Eleveld NG 2015

7/18/2019 Eleveld NG 2015

1/10

NATURE GENETICS ADVANCE ONLINE PUBLI CATION 1

ART I CLES

Neuroblastoma is a pediatric tumor of the peripheral sympatheticnervous system. Tumor behavior varies from spontaneous regressionto incurable progression. Patients with high-risk neuroblastoma havea survival rate of less than 50% (refs. 1,2), despite extensive treat-ment involving chemotherapy, surgery, radiation therapy and immu-notherapy. In a majority of patients, an initial response to therapy isobserved; however, up to 60% of these patients subsequently relapse

with therapy-resistant tumors35. Genetic alterations includingMYCNamplification and segmental chromosome alterations such as 1pdeletion, 11q deletion or 17q gain are associated with poor prognosis69;however, it is unknown which genetic defects are associated withdisease relapse.

Thus far, DNA obtained from 389 primary neuroblastomasat the time of diagnosis has been profiled by next-generation

Relapsed neuroblastomas show frequent RAS-MAPKpathway mutationsThomas F Eleveld1,29, Derek A Oldridge24,29, Virginie Bernard5,29, Jan Koster1, Leo Colmet Daage5,6,Sharon J Diskin24, Linda Schild1, Nadia Bessoltane Bentahar5, Angela Bellini6, Mathieu Chicard6, Eve Lapouble7,Valrie Combaret8, Patricia Legoix-N5, Jean Michon9, Trevor J Pugh10, Lori S Hart2,3, JulieAnn Rader2,3,Edward F Attiyeh24, Jun S Wei11, Shile Zhang11, Arlene Naranjo12, Julie M Gastier-Foster13,14, Michael D Hogarty24,Shahab Asgharzadeh1517, Malcolm A Smith18, Jaime M Guidry Auvil19, Thomas B K Watkins20,Danny A Zwijnenburg1, Marli E Ebus1, Peter van Sluis1, Anne Hakkert1, Esther van Wezel21,22,C Ellen van der Schoot21,22, Ellen M Westerhout1, Johannes H Schulte2326, Godelieve A Tytgat27,M Emmy M Dolman1, Isabelle Janoueix-Lerosey28, Daniela S Gerhard14, Huib N Caron27, Olivier Delattre28,Javed Khan11, Rogier Versteeg1, Gudrun Schleiermacher6,9,28,30, Jan J Molenaar1& John M Maris24,30

The majority of patients with neuroblastoma have tumors that initially respond to chemotherapy, but a large proportion willexperience therapy-resistant relapses. The molecular basis of this aggressive phenotype is unknown. Whole-genome sequencingof 23 paired diagnostic and relapse neuroblastomas showed clonal evolution from the diagnostic tumor, with a median of 29somatic mutations unique to the relapse sample. Eighteen of the 23 relapse tumors (78%) showed mutations predicted to activatethe RAS-MAPK pathway. Seven of these events were detected only in the relapse tumor, whereas the others showed clonalenrichment. In neuroblastoma cell lines, we also detected a high frequency of activating mutations in the RAS-MAPK pathway(11/18; 61%), and these lesions predicted sensitivity to MEK inhibition in vitroand in vivo. Our findings provide a rationale forgenetic characterization of relapse neuroblastomas and show that RAS-MAPK pathway mutations may function as a biomarker for

new therapeutic approaches to refractory disease.

1Department of Oncogenomics, Academic Medical Center of the University of Amsterdam, Amsterdam, the Netherlands. 2Division of Oncology, Childrens Hospitalof Philadelphia, Philadelphia, Pennsylvania, USA. 3Center for Childhood Cancer Research, Childrens Hospital of Philadelphia, Philadelphia, Pennsylvania, USA.4Department of Pediatrics, Perelman School of Medicine at the University of Pennsylvania, Philadelphia, Pennsylvania, USA. 5ICGEX Platform, Institut Curie, Paris,

France. 6Laboratory RTOP (Recherche Translationelle en Oncologie Pdiatrique), Transfer Department, Institut Curie, Paris, France. 7Unit de Gntique Somatique,

Institut Curie, Paris, France. 8Centre Lon-Brard, Laboratoire de Recherche Translationnelle Lyon, Lyon, France. 9Dpartement de Pdiatrie, Institut Curie, Paris,France. 10Princess Margaret Cancer Centre, University Health Network; Department of Medical Biophysics, University of Toronto, Toronto, Ontario, Canada.11Pediatric Oncology Branch, Oncogenomics Section, Center for Cancer Research, US National Institutes of Health, Gaithersburg, Maryland, USA. 12Department of

Biostatistics, University of Florida, Childrens Oncology Group (COG), Gainesville, Florida, USA. 13The Ohio State University College of Medicine, Columbus, Ohio,USA. 14Biopathology Center, Nationwide Childrens Hospital, Columbus, Ohio, USA. 15Division of Hematology/Oncology, Childrens Hospital Los Angeles, Los

Angeles, California, USA. 16Saban Research Institute, Childrens Hospital Los Angeles, Los Angeles, California, USA. 17Keck School of Medicine, University ofSouthern California, Los Angeles, California, USA. 18Cancer Therapy Evaluation Program, National Cancer Institute, Bethesda, Maryland, USA. 19Office of Cancer

Genomics, National Cancer Institute, Bethesda, Maryland, USA. 20Translational Cancer Therapeutics Laboratory, Cancer Research UK, London, UK. 21Department

of Experimental Immunohematology, Sanquin Research, Amsterdam, the Netherlands. 22Landsteiner Laboratory, Academic Medical Center of the University of

Amsterdam, Amsterdam, the Netherlands. 23Department of Pediatric Oncology and Hematology, Charit University Medicine, Berlin, Germany. 24German Consortium

for Translational Cancer Research (DKTK), Heidelberg, Germany. 25Department of Pediatric Oncology and Hematology, University Childrens Hospital Essen, Essen,

Germany. 26Translational Neuro-Oncology, West German Cancer Center (WTZ), University Hospital Essen, Essen, Germany. 27Department of Pediatric Oncology,

Emma Childrens Hospital, Academic Medical Center of the University of Amsterdam, Amsterdam, the Netherlands. 28INSERM U830, Laboratoire de Gntique

et Biologie des Cancers, Institut Curie, Paris, France. 29These authors contributed equally to this work. 30These authors jointly supervised this work.

Correspondence should be addressed to G.S. ([email protected]), J.J.M. ([email protected]) or J.M.M. ([email protected]).

Received 15 January; accepted 11 May; published online 29 June 2015; doi:10.1038/ng.3333
http://www.nature.com/doifinder/10.1038/ng.3333http://www.nature.com/doifinder/10.1038/ng.3333

7/18/2019 Eleveld NG 2015

2/10

2 ADVANCE ONLINE PUBLI CATION NATURE GENETICS

ART I CLES

sequencing techniques1013. These studies documented a relativepaucity of somatic mutations, with activating mutations in ALKand inactivating mutations inATRXbeing most frequent but withmutations in each of these genes occurring in less than 10% ofthe cases studied. These studies challenge the notion of precision med-icine based on somatic genetic alterations in primary neuroblastomatumors alone. Notably, none of these large-scale studies considered

the tumor genome at relapse, partly because patients are rarely sub-jected to a tumor biopsy at the time of disease progression, as currentdiagnostic radiology techniques such as metaiodobenzylguanidinescintigraphy are highly sensitive and specific14.

Recently, sequencing of theALKlocus in neuroblastomas at thetime of relapse identified 14 activating mutations in 54 cases (26%)15,suggesting that the frequency ofALKaberrations is higher in relapsedneuroblastoma genomes. This implies selection of tumor cells withalterations in genes that mediate neuroblastoma relapse. Therefore,to identify genetic alterations associated with relapsed neuroblas-toma, we performed whole-genome sequencing of 23 triplet samplesof primary tumor, relapsed tumor and constitutional DNA.

