Multiregion Whole-Exome Sequencing Uncovers the Genetic

Published OnlineFirst May 23, 2016; DOI: 10.1158/0008-5472.CAN-15-3476 Cancer Molecular and Cellular Pathobiology Research

Multiregion Whole-Exome Sequencing Uncovers the Genetic Evolution and Mutational Heterogeneity of Early-Stage Metastatic Melanoma Katja Harbst1, Martin Lauss1, Helena Cirenajwis1, Karolin Isaksson2, Frida Rosengren1, Therese Torngren€ 1, Anders Kvist1, Maria C. Johansson1, Johan Vallon-Christersson1, Bo Baldetorp1, Åke Borg1,Ha kan Olsson1,3, Christian Ingvar2,4, Ana Carneiro1,3, and Goran€ Jonsson€ 1

Abstract

Cancer genome sequencing has shed light on the underlying occur frequently. In addition, we found that the proportion of genetic aberrations that drive tumorigenesis. However, current ultraviolet B (UVB) radiation-induced C>T transitions dif- sequencing-based strategies, which focus on a single tumor biop- fered signiﬁcantly (P < 0.001) between early and late muta- sy, fail to take into account intratumoral heterogeneity. To address tion acquisition, suggesting that different mutational process- this challenge and elucidate the evolutionary history of melano- es operate during the evolution of metastatic melanoma. ma, we performed whole-exome and transcriptome sequencing of Finally, clinical history reports revealed that patients harbor- 41 multiple melanoma biopsies from eight individual tumors. ing a high degree of mutational heterogeneity were associated This approach revealed heterogeneous somatic mutations in the with more aggressive disease progression. In conclusion, our range of 3%–38% in individual tumors. Known mutations in multiregion tumor-sequencing approach highlights the genet- melanoma drivers BRAF and NRAS were always ubiquitous ic evolution and non-UVB mutational signatures associated events. Using RNA sequencing, we found that the majority of with melanoma development and progression, and may pro- mutations were not expressed or were expressed at very low levels, vide a more comprehensive perspective of patient outcome. and preferential expression of a particular mutated allele did not Cancer Res; 76(16); 1–10. 2016 AACR.

Introduction confined to distinct regions (2). Lung adenocarcinomas have also been analyzed using a similar multiregion approach (3, 4). Intratumor heterogeneity (ITH) has recently been thoroughly Although lung adenocarcinomas are also characterized by ITH, examined in cancer using whole-exome sequencing (WES) of 76% of all mutations were identified in all tumor regions. multiple regions of a sample (1–8). ITH could have major Furthermore, somatic mutations acquired during the progres- implications for clinical strategies concerning biopsy represen- sion of the tumors tended to be less smoking induced and tation of the tumor, clonality of alterations in targetable genes, instead had evidence of being acquired through upregulation of and drug resistance. In renal cell carcinoma, substantial ITH APOBEC cytidine deaminases (9). Other studies using multi- was demonstrated with up to 70% of mutations occurring in a region sequencing of tumors revealed different extent of het- heterogeneous pattern across spatially separated regions in erogeneity depending on the tumor type, for example, close to individual tumors. Several known cancer gene mutations were 100% heterogeneous genetic alterations in multifocal prostate cancer (6, 7), 55% in esophageal adenocarcinoma (8), and 49% 1Division of Oncology and Pathology, Department of Clinical Sciences, in ovarian cancer (5). Overall, these data point to the fact that Lund University, Lund, Sweden. 2Department of Surgery, Ska ne Uni- single-biopsy strategies may be inadequate to inform on muta- versity Hospital, Lund, Sweden. 3Department of Oncology, Skane tion status and may lead to significant molecular information University Hospital, Lund, Sweden. 4Division of Surgery, Department of Clinical Sciences, Lund University, Lund, Sweden. being missed. We recently found evidence of intrapatient tumor heteroge- Note: Supplementary data for this article are available at Cancer Research neity, using targeted gene sequencing and gene expression Online (http://cancerres.aacrjournals.org/). microarray analysis, in treatment-na€ve metastatic melanoma € K. Harbst, M. Lauss, A. Carneiro, and G. Jonsson contributed equally to this patients (10). Furthermore, individual metastases have been article. shown to be founded by multiple cell populations in the Corresponding Author: Goran€ Jonsson,€ Division of Oncology and Pathology, primary melanoma (11) highlighting the complexity of meta- Department of Clinical Sciences, Lund University, Lund, Sweden. Phone: 464- static progression. Although few treatments have been available 6222-0383; Fax: 464-614-7327; E-mail: [email protected] in metastatic melanoma, the recent development of inhibitors doi: 10.1158/0008-5472.CAN-15-3476 targeting BRAF and MEK changed the clinical outcome of 2016 American Association for Cancer Research. patients with metastatic disease (12). However, in most cases,

www.aacrjournals.org OF1

Harbst et al.

Table 1. Clinical features of patients included in the study Overall survival Number of Patient Disease state at surgery Treatment (after surgery) (months from surgery) Patient status samples Mm1641 Lymph node and lung metastases BRAF inhibitor 6 Dead of disease 7 Mm1702 Lymph node and lung metastases Dacarbazine 5 Dead of disease 3 Mm1765 Primary melanoma No treatment NA NA 4 Mm1767 Lymph node metastasis No treatment 12 Alive 4 Mm1772 Lymph node and skeletal metastases Ipilimumab, Pembrolizumab 11 Alive 4 Mm1837a Primary melanoma Nivolumab 6 Dead of disease 3 Mm1749b In-transit metastasis Ipilumumab 48 Dead of disease 9 Mm1742c Lymph-node Temozolomide, Pembrolizumab, Ipilimumab 26 Dead of disease 3 aThis patient died within one month of stage IV diagnosis. bNine in-transit metastases were obtained from this patient. cThis patient was treated with temozolomide during surgery.

