosphatidylinositol-3-kinase (PI3K) pathways. Moreover, we demonstrate that the importance of c-Jun depends on melanoma stage and mutation status of the tumor suppressor PTEN. Our study reveals that activation of c-Jun overrules the tumor suppressive effect of PTEN in early melanoma development. These findings help to understand the relevance of c-Jun within cancer pathways in different melanoma cell types, especially in relation to MAPK and PI3K pathways, which are commonly deregulated in melanomas. Consequently, targeting c-Jun in PTEN+ melanoma cells may represent a promising therapeutic strategy to inhibit survival of melanoma cells to prevent the development of a metastatic phenotype.


Introduction
Melanoma is a highly aggressive and heterogeneous type of cancer, and its incidence is growing faster than any other cancer entity. In recent decades, many altered pathways regulating the development and progression of melanoma and the high migratory and invasive potential of melanoma cells have been identified, but a detailed molecular understanding of this disease is largely lacking.
One crucial step in melanoma development and progression is the deregulation of cancer-supporting transcription factors, especially activating protein-1 (AP-1) transcription factors, including the c-Jun, JDP, c-Fos, FRA, and MAF subfamilies 1 . AP-1 proteins bind to the classical palindromic recognition sequence 5′-TGA(C/G)TCA-3′ and regulate target gene expression, leading to deregulation of cancer-relevant signaling pathways. Thus, AP-1 transcription factor dimers play an important role in different cancer types, including malignant melanoma [1][2][3][4][5] . A main characteristic of AP-1 complexes in the cell is their heterogeneity in dimer composition. This heterogeneity is caused by the fact that multiple AP-1 subunits can be expressed simultaneously, and different dimer compositions lead to the transcriptional regulation of different target genes. Various studies demonstrated that differences in AP-1 dimer compositions cause altered specificity in their binding site selection 6 . However, it remains unclear which direct target genes of AP-1 homodimer or heterodimer cause the functional effects that support melanomagenesis.
The AP-1 family member c-Jun is a main regulator of melanoma progression 4,7,8 , acts by regulating target genes supporting proliferation and migration of cancer cells, and thus promotes the malignant phenotype. We have previously demonstrated that the microRNA miR-125b directly regulates the transcription factor c-Jun, affecting the proliferative and migratory potential of melanoma cells 9 . Furthermore, we identified an alternative regulatory pathway of c-Jun in melanoma cells leading to upregulation of c-Jun activity via the loss of the cell-adhesion molecule E-cadherin 10,11 . Despite its role in the aforementioned pathophysiological processes, only a few specific c-Jun target genes have been identified to date. Previous chromatin immunoprecipitation (ChIP) studies focusing on AP-1/c-Jun in nonmelanoma cell types identified a few cancer-related target genes of c-Jun. Deregulation of c-Jun is one of the most important events in malignant melanoma and many other cancer entities, but the functional relevance of c-Jun deregulation and its molecular effects on target gene expression have not been determined in detail to date.
Another crucial event supporting cancer cell survival is the upregulation of PI3K/AKT signaling activity in various cancer types, which can largely be attributed to deregulation of the AKT negative regulator, phosphatase, and tensin homolog (PTEN) phosphatase 12 . Mutations in the PTEN gene and thus the loss of this tumor suppressor protein are prevalent in melanoma and lead to upregulation of AKT activity. Activated AKT protects cells from apoptosis by phosphorylating and inactivating proapoptotic substrates 13 . Mutational inactivation of PTEN frequently occurs in human cancers; however, early molecular tumor-promoting mechanisms in PTEN expressing (PTEN + ) cancer cells, such as those in early developing melanoma, remain unclear.
In this study, we demonstrate that c-Jun upregulation in melanoma cells depends on melanoma stages (primary tumor (PT) vs. metastasis (MET)) and PTEN expression status.