RESULTS

Clonal evolutionWe sequenced the genomes of 23 triplet samples constituting pri-mary and relapse neuroblastomas and lymphocytes (SupplementaryTable 1). Tumors were of all stages and had variable outcome, withthe only eligibility criterion being the availability of high-qualityDNA from the triplet samples. There was a roughly equal distributionof cases among low-, intermediate- and high-risk groups5. Themedian time from diagnosis to relapse was 11.3 months (range of190 months). Twenty-one of the 23 subjects in this study receivedchemotherapy before relapse, and 8 also received radiation therapy,according to internationally accepted treatment protocols (Table 1).Thus, all patients underwent similar chemotherapy regimens, withhigh-risk patients additionally receiving radiation therapy and high-

dose chemotherapy with stem cell rescue. No patient on this studyreceived targeted inhibitors to any oncogenic pathway between thetime of diagnosis and relapse. The majority of low-risk cases receivedchemotherapy because of the site and/or size of the primary tumor.

We analyzed the sequence data for somatic mutations resulting inamino acid changes or located in splice-site regions within 3 bp of anexon, as well as for focal structural aberrations of regions contain-ing five genes or fewer (Supplementary Tables 24). There was amedian of 14 more mutations in the relapsed samples than in thesamples at diagnosis (Fig. 1and Supplementary Fig. 1). On aver-age, 28% of the mutations detected in the primary tumor were alsodetected at relapse, showing that the primary and relapsed tumorswere of common descent. To gain more insight into the clonalarchitecture, we estimated the cancer cell fraction (CCF) of all

somatic mutations using a customized reimplementation of a pre-viously described Bayesian approach16, which infers CCF frommutant allele fractions determined by sequencing and accountsfor contamination and locus-specific copy number. This analy-sis yielded a median CCF of 61% for mutations detected in theprimary tumor but lost in the relapse tumor, in comparison to amedian CCF of 90% for primary tumor mutations shared with therelapse tumor (Supplementary Fig. 2a), a pattern that is consistentwith subclonal outgrowth of the relapsed tumor. Furthermore, weestimated a median CCF of 83% for all relapse tumor mutations incomparison to a median CCF of 75% for all primary tumor muta-tions, indicating clonal enrichment of a subset of mutations atrelapse (Supplementary Fig. 2b). Comparison of genes affected by

smaller structural events and chromosomal copy number altera-tions in primary and paired relapse tumors showed similar results,with a subset of aberrations shared but many detected only in theprimary or relapse tumor (Fig. 1c,d, Supplementary Fig. 3 andSupplementary Table 4).

Enrichment of mutations that activate RAS-MAPK signaling

Unbiased pathway analysis

17

using whole-genome sequencingdata from the relapse samples on a per-patient basis identified astrong enrichment (P= 6.1 107) for mutations in genes associ-ated with RASmitogen-activated protein kinase (MAPK) signal ing(Supplementary Table 5). We next filtered the identified mutatedgenes against the Cancer Gene Census18and subsequently focusedon hotspot regions by selecting mutations annotated in the Catalogueof Somatic Mutations in Cancer (COSMIC) to identify events thatare well annotated to activate this pathway. Fifteen of 23 relapse sam-ples contained somatic mutations that met these criteria. In addi-tion, three relapse samples showed structural alterations involvingthese RAS-MAPK pathway genes; thus, we detected aberrations inthis pathway in 18 of 23 relapse samples (78%), and all alterationswere consistent with pathway activation (Table 2). Eleven mutations

activating RAS-MAPK signaling that were present in primary tumorswere all preserved in the corresponding relapse tumors. Seven muta-tions were not detectable in the primary tumor at the sequencingdepth achieved, and we therefore employed ultra-deep sequencing todetermine whether these mutations were present in fractions underthe whole-genome sequencing detection limit. TheALKmutation inN607 was found in the primary tumor, whereas the other mutationswere undetectable (Supplementary Fig. 4a). We assayed structuralaberrations using PCR-based methods. Only theALKrearrangementin N790 was shown to be present at low frequency in the primarytumor (Supplementary Fig. 4b).

ALK aberrations occurred in ten relapse tumors and were alsodetected by whole-genome sequencing in seven of the corresponding

primary tumors. All detected single-nucleotide variants (SNVs) havebeen proven to constitutively activate this receptor tyrosine kinaseknown to activate RAS-MAPK signaling19. Furthermore, one relapsesample showed a de novoamplification giving rise to a PPM1G-ALKfusion gene, which activated the RAS-MAPK pathway when expressedin neuroblastoma cell lines (Supplementary Fig. 5).

Two tumors showed relapse-specific inactivation of the NF1tumor-suppressor gene, through homozygous deletion in one caseand heterozygous deletion combined with a splice-site mutation inthe other (Supplementary Fig. 6). NF1inactivation has been reportedin neuroblastoma and confers activation of RAS-MAPK signalingand resistance to retinoic acid20. One pair of primary and relapsetumors showed a heterozygous mutation in PTPN11.Mutations inPTPN11activate RAS-MAPK signaling, and the identified mutation

encoding a p.Ala72Thr substitution has been reported in leukemiaand neuroblastoma12,21.

One relapse tumor showed a tandem duplication in the BRAFgenethat was not detected in the primary tumor. This rearrangement leadsto expression of a BRAFtranscript that encodes an elongated proteinwith two kinase domains. Expression of this BRAFgene with a tandemduplication in a neuroblastoma cell line induced activation of the RAS-MAPK pathway (Supplementary Fig. 7). Taken together, the somaticmutations detected in this case series shown to activate the RAS-MAPKpathway were mutually exclusive, with no case having two somaticevents known to hyperactivate this growth-promoting pathway.

We hypothesized that mutations activating RAS-MAPK signalingexhibit relapse-specific enrichment due to treatment. We therefore

7/18/2019 Eleveld NG 2015

3/10


ART I CLES

performed a clonality analysis, comparing CCF estimates for RAS-

MAPK mutations in paired primary (CCFp) and relapse (CCFr)tumors (Fig. 2and Supplementary Fig. 8). RAS-MAPK mutationswere almost universally present within major subclonal populationsat relapse, as indicated by CCFr> 0.5 with probability >90% underthe posterior distribution for 14 of 15 relapse tumors. In 7 of 15tumor pairs, there was strong evidence of relapse-specific enrich-ment of RAS-MAPK mutationsincluding mutations in ALK(4 pairs), HRAS(1 pair), KRAS(1 pair) and NF1 (1 pair)basedon a criterion of CCFr> CCFpwith probability >90% for each pair.By contrast, the probability that CCFr> CCFp fell within 2080%bounds for the remaining eight pairs, indicating that RAS-MAPKmutations that were already present in the primary tumor wereretained at relapse. Collectively, these results support RAS-MAPKmutations as somatic drivers that undergo positive selection over

the course of neuroblastoma treatment.

Chromosomal aberrationsThree relapse samples showed homozygous deletions in the CDKN2Alocus, encoding the tumor-suppressor proteins p14ARFand p16, whereasboth alleles of CDKN2A were present in the corresponding primarytumors (Supplementary Fig. 9). CDKN2A deletions were previouslyreported as frequent events in neuroblastoma relapse22. We detected otherrelapse-specific segmental chromosome defects, including loss of 6q (fivecases) and loss of 17p (three cases). Furthermore, we detected relapse-specific aberrations that are frequently detected in primary neuroblastomaand are associated with poor prognosis, including loss of chromosomes 1p(one case) and 11q (three cases) (Fig. 1cand Supplementary Fig. 3)7,9.

