mutual exclusivity analysis of genetic and epigenetic ... · methylation (reviewed in ref. 6)....

Cell Cycle and Senescence

Mutual Exclusivity Analysis of Genetic and EpigeneticDrivers in Melanoma Identifies a Link Between p14ARF

and RARb Signaling

Christina Dahl1, Claus Christensen1, G€oran J€onsson2, Anders Lorentzen1, Mette Louise Skjødt1,Åke Borg2, Graham Pawelec3, and Per Guldberg1

AbstractMelanoma genomes contain thousands of alterations including: mutations, copy number alterations, structural

aberrations, and methylation changes. The bulk of this variation is stochastic and functionally neutral, with only asmall minority representing "drivers" that contribute to the genesis and maintenance of tumors. Drivers are oftendirectly or inversely correlated across tumors, reflecting the molecular and regulatory signaling pathways in whichthey operate. Here, a profile of genetic and epigenetic drivers in 110 human melanoma cell lines was generated andsearched for non-random distribution patterns. Statistically significant mutual exclusivity was revealed amongcomponents of each of the p16INK4A-CDK4-RB, RAS-RAF-MEK-ERK and PI3K-AKT signaling pathways. Inaddition, an inverse correlation was observed between promoter hypermethylation of retinoic acid receptor b(RARB) and CDKN2A alterations affecting p14ARF (P < 0.0001), suggesting a functional link between RARbsignaling and the melanoma-suppressive activities of p14ARF. Mechanistically, all-trans retinoic acid (ATRA)treatment increased the expression of p14ARF in primary human melanocytes and the steady-state levels of p14ARF

in these cells were shown to be regulated via RARb. Furthermore, the ability of ATRA to induce senescence isreduced in p14ARF-depleted melanocytes, and we provide proof-of-concept that ATRA can induce irreversiblegrowth arrest in melanoma cells with an intact RARb-p14ARF signaling axis, independent of p16INK4A and p53status.

Implications: These data highlight the power of mutual exclusivity analysis of cancer drivers to unravel molecularpathways and establish a previously unrecognized cross-talk betweenRARb and p14ARFwith potential implicationsfor melanoma treatment. Mol Cancer Res; 11(10); 1166–78. �2013 AACR.

IntroductionMelanoma develops from pigment-producing melanocy-

tic cells and is themost aggressive formof skin cancer, causingnearly 50,000 deaths worldwide each year. The incidence ofmelanoma has increased markedly over the past 50 years andcontinues to increase in many Western populations. Early

detection and surgical excision of the primary tumor can becurative for the majority of patients, but advanced, dissem-inated forms of the disease are associated with a dismalprognosis (1). Until recently, the treatment options formetastatic melanoma were limited because of the intrinsicresistance of melanoma cells to conventional forms of anti-cancer therapy. Significant progress has come from differentlines of treatment, including the combination of dacarbazinewith ipilimumab (a monoclonal antibody against CTLA-4),the combination of a gp100peptide vaccinewith interleukin-2, monotherapy with trametinib [a mitogen-activated pro-tein/extracellular signal–regulated kinase (MEK) inhibitor],and monotherapy with the BRAF inhibitors vemurafenib(also knownas PLX4032) and dabrafenib (reviewed in ref. 2).In phase III clinical trials, these treatments have all beenshown to induce complete or partial tumor regression andincreased survival in patients with metastatic melanoma.However, intrinsic or acquired resistance to these treatmentoptions remains a major concern.A milestone in melanoma research was the discovery of

BRAF mutation as a frequent and early event in melanomadevelopment (3). Extensive work has shown that BRAFmutations (predominantly p.V600E) occurring in more

Authors' Affiliations: 1Danish Cancer Society Research Center, Copen-hagen, Denmark; 2Department of Oncology, University Hospital, Lund,Sweden; and 3Department of Internal Medicine II, University of T€ubingen,T€ubingen, Germany

Note: Supplementary data for this article are available at Molecular CancerResearch Online (http://mcr.aacrjournals.org/).

Current address for A. Lorentzen: Eucaryotic Cell Biology Research Group,Department of Science, Roskilde University, Roskilde, Denmark.Current address for M.L. Skjødt: Department of Mammalian Cell Technol-ogy, Biopharmaceutical Research Unit, Novo Nordisk A/S, Ma

�løv,

Denmark.

Corresponding Author: Per Guldberg, Danish Cancer Society ResearchCenter, Strandboulevarden 49, DK-2100 Copenhagen, Denmark. Phone:45-3525-7500; Fax: 45-3525-7721; E-mail: [email protected]

doi: 10.1158/1541-7786.MCR-13-0006

�2013 American Association for Cancer Research.

MolecularCancer

Research

Mol Cancer Res; 11(10) October 20131166

on October 5, 2020. © 2013 American Association for Cancer Research. mcr.aacrjournals.org Downloaded from

Published OnlineFirst July 12, 2013; DOI: 10.1158/1541-7786.MCR-13-0006

http://mcr.aacrjournals.org/

than 50%ofmelanomas lead to constitutive activation of theRAS–RAF–MEK–ERK signaling pathway (3) and confer astate of "oncogene addiction" on melanoma cells (4, 5).These findings provided important insights into the biologyof melanoma and paved the way for the development oftherapeutic small-molecule drugs directed against the BRAFoncogene (reviewed in ref. 2). The work leading to thediscovery of BRAFmutations had been preceded by decades-long candidate gene approaches, which identified othergenes recurrently mutated in melanoma, including NRAS,the CDKN2A locus (also known as INK4A/ARF) and PTENas some of the most frequent targets (reviewed in ref. 6). Inaddition to these mutational events, other types of molecularalteration have been shown to contribute to melanomadevelopment and progression, including gains of the KIT,MITF, CCND1, and MYC proto-oncogenes and transcrip-tional silencing of tumor suppressors by promoter hyper-methylation (reviewed in ref. 6).Data from recent whole-genome or -exome sequencing of

melanomas (7–12), together with genome-wide studies ofcopy number variations (13–15) and DNA methylationchanges (16), have added new layers of complexity to ourunderstanding of the melanoma genome. An importantchallenge arising from these studies is to identify "driver"events that contribute to melanoma genesis and mainte-nance, and to distinguish them from functionally neutral"passenger" events. Another challenge is to identify themolecular signaling pathways that are perturbed by driverevents. One increasingly used approach to address thesechallenges is mutual exclusivity analysis. The rationale forthis approach is that, although the patterns of genomicalterations in individual tumors are exceedingly diverse andcomplex, drivers will tend to cluster within a limited numberof biologic pathways that are essential for the tumorigenicprocess (17). Once a component of a pathway becomesaltered through a specific genomic alteration and confers aselective advantage on the tumor cell, second hits in the samepathway will provide no additional advantage and thus willbe less likely to accumulate. The consequences of suchselection processes are that two hits in the same pathwayrarely occur in the same tumor, and that different eventswithin a pathway are inversely correlated across tumors.Genomic analyses of melanoma have shown mutual

exclusivity between BRAFV600E and NRAS mutations, bothleading to activation of the RAS–RAF–MEK–ERK pathway(3), between NRAS mutations and inactivating events inPTEN, both leading to activation of the PI3K–AKT path-way (18), and among mutations in genes encoding compo-nents of the p16INK4A–CDK4–RB senescence barrier (19).Less is known about the functional significance of other, lessfrequent alterations. Moreover, integrated analyses of thevarious types of genome alterations, including copy numbervariations and DNA-methylation changes, have not beenconducted. Here, we have examined the status of knowngenetic and epigenetic drivers in 110 human melanoma celllines and assessed their pairwise relationships. This approachconfirmed the examples of mutual exclusivity describedearlier and further uncovered an inverse correlation between

epigenetic silencing of the retinoic acid receptor b (RARb)and genetic inactivation of the melanoma suppressor,p14ARF. Functional studies in melanocyte and melanomamodels suggest that p14ARF is a downstream target of RARbsignaling, and that the status of RARb and p14ARF inindividual melanomasmay predict the response to treatmentwith retinoic acid, the biologically active form of vitamin A.

