Whole-exome and targeted gene sequencing of gallbladder carcinoma identifies recurrent mutations in the ErbB pathway

Page 1

872 VOLUME 46 | NUMBER 8 | AUGUST 2014 Nature GeNetics

Individualswithgallbladdercarcinoma(GBC),themostaggressivemalignancyofthebiliarytract,haveapoorprognosis.HerewereporttheidentificationofsomaticmutationsforGBCin57tumor-normalpairsthroughacombinationofexomesequencingandultra-deepsequencingofcancer-relatedgenes.ThemutationpatternisdefinedbyadominantprevalenceofC>TmutationsatTCNsites.Geneswithasignificantfrequency(falsediscoveryrate(FDR)<0.05)ofnon-silentmutationsincludeTP53(47.1%),KRAS(7.8%)andERBB3(11.8%).Moreover,ErbBsignaling(includingEGFR,ERBB2,ERBB3,ERBB4andtheirdownstreamgenes)isthemostextensivelymutatedpathway,affecting36.8%(21/57)oftheGBCsamples.MultivariateanalysesfurthershowthatcaseswithErbBpathwaymutationshaveaworseoutcome(P=0.001).ThesefindingsprovideinsightintothesomaticmutationallandscapeinGBCandhighlightthekeyroleoftheErbBsignalingpathwayinGBCpathogenesis.

GBC is a rare neoplasm with an incidence of 2.5 in 100,000 indi-viduals1,2. Risk factors for its development include the presence of gallstones, chronic inflammation and the presence of anomalous

pancreatobiliary ductal junctions3. Individuals with GBC often do not show early symptoms and may receive late diagnoses and unsat-isfactory treatments, leading to poor prognosis. The median survival time for individuals with GBC is less than 1 year4. GBC cells are gen-erally highly metastatic, but little is known about their pathogenesis.

Whole-exome and targeted gene sequencing of gallbladder carcinoma identifies recurrent mutations in the ErbB pathwayMaolan Li1,2,13, Zhou Zhang1–3,13, Xiaoguang Li4,13, Junyi Ye5, Xiangsong Wu1,2, Zhujun Tan1,2, Chang Liu6, Baiyong Shen7, Xu-An Wang1,2, Wenguang Wu1,2, Daizhan Zhou3, Di Zhang3, Ting Wang3, Bingya Liu7, Kai Qu6, Qichen Ding1,2, Hao Weng1,2, Qian Ding1,2, Jiasheng Mu1,2, Yijun Shu1,2, Runfa Bao1,2, Yang Cao1,2, Peizhan Chen4, Tianyu Liu1,2, Lin Jiang1,2, Yunping Hu1,2, Ping Dong1,2, Jun Gu1,2, Wei Lu1,2, Weibin Shi1,2, Jianhua Lu1,2, Wei Gong1,2, Zhaohui Tang1,2, Yong Zhang1,2, Xuefeng Wang1,2, Y Eugene Chin8, Xiaoling Weng5, Hong Zhang5, Wei Tang9, Yonglan Zheng10, Lin He3,5,11, Hui Wang4,12,14, Yun Liu5,14 & Yingbin Liu1,2,14

1Department of General Surgery, Xinhua Hospital affiliated to Shanghai Jiao Tong University School of Medicine, Shanghai, China. 2Institute of Biliary Tract Disease, Shanghai Jiao Tong University School of Medicine, Shanghai, China. 3Key Laboratory for the Genetics of Developmental and Neuropsychiatric Disorders (Ministry of Education), Bio-X Center, Shanghai Jiao Tong University, Shanghai, China. 4Key Laboratory of Food Safety Research, Institute for Nutritional Sciences, Shanghai Institutes for Biological Sciences, Chinese Academy of Sciences, University of Chinese Academy of Sciences, Shanghai, China. 5Institutes of Biomedical Sciences, Fudan University, Shanghai, China. 6Department of Hepatobiliary Surgery, The First Affiliated Hospital of Medical College, Xi′an Jiaotong University, Xi’an, China. 7Department of General Surgery, Ruijin Hospital, Shanghai Jiao Tong University School of Medicine, Shanghai, China. 8Institute of Health Sciences, Shanghai Institutes for Biological Sciences, Chinese Academy of Sciences, Shanghai, China. 9Laboratory of Human Carcinogenesis, Center for Cancer Research, National Cancer Institute, US National Institutes of Health, Bethesda, Maryland, USA. 10Department of Medicine, The University of Chicago, Chicago, Illinois, USA. 11Women’s Hospital, Zhejiang University School of Medicine, Hangzhou, China. 12Key Laboratory of Food Safety Risk Assessment, Ministry of Health, Beijing, China. 13These authors contributed equally to this work. 14These authors jointly directed this work. Correspondence should be addressed to Yingbin Liu ([email protected]), Yun Liu ([email protected]) or H. Wang ([email protected]).

Received 23 March; accepted 11 June; published online 6 July 2014; doi:10.1038/ng.3030

A>C

A>G

A>T

C>A

C>G

C>T

0.200.15

0.100.05

Mut

atio

ns p

er M

b

0

A_AA_CA_GA_TC_AC_CC_GC_TG_AG_CG_GG_TT_AT_CT_GT_T

Figure 1 Somatic SNV signature in GBC. The 4,592 somatic SNVs (including 1,450 non-silent and 3,142 silent SNVs) found in 32 GBC exomes were divided into 96 subgroups defined by substitution class and adjacent bases. Each column around the circle represents a subgroup, and the height of each column shows mutation frequency per Mb.

l e t t e r snp

g©

2014

Nat

ure

Am

eric

a, In

c. A

ll rig

hts

rese

rved

.