RAS-MAPK pathway mutations and sensitivity to MEK inhibition

To determine whether neuroblastoma cell lines contain RAS-MAPKmutations, we analyzed whole-genome sequencing data from aseries of human-derived neuroblastoma cell lines for mutations in

ALK, NRAS, HRAS, KRAS, BRAF, PTPN11and NF1. Eleven of the18 cell lines showed such mutations (Supplementary Fig. 10 andSupplementary Table 6).

We tested our cell line panel for sensitivity to the MEK inhibitorstrametinib, cobimetinib and binimetinib, to determine the relationshipbetween mutation status and drug sensitivity. The data showed clus-tering of the cell lines into four groups with increasing sensitivityto MEK inhibition: (i) lines without RAS-MAPK mutations; (ii) lineswith ALKmutations; (iii) lines with NF1 mutations; and (iv) lineswith RAS gene or BRAFmutations (Supplementary Fig. 11). In thecell lines with mutated RAS genes or BRAF, MEK inhibitor treatment

caused almost complete cell cycle arrest at low nanomolar concentra-tions, whereas in the NF1- andALK-mutated lines the effect on cellcycle inhibition was less robust (data not shown). When the sensitiv-ity of the cell lines was expressed as the concentration at which cellgrowth was inhibited by 50% (GI50), there were significant differencesbetween cell lines with and without RAS-MAPK mutations (Fig. 3ac).GI50values for the three different compounds were highly correlatedin the cell line panel (Supplementary Fig. 12), suggesting an on-targeteffect (r2= 0.490.79; P< 0.01). A relationship between mutationstatus and sensitivity to MEK inhibition was also observed in anindependent published data set23(Supplementary Fig. 13).

To validate that ALK and RAS gene mutations directly activatethe RAS-MAPK pathway in neuroblastoma cells, we induced

Table 1 Clinical characteristics of the patient cohort

Number Patient ID

Risk

stratification

(INRG)

Stage

(INSS)

MYCN

status Sex

Age at

diagnosis

(months)

Time to

relapse

(months

after

diagnosis)

Time to

last report

(months

after

diagnosis) Status

Treatment (between

diagnosis and relapse)

Diagnosis

location

Relapse

location

Radiation

therapy

Chemo-

therapy Surgery

1 FR_NB1178 High 4 Non-amplified M 30 21 24 Dead Y Y Y Retroperitoneum Liver

2 FR_NB1269 High 4 Amplified M 14 9 11 Dead N Y Y Retroperitoneum Retroperitoneum

3 FR_NB1382 High 4 Amplified M 4 50 64 Dead N Y Y Abdomen Abdomen 4 NL_N774 High 4 Non-amplified F 83 6 10 Dead Y Y Y Adrenal gland Abdomen

5 US_PATNKP High 4 Non-amplified M 113 20 51 Dead Y Y Y Retroperitoneum Pelvis

6 US_PASGAP High 4 Non-amplified M 44 42 51 Dead Y Y Y Adrenal gland Soft tissue, skull

7 NL_N790 High 3 Amplified F 42 63 110 Dead Y Y Y Adrenal gland Liver

8 US_PASHFA High 3 Amplified F 13 7 11 Dead N Y Y (biopsy

only)

Adrenal gland Abdomen

9 FR_NB804 Intermediate 4 Non-amplified F 2 26 56 Alive N Y Y Subcutaneous

nodule

Orbita

10 NL_N607 Intermediate 4 Non-amplified F 3 7 84 Alive Y Y N Liver Orbita

11 US_PARHAM Intermediate 4 Non-amplified F 11 1 81 Dead N Y N Pelvis Pelvis

12 US_PATYIL Intermediate 4 Non-amplified F 11 8 16 Dead N Y Y Abdomen Pararenal

13 NL_N571 Intermediate 3 Non-amplified M 49 12 16 Dead N Y Y Adrenal gland Abdomen

14 US_PASNPG Intermediate 3 Non-amplified F 10 10 63 Alive N Y Y Retroperitoneum Paraspinal

15 US_PARBAJ Intermediate 3 Non-amplified M 1 10 88 Alive N Y Y Retroperitoneum Abdomen

16 US_PAUDDK Intermediate 3 Non-amplified M 12 11 38 Alive N Y Y (biopsy

only)

Pelvis Pelvis

17 FR_NB1224 Low 2 Non-amplified M 15 8 18 Alive N Y N Mediastinum Mediastinum

18 FR_NB0175 Low 2 Non-amplified M 98 90 103 Dead N N Y Retroperitoneum Retroperitoneum

19 FR_NB308 Low 2 Non-amplified F 2 21 91 Alive N Y N Abdomen Abdomen

20 NL_N041 Low 2 Non-amplified F 109 72 92 Dead Y Y Y Abdomen Abdomen

21 US_PAPVEB Low 2 Non-amplified M 57 9 40 Dead N N N Adrenal gland Bone marrow

22 NL_N789 Low 1 Non-amplified F 124 78 168 Alive Y Y Y Adrenal gland Lymph node

23 FR_NB399 Low 4s Non-amplified M 0 7 134 Dead N Y N Subcutaneous

nodule

Liver

M, male; F, female; Y, yes; N, no; INRG, International Neuroblastoma Risk Group; INSS, International Neuroblastoma Staging System.

7/18/2019 Eleveld NG 2015

4/10


ART I CLES

PNB1178

d 1

NB1269

NB1382

N774

PATNKP

PASGAP

N790

PASHFA

NB804

N607

PARHAM

PATYIL

N571

PASNPG

PARBAJ

PAUDDK

NB1224

NB0175

NB308

N041

PAPVEB

N789NB399

RPRPRPRPRPRPRPRPRPRPRPRPRPRPRPRPRPRPRPRPR

PRPR

2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 X Y

2 0

log2(ratio)

2

250b c

SharedPrimary unique

Relapse unique200

150

Numberofsomaticmutations

100

50

NB1178

NB1269

NB1382

N774

PATNKP

PASGAP

N790

PASHFA

NB804

N607

PARHAM

PATYIL

N571

PASNPG

PARBAJ

PAUDDK

NB1224

NB0175

NB308

N041

PAPVEB

N789

NB399

0

70SharedPrimary unique

Relapse unique

PatientPatient

Numberofgenesaffectedby

structuralvariants

NB1178

NB1269

NB1382

N774

PATNKP

PASGAP

N790

PASHFA

NB804

N607

PARHAM

PATYIL

N571

PASNPG

PARBAJ

PAUDDK

NB1224

NB0175

NB308

N041

PAPVEB

N789

NB399

60

50

40

30

20

10

0

Primarya

TAFIL

C9

orf

3

PRTF

DC

1

ZNF484

PALM2

SH3GLB2

BM

11

DNMBP

SORCS1

MPEG

1

DSCA

ML1

KRT8

6

C14orf49

VPS18 GPR176

FURIN

CES2EDC4

ZNF778

MED11

DNAH9

KIF1C

KSR1

ALOX128

RUNDC1LYZL6

ABCA6

DSG1

ZNF566

CEP250

SLC2A10

POLDIP3

SMC1A

SA

GE1

CCL4NF1

IL15RA

ARHGEF5

CYP3A7

LOC100289187

NPCIL1

PDE78

KIAA1244

PNPLA1

GCM1

HISTIHIT

SH3TC2

KIF3A

CMYA5

TLL1

ARSJ

UGT2A3

DCAF4L1

FRYL TMPRSS11F

CORIN

SERPI

NI2PAQR

9

CELSR3FA

M198A

UGT1

A1SC

N2A

LY75

CCDC

93

DNMT3

A

ROCK2

OR2B11

IGFN1

PKP1

TMEM81

COL

24A1

ZMYM1

FARS

B

DNAH

12SETD

2UBA7

PIK3CA

NR2E1

LRGUK

FLG

FLG

LAMC1

2

3

4

5

6

7

8

FSIP2

BSN

ZNF7

17

PIK3CA

ZBTB49

SLC6A3

MUC3A

TRMT12

1

Relapse

9

SVE

P1

SEC61A2

ZC3H12C

ARHG

AP32

10

11

12

13

14

15

16

18

19

20

21

17

SLC30A4

MAP2K3

LLGL2

CCDC165

MUC16

GNAS

APOL1

22

x

y

2

3

BSN

FSIP2

4

5

6

7

8

1

910

11

ZC3H12C

SEC

61A2

12

13

14

15SLC30A4

16

18CCDC165

19

20

21

17

22

xy

Figure 1 Mutational spectra of diagnostic and relapsed neuroblastomas. (a) Circos plots showing structural variations and somatic mutations in the