Materials and MethodsCells and reagentsMelanoma cell lines were obtained from The European

Searchable Tumour Line Database (ESTDAB; http://www.ebi.ac.uk/ipd/estdab) and maintained in RPMI-1640 medi-um containing 10% FBS and antibiotics at 37�C and 5%CO2. The 110 cell lines used in this study were previouslycharacterized for surface markers and tumor-associated anti-gens, and wereDNA fingerprinted to control for duplicationand cross-contamination (20). Primary human melanocyteswere obtained from Invitrogen and maintained in Medium254 containing human melanocyte growth supplement(Invitrogen) at 37�C and 5% CO2. All-trans retinoic acid(ATRA) and 5-aza-20-deoxycytidine were purchased fromSigma-Aldrich. LE135 was purchased from Santa CruzBiotechnology.

Mutations and homozygous deletionsGenomic DNA from melanoma cell lines was isolated

using standard procedures. Sequence analysis of CDKN2Awas conducted as described previously (21). Pyrosequencingof the BRAF c.1799T>A mutation was conducted using thePyroMark Q24 platform (Qiagen) and previously describedprimers (22). Mutation analysis of BRAF (exons 11 and 15),NRAS (exons 2 and 3), KIT (exons 9, 11, 13, and 17),PIK3CA (exons 9 and 20), PIK3R1 (exon 9, 10, and 12),IDH1 (exon 4), FGFR1 (exon 7), GNAQ (exon 5), TP53(exons 4–9), CTNNB1 (exon 3), CDK4 (exon 2), RB1(exons 1–27), and PTEN (exons 1–9) was conducted usinga combination of PCR and denaturing gradient gel electro-phoresis followed by direct sequence analysis. Primers arelisted in Supplementary Table S5. Analysis ofRB1mutationswas restricted to cell lines with wild-type CDKN2A andCDK4. Homozygous CDKN2A and PTEN deletions weredetermined as described previously (23, 24).

Copy number assessmentCopy numbers of MITF, MYC, AKT3, CDK4, and

CCND1were determined by quantitative PCR (qPCR) usingthe LightCycler 2.0 instrument (Roche) and the FastStartDNA MasterPLUS SYBR Green I Kit (Roche). The LINE-1retrotransposon was used as a reference gene for all copynumber analyses. A mixture of DNA from 6 human healthydonors was used as a diploid control. Copy number gain wasdefined as a relative copy number of � 4. Primer sequencesand assay conditions are listed in Supplementary Table S5.

Bisulfite treatment and methylation analysisGenomic DNA (500 ng) was bisulfite-converted accord-

ing to the standard procedures (25) or using the EZ DNA

p14ARF as a Target of Retinoic Acid Signaling

www.aacrjournals.org Mol Cancer Res; 11(10) October 2013 1167




Methylation-Gold Kit (Zymo Research). The methylationstatus of the INK4A, APC, RARB, PYCARD, RASSF1A, andIGFBP7 promoter regions was determined by mehylation-specific melting curve analysis (MS-MCA; ref. 26) using theLightCycler 1.1 and 2.0 instruments (Roche) and theFastStart DNA Master SYBR Green I Kit (Roche). Primersequences and PCR conditions are listed in SupplementaryTable S5. Enzymatically methylated DNA (CpGenomeUniversal Methylated DNA; Millipore) was used as a meth-ylation-positive control.

Quantitative reverse transcription PCRTotal RNA was isolated from cells using the RNeasy Kit

(Qiagen) and quantified using a NanoDrop spectrophotom-eter (NanoDrop Technology, Inc.). Only RNA sampleswith an A260/A280 ratio of 1.8 or higher were used. Reversetranscription was conducted in reactions containing 5 mg oftotal RNA, oligo(dT) and random hexamers as primers andSuperScript III reverse transcriptase (Invitrogen). Quanti-tative reverse transcription PCR (qRT-PCR) was conductedusing the Roche LightCycler 2.0 and the FastStart DNAMasterPLUS SYBR Green I Kit (Roche). Primer sequencesand assay conditions are listed in Supplementary Table S5.

ImmunoblottingCytoplasmic and nuclear extracts were fractionated using

the NE-PER Nuclear and Cytoplasmic Extraction ReagentKit (Pierce) and analyzed independently. Protein concen-tration was measured using the Coomassie Plus ProteinAssay Reagent (Pierce). p14ARF expression was analyzed by12%NuPAGE gels usingMES running buffer (Invitrogen).Membranes were incubated overnight with 2 mg/mL anti-p14ARF antibody (Clone 14P03; Labvision) followed by 2hours of incubation at room temperature.

Transfection and RNA interferenceNormal human melanocytes were transfected using the

Amaxa nucleofector programU-024 as recommended by themanufacturer. To obtain effective transient RNA interfer-ence, melanocytes were transfected twice using 1 mmol/LON-TARGETplus SMARTpool siRNA to RARb (Dhar-macon). ON-TARGETplus siCONTROL Nontargetingsi-RNA (Dharmacon) was used as control.

p14ARF-depleted melanocytesHuman melanocytes stably expressing short hairpin RNA

(shRNA) targeting CDKN2A will be described in detailelsewhere (Christensen and Guldberg, manuscript in prep-aration). In brief, melanocytes were transfected with pSuperplasmids containing resistance to G418 (5), and two inde-pendent cultures of stable transfectants were generated:HEM-1, which expresses shRNA to a target sequence withinCDKN2A exon 2, andHEM-2, which expresses shRNA to atarget within CDKN2A exon 1a and has spontaneously lostp14ARF. Both cultures were maintained in Medium 254containing human melanocyte growth supplement (Invitro-gen), stem cell factor (100 ng/mL; Stratmann Biotech), andendothelin-1 (10 ng/mL; Sigma-Aldrich).

Senescence-associated b-galactosidase assayCellular senescence was assessed by using the Senescence

b-Galactosidase Staining Kit (Cell Signaling Technology)according to the manufacturer's instructions. Senescence-associated b-galactosidase (SA-b-Gal)–positive cells wereidentified by their blue-green cytoplamatic staining under�200 magnification and a total of at least 70 cells werecounted from 10 random fields in each experiment. Allmicroscopic fields were scored independently by 2 personsfor determination of the mean percentage of positivelystained cells.

Growth ratesCells were seeded at low density in 25 cm2

flasks andallowed to attach overnight. The next day, the cells wereincubated with 10 mmol/L ATRA or vehicle [dimethylsulfoxide (DMSO)]. Media was replaced every second dayfor 16 days when control population treated with vehiclereached confluence. Cells were counted with Trypan blueexclusion at day 0, 4, 8, 12, and 16.

Growth inhibition assaysCells were seeded into 24-well plates (1.5–2.5� 103 cells/

well) and allowed to attach overnight. Themediumwas thenreplacedwith 10mmol/LATRA (Sigma) or vehicle (DMSO)and grown in the presence or absence of the drug for 14 to 19days. Each cell line was fixed at the same time and stainedwith crystal violet. Crystal violet was subsequently extractedand the absorbance was read at 595 nm in a microtiter platespectrophotometer. Relative absorbance was normalized tothe control cells treated with vehicle.