C-Jun and the tumor suppressor PTEN exhibit a positive correlation in melanoma cells
Given that melanomas gain different properties during their progression, we selected six different melanoma cell lines (Sbcl-2, WM3211, WM793, WM1366, WM1158, WM9) categorized by tumor stage (PT, MET) and their previously described BRAF and PTEN mutation status 14 Supplementary Fig. 2). We first performed functional assays to analyze the proliferative ( Supplementary  Fig. 1Ai) and/or migratory potential ( Supplementary  Fig. 1Aii) of four melanoma cell lines exhibiting different tumor stages, BRAF mutation statuses, and PTEN deletions. Interestingly, based on the mentioned characteristics, a distinct classification based on proliferative and/or migratory behavior was not possible. Thus, we further focused on c-Jun-dependent differences of gene expression among the analyzed melanoma cell lines.
We first analyzed differentially expressed genes in malignant melanoma by performing RNA-Sequencing (RNA-seq) with all aforementioned melanoma cell lines and normal human epidermal melanocytes (NHEM). We assigned tumor stage and mutation status to our samples and performed principal component analyses (PCA) to construct low-dimensional representations of the expression profiles of the complete RNA-Seq data set. The PCA results clearly demonstrated sample clustering based on PTEN and BRAF status (Fig. 1a).
Analysis of mRNA expression levels of both c-Jun and PTEN in all used melanoma cell lines (compared to NHEMs) revealed high c-Jun expression levels in PTEN WT cells that were gradually reduced in PTEN HemDel and PTEN HomDel cells (Fig. 1b). PTEN mRNA expression follows the corresponding copy numbers of the used cell lines, demonstrating high PTEN mRNA levels in NHEMs and PTEN WT that continuously decrease in PTEN HemDel and PTEN HomDel (Fig. 1c). We exemplarily confirmed c-Jun and PTEN mRNA expression via qRT-PCR analysis in one PTEN WT (WM1366) and one PTEN HomDel (WM9) melanoma cell line (Fig. 1d, e) compared to NHEMs. In accordance with these findings, Western blot analyses of PTEN WT melanoma cells exhibited high c-Jun protein levels, which decreased with the gradual loss of PTEN (Fig. 1f, Supplementary Fig. 1B). Due to the fact that PTEN is known to inhibit AKT activity (predominantly AKT3 in melanoma) and thus promotes melanoma cell apoptosis, loss of PTEN is a crucial event during melanoma development and progression. Hence, we further analyzed AKT1 and AKT3 mRNA expression by qRT-PCR in both PTEN + and PTEN − melanoma cells. Interestingly, no differences in AKT expression between PTEN WT and PTEN HomDel cells were detected (Fig. 1g), supporting the role of loss of PTEN on AKT activity.
Consequently, we hypothesize that PTEN + melanoma cells in early melanoma development utilize c-Jun to overcome the tumor-suppressive effect of PTEN.