Page 2

Nature GeNetics VOLUME 46 | NUMBER 8 | AUGUST 2014 873

Thus far, most studies have focused on a limited number of candidate genes2,5. A recent study analyzed exome sequence from nine GBC tumors and identified TP53 as a significantly mutated gene6. Knowledge of the spectrum of somatic mutations in GBC remains incomplete.

To profile the somatic mutation spectrum in GBCs, we performed whole-exome sequencing on 32 pairs of clinically and pathologi-cally characterized GBC tissues and case-matched normal tissues (Supplementary Fig. 1 and Supplementary Tables 1–3). Using strin-gent criteria, we identified 1,450 somatic single-nucleotide variations (SNVs) and 34 somatic insertions or deletions (indels) that were pre-dicted to alter protein-coding sequence (defined by high or moderate impact in SnpEff7; Supplementary Table 4). The overall mutation rate was 1.42 mutations per Mb (Supplementary Table 5).

The nucleotide mutation pattern can be indicative of specific muta-genesis mechanisms occurring in tumor cells8–10. Here we found that a C>T/G>A change was the most frequent substitution type, com-prising 41.6% of all somatic SNVs observed in GBC. Moreover, the trinucleotide signature was dominated by C>T mutations at TCN

sites, especially at TCA sites (Fig. 1). TCA is the optimum motif for APOBEC3B8,11. The APOBEC mutation pattern is also found in bladder, breast, cervical, head and neck, and lung cancers12. Its presence in GBC indicates the important roles of APOBEC cytidine deaminase mutagenesis in the pathogenesis of GBC.

There were 23 genes with 3 or more nonsynonymous somatic mutations, including TP53, CS, ERBB3 and ERBB2. However, only TP53 and CS were found to be statistically significant by MutSigCV10 (Supplementary Table 6). We also found somatic mutations enriched in the ErbB signaling pathway, the non-homologous end-joining pathway and other pathways associated with cancer cell signaling transduction, although none of these achieved statistical significance, as determined by MuSiC path-scan13 (Supplementary Table 7).

Because, in almost all types of cancer, recurrently mutated genes are restricted to a small panel of cancer-related genes10,14, we designed a cancer gene panel, including 283 candidate genes that were fre-quently mutated in our exome sequencing (such as TP53, CS, ERBB3 and ERBB2) or have frequently been implicated in multiple types

Modifier

Silent

Male

Negative

Negative

Neck

Female

Positive

Positive

Fundus/body

Non-silent

80

60

40

Mut

atio

n nu

mbe

r

20

0Sex

Lymph node metastasis

Gallbladder stonesTumor location

30 20 10Mutation number

0

47.1%

7.8%

11.8%

5.9%

9.8%

5.9%

3.9%

5.9%

3.9%

5.9%

3.9%

5.9%

5.9%

3.9%

3.9%

3.9%

3.9%

XH

DG

44X

HD

G67

XH

DG

46X

HD

G35

XH

DG

10X

HD

G19

XH

DG

21X

HD

G06

XH

DG

04X

HD

G45

XH

DG

58X

HD

G53

XH

DG

27X

HD

G17

XH

DG

08X

HD

G32

XH

DG

48X

HD

G47

XH

DG

55X

HD

G66

XH

DG

65X

HD

G22

XH

DG

52X

HD

G63

XH

DG

30X

HD

G59

XH

DG

38X

HD

G54

XH

DG

25X

HD

G50

XH

DG

51X

HD

G03

XH

DG

20X

HD

G34

XH

DG

37X

HD

G64

XH

DG

02X

HD

G05

XH

DG

33X

HD

G62

XH

DG

01X

HD

G07

XH

DG

09X

HD

G31

XH

DG

29X

HD

G31

XH

DG

31X

HD

G49

XH

DG

56X

HD

G60

XH

DG

61

3.9%

3.9%

3.9%

3.9%

3.9%

3.9%

2.0%

3.9%

3.9%

5.9%

5.9%

Modifier Synonymous Missense Splicing site Nonsense

TP53

KRAS

ERBB3

CDKN2A

ERBB2

BRAF

RNF43

PIK3CA

SF3B1

AKAP11

BRCA1

FBXW7

GNAS

SMARCB1

MAP2K4

CPNE4

POLE

GLTSCR1

NALCN

ARID1B

NF1

RB1

SMAD4

RGPD3

EGFR

FLG

CSMD3

LAMA2

0 2 4 6–log10 (P value)

8

>8

Figure 2 Significantly mutated genes in GBC. The top bar plot shows the somatic mutation frequency for tumors from each case. The bottom middle plot shows the mutation status of the recurrently mutated genes for each tumor. Somatic mutations are colored according to functional class. The bottom left plot shows the mutation count for each individual gene. The bottom right bar plot shows the significance of each gene; −log (P ) values are shown in light gray, and −log (q) values are shown in dark gray. The red vertical line represents a P value of 0.05.

l e t t e r snp

g©

2014

Nat

ure

Am

eric

a, In

c. A

ll rig

hts

rese

rved

.

Page 3

874 VOLUME 46 | NUMBER 8 | AUGUST 2014 Nature GeNetics

l e t t e r s

of cancer (Catalogue of Somatic Mutations in Cancer (COSMIC); Supplementary Table 8). These genes were sequenced in 51 pairs of GBC-normal tissue samples (25 of which overlapped with the tissues analyzed by whole-exome sequencing; Supplementary Fig. 2 and Supplementary Table 2) by ultra-deep targeted sequencing with a mean depth of 367× (Supplementary Table 9). We identified a total of 271 non-silent SNVs (Supplementary Tables 10 and 11), resulting in an overall mutation rate of 4.53 mutations per Mb. Three genes (TP53, KRAS and ERBB3) were significantly mutated in 51 GBC tumors (MutSigCV FDR < 0.05; Fig. 2 and Supplementary Table 12).