Cell proliferationBromodeoxyuridine (BrdUrd) incorporation was deter-

mined using the BrdU Cell Proliferation Assay Kit (CellSignaling). In brief, cells were seeded in a 96-well format (2� 103 cells/well) and cultured for 24 hours in the presence of10 mmol/L BrdUrd. Cells were fixed, denaturated, andreacted with detection antibody solution for 1 hour andhorseradish peroxidase (HRP)-conjugated secondary anti-body for 30 minutes. 3,30,5,50-Tetramethylbenzidine(TMB) substrate was added for 25 minutes, and absorbancewas measured at 450 nm.

Cell-cycle analysisCell-cycle analysis was conducted by flow cytometry

(FACSVerse BD and FACSDiva BD). Briefly, 106 cells weretrypsinized, fixed in 70% ethanol, washed and resuspendedin PBS containing 10 mg/mL propidium iodide (Sigma) and1 mg/mL RNase (Sigma). A total of 20,000 events werecollected per sample and cell-cycle profiles were obtainedusing the FlowJo V10 software. When examining the cell-cycle distribution, sub-G1 and polyploid cells were excludedfrom the analysis. G1-, S-, andG2-phase gates were applied tohistograms by manually defined distribution peaks.

Statistical analysisAll calculations were carried out in R (http://www.r-proj-

ect.org). Mutation, deletion, amplification, and promoter

Dahl et al.

Mol Cancer Res; 11(10) October 2013 Molecular Cancer Research1168




hypermethylation eventswere transformed into dichotomousvariables and analyzed for correlations using the Pearsoncorrelation coefficient. The significance level of these correla-tions was determined using the Fisher exact test.

ResultsGenetic and epigenetic drivers in human melanoma celllinesDNA from 110 unrelated melanoma cell lines was sys-

tematically examined for specific mutation, deletion, ampli-fication, and promoter hypermethylation events. A list ofgenes and the frequencies and types of alterations are shownin Fig. 1, and detailed information is provided in Supple-mentary Tables S1–S4. A total of 646 alterations wereidentified in this series ofmelanoma cell lines, with an averageof 5.9 (range, 0–10) alterations per cell line. One cell line(ESTDAB-097) carried none of the drivers investigated, andESTDAB-098 cells carried low copy number gains ofCCND1 and MITF as the only changes.Mutations were identified in known hot-spot regions of

10 different proto-oncogenes (Fig. 1; Supplementary TablesS1 and S2). The most frequent target was BRAF, whichharbored mutations in 75 of the cell lines (68%), with p.V600E accounting for 66 of these events (88%). MutationsinNRAS were found in 18 of the cell lines (16%), and 16 ofthese mutations affected residue Q61. Mutations in theremaining proto-oncogenes were less frequent, affecting

residues K22 or R24 of CDK4 (N ¼ 8), residues D32 orS45 of CTNNB1 (N ¼ 4), residue R132 of IDH1 (N ¼ 3),residues E545 orQ546 ofPIK3CA (N¼ 2), residueQ209 ofGNAQ (N¼ 2), residue L576 of KIT (N¼ 1), residue P252of FGFR1 (N ¼ 1), and residues 464–468 of PIK3R1 (N ¼1). The two cell lines harboring GNAQ mutations wereestablished from uveal melanomas, consistent with previousreports showing thatGNAQmutations occur at high rates inthis melanoma subtype (27).Mutations and/or deletions of the tumor suppressors

CDKN2A, PTEN, and TP53 were identified in 93 (85%),33 (30%), and 26 (24%) of the cell lines, respectively (Fig. 1and Supplementary Table S1). The majority of CDKN2Aalterations were predicted to affect both p16INK4A andp14ARF; however, subsets of alterations were specific for eitherprotein, including deletion (N ¼ 2) or mutation (N ¼ 1) ofexon 1a (specific for p16INK4A), deletion of exon 1b (specificfor p14ARF; N ¼ 1), and missense mutations in exon 2potentially affecting either p16INK4A (N ¼ 3) or p14ARF

(N ¼ 1). One specific CDKN2A mutation affecting bothp16INK4A and p14ARF (p.R80X/p.P94L) was found in five ofthe cell lines. In the following, the gene symbols INK4A andARF will be used to specify, when appropriate, whetherCDKN2A alterations have a predicted effect on p16INK4A andp14ARF, respectively.A total of 73 copy number gains were identified, with

frequencies of 25% for CCND1, 19% for MITF, 15%

Figure 1. Profile of genetic andepigenetic alterations in melanomacell lines (N ¼ 110).






for MYC, 5% for AKT3, and 3% for CDK4 (Fig. 1 andSupplementary Table S3). None of the AKT3 gains wasgreater than 5.A total of 255 promoter hypermethylation events were

identified, with frequencies of 56% for IGFBP7, 55% forRASSF1A, 53% for PYCARD, 45% for RARB, 13% forAPC, and 11% for INK4A (Fig. 1 and Supplementary TableS4). Overall, 99 of the cell lines (90%) were hypermethy-lated in one or more of the six promoter regions, 77 (70%)had two or more hypermethylated genes, and 50 (45%) hadthree or more hypermethylated genes.

Pathway analysis and pairwise associationsTable 1 lists the genomic alterations grouped according to

known molecular pathways. Genes encoding components ofthe RAS–RAF–MEK–ERK pathway were altered in 86% ofthe cell lines, the majority of which (96%) carried mutationsin BRAF and/or NRAS. Consistent with previous evidenceshowing that activating mutations in KIT, FGFR1, andGNAQ can activate the RAS–RAF–MEK–ERK pathway,cell lines carrying mutations in either of these genes carried

wild-type copies ofBRAF andNRAS. The p16INK4A–CDK4–RB senescence barrier was affected in 91% of the cell lines,either through inactivating INK4A events, RB1 mutationsand/or amplification, or mutation of CDK4; genetic eventsaffecting the PI3K–AKT signaling pathway were found in53% of the cell lines, with alterations in PTEN and NRASaccounting for 86% these events; and the p53–p14ARF axiswas affected in 79% of the cell lines, either through TP53mutation, ARF mutation/deletion, or both.For genes that were altered in at least four of themelanoma

cell lines, we conducted pairwise comparisons using thePearson correlation coefficient (Table 2). Negative associa-tions were found between BRAF and NRAS mutations,which only coexisted in two cell lines harboring non-p.V600E BRAF mutations (P ¼ 5.8 � 10�8), CDK4 andINK4A alterations (P¼ 2.2� 10�5), andNRAS and PTENalterations (P ¼ 0.012). PTEN alterations correlated posi-tively with BRAF mutations (P ¼ 9.6 � 10�5).Among other statistically significant direct and inverse

correlations (see Table 2), one particularly interesting find-ing was an inverse correlation between CDNK2A alterationsand hypermethylation of the RARB promoter. Notably, thisassociation was much stronger for ARF alterations (P ¼ 2.7� 10�5) than for INK4A alterations (P¼ 0.025). Of the 110cell lines, 52 harboredmutation or deletion ofARF andwild-type RARB (47%), 26 had RARB promoter hypermethyla-tion andwild-typeARF (24%), and 23 harbored alteration ofboth genes (21%). Only nine cell lines carried both wild-type ARF and RARB (8%). Notably, while RARB hyper-methylation was under-represented in cell lines with ARFalterations, it was over-represented in cell lines carryingCDKN2A alterations that specifically target the expressionor function of p16INK4A, for example, INK4A promoterhypermethylation events andmutations and deletions affect-ing exon 1a (14 of 18; x2 test; P < 0.05).