primary and corresponding relapse tumors from PASGAP. The inner ring represents copy number variations (red, gain; green, loss) identified on the

basis of coverage of the tumor and lymphocyte genomes. The lines traversing the ring indicate inter- and intrachromosomal rearrangements identified

by discordant mate pairs from paired-end reads. Aberrations are colored according to their presence (gray, only detected in the primary tumor; black,

detected in the primary and relapse tumors; blue, only detected in the relapse tumor; red, events predicted to activate RAS-MAPK signaling). (b) The

number of nonsynonymous mutations identified by whole-genome sequencing in 23 primary tumors and their corresponding relapse tumors. Mutations

identified in both primary and relapse tumors are shown in gray, whereas mutations unique to the primary or relapse tumor are shown in blue and yellow,

respectively. (c) The number of structural variants identified by whole-genome sequencing in 23 primary tumors and their corresponding relapse tumors

using the same color scheme as in b. (d) Coverage plots displaying the structural variations in primary and relapse tumors. Coverage was calculated for

1-Mb bins along the genome and normalized to the coverage in patient lymphocyte DNA. Color intensity reflects the magnitude of the gain or loss (blue,

loss; red, gain) in the associated chromosomal location. P, primary; R, relapse.

7/18/2019 Eleveld NG 2015

5/10


ART I CLES

the expression of an ALK Phe1174Leu and an NRAS Gln61Lysmutant in two cell lines that did not harbor RAS-MAPK mutations.

Expression of either mutated protein caused activation of the pathway(Supplementary Fig. 14). We have shown previously that knockdownof NF1causes hyperactivated RAS-MAPK signaling in neuroblastomacell lines20.

We next treated various humanneuroblastoma-derived cell line xenograftmodels, representing the four groupsdescribed above, with the MEK inhibitorbinimetinib. SK-N-AS xenografts, whichharbor an NRAS mutation encodingp.Gln61Lys, showed inhibition of tumor

growth and increased survival when treatedwith binimetinib, in a dose-dependentfashion (Fig. 3d). NBL-S xenografts have aninactivating mutation in one allele ofNF1andan almost complete absence of NF1 proteinexpression (Supplementary Fig. 15); thesexenografts also showed inhibition of growthwith treatment. Conversely, treatment ofKelly and IMR-5 xenografts had no effecton tumor growth. IMR-5 cells do not haveRAS-MAPK pathway mutations detectable bywhole-exome sequencing (data not shown),whereas Kelly cells harbor anALKmutationencoding a p.Phe1174Leu substitution.

We then determined whether inhibitionof cell growth corresponds with inhibi-tion of the RAS-MAPK pathway in the celllines that were used for the mouse xenograftexperiments. We treated the cell lines withincreasing concentrations of binimetinibin vitrofor 24 h and screened for ERK phos-phorylation (Supplementary Fig. 16). Thethree lines with RAS-MAPK mutationsshowed phosphorylation of ERK underuntreated conditions, reaffirming that these

mutations indeed lead to activation of the RAS-MAPK pathway. Uponexposure to binimetinib, SK-N-AS and NBL-S cells showed a dose-

dependent decrease in the levels of phosphorylated ERK, whereasKelly cells showed no significant change. These data suggest that theminimal effect of MEK inhibition in vitroand the absence of an effectin vivofor the Kelly cell line may be due, at least in part, to continued

Table 2 RAS-MAPK pathway mutations in relapsed neuroblastomas

Genomic aberrations in RAS-MAPK pathway in relapse tumors

Number Patient ID Gene Genomic event Impact Event type

COSMIC

ID

Detected

in primary

tumora

1 FR_NB1178

2 FR_NB1269 ALK Somatic mutation p.Leu1196Met Activating 99137 Yes

3 FR_NB1382 ALK Somatic mutation p.Tyr1278Ser Activating 28058 No

4 NL_N774 PTPN11 Somatic mutation p.Ala72Thr Activating 13014 Yes

5 US_PATNKP FGFR1 Somatic mutation p.Asn546Lys Activating 19176 Yes

6 US_PASGAP NF1 Somatic mutation

(splice donor) +

hemizygous deletion

Inactivating No

7 NL_N790 ALK Amplification and

fusion

Activating No

8 US_PASHFA

9 FR_NB804

10 NL_N607 ALK Somatic mutation p.Phe1174Leu Activating 28055 No

11 US_PARHAM ALK Somatic mutation p.Arg1275Gln Activating 28056 Yes

12 US_PATYIL NRAS Somatic mutation p.Gln61Lys Activating 580 Yes

13 NL_N571 NF1 Homozygous deletion Inactivating No

14 US_PASNPG ALK Somatic mutation p.Phe1174Ile Activating 28491 Yes

15 US_PARBAJ HRAS Somatic mutation p.Gln61Lys Activating 496 No

16 US_PAUDDK

17 FR_NB1224 ALK Somatic mutation p.Arg1275Gln Activating 28056 Yes

18 FR_NB0175 ALK Somatic mutation p.Tyr1278Ser Activating 28058 Yes

19 FR_NB308 ALK Somatic mutation p.Phe1174Leu Activating 28061 Yes

20 NL_N041 BRAF Tandem duplication

of catalytic domain

Activating No

21 US_PAPVEB KRAS Somatic mutation p.Gly12Asp Activating 521 Yes

22 NL_N789

23 FR_NB399 ALK Somatic mutation p.Arg1275Gln Activating 28056 Yes

aBased on whole-genome sequencing analysis.

1.00

0.75

C

CF

posteriorprobabilitydensity

0.50

0.25

NB1269

ALK(L1196M)

N774

PTPN

11(A72T)

PATNKP

FGFR

1(N54

6K)

PATYIL

NRAS

(Q61K)

NB399

ALK(R1275

Q)

PAPVEB

KRAS

(G12D)NB

308

ALK(F1174L)

NB0175

ALK(Y1278

S)NB

1224

ALK(R1275

Q)

PARB

AJ

HRAS

(Q61K)

PASNPG

ALK(F1174I)

PARH

AM

ALK(R1275

Q)N607

ALK(F1174L)

PASG

AP

NF1(Splice)

NB1382

ALK(Y1278

S)

0

Primary

Relapse

Figure 2 RAS-MAPK pathway mutations reside within major relapsed neuroblastoma subclones. Each of the 15 panels represents a primary-relapse pair

with a corresponding RAS-MAPK pathway mutation. Posterior distributions of CCF were computed by the method of Carter and colleagues16and are

represented by violin plots, with black dots positioned at the distribution medians. Four primary-relapse pairs (NB1382, PASGAP, N607 and PARBAJ)

possess relapse-specific RAS-MAPK pathway mutations that are undetectable in the corresponding primary tumors by whole-genome sequencing. Three

additional pairs (NB1224, NB308 and PAPVEB) also show evidence of relapse-specific enrichment of RAS-MAPK mutations based on the criterion

that the probability of CCFr > CCFp was >90% for each pair. CCF estimates for ALKmutations in NB308 reflect p.Phe1174Leu (c.352C>A) and

p.Phe1174Leu (c.352C>G) substitutions in the primary and relapse samples, respectively, as previously described15.

