ampullary cancers harbor elf3 tumor suppressor gene
TRANSCRIPT
Article
Ampullary Cancers Harbor
ELF3 Tumor SuppressorGene Mutations and Exhibit Frequent WNTDysregulationGraphical Abstract
Highlights
d Three periampullary tumor types share a common molecular
blueprint
d Frequent WNT pathway disruption can be a potential
therapeutic target
d ELF3 is a frequently mutated tumor suppressor gene of
periampullary tumors
Gingras et al., 2016, Cell Reports 14, 907–919February 2, 2016 ª2016 The Authorshttp://dx.doi.org/10.1016/j.celrep.2015.12.005
Authors
Marie-ClaudeGingras, Kyle R. Covington,
David K. Chang, ..., Andrew V. Biankin,
David A. Wheeler, Richard A. Gibbs
[email protected] (M.-C.G.),[email protected] (D.A.W.)
In Brief
Gingras et al. compare the genomic
profiles of ampullary, distal bile duct, and
duodenal adenocarcinomas. They find
disruption of the WNT pathway, which
could be used as a tumor
subclassification for targeted therapy, a
high frequency of ELF3 inactivating
mutations, and microsatellite instability.
Cell Reports
Article
Ampullary Cancers Harbor ELF3 Tumor SuppressorGeneMutations andExhibit FrequentWNTDysregulationMarie-Claude Gingras,1,2,3,* Kyle R. Covington,1 David K. Chang,4,5,6,7 Lawrence A. Donehower,1,8 Anthony J. Gill,6,9,10
Michael M. Ittmann,3,11,12 Chad J. Creighton,1,3 Amber L. Johns,6 Eve Shinbrot,1 Ninad Dewal,1 William E. Fisher,2,3,13
Australian Pancreatic Cancer Genome Initiative, Christian Pilarsky,14 Robert Gr€utzmann,15 Michael J. Overman,16
Nigel B. Jamieson,4,5,17 George Van Buren II,2,3,13 Jennifer Drummond,1 Kimberly Walker,1 Oliver A. Hampton,1 Liu Xi,1
Donna M. Muzny,1 Harsha Doddapaneni,1 Sandra L. Lee,1 Michelle Bellair,1 Jianhong Hu,1 Yi Han,1 Huyen H. Dinh,1
Mike Dahdouli,1 Jaswinder S. Samra,10,18 Peter Bailey,4 Nicola Waddell,19,20 John V. Pearson,19,20 Ivon Harliwong,19
Huamin Wang,21 Daniela Aust,22 Karin A. Oien,4,23 Ralph H. Hruban,24 Sally E. Hodges,2,13 Amy McElhany,2,13
Charupong Saengboonmee,1,25 Fraser R. Duthie,4,23 Sean M. Grimmond,4,19 Andrew V. Biankin,4,5,6,7
David A. Wheeler,1,3,* and Richard A. Gibbs11Department of Molecular and Human Genetics, Human Genome Sequencing Center, Baylor College of Medicine, Houston,
TX 77030, USA2Michael DeBakey Department of Surgery, Baylor College of Medicine, Houston, TX 77030, USA3Dan L. Duncan Cancer Center, Baylor College of Medicine, Houston, TX 77030, USA4Wolfson Wohl Cancer Research Centre, Institute for Cancer Sciences, University of Glasgow, Garscube Estate, Bearsden,
Glasgow G61 1BD, UK5West of Scotland Pancreatic Unit, Glasgow Royal Infirmary, Glasgow G31 2ER, UK6The Kinghorn Cancer Centre and the Cancer Research Program Garvan Institute of Medical Research, Darlinghurst, Sydney, NSW 2010,Australia7South Western Sydney Clinical School, Faculty of Medicine, University of New South Wales, Liverpool, NSW 2170, Australia8Department of Molecular Virology and Microbiology, Baylor College of Medicine, Houston, TX 77030, USA9Department of Anatomical Pathology, Royal North Shore Hospital, St Leonards, Sydney, NSW 2065, Australia10Sydney Medical School, University of Sydney, Sydney, NSW 2006, Australia11Department of Pathology and Immunology, Baylor College of Medicine, Houston, TX 77030, USA12Michael E. DeBakey Department of Veterans Affairs Medical Center, Houston, TX 77030, USA13The Elkins Pancreas Center at Baylor College of Medicine, Houston, TX 77030, USA14Department of Surgery, TU Dresden, 01307 Dresden, Germany15Department of Surgery, Universitatsklinikum Erlangen, 91054 Erlangen, Germany16Department of Gastrointestinal Medical Oncology, University of Texas MD Anderson Cancer Center, Houston, TX 77030, USA17Academic Unit of Surgery, Institute of Cancer Sciences, Glasgow Royal Infirmary, Level 2, New Lister Building, University of Glasgow,
Glasgow G31 2ER, UK18Department of Surgery, Royal North Shore Hospital, St Leonards, Sydney, NSW 2065, Australia19Queensland Centre for Medical Genomics, Institute for Molecular Bioscience, The University of Queensland, St Lucia, Brisbane, QLD 4072,Australia20QIMR Berghofer Medical Research Institute, Herston, Brisbane, QLD 4006, Australia21Department of Pathology, University of Texas MD Anderson Cancer Center, Houston, TX 77030, USA22Department of Pathology, TU Dresden, 01307 Dresden, Germany23Department of Pathology, Southern General Hospital, Greater Glasgow and Clyde NHS, Glasgow G51 4TF, UK24Department of Pathology, The Sol Goldman Pancreatic Cancer Research Center, the Johns Hopkins University School of Medicine,
Baltimore, MD 21231, USA25Department of Biochemistry and Liver Fluke and Cholangiocarcinoma Research Center, Faculty of Medicine, Khon Kaen University,
Khon Kaen 40002, Thailand
*Correspondence: [email protected] (M.-C.G.), [email protected] (D.A.W.)
http://dx.doi.org/10.1016/j.celrep.2015.12.005This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).
SUMMARY
The ampulla of Vater is a complex cellular environ-ment from which adenocarcinomas arise to form agroup of histopathologically heterogenous tumors.To evaluate the molecular features of these tumors,98 ampullary adenocarcinomas were evaluated andcompared to 44 distal bile duct and 18 duodenal ad-enocarcinomas. Genomic analyses revealed muta-tions in the WNT signaling pathway among half of
C
the patients and in all three adenocarcinomas irre-spective of their origin and histological morphology.These tumors were characterized by a high fre-quency of inactivating mutations of ELF3, a highrate of microsatellite instability, and common focaldeletions and amplifications, suggesting commonattributes in the molecular pathogenesis are at playin these tumors. The high frequency ofWNT pathwayactivating mutation, coupled with small-molecule in-hibitors of b-catenin in clinical trials, suggests future
ell Reports 14, 907–919, February 2, 2016 ª2016 The Authors 907
treatment decisions for these patientsmay be guidedby genomic analysis.
INTRODUCTION
Though the pancreas, bile duct, and intestinal duodenum share
common embryologic origins in the ventral endoderm, the
adenocarcinomas arising in this region presumably originate
from different epithelial cellular constituents present at the site
(Zaret and Grompe, 2008). These tumors have been described
in many different ways: intra-ampullary, periampullary, intra-
ampullary papillary-tubular neoplasm, ampullary/ductal, periam-
pullary-duodenal, and ampullary/not otherwise specified. The tu-
mors clearly separated from the ampulla of Vater and localized in
the bile duct, duodenum, or pancreatic duct have been identified
as distal cholangiocarcinomas or distal bile duct (CAC), duodenal
(DUOAC), or pancreatic ductal (PDAC) adenocarcinomas.
As recommended in the AJCC seventh edition 2009 staging
system (Edge et al., 2009), the current subtype classification of
ampullary adenocarcinoma (AMPAC) is based on the anatomical
location from which the tumor is thought to arise (Edge et al.,
2009), sometimes supplemented by histopathology and expres-
sion of differential markers (Adsay et al., 2012; Chang et al.,
2013; Ehehalt et al., 2011; Morini et al., 2013). This classification
is subjective and prone to inter-observer variability and can
significantly impact treatment selection and therapeutic devel-
opment (Amptoulach et al., 2011; Heinrich and Clavien, 2010;
Romiti et al., 2012; Westgaard et al., 2013). Current treatment
approaches do not distinguish patients based on subtypes, yet
tumors may arise from at least the three epithelia that converge
at that site, and somemay arise from the ampulla itself, where lit-
tle is known of the specialized epithelium that may be present.
Malignancies that arise from different cellular origins often
have vastly differing sensitivities to therapeutics. Post hoc ana-
lyses of clinical trials using histopathological criteria have not
discerned such a difference and likely represent the inaccuracy
of such a classifier. However, as most therapeutic development
is focused on agents that target specific molecular mechanisms,
a molecular characterization that would allow selection of pa-
tients for specific therapies would facilitate therapeutic develop-
ment with the aim of improving outcomes and alleviate the
impact of an inaccurate subjective classification.
For this study, we have assembled a large cohort of AMPAC
with nearby DUOAC and CAC for comparison. We show that
tumors from the duodenum, ampulla of Vater, and distal bile
duct exhibit a common spectrum of features irrespective of
their morphology, marker expression, and cellular origin. Here,
we use the term ‘‘periampullary tumors’’ in this study to refer to
the three tumor types of AMPAC, DUOAC, and CAC collectively,
as defined by the AJCC seventh edition 2009 staging system
(Edge et al., 2009), excluding cases that clearly arise from the
pancreas (pancreatic adenocarcinoma [PDAC]).