C-Jun significantly influences gene expression in malignant melanoma
To further explore the c-Jun influences gene expression in malignant melanoma, we mapped its binding sites using ChIP-Sequencing (ChIP-seq). First, we identified genes associated with c-Jun peaks in each melanoma cell line and compared their mRNA expression to normal NHEMs. Our analysis showed a significant induction of c-Jun peak-associated genes in each sequenced c-Jun expressing melanoma cell line (Sbcl2, WM3211, WM1366, WM793, and WM1158), suggesting a positive regulatory activity of c-Jun (Fig. 2a). Moreover, analysis of the resulting c-Jun ChIP-Seq peaks confirmed the melanoma cell line clustering detected via RNA-Seq (Fig. 1a). Most interestingly, our c-Jun ChIP-Seq data clearly   To further study the impact of c-Jun on the activity status of bound regulatory elements, we compared the local deposition of H3K27ac, a histone modification associated with active regulatory sites, at c-Jun peaks (Supplementary Table 2) in melanoma cells compared to NHEMs. As shown in heatmap representations and corresponding histograms, c-Jun-bound regions show a higher local acetylation in melanoma compared to melanocytes (Fig. 2c).
Taken together, our data demonstrate that gene expression differences between all analyzed melanoma cell lines depend on c-Jun and PTEN expression/activity. These findings indicate that c-Jun plays a crucial role in early melanoma development.
Functional annotation reveals an enrichment of PI3K/AKT signaling in PTEN + /c-Jun + melanoma cells As a next step, we were interested in the functional relevance of all differentially expressed genes (based on RNA-Seq) and the subset of differentially expressed genes associated with a c-Jun peak. We performed differential expression analysis of the RNA-Seq data described above and compared the gene expression status of PTEN WT , PTEN HemDel , or PTEN HomDel cells with NHEMs. The comparison resulted in three different gene sets with differentially expressed genes represented as heat maps (Fig. 3a, total diffgenes). These gene sets were subsequently merged with c-Jun ChIP-Seq peak annotation data to identify those genes differentially expressed, which are regulated by c-Jun (Fig. 3b, diffgenes Jun). The amount of total differentially expressed genes and differentially expressed genes associated with a c-Jun peak of all generated gene sets are illustrated in Supplementary Table 3. Regarding PTEN HomDel cells (lacking c-Jun), we further explored the total differentially expressed genes.
These newly generated data sets, including the total differentially expressed genes and differentially expressed genes associated with a c-Jun peak, were used to perform detailed functional annotation clustering with the bioinformatic tools Database for Annotation, Visualization and Integrated Discovery (DAVID; v 6.8) and Gene Set Enrichment Analysis (GSEA) tool, respectively, to identify enriched gene sets within each data set and their molecular functions.
Functional annotation analysis via DAVID 15,16 resulted in several melanoma-/cancer-specific enriched Kyoto Encyclopedia of Genes and Genomes (KEGG) pathways (Table 1). Interestingly, functional annotation of the differentially expressed genes associated with a c-Jun peak yielded similar enriched pathways in PTEN WT and PTEN HemDel cells as the total differentially expressed genes of PTEN HomDel cells. Our analysis revealed c-Jun to be a regulator of genes involved in the PI3K/AKT signaling pathway in PTEN + melanoma cells, indicating that c-Jun mainly supports anti-apoptotic and pro-survival processes in early (PTEN WT and PTEN HemDel ) melanoma development.