the success is limited by the development of resistance in the Illumina Paired-End Sequencing Library Protocol (Agilent Tech- majority of patients within 6–9 months from treatment initi- nologies), using Clinical Research Exome (CRE) capture oligo ation. Several groups have studied whether genetic heteroge- panel. For Mm1641, libraries were prepared using the following neity can explain therapy resistance (4, 13, 14) and further three platforms: SureSelect All Exon V5þUTR (Agilent Techno- highlighting the complex heterogeneity of therapy resistance, it logies); TruSeqProtocol with TruSeq Exome Enrichment (Illu- has been shown that different mechanisms of resistance occur mina); and Nextera Rapid Capture Protocol with Expanded simultaneously in a single melanoma patient (15). Whether Exome (Illumina). All biopsies were sequenced with a physical these genetic lesions exist in small subclones prior to treatment coverage of at least 120 with the exception of Mm1641 (cov- initiation still remains unclear. erage 220). Libraries were sequenced on a HiSeq 2500 using To address these challenges, we applied multiregion DNA and normal or rapid run mode. RNA sequencing to melanoma tumors from eight patients. Our fi € ndings shed light on the extent of ITH in treatment-na ve WES data analysis melanoma, help to separate early from late genetic events, and Reads were mapped to the human reference genome (hg19) decipher mutational spectra occurring during the course of early with decoy using Novoalign (Novocraft Technologies). Postalign- metastatic disease. ment processing is described in the Supplementary Information. For variant calling, MuTect (16) and VarScan (17) were used, and Patients and Methods consensus calls between the two in the coding parts of the genes by Annovar annotation (18) were retained (see Supplementary Patient selection Information). Copy number log ratios of sequenced exons were Patients (n ¼ 8; Table 1) were selected for the study by a surgeon generated from bam files of tumor–normal pairs using CONTRA or oncologist based on the size of the tumor to allow for sampling 2.03, with default parameters (19). Exons with insufficient cov- of each quadrant of the tumor (see Supplementary Information). erage in the normal sample were removed. Copy number data The biopsies were taken at surgery and immediately stored at were segmented using GLAD (20). 80C (see Supplementary Fig. S1 for study workflow). We collected frozen tissue from multiple spatially separated tumor regions (Supplementary Table S1). Six patients had metastatic Phylogenetic and subclonality analysis disease at the time of surgery, and for two patients (Mm1765, Phylogenetic trees of single-nucleotide variation (SNV) data Mm1837) the primary melanoma was biopsied. In one case were generated using the maximum parsimony method. ABSO- fi (Mm1742), the inguinal metastasis was removed during temo- LUTE (21) was used for the identi cation of cancer cell fractions of zolomide treatment due to local progress and should be consid- SNVs with support from segmented copy number data and FACS ered separately in all subsequent analyses. All other cases were ploidy analysis (for details see Supplementary Information). untreated at the time of surgical removal of tumor. For patient Mm1702, only three different core-needle biopsies Gene expression analysis were analyzed. Blood samples were available from all patients Gene expression levels of multiregion samples were assayed except for Mm1837, for whom normal skin was used as matched using the Illumina HT12 microarray platform according to normal control. All patients consented and the study was the manufacturer's instructions at the Swegene Center for approved by the Regional Ethical Committee (Dnr. 101/2013). Integrative Biology at Lund University. Data analysis included integration of the multiregion samples with a reference set of Nucleic acid extraction 214 melanomas (22) and subtype classification (Supplemen- DNA and RNA were extracted from tumor and normal skin tary Information). using AllPrep DNA/RNA Mini Kit (Qiagen). DNA was extracted from blood using QIAamp DNA Blood Mini Kit (Qiagen). Allele-specific expression at mutation sites RNA-seqwasperformedusingtheIlluminaNextSeqforthe Whole-exome sequencing multiregion samples largely as described by Saal and collea- Tumor and matched normal DNA samples were subjected to gues (23) but using Illumina TruSeq Stranded mRNA Library library preparation. All sample libraries except Mm1641 were Prep Kit and NextSeq500. Mm1641:R1 and Mm1749:R6 were prepared according to SureSelect Target Enrichment System for omitted due to lack of input mRNA. For details on RNA-seq

OF2 Cancer Res; 76(16) August 15, 2016 Cancer Research

Genetic Evolution of Early-Stage Metastatic Melanoma

and allele-speciﬁc expression analysis see Supplementary Table 1). Although intriguing, larger series are required to Information. determine the usage of ITH as a biomarker in melanoma.

Analysis of regional subclones and mapping melanoma tumor Results evolution by phylogenetic tree analysis Analysis of intratumor heterogeneity in melanoma tumors Next, we mapped driver genes from Catalogue of Somatic WES was performed on 41 tumor samples obtained from Mutations in Cancer (COSMIC) Cancer Gene Census list (26) eight patients with an average target depth of 120 per sample to determine whether mutations in previously identified cancer (220 for Mm1641). Overall, three samples were excluded due driver genes were predominantly trunk or branch mutations to high normal tissue admixture (see Supplementary Informa- within the phylogenetic tree. We found 129 mutations in cancer tion). We called a mutation being present in a tumor if a driver genes with 113 mutations (88%) being trunk mutations nucleotide substitution was present in at least 10% of reads while only 16 (12%) were branch mutations (Fig. 2). Mutations in at least one tumor region. We then determined the regional in the melanoma driver genes, BRAF, NRAS,andNF1,were distribution of nonsynonymous mutations based on the WES always present on the trunk consistent with the critical role of data. We found on average 489 (range 45–814) nonsynon- constitutive activation of the MAPK pathway in the majority ymous mutations per tumor whereof 13% (range 2.7%–37.8%) of melanomas. Although the majority of driver mutations of the nonsynonymous somatic mutations were heterogeneous, detected in these melanomas appear to be present in all that is, not detected across all sampled regions in the individual spatially separated regions (Figs. 1, 2A, and B), we also tumors (Fig. 1). Notably, mutational ITH was not obvious at observed parallel evolution of subclones within the melanoma. the morphologic level, with uniform appearance of the tumor For example, Mm1641 harbored three distinct and spatially throughout its regions (Supplementary Fig. S2). In melanoma, separated activating mutations in PIK3CA (encoding p.R38H, three genes (BRAF, NRAS,andNF1), all key components of the p.V344G, and p.H1047R). Furthermore, a PIK3R1 (encoding MAPK pathway, are frequently mutated and overall demon- p.Q384X) truncating mutation was detected as part of the strate a mutually exclusive pattern (24). We found BRAF V600E dominant clone in a distinct region (R5) wild-type for the mutation in two patients, three cases harbored NRAS Q61 three different PIK3CA mutations. In addition, a mutation in mutation, one case had both NRAS Q61 and NF1 mutation, the phosphatase domain of PTEN (p.K125E) was present at and one case had both KIT and NF1 mutation. KIT is rarely high allele frequencies in all regions except R1. This mutation mutated in cutaneous melanoma; however, it is a known alters a highly conserved amino acid residue and was predicted oncogenic driver in acral and mucosal melanoma (25). These "probably damaging" by PolyPhen-2 with a score of 0.999. mutations were present across the tumor regions at similar Importantly, R1 harbored the H1047R PIK3CA kinase domain allele frequencies (Supplementary Table S1). The Mm1767 mutation frequently found to be mutually exclusive to PTEN tumor, which did not receive treatment regimens prior to mutations (27). The R7 region demonstrated the highest degree surgical removal, had the lowest degree of mutational hetero- of subclonality that presumably includes the clones from R5 geneity with 546 nonsynonymous somatic mutations, of which and R1 due to subclonal presence of mutations in PIK3CA only 2.7% were not detected in all spatially separated regions. (p.H1047R), PTEN,andPIK3R1 (Fig. 2C and Supplementary Three patients (Mm1837, Mm1641, Mm1702) died of their Figs. S4–S8). Overall, parallel evolution converging on distinct disease within six months of the analyzed biopsy (Table 1). genes in the same pathway indicates a strong selection for These had higher degree of mutational heterogeneity as com- aberrationsinthesegenesintheprogressionofmelanoma. pared with the other patients (P ¼ 0.03, t test). Seven spatially Although the presence of private mutations was observed in all separated regions from a lymph node metastasis from patient analyzed tumors, there was no evidence of subclonality of Mm1641 exhibited 14.2% mutational heterogeneity. At the cancer driver genes in the other patients. time of surgery, the patient had an aggressive disease and It is well established that most somatic mutations acquired in developed lung metastases. The patient subsequently received melanomas are C>T transitions preceded by a pyrimidine repre- BRAF inhibitor therapy; however,aftersixmonthsontherapy, senting UVB-induced mutations. We found that approximately the patient had a progressive disease course. From Mm1702, 80% of all trunk mutations harbored the UVB signature, while three core-needle biopsies in a lymph node metastasis were there was a span of 20%–70% of branch mutations having the analyzed and revealed 20.9% mutational heterogeneity. In UVB signature (Fig. 3; P < 0.001). On the other hand, T>G addition, FACS ploidy analysis revealed one hypo- and one transversions characteristic of UVA mutation signature (28) were hyper-tetraploid subclone in all tumor regions (Supplementary significantly increased among the branch mutations as compared Fig. S3; Supplementary Table S2). At the time of diagnosis the with trunk mutations (P < 0.001, Fisher exact test). These findings patient presented with distant metastases to the lungs and indicate that different mutational processes are operative during shortly thereafter the patient developed meningeal carcinoma- the progression of melanoma (Supplementary Fig. S9). Notably, tosis. The third patient (Mm1837) had 45 somatic mutations, the only case biopsied during treatment (Mm1742) did not show which exhibited 37.8% mutational heterogeneity. The patient a decrease of UVB- or increase of UVA-induced mutations (Fig. 3). was diagnosed with a large primary melanoma (Breslow thick- Finally, we constructed phylogenetic trees for these tumors ness 18 mm) of a nondefined pathologic subtype that was by maximum parsimony analysis (Fig. 4). We observed a MLANA negative, and developed an aggressive metastatic dis- branched evolutionary pattern in tumor Mm1641 where seven ease six months after initial diagnosis. Two patients are cur- distinct regions were subjected to WES. In contrast, Mm1749 rently alive, of which, one received both ipilumumab and showed limited degree of branching, even though these were pembrolizumab. These two patients had the lowest degree different in-transit metastases removed simultaneously. Inclu- of heterogeneous mutations (2.7% and 5.4%, respectively; sion of three additional asynchronous in-transit metastases