As expected, TP53 and KRAS were recurrently mutated, with muta-tion rates of 47.1% and 7.8%, respectively (Fig. 2 and Supplementary Fig. 3). Mutations of TP53 are common molecular changes, estimated to occur in over 50% of human tumors, including in GBC14. Activating mutations of the KRAS proto-oncogene have been investigated in GBC pathogenesis; most of these mutations occurred in codon 12 of the gene15. Similarly, we found four mutations of KRAS, all of which were located in codon 12 or 13 (encoding p.Gly12Arg, p.Gly12Val and p.Gly13Asp; Supplementary Fig. 3). These mutations lead to consist-ent activation of KRAS and incessant signals for cell growth16.

Notably, we found somatic mutations in ERBB3 (11.8%), ERBB2 (9.8%), EGFR (also known as ERBB1; 3.9%) and ERBB4 (3.9%) (Fig. 2). The ErbB receptors have been implicated in the development of

several types of tumors17,18, and EGFR and ERBB2 overexpression has been implicated in GBC progression19,20. However, activating mutations of ERBB3 and ERBB2 have not been described previously in GBC. Five of six mutations in ERBB3 and three of five mutations in ERBB2 affected the extracellular domain (ECD; Fig. 3a), as observed in other tumors17. Among the mutations affecting the ECD, those mapping to codon 104 of ERBB3, an established oncogenic mutation17, were most frequent (4/5), indicating that codon 104 is a mutational hotspot of ERBB3 in GBC (Fig. 3a). The ERBB2 mutations mostly affected subdomain II of the ECD, which is required for partnering with other ErbB family members21,22 (Fig. 3a). Mutations affecting subdomain II have been implicated in the failure of antibody-based ERBB2-targeted therapy23.

To determine the oncogenic effects of the ERBB3 and ERBB2 mutations, we overexpressed the mutants of ERBB3 (Val104Leu, Arg426Trp and Val1035Asp; Fig. 3b) and ERBB2 (Glu265Lys, Gly292Arg, Ser310Tyr and Val842Ile; Fig. 3c) in human gallbladder cancer cell lines (GBC-SD, NOZ and OCUG-1). Overexpression of each ERBB3 or ERBB2 mutant resulted in a significant increase in proliferation in at least one cell line compared with mock transfection or expression of wild-type ERBB3 or ERBB2 (P < 0.05; Fig. 3b,c and Supplementary Fig. 4). Considering that there are no immortalized gallbladder epithelial cells available as a research model, we used the

Ctr

lW

TV

104L

R42

6WV

1035

D

Ctr

lW

TV

104L

R42

6WV

1035

D

200

150

100*

**§

*** ** **

§ §§§§

***

Cel

l gro

wth

(%

of c

ontr

ol)

50

0

GBC-SDNOZ

Ctr

lW

TV

104L

R42

6WV

1035

D

Ctr

lW

TV

104L

R42

6WV

1035

D

200

ERBB3ERBB3 + ERBB2 WT

150

100 ** ****§§§

Cel

l gro

wth

(%

of c

ontr

ol)

50

0

*** *

§§

**§§**

§§

**§§

**§§

**§§

**§

Ctr

lW

TE

265K

G29

2RS

310Y

V84

2I

Ctr

lW

TE

265K

G29

2RS

310Y

V84

2I

200

150

100

Cel

l gro

wth

(%

of c

ontr

ol)

50

0

GBC-SDNOZ

§§

****

*

Ctr

lW

TE

265K

G29

2RS

310Y

V84

2I

Ctr

lW

TE

265K

G29

2RS

310Y

V84

2I

200

ERBB2ERBB2 + ERBB3 WT

150

100

Cel

l gro

wth

(%

of c

ontr

ol)

50

0

20

NH2

ERBB3

a

190

V104L

R426W V1035DV104LV104LV104M

I II III IV TKD SD

309 500 643 709 966 1,342

COOH

Extracellular domain Intracellular domain

NH2

ERBB2 E265KG292R

R678Q V842IS310Y

I II III IV TKD SD COOH

22 196Extracellular domain Intracellular domain

321 489 653 720 987 1,255

b

ERBB3 Ctr

l vec

tor

WT

V10

4L

R42

6W

V10

35D

Ctr

l vec

tor

WT

V10

4LR

426W

V10

35D

Flag-ERBB3

ERBB2

β-actin

GBC-SD NOZ

c

ERBB2 Ctr

l vec

tor

WT

G29

2RE

265K

S31

0Y

V84

2I

Ctr

l vec

tor

WT

G29

2RE

265K

S31

0Y

V84

2I

Myc-ERBB2

ERBB3

β-actin

GBC-SD NOZ

d

ERBB3 Ctr

l vec

tor

WT

V10

4L

R42

6W

V10

35D

Ctr

l vec

tor

WT

+ ERBB2 WT

V10

4L

R42

6W

V10

35D

Flag-ERBB3

ERBB2

β-actin

HEK293T

+ ERBB3 WT

HEK293T

e

ERBB2 Ctr

l vec

tor

WT

G29

2RE

265K

S31

0Y

V84

2I

Ctr

l vec

tor

WT

G29

2RE

265K

S31

0Y

V84

2I

ERBB2

Flag-ERBB3

β-actin

Cel

l gro

wth

(%

of c

ontr

ol)