7/18/2019 Eleveld NG 2015

6/10


ART I CLES

ERK phosphorylation. To confirm that theresponse we observed in vivowas also due

to target inhibition, we analyzed ERK phos-phorylation in the xenografts that respondedto treatment and demonstrated a dose-dependent decrease in the levels of phos-phorylated ERK (Supplementary Fig. 17).

DISCUSSIONRecent studies using next-generation sequencing methodologies indi-cate, first, a low mutation rate in primary neuroblastoma tumors, witha mean of 15 nonsynonymous mutations per sample, and, second,a paucity of recurrently affected genes1013. Because relapsed high-risk neuroblastoma tumors often demonstrate substantial resistanceto chemotherapy and are typically lethal, we hypothesized that the

mutational spectrum of relapsed disease would be different than thatfor primary tumors and might mediate therapy resistance.In this study, we characterized the genomes of 23 relapsed neu-

roblastomas and compared each to the genome of the correspondingprimary tumor. We show that the relapsed tumors generally containmore mutations and structural aberrations and that clonal selectiontakes place between the primary tumor and the relapse tumor. Wefound that 18 of 23 relapse tumors harbored mutations predicted tohyperactivate the RAS-MAPK signaling pathway and that cell linescontaining similar mutations show sensitivity to inhibition of MEK,a downstream node in the canonical growth-promoting pathway.These results provide a strong rationale for recommending biopsyand genomic characterization of relapsed neuroblastoma tumors andfor prioritizing the clinical testing of MEK inhibition strategies in the

treatment of relapsed neuroblastoma.Because of the sensitivity and specificity of modern imaging

modalities, the diagnosis of neuroblastoma relapse rarely requires atumor biopsy. In addition, until recently, there has been no realisticpotential for therapeutic benefit based on biopsy results, and thereforevery few relapse neuroblastoma samples are available for study. Herewe collected high-quality material from 23 primary-relapse tumorpairs across the spectrum of neuroblastoma phenotypes, includingthose assigned to high-, intermediate- and low-risk groups5. Theonly inclusion criterion for this study was the availability of matchedsamples, but some cases may have been biopsied owing to anunusual clinical course. Indeed, there were a fairly high numberof intermediate- and low-risk tumors that normally do not show

frequent relapse5. However, the frequency of RAS-MAPK mutationsdid not differ among the groups, so it is unlikely that the over-representation analysis is inf luenced by this bias.

The high frequency of RAS-MAPK pathway mutations at diagnosisin this cohort was unexpected, as such high frequencies were notreported in whole-genome sequencing series of primary tumors1013.It is possible that the presence of these lesions in a diagnostic sampleis a biomarker of a more aggressive clinical course and a higher likeli-hood of relapse. These findings need to be validated in a prospectivepatient cohort.

We detected several recurrent structural aberrations at the time ofrelapse. Partial loss of chromosome 6q was observed in five relapsesamples, and homozygous deletion of CDKN2Awas found in threerelapse samples. Both events are infrequent in primary neuroblastomaand present interesting targets for further research. Furthermore,neuroblastoma-associated aberrations such as loss of 1p and 11qwere observed in the relapse tumor and not in the primary tumor,indicating that these events might not be tumor initiating but ratherare crucial steps in neuroblastoma tumor evolution24.

Events affecting RAS-MAPK signaling were detected in 18 of23 relapse samples. In four cases, we identified structural variants,highlighting the benefit of whole-genome sequencing for detectionof the full spectrum of genetic alterations. In 7 of these 18 cases,the mutations were observed only in the relapse tumor, which indi-

cates that analysis of primary tumor samples is not sufficient to guidethe choice of treatment for neuroblastoma relapses. These findingsare in line with the de novooccurrence ofALKmutations reportedpreviously15.

The observation that several RAS-MAPK mutations were presentin the relapse tumor but not in the corresponding primarytumor favors a model where subclones with secondary drivermutations expand over time, possibly under the selective pressureof chemotherapy, as was recently described for chronic lymphocyticleukemia25. Whether these mutations occurred between diagnosisand relapse, were present at levels below detection limits or wereundetectable owing to spatial heterogeneity in the primary tumorremains to be determined.

2

a b c

22

2

2

1

1

1

1log(GI50

)(M)

log(GI50

)(M)

log(GI50

)(M)

RASgene

orRAF

RASgene

orRAF

RASgene

orRAF

0

00

P= 0.0039

Binimetinib

**

*

*** ***

***

***

****

*

NF1

NF1

NF1

ALK

ALK

ALK

Notm

utated

Notm

utated

Notm

utated

11

3

3

4

P= 0.0135P= 0.0023

Trametinib Cobimetinib

2 2

20

2 2

20

1 1 1 1

1010

0

0

0

0 10 1020 20

0 0

00

NRASmutated(SK-N-AS)

Tumor

volume(cm

3) 3

d3

30 30 30 30

3 3

40 40 4040 50

20100

0 0 20 20 20

2020

10 10 1030 30 30 3040 40 40 4050 50 50 50

50 50 50

NF1mutated(NBL-S)

ALKmutated(Kelly)

Non-mutated(IMR-5)

Percent

survival

60

80

100

00

00

0

20 2040 40 40 40

60 60 60

60 60 6060

80 80 80100 100 100

Day after

start of treatment

Day after

start of treatment

Day after

start of treatment

Day after

start of treatment

VehicleBinimetinib,

3 mg/kgBinimetinib,

30 mg/kg

VehicleBinimetinib,3 mg/kg

Binimetinib,30 mg/kg

*

******

*

**

**

******** **

******

Figure 3 Sensitivity of neuroblastoma cell

line models to MEK inhibition therapy.

(ac) GI50values for a panel of neuroblastoma cell

lines treated with binimetinib (a), trametinib (b)

and cobimetinib (c). Cell lines are grouped

according to mutation status (Supplementary

Table 6). Red bars represent the mean GI50

value for each group.Pvalues were derived

from Kruskal-Wallis tests. Students ttests

were performed to determine differences

between non-mutated and mutated groups:

*P< 0.05, **P< 0.01, ***P< 0.001.

(d) Human neuroblastoma-derived SK-N-AS,

NBL-S, Kelly and IMR-5 xenografts were

treated with binimetinib (3 mg/kg or

30 mg/kg) or vehicle by oral gavage twice

daily. Each cohort consisted of ten mice, and

tumor volumes (cm3) and percent survival

are shown. Error bars, s.e.m.; significance

is denoted: *P< 0.05.

7/18/2019 Eleveld NG 2015

7/10


ART I CLES

It has been firmly established that mutations in the RAS-MAPKpathway can occur as resistance mechanisms against treatment withtargeted kinase inhibitors26; however, no targeted inhibitors were usedin the treatment of our patient cohort. Mutations in the RAS-MAPKpathway may also be associated with resistance to conventionalcytotoxic therapies in neuroblastoma, but more research is neededto establish the molecular basis of this phenotype.

TheALKgene was mutated in ten relapse samples, and it is knownthat the most frequentALKmutations in neuroblastoma activate theRAS-MAPK signaling pathway27. The results of our xenograft thera-peutic studies indicate that single-agent treatment with a MEK inhibitormight not be effective inALK-mutated tumors. However, the findingthatALK-mutated cell lines consistently showed some sensitivity toMEK inhibition in vitrosuggests that activated RAS-MAPK signalingdoes have a role inALK-mutated neuroblastoma and warrants furtherinvestigation on the use of MEK inhibitors in combination therapies.ALK inhibitors have proven to be effective in the treatment ofALK-mutated tumors28, but some mutations are associated with resistanceto currently available ALK inhibitors19. Therefore, combined MEKand ALK inhibition might improve response in tumors containingsuch mutations. Combination of these inhibitors with ones targeted

against other pathways that are activated in ALK-mutated neuro-blastoma, such as phosphoinositide 3-kinase (PI3K) and mTOR29,30signaling, might also improve therapeutic efficacy.