Retinoic acid induces p14ARF expression in humanmelanocytes via RARbPrevious studies have shown that human epidermal mel-

anocytes are growth inhibited byATRA (28, 29) and that theexpression of the RARb2 isoform [the predominant receptormediating the proliferation inhibitory effects of ATRA (30)]is often lost in melanomas due to promoter hypermethyla-tion. As our genomic data suggested a link between RARBand ARF in melanoma, we conducted a series of experimentsto address whether p14ARF is a downstream target gene of theRARb-mediated signaling cascade. We first asked whetherthe expression of p14ARF in melanocytes is altered uponshort-term exposure to ATRA. In two independent culturesof primary human melanocytes, treatment with 10 mmol/LATRA for 48 hours led to upregulation of RARb2, a knowndirect target of retinoic acid (ref. 31; and data not shown).Interestingly, in the same cells, ATRA also increased theexpression of p14ARF at both themRNA and protein levels ina dose-dependent manner (Fig. 2A). To examine whetherthis ability of ATRA to induce p14ARF expression is medi-ated via RARb, we cultured melanocytes in the presence orabsence of the RARb antagonist LE135 (32). As shown

Table 1. Molecular pathway distribution ofgenome alterations in melanoma cell lines(N = 110)

Pathway Gene(s)No. of celllines (%)

RAS–RAF—MEK—ERK BRAF 73 (66)NRAS 16 (15)KIT 1 (1)FGFR1 1 (1)GNAQ 2 (2)BRAF þ NRAS 2 (2)WT 15 (14)

p16INK4A-CDK4-RB INK4A 88 (80)CDK4 8 (7)RB1 1 (1)INK4A þ CDK4 3 (3)WT 10 (9)

PI3K—AKT PTEN 30 (27)NRAS 16 (15)KIT 1 (1)FGFR1 1 (1)PIK3CA 1 (1)PIK3R1 1 (1)AKT3 4 (4)NRAS þ PTEN 1 (1)NRAS þ PIK3CA 1 (1)PTEN þ AKT3 2 (2)WT 52 (47)

p53–p14ARF TP53 12 (11)ARF 61 (55)TP53 þ ARF 14 (13)WT 23 (21)

Dahl et al.





Tab

le2.

Assoc

iatio

nsam

ongge

netic

andep

igen

etic

driv

ersin

melan

omace

lllines

(N¼

110)

AKT3

APC

ARF

BRAF

CCND1

CDK4

MYC

CTNNB1

IGFB

P7

INK4A

MITF

NRAS

PTEN

PYCARD

RARB

RASSF1

ATP53

AKT3

0.03

�0.01

0.08

�0.04

�0.08

0.01

�0.05

�0.11

0.00

�0.01

�0.11

0.02

�0.01

0.03

�0.02

0.05

APC

�0.21

�0.03

�0.03

�0.13

0.07

�0.07

0.12

0.03

�0.12

0.20

�0.19

0.09

0.26

��0.02

�0.02

ARF

0.20

�0.03

�0.23�

0.23

�0.03

0.03

0.57

��

�0.12

�0.12

0.19

�0.10

�0.41�

��0.08

�0.17

BRAF

0.21

�0.03

0.06

0.13

�0.09

0.05

0.18

�0.54�

��0.36

��

�0.14

�0.02

�0.04

�0.17

CCND1

0.16

0.06

0.23

�0.12

�0.07

0.05

�0.20

0.09

�0.05

0.00

0.01

0.08

CDK4

�0.14

0.10

0.05

�0.49�

��0.15

�0.15

�0.02

0.01

0.19

0.06

�0.04

MYC

0.06

0.16

0.12

�0.07

�0.18

0.12

0.13

0.05

�0.04

0.07

CTN

NB1

�0.02

�0.04

0.03

�0.09

0.08

�0.11

�0.08

�0.12

0.01

IGFB

P7

0.13

0.10

�0.01

0.10

0.34

��

0.09

0.08

0.06

INK4A

�0.21�

0.01

0.09

0.10

�0.22�

�0.03

�0.09

MITF

�0.15

�0.02

�0.14

0.08

�0.02

0.06

NRAS

�0.24�

0.07

0.05

0.11

0.10

PTE

N0.06

�0.31�

�0.08

�0.08

PYCARD

�0.03

0.23

�0.01

RARB

�0.17

0.10

RASSF1

A�0

.09

TP53

NOTE

:Correlatio

nssh

ownarefco

efficien

tsforp

airsof

dicho

tomou

sva

riables.Correlatio

nsforw

hich

P<0.05

,P<0.01

,orP

<0.00

1aresh

ownwith

asing

le,d

ouble,o

rtrip

leas

teris

k,resp

ectiv

ely.






in Fig. 2B, LE135 effectively blocked the ATRA-inducedexpression of p14ARF. As LE135may also act as an antagonistof RARa, we extended our analysis by transiently transfect-ing normal humanmelanocytes with either siRNAs targetingRARb or control siRNAs. As shown in Fig. 2C, knockdownof RARb resulted in a more than 60% reduction in thesteady-state expression levels of p14ARF 48 hours posttrans-fection. Collectively, these data suggest that RARb signalingcontributes to the regulation of p14ARF expression inmelanocytes.

Cellular responses to ATRA in normal and p14ARF

-depleted melanocytesTo characterize in greater detail the cellular response of

normal melanocytes to ATRA, we first assessed the effects oncell proliferation bymeasuring the rate of DNA synthesis. Asshown in Fig. 3A, BrdUrd incorporation was reduced by50% after 10 days of growth in the presence of ATRA.Furthermore, this effect could be rescued by LE135, sug-gesting that the growth-inhibitory effects of ATRA are

mediated via RARb. To determine the effect on cell-cycledistribution, melanocytes were treated with ATRA for 10days and analyzed by fluorescence-activated cell sorting(FACS) analysis. These experiments showed an accumula-tion of cells in G1-phase of the cell cycle and a correspondingdecrease in the fraction of cells in S- andG2-phases (Fig. 3B).We next measured the activity of SA-b-Gal, a hallmark

of senescence (33). Treatment of two independent cultures ofmelanocytes with ATRA for 2 weeks increased the fraction ofSA-b-Gal–positive cells by 300% to 600% (Fig. 3C). Assenescence induced by oncogenic stimuli has been shown tobe accompanied by higher p14ARF expression in some celltypes (34) and as ATRA increased the expression of p14ARF innormal humanmelanocytes (Fig. 2A),we investigated the roleof p14ARF inmediatingATRA-induced growth arrest. To thisend, we used two independent lines of melanocytes thatexpress shRNAs againstCDKN2A sequences and express lowlevels of p14ARF (HEM-1 and -2; seeMaterials andMethods).When SA-b-Galwas used as a surrogatemarker of senescence,the response to ATRA was significantly reduced in

Figure2. ATRA induces p14ARF expression in humanmelanocytes viaRARb. A, expression levels of p14ARF in normal humanmelanocytes exposed toATRA for48 hours. Left, immunoblotting of a representative melanocyte culture. b-Actin was used as control for protein loading. Right, qRT-PCR. The p14ARF

expression levelswere normalized to porphobilinogendeaminase (PBGD) expression. The data represent themeanþSDof two independent experiments. B,inhibition of ATRA-induced p14ARF expression by the RARb antagonist LE135. Cells were grown for 48 hours in the presence or absence of ATRA (10 mmol/L)and in the presenceor absence of LE135 (1mmol/L). ThemRNAexpression levels of p14ARFweremeasuredby qRT-PCRandnormalized toPBGDexpression.The data represent the mean þ SD of two independent experiments. C, expression levels of RARb2 and p14ARF in normal human melanocytes transientlytransfectedwith siRNAs targetingRARb. TheRARb2andp14ARF expression levelsweredeterminedbyqRT-PCR48hours posttransfection andnormalized toPBGD expression. Expression in cells transfected with control siRNAs was set at 100%. The data represent themeanþSD of two independent experiments.