RESULTS
In order to develop a molecular taxonomy for periampullary can-
cers and define subtypes with clinical relevance, we performed
908 Cell Reports 14, 907–919, February 2, 2016 ª2016 The Authors
exome sequencing and copy-number analysis of 160 cancers
arising in the periampullary region, 62 of these clearly arising
from either the bile duct (n = 44) or the duodenum (n = 18) and
98 for which the epithelium of origin could not be clearly defined
morphologically (AMPAC). Mutations were validated by deep
and ultra-deep sequencing on a limited target region consisting
of 71 recurrently mutated genes. RNA sequencing (RNA-seq)
was performed on 30 patients: a 28-patient subset of the 98
ampullary tumors and a two-patient subset of the 18 duodenal
tumors.
Clinical Characteristics and SubtypingThe clinical characteristics of our patient cohort are described in
Table S1A. In this study, the anatomical primary site of origin of
all tumors was defined using the AJCC seventh edition 2009
staging system (Edge et al., 2009). In addition, the tumors were
also classified independently by cellular morphology and immu-
nohistochemistry (IHC) staining (see Experimental Procedures)
into intestinal, pancreatobiliary, or mixed subtypes (Table S1B).
Since treatment may be determined based on subtypes defined
by the combination of morphology and IHC even if these mea-
sures are somewhat subjective, it was an important objective
of our study to assess the reliability and meaning of these sub-
types. Subtyping according to IHC, the AMPAC tumors were
51% pancreatobiliary and 34% intestinal, with the remainder
mixed. CAC was dominated by the pancreatobiliary subtype,
86% as expected; however, 11% of CAC exhibited an intestinal
phenotype. In DUOAC, the intestinal subtype was 44%, with
22% pancreatobiliary and the remainder mixed.
By histological morphology, a smaller proportion of each
tumor type was classified as pancreatobiliary (AMPAC, 37%;
CAC, 77%; and DUOAC, 6%). The two methods of classifica-
tion yielded concordant subtypes only 62% of the time for
AMPAC tumors, 77% of the time in CAC, and 53% of the time
in DUOAC. Although the two methods often disagreed, all three
tumor types included in their numbers concordant cases of
all three subtypes. Thus, tumors originating in each organ site
in the periampullary region may be classified as any of the
three subtypes, though this classification system is rarely
applied to DUOAC or CAC tumors. These tumors were analyzed
by genomic methods to further characterize their molecular
properties.
Mutation AnalysisExomes were sequenced to an average of 120-fold coverage re-
sulting in 28,795 mutations across 152 patients. Eight additional
patients were sequenced with targeted custom sequencing
and were included in the study (see Supplemental Experimental
Procedures, ‘‘Sequencing design and Mutation analysis,’’ and
Tables S1C–S1E). Microsatellite instable phenotypes were
observed in 12 patients representing each organ cohort (Fig-
ure 1), accounting for 18,572 of the whole-exome sequencing
(WES) discovery set. Using a method we developed based on
the enzyme slippage of the homopolymer region (E.S., unpub-
lished data), we identified two other patients among the targeted
sequencing set (Figure S1A).
Excluding microsatellite instable (MSI) tumors and correc-
ting for tumor purity, the median mutation rate did not vary
A B
C D
Figure 1. Mutation Frequencies and MSI
Characteristics
(A) Mutation frequencies for all patients by
anatomical site (D, DUOAC; A, AMPAC; C, CAC)
and subtype (I, intestinal; M, mixed; PB, pan-
creatobiliary). Black dots, microsatellite stable
(MSS); red dots, microsatellite instable (MSI).
(B) Germline mutations in MMR gene associated
with Lynch syndrome were detected in 66% of the
MSI samples. Survival (m) is in months; black tile,
patient died of disease; white tiles, patient alive;
Lynch Mutation Freq, frequency each gene is
observed in Lynch syndrome patients; blue tiles,
missense mutations; green, frameshift mutations;
red, nonsense mutations; ‘‘L’’ = known Lynch
syndrome mutation; ‘‘d’’ deleterious mutation by
PolyPhen2.
(C) Kaplan-Meier plot for survival based on MSI
status in AMPAC (log rank p = 0.04, n = 96).
(D) Kaplan-Meier plot for survival based on MSI
status in all periampullary tumors (p = 0.0028,
n = 160).
See also Figures S1A–S1C and Tables S1A–S1E.
significantly across the AMPAC, CAC, and DUOAC (3.8, 4.6, and
4.7 per Mb, respectively) but was clearly distinct from the MSI
mutation rate (68, 127, and 108 per Mb, respectively) (Figures
1A, S1B, and S1C). Two-thirds of the hypermutated WES sam-
ples had germline mutation in genes associated with Lynch syn-
drome. Interestingly, PMS2, a gene that accounts for less than
5% of Lynch syndrome patients overall (Thompson et al.,
2004) (OMIM #600259), was mutated in one half of our MSI pa-
tients (Figure 1B). Although MSI was more common in DUOAC
than CAC patients, every morphologic category harbored at
least one PMS2 germline mutation in this study. Leaving aside
germline contribution, the overall frequency of MSI in AMPAC
was 3%. MSI appeared to confer a survival advantage in
AMPAC, as it does in other gastrointestinal (GI) cancers, as all
six AMPAC patients were alive ranging from 2 to 8 years after
diagnosis (p = 0.04 with a lack of negative event) (Figure 1C).
Taking all three anatomical sites into consideration, MSI have
better survival, p < 0.0021 (Figure 1D).
Non-negative matrix factorization was used to evaluate the
mutation signatures associated with periampullary tumors. We
identified five prominent signatures, out of 21 observed (Figures
2A, S2A, and S2B; Table S2). The most common signature was
C > T at CpG islands (#6). Indeed, this signature is most common
across all tumor types. A few CAC and AMPAC tumors had a
strong T > G, > C signature (#7) associated with the digestive
track tumors and consistent with DNA damage and exposure
to arsenic (Martinez et al., 2013). A C > G signature (#4) charac-
teristic of DNA damage by APOBEC enzymes was also present
in a few patients (Roberts et al., 2013).
Cell Reports 14, 907–919
We observed signature #1 at greater
than 20% of the total signature in 9.6%
of our entire tumor set (6% AMPAC and
21% CAC). Signature #1 is characterized
by AC, AT > AN and is enriched in non-
transcribed regions of the genome in
samples from several cancer types (PDAC, medulloblastoma,
breast tumor, AML, and CLL). However, signature #1 was also
observed in the coding region of 18 out of 486 hepatocellular car-
cinoma (4%) and 31 out of 450 colorectal carcinoma (CRC) (7%)
(Lawrence et al., 2013; Totoki et al., 2014; K.R.C., unpublished
data). Whereas none of the known signatures have yet been
associated with a difference in outcome, signature #1 was asso-
ciated with poor outcomes in our study set (multivariate Cox pro-
portional hazards p = 0.02) (Figure 2B).
The analysis of the periampullary tumors, excluding MSI pa-
tients, revealed 19 genes mutated significantly above back-
groundusingMutSig-CV (Lawrence et al., 2013) (Figure 3A; Table
S3A). Considering the ratio of inactivating tomissensemutations,
an additional three genes were brought in to the significantly
mutated gene list (Table S3B) including PBRM1, RECQL4, and
KDM6A. Gene expression data confirmed that the variants
harboring missensemutation in the driver genes were expressed
between 85% and 88% of the time (Table S3C).
Most interestingly, ELF3 a transcriptional regulator of TGFBR2
was mutated in 10.6% of the periampullary tumors with predom-
inantly inactivating frameshift or nonsense mutations (Figure 3B).
This mutation frequency is three times higher than in any other
cancer (Table S3D) (Cerami et al., 2012; Gao et al., 2013; Law-
rence et al., 2014) (http://www.cBioPortal.org). In agreement
with our finding, ELF3 mutations were found in 9.5% of extrahe-
pathic CAC in a recent study of 74 samples with four inactivating
mutations out of seven (Nakamura et al., 2015). ELF3 mutation
occurred 71%of the timewithWNTpathwaymutations in all three
periampullary groups (Figure S3A). (chi-square test, p = 0.02).
, February 2, 2016 ª2016 The Authors 909
A
B
Figure 2. Mutation Signature in Periampullary Tumors
(A) Heatmap of five dominant mutation signatures from NMF analysis of mutation spectrum for each subject. Intensity indicates the proportion of mutations for
that subject attributed to the indicated signature. Subjects are sorted first by signature 1, then signature 6 from the highest to the lowest value. Only signatures
with high penetrance are shown.
(B) Kaplan-Meier curve of survival in this cohort stratified by signature 1 levels (high, red line: signature 1 component >10% of all mutations; low, black line:
otherwise, multivariate Cox proportional hazards p = 0.001).
See also Figures S2A and S2B and Table S2.
Considering the 44 CAC alone, four genes were significantly
mutated in this cancer: TP53, KRAS, SMAD4, and CDKN2A
with the highest mutation incidence in TP53. Whereas intrahe-
patic CAC tumors frequently harbor BAP1, IDH1, and IDH2
(Nakamura et al., 2015), these were absent with the exception
of a single IDH1 hotspot mutation in the periampullary CAC.
This is in agreement with Nakamura et al. (2015), where no
IDH1 mutations could be detected among 74 extrahepatic tu-
mors (compared to a 5% mutation rate in intrahepatic tumor)
and a less than 3% BAP1 mutation rate was found in extrahe-
910 Cell Reports 14, 907–919, February 2, 2016 ª2016 The Authors
patic tumors (compared to a 12.4%mutation rate in intrahepatic
tumor).