www.aacrjournals.org Cancer Res; 76(16) August 15, 2016 OF3

Harbst et al.

Mm1749 Mm1641 Mm1765 Mm1702 Mm1767 Mm1742 1 2 3 4 5 6 M1M2M3 1 2 3 4 5 6 7 1 2 3 4 1 2 3 1 2 3 4 1 2 3 TNFRSF17 MUC4 ILVBL GATA1 TP53 XPC NUTM2A MED12 WWC2 NEXN RPL5 ERC1 NCAN CIC KAT6A PRDM16 ROS1 PTEN GPHN EPPK1 TPR C1QTNF2 EWRS1 KIAA1549 NF1 KIAA0355 CDKN2A TBX3 CDK12 NFKMIE CDK4 CDRT15L2 NRAS TSPAN2 FRMPD3 HCLS1 ARID2 MAML2 NF1 TM7SF2 PPP6C KMT2D VPS52 BRAF NDUFA13 CREB1 PIP5K1C PTPRB SMARCA4 OR56A1 BRAF ITK ZBTB16 FGFR2 LRIG3 RET PTPLAD2 BRAF IKBKB NRAS ZFP92 PKDREJ ARID1A CIITA ADAT1 MED1 CBLC JAK3 IKZF1 CRTC3 CIITA MYH6 LCK NEB WT1 CNBD1 KIAA1549 BBX ASXL1 PTPRB TOPAZ1 PML AHNAK2 TPRX1 PTPRB SYK CRELD1 CRTC1 CNOT2 CRYBG3 ATP2B3 UTP14A CDH11 TBC1D8B DNASE1L1 GRIN2A TRANK1 MAFB CSF2RA PARP15 GRPR TMEM5 AIFM1 STAG2 RB1 ITGB6 CARD11 FBXW7 POU6F1 ROS1 NOP14 FAM111A GRIN2A PPP2R3C PVRL4 BCOR SND1 ANKRD30B LSS ELF4 CCNE1 PER3 PIK3CA HIST1H3B PER3 SLC25A5 UBR5 GRIA4 GAB3 NPHS1 G6PD WDFY4 BHLHB9 FGFR2 EP300 IVL PTPRK FRMD7 BRCA1 BEND3 PNMA5 TET1 GPRASP1 MED12 MITF CA7 USP26 HEATR1 TAF1 CCDC120 GNAS UBQLN2 FGFR4 JAK3 PTPRB FRMPD4 NF1 IGFBP2 FANCB HDGFRP2 SETBP1 PORCN LIFR RFPL4A DRP2 MYH11 ACSS3 NWD1 REPS2 UBAP2L GOPC RBM5 C2orf44 CXorf67 FGFR2 CFAP44 NRAS NOTCH2 PLS3 FAM115A WDR45 DSPP NEBL JAK3 TBK1 OVGP1 PATE1 SLC45A3 IMPA1 ELN LRRC45 OPN5 SLC22A24 LRRCC1 TNFSF12 CPAMD8 SATB1 KIT LRCOL1 HSPG2 COL3A1 RPL5 TIAM2 FIZ1 ALK MMS22L TMEM63C GLP1R ZNF385C Mm1772 TNNT3 FAM170A PLXNA4 PRDM16 CSMD1 KPNB1 VPS13D HDAC2 1 2 3 4 GLYATL2 TCF12 CUL1 PPFIBP1 IMPG1 CACNA1I FOXO3 ABCA2 ELMO3 OR10G9 ZFYVE27 PIK3CA PKD1 NYAP2 MSN CUX2 FBRSL1 MYH11 UTP6 CECR1 CCDC30 KIF19 ZIC5 LRRC7 TBCA ZNF730 CDK4 SLC35A1 FAM153B DPP9 HAUS3 PLD3 TDG ANKZF1 EFCC1 POU3F1 PLEKHN1 FAM178B RPAP3 CBL CNTNAP2 SFT2D3 DNAH6 BPTF PIK3CA ROCK2 RBMXL1 C20orf144 RBM27 CRIPAK IDH1 KIAA2018 GPD2 NWD2 TCHH NF1 SLC25A27 ARL6IP4 ANKRD62 DPYSL4 ADCK1 NOTCH2 ADAM32 NHSL2 NEUROG3 SLC38A11 RBM5 ADAMTS19 PCDHB10 ZNF831 PIK3R1 PTPRC MED24 DND1 SLC2A3 PODXL Wildtype LRRIQ3 CELSR1 LTBP3 SLC2A146 USH2A B3GALT6 SERF1A INPP4A GUCY2C POU3F3 HGS ELN Mutaon CELA1 MST1 GALNT8 NKX2-4 EXOC6B PLBD1 PRR23A PSKH2 SLITRK2 KCNS2 TGFBR2 Mm1837 TAS2R7 SEL1L2 ZNF529 DEDD PBRM1 APC2 GRXCR1 1 2 3 CHRNE DESI2 PCDHAC2 KMTD2 TNC GAD2 CCDC88C NOTCH2 RANBP17 ADM5 DTX3L AARS2 PHF2 ABCA13 NRAS ZNF23 CARD11 PCDHA6 AEBP1 ANK3 EDEM1 LTN1 ORT27 FCGBP MYH3 CELA1 MSGN1 IGSF3 LGALS12 TWIST1 MSGN1 SSTR1 BCR NR1D1 NCKAP5L KRTAP4-11 C9orf50 DSE IL37 PCDHB10 ZNF681 SEZ6 MTMR4 RNPC3 FRMD3 PRKRA CDK2 ALMS1 NPAS3 COL8A2 KCNC1 SCN8A ZIC2 RBM15 NUMBL ANKRD34C FAM83H ADAD1 DCTN5 IFFO1 ANKLE1 HMMR CHTF18 PLXNA2 SCAF1 NLRP4 UBE2Q2 SLC38A3 DDHD1 SIPA1L1 IL7R ADAM21 FAM169A XKR4 TBC1D2 EPS8L2 NPY2R VCY COQ6 TLDC1 CNOT1 MAP3K9 C9orf43 ALDH2 AFAP1L1 NAGK PCDHA13 MPHOSPH8 HMCN1 SLC35F4 MED15 USP25 NDC1 SEC16A LRIG3 JAK2 DNAH6 SALL3 SESTD1 ACE2 PPP2R2B ADRM1 LAMA1 TBP GDF1 NOC3L PDGFRA AKAP9 ZBBX MYH14 ANKRD31 DCC VPS13C NUMA1 JAKMIP2 VTN ILF3 Case Nonsynonomous Heterogeneous LPA HEXIM1 EXOC3L4 CGREF1 mutaons mutaons (%) DTX3L NEURL1B HERC1 TPP2 TOP1 FGF23 Mm1749 814 7.2 TRIM43 SLC2A3 LRBA FGFRL1 Mm1641 380 14.2 COL7A1 CRIPAK APOBR ZNF358 MX2 Mm1772 367 5.4 TIAM1 KRTAP11-1 KRT8 Mm1765 742 5.9 PLA2G4D SLC26A10 Mm1702 541 20.9 GIT2 ATP13A3 KDM2B Mm1767 546 2.7 FAM71D Mm1742 473 6.3 Mm1837 45 37.8