Dahl et al.





p14ARF-depletedcells comparedwithnormal counterpart cells(Fig. 3C and D). In conclusion, treatment of normal humanmelanocytes with ATRA is associatedwith loss of proliferativeactivity,G1 arrest, and increased activity of SA-b-Gal, and thisresponse may at least in part depend on p14ARF.

ATRA induces irreversible G1 arrest in melanoma cellswith an intact RARb–p14ARF axisThemajority of humanmelanomas are inherently resistant

or weakly sensitive to ATRA (35). Part of this resistance canbe explained by epigenetic silencing of RARb2, which isfound in 20% to 70% of melanoma cell lines (reviewed inref. 6; and this study). However, resistance mechanisms inmelanoma cells expressing RARb2 have not been elucidated.Given the strong genetic and functional links between RARband p14ARF, we addressed whether the status of RARB andARF could predict the sensitivity of melanoma to ATRA.

Specifically, we asked whether melanoma cells with an intactRARb–p14ARF axis are sensitive to ATRA. From the ninemelanoma cell lines identified with an unmethylated RARBpromoter and wild-type ARF, we excluded ESTDAB-097and ESTDAB-098 as they grew slowly in culture and carriedno melanoma-specific genome alterations, and ESTDAB-179, which did not express p14ARF even though noCDKN2A alterations were identified. All the remaining sixcell lines expressed p14ARF and RARb2 mRNA (data notshown). Consistent with the data obtained in melanocytes(Fig. 2A), treatment with ATRA increased the expressionlevels of p14ARF in these cell lines by 1.5- to 2.5-fold (data notshown). Interestingly, all six cell lines responded to treatmentwith ATRA (Fig. 4). Two of these lines harbored INK4Aalterations, one harbored a TP53 mutation, and one har-bored defects in both genes, suggesting that p16INK4A andp53 are dispensable for execution of ATRA-induced growth

Figure 3. Cellular responses to ATRA in normal and p14ARF-depletedmelanocytes. A, BrdUrd incorporation in primary humanmelanocytes grown for 48 hoursin the presence or absence of ATRA (10 mmol/L) and in the presence or absence of LE135 (1 mmol/L). B, FACS analysis showing representative cell-cycleprofiles of normal human melanocytes cultured for 10 days in the presence or absence of ATRA (10 mmol/L). C, ATRA-induced senescence in twocultures of normal human melanocytes (WT-1 and -2) and two independent pools of melanocytes expressing shRNAs against CDKN2A sequences (HEM-1and -2). Cells were treated with ATRA (10 mmol/L) for 2 weeks, and the number of senescent cells was determined by SA-b-Gal positivity. Columnsrepresent percentageof SA-b-Gal–positive cells. D, the samedata as in (C), butwith the increase in senescence in normalmelanocytes set at 100%.Statisticalanalysis was conducted using a paired t test. �, P < 0.05; ns, nonsignificant.






inhibition. Cell lines harboring RARB promoter hyper-methylation (with p14ARF expression) or ARF deletion (withRARb2 expression) had a poor response to ATRA (Fig. 4A

and C). Collectively, these results suggest that melanomaswith an intact RARb–p14ARF axis would potentially besusceptible to treatment with ATRA.

Figure 4. ATRA-mediated growth inhibition of melanoma cells with an intact RARb–p14ARF axis. A, dose-dependent ATRA sensitivity of melanoma cells withdifferent ARF and RARB status. Cells were treated with increasing concentrations of ATRA for 14 to 19 days. For each cell line, all dishes were fixed atthe same time, stained with crystal violet, and photographed. B, time-dependent effect of ATRA treatment on growth rate in ESTDAB-196 cells. Cells wereseeded at low density in 25-cm2

flasks and fresh medium with ATRA or vehicle was added every 48 hours for 16 days. Cells were counted with Trypan blueexclusion. Mean cell number þ SD of duplicate samples are shown. The growth curves shown are representative of three independent experiments. C,growth inhibition assays of melanoma cells. Cells were grown in the presence or absence of 10 mmol/L ATRA for 14 to 19 days and were then fixedand stainedwith crystal violet. The absorbancewas readat 595nm in amicrotiter plate spectrophotometer. The absorbance valuesof cells treatedwith vehiclewere set at 100%. Shown is the average þ SD of two independent experiments carried out in triplicate. Mut, mutated; Del, deleted; Met, methylated;WT, wild-type. D, expression levels of RARb2 in two melanoma cell lines (ESTDAB-013 and ESTDAB-079) with RARB promoter hypermethylation aftertreatment with 1 mmol/L 5-aza-CdR for 4 days. E, ATRA-induced growth inhibition of 5-aza-CdR–treatedmelanoma cells. Cells were grown in the presence orabsence of ATRA (10 mmol/L) and in the presence or absence of 5-aza-CdR (1 mmol/L) for 14 to 17 days. Fixation, staining, and absorbance measurementwere done as described in (C). The absorbance values of cells treated with vehicle were set at 100%. Shown is the average þ SD of two independentexperiments carried out in triplicate. Statistical analysis was conducted using a paired t test. �, P < 0.05; ns, nonsignificant.

Dahl et al.





In the context of our findings, a pertinent question iswhether restoration of RARb expression is able to restoreATRA sensitivity in melanoma cells. To address this ques-tion, we first treated two melanoma cell lines with RARBpromoter hypermethylation and wild-type ARF with theDNA-hypomethylating agent 5-aza-20-deoxycytidine (5-aza-CdR). As shown in Fig. 4D, treatment with 1 mmol/L 5-aza-CdR for 4 days reactivated RARb2 mRNA expres-sion in both cell lines. Furthermore, in contrast to untreatedcells, the growth of 5-aza-CdR–treated cells was significantlyinhibited by ATRA, suggesting that derepression of RARb2can restore ATRA sensitivity (Fig. 4E).We made several attempts to stably deplete RARb- and

p14ARF-expressingmelanoma cell lines of p14ARF to be able tostudy more directly the role of this protein in long-termcellular repsonses to ATRA. However, although we used thesame shRNAvectors that efficiently knocked downp14ARF innormal melanocytes (Fig. 2; and data not shown) and stableclones were readily generated with the empty vector, we wereunable to generate clones expressing shRNA vectors targetingp14ARF in any of these six cell lines (data not shown).Cell-cycle analysis of RARb- and p14ARF-expressing mel-

anoma cells treated with ATRA showed a decrease in thefraction of cells in S- and G2-phases of the cell cycle and anincrease in the fraction of cells inG1-phase (Fig. 5A). Analysisof the sub-G1 fraction didnot show an increase in the numberof apoptotic cells in ATRA-treated cultures (data not shown).After 3 weeks of growth in the presence of ATRA, BrdUrdincorporation was reduced by 50% (Fig. 5A), similar to whatwas observed in primary melanocytes (Fig. 2A). Cells werethen released and cultured for 1 week in medium without

exogenously added ATRA. As shown in Fig. 5B, BrdUrdincorporation was not increased after the release. Together,these results suggest that ATRA induces irreversible G1 arrestin melanoma lines with an intact RARb–p14ARF axis.