Alteration of Key Signaling PathwaysThe significantlymutated genes defined five pathways in periam-
pullary tumors: TP53/cell division, RAS/PI3K, WNT, TGF-b,
and chromatin remodeling pathways. We combined the point
mutations and copy-number alterations (CNA) changes at
the gene level within these five pathways to assess the impact
of these pathways among the three anatomical sites (Figures
Figure 3. Significantly Mutated Genes in
Non-MSI Periampullary Tumors
(A) Significantly mutated genes are displayed by
FDR value (MutSigCV). Genes with FDR < 0.1 are
located in the left panel, genes with FDR > 0.1 but
significantly inactivated are in the middle panel,
and genes slightly under the significant threshold
of the significantly mutated gene (SMG) list are in
the right panel. The amount of samples for each
tumor type is stacked.
(B) ELF3 inactivating mutations were distributed
along the entire gene characteristic of a tumor
suppressor (q < 1.6 3 10�11). All the mutations
found in the study are represented in the figure,
each mutation being found in one patient.
See also Tables S3A and 3B–S3D.
4A and 4B). The similarities and differences in gene mutations
per tumor types and subtypes are illustrated in Figures S3A
and S3B.
TheWNT pathway was mutated in 46% of patients overall but
was clearly differentially mutated across the three tumor types,
being more frequently mutated in DUOAC (72%) than in AMPAC
(49%) or CAC (30%) (chi-square p < 0.05) (Tables S4A and S4B).
This predominance of WNT pathway mutation in DUOAC was
due mainly to more frequent mutations of APC and SOX9.
Whereas the TP53, RAS, TGF-b signaling and chromatin remod-
eling pathways are deregulated in many tumor types, the WNT
pathway deregulation is reported only in gastrointestinal tumors
(Biankin et al., 2012; Cancer Genome Atlas, 2012). We reasoned
that grouping the patients by our histological classification might
enrichWNTmutation in the intestinal subtype relative to the pan-
creatobiliary subtype. As expected, the intestinal subtype had
67% WNT pathway alterations compared to pancreatobiliary
with 30% WNT alterations, very close to the WNT frequency
based on anatomical site (Figures 4B, S3A, and S3B; Tables
S4A and S4B). Although we observe a gradient ofWNT pathway
disruption in tumors as their anatomical site moves away from
the GI tract, WNT mutation is still frequent in CAC, or ‘‘pancrea-
tobiliary’’ subtype tumors.
TGFBR2 was also more frequently mutated in DUOAC than
AMPAC and CAC, but this may have been secondary to MSI,
which was in higher proportion in DUOAC. TGFBR2 harbors an
A homopolymer run of eight bases that is a frequent target of
Cell Reports 14, 907–919
mutation in MSI patients, and 5 of the 12
TGFBR2 mutations were at this site.
Interestingly, SMAD4, a gene frequently
mutated in PDAC, was the most
commonly mutated gene of the TGF-b
pathway in AMPAC and CAC, the tissue
sites in closest proximity to the pancreas.
Mutant KRAS was the major RAS
signaling oncogene in all three tumor
types. Overall, the RTK/RAS/PI3K path-
way was activated in all periampullary pa-
tients at a statistically similar rate ranging
from 84% to 94% among the three tumor
types (Tables S4A and S4B).
Alterations in the SWI/SNF chromatin remodeling pathway
were observed most frequently in ARID1A and ARID2. Overall,
mutations in the SWI/SNF complex were equally frequent in
the three tumor types.
Pathway Mutation Correlates with Disease OutcomeMultivariate analyses on the periampullary tumors as a group
showed mutations in the TGF-b pathway are associated with
better overall survival (multivariate Cox proportional hazard
p = 0.0059, HR = 0.42) independent of stage, gender, subtype,
and MSI status (multivariate Cox proportional hazard p =
0.029). Mutations in the PI3K pathway were also associated
with better overall survival (multivariate-Cox proportional haz-
ards p = 0.036, HR = 0.43) (Figure S3C). Mutations in TP53,
KRAS, WNT, and chromatin remodeling pathways showed no
significant association with outcomes in multivariate modeling.
Interestingly, TGF-b pathway mutations were also negatively
associated with mutation signature 1 (multivariate ANOVA p =
0.02), possibly explaining the association with outcomes. How-
ever, the contribution of signature 1 to outcomes was still signif-
icant when considering TGF-b pathway mutations in the model,
indicating that these two effects are not entirely redundant.
RNA ExpressionRNA expression was analyzed in 28 AMPAC and 2 DUOAC. Due
to the high frequency of mutation in WNT and the current devel-
opment of therapeutic agents targeting b-catenin, we evaluated
, February 2, 2016 ª2016 The Authors 911
Figure 4. Major Altered Pathways in Periampullary Tumors
(A) Frequency of changes defined by somatic mutations or copy-number loss or gain is expressed as a percentage of cases for each gene. Inactivation (blue) or
activation (red) is graded in intensity by percent of patients affected.
(B) Genetic alterations in the significantly mutated genes grouped by pathway are illustrated for each patient. Note WNT and PI3K signaling pathways could be
found in the three tumor types and in each of their subtypes, including the pancreatobiliary subtype.
See also Figures S3A and S3B, wherein mutations in each gene are grouped by tumor type and subtype, and Figure S3C and Tables S4A and S4B.
912 Cell Reports 14, 907–919, February 2, 2016 ª2016 The Authors
Figure 5. Relative RNA Expression of WNT
Antagonist, Agonist, and Target Genes
Tumors were split between those with and without
WNT canonical pathway mutations as shown in
the mutation panel. The level of RNA expression
for each gene can be visualized in the heat map
and the average expression of all the genes is
summarized in the lower panel. See also Figure S4
for fusions.
the expression data using a previously developedWNT signature
that included WNT antagonist, WNT agonist, and WNT target
genes (Donehower et al., 2013). An increase in expression in
these three gene groups as a result of the WNT pathway dereg-
ulation was noticed in colorectal cancer (Donehower et al.,
2013). This could be explained by the fact that CTNNB1 activa-
tion resulted in an increased expression of targeted genes and
the unrestricted WNT signaling set up a negative feedback
loop of the WNT antagonist genes attempting to shut down
signaling. In this study, we divided the patients into WNT
mutated and those without (Figure 5, mutation panel). We then
looked at the relative RNA expression in the two tumor groups
forWNT antagonists, WNT agonists, andWNT targets (Figure 5,
middle panel). The tumors with WNT mutations trend signifi-
cantly toward higher overall WNT gene expression (p < 0.001)
(Figure 5, lower panel). The WNT gene expression profile was
also increased in some of the WNT non-mutated patients, indi-
cating that some other mechanism affecting the WNT pathway
might be at play.
Fusion analysis identified two noticeable non-recurrent fu-
sions: SLC45A3-ELK4 used as a prognosis marker in prostate
Cell Reports 14, 907–919
cancer, where its expression is elevated
(Kumar-Sinha et al., 2012; Ren et al.,
2014), and a LINE-MET fusion in a patient
without any KRAS or TP53 driving
mutations and a high MET expression
(Figure S4; Table S5). LINE element
insertions are found in PDAC, colon,
hepatocellular, esophageal, and lung
carcinoma (Paterson et al., 2015; Rodi�c
et al., 2015).
Copy-Number AlterationThe majority of CNAs involved entire
chromosomes or chromosome arms as
opposed to focal events, which are com-
mon in gastrointestinal tumors. Arm-level
deletions outnumbered amplifications
across all tumors (Figure 6A). The three
tumor types shared four arm-level ampli-
fications and nine arm-level deletions.
AMPAC shared amplification of 1q and
deletion of 1p and 8p CAC (Table S6A).
AMPAC shared no events specifically
with DUOAC, making AMPAC marginally
more similar to CAC in its CNA pattern.
AMPAC also had two unique amplica-
tions on 5p and 6p, whereas 3q amplification was unique to
CAC, and 6p was unique to DUOAC.
A combined GISTIC analysis revealed as expected a focal
deletion of 9p23.1, involving CDKN2A (Table S6B). A focal
deletion in chromosome 9 removed the promotor and 50 end of
KDM4C (Figure 6B). Although present in every tumor type, it
was only statistically significant in AMPAC (Table S6B). This
deletion resulted in a significant decrease in expression of
KDM4C as well as the upstream UHRF2 (Figure 6B, inserts).
Interestingly, overexpression of both genes has been associated
with a pro-growth effect on colon cancer cells (KDM4C) and a
much lower disease-free survival and overall survival in patients
with colon cancer (UHRF2) (Lu et al., 2014; Kim et al., 2014).
KDM4C forms also complexes with b-catenin (Kim et al., 2014;
Yamamoto et al., 2013).
DISCUSSION
This study compares the genetic constitution of ampullary can-
cer with two nearby tumor types with pathologic classification.
Historically, ampullary cancers have been classified as intestinal
, February 2, 2016 ª2016 The Authors 913
Figure 6. Copy-Number Alteration
(A) Nexus GISTIC analysis of copy-number alteration by anatomical site. Upper blue panel shows copy-number gains and the lower red panel shows copy-
number losses for each tumor type. Blue arrows demark changes characteristic of a given anatomical site.
(legend continued on next page)
914 Cell Reports 14, 907–919, February 2, 2016 ª2016 The Authors
or pancreatobiliary subtypes based on immunohistochemistry
and/or cellular morphology. The genomic analysis mirrored
these results by the finding that some ampullary tumors exhibit
properties of intestinal tumors such as microsatellite instability,
ELF3 mutation, and disruption of WNT signaling.