We also detect mutations in NF1, BRAF, PTPN11, FGFR1and thethree RAS genes, and all lesions in the RAS-MAPK pathway weremutually exclusive. Cell lines with RAS-MAPK mutations showmoderate to high sensitivity to MEK inhibitors, and treatment withthe inhibitor binimetinib of SK-N-AS xenografts, which contain anNRASmutation, as well as NBL-S cells, which have loss ofNF1, resultsin significant therapeutic efficacy. These findings suggest that MEKinhibition might be of clinical benefit in the treatment of neuroblas-toma relapses containing RAS-MAPK mutations.

Our study provides a strong rationale for performing biopsies

and genomic characterization of relapsed neuroblastomas, withthe prospect of direct patient benefit. We show that RAS-MAPKpathway mutations occur frequently in relapsed neuroblastoma andpropose matching such pathway lesions to novel therapies such as MEKinhibition. A prospective clinical trial in which biopsy and next-generation sequencing are used for the selection of a targeted therapywill be required to determine the clinical implication of this newlydefined genetic landscape of relapsed neuroblastoma.

URLs.R2 bioinformatics platform, http://r2.amc.nl/ ; Exome VariantServer (EVS), http://evs.gs.washington.edu/EVS/ .

METHODSMethods and any associated references are available in the online

version of the paper.

Accession codes. The whole-genome sequencing data have beendeposited at the European Genome-phenome Archive (EGA) underaccession EGAS00001001184for the French cases and under acces-sion EGAS00001001183for the Dutch cases. Sequence data for theUS cases are available in the database of Genotypes and Phenotypes(dbGaP) under accessionphs000467. Whole-exome sequencing datafor neuroblastoma cell lines are available through the Sequence ReadArchive (SRA) under BioProject PRJNA282395.

Note: Any Supplementary Information and Source Data files are available in theonline version of the paper.

ACKNOWLEDGMENTSThe authors would like to thank N. Ross for sample preparation and qualitycontrol. This study was supported in part by US National Institutes of Healthgrants RC1MD004418 to the TARGET consortium and CA98543 and CA98413to the Childrens Oncology Group and was supported in part by a grant throughoutthe University of Pennsylvania Genome Frontiers Institute. In addition, thisproject was funded in part with federal funds from the National Cancer Institute,US National Institutes of Health, under contract HHSN261200800001E.The content of this publication does not necessarily reflect the views or policies

of the US Department of Health and Human Services nor does mention of tradenames, commercial products or organizations imply endorsement by theUS government. The binimetinib xenograft studies were supported through aresearch collaborative agreement between the Childrens Hospital of Philadelphiaand Novartis Pharmaceuticals. In France, this study was supported by theAnnenberg Foundation and the Nelia and Amadeo Barletta Foundation. Fundingwas also obtained from SiRIC/INCa (grant INCa-DGOS-4654) and from theComit dEvaluation et Suivi des Projets de Recherche de Transfert (CEST) ofInstitut Curie. This study was also funded by the Associations Enfants et Sant,Association Hubert Gouin Enfance et Cancer, Les Bagouz Manon and Les Amisde Claire. Deep sequencing experiments were conducted on the ICGex next-generation sequencing platform at the Institut Curie funded by the EQUIPEXInvestissements dAvenir program (ANR-10-EQPX-03) and ANR10-INBS-09-08from the Agence Nationale de le Recherche and by the CanceropleIle-de-France. In the Netherlands, this study was supported by grants from theVilla Joep Foundation, the Kinderen Kankervrij Foundation and the NetherlandsCancer Foundation.

AUTHOR CONTRIBUTIONSJ.M.M., J.J.M. and G.S. co-coordinated the project. T.F.E., D.A.O., V.B., J. Koster,R.V., G.S., J.J.M. and J.M.M. prepared the manuscript. D.A.O., J. Koster, L.C.D.,V.B., N.B.B., P.L.-N., S.J.D., T.J.P., E.F.A., J.S.W., S.Z., S.A., T.B.K.W., D.A.Z. andA.N. generated microarray and/or next-generation sequencing data and/orperformed computational data analysis. T.F.E., L.S.H., J.R. and J.M.M. conceiveddrug testing and gene manipulation experiments and analyzed the data. P.v.S.,M.E.E., E.L., V.C., A.B. and M.C. performed sample preparation. L.S.H., J.R., L.S.,A.H., M.E.M.D. and T.F.E. contributed to in vitrodrug testing and manipulationexperiments. A.B., M.C., P.L.-N., D.A.O., T.J.P., E.v.W. and C.E.v.d.S. performedsubclonal detection of mutational and structural variants. J.M.G.-F., M.D.H.,E.M.W., J.H.S., G.A.T., H.N.C., J.M. and V.C. contributed patient tissue and/orother biological reagents. M.A.S., J.M.G.A., D.S.G., J. Khan and J.M.M. organizedthe US neuroblastoma TARGET consortium sequencing effort. J.J.M., R.V., J.Koster. and H.N.C. organized the Dutch consortium sequencing effort. G.S., J.M.,

I.J.-L. and O.D. organized the French consortium sequencing effort.

COMPETING FINANCIAL INTERESTSThe authors declare competing financial interests: details are available in the onlineversion of the paper.

Reprints and permissions information is available online at http://www.nature.com/

reprints/index.html.

1. Hero, B. et al.Localized infant neuroblastomas often show spontaneous regression:

results of the prospective trials NB95-S and NB97. J. Clin. Oncol.26, 15041510

(2008).

2. Park, J.R. et al. Childrens Oncology Groups 2013 blueprint for research:

neuroblastoma. Pediatr. Blood Cancer60, 985993 (2013).

3. Simon, T. et al. Treatment and outcomes of patients with relapsed, high-risk

neuroblastoma: results of German trials. Pediatr. Blood Cancer 56, 578583

(2011).

4. Maris, J.M. Recent advances in neuroblastoma. N. Engl. J. Med.362, 22022211

(2010).5. Cohn, S.L. et al.The International Neuroblastoma Risk Group (INRG) classification

system: an INRG Task Force report. J. Clin. Oncol.27, 289297 (2009).

6. Bown, N. et al.Gain of chromosome arm 17q and adverse outcome in patients

with neuroblastoma. N. Engl. J. Med.340, 19541961 (1999).

7. Attiyeh, E.F.et al.Chromosome 1p and 11q deletions and outcome in neuroblastoma.

N. Engl. J. Med.353, 22432253 (2005).

8. Seeger, R.C. et al.Association of multiple copies of the N-myconcogene with rapid

progression of neuroblastomas. N. Engl. J. Med.313, 11111116 (1985).

9. Caron, H. et al.Allelic loss of chromosome 1p as a predictor of unfavorable outcome

in patients with neuroblastoma. N. Engl. J. Med.334, 225230 (1996).

10. Cheung, N.K. et al.Association of age at diagnosis and genetic mutations in patients

with neuroblastoma. J. Am. Med. Assoc.307, 10621071 (2012).

11. Molenaar, J.J. et al.Sequencing of neuroblastoma identifies chromothripsis and

defects in neuritogenesis genes. Nature483, 589593 (2012).