700600500400300200100

00 2 4

Days

NOZ

******

**

6 Cel

l gro

wth

(%

of c

ontr

ol)

400

300

200

100

00 2 4

Days

OCUG-1

**

******

6

f

Cel

l gro

wth

(%

of c

ontr

ol)

700600500400300200100

00 2 4

Days

GBC-SD

***

6

Blank Ctrl siRNA ERBB3 siRNAERBB2 siRNA

Figure 3 Somatic alterations of ERBB3 and ERBB2 and their oncogenic effects on normal and GBC cells. (a) ERBB3 and ERBB2 somatic non-silent alterations (inverted triangles) are depicted over the affected protein domains. I–IV, subdomains of the extracellular domain; TKD, tyrosine kinase domain; SD, C-terminal signaling domain. (b,c) GBC cells (GBC-SD and NOZ) were transiently transfected with vector expressing wild-type (WT) or mutant ERBB3 (b) or wild-type or mutant ERBB2 (c). The expression of relevant proteins was assessed by protein blotting 72 h after transduction (top), and cell viability was detected by thiazolyl blue tetrazolium bromide (MTT) assay 96 h after transduction (bottom). Ctrl, pCMV6 control vector. (d,e) Non-malignant HEK293T cells were transfected with vector expressing wild-type or mutant ERBB3 (d) or wild-type or mutant ERBB2 (e), either alone or in combination with wild-type ERBB2 or ERBB3, and subjected to protein blotting analysis (top) and MTT assay (bottom). Cell proliferation assays were repeated three times in GBC-SD and NOZ cells and six times in HEK293T cells. Data represent mean ± s.e.m. (*P < 0.05, **P < 0.01 with respect to cells transfected with the control vector; §P < 0.05, §§P < 0.01 with respect to cells expressing wild-type ERBB3 or ERBB2, as indicated). (f) Evaluation of cell viability in response to small interfering RNA (siRNA) for ERBB2 and ERBB3 in three types of GBC cells (GBC-SD, NOZ and OCUG-1) at the indicated times after transfection. The data shown are representative of values from three independent experiments (mean ± s.e.m.; *P < 0.05, **P < 0.01 relative to the control siRNA group). Throughout the figure, unpaired Student’s t tests were used to calculate all P values.

npg

© 2

014

Nat

ure

Am

eric

a, In

c. A

ll rig

hts

rese

rved

.

Page 4

Nature GeNetics VOLUME 46 | NUMBER 8 | AUGUST 2014 875

l e t t e r s

non-malignant HEK293T cell line (ERBB2 and ERBB3 non-expressing) as a normal cell type to further confirm the oncogenic effects of the ErbB alterations. We found that expression of ERBB3 alone had no effect on cell proliferation (Fig. 3d). However, expression of ERBB3 mutants in combination with wild-type ERBB2 significantly promoted cell proliferation in HEK293T cells (Fig. 3d), consistent with ERBB3 preferentially forming heterodimers with ERBB2 to acquire signaling potential, owing to its defective kinase domain24. Similarly, expression of wild-type or mutant ERBB2 resulted in a slight increase in the proliferation of HEK293T cells, and the effect was significantly enhanced when coexpressing ERBB2 mutants and wild-type ERBB3 (Fig. 3e). Next, we used RNA interference (RNAi) to mediate loss of function and thereby further confirm the oncogenic behaviors of the ErbB family members in the GBC cell lines (Supplementary Fig. 5). Notably, silencing of EGFR, ERBB2 and ERBB3 inhibited the growth of four lines of GBC cells in a time-dependent manner (Fig. 3f and Supplementary Fig. 6) and impaired cell motility (Supplementary Fig. 7). ERBB4 was not investigated because of its low expression in these cell lines. Moreover, in GBC, mutations of genes in the ErbB family were mutually exclu-sive (Fig. 4a). These data show the oncogenic potential of the ErbB mutants, and the presence of these activating ErbB alterations high-lights the importance of ErbB family receptors in GBC development and progression.

We also found non-silent mutations in 25 genes recurrently mutated in GBC, including in CDKN2A (5.9%) and BRAF (5.9%), although the mutation rates were only marginally significant (MutSigCV P < 0.05; Fig. 2 and Supplementary Table 12). Intriguingly, the majority of the recurrently mutated genes in GBC did not overlap with previously reported genetic lesions such as those affecting chromatin-remodeling genes, including BAP1, ARID1A and PBRM1, in cholangiocarcino-mas6. These results suggest that, although both tumor cells originate from the biliary epithelium, they are genetically distinct.

We looked for enrichment of somatic mutations in the core canoni-cal pathways in GBC through Genome MuSiC path-scan analysis13. The ErbB signaling pathway was identified as the most significantly mutated pathway (FDR of ~9.2 × 10−14; Supplementary Table 13). Gene expression studies previously implicated this pathway in GBC pathogenesis19,20,25,26.

Combining the data from whole-exome and targeted sequencing, which resulted in a total of 57 samples, we found that 36.8% of GBCs (21/57) had protein sequence–altering mutations in the ErbB pathway. In total, we detected 37 non-silent somatic mutations of 15 genes in the ErbB pathway (ERBB3, ERBB2, ERBB4, EGFR, KRAS, NRAS, HRAS, PIK3CA, BRAF, MAP2K4, MAPK10, SRC, MYC, NRG1 and SOS2) in GBC (Fig. 4a,b). Most of the mutations in the pathway were mutually exclusive, although some were not, especially mutations

in KRAS, BRAF and PIK3CA. These are well-known oncogenes that are involved in multiple cancer-related pathways. Because of the inclusion of these genes, the statistical significance of the ErbB pathway could be overestimated.