We found that the classification approaches of the three peri-
ampullary tumors are often discordant with one another. No
unique molecular characteristics were specifically associated
with one tumor subtype or one tumor type. Interestingly patients
from each tumor type and subtype exhibited alterations in WNT
pathway genes, including nearly one-fourth of the CAC tumors
and one-fourth of the pancreatobiliary tumors. Other studies
using subtype classification different from ours report WNT
pathway mutations in AMPAC pancreatobiliary (PB) subtype
(Achille et al., 1996; Hechtman et al., 2015). Transcriptional
changes in AMPAC tumors in WNT signaling genes was
increased, as expected, in tumors with WNT mutation, reinforc-
ing a molecular dichotomy. With half of the patients across the
three tumor types harboring WNT mutation, this could impact
greatly the choice of treatment since several WNT pathway tar-
geted therapies are in development. Ampullary, duodenal and
distal bile duct adenocarcinoma could be regarded as a WNT
± entity from the perspective of treatment. Thus, the molecular
data suggest that clinical testing for WNT signaling status might
be beneficial to patients in the near future, making this a stepping
stone to personalized medicine.
The identification of ELF3 as a significantly mutated gene with
an inactivating mutation pattern is also of interest. It was re-
ported at lower frequency in bladder and biliary tract cancers,
but not in any other cancer so far. ELF3 encodes an ETS-domain
transcription factor. By interacting with promoter regions, ELF3
is implicated in the regulation of several genes during epithelial
cell differentiation (Oliver et al., 2012). One of the genes transac-
tivated by ELF3 is TGFBR2, a prime initiator of TGF-b signaling, a
pathway with a dual role in tumorigenesis, suppressing tumor
progression at early stages but enhancing invasion and metas-
tasis at later stages (Roberts and Wakefield, 2003). The tumor-
suppressor antiproliferative function of ELF3 was previously
noted in studies on colorectal, prostate, and oral squamous can-
cer cells (Iwai et al., 2008; Lee et al., 2003, 2008; Shatnawi et al.,
2014) andmore recently in biliary tract cancer cell line (Nakamura
et al., 2015). Such studies showed that ELF3 directly binds to the
promoter region of EGR1 (Lee et al., 2008) and TGFBR2 (Lee
et al., 2003), increasing the transcription of these two tumor
suppressor genes in CRC, whereas ELF3 binding to androgen
receptor (AR) (Shatnawi et al., 2014) and matrix metalloprotei-
nase-9 (MMP9) (Iwai et al., 2008) promoters suppressed the
transcriptional activity of these tumor growth- and invasive-
ness-promoting genes in prostate and squamous cancers,
respectively. However, recent observations also suggested an
oncogenic functional role in CRC development when ELF3 is
amplified, and its upregulated expression correlated with cancer
progression and decreased patient survival (Wang et al., 2014).
(B) Focal deletion in the promotor region and at the 50 end of KMD4C impacts its an
(Integrative Genomics Viewer, Broad Institute). Deletions are in blue, and amplifica
groups: samples with (1) or without (2) focal deletion. The color ladder on the righ
See also Tables S6A and S6B.
C
A WNT-independent CTNNB1 transactivation facilitating tumor
development was also reported (Wang et al., 2014). Such dual
function has also been observed in breast and prostate cancer
(Longoni et al., 2013; Oliver et al., 2012; Shatnawi et al., 2014).
It could be argued that when ELF3 inactivation occurs early in tu-
mor development, it provides a moderate growth advantage by
suppressing TGF-b signaling. The fact that we found ELF3muta-
tion in a duodenal adenoma with intraepithelial neoplasia and
dysplasia components (DUOAC 707) and that 75%of the tumors
with ELF3mutation were lower-grade tumors (stage I or II) could
support this hypothesis. The ELF3 functional switch might
depend on tumor stage and expression of other factors and/or
be associated with its expression level, some genes being trans-
activated only when ELF3 is overexpressed. In any case, ELF3 is
implicated in the development of periampullary tumors, and its
exact functional role during periampullary tumor development
will need to be investigated further.
EXPERIMENTAL PROCEDURES
Clinical Data
A total of 160 tumors (98 AMPAC, 44 CAC, and 18 DUOAC) were collected by
the different groups participating in this study: Australian Pancreatic Cancer
Genome Initiative (APGI), Baylor College of Medicine Elkins Pancreatic Center
(BCM) as a member of The Cancer Research Banking, MD Anderson Cancer
Center (MDA), and Technical University of Dresden (TUD). Ethical approval
was obtained from each of these institution’s research ethic boards. All
patients underwent surgical pancreatoduodenectomy with curative intent
without known residual disease. Clinical data variables including race, sex,
age, familial history, operative procedure, pathological findings, and survival
from the date of initial surgery to the date of death or last follow up are pre-
sented in Table S1A.
Tumor Classification
A section of the tumor was fixed in formaldehyde and embedded in paraffin
(FFPE). H&E-stained and immunohistochemistry slides from the FFPE tissue
were examined by the pathologists from the original site of collection to
confirm diagnosis of the specific tumor section and to grade expression of
subtype markers. All slides were then centrally reviewed by a single patholo-
gist (A.G.) who was blinded to all clinical, molecular, and pathological data
at the time of review and scoring. The distinction between pancreatic, biliary,
ampullary or intestinal carcinomawas based on the anatomical site fromwhich
the carcinoma was thought to arise using the guidelines recommended in the
AJCC seventh edition 2009 staging manual (Edge et al., 2009).
Histology and Morphology
Tumors were classified as pancreatobiliary, intestinal, or mixed morphological
subtype based on the cellular morphology. A morphology similar to colorectal
adenocarcinoma (tall often pseudostratified columnar epithelium with oval
nuclei forming elongated glands) was defined as intestinal type. Morphology
similar to pancreatobiliary carcinoma (small solid nest of cells with rounded
nuclei surrounded by desmoplastic stroma and forming simple or branching
rounded glands) was defined as pancreatobiliary type. Mixed histological
types contained a mixture of both intestinal and pancreatobiliary types with
80% or less of the cells with either morphology. Grade of differentiation was
also noted aswell as presence of adenoma, signet-ring cells, ormucinous cells
(Table S1B).
dUHRF2 expression. Human Omni 2.5 SNP array results were analyzed in IGV
tions are in red. Gene expressionwas analyzed by dividing the samples into two
t indicates the tumor type (pink, AMPAC; purple, CAC; and orange, DUOAC).
ell Reports 14, 907–919, February 2, 2016 ª2016 The Authors 915
Immunohistochemistry Staining
FFPE sections were stained with antibodies against MUC1 and CDX2 (see
Supplemental Experimental Procedures). This two-antibody panel has previ-
ously been validated by our group to predict prognosis in ampullary carcinoma
(Chang et al., 2013), and the methods we used were the same as employed in
that study. Briefly, expression was evaluated by estimating the percentage of
positively stained carcinoma cells and the intensity of the staining (0 absent, 1+
weak, 2+, 3+ strong). H scores were calculated for bothmarkers bymultiplying
the percentage of stained cells by the intensity of the staining. The ratio of the
CDX2/MUC1 H score defined the subtypes: a ratio of 2 and above and smaller
than 0.5 were considered intestinal and pancreatobiliary, respectively. Inter-
mediate values were associated to a mixed subtype (Table S1B).
Nucleic Acid Isolation
Samples were retrieved and had full face sectioning performed in OCT embed-
ding media to verify the presence of carcinoma in the sample to be sequenced
and to estimate the percentage of malignant epithelial nuclei in the sample
relative to stromal nuclei. Macrodissection was performed if possible to excise
areas of non-malignant tissue. DNA and RNA extraction was performed at the
center of collection following their own protocol with all samples being tracked
using unique identifiers though out the process (see Supplemental Methods).
DNAwas shipped and quantified at BCM-Human Genome Sequencing Center
(HGSC) using the PicoGreen DNA Assay.
SNP Array Assays
SNP arrays were processed at the HGSC for each sample using the Illumina
Infinium LCG Assay according to the manufacturer’s guidelines. Specifically,
assays were performed with Human Omni2.5-8 BeadChips (Illumina, catalog
no.WG-311-2513), interrogating 2.5million SNP loci with aMAF detection limit
of 1% (see Supplemental Experimental Procedures). SNP calls were collected
using Illumina’s GenomeStudio software (version 2011.1) in which standard
SNP clustering and genotyping were performed with the default settings rec-
ommended by the manufacturer. Data from samples that met a minimum SNP
call rate of 0.9 were considered passing and were included in subsequent an-
alyses. Results were analyzed on Nexus (BioDiscovery).
Sequencing
Library preparation, whole (Bainbridge et al., 2011) and targeted exome cap-
ture, and regular and ultra-deep sequencing on HiSeq 2000 platform are
detailed in Supplemental Experimental Procedures. In brief, 152 samples
were whole-exome sequenced and their mutations validated with a custom
design targeted exome capture. The targeted capture consisted of a panel
of 71 genes covering 0.25 Mb, and the probes were designed by Nimblegen
(genes are listed in Supplemental Experimental Procedures). These genes
were selected on the basis they were significantly mutated and/or had
high impact in the development of AMPAC, PDAC, DUOAC, CAC, and other
pancreatic tumor types. The selective targeted capture was also used in dis-
covery on eight samples received at a later date (six samples) or of low purity
(two samples with <10% tumor). The mutations identified with the targeted
capture were validated with ultra-deep (single-molecule reconstruction)
sequencing.