12. Pugh, T.J. et al.The genetic landscape of high-risk neuroblastoma. Nat. Genet.45,

279284 (2013).
http://r2.amc.nl/http://evs.gs.washington.edu/EVS/http://www.nature.com/doifinder/10.1038/ng.3333http://www.nature.com/doifinder/10.1038/ng.3333https://www.ebi.ac.uk/ega/studies/EGAS00001001184https://www.ebi.ac.uk/ega/studies/EGAS00001001183http://www.ncbi.nlm.nih.gov/projects/gap/cgi-bin/study.cgi?study_id=phs000467.v10.p3http://www.ncbi.nlm.nih.gov/projects/gap/cgi-bin/study.cgi?study_id=phs000467.v10.p3http://www.ncbi.nlm.nih.gov/bioproject/?term=PRJNA282395http://www.nature.com/doifinder/10.1038/ng.3333http://www.nature.com/doifinder/10.1038/ng.3333http://www.nature.com/doifinder/10.1038/ng.3333http://www.nature.com/reprints/index.htmlhttp://www.nature.com/reprints/index.htmlhttp://www.nature.com/reprints/index.htmlhttp://www.nature.com/reprints/index.htmlhttp://www.nature.com/doifinder/10.1038/ng.3333http://www.nature.com/doifinder/10.1038/ng.3333http://www.nature.com/doifinder/10.1038/ng.3333http://www.ncbi.nlm.nih.gov/bioproject/?term=PRJNA282395http://www.ncbi.nlm.nih.gov/projects/gap/cgi-bin/study.cgi?study_id=phs000467.v10.p3https://www.ebi.ac.uk/ega/studies/EGAS00001001183https://www.ebi.ac.uk/ega/studies/EGAS00001001184http://www.nature.com/doifinder/10.1038/ng.3333http://www.nature.com/doifinder/10.1038/ng.3333http://evs.gs.washington.edu/EVS/http://r2.amc.nl/

7/18/2019 Eleveld NG 2015

8/10


ART I CLES

13. Sausen, M. et al. Integrated genomic analyses identify ARID1A and ARID1B

alterations in the childhood cancer neuroblastoma. Nat. Genet.45, 1217 (2013).

14. Taggart, D., Dubois, S. & Matthay, K.K. Radiolabeled metaiodobenzylguanidine for

imaging and therapy of neuroblastoma. Q. J. Nucl. Med. Mol. Imaging52, 403418

(2008).

15. Schleiermacher, G. et al. Emergence of new ALK mutations at relapse of

neuroblastoma. J. Clin. Oncol.32, 27272734 (2014).

16. Carter, S.L. et al. Absolute quantification of somatic DNA alterations in human

cancer. Nat. Biotechnol.30, 413421 (2012).

17. Boca, S.M., Kinzler, K.W., Velculescu, V.E., Vogelstein, B. & Parmigiani, G. Patient-

oriented gene set analysis for cancer mutation data. Genome Biol.11, R112 (2010).18. Futreal, P.A. et al.A census of human cancer genes. Nat. Rev. Cancer4, 177183

(2004).

19. Bresler, S.C. et al. ALK mutations confer differential oncogenic activation and

sensitivity to ALK inhibition therapy in neuroblastoma. Cancer Cell26, 682694

(2014).

20. Hlzel, M. et al. NF1 is a tumor suppressor in neuroblastoma that determines

retinoic acid response and disease outcome. Cell142, 218229 (2010).

21. Tartaglia, M. et al. Diversity and functional consequences of germline and

somatic PTPN11 mutations in human disease. Am. J. Hum. Genet.78, 279290

(2006).

22. Carr-Wilkinson, J. et al.High frequency of p53/MDM2/p14ARFpathway abnormalities

in relapsed neuroblastoma. Clin. Cancer Res.16, 11081118 (2010).

23. Garnett, M.J. et al.Systematic identification of genomic markers of drug sensitivity

in cancer cells. Nature483, 570575 (2012).

24. Schleiermacher, G. et al. Accumulation of segmental alterations determines

progression in neuroblastoma. J. Clin. Oncol.28, 31223130 (2010).

25. Landau, D.A. et al. Evolution and impact of subclonal mutations in chronic

lymphocytic leukemia. Cell152, 714726 (2013).

26. Long, G.V. Increased MAPK reactivation in early resistance to dabrafenib/trametinib

combination therapy of BRAF-mutant metastatic melanoma. Nat. Commun.5, 5694

(2014).27. George, R.E. et al. Activating mutations in ALK provide a therapeutic target in

neuroblastoma. Nature455, 975978 (2008).

28. Moss, Y.P. et al. Safety and activity of crizotinib for paediatric patients with

refractory solid tumours or anaplastic large-cell lymphoma: a Childrens Oncology

Group phase 1 consortium study. Lancet Oncol.14, 472480 (2013).

29. Berry, T. et al.The ALKF1174Lmutation potentiates the oncogenic activity of MYCN

in neuroblastoma. Cancer Cell22, 117130 (2012).

30. Moore, N.F. et al. Molecular rationale for the use of PI3K/AKT/mTOR pathway

inhibitors in combination with crizotinib in ALK-mutated neuroblastoma. Oncotarget

5, 87378749 (2014).

7/18/2019 Eleveld NG 2015

9/10

NATURE GENETICSdoi:10.1038/ng.3333

ONLINE METHODSSample collection and patient selection.The inclusion criteria for this studywere histopathological confirmation of neuroblastoma at original diagnosisand the presence of biopsy material from a subsequent relapse specimen.Patients were included in this study after obtaining informed consent fromparents or guardians, with oversight from the ethics committees Comitde Protection des Personnes Sud-Est IV, reference L07-95/L12-171, andComit de Protection des Personnes Ile-de-France, reference 0811728

in France, the review board at the Childrens Hospital of Philadelphia andreview boards at other Childrens Oncology Group sites that submitted samplesfor patients on this study in the United States, and the review board of theAcademic Medical Center of the University of Amsterdam in the Netherlands.Somatic ALK mutation status has been reported for samples FR_NB0175,FR_NB308, FR_NB399, FR_NB1224, FR_NB1269 and FR_NB13821 (ref.15).

Whole-genome sequencing, summary.Whole-genome sequencing was per-formed by Complete Genomics to an average coverage of 50 per sample (forDutch and US patient material and cell line material) 31 or using IlluminaHiSeq 2500 instruments to an average coverage of 80 per sample (for Frenchpatient material). Material for each patient (lymphocytes, primary tumorand relapse tumor) was in all cases sequenced using the same sequencingplatform. Details on sequencing results are given in Supplementary Table 1.Estimation of normal tissue contamination and whole-genome segmentation

was performed on all primary-normal and relapse-normal pairs with Sequenzasoftware v2.1.0 using both binned coverage ratio data and SNV allelic ratiosas input32. Coverage-based copy number plots were generated as previouslydescribed11, with the exception that values were corrected for ploidy and purityas determined by Sequenza. The R2 bioinformatics platform and Circos33wereused for analysis and visualization. Subclonal detection of variants is describedin the Supplementary Note.

Whole-genome sequencing, Complete Genomics.Potential somatic variantswere determined with the CallDif f algorithm with somatic output as availablein the CGAtools v1.3.0 package, maintained by Complete Genomics. Everytumor was compared to its matched blood sample across the genome. Thesomatic output files were then filtered to regions where coding sequencesare defined in the UCSC refflat. Silent mutations were subsequently removedfrom the analysis. In addition, we determined the presence of somatic splice-

site variants in the three bases surrounding exons as defined by UCSC refflatdata. Variants with a somatic score >0.05 (for NL_041, NL_571 and NL_N607)or Somatic Quality high (the remainder of the patients) were included inthe analysis.