Figure 1. Regional distribution of mutations found in multitumor regions of eight melanoma patients. Presence (yellow) or absence (blue) of each mutation is marked for each tumor region. Insets provide gene symbols for the heterogeneous mutations in each case. Cancer Gene Census genes are marked in red. The table shows numbers of nonsynonymous mutations and percentage of heterogeneous mutations in each case. Notably, for Mm1742, the melanoma metastasis was removed during temozolomide treatment.

supported the limited branched evolution in Mm1749. In the Transcriptome analysis of intratumor heterogeneity phylogenetic tree it is clearly observed that M1-3 is closely We then performed microarray-based gene expression proﬁling intermingled with R1-6. In conclusion, the branched evolu- of all samples to determine the transcriptional heterogeneity tionary pattern indicates that a linear progression model does within individual tumors. Importantly, we pooled the data from not sufﬁciently explain melanoma evolution. the multiregion samples with a reference set of 214 metastatic

OF4 Cancer Res; 76(16) August 15, 2016 Cancer Research

Genetic Evolution of Early-Stage Metastatic Melanoma

Number of cancer driver Fracon of trunk vs. non-trunk ABCgene mutaons cancer driver gene mutaons Subclonality analysis of Mm1641

Cancer gene mutaons ● PIK3CA (V344G) ● Nontrunk mutaons ●● PTEN ● PTEN PIK3CA (V344G) BRAF Other somac mutaons BRAF 120 Trunk mutaons 100

100 80 R4 Subclonal fracon R4 Subclonal fracon PIK3CA (H1047R) PIK3R1 PIK3CA (H1047R) ● PIK3CA (R38H) ● ● PIK3R1 ● PIK3CA (R38H) 0.0 0.2 0.4 0.6 0.8 1.0 80 0.0 0.2 0.4 0.6 0.8 1.0 60 0.0 0.2 0.4 0.6 0.8 1.0 0.0 0.2 0.4 0.6 0.8 1.0 R1 Subclonal fracon R2 Subclonal fracon 60 PTEN ● 1.0 BRAF ● ● PIK3R1 Fracon (%) Fracon 40 BRAF 40

PIK3CA (H1047R) Number of cancer driver mutaons driver Number of cancer 20 ● 20 ● PTEN

● R5 Subclonal fracon PIK3R1 R7 Subclonal fracon PIK3CA (H1047R) PIK3CA (V344G) PIK3CA (V344G) ● PIK3CA (R38H) ● ● ● PIK3CA (R38H) 0.0 0.2 0.4 0.6 0.8 0.0 0.2 0.4 0.6 0.8 1.0 0 0 All Mm1749Mm1641Mm1772Mm1765Mm1702Mm1767Mm1837Mm1742 T NT T NT T NT T NT T NT T NT T NT T NT T NT 0.0 0.2 0.4 0.6 0.8 1.0 All Mm1749Mm1641Mm1772Mm1765Mm1702Mm1767Mm1837Mm1742 0.0 0.2 0.4 0.6 0.8 1.0 R1 Subclonal fracon R2 Subclonal fracon