Data Analysis
Primary Data Analysis
Initial sequence analysis was performed using the HGSC Mercury analysis
pipeline (https://www.hgsc.bcm.edu/software/mercury). First, the primary
analysis software on the instrument produces .bcl files that are transferred
off-instrument into the HGSC analysis infrastructure by the HiSeq Real-time
Analysis module. Once the run is complete and all .bcl files are transferred,
Mercury runs the vendor’s primary analysis software (CASAVA), which demul-
tiplexes pooled samples and generates sequence reads and base-call
confidence values (qualities). The next step is the mapping of reads to the
GRCh37 Human reference genome (http://www.ncbi.nlm.nih.gov/projects/
genome/assembly/grc/human/) using the Burrows-Wheeler aligner (Li and
Durbin, 2009) (BWA; http://bio-bwa.sourceforge.net/) and producing a BAM
(Li et al., 2009) (binary alignment/map) file. The third step involves quality reca-
libration (using GATK; DePristo et al., 2011; https://www.broadinstitute.org/
916 Cell Reports 14, 907–919, February 2, 2016 ª2016 The Authors
gatk/) and, where necessary, the merging of separate sequence-event
BAMs into a single sample-level BAM. BAM sorting, duplicate-read marking,
and realignment to improve in/del discovery all occur at this step.
Cancer Data Analysis
Primary BAM files were separately run through Atlas-SNP (Shen et al., 2010),
Atlas-Indel, and PInDel (Ye et al., 2009). Data were aggregated for each tumor/
normal pair, and variants were cross-checked for each tissue pair. Variant
annotation was performed using Annovar (Wang et al., 2010), COSMIC (Forbes
et al., 2011), and dbSNP (Sherry et al., 2001). Variant filtering was performed to
remove low-quality variants. Cohort-level data processing was performed to
remove additional false somatic calls by filtering against a cohort of normal
tissues.
Ultra-deep Sequencing Analysis
Duplicate reads were aggregated and consensus variants were defined as the
variant being present in 90% of the reads contributing to both halves of the
duplex molecule. Subsequent filtering was employed to remove variants in
which there was either mapping error (tested using BLAST) or sequence error
non-consensus block rate of 50% (Altschul et al., 1990). Variants detected in
this way were annotated using ANNOVAR, COSMIC, and dbSNP annotations
(Forbes et al., 2011; Sherry et al., 2001; Wang et al., 2010).
Mutational Signature
Mutation signatureswere generated froma set of over 6,000 somaticmutations
across a range of cancer types using non-smooth non-negative matrix factor-
ization (nsNMF) (K.R.C., unpublished data; Pascual-Montano et al., 2006). The
solution resulted in 21distinctmutational signatures, similar to those previously
reported (Alexandrov et al., 2013; Gaujoux and Seoighe, 2010), many of which
could be correlated with previously published mutational modes, including
APOBEC, UV radiation exposure, POLE hypermutation (Lawrence et al.,
2013), and CpG island mutation. Mutations for this cohort where compared
against the solved NMF to generate a mutational decomposition for each of
the tumor samples. Samples were aggregated and compared using hierarchi-
cal clustering and other correlative statistics to clinical covariates.
Tumor Purity and Normalization Mutation Rate
Tumor purity was estimated using ASCAT and the tumor variant allelic fraction
of driver genes. The average of both analyses was plotted against the number
of mutations in each tumor, and the slope value was used to approximate the
number of mutations that would have been identified in 100% tumor cellularity
(Figure S1A).
Significantly Altered Genes
Several approaches were taken to dissect genes and pathways which
were mutated more often than by chance in this dataset. We used the final
MAF file (Tables S6A and 6B) to calculate significantly mutated genes using
MutSig-CV and an inactivation bias test (Lawrence et al., 2013; Totoki et al.,
2014).
Microsatellite Instability Coefficient
For details, see Supplemental Experimental Procedures.
Multivariable Cox Analysis
Cox proportional hazards analysis was performed using the survival (Ther-
neau, 2000) package in R (R Development Core Team, 2008). We included
age at diagnosis, gender, stage, grade, tumor type, histologic subtype (IHC),
and mutation status (WNT, TGFB, TP53, KRAS, PI3K, and chromatin remodel-
ing) in the multivariate Cox analysis. Country of origin and ethnicity were not
included as covariates since they had no associated effect with survival.
RNA-Seq
Total RNA was prepared using the AllPrep RNA/DNA isolation kit (QIAGEN).
RNA integrity was confirmed (RIN > 7.0) on a Bioanalyzer (Agilent). RNA-seq
libraries were prepared using the TruSeq Stranded total RNA LT library kit
(Illumina) following the manufacturer’s instructions. 100-base-pair-end
sequencing was then performed to a minimum depth of 50 million reads of
each sample on an Illumina HiSeq2000 sequencer.
Transcript Expression Analysis
Gene Expression of the WNT Pathway
The profiles in the RNA-seq dataset were quantile normalized, and log-trans-
formed expression values were then centered to SDs from the median across
sample profiles. Tumors were split into two groups (those withWNT canonical
pathway mutations and those without such mutations) and were scored for
relative activity in theWNT pathway. The gene signature score within each tu-
mor profile was defined as the average of the centered values for the WNT
signature genes.
Fusion Analysis
The deFuse software version 0.6.1 (McPherson et al., 2011) with default set-
tings was used to detect fusion genes. The deFuse results were further filtered
by removing identified read through fusions, selecting coding regions, select-
ing in-frame (open reading frame) genes and selecting samples with a deFuse
confidence score of >80%. This filtering resulted in a list of candidate fusion
genes. To characterize these candidate fusion genes, we took each spanning
junction read and using the BLAT tool in UCSC genome browser examined
where the reads mapped. The fusions that mapped with 100% identity to
each part of the identified fusion (gene1 or gene2) were selected for further
analysis. This filter removed genes that mapped to multiple locations. Next,
each RNA BAM from candidate fusion genes was examined in IGV, looking
for stacked soft clipped reads, changes in coverage, at the identified fusion
breakpoints. The sequence of each soft clipped read was brought into the
UCSC genome browser and mapped using BLAT. Only fusions that had reads
that matched (100%) the identified fusion genes were considered further.
ACCESSION NUMBERS
The accession number for the sequence data reported in this paper is dbGAP:
PRJNA280134.
SUPPLEMENTAL INFORMATION
Supplemental Information includes Supplemental Experimental Procedures,
four figures, and six tables and can be found with this article online at http://
dx.doi.org/10.1016/j.celrep.2015.12.005.
CONSORTIA
Themembers of Australian Pancreatic Cancer Genome Initiative are Andrew V.
Biankin, Amber L. Johns, Amanda Mawson, David K. Chang, Christopher J.
Scarlett, Mary-Anne L. Brancato, Sarah J. Rowe, SkyeH. Simpson,MonaMar-
tyn-Smith, Michelle T. Thomas, Lorraine A. Chantrill, Venessa T. Chin, Angela
Chou, Mark J. Cowley, Jeremy L. Humphris, Marc D. Jones, R. Scott Mead,
Adnan M. Nagrial, Marina Pajic, Jessica Pettit, Mark Pinese, Ilse Rooman,
Jianmin Wu, Jiang Tao, Renee DiPietro, Clare Watson, Angela Steinmann,
Hong Ching Lee, Rachel Wong, Andreia V. Pinho, Marc Giry-Laterriere, Roger
J. Daly, Elizabeth A. Musgrove, and Robert L. Sutherland (Garvan Institute of
Medical Research, Australia); Sean M. Grimmond, Nicola Waddell, Karin S.
Kassahn, David K. Miller, Peter J. Wilson, Ann-Marie Patch, Sarah Song,
Ivon Harliwong, Senel Idrisoglu, Craig Nourse, Ehsan Nourbakhsh, Suzanne
Manning, Shivangi Wani, Milena Gongora, Matthew Anderson, Oliver Holmes,
Conrad Leonard, Darrin Taylor, ScottWood, Christina Xu, Katia Nones, J. Lynn
Fink, Angelika Christ, Tim Bruxner, Nicole Cloonan, Felicity Newell, John V.
Pearson, Peter Bailey, Michael Quinn, Shivashankar Nagaraj, Stephen Kazak-
off, Nick Waddell, Keerthana Krisnan, Kelly Quek, and David Wood (Queens-
land Centre for Medical Genomics and Institute for Molecular Biosciences,
Australia); Jaswinder S. Samra, Anthony J. Gill, Nick Pavlakis, Alex Guminski,
and Christopher Toon (Royal North Shore Hospital, Australia); Ray Asghari,
Neil D. Merrett, Darren Pavey, and Amitabha Das (Bankstown Hospital,
Australia); Peter H. Cosman, Kasim Ismail, and Chelsie O’Connnor (Liverpool
Hospital, Australia); Vincent W. Lam, DuncanMcLeod, Henry C. Pleass, Arthur
Richardson, and Virginia James (Westmead Hospital, Australia); James G.
Kench, Caroline L. Cooper, David Joseph, Charbel Sandroussi, Michael Craw-
ford, and James Gallagher (Royal Prince Alfred Hospital, Australia); Michael
Texler, Cindy Forest, Andrew Laycock, Krishna P. Epari, Mo Ballal, David R.
Fletcher, and Sanjay Mukhedkar (Fremantle Hospital, Australia); Nigel A.