DiscussionWe have generated a profile of genetic and epigenetic

drivers in human melanoma-derived cell lines and searchedfor pairwise associations across all genes to identify possiblefunctional interplays. The use of established cell lines for thistype of approach has both advantages and disadvantages,which should be taken into account when interpreting ourdata. A major advantage of using cell line DNA as a templateis that some genomic changes, particularly homozygousdeletions and copy number gains, can be determined withgreater accuracy as they will not bemasked by contaminatingnormal tissue. Although cell lines may contain genomicartifacts introduced as a result of cell culturing, previousstudies have shown that cell lines usually are representative ofthe tumors from which they originate (36). The frequenciesof genome alterations found in our study are generally higherthan, but still compatible with the frequencies observed inmelanoma biopsies (6). A potential limitation of studiesbased on cancer cell lines, including ours, is that they areconfounded by a bias in favor of tumors that are able to growunder standard culture conditions and thus may havespecific cellular properties and genomic changes. Accord-ingly, the collection ofmelanoma cell lines used in our study,although relatively large in number, may not be clinicallyrepresentative.

Figure 5. Cellular responses to ATRA inmelanoma cells lines expressing RARb and p14ARF. A, representative FACS analyses showing the proportions of cellsin the different stages of the cell cycle after 3 of weeks of growth in the presence or absence of ATRA (10 mmol/L). B, DNA synthesis measured by BrdUrdincorporation. Cells were grown in the presence or absence of ATRA (10 mmol/L) for 3 weeks and then for 1 week without ATRA (release).






Previouswork on smaller collections ofmelanoma cell lines(37, 38) and uncultured melanomas have found a number ofdriver associations, which we confirmed in this study, includ-ing negative associations between BRAF and NRAS muta-tions, CDK4 and INK4A alterations, and NRAS and PTENalterations, and a positive correlation between PTEN andBRAF alterations. The functional implications of these asso-ciations have all been corroborated in various experimentalmodels (3, 19, 39). In addition to the above associations, aprevious study (38) suggested a negative association betweenARF (exon 1b) and TP53 alterations, consistent with theknown role of p14ARF as a positive regulator of p53 (34) andwith the consensus view that the relatively low frequency ofTP53 mutations in melanoma can be explained by a highfrequency of inactivatingARF events. An association betweenARF andTP53 alterations didnot reach statistical significancein our study, and thus it remains unclear whether p14ARF andp53 operate in the same or different molecular pathways inmelanoma. In support of the latter, previous work has shownthat p19Arf (the mouse homolog of p14ARF) can function as amelanoma suppressor by inducing p53-independent senes-cence (40).As the most important novel finding in our study, we

uncovered a strong inverse correlation between hypermethy-lation of the RARB promoter and ARF alterations. Thisassociation was particularly relevant as little is known aboutthe melanoma-suppressive properties of RARb, and asgenomic alterations of other core components of the retinoicacid signaling pathway have not been described in melano-ma. To corroborate the link between RARb and p14ARF

suggested by the genomic data, we first investigated whetherthe regulation of p14ARF expression was influenced by thechanges in RARb expression or ligand concentrations. Incultures of normal humanmelanocytes, steady-state levels ofp14ARF were increased at both the mRNA and protein levelupon short-term exposure to ATRA, whereas they werereduced in response to depletion or chemical inhibition ofRARb. These findings are consistent with an important roleof RARb signaling in the regulation of p14ARF expression inmelanocytes and would, to our knowledge, identify retinoicacid as the first naturally occurring inducer of p14ARF.Previous studies have shown that p14ARF expression can beinduced by hyperproliferative stimuli such as those elicitedby the Myc and E1A oncogenes (34), providing an expla-nation for the tumor-specific properties and expressionpatterns of p14ARF. Several reports have suggested thatp14ARF and p19Arf are not expressed in most normal cells,including cultured human melanocytes (41). However, ourdata show that ATRA can induce the expression of p14ARF innormal melanocytes in the nanomolar range, that is, atconcentrations similar to the concentrations of retinoids innormal epidermis (42). Considering the important role ofretinoids and RARs in embryonic development and cellulardifferentiation (43), p14ARF expression in melanocytes maydepend on the differentiation state of these cells, which is, atleast in part, controlled by retinoic acid (44). A functionalrole of p14ARF in normal differentiated melanocytes remainsunknown.

The implication of p14ARF as a melanoma suppressor hasbeen controversial and notoriously difficult to study, primar-ily due to the unusual structure of theCDKN2A locus, whichencodes p14ARF and p16INK4A in different reading frames(34). Although a critical role for p16INK4A in replicative andoncogene-induced melanocyte senescence has been welldocumented both in vitro and in vivo (33, 45, 46), thesestudies have provided little insight into a possible role ofp14ARF. Studies of cancer-associated CDKN2A mutationpatterns have suggested that exon 2 mutations preferentiallytarget p16INK4A function (38) and thus that p14ARF iscoincidentally inactivated in a high proportion of melanomasdue to the strong selection pressure against p16INK4A. Tobetter understand the association between RARb signalingand the melanoma-suppressive properties of the CDKN2Alocus, we conducted a series of experiments in the settings ofmelanocytes and melanoma cells. Consistent with previouswork (28, 29), we showed that ATRA could induce senes-cence in normal human melanocytes. This response wasdependent on CDKN2A, as senescence was less pronouncedin cells depleted of p14ARF and p16INK4A. Furthermore,while melanoma cell lines are generally resistant to retinoids(35), cell lines with wild-type copies of RARB and ARFresponded to treatment with ATRA. Importantly, some ofthe ATRA-responsive cell lines harbored INK4A and/orTP53 alterations, suggesting that the response to ATRA isindependent of the functions of p16INK4A and p53.p14ARF is lost in a great proportion of melanomas and is

considered as a bona fide tumor suppressor. In this context,it was unexpected that stable depletion of p14ARF was notcompatible with continued growth of melanoma cells withan intact RARb–p14ARF axis. This finding may suggest thatp14ARF is essential for cell survival in some cases of mel-anoma and thus represents an example of nononcogeneaddiction, as has been described for a number of genesrequired for cancer cells to tolerate cellular stresses (47).Despite this unresolved issue, the collected data from ourgenetic and functional studies suggest that RARb andp14ARF operate in the same molecular pathway, and thatthe melanoma-suppressive function of p14ARF may berelated to its role in this pathway.Our findings may have clinical implications. Topical

treatment with retinoic acid has been found to be effectivein dysplastic nevi (48) and resulted in tumor regression in2 patients with cutaneous metastatic melanoma (49).However, the clinical activity of retinoids for systemictreatment of melanoma is limited with only few patientsshowing a positive response (50–52). Currently, no pre-dictive markers of ATRA responsiveness are available forclinical use. Our data showing that melanoma cell lines withwild-type copies of RARB and ARF respond to treatmentwith ATRA suggest that the status of these genes identifies agroup of patients who may benefit from therapies includingretinoids.In conclusion, our data suggest that RARb and p14ARF

may be linked in a signaling pathway that is operating inmelanocytic cells and is lost in the majority of melanomas.This is an example of how integrated genomic analysis of

Dahl et al.





cancer may be used to map individual drivers to specificmolecular pathways and to reveal previously unrecognizedmolecular cross-talk. The set of melanoma drivers examinedhere is not exhaustive. Indeed, recent sequencing studieshave identified several additional putative melanoma drivers,albeit at low frequency (<10%; refs. 7–12). Moreover, ourunderstanding of the importance of the multiple methyla-tion changes present in melanoma genomes is still at an earlystage. The collection of melanoma cell lines characterizedhere may be useful for further unraveling the hierarchy ofdriver genes and molecular pathways in melanoma. Asexemplified by the identification of a subset of melanomasthat respond to retinoic acid, these cell lines may also be ofvalue for studying the genomic basis of therapeutic sensi-tivity and resistance in melanoma and for identifying newtargets for intervention. With the realization that intrinsicand acquired resistance is an inherent problem of currentmelanoma treatment options (2), the need for developingnew rational treatment regimens that can effectively controlthis malignancy is urgent.