Figure 2. Analysis of cancer gene mutations and subclonality in melanoma tumors. A, bar plot displaying the number of mutations in cancer census genes (n ¼ 129). Mutations are divided in trunk (T, shared mutations present in all regions) and nontrunk (NT, mutations not present in all regions). B, fractions of cancer gene mutations among trunk (T) and nontrunk (NT) mutations demonstrating that only a fraction of all T and NT mutations are alterations in cancer driver genes. C, mutation plot demonstrating the subclonality analysis in different regions of tumor Mm1641 that displayed the most distinct subclonal pattern. The subclonal fraction is obtained from the ABSOLUTE analysis. Subclonal fractions of all mutations in two regions are plotted on the x-andy-axis. Melanoma driver mutations such as BRAF mutations and mutations in the PI3K pathway are indicated in the plots. melanoma tumors previously profiled using the same platform, levels of CTLA4 and these two regions formed a distinct branch allowing to relate the gene expression levels of the ITH regions in that seemed to have evolved later than R3 and R4, suggesting that relation to the full melanoma expression space (22). We deter- some clones may evolve to escape the immune response (Figs. 3 mined the intratumoral expression of a subtype signature shown and 5). Gene copy number values extracted from the WES data to classify melanomas into four molecular subgroups: high- were used to determine an additional layer of ITH and unveiled immune (associated with a good prognosis), microphthalmia- additional information using GISTIC regions from the melanoma associated transcription factor (MITF)-low proliferative (associ- Cancer Genome Atlas study (24). The amount of copy number ated with a poor prognosis), MITF-high pigmentation (associated changes per tumor varied between patients. For example, with a poor prognosis), and normal-like. In 4 of 8 patients, we Mm1765 harbored copy number changes exclusively on chromo- could observe gene expression–based ITH when taking into somes 1q, 6p24.3, and 9p23 with a heterogeneous pattern at account genes that have been used to develop the subtype clas- chromosome 9p23 (Fig. 5). In Mm1749, several copy number sification. Inconsistent with the phylogenetic tree analysis, we changes were identified with three in-transit metastases lacking a observed that two regions evolutionary similar by mutation CDKN2A homozygous deletion detected in all the other six analysis could be assigned to different gene expression subtypes metastases. Mm1837 harbored the highest degree of mutational (Fig. 5). For example, based on phylogenetic tree analysis in ITH; however, we found no gene copy number differences. Mm1641, R1, and R7 displayed genetic divergence. However, Overall, this suggests that ITH may independently be dictated by based on gene expression subtype analysis, R3 and R5 demon- mutations, DNA copy number changes, or gene expression. strated different gene expression classification. Such differences were also observed patients Mm1702, Mm1749, and Mm1742. RNA-seq reveals allele-specific expression Next, we investigated mRNA levels of 10 genes involved in Finally, we performed RNA sequencing to determine spatially melanocyte differentiation (MITF, MLANA, TYR), immune check- restricted mutation allele–specific expression. The majority of point (CTLA4, PD1), T-cell functions (CD3, CD247, CD8A), and mutated genes had low or absent expression, as only about cell proliferation (CCNE1, AURKA; Fig. 5). MITF and its target 40% of mutated genes were expressed and <20% had >20 reads genes (MLANA and TYR) displayed intratumor gene expression (Fig. 6A). To determine allele-specific expression, we used the >20 differences and showed agreement with the subtype classification. read cutoff to obtain robust frequencies of the mutated allele, and Furthermore, mRNA levels of immune checkpoint genes (CTLA4 correlated the variant allele frequency (VAF) of the RNA to the VAF and PD1) showed some consistency with the phylogenetic tree of the DNA. In line with Castle and colleagues (29), we observed a analysis. For example, R1 and R2 from Mm1765 had decreased good correlation between RNA and DNA VAF for the majority of

www.aacrjournals.org Cancer Res; 76(16) August 15, 2016 OF5

Harbst et al.

P < 0.001 Mm1641 Mm1702 Mm1749 Mm1765 Mm1767 Mm1772 Mm1837 Mm1742* 322 43 644 137 1113 73 1085 54 813 12 531 22 51 13 641 32 1.0 Y−C>T R−C>T T>C 0.8 C>A C>G T>A T>G 0.6

0.4 Proporon

0.2

0.0 Trunk Trunk Trunk Trunk Trunk Trunk Trunk Trunk Trunk Branches Branches Branches Branches Branches Branches Branches Branches Branches

Figure 3. Mutational signatures in melanoma revealed by multiregion sequencing. Fraction of each kind of base substitution is shown for trunk (shared) and branch (heterogeneous) SNVs for each patient. The two left columns represent mutations combined from all patients. A Fisher exact test was used to determine differences in mutational spectrum between trunk and branch mutations. Color codes reﬂect the different base substitutions, with C>T mutations divided into two groups: Y-C>T denotes C-to-T mutations at sites preceded by a pyrimidine (Y), and R-C>T denotes those preceded by a purine (R). , the patient treated with temozolomide at the time of surgery.

mutations, that is, the mutated allele was expressed according to 12% of cancer driver mutations to be affected by ITH, indicating its frequency in the DNA (Fig. 6B). As an exception from this rule that heterogeneous mutations were predominantly passenger and in accordance with the concept of nonsense-mediated mRNA mutations. Moreover, ITH seemed to be independent of the decay, we found that alleles harboring stop mutations had sig- number of biopsies analyzed, as the tumor with only three regions nificantly lower expression as compared with alleles harboring analyzed showed the highest degree of ITH, suggesting that the synonymous or missense mutations (Fig. 6C). As expected, a degree of ITH could vary in melanoma. significant fraction of genes with mutations at chromosomal loci In line with other melanoma sequencing studies (9, 30), we undergoing loss of heterozygosity (LOH) demonstrated prefer- detected UVB-induced C>T transitions. Our analysis confirms that ential expression of the mutated allele. Subsequently, we used a this signature is enriched early in melanoma tumor evolution binomial test to identify genes with preferential expression of the occurring in all tumor regions. We found a significant reduction of mutated allele and identified a limited number of genes. Only the UVB-induced mutational signature in branch (nontrunk) TNC had a branch mutation with preferential expression of the mutations, suggesting that mutations arising later in the progres- mutated allele in three different in-transit metastases from patient sion of melanoma may occur as a result of other mutational Mm1749. The vast majority of the genes harboring nontrunk processes. This is in line with recent studies in lung cancer, where mutations were expressed at similar mRNA levels in tumor later acquired mutations lack the tobacco-smoking signature regions with and without the mutation (Supplementary Fig. (3, 4). Furthermore, with the exception of one patient in current S10). We found several genes with trunk mutations, however, study none received any systemic treatment prior to surgical with a preferential expression of mutated allele in only one of the removal of tumor; therefore, the switch in mutational processes spatially separated regions. For example, Mm1765 harbored a KIT is likely due to evolutionary changes occurring during melanoma D816V mutation found at equal DNA allele frequencies in all progression and not due to extrinsic factors. This further high- spatially separated regions, with preferential overexpression of the lights the benefit of multiregion sequencing as different muta- mutated allele only in R3 (Fig. 6D and Supplementary Figs. S11– tional processes operate during tumor evolution. S15). Collectively, our data suggest that preferential expression of We confirmed previous findings consistent with activating the mutated allele occurs; however, it does not seem to be a somatic mutations in the MAPK pathway being driver events in general mechanism of melanomagenesis. the majority of melanomas (24). We showed that constitutive activation of BRAF, KIT, and NRAS through somatic mutations was ubiquitous in seven patients. The last patient was wild-type Discussion for BRAF, NRAS, NF1, and KIT; however, the tumor harbored a We have identified ITH of somatic mutations, DNA copy nonkinase domain PDGFRA mutation. In contrast, we found number, and transcriptomic changes in multiple regions from evidence of genetic mutations in the PI3K pathway occurring as eight melanoma tumors and demonstrated that 3%–38% of nontrunk events. In particular, we found parallel evolution of somatic mutations were not identified in all regions of the tumors. three distinct activating PIK3CA mutations (p.V344G, p.R38H, In contrast to renal cell cancer, where approximately 75% of and p.H1047R) and a stop mutation in PIK3R1 in a large lymph cancer driver mutations are affected by ITH (2), we found only node metastasis (Mm1641). Mutational convergence affecting