Spry, Bastiaan DeBoer, and Ming Chai (Sir Charles Gairdner Hospital,
Australia); Nikolajs Zeps, Maria Beilin, and Kynan Feeney (St John of God
Healthcare, Australia); NanQNguyen, AndrewR. Ruszkiewicz, ChrisWorthley,
C
Chuan P. Tan, and Tamara Debrencini (Royal Adelaide Hospital, Australia);
John Chen, Mark E. Brooke-Smith, and Virginia Papangelis (Flinders Medical
Centre, Australia); Henry Tang and Andrew P. Barbour (Greenslopes Private
Hospital, Australia); Andrew D. Clouston and Patrick Martin (Envoi Pathology,
Australia); Thomas J. O’Rourke, Amy Chiang, Jonathan W. Fawcett, Kellee
Slater, Shinn Yeung, Michael Hatzifotis, and Peter Hodgkinson (Princess Alex-
andria Hospital, Australia); Christopher Christophi, Mehrdad Nikfarjam, and
Angela Mountain; Victorian Cancer Biobank (Australia); James R. Eshleman,
Ralph H. Hruban, Anirban Maitra, Christine A. Iacobuzio-Donahue, Richard
D. Schulick, Christopher L. Wolfgang, Richard A. Morgan, and Mary Hodgin
(Johns Hopkins Medical Institutes, USA); Aldo Scarpa, Rita T. Lawlor, Stefania
Beghelli, Vincenzo Corbo,Maria Scardoni, andClaudio Bassi (ARC-Net Centre
for AppliedResearch onCancer, Italy); Margaret A. Tempero (University of Cal-
ifornia, San Francisco, USA); David K. Chang (Garvan Institute of Medical
Research, Australia; University of Glasgow, UK; and Greater Glasgow and
Clyde National Health Service, UK); Sean M. Grimmond and Craig Nourse
(University of Glasgow, UK; Queensland Centre for Medical Genomics and
Institute for Molecular Biosciences, Australia); Elizabeth A. Musgrove (Univer-
sity of Glasgow, UK); Marc D. Jones (Garvan Institute of Medical Research,
Australia; and, University of Glasgow, UK); and Nigel B. Jamieson, Fraser R.
Duthie, and Janet S. Graham (University of Glasgow, UK; andGreater Glasgow
and Clyde National Health Service, UK).
AUTHOR CONTRIBUTIONS
The research network comprises Baylor College of Medicine Human Genome
Sequencing Center, Australian Pancreatic Cancer Genome Initiative, MD An-
derson Cancer Center, TU Dresden, and Glasgow as part of the International
Cancer Genome Consortium study. Each center contributed biospecimens
collected at affiliated hospitals and processed at each biospecimen core
resource centre. Data generation and analyses were performed by the Human
Genome Sequencing Center. Investigator contributions are as follows: project
leader: M.C.G., C.P., R.G., M.J.O., S.M.G., A.V.B., D.A.W., and R.A.G.; writing
team: M.C.G. and D.A.W.; bioinformatics/databases: J.D., K.W., O.A.H., L.X.,
M.D., and P.B.; sequencing: D.M.M, H.D., S.L.L., M.B., J.H., Y.H., and H.H.D.;
data analysis: M.C.G., K.R.C., D.K.C., L.A.D., C.J.C, E.S., N.D., J.S.S., J.V.P.,
and C.S.; surgery: D.K.C, W.E.F, R.G., N.B.J., and G.V.B.; sample collection
and clinical annotation: A.L.J., C.P., S.E.H., and A.M.; sample processing
and quality control: M.C.G., A.L.J., C.P., R.G., S.L.L., and H.W.; and pathology
assessment: A.J.G., M.M.I., H.W., D.A., K.O., R.H.H., and F.R.D.
ACKNOWLEDGMENTS
We acknowledge the following funding support: HGSC-BCM: NHGRI U54
HG003273 and CPRIT grant RP101353-P7 (Tumor Banking for Genomic
Research and Clinical Translation Site 1); D.W.A. CPRIT grant RPI21018;
M.M.I. and C.J.C.: Dan L. Duncan Cancer Center NIH P30 Cancer Center
support grant (P30 CA125123) supporting the BCM Human Tissue Acquisi-
tion and Pathology Core and the Biostatistics and Bioinformatics Core;
M.J.O.: Kavanagh Family Foundation. Australian Pancreatic Cancer
Genome Initiative: National Health and Medical Research Council of
Australia (NHMRC; 631701, 535903, 427601); Queensland Government
(NIRAP); University of Queensland; Institute for Molecular Bioscience; Can-
cer Research UK (C596/A18076, C29717/A17263); University of Glasgow;
Australian Government: Department of Innovation, Industry, Science and
Research (DIISR); Australian Cancer Research Foundation (ACRF); Cancer
Council NSW: (SRP06-01, SRP11-01. ICGC); Cancer Institute NSW: (10/
ECF/2-26; 06/ECF/1-24; 09/CDF/2-40; 07/CDF/1-03; 10/CRF/1-01, 08/
RSA/1-15, 07/CDF/1-28, 10/CDF/2-26,10/FRL/2-03, 06/RSA/1-05, 09/RIG/
1-02, 10/TPG/1-04, 11/REG/1-10, 11/CDF/3-26); Garvan Institute of Medical
Research; Avner Nahmani Pancreatic Cancer Research Foundation; Howat
Foundation; R.T. Hall Trust; Petre Foundation; Philip Hemstritch Foundation;
Gastroenterological Society of Australia (GESA); American Association for
Cancer Research (AACR) Landon Foundation – INNOVATOR Award; Royal
Australasian College of Surgeons (RACS); Royal Australasian College of
Physicians (RACP); Royal College of Pathologists of Australasia (RCPA); Ital-
ian Ministry of Research (Cancer Genome Project FIRB RBAP10AHJB);
ell Reports 14, 907–919, February 2, 2016 ª2016 The Authors 917
Associazione Italiana Ricerca Cancro (12182); Fondazione Italiana Malattie
Pancreas – Ministero Salute (CUP_J33G13000210001); Wilhelm Sander Stif-
tung 2009.039.2; NIH grant P50 CA62924. See the Supplemental Informa-
tion for further acknowledgments.
Received: April 14, 2015
Revised: October 30, 2015
Accepted: November 19, 2015
Published: January 21, 2016
REFERENCES
Achille, A., Scupoli, M.T., Magalini, A.R., Zamboni, G., Romanelli, M.G., Orlan-
dini, S., Biasi, M.O., Lemoine, N.R., Accolla, R.S., and Scarpa, A. (1996). APC
gene mutations and allelic losses in sporadic ampullary tumours: evidence of
genetic difference from tumours associated with familial adenomatous polyp-
osis. Int. J. Cancer 68, 305–312.
Adsay, V., Ohike, N., Tajiri, T., Kim, G.E., Krasinskas, A., Balci, S., Bagci, P.,
Basturk, O., Bandyopadhyay, S., Jang, K.T., et al. (2012). Ampullary region
carcinomas: definition and site specific classification with delineation of four
clinicopathologically and prognostically distinct subsets in an analysis of 249
cases. Am. J. Surg. Pathol. 36, 1592–1608.
Alexandrov, L.B., Nik-Zainal, S., Wedge, D.C., Aparicio, S.A., Behjati, S., Bian-
kin, A.V., Bignell, G.R., Bolli, N., Borg, A., Børresen-Dale, A.L., et al.; Australian
Pancreatic Cancer Genome Initiative; ICGC Breast Cancer Consortium; ICGC
MMML-Seq Consortium; ICGC PedBrain (2013). Signatures of mutational pro-
cesses in human cancer. Nature 500, 415–421.
Altschul, S.F., Gish, W., Miller, W., Myers, E.W., and Lipman, D.J. (1990). Basic
local alignment search tool. J. Mol. Biol. 215, 403–410.
Amptoulach, S., Josefsson, A., Kavantzas, N., and Kalaitzakis, E. (2011).
Adenocarcinoma of the ampulla of Vater: does the histopathologic type mat-
ter? Scand. J. Gastroenterol. 46, 1401–1403.
Bainbridge, M.N., Wang, M., Wu, Y., Newsham, I., Muzny, D.M., Jefferies, J.L.,
Albert, T.J., Burgess, D.L., and Gibbs, R.A. (2011). Targeted enrichment
beyond the consensus coding DNA sequence exome reveals exons with
higher variant densities. Genome Biol. 12, R68.
Biankin, A.V., Waddell, N., Kassahn, K.S., Gingras, M.C., Muthuswamy, L.B.,
Johns, A.L., Miller, D.K., Wilson, P.J., Patch, A.M., Wu, J., et al.; Australian
Pancreatic Cancer Genome Initiative (2012). Pancreatic cancer genomes
reveal aberrations in axon guidance pathway genes. Nature 491, 399–405.
Cancer Genome Atlas, N.; Cancer Genome Atlas Network (2012). Comprehen-
sive molecular characterization of human colon and rectal cancer. Nature 487,
330–337.
Cerami, E., Gao, J., Dogrusoz, U., Gross, B.E., Sumer, S.O., Aksoy, B.A., Ja-
cobsen, A., Byrne, C.J., Heuer, M.L., Larsson, E., et al. (2012). The cBio cancer
genomics portal: an open platform for exploring multidimensional cancer
genomics data. Cancer Discov. 2, 401–404.
Chang, D.K., Jamieson, N.B., Johns, A.L., Scarlett, C.J., Pajic, M., Chou, A.,
Pinese, M., Humphris, J.L., Jones, M.D., Toon, C., et al. (2013). Histomolecular
phenotypes and outcome in adenocarcinoma of the ampulla of vater. J. Clin.