The GSEA tool utilizes the expression status (read counts of RNA-Seq) for functional annotation clustering and thus identifies genes that are upregulated or downregulated.  Supplementary Fig. 1B. GAPDH serves as loading control. g AKT1 and AKT3 mRNA expression analysis via qRT-PCR normalized to ß-actin. The box-plots show the mean ± s.d. of read counts of two independent experiments (***P < 0.001; **P < 0.01; *P < 0.5). The bars show the mean ± s.d. of three independent experiments; measurements were performed in triplicate (*P < 0.5)  Figure 4a shows the overlap of the identified differentially expressed PI3K/AKT-related genes (RNA-Seq) and the c-Jun-regulated genes detected by ChIP-Seq in both subgroups, PTEN WT (9.94% diff. exp.; 5.11% c-Jun regulated) and PTEN HemDel (7.95% differentially expressed; 2.84% c-Jun regulated). Figure 4b displays the expression status of four differentially expressed PI3K/ AKT-related genes (PDGFB and CCND1 regulated by c-Jun in PTEN WT , CDK6, and BCL2 regulated by c-Jun in PTEN HemDel ), the associated c-Jun peaks (red arrows; ChIP-Seq) and the H3K27 acetylation status around the c-Jun peaks (blue), illustrating the differences in transcriptionally active genomic regions between NHEMs and melanoma cell lines. Moreover, we confirmed our findings by quantitative mRNA expression analysis of the identified differentially expressed c-Jun target genes (FGF5, CCND1, EGFR, PDGFB, LAMB3) after c-Jun siRNA transfection ( Supplementary Fig. 3B). Our data clearly reveal the potential of the transcription factor c-Jun to either upregulated or downregulate target gene expression. All identified differentially expressed genes and those regulated by c-Jun of the analyzed melanoma subsets are presented in Supplementary Table 4. The previously described functions of each identified c-Jun regulated differentially expressed gene in melanoma are depicted in Supplementary Table 5.
Validation of c-Jun-dependent anti-apoptotic and prosurvival effects in early melanoma cells To validate our findings, we performed knockdown experiments. First, we determined the total protein    (Fig. 5a, b). We observed increased phospho-AKT (P-AKT) in WM9 melanoma cells lacking PTEN and c-Jun expression. Thus, we hypothesized that c-Jun expression and transcriptional activity promotes survival in PTEN-expressing melanoma cells. Interestingly, AKT-siPool transfection in PTEN−/c-Jun -WM9 melanoma cells to mimic PTEN re-expression increased c-Jun protein expression (Fig. 5c, d). Moreover, we could validate our prior findings after c-Jun-siPool transfection in WM1366. Western blot analysis depicted a decreased AKT activity (P-AKT) (Fig. 5e, f) and inhibition of c-Jun after siRNA transfection resulted in a significant decrease in cell number compared to sictrl-transfected cells (Fig. 5g). Transfection efficiencies are presented in Supplementary Fig. 6. Importantly, these findings could also be validated by analyzing TCGA-sequencing data (https://www.cancer. gov/tcga) of a cohort of 471 melanoma patients (TCGA, Project ID: TCGA-SKCM). These data were categorized by the PTEN copy number alteration (PTEN WT , PTEN-HemDel, PTENHomDel ) and further analyzed regarding to the c-Jun expression status by using the cBioPortal 18,19 . The data reveal a significant downregulation of c-Jun in patients with PTEN HomDel and an almost unchanged c-Jun expression pattern in those with PTEN HemDel compared to PTEN WT patients, supporting our findings (Fig. 5g).  (Fig. 6a). Further, we investigated overall survival of 471 skin cutaneous melanoma patients based on the PTEN and c-Jun expression status, and could clearly show that melanoma patients with alterations in both, c-Jun and PTEN, belong to the high risk group, whereas patients with either PTEN WT or c-Jun WT belong to the low risk group (Fig. 6b, c).
In summary, we identified a new, functional relevant relationship between the oncogenic transcription factor c-Jun and the tumor suppressor PTEN in malignant melanoma and a mechanism explaining the reduction in apoptosis and induction of tumor cell growth in PTEN + melanoma cells, thus enabling the development of a malignant phenotype.