OF6 Cancer Res; 76(16) August 15, 2016 Cancer Research

Genetic Evolution of Early-Stage Metastatic Melanoma

Mm1749 Mm1742 Mm1641 Mm1767 1,149 641 476 813

BRAF NRAS NRAS FBXW7 TP53 NF1 ARID2 BRAF MITF CDKN2A PIK3CA (R38H) PPP6C BRCA1 PTEN HMZ R5 R2 R4 PIK3R1 R3 R6 R2 R6 R1 PIK3CA (H1047R) IDH1 R7 R1 Figure 4. R3 R5 R2 Phylogenetic trees of mutation R5 R4 evolution from different regions of M1 R3 eight melanoma patients generated R4 R2 R1 M2 by maximum parsimony analysis from M3 Mm1837 Mm1772 exome data. The number of somatic 42 531 R1 mutations common to all tumor regions is indicated at the top of each tree. The different tumor regions (R) Mm1765 are indicated at the ends of the branches. In case Mm1749, M1-3 1,086 represent three asynchronous in- NRAS transit metastases. Length of branch KMTD2 Mm1702 PDGFRA reﬂects number of mutations. Blue, the CDKN2A hmz stem of the tree with all shared 643 mutations; yellow, branches shared by NF1 at least two regions; red branch, KIT individual regions. Also indicated in CDK4 NRAS the phylogenetic tree are mutations CDK4 in melanoma cancer driver genes. RB1 TOP1

R4 R3 R3 R4 R5 R2 R2 R1 R1 JAK2 R3 NOTCH2 R4 R4 R2

R1 R2 different PI3K members in single melanoma tumors implicate Although it is difﬁcult to discern the full extent of ITH at all that activation of this pathway is involved in later stages of levels using multiregion sequencing only, we observed that metastatic melanoma progression. In addition, activation of the patients having melanoma tumors with high degree of ITH as PI3K pathway via somatic mutations may be a reason for acquired revealed by our approach experienced a more aggressive disease resistance to BRAF inhibition as previously described (31). course. These ﬁndings are further supported by data from similar Indeed, patient Mm1641 received BRAF inhibition after surgery studies in lung cancer, esophageal adenocarcinoma, and pediatric but subsequently developed resistance to the therapy. tumors (4, 8, 32). However, given that this study included a

www.aacrjournals.org Cancer Res; 76(16) August 15, 2016 OF7

Harbst et al.

A Mm1641 Mm1765 Mm1702 Mm1749 Mm1837 Mm1772 Mm1742 Mm1767 R1 R2 R3 R4 R1 R2 R4 R5 M1 M2 M3 R3 R6 R4 R1 R2 R3 R3 R2 R1 R1 R2 R4 R5 R1 R4 R6 R7 R2 R3 R5 R4 R5 R2 R1 R2 R3

B Subtype

C MITF TYR MLANA PD1 CTLA4 CD3D CD247 CD8A CCNE1 Gene expression AURKA D 1p36.31 1p22.1 1p12 1q21.3 1q44 2q37.3 3p13 3q13.31 4q12 4q34.3 5p15.31 5q12.1 5q31.3 6p24.3 6q12 6q22.31 6q26 7p22.1 7q34 8q11.21 8q24.21 9p23 9p21.3 10p15.3

DNA Copy number 10q23.31 10q26.3 11q13.3 11q23.3 12q14.1 12q15 12q23.3 14q23.3 15q13.3 15q21.1 15q26.2 16p13.3 16q12.1 17q25.3 19p13.3 19p13.2 22q13.2

Figure 5. Gene expression and DNA copy number aberrations analysis for the eight melanoma patients. For each tumor site/patient (A), the following information is shown (panels from top): B, molecular subtype [high-immune (orange), pigmentation (green), proliferative (red), normal-like (blue), and unclassiﬁed (gray; ref. 22). C, expression of pigmentation, immune response, and proliferation genes (green, downregulation; red, upregulation). D, copy number levels of GISTIC regions deﬁned by the The Cancer Genome Atlas melanoma study (24). For copy number aberrations, blue denotes loss of genetic material and red denotes gain thereof.

limited number of patients all treated differently, additional larger that there is considerable variation in prognostic gene expression studies are required to further conﬁrm these data given the signatures within individual melanoma tumors in line with signiﬁcance of prognostic markers in melanoma. results from studies in renal cell cancer (2). However, the analyzed Addressing an additional layer of genomic heterogeneity we gene signatures are partly dependent on the tumor microenvi- investigated the expression of mutated alleles using RNA sequenc- ronment, highlighting another layer of genomic complexity with- ing. In line with the results of Castle and colleagues (29), we found in melanoma tumors. a good correlation (corr. ¼ 0.57) between DNA and RNA VAFs. Taken together, we have used a multiregion sequencing Our molecular exploration of ITH in melanoma further showed approach in early-stage treatment-na€ve metastatic melanomas

OF8 Cancer Res; 76(16) August 15, 2016 Cancer Research

Genetic Evolution of Early-Stage Metastatic Melanoma

A 100 100 80 80 60 60 40 40 20 20 % Mutaons with reads>1 % Mutaons 0 with reads>20 % Mutaons 0 T T NT NT Mm1641 T Mm1702 T Mm1742 T Mm1749 T Mm1765 T Mm1767 T Mm1772 T Mm1837 T Mm1641 T Mm1702 T Mm1742 T Mm1749 T Mm1765 T Mm1767 T Mm1772 T Mm1837 T Mm1641 NT Mm1702 NT Mm1742 NT Mm1749 NT Mm1765 NT Mm1767 NT Mm1772 NT Mm1837 NT Mm1641 NT Mm1702 NT Mm1742 NT Mm1749 NT Mm1765 NT Mm1767 NT Mm1772 NT Mm1837 NT B C Mutaons with >20 reads, DNA and RNA