Oncol. 31, 1348–1356.
DePristo, M.A., Banks, E., Poplin, R., Garimella, K.V., Maguire, J.R., Hartl, C.,
Philippakis, A.A., del Angel, G., Rivas, M.A., Hanna, M., et al. (2011). A frame-
work for variation discovery and genotyping using next-generation DNA
sequencing data. Nat. Genet. 43, 491–498.
Donehower, L.A., Creighton, C.J., Schultz, N., Shinbrot, E., Chang, K., Gunar-
atne, P.H., Muzny, D., Sander, C., Hamilton, S.R., Gibbs, R.A., andWheeler, D.
(2013). MLH1-silenced and non-silenced subgroups of hypermutated
colorectal carcinomas have distinct mutational landscapes. J. Pathol. 229,
99–110.
Edge, S.E., Byrd, D.R., and Compton, C.C. (2009). AJCC Cancer Staging
Manual, Seventh Edition (New York: Springer).
Ehehalt, F., R€ummele, P., Kersting, S., Lang-Schwarz, C., R€uckert, F., Hart-
mann, A., Dietmaier, W., Terracciano, L., Aust, D.E., Jahnke, B., et al. (2011).
918 Cell Reports 14, 907–919, February 2, 2016 ª2016 The Authors
Hepatocyte nuclear factor (HNF) 4a expression distinguishes ampullary can-
cer subtypes and prognosis after resection. Ann. Surg. 254, 302–310.
Forbes, S.A., Bindal, N., Bamford, S., Cole, C., Kok, C.Y., Beare, D., Jia, M.,
Shepherd, R., Leung, K., Menzies, A., et al. (2011). COSMIC: mining complete
cancer genomes in the Catalogue of Somatic Mutations in Cancer. Nucleic
Acids Res. 39, D945–D950.
Gao, J., Aksoy, B.A., Dogrusoz, U., Dresdner, G., Gross, B., Sumer, S.O., Sun,
Y., Jacobsen, A., Sinha, R., Larsson, E., et al. (2013). Integrative analysis of
complex cancer genomics and clinical profiles using the cBioPortal. Sci.
Signal. 6, pl1.
Gaujoux, R., and Seoighe, C. (2010). A flexible R package for nonnegative ma-
trix factorization. BMC Bioinformatics 11, 367.
Hechtman, J.F., Liu, W., Sadowska, J., Zhen, L., Borsu, L., Arcila, M.E., Won,
H.H., Shah, R.H., Berger, M.F., Vakiani, E., et al. (2015). Sequencing of 279
cancer genes in ampullary carcinoma reveals trends relating to histologic sub-
types and frequent amplification and overexpression of ERBB2 (HER2). Mod.
Pathol. 28, 1123–1129.
Heinrich, S., and Clavien, P.A. (2010). Ampullary cancer. Curr. Opin. Gastroen-
terol. 26, 280–285.
Iwai, S., Amekawa, S., Yomogida, K., Sumi, T., Nakazawa, M., Yura, Y.,
Nishimune, Y., and Nozaki, M. (2008). ESE-1 inhibits the invasion of oral
squamous cell carcinoma in conjunction with MMP-9 suppression. Oral Dis.
14, 144–149.
Kim, T.D., Fuchs, J.R., Schwartz, E., Abdelhamid, D., Etter, J., Berry, W.L., Li,
C., Ihnat, M.A., Li, P.K., and Janknecht, R. (2014). Pro-growth role of the
JMJD2C histone demethylase in HCT-116 colon cancer cells and identification
of curcuminoids as JMJD2 inhibitors. Am. J. Transl. Res. 6, 236–247.
Kumar-Sinha, C., Kalyana-Sundaram, S., and Chinnaiyan, A.M. (2012).
SLC45A3-ELK4 chimera in prostate cancer: spotlight on cis-splicing. Cancer
Discov. 2, 582–585.
Lawrence, M.S., Stojanov, P., Polak, P., Kryukov, G.V., Cibulskis, K., Siva-
chenko, A., Carter, S.L., Stewart, C., Mermel, C.H., Roberts, S.A., et al.
(2013). Mutational heterogeneity in cancer and the search for new cancer-
associated genes. Nature 499, 214–218.
Lawrence, M.S., Stojanov, P., Mermel, C.H., Robinson, J.T., Garraway, L.A.,
Golub, T.R., Meyerson, M., Gabriel, S.B., Lander, E.S., and Getz, G. (2014).
Discovery and saturation analysis of cancer genes across 21 tumour types.
Nature 505, 495–501.
Lee, H.J., Chang, J.H., Kim, Y.S., Kim, S.J., and Yang, H.K. (2003). Effect of
ets-related transcription factor (ERT) on transforming growth factor (TGF)-
beta type II receptor gene expression in human cancer cell lines. J. Exp.
Clin. Cancer Res. 22, 477–480.
Lee, S.H., Bahn, J.H., Choi, C.K., Whitlock, N.C., English, A.E., Safe, S., and
Baek, S.J. (2008). ESE-1/EGR-1 pathway plays a role in tolfenamic acid-
induced apoptosis in colorectal cancer cells. Mol. Cancer Ther. 7, 3739–3750.
Li, H., and Durbin, R. (2009). Fast and accurate short read alignment with Bur-
rows-Wheeler transform. Bioinformatics 25, 1754–1760.
Li, H., Handsaker, B., Wysoker, A., Fennell, T., Ruan, J., Homer, N., Marth, G.,
Abecasis, G., and Durbin, R.; 1000 Genome Project Data Processing Sub-
group (2009). The Sequence Alignment/Map format and SAMtools. Bioinfor-
matics 25, 2078–2079.
Longoni, N., Sarti, M., Albino, D., Civenni, G., Malek, A., Ortelli, E., Pinton, S.,
Mello-Grand, M., Ostano, P., D’Ambrosio, G., et al. (2013). ETS transcription
factor ESE1/ELF3 orchestrates a positive feedback loop that constitutively
activates NF-kB and drives prostate cancer progression. Cancer Res. 73,
4533–4547.
Lu, S., Yan, D., Wu, Z., Jiang, T., Chen, J., Yuan, L., Lin, J., Peng, Z., and Tang,
H. (2014). Ubiquitin-like with PHD and ring finger domains 2 is a predictor of
survival and a potential therapeutic target in colon cancer. Oncol. Rep. 31,
1802–1810.
Martinez, V.D., Thu, K.L., Vucic, E.A., Hubaux, R., Adonis, M., Gil, L.,
MacAulay, C., Lam, S., and Lam, W.L. (2013). Whole-genome sequencing
analysis identifies a distinctive mutational spectrum in an arsenic-related lung
tumor. J. Thorac. Oncol. 8, 1451–1455.
McPherson, A., Hormozdiari, F., Zayed, A., Giuliany, R., Ha, G., Sun, M.G.,
Griffith, M., Heravi Moussavi, A., Senz, J., Melnyk, N., et al. (2011). deFuse:
an algorithm for gene fusion discovery in tumor RNA-Seq data. PLoS Comput.
Biol. 7, e1001138.
Morini, S., Perrone, G., Borzomati, D., Vincenzi, B., Rabitti, C., Righi, D., Castri,
F., Manazza, A.D., Santini, D., Tonini, G., et al. (2013). Carcinoma of the
ampulla of Vater: morphological and immunophenotypical classification pre-
dicts overall survival. Pancreas 42, 60–66.
Nakamura, H., Arai, Y., Totoki, Y., Shirota, T., Elzawahry, A., Kato, M., Hama,
N., Hosoda, F., Urushidate, T., Ohashi, S., et al. (2015). Genomic spectra of
biliary tract cancer. Nat. Genet. 47, 1003–1010.
Oliver, J.R., Kushwah, R., and Hu, J. (2012). Multiple roles of the epithelium-
specific ETS transcription factor, ESE-1, in development and disease. Lab.
Invest. 92, 320–330.
Pascual-Montano, A., Carazo, J.M., Kochi, K., Lehmann, D., and Pascual-Mar-
qui, R.D. (2006). Nonsmooth nonnegative matrix factorization (nsNMF). IEEE
Trans. Pattern Anal. Mach. Intell. 28, 403–415.
Paterson, A.L., Weaver, J.M., Eldridge, M.D., Tavare, S., Fitzgerald, R.C., and
Edwards, P.A.; OCCAMs Consortium (2015). Mobile element insertions are
frequent in oesophageal adenocarcinomas and can mislead paired-end
sequencing analysis. BMC Genomics 16, 473.
R Development Core Team (2008). R: A Language and Environment for Statis-
tical Computing (R Foundation for Statistical Computing).
Ren, G., Zhang, Y., Mao, X., Liu, X., Mercer, E., Marzec, J., Ding, D., Jiao, Y.,
Qiu, Q., Sun, Y., et al. (2014). Transcription-mediated chimeric RNAs in pros-
tate cancer: time to revisit old hypothesis? Omics 18, 615–624.
Roberts, A.B., and Wakefield, L.M. (2003). The two faces of transforming
growth factor beta in carcinogenesis. Proc. Natl. Acad. Sci. USA 100, 8621–
8623.
Roberts, S.A., Lawrence, M.S., Klimczak, L.J., Grimm, S.A., Fargo, D., Stoja-
nov, P., Kiezun, A., Kryukov, G.V., Carter, S.L., Saksena, G., et al. (2013). An
APOBEC cytidine deaminase mutagenesis pattern is widespread in human
cancers. Nat. Genet. 45, 970–976.