Disclosure of Potential Conflicts of InterestNo potential conflicts of interest were disclosed.

Authors' ContributionsConception and design: C. Dahl, P. GuldbergDevelopment of methodology: C. Dahl, C. Christensen, P. GuldbergAcquisition of data (provided animals, acquired and managed patients, providedfacilities, etc.): C. Dahl, G. J€onsson, A. Lorentzen, M.L. Skjødt, G. Pawelec, P.GuldbergAnalysis and interpretation of data (e.g., statistical analysis, biostatistics, compu-tational analysis): C. Dahl, C. Christensen, M.L. Skjødt, P. GuldbergWriting, review, and/or revision of the manuscript: C. Dahl, C. Christensen, G.J€onsson, M.L. Skjødt, A. Lorentzen, Å. Borg, G. Pawelec, P. GuldbergAdministrative, technical, or material support (i.e., reporting or organizing data,constructing databases): C. Dahl, G. PawelecStudy supervision: P. Guldberg

AcknowledgmentsThe authors thank Vibeke Ahrenkiel for technical assistance, and Ulrik Ralfkiær

and Birgitte Grum-Schwensen for help with cell-cycle analysis.

Grant SupportThis study was supported by grants from the Danish Cancer Society, the Neye

Foundation and the Danish Cancer Research Foundation. The cell lines used here arefrom the EU Research Infrastructure Support program (contract QLRI-CT-2001-01325).

The costs of publication of this article were defrayed in part by the payment of pagecharges. This article must therefore be herebymarked advertisement in accordance with18 U.S.C. Section 1734 solely to indicate this fact.

Received January 7, 2013; revised June 25, 2013; accepted June 30, 2013;published OnlineFirst July 12, 2013.

References1. Tsao H, Atkins MB, Sober AJ. Management of cutaneous melanoma.

N Engl J Med 2004;351:998–1012.2. Flaherty KT, Hodi FS, Fisher DE. From genes to drugs: targeted

strategies for melanoma. Nat Rev Cancer 2012;12:349–61.3. Davies H, Bignell GR, Cox C, Stephens P, Edkins S, Clegg S, et al.

Mutationsof theBRAFgene inhumancancer.Nature2002;417:949–54.4. Hingorani SR, Jacobetz MA, Robertson GP, Herlyn M, Tuveson DA.

Suppression of BRAF(V599E) in human melanoma abrogates trans-formation. Cancer Res 2003;63:5198–202.

5. Christensen C, Guldberg P. Growth factors rescue cutaneous mela-noma cells from apoptosis induced by knock-down of mutated(V600E) B-RAF. Oncogene 2005;24:6292–302.

6. Dahl C, Guldberg P. The genome and epigenome of malignant mel-anoma. APMIS 2007;115:1160–75.

7. Wei X, Walia V, Lin JC, Teer JK, Prickett TD, Gartner J, et al. Exomesequencing identifiesGRIN2Aas frequentlymutated inmelanoma.NatGenet 2011;43:442–6.

8. Stark MS, Woods SL, Gartside MG, Bonazzi VF, Dutton-Regester K,Aoude LG, et al. Frequent somaticmutations inMAP3K5 andMAP3K9in metastatic melanoma identified by exome sequencing. Nat Genet2012;44:165–9.

9. Nikolaev SI, Rimoldi D, Iseli C, Valsesia A, Robyr D, Gehrig C, et al.Exome sequencing identifies recurrent somatic MAP2K1 andMAP2K2 mutations in melanoma. Nat Genet 2012;44:133–9.

10. Berger MF, Hodis E, Heffernan TP, Deribe YL, Lawrence MS, Proto-popovA, et al.Melanomagenomesequencing reveals frequentPREX2mutations. Nature 2012;485:502–6.

11. Krauthammer M, Kong Y, Ha BH, Evans P, Bacchiocchi A, McCuskerJP, et al. Exome sequencing identifies recurrent somatic RAC1 muta-tions in melanoma. Nat Genet 2012;44:1006–14.

12. Hodis E, Watson IR, Kryukov GV, Arold ST, Imielinski M, Theurillat JP,et al. A landscape of driver mutations in melanoma. Cell 2012;150:251–63.

13. Curtin JA, Fridlyand J, Kageshita T, Patel HN, Busam KJ, Kutzner H,et al. Distinct sets of genetic alterations in melanoma. N Engl J Med2005;353:2135–47.

14. Stark M, Hayward N. Genome-wide loss of heterozygosity and copynumber analysis in melanoma using high-density single-nucleotidepolymorphism arrays. Cancer Res 2007;67:2632–42.

15. J€onsson G, Dahl C, Staaf J, Sandberg T, Bendahl PO, Ringner M, et al.Genomic profiling of malignant melanoma using tiling-resolution arrayCGH. Oncogene 2007;26:4738–48.

16. Koga Y, Pelizzola M, Cheng E, Krauthammer M, Sznol M, Ariyan S,et al. Genome-wide screen of promoter methylation identifies novelmarkers in melanoma. Genome Res 2009;19:1462–70.

17. Vogelstein B, Kinzler KW. Cancer genes and the pathways theycontrol. Nat Med 2004;10:789–99.

18. TsaoH, Zhang X, Fowlkes K, Haluska FG. Relative reciprocity of NRASand PTEN/MMAC1 alterations in cutaneous melanoma cell lines.Cancer Res 2000;60:1800–4.

19. Bartkova J, Lukas J, Guldberg P, Alsner J, Kirkin AF, Zeuthen J, et al.The p16-cyclin D/Cdk4-pRB pathway as a functional unit frequentlyaltered in melanoma pathogenesis. Cancer Res 1996;56:5475–83.

20. Robinson J, Roberts CH, Dodi IA, Madrigal JA, Pawelec G, Wedel L,et al. The European Searchable Tumour Line Database. Cancer Immu-nol Immunother 2009;58:1501–6.

21. Borg A, Sandberg T, Nilsson K, Johannsson O, Klinker M, Masback A,et al. High frequency of multiple melanomas and breast and pancreascarcinomas in CDKN2A mutation-positive melanoma families. J NatlCancer Inst 2000;92:1260–6.

22. Lade-Keller J, Romer KM, Guldberg P, Riber-Hansen R, Hansen LL,Steiniche T, et al. Evaluation of BRAF mutation testing methodologiesin formalin-fixed, paraffin-embedded cutaneous melanomas. J MolDiagn 2013;15:70–80.

23. Grønbæk K, de Nully Brown P, Møller MB, Nedergaard T, Ralfkiaer E,Møller P, et al. Concurrent disruption of p16INK4a and the ARF-p53pathway predicts poor prognosis in aggressive non-Hodgkin's lym-phoma. Leukemia 2000;14:1727–35.

24. Guldberg P, thor Straten P, Birck A, Ahrenkiel V, Kirkin AF, Zeuthen J.Disruption of the MMAC1/PTEN gene by deletion or mutation is afrequent event in malignant melanoma. Cancer Res 1997;57:3660–3.

25. Clark SJ, Harrison J, Paul CL, Frommer M. High sensitivity mapping ofmethylated cytosines. Nucleic Acids Res 1994;22:2990–7.

26. Worm J, Aggerholm A, Guldberg P. In-tube DNA methylation profilingby fluorescence melting curve analysis. Clin Chem 2001;47:1183–9.

27. Van Raamsdonk CD, Bezrookove V, Green G, Bauer J, Gaugler L,O'Brien JM, et al. Frequent somatic mutations of GNAQ in uvealmelanoma and blue naevi. Nature 2009;457:599–602.