Discussion
In previous studies, we demonstrated that the transcription factor c-Jun plays a critical role in the development and progression of malignant melanoma 3,5,9,10 . However, the detailed molecular mechanism of c-Jun's influence on melanoma progression and development remains elusive given that only a subset of target genes is known. Previous studies have demonstrated that cancerrelevant genes, such as cyclin D1, p53, and INK4A 20,21 , are regulated by c-Jun. Moreover, we demonstrate that tumor-relevant genes, such as FosB, WEE1, PVR, MAP1LC3B, and LGALS3, are deregulated in melanoma by direct regulation via c-Jun 5 .
In this study, we investigated the detailed molecular role of the transcription factor c-Jun in melanoma cell lines representing different tumor stages and mutation status as determined by RNA-Seq and ChIP-Seq experiments. Differential gene expression in melanoma compared to NHEMs differs according to melanoma stage and mutation status of the sequenced melanoma cell lines, indicating various possible mechanisms leading to a malignant phenotype. Consistent with our findings, Haqq and colleagues 22 also discovered gene expression differences between radial growth phase (RGP) and vertical growth phase (VGP) melanoma cells. It is well established that critical tumorrelevant signaling pathways, such as MAPK and PI3K, are deregulated in melanoma cells, leading to a high proliferative, migratory, and invasive potential of these cells and overcoming apoptosis 23,24 . Although oncogenic key players, such as BRAF, NRAS, p53, and PTEN, within these pathways are well characterized, the detailed molecular mechanisms remain elusive.
Our analysis revealed that c-Jun expression status correlates with PTEN expression in melanoma cells, indicating a dependency of PTEN and c-Jun to allow the development of a malignant phenotype. Computational clustering of our sequencing data suggested a dependency of c-Jun activity in melanoma cells based on the presence of the tumor suppressor PTEN. Moreover, the number of c-Jun peaks decreased as PTEN expression decreased and was ultimately lost. Mutations in members of the PI3K pathway have been extensively studied in many cancer entities. In melanoma, PTEN inactivation could be identified as the crucial step activating the PI3K pathway [25][26][27] . Consistently, recent studies also demonstrate a correlation between c-Jun and PTEN in nonmelanoma cancer. Hettinger and colleagues identified a direct downregulation of PTEN by c-Jun, leading to an upregulation of the AKT survival pathway in several cancer cells, such as pancreatic cells, glioblastoma cells, and transformed kidney epithelial cells 28 . Moreover, these results could also be confirmed in lung cancer cells 13 . However, the described PTEN and c-Jun expression pattern and binding site of c-Jun by Hettinger et al. was not revealed in our data, suggesting a different molecular relation between PTEN and c-Jun in melanoma. Our data clearly revealed a functional correlation of c-Jun and PTEN in melanoma cells. In accordance with this finding, recent studies identified a crosstalk mechanism, whereby constitutive ERK activity suppresses PTEN in melanoma models and is mediated by modulation of c-Jun 29 . Furthermore, crosstalk between JNK and PTEN and thus the PI3K/AKT pathway could be detected in invasive adenocarcinoma of the prostate 30 . Interestingly, a mechanism exists that links ERK to JNK signaling in human melanoma. Hyperactivated ERK increases the stability of c-Jun protein and induces c-Jun transcription by CREB activation. C-Jun subsequently induces RACK1 transcription, which enhances JNK activation. Consequently, c-Jun is further stabilized and activated by JNK 31 .
Taken together, both identified crosstalk mechanisms support the hypothesis that c-Jun upregulation is the main link between the ERK/JNK-signaling pathways and the PI3K pathway, to overcome the tumor suppressive effect of PTEN, which would otherwise counteract the transformation process by inducing apoptosis.
Functional annotation analyses via multiple tools confirmed our findings. Analyses of the identified differentially expressed genes of melanoma cells resulted in a PTEN-independent enrichment of PI3K/AKT-signaling targets. Interestingly, we detected no increased AKT activity in PTEN-expressing melanoma cells and thus investigated if c-Jun regulates PI3K/AKT-related genes to promote malignancy. Indeed, our data revealed direct regulation of PI3K/AKT targets via c-Jun but not AKT itself, suggesting an overruling effect of c-Jun in the presence of PTEN during early melanoma development. Previous studies described that increased levels of c-Jun in melanoma cell lines coincide with upregulation of phosphoinositide-dependent kinase 1 (PDK1) and phosphorylation of protein kinase C (PKC) and AKT 32 . Although, prior studies reported that c-Jun is a negative regulator inhibiting PTEN, detailed mechanisms by which c-Jun inhibits PTEN and thus indirectly increases AKT activity remain to be established 33,34 . Our data clearly showed no detectable alterations in AKT expression and activity in c-Jun and PTEN-coexpressing melanoma cells. Thus, c-Jun seems to play a crucial role in the presence of PTEN to enable the development of a malignant phenotype, which could be confirmed by various knockdown experiments. Previous data from Chen and colleagues support our results, revealing that PTEN inactivation results in cellular senescence in mouse embryonic fibroblasts 35 . Given that melanomas preferentially activate the PI3K pathway through inactivation of PTEN, it can be hypothesized that c-Jun expression in early melanoma stages protects melanoma cells with decreasing PTEN levels from this growth arrest.
Taken together, our study confirms a crucial role of the transcription factor c-Jun in melanoma development and progression in PTEN + melanoma cells to overcome apoptosis and promote malignancy not only in vitro but also in vivo. TCGA data analysis by cBioPortal confirm and support our latest findings. Thus, the detection of c-Jun and PTEN coexpression in melanoma represents a promising diagnostic marker for highly aggressive melanoma cells with the potential to result in MET with the loss of both c-Jun and PTEN.
Further investigations will be necessary to determine the detailed tumor-promoting consequences of c-Jun activity in malignant melanoma.