In assessing the relationship of these mutations to the clinicopatho-logical characteristics of individuals with GBC, we found a slightly higher rate of ErbB pathway mutations (P = 0.049) in tumors located on the gallbladder neck, but none of the other characteristics were sig-nificantly associated (Supplementary Table 14). Moreover, a striking difference in overall survival time was observed between cases with mutations in the ErbB pathway and those without: cases with ErbB pathway mutations had lower overall survival rates (median of 8 months versus 13 months; P = 0.012; Fig. 4c and Supplementary Table 15). To obtain a more precise estimate for prognosis, we used a multi-variate Cox proportional hazards model. As expected, tumor-node- metastasis (TNM) stage was negatively correlated with post-operative survival time (P = 0.035; Supplementary Table 16). Besides TNM stage, we observed a tenfold increase in mortality risk for cases with ErbB pathway mutations (hazard ratio = 10.4, 95% confidence interval = 1.9–14.1; P = 0.001; Supplementary Table 16), suggesting that these mutations were associated with a clinical prognosis.

Overall, the present study identified in a systematic way the somatic mutation framework of GBC, discovered a high rate of C>T/G>A transitions in GBC, and elucidated GBC driver genes and a core pathway (ErbB signaling). Of note, mutations in genes in the ErbB pathway were associated with poor prognosis in individuals with GBC. These results suggest that patients harboring mutations in the ErbB pathway might benefit from targeted therapies presently available or in development.

URLs. COSMIC, http://cancer.sanger.ac.uk/cancergenome/projects/cosmic/; Genomics of Drug Sensitivity in Cancer, http://www.cancer-rxgene.org/.

b 3.9%ERBB1 (EGFR)

9.8%ERBB2 (HER2)

11.8%ERBB3

ErbB family

P P P P

7.8%2.0%

2.0%2.0%3.1% 5.9%

5.9% 3.9%

3.9%

COX2

MAPK10

MAP2K4BRAF FAK AKT

PIK3CASRCSOS2NRAS

HRAS

KRAS

SHCGRB2

MYC STATs

Regulation of gene transcription

Proliferation Survival Angiogenesis Invasion

Nucleus

6.3%

3.9%ERBB4

NRG13.1%

c

P = 0.012

Wild type

0.8

1.0

0.6

0.4

0.2

0

0 10.00 20.00

Months after surgery

Ove

rall

surv

ival

30.00 40.00

MutantWild type, censoredMutant, censored

a Samples (21/57)ERBB3

EGFRKRAS

PIK3CAMAPK10

MYC

HRASNRG1BRAF

NRASSOS2

Silent Missense Nonsense

MAP2K4

SRC

ERBB4ERBB2

Figure 4 Somatic mutations of the ErbB signaling pathway in GBC. (a) Shown are the mutation status of genes of the ErbB pathway in the 21 GBCs that carry at least 1 non-silent mutation. Mutations in the ErbB family are mutually exclusive. (b) The key genes of the ErbB signaling pathway with mutation frequencies in GBC are shown. The frequencies for EGFR, ERBB2, ERBB3, ERBB4, SRC, KRAS, HRAS, NRAS, BRAF, PIK3CA, MAP2K4 and MYC were estimated from 51 samples analyzed by targeted sequencing. The frequencies for NRG1, SOS2 and MAPK10 were estimated from 32 samples analyzed by exome sequencing. (c) The impact of mutations in the ErbB pathway on clinical outcome is shown. Cases with mutations in the ErbB pathway (n = 21) demonstrate worse overall survival than cases without mutations in the ErbB pathway (n = 36).

npg

© 2

014

Nat

ure

Am

eric

a, In

c. A

ll rig

hts

rese

rved

.

Page 5

876 VOLUME 46 | NUMBER 8 | AUGUST 2014 Nature GeNetics

METHoDsMethods and any associated references are available in the online version of the paper.

Accession codes. Sequence data are available at the European Genome-phenome Archive (EGA), which is hosted by the European Bioinformatics Institute (EBI), under accession number EGAS00001000853.

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

ACKnoWLEDGMEnTSThis study was supported by the National Natural Science Foundation of China (81172026, 81272402, 81301816, 81172029, 81370728, 81125020, 81328022 and 81302507), the National High-Technology Research and Development Program (863 Program, 2012AA022606; 2012BAK01B00), the Foundation for Interdisciplinary Research of Shanghai Jiao Tong University (YG2011ZD07), the Shanghai Science and Technology Commission Intergovernmental International Cooperation Project (12410705900), the Shanghai Science and Technology Commission Medical-Guiding Project (12401905800), the China Postdoctoral Science Foundation (2013M541513), the Program for Changjiang Scholars and the Leading Talent program of Shanghai.

AUTHoR ConTRIBUTIonSH. Wang, Yun Liu and Yingbin Liu conceived the study. C.L., B.S., B.L., Y.E.C., L.H., H. Wang, Yun Liu and Yingbin Liu directed the study. M.L., Z.Z., X.L., H. Wang, Yun Liu and Yingbin Liu contributed to the project design. M.L., X.L., D. Zhou, T.W., X. Wu, X.-A.W. and Qichen Ding performed experiments. Z.Z., J.Y., D. Zhang, X. Weng and H.Z. performed bioinformatics data analysis. W.W., K.Q., H. Weng, Qian Ding, P.C., T.L., Y.H. and W.L. contributed samples, data and comments on the manuscript. M.L., Z.Z., X.L., Z. Tan, J.M., W.G., W.T. and Y. Zheng analyzed and interpreted data. Y.S., R.B., Y.C., L.J., P.D., J.G., W.S., J.L., Z. Tang, Y. Zhang and X. Wang contributed reagents, materials and/or analysis tools. M.L., Z.Z. and X.L. wrote the manuscript.