P = 1e−48 VAF of RNA VAF −0.5 0 0.5 cor = 0.57 VAF Expression: observed − expected VAF 0 0.2 0.4 0.6 0.8 1.0 0 0.2 0.4 0.6 0.8 1.0 Nonsynonymous Synonymous Stopgain VAF of DNA D Mm1765 R1 Mm1765 R2 Mm1765 R3 Mm1765 R4

1.0 KIT 0.8

KIT KIT KIT 0.4 0.6 0.8 1.0 VAF of RNA VAF VAF of RNA VAF VAF of RNA VAF VAF of RNA

0.2 cor = 0.24 cor = 0.28 cor = 0.31 cor = 0.28 0 0.2 0.4 0.6 0.8 1.0 0 0.2 0.4 0.6 0 0.2 0.4 0.6 0.8 1.0 0 0 0.2 0.4 0.6 0.8 1.0 0 0.2 0.4 0.6 0.8 1.0 0 0.2 0.4 0.6 0.8 1.0 0 0.2 0.4 0.6 0.8 1.0 VAF of DNA VAF of DNA VAF of DNA VAF of DNA

Figure 6. RNA sequencing of melanoma samples from eight patients. A, fraction of all mutations present in >1 read (left) and >20 reads (right). B, correlation plot between DNA and RNA VAF across all samples. C, analysis of VAF expression versus expected expression for synonymous, nonsynonymous, and stop mutations in all samples. D, correlation plots in individual regions in Mm1765, which was the only sample displaying preferential expression of a cancer driver gene. Preferential expression of the KIT mutated allele observed only in R3 is indicated in the plots. Red, signiﬁcant values using a binomial test as described in Supplemental information. Cor., Spearman correlation value between DNA and RNA VAF.

andshowedthatITHwasdirectedbyeithermutationalorgene our results implicate that there are other mutational processes expression heterogeneity, suggesting a dual layer of genomic operating during the evolution of metastatic melanoma. Thus, complexity in the heterogeneity of melanoma tumors. This early melanoma progression may be dependent on UVB- study represents the first comprehensive genomic investigation induced mutagenesis, while later metastatic evolution is asso- of intratumor heterogeneity in metastatic melanoma and com- ciated with additional non-UVB mutational processes. Impor- plements the findings by Shain and colleagues on genetic tantly, our results further indicate that mutation based ITH may evolution of melanoma from precursor lesions (33). Although be a prognostic biomarker, as patients with high degree of the majority of mutations in cancer driver genes such as mutational heterogeneity presented with an aggressive disease BRAFV600E occurred ubiquitously throughout the tumor, course. Extension of research to epigenetic and clinical assess- mutations in the PI3K pathway demonstrated a heterogeneous ment in much larger cohorts, preferably with repeated biopsies pattern, suggesting that geneticeventsoccurringinthispathway at relapse after treatment, is needed to fully understand the represent later acquired mutations in metastatic melanoma clinical and biologic impact of ITH in melanoma. evolution. Furthermore, Shain and colleagues (33) found that UVB is the main factor in both the initiation of precursor Disclosure of Potential Conflicts of Interest lesions and their progression to primary melanoma; however, No potential conflicts of interest were disclosed.

www.aacrjournals.org Cancer Res; 76(16) August 15, 2016 OF9

Harbst et al.

Authors' Contributions Administrative, technical, or material support (i.e., reporting or organizing Conception and design: K. Harbst, A. Borg, C. Ingvar, A. Carneiro, data, constructing databases): A. Carneiro, G. Jonsson G. Jonsson Study supervision: K. Harbst, C. Ingvar, A. Carneiro, G. Jonsson Development of methodology: K. Harbst, M. Lauss, G. Jonsson Acquisition of data (provided animals, acquired and managed pati- Grant Support ents, provided facilities, etc.): K. Harbst, K. Isaksson, F. Rosengren, T. The Swedish Research Council, the Swedish Cancer Society, the Berta Torngren,€ M.C. Johansson, B. Baldetorp, A. Borg, C. Ingvar, A. Carneiro, Kamprad Foundation, The Gunnar Nilsson Cancer foundation, the Gustav Vth G. Jonsson Jubilee Foundation, and Governmental Funding of Clinical Research within Analysis and interpretation of data (e.g., statistical analysis, bio- National Health Service. statistics, computational analysis): K. Harbst, M. Lauss, H. Cirenajwis, The costs of publication of this article were defrayed in part by the pay- A. Kvist, J. Vallon-Christersson, B. Baldetorp, H. Olsson, A. Carneiro, ment of page charges. This article must therefore be hereby marked advertise- G. Jonsson ment in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. Writing, review, and/or revision of the manuscript: K. Harbst, M. Lauss, H. Cirenajwis, K. Isaksson, J. Vallon-Christersson, H. Olsson, C. Ingvar, A. Carneiro, Received January 3, 2016; revised March 10, 2016; accepted March 28, 2016; G. Jonsson published OnlineFirst May 23, 2016.