Comparisons of structural variants between tumor and lymphocyte genomeswere performed with the JunctionDiff and Junction2Event algorithms fromCGAtools. These somatic events were filtered to remove events matching thefollowing criteria: events annotated as artifacts, footprints smaller than 70bases, less than 10 discordant mate pairs, under-represented repeats and pres-ence in a set of baseline genomes (as provided on the website of CompleteGenomics (B36baseline-junctions.tsv)). Of the remaining entries, we kept thefollowing events: (i) exon_bites, where both ends of a junction were within thesame gene and in addition affect an exonic sequence, (ii) breaks by inversion,where both ends of a junction land within a gene, thereby damaging bothgenes but leaving the genes in between unaffected, (iii) potential fusion genes,

which are strand matched, where both ends of a junction landed within a geneand the resulting end product fit in terms of orientation of both genes, and(iv) regions (deletions or (tandem) duplications) of up to 1 Mb, containingup to five genes.

Whole-genome sequencing, Illumina.Whole-genome sequencing was per-formed using an Illumina HiSeq 2500, with 90-bp paired-end reads for sixtumors and 100-bp paired-end reads for the remaining two tumors. Afteralignment with hg19 using Burrows-Wheeler Aligner (BWA)34, bam files werecleaned according to Genome Analysis Toolkit (GATK) recommendations35.

Variant calling was performed in parallel using three variant callers: GATK2.2-16, SAMtools 0.1.18 and MuTect 1.1.430 (refs. 3436). ANNOVAR v2012-10-23 (ref. 37) with COSMIC v64 and dbSNP v137 was used for annotation.SNVs with a quality under 30, a depth of coverage under 6 or with fewer than

2 reads supporting the variant were filtered out, as were variants reported inmore than 1% of the population in the 1000 Genomes Project38or ExomeSequencing Project (Exome Variant Server, National Heart, Lung, and BloodInstitute (NHLBI) GO Exome Sequencing Project (ESP)).

Variants were then filtered to those regions where coding sequences aredefined or to variants in the three bases surrounding exons. Silent mutationswere subsequently removed from the analysis. Tumor and corresponding con-stitutional genomes were compared using the SAMtools mpileup algorithm34,

and non-somatic variants were discarded from the analysis.Structural variants, including deletions, inversions, tandem duplicationsand translocations, were analyzed using DELLY v0.5.5 with standard param-eters39. In tumors, at least ten supporting reads were required to make a call,and five supporting reads were required for sample NB0175 with a cover-age of only 40. To predict structural variants in constitutional samples forsubsequent somatic filtering, only two supporting reads were required. Toidentify somatic events, all the structural variants in each normal sample werefirst flanked by 500 bp in both directions and any structural variant called ina tumor sample that was in the combined flanking regions of the respectivenormal sample was removed. Deletions with more than five genes affected orlarger than 1 Mb in size and inversions or tandem duplications covering morethan four genes were removed. We focused on exonic and splicing events fordeletions, inversions and tandem duplications. For translocations, we keptall structural variants that occurred in intronic, exonic, 5UTR, upstream or

splicing regions.

Clonality analysis. Estimation of the CCF of somatic mutations was per-formed using the Bayesian method of Carter et al.to infer posterior intervalswithout clustering for comparison25,40. Namely, we assumed that the expectedallele fraction of a mutation in a sample of tumor purity , total somatic copynumber qand mutation multiplicity scould be expressed as a function of theCCF c, wheref(c) = cs/(2(1 ) + q). Given a uniform prior on c, the pos-terior density of cis therefore proportional to Binom(a|N,f(c)), where ais the

variant read count and Nis the total read count. While and qare estimatedby Sequenza, the mutation multiplicity sis generally not known. However,under the parsimonious assumption that a mutation occurs only once withina tumors evolutionary history, we can bound sby 1 smq, where misthe major allelic copy number of the mutation locus, which is estimated bySequenza. We therefore modeled the posterior distribution under two assump-

tions: s= 1 (biased toward higher clonality estimates) and s= m(biased towardlower clonality estimates).

Estimation of CCF was also performed using PyClone v0.12.7 as an alterna-tive method for comparison41. For each primary or relapse tumor, PyClone wasrun on all somatic coding point mutations using the parental_copy_numbermethod and pyclone_beta_binomial density, with estimates of tumor purityand allelic copy number from Sequenza provided as necessary inputs. TheMarkov chain Monte Carlo (MCMC) step of PyClone was run for 10,000iterations with burn-in and thin parameters set to 1,000 and 10, respectively,resulting in 900 independent samples from the posterior distribution of CCFper mutation. Otherwise, default options for PyClone were used.

Cancer mutation analysis. To identify pathways or processes thatwere frequently affected in neuroblastoma relapse tumors, we used theCancerMutationAnalysis R package17. Somatic mutations detected only in

the relapse sample and detected in the relapse and primary samples for alltumors were used as input. This algorithm is not suitable for the analysisof structural variants, so these were not included. Pvalues were generatedusing the permutation null method without heterogeneity and signify theenrichment of mutated genes associated with a certain Gene Ontology (GO)Biological Process category across all relapse tumors.

Cell lines.All cell lines were cultured in DMEM or RPMI-1640 supplementedwith 10% FBS, 20 mM -glutamine, 10 U/ml penicillin and 10 g/ml strepto-mycin and maintained at 37 C under 5% CO2. Cell line identities are regularlyconfirmed by short tandem repeat (STR) profiling using the PowerPlex16system and GeneMapper software (Promega). Cell lines are regularly screenedfor the presence of mycoplasma. Inducible expression of RAS-MAPK mutantsis described in the Supplementary Note.

7/18/2019 Eleveld NG 2015

10/10

NATURE GENETICS doi:10.1038/ng.3333

Cell viability assays in response to MEK inhibition.For cell viability assays,2.525 103cells were seeded in 50 l in 96-well plates 1 d before treatmentwith one of three MEK inhibitors. Binimetinib, trametinib or cobimetinib(Selleckchem) was added in a seven-point fivefold dilution series, and cell

viability was assayed by MTT assay (Sigma) after 72 h, as describedpreviously42. All experiments were performed in triplicate, and values werecompared to those for solvent-treated controls.

GI50 values were calculated by 100 (T T0)/(C T0) for every drug

concentration, where T is the optical density for a certain drug concentra-tion at 72 h, T0 is the optical density at 0 h (before adding the drug) andCis the optical density of solvent-treated controls at 72 h. Curves were fittedon the data points using nonlinear regression in GraphPad 5 (log(inhibitor)

versus response variable slope), and GI50 values were interpolated fromthese curves. If curves did not reach 50% growth inhibition, the GI50valuewas set at 10 M.

Whole-exome sequencing of neuroblastoma cell lines. Whole exomesequencing of the neuroblastoma cell lines SK-N-AS, NBL-S, Kelly andIMR-5 was performed using in-solution hybrid capture43 followed byIllumina sequencing, as described previously12. Cell lines were obtainedfrom the Neuroblastoma Cell Line Repository at the Childrens Hospital ofPhiladelphia.

MEK inhibition in mouse xenotransplants. The human neuroblastoma-derived cell lines SK-N-AS, NBL-S, Kelly and IMR-5 were xenotransplantedsubcutaneously into female C.B-17 SCID mice at 57 weeks of age. Once theengrafted tumors reached 200 mm3, mice were treated orally with binimetinib(Novartis) at 3 mg/kg (n= 10) or 30 mg/kg (n= 10) or received vehicle only(n= 10) by simple randomization. Mice were treated twice daily, and tumorsize was monitored three times weekly; all investigators other than the mousetechnician were blinded to group allocation and study outcomes until all micecompleted a trial. Tumor burden was determined according to the formula

(/6)d3, where d represents the mean tumor diameter obtained by calipermeasurement. Statistical analysis was performed using a two-tailed t testat each time point, with Pvalues

Documents

Eleveld NG 2015