CoMPETInG FInAnCIAL InTERESTSThe authors declare no competing financial interests.

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

1. Randi, G., Franceschi, S. & La Vecchia, C. Gallbladder cancer worldwide: geographical distribution and risk factors. Int. J. Cancer 118, 1591–1602 (2006).

2. Srivastava, K., Srivastava, A., Sharma, K.L. & Mittal, B. Candidate gene studies in gallbladder cancer: a systematic review and meta-analysis. Mutat. Res. 728, 67–79 (2011).

3. Wolpin, B.M. & Mayer, R.J. A step forward in the treatment of advanced biliary tract cancer. N. Engl. J. Med. 362, 1335–1337 (2010).

4. Boutros, C., Gary, M., Baldwin, K. & Somasundar, P. Gallbladder cancer: past, present and an uncertain future. Surg. Oncol. 21, e183–e191 (2012).

5. Maurya, S.K., Tewari, M., Mishra, R.R. & Shukla, H.S. Genetic aberrations in gallbladder cancer. Surg. Oncol. 21, 37–43 (2012).

6. Jiao, Y. et al. Exome sequencing identifies frequent inactivating mutations in BAP1, ARID1A and PBRM1 in intrahepatic cholangiocarcinomas. Nat. Genet. 45, 1470–1473 (2013).

7. Cingolani, P. et al. A program for annotating and predicting the effects of single nucleotide polymorphisms, SnpEff: SNPs in the genome of Drosophila melanogaster strain w1118; iso-2; iso-3. Fly (Austin) 6, 80–92 (2012).

8. Alexandrov, L.B. et al. Signatures of mutational processes in human cancer. Nature 500, 415–421 (2013).

9. Li, M. & Liu, Y. The applications of exome sequencing in the study of gastrointestinal cancer. Chin. J. Pract. Surg. 33, 414–416 (2013).

10. Lawrence, M.S. et al. Mutational heterogeneity in cancer and the search for new cancer-associated genes. Nature 499, 214–218 (2013).

11. Roberts, S.A. et al. An APOBEC cytidine deaminase mutagenesis pattern is widespread in human cancers. Nat. Genet. 45, 970–976 (2013).

12. Kuong, K.J. & Loeb, L.A. APOBEC3B mutagenesis in cancer. Nat. Genet. 45, 964–965 (2013).

13. Dees, N.D. et al. MuSiC: identifying mutational significance in cancer genomes. Genome Res. 22, 1589–1598 (2012).

14. Wong, K.M., Hudson, T.J. & McPherson, J.D. Unraveling the genetics of cancer: genome sequencing and beyond. Annu. Rev. Genomics Hum. Genet. 12, 407–430 (2011).

15. Wistuba, I.I. & Gazdar, A.F. Gallbladder cancer: lessons from a rare tumour. Nat. Rev. Cancer 4, 695–706 (2004).

16. Croce, C.M. Oncogenes and cancer. N. Engl. J. Med. 358, 502–511 (2008).17. Jaiswal, B.S. et al. Oncogenic ERBB3 mutations in human cancers. Cancer Cell

23, 603–617 (2013).18. Desai, M.D., Saroya, B.S. & Lockhart, A.C. Investigational therapies targeting the

ErbB (EGFR, HER2, HER3, HER4) family in GI cancers. Expert Opin. Investig. Drugs 22, 341–356 (2013).

19. Nakazawa, K. et al. Amplification and overexpression of c-erbB-2, epidermal growth factor receptor, and c-met in biliary tract cancers. J. Pathol. 206, 356–365 (2005).

20. Kiguchi, K. et al. Constitutive expression of ErbB-2 in gallbladder epithelium results in development of adenocarcinoma. Cancer Res. 61, 6971–6976 (2001).

21. Baselga, J. & Swain, S.M. Novel anticancer targets: revisiting ERBB2 and discovering ERBB3. Nat. Rev. Cancer 9, 463–475 (2009).

22. Sliwkowski, M.X. Ready to partner. Nat. Struct. Biol. 10, 158–159 (2003).23. Franklin, M.C. et al. Insights into ErbB signaling from the structure of the

ErbB2-pertuzumab complex. Cancer Cell 5, 317–328 (2004).24. Macdonald-Obermann, J.L., Adak, S., Landgraf, R., Piwnica-Worms, D. & Pike, L.J.

Dynamic analysis of the epidermal growth factor (EGF) receptor–ErbB2-ErbB3 protein network by luciferase fragment complementation imaging. J. Biol. Chem. 288, 30773–30784 (2013).

25. Pignochino, Y. et al. Targeting EGFR/HER2 pathways enhances the antiproliferative effect of gemcitabine in biliary tract and gallbladder carcinomas. BMC Cancer 10, 631 (2010).

26. Goldin, R.D. & Roa, J.C. Gallbladder cancer: a morphological and molecular update. Histopathology 55, 218–229 (2009).

l e t t e r snp

g©

2014

Nat

ure

Am

eric

a, In

c. A

ll rig

hts

rese

rved

.