28. Fligiel SE, Inman DR, Talwar HS, Fisher GJ, Voorhees JJ, Varani J.Modulation of growth in normal andmalignant melanocytic cells by all-trans retinoic acid. J Cutan Pathol 1992;19:27–33.

29. Fan J, Eastham L, Varney ME, Hall A, Adkins NL, Chetel L, et al.Silencing and re-expression of retinoic acid receptor beta2 in humanmelanoma. Pigment Cell Melanoma Res 2010;23:419–29.

30. Faria TN, Mendelsohn C, Chambon P, Gudas LJ. The targeted disrup-tion of both alleles of RARbeta(2) in F9 cells results in the loss of retinoicacid-associated growth arrest. J Biol Chem 1999;274:26783–8.

31. de Th�e H, Vivanco-Ruiz MM, Tiollais P, Stunnenberg H, Dejean A.Identification of a retinoic acid responsive element in the retinoic acidreceptor beta gene. Nature 1990;343:177–80.

32. Li Y, Hashimoto Y, Agadir A, Kagechika H, Zhang X. Identification ofa novel class of retinoic acid receptor beta-selective retinoidantagonists and their inhibitory effects on AP-1 activity and retinoicacid-induced apoptosis in human breast cancer cells. J Biol Chem1999;274:15360–6.

33. Michaloglou C, Vredeveld LC, Soengas MS, Denoyelle C, Kuilman T,van der Horst CM, et al. BRAFE600-associated senescence-like cellcycle arrest of human naevi. Nature 2005;436:720–4.

34. Kim WY, Sharpless NE. The regulation of INK4/ARF in cancer andaging. Cell 2006;127:265–75.

35. Niles RM. The use of retinoids in the prevention and treatment of skincancer. Expert Opin Pharmacother 2002;3:299–303.

36. Jones S, Chen WD, Parmigiani G, Diehl F, Beerenwinkel N, Antal T,et al. Comparative lesion sequencing provides insights into tumorevolution. Proc Natl Acad Sci U S A 2008;105:4283–8.

37. Daniotti M, Oggionni M, Ranzani T, Vallacchi V, Campi V, Di SD, et al.BRAF alterations are associated with complex mutational profiles inmalignant melanoma. Oncogene 2004;23:5968–77.

38. Yang G, Rajadurai A, Tsao H. Recurrent patterns of dual RB and p53pathway inactivation in melanoma. J Invest Dermatol 2005;125:1242–51.

39. Vredeveld LC, Possik PA, Smit MA, Meissl K, Michaloglou C, HorlingsHM, et al. Abrogation of BRAFV600E-induced senescence by PI3Kpathway activation contributes to melanomagenesis. Genes Dev2012;26:1055–69.

40. Ha L, Ichikawa T, AnverM, Dickins R, Lowe S, Sharpless NE, et al. ARFfunctions as amelanoma tumor suppressor by inducing p53-indepen-dent senescence. Proc Natl Acad Sci U S A 2007;104:10968–73.

41. Haferkamp S, Scurr LL, Becker TM, Frausto M, Kefford RF, Rizos H.Oncogene-induced senescence does not require the p16(INK4a) orp14ARF melanoma tumor suppressors. J Invest Dermatol 2009;129:1983–91.

42. Duell EA,KangS, Voorhees JJ. Retinoic acid isomersapplied to humanskin in vivo each induce a 4-hydroxylase that inactivates only transretinoic acid. J Invest Dermatol 1996;106:316–20.

43. Tang XH, Gudas LJ. Retinoids, retinoic acid receptors, and cancer.Annu Rev Pathol 2011;6:345–64.

44. Inoue Y, Hasegawa S, Yamada T, Date Y, Mizutani H, Nakata S, et al.Bimodal effect of retinoic acid on melanocyte differentiation identifiedby time-dependent analysis. Pigment Cell Melanoma Res 2012;25:299–311.

45. SviderskayaEV,Gray-Schopfer VC,Hill SP, Smit NP, Evans-Whipp TJ,Bond J, et al. p16/cyclin-dependent kinase inhibitor 2A deficiency inhumanmelanocyte senescence, apoptosis, and immortalization: pos-sible implications for melanoma progression. J Natl Cancer Inst2003;95:723–32.

46. DhomenN,Reis-Filho JS, daRochaDS,HaywardR,SavageK,DelmasV, et al. Oncogenic Braf induces melanocyte senescence and mela-noma in mice. Cancer Cell 2009;15:294–303.

47. Solimini NL, Luo J, ElledgeSJ.Non-oncogene addiction and the stressphenotype of cancer cells. Cell 2007;130:986–8.

48. Halpern AC, Schuchter LM, Elder DE, Guerry D, Elenitsas R, Trock B,et al. Effects of topical tretinoin on dysplastic nevi. J Clin Oncol1994;12:1028–35.

49. Levine N, Meyskens FL. Topical vitamin-A-acid therapy for cutaneousmetastatic melanoma. Lancet 1980;2:224–6.

50. GoodmanGE. Phase II trial of retinol in patients with advanced cancer.Cancer Treat Rep 1986;70:1023–4.

51. Meyskens FL Jr, Liu PY, Tuthill RJ, Sondak VK, Fletcher WS, JewellWR, et al. Randomized trial of vitaminA versus observation as adjuvanttherapy in high-risk primary malignant melanoma: a Southwest Oncol-ogy Group study. J Clin Oncol 1994;12:2060–5.

52. Richtig E, Soyer HP, Posch M, Mossbacher U, Bauer P, Teban L, et al.Prospective, randomized, multicenter, double-blind placebo-con-trolled trial comparing adjuvant interferon alfa and isotretinoin withinterferon alfa alone in stage IIA and IIB melanoma: European Coop-erative Adjuvant Melanoma Treatment Study Group. J Clin Oncol2005;23:8655–63.

Dahl et al.





2013;11:1166-1178. Published OnlineFirst July 12, 2013.Mol Cancer Res Christina Dahl, Claus Christensen, Göran Jönsson, et al.

Signalingβ and RARARFMelanoma Identifies a Link Between p14Mutual Exclusivity Analysis of Genetic and Epigenetic Drivers in

Updated version

10.1158/1541-7786.MCR-13-0006doi:

Access the most recent version of this article at:

Material

Supplementary

http://mcr.aacrjournals.org/content/suppl/2014/02/17/1541-7786.MCR-13-0006.DC1

Access the most recent supplemental material at:

Cited articles

http://mcr.aacrjournals.org/content/11/10/1166.full#ref-list-1

This article cites 52 articles, 15 of which you can access for free at:

E-mail alerts related to this article or journal.Sign up to receive free email-alerts

Subscriptions

Reprints and

[email protected]

To order reprints of this article or to subscribe to the journal, contact the AACR Publications Department at

Permissions

Rightslink site. Click on "Request Permissions" which will take you to the Copyright Clearance Center's (CCC)

.http://mcr.aacrjournals.org/content/11/10/1166To request permission to re-use all or part of this article, use this link



http://mcr.aacrjournals.org/lookup/doi/10.1158/1541-7786.MCR-13-0006

http://mcr.aacrjournals.org/content/suppl/2014/02/17/1541-7786.MCR-13-0006.DC1

http://mcr.aacrjournals.org/content/11/10/1166.full#ref-list-1

http://mcr.aacrjournals.org/cgi/alerts

mailto:[email protected]

http://mcr.aacrjournals.org/content/11/10/1166


mutual exclusivity analysis of genetic and epigenetic ... · methylation (reviewed in ref. 6)....

Documents