Analysis of gene expression by quantitative PCR
cDNAs of total RNA fractions were generated using Super-Script II Reverse Transcriptase Kit (Invitrogen, Groningen, The Netherlands). qRT-PCR was performed on a LightCycler (Roche, Mannheim, Germany) as described previously 5 . Annealing and melting temperatures were optimized for each primer set ( Table 2). ß-Actin was used for normalization.

RNA-Seq library preparation and mapping
RNA-Seq samples and libraries were prepared as described previously 37 . Library preparation was performed with at least two biological replicates. Sequencing was performed according to the paired-end RNA sequencing protocols from Illumina on a HiSeq2000 with paired-end module (Illumina, Inc.). Fifty samples were sequenced from each side of a fragment~100 bp long with an average number 20 million reads per sample.
Paired-end reads were aligned to the human reference genome sequence (hg38) using the STAR alignment software (v 2.5.2a) 38 . Only reads that mapped to a single unique location were considered for further analysis. Counts for RNA-Seq reads were calculated using the feature-counts software (v 1.4.6-p5).

Chromatin immunoprecipitation
ChIP was performed with the human melanoma cell lines Sbcl-2, WM3211, WM1366, WM793, WM1158, WM9 and NHEM that had not exceeded six passages. In terms of ChIP experiments with transcription factors, we performed double crosslinking with disuccinimidyl glutarate (DSG, Thermo Scientific, Rockford, USA) before a fixation step with formaldehyde. Chromatin from 10 × 10 6 cells of each cell line was crosslinked in 1/10th volume of fixation buffer (50 mM HEPES/KOH, pH 7.9, 11% formaldehyde) for 10 min at room temperature and quenched by 0.125 M glycine. After two washes with PBS and PMSF, the cells were scraped and centrifuged at 4°C and 3500×g for 10 min. The supernatant was discarded, and the pellet was resuspended in 15 ml of lysis buffer 1 (5 mM PIPES [pH 8.0], 85 mM KCl, 0.5% NP-40, 1x Roche Complete, EDTA-free protease inhibitor) and incubated for 10 min on ice. After centrifugation of 5 min at 4°C and 3500×g, the supernatant was discarded, and the pellet resuspended in 15 ml lysis buffer 2 (50 mM Tris-HCl, pH 8.0, 10 mM EDTA, 1% SDS, 1× Roche Complete, EDTA-free protease inhibitor). The suspension was then incubated on ice for an additional 10 min and was examined under the microscope for quality assessment of nuclear prep. The nuclei were pelleted at 4°C and 3500×g for 5 min. The supernatant was discarded, and the pellet was resuspended in the desired volume of sonication buffer (10 × 10 6 cells in 450 µl sonication buffer: 16.7 mM Tris-HCl [pH 8.0], 1.2 mM EDTA, 167 mM NaCl, 0.01% SDS, 1.1% Triton X-100, 1× Roche Complete, EDTAfree protease inhibitor). Cross-linked chromatin was sheared to an average DNA fragment size of~400-600 bp using a Branson Sonifier 250 (Danbury, CT). After centrifugation, 2% of supernatant was used as input. After preclearing with Sepharose CL-4B beads (Sigma) for 2 h, chromatin samples from 10 × 10 6 cells were immunoprecipitated overnight with 5 μg of rabbit polyclonal antibody anti-c-Jun (Santa Cruz Biotechnology Inc.) or anti-H3K37ac (Abcam). Immunocomplexes were recovered by 3 h of incubation with protein A-Sepharose beads (GE Healthcare) at 4°C. These beads were washed thrice and resuspended in 1 mM EDTA and 10 mM Tris-HCl, pH 8.

ChIP-seq and mapping
DNA from ChIP (10-50 ng) was adapter ligated and polymerase chain reaction amplified according to the manufacturer's instructions (Illumina, San Diego, USA). ChIP fragments were sequenced for 36 cycles on Illumina Genome Analyzer according to the manufacturer´s protocol. Sequence tags of all experiments were mapped to the current human reference sequence (GRCh8/hg38) using Bowtie 2 (v 2.2.7) 39 , and only uniquely mapped tags were used for downstream analysis. Tag counts were normalized to 10 7 specifically mapped tags.