Page 6

Nature GeNeticsdoi:10.1038/ng.3030

oNLINEMETHoDsStudy subjects. The study population consisted of 57 Chinese subjects (32 recruited for whole-exome sequencing and 51 recruited for targeted deep sequencing; Supplementary Fig. 2a and Supplementary Tables 1 and 2) with GBC who underwent cholecystectomy without previous radiotherapy or chemotherapy from 2008 to 2012. Of the subjects, 23 were male (40.4%), and the mean age was 60.7 years (range of 37 to 86 years). The histopathological subtypes of the 57 adenocarcinomas included 5 well-differentiated, 36 moder-ately differentiated and 16 poorly differentiated adenocarcinomas. According to the American Joint Committee on Cancer (AJCC 7th Edition) staging system, 3 (5.3%), 30 (52.6%), 16 (28.1%) and 8 (14.0%) cases were diagnosed as stage II, IIIA, IIIB and IVB disease, respectively. Tissue samples were bisected, with half of the tissue stored in liquid nitrogen and half retained for histological confirmation. All diagnoses of GBC were confirmed by histopathological examination, and adjacent control samples were confirmed to be free of tumor cells. DNA samples isolated from GBCs with more than 60% tumor content and their matched normal tissues were subjected to exome sequenc-ing and targeted sequencing. All subjects provided written informed consent for genetic analysis under a protocol approved by the Ethics Committee of Xinhua Hospital, affiliated with the Shanghai Jiao Tong University School of Medicine.

Whole-exome sequencing. Genomic DNA libraries were prepared using pro-tocols recommended by Illumina. Whole-exome enrichment was performed using the TruSeq Exome Enrichment kit (Illumina). Captured DNA libraries were sequenced with the Illumina HiSeq 2500 Genome Analyzer, yielding 200 (2 × 100) base pairs from the final library fragments.

Ultra-deep targeted gene sequencing. Ultra-deep targeted sequencing included 283 ‘tumor-related genes’ in the target enrichment panel. These genes were selected according to the following criteria: (i) genes recurrently mutated in the 32 GBC exomes; (ii) genes identified as high priority in the COSMIC database; (iii) genes related to drug sensitivity; and (iv) genes screened in our study of gastrointestinal cancer exome sequencing (Yun Liu, D. Zhou, Z.Z., Yingbin Liu & K. Zhao et. al., unpublished data and Supplementary Table 8). Targeted gene enrichment was performed with the TruSeq Custom Enrichment kits (Illumina).

Next-generation sequencing data processing and mutation calling. Sequencing reads were trimmed and filtered with Trimmomatic27. The result-ing reads were aligned to the hg19 reference genome using Burrows-Wheeler Aligner (BWA)28, and the Genome Analysis Toolkit (GATK)29 was used for base quality score recalibration, indel realignment and duplicate removal. The MuTect30 algorithm was used to identify somatic SNVs in whole-exome and targeted gene sequencing data. MuTect identifies candidate somatic SNVs by Bayesian statistical analysis of bases and their qualities in the tumor and normal BAM files at a given genomic locus. We required a minimum of ten reads covering a site in the tumor and eight reads in the normal sample to declare that a site was adequately covered for mutation calling. Default param-eters were used for exome data. To adapt for the ultra-high depth of targeted sequencing, we set maximum alternative alleles in normal count at 10 and maximum alternative alleles in normal fraction at 0.05. To filter out likely false positives, we performed a further filtration step requiring that the alternative allele proportion in the normal sample be less than 0.30 of that in the tumor sample. A total of 26 pairs of GBC samples were processed using both whole-exome sequencing and targeted sequencing. Of the exome SNVs, 76.3% (71 of 93) were validated by ultra-deep targeted sequencing. We detected 22 and 96 mutations only by exome or targeted sequencing (Supplementary Fig. 2). The Pindel31 algorithm was used to identify indels with default parameters. We excluded inversions and large indels (>100 bp), as they were difficult to verify, and then used the following criteria to filter indels: (i) depth in tumor tissue of ≥10; (ii) depth in normal tissue of ≥8; (iii) alternative allele in normal tissue of 0; and (iv) alternative allele in tumor tissue of ≥5 (only applied in targeted sequencing). Somatic SNV and indel results were then combined and compared to the COSMIC database. Mutation functions were predicted using SnpEff7, PolyPhen32, PROVEAN33 and SIFT33. The four algorithms showed little difference in predicting non-silent mutations (altering protein sequence) but showed considerable differences in predicting the effects of amino acid

changes. To maintain data integrity, we retained all the non-silent mutations predicted by SnpEff and did not perform further filtering.

Gene and pathway mutational significance analysis. All somatic mutations were stored in a MAF file and analyzed by MutSigCV10, with default covari-ates tables, to calculate gene mutational significance. MutSigCV is designed to identify cancer driver genes and is able to eliminate most of the apparent false positive genes by incorporating mutational heterogeneity into the analyses. Genes with q (FDR) ≤ 0.1 were considered to be significantly mutated. To identify known cellular pathways with significant accretions with somatic mutations in GBC, we performed path-scan analyses34 in the MuSiC suite13. Path-scan tests for the scenario where pathway mutations collectively contribute to tumor development. We used the canonical pathways defined by the Kyoto Encyclopedia of Genes and Genomes (KEGG) database. Because of the overwhelming abundance of TP53 mutations, we performed the analysis using default parameters but excluding TP53 mutations.

Sanger sequencing validation. A total of 111 somatic SNVs and indels identified by next-generation sequencing were verified by Sanger sequencing. Excluding the loci that failed in PCR, 81 of 90 SNVs (90%) and 7 of 9 indels (78%) were successfully validated. For exome sequencing, 42 of 47 SNVs (89%) were validated; for targeted sequencing, 64 of 68 SNVs (94%) were validated (Supplementary Table 17).