References 1. Gerlinger M, Rowan AJ, Horswell S, Larkin J, Endesfelder D, Gronroos E, 17. Koboldt DC, Zhang Q, Larson DE, Shen D, McLellan MD, Lin L, et al. et al. Intratumor heterogeneity and branched evolution revealed by multi- VarScan 2: somatic mutation and copy number alteration discovery in region sequencing. N Engl J Med 2012;366:883–92. cancer by exome sequencing. Genome Res 2012;22:568–76. 2. Gerlinger M, Horswell S, Larkin J, Rowan AJ, Salm MP, Varela I, et al. 18. Wang K, Li M, Hakonarson H. ANNOVAR: functional annotation of genetic Genomic architecture and evolution of clear cell renal cell carcinomas variants from high-throughput sequencing data. Nucleic Acids Res 2010; defined by multiregion sequencing. Nat Genet 2014;46:225–33. 38:e164. 3. de Bruin EC, McGranahan N, Mitter R, Salm M, Wedge DC, Yates L, et al. 19. Li J, Lupat R, Amarasinghe KC, Thompson ER, Doyle MA, Ryland GL, et al. Spatial and temporal diversity in genomic instability processes defines lung CONTRA: copy number analysis for targeted resequencing. Bioinformatics cancer evolution. Science 2014;346:251–6. 2012;28:1307–13. 4. Zhang J, Fujimoto J, Zhang J, Wedge DC, Song X, Zhang J, et al. Intratumor 20. Hupe P, Stransky N, Thiery JP, Radvanyi F, Barillot E. Analysis of array CGH heterogeneity in localized lung adenocarcinomas delineated by multi- data: from signal ratio to gain and loss of DNA regions. Bioinformatics region sequencing. Science 2014;346:256–9. 2004;20:3413–22. 5. Bashashati A, Ha G, Tone A, Ding J, Prentice LM, Roth A, et al. Distinct 21. Carter SL, Cibulskis K, Helman E, McKenna A, Shen H, Zack T, et al. evolutionary trajectories of primary high-grade serous ovarian cancers Absolute quantification of somatic DNA alterations in human cancer. Nat revealed through spatial mutational profiling. J Pathol 2013;231:21–34. Biotechnol 2012;30:413–21. 6. Boutros PC, Fraser M, Harding NJ, de Borja R, Trudel D, Lalonde E, et al. 22. Cirenajwis H, Ekedahl H, Lauss M, Harbst K, Carneiro A, Enoksson J, et al. Spatial genomic heterogeneity within localized, multifocal prostate cancer. Molecular stratification of metastatic melanoma using gene expression Nat Genet 2015;47:736–45. profiling: Prediction of survival outcome and benefit from molecular 7. Cooper CS, Eeles R, Wedge DC, Van Loo P, Gundem G, Alexandrov LB, et al. targeted therapy. Oncotarget 2015;6:12297–309. Analysis of the genetic phylogeny of multifocal prostate cancer identifies 23. Saal LH, Vallon-Christersson J, Hakkinen J, Hegardt C, Grabau D, Winter C, multiple independent clonal expansions in neoplastic and morphologi- et al. The Sweden Cancerome Analysis Network - Breast (SCAN-B) Initia- cally normal prostate tissue. Nat Genet 2015;47:367–72. tive: a large-scale multicenter infrastructure towards implementation of 8. Murugaesu N, Wilson GA, Birkbak NJ, Watkins TB, McGranahan N, Kumar breast cancer genomic analyses in the clinical routine. Genome Med S, et al. Tracking the genomic evolution of esophageal adenocarcinoma 2015;7:20. through neoadjuvant chemotherapy. Cancer Discov 2015;5:821–31. 24. The Cancer Genome Atlas Network. Genomic classification of cutaneous 9. Alexandrov LB, Nik-Zainal S, Wedge DC, Aparicio SA, Behjati S, Biankin AV, melanoma. Cell 2015;161:1681–96. et al. Signatures of mutational processes in human cancer. Nature 2013; 25. Curtin JA, Busam K, Pinkel D, Bastian BC. Somatic activation of KIT in 500:415–21. distinct subtypes of melanoma. J Clin Oncol 2006;24:4340–6. 10. Harbst K, Lauss M, Cirenajwis H, Winter C, Howlin J, Torngren T, et al. 26. Futreal PA, Coin L, Marshall M, Down T, Hubbard T, Wooster R, et al. A Molecular and genetic diversity in the metastatic process of melanoma. census of human cancer genes. Nat Rev Cancer 2004;4:177–83. J Pathol 2014;233:39–50. 27. Saal LH, Holm K, Maurer M, Memeo L, Su T, Wang X, et al. PIK3CA 11. Sanborn JZ, Chung J, Purdom E, Wang NJ, Kakavand H, Wilmott JS, et al. mutations correlate with hormone receptors, node metastasis, and ERBB2, Phylogenetic analyses of melanoma reveal complex patterns of metastatic and are mutually exclusive with PTEN loss in human breast carcinoma. dissemination. Proc Natl Acad Sci U S A 2015;112:10995–1000. Cancer Res 2005;65:2554–9. 12. Chapman PB, Hauschild A, Robert C, Haanen JB, Ascierto P, Larkin J, et al. 28. Drobetsky EA, Turcotte J, Chateauneuf A. A role for ultraviolet A in solar Improved survival with vemurafenib in melanoma with BRAF V600E mutagenesis. Proc Natl Acad Sci U S A 1995;92:2350–4. mutation. N Engl J Med 2011;364:2507–16. 29. Castle JC, Loewer M, Boegel S, Tadmor AD, Boisguerin V, de Graaf J, et al. 13. Landau DA, Carter SL, Stojanov P, McKenna A, Stevenson K, Lawrence MS, Mutated tumor alleles are expressed according to their DNA frequency. Sci et al. Evolution and impact of subclonal mutations in chronic lymphocytic Rep 2014;4:4743. leukemia. Cell 2013;152:714–26. 30. Hodis E, Watson IR, Kryukov GV, Arold ST, Imielinski M, Theurillat JP, et al. 14. Mroz EA, Rocco JW. MATH, a novel measure of intratumor genetic A landscape of driver mutations in melanoma. Cell 2012;150:251–63. heterogeneity, is high in poor-outcome classes of head and neck squamous 31. Haarberg HE, Smalley KS. Resistance to Raf inhibition in cancer. Drug cell carcinoma. Oral Oncol 2013;49:211–5. Discov Today Technol 2014;11:27–32. 15. Kemper K,KrijgsmanO, Cornelissen-Steijger P, Shahrabi A,Weeber F, Song JY, 32. Mengelbier LH, Karlsson J, Lindgren D, Valind A, Lilljebjorn H, Jansson C, et al. Intra- and inter-tumor heterogeneity in a vemurafenib-resistant mela- et al. Intratumoral genome diversity parallels progression and predicts noma patient and derived xenografts. EMBO Mol Med 2015;7:1104–18. outcome in pediatric cancer. Nat Commun 2015;6:6125. 16. Cibulskis K, Lawrence MS, Carter SL, Sivachenko A, Jaffe D, Sougnez C, et al. 33. Shain AH, Yeh I, Kovalyshyn I, Sriharan A, Talevich E, Gagnon A, et al. The Sensitive detection of somatic point mutations in impure and heteroge- genetic evolution of melanoma from precursor lesions. N Engl J Med neous cancer samples. Nat Biotechnol 2013;31:213–9. 2015;373:1926–36.

OF10 Cancer Res; 76(16) August 15, 2016 Cancer Research

Multiregion Whole-Exome Sequencing Uncovers the Genetic Evolution and Mutational Heterogeneity of Early-Stage Metastatic Melanoma

Katja Harbst, Martin Lauss, Helena Cirenajwis, et al.

Cancer Res Published OnlineFirst May 23, 2016.

Updated version Access the most recent version of this article at: doi:10.1158/0008-5472.CAN-15-3476

Supplementary Access the most recent supplemental material at: Material http://cancerres.aacrjournals.org/content/suppl/2016/05/21/0008-5472.CAN-15-3476.DC1

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

Reprints and To order reprints of this article or to subscribe to the journal, contact the AACR Publications Subscriptions Department at [email protected].

Permissions To request permission to re-use all or part of this article, use this link http://cancerres.aacrjournals.org/content/early/2016/07/28/0008-5472.CAN-15-3476. Click on "Request Permissions" which will take you to the Copyright Clearance Center's (CCC) Rightslink site.