Sequencing data analysis
The raw RNA-Seq counts were used for differential gene expression analysis that was performed using DESeq2 (v 1.14.1) 40 . Differentially expressed genes with a false-discovery rate (FDR) <0.05 were regarded as statistically significant. Normalization was performed by library size based on the raw counts.
Genome Ontology annotation and RNA-Seq reads annotation were performed using scripts provided by STAR (based on GENCODE V24).
Analysis of mapped ChIP-seq tags was performed using HOMER, which is freely available at http://biowhat.ucsd. edu/homer/ 41 . The ChIP-Seq sequence data were assessed for abnormal GC content, excessive clonal amplification (multiple tags starting at the same genomic position), inherent sequence bias in the sequenced tags and surrounding genomic positions and contamination with plasmid/cDNA sequences. The identification of ChIP-Seq peaks (bound regions) was performed using a custom approach (HOMER) that combines features of previously published methods. Peaks were defined at a 0.001 Attachment assay using the xCELLigence system Attachment was determined using the xCELLigence System of Roche Diagnostics (Penzberg, Germany) as described previously 42 . This instrument monitors the behavior of the cells in real time by measuring the electrical impedance across interdigitated microelectrodes covering the bottom of 'E-plates'. Electrode impedance is displayed as cell index values.

Cell proliferation assay
Melanoma cells (4000 cells/well) were seeded in a 96well plate, the medium was aspirated and cells were incubated with the dye-binding solution from the CyQUANT ® NF Cell Proliferation Assay Kit (Invitrogen) according to the manufacturer's protocol. Finally, fluorescence was measured in an EMax Microplate Reader (MWG Biotech, Ebersberg, Germany) after 1 and 72 h.

Migration assay
Migration potential of melanoma cells was quantified using the Cultrex 96-Well Cell Migration Assay (Trevigen, Gaithersburg, USA) according to the manufacturer's instructions as described previously 43 . Briefly, melanoma cells were seeded into the upper compartment of the provided 96-well plate (20,000 cells/well) in RPMI. The lower compartment was filled with RPMI supplemented with conditioned medium from fibroblasts and 20% (directed migration) or 0% (undirected migration) FCS, respectively, as chemoattractants. After incubation at 37°C for 4 h cell migration was quantified by fluorimetry with an EMax Microplate Reader (MWG Biotech, Ebersberg, Germany).

GO analysis
The Database for Annotation, Visualization, and Integrated Discovery (DAVID; v 6.8) 44,45 was used to investigate biological meaning of the obtained differential expressed gene lists by DESeq2 of our RNA-Seq data and also GO terms for differentially expressed genes potentially regulated by c-Jun. The STRING database (v 10.5) 46,47 was used to visualize predicted associations of differentially expressed genes or differentially expressed genes potentially regulated by c-Jun, respectively. Moreover, enriched GO terms of different gene sets were also identified by STRING.
GREAT 50,51 was used to identify GO terms of noncoding genomic regions (ChIP-Seq peaks) by analyzing the annotations of the nearby genes. The association rule was set as follows: proximal, 10 kb upstream and 10 kb downstream (any gene in this interval relative to input regions is included); plus distal, up to 1000 kb (if no gene is present in the proximal interval, the closest gene in this distal interval is included).

Statistical analysis
In the bar graphs, results are expressed as mean ± s.d. (range). Comparisons between qRT-PCR data were made using the Student's unpaired t-test. A P-value of <0.05 was considered as statistically significant (NS: not significant, *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001). Group comparisons of Next Generation Sequencing data were made by the non-parametric Wilcoxon test and of TCGA patient data by the Welch's t-test after removing identified outliers. All calculations were performed using the Graph-Pad Prism Software (V8, GraphPad Software, Inc., San Diego, CA, USA) or RStudio (Version: 1.1.456, RStudio, Inc., Boston, MA, URL, http://www.rstudio.com).