Cell culture and reagents. The human GBC cell lines GBC-SD and SGC-996 were purchased from the Shanghai Institutes for Biological Sciences (Shanghai, China). The OCUG-1 and NOZ cell lines were obtained from the Health Science Research Resources Bank (Osaka, Japan). HEK293T cells were obtained from the American Type Culture Collection (ATCC). SGC-996 cells were cultured in RPMI-1640 medium, GBC-SD and OCUG-1 cells were cultured in DMEM, and NOZ cells were cultured in William’s medium E, with all media containing 10% FBS, 100 units/ml penicillin and 100 mg/ml streptomycin. Cells were incubated at 37 °C with 5% CO2 and tested negative for mycoplasma infection by RT-PCR. The primary antibodies against ERBB2 (2242), ERBB3 (4754), Myc tag (2276) and Flag tag (2368) were all obtained from Cell Signaling Technology and used at a dilution of 1:1,000. Thiazolyl blue tetrazolium bromide (M5655, MTT) was purchased from Sigma.

ErbB mutation scanning in GBC cell lines. ErbB family members were sequenced in four GBC cell lines (GBC-SD, NOZ, OCUG-1 and SGC-996). After known SNPs were removed, only the NOZ cell line was found to carry three nonsynonymous variants in the ErbB family (encoding p.Val524Ile in EGFR, p.Ser1050Leu in ERBB2 and p.Lys881Gln in ERBB4). None of these alterations overlapped with the somatic changes found in GBC cases or in the COSMIC database.

Effect of loss of function mediated by RNA interference on cell viability and migration. For siRNA-mediated silencing of ErbB receptors, we used the targeting siRNA sequences listed in Supplementary Table 18. The scrambled sequence served as a negative control. These double-stranded siRNA duplexes were synthesized by the Genepharma Company. To determine cell viability, 2,000–3,000 cells per well were plated into 96-well plates and later transfected with targeted siRNA using the Lipofectamine RNAiMAX assay according to the manufacturer’s instructions. The MTT assay was performed to determine cell viability at days 0, 2, 4 and 6 after transfection. Cell viability was assessed by the absorbance of the resultant purple formazan product at 490 nm. For cell migration, cells were transfected with siRNA against a targeted gene and, 48 h later, were plated onto Transwell Permeable Support inserts with an 8-µm micro-porous membrane (Corning Costar). Culture medium containing 10% FBS was used as a chemoattractant in the lower compartment. Within 6–8 h, cells that migrated to the bottom membranes of the Transwells were stained and counted in five random 100× fields. The experiments were repeated at least three times.

Cloning and transduction of mutant or wild-type ERBB2 and ERBB3 into cell lines. ERBB2 or ERBB3 somatic mutations were introduced into the PCMV6 vector bearing the ERBB2 or ERBB3 ORF and encoding a C-terminal fusion of Myc/DDK tag using the Fast Mutagenesis system kit (Beijing TransGen Biotech). The constructs obtained were verified using

npg

© 2

014

Nat

ure

Am

eric

a, In

c. A

ll rig

hts

rese

rved

.

Page 7

Nature GeNetics doi:10.1038/ng.3030

Sanger sequencing and were transfected into GBC cells or HEK293T cells using FuGENE HD (Promega). Transgene expression was verified by protein blot analysis for ERBB2 or ERBB3. Cell viability was assessed 96 h after transduction, as described above.

Statistical analyses. Statistical analyses were conducted with SPSS software version 18.0 (SPSS, Inc.). The association of mutations with clinical character-istics was analyzed by the χ2 test (Fisher’s exact test for low cell counts). The Kaplan-Meier test was used for univariate survival analysis. Multivariate Cox proportional hazards models were used to calculate the hazard ratios and 95% confidence intervals of the associations between risk factors and survival. The two-tailed Student’s t test was used to evaluate cell viability and migration with at least three independent experiments. Results were considered statistically significant with P < 0.05.

27. Lohse, M. et al. RobiNA: a user-friendly, integrated software solution for RNA-Seq–based transcriptomics. Nucleic Acids Res. 40, W622–W627 (2012).

28. Li, H. & Durbin, R. Fast and accurate short read alignment with Burrows-Wheeler transform. Bioinformatics 25, 1754–1760 (2009).

29. McKenna, A. et al. The Genome Analysis Toolkit: a MapReduce framework for analyzing next-generation DNA sequencing data. Genome Res. 20, 1297–1303 (2010).

30. Cibulskis, K. et al. Sensitive detection of somatic point mutations in impure and heterogeneous cancer samples. Nat. Biotechnol. 31, 213–219 (2013).

31. Ye, K., Schulz, M.H., Long, Q., Apweiler, R. & Ning, Z. Pindel: a pattern growth approach to detect break points of large deletions and medium sized insertions from paired-end short reads. Bioinformatics 25, 2865–2871 (2009).

32. Adzhubei, I.A. et al. A method and server for predicting damaging missense mutations. Nat. Methods 7, 248–249 (2010).

33. Choi, Y., Sims, G.E., Murphy, S., Miller, J.R. & Chan, A.P. Predicting the functional effect of amino acid substitutions and indels. PLoS ONE 7, e46688 (2012).

34. Wendl, M.C. et al. PathScan: a tool for discerning mutational significance in groups of putative cancer genes. Bioinformatics 27, 1595–1602 (2011).

npg

© 2

014

Nat

ure

Am

eric

a, In

c. A

ll rig

hts

rese

rved

.