Rodi�c, N., Steranka, J.P., Makohon-Moore, A., Moyer, A., Shen, P., Sharma,
R., Kohutek, Z.A., Huang, C.R., Ahn, D., Mita, P., et al. (2015). Retrotransposon
insertions in the clonal evolution of pancreatic ductal adenocarcinoma. Nat.
Med. 21, 1060–1064.
Romiti, A., Barucca, V., Zullo, A., Sarcina, I., Di Rocco, R., D’Antonio, C., La-
torre, M., and Marchetti, P. (2012). Tumors of ampulla of Vater: A case series
and review of chemotherapy options. World J. Gastrointest. Oncol. 4, 60–67.
C
Shatnawi, A., Norris, J.D., Chaveroux, C., Jasper, J.S., Sherk, A.B.,
McDonnell, D.P., and Giguere, V. (2014). ELF3 is a repressor of androgen
receptor action in prostate cancer cells. Oncogene 33, 862–871.
Shen, Y., Wan, Z., Coarfa, C., Drabek, R., Chen, L., Ostrowski, E.A., Liu, Y.,
Weinstock, G.M., Wheeler, D.A., Gibbs, R.A., and Yu, F. (2010). A SNP
discovery method to assess variant allele probability from next-generation
resequencing data. Genome Res. 20, 273–280.
Sherry, S.T., Ward, M.H., Kholodov, M., Baker, J., Phan, L., Smigielski, E.M.,
and Sirotkin, K. (2001). dbSNP: the NCBI database of genetic variation.
Nucleic Acids Res. 29, 308–311.
Therneau, T.M. (2000). Modeling Survival Data: Extending the Cox Model
(Springer-Verlag).
Thompson, E., Meldrum, C.J., Crooks, R., McPhillips, M., Thomas, L., Spigel-
man, A.D., and Scott, R.J. (2004). Hereditary non-polyposis colorectal cancer
and the role of hPMS2 and hEXO1 mutations. Clin. Genet. 65, 215–225.
Totoki, Y., Tatsuno, K., Covington, K.R., Ueda, H., Creighton, C.J., Kato, M.,
Tsuji, S., Donehower, L.A., Slagle, B.L., Nakamura, H., et al. (2014). Trans-
ancestry mutational landscape of hepatocellular carcinoma genomes. Nat.
Genet. 46, 1267–1273.
Wang, K., Li, M., and Hakonarson, H. (2010). ANNOVAR: functional annotation
of genetic variants from high-throughput sequencing data. Nucleic Acids Res.
38, e164.
Wang, J.L., Chen, Z.F., Chen, H.M., Wang, M.Y., Kong, X., Wang, Y.C., Sun,
T.T., Hong, J., Zou, W., Xu, J., and Fang, J.Y. (2014). Elf3 drives b-catenin
transactivation and associates with poor prognosis in colorectal cancer. Cell
Death Dis. 5, e1263.
Westgaard, A., Pomianowska, E., Clausen, O.P., and Gladhaug, I.P. (2013).
Intestinal-type and pancreatobiliary-type adenocarcinomas: how does ampul-
lary carcinoma differ from other periampullary malignancies? Ann. Surg.
Oncol. 20, 430–439.
Yamamoto, S., Tateishi, K., Kudo, Y., Yamamoto, K., Isagawa, T., Nagae, G.,
Nakatsuka, T., Asaoka, Y., Ijichi, H., Hirata, Y., et al. (2013). Histone demethy-
lase KDM4C regulates sphere formation by mediating the cross talk between
Wnt and Notch pathways in colonic cancer cells. Carcinogenesis 34, 2380–
2388.
Ye, K., Schulz, M.H., Long, Q., Apweiler, R., and Ning, Z. (2009). Pindel: a
pattern growth approach to detect break points of large deletions andmedium
sized insertions from paired-end short reads. Bioinformatics 25, 2865–2871.
Zaret, K.S., andGrompe,M. (2008). Generation and regeneration of cells of the
liver and pancreas. Science 322, 1490–1494.
ell Reports 14, 907–919, February 2, 2016 ª2016 The Authors 919
Cell Reports
Supplemental Information
Ampullary Cancers Harbor the Tumor Suppressor
Gene ELF3 and Exhibit Frequent WNT Dysregulation
Marie-Claude Gingras, Kyle R. Covington, David K. Chang, Lawrence A. Donehower,
Anthony J. Gill, Michael M. Ittmann, Chad J. Creighton, Amber L. Johns, Eve Shinbrot,
Ninad Dewal, William E. Fisher, Australian Pancreatic Cancer Genome Initiative,
Christian Pilarsky, Robert Grützmann, Michael J. Overman, Nigel B. Jamieson, George
Van Buren II, Jennifer Drummond, Kimberly Walker, Oliver A. Hampton, Liu Xi, Donna
M. Muzny, Harsha Doddapaneni, Sandra L. Lee, Michelle Bellair, Jianhong Hu, Yi Han,
Huyen H. Dinh, Mike Dahdouli, Jaswinder S. Samra, Peter Bailey, Nicola Waddell, John
V. Pearson, Ivon Harliwong, Huamin Wang, Daniela Aust, Karin A. Oien, Ralph H.
Hruban, Sally E. Hodges, Amy McElhany, Charupong Saengboonmee, Fraser R. Duthie,
Sean M. Grimmond, Andrew V. Biankin, David A. Wheeler, and Richard A. Gibbs
A
B C
Figure S1. Related to Figure 1. A. Microsatellite instability coefficient. MSI tumors can be identified by the
frequency of INDEL mutations in homopolymer regions of specific genes (See Supplemental method). * patient
sequenced with targeted sequencing only. B. Correction factor of cellular purity. The average purity was plotted
against the number of mutations in each tumor samples. The normalized number of mutations for a 100% tumor
cellularity was estimated with the slope value. C. Mutation frequency. Unnormalized versus normalized mutation
frequencies for all patients by anatomical site and subtype. Black dots, non MSI; red dots, MSI; red diamonds, non
MSI average.
Supplemental Figures
A
B
Figure S2. Related to Figure 2. A. Mutation signature in periampullary tumors. Heat map of the 21 mutation signatures
from non-negative factorization analysis of mutation spectrum for each subject. Intensity indicates the proportion of
mutations for that subject attributed to the indicated signature. Subjects are sorted by hierarchical clustering. B. Mutation
signature comparison. Non-negative factorization analysis of mutation spectrum of the 5 dominant mutation signatures for
each subject in tumor type (A: AMPAC; C: CAC; D: DUOAC) and subtype (I: intestinal; M mixed; PB: pancreatobiliary).
Figure S3. Related to Figure 4. Somatic mutations in major altered pathways in the tumors grouped by types
(A) and subtypes (B). Genetic alterations in the significantly mutated genes grouped by pathway are illustrated for
each patient. Note PI3K and WNT signaling pathways could be found in each tumor type and subtypes, including the
pancreatobilliary subtype. C. Pathway alteration effect on survival. Kaplan-Meier curve of survival among patient
with or without TGF-β and PI3K altered pathway. (mutated, red; non mutated, black)
TGF-B pathway mutations C
Subtype IHC markers
I I I I I I I I I I I I I I
Tumor type
Subtype morphology
APC
CTNNB1
SOX9
AXIN1
FBXW7
TGFBR2
ACVR1B
ELF3
SMAD4
SMAD3
PIK3CA
PIK3R1
PTEN
ERBB2
GNAS
ARID1A
ARID2
swi/snf ARID1B
SMARCA4
SMARCA2
PBRM1
KRAS
NF1
TP53
ATM
CDKN2A
PancreatobiliaryIntestinal Mixed
KRAS
signaling
TP53
signaling
TGFB
signaling
PIK
signaling
WNT
signaling
Tumor type
I I I I I I I I I I I I I I
Subtype morphology
Subtype IHC markers
APC
CTNNB1
SOX9
AXIN1
FBXW7
TGFBR2
ACVR1B
ELF3
SMAD4
SMAD3
PIK3CA
PIK3R1
PTEN
ERBB2
GNAS
ARID1A
ARID2
ARID1B
SMARCA4
SMARCA2
PBRM1
KRAS
NF1
TP53
ATM
CDKN2A
CAC
KRAS
signaling
TP53
signaling
DUOAC
WNT
signaling
TGFB
signaling
PIK
signaling
swi/snf
AMPACA
B
PI3K pathway mutations
I MSI Adenoma Purity<10% No driver mutation
Duodenal Ampullary Cholangiocarcinoma
Intestinal Mixed Pancreatobiliary
Nonsense INDEL Splice site Missense
Inactivated gene smg gene
Figure S4. Related to Figure 5. ELK4 and MET fusions. The exonic regions of each gene are represented and
numbered with the protein domain illustrated behind. FAM227B had a LINE element located in the intron that was
present in the fusion RNA.
Table S2. Related to Figure 2. Most prevalent Mutation Signature. Signatures present over 20%
of the total signature were considered as prevalent.
Related to Figure 1 and uploaded as an Excel file:
Table S1a. Summary of clinico-pathological variables and outcome listed by tumor type.
Table S1b. Tumor General and Subtyping information.
Table S1c. WES discovery and validation. 152 samples mutation file (maf).
Table S1d. Custom design targeted capture sequencing discovery and validation. Eight samples
mutation file (maf).
Table S1e. Mutation file (maf) of false negative undetected with the primary discovery method
Dominant ConcomitantDUOAC
n=14
CAC
n=38
AMPAC
n=38
1 _ _ 2 1
1 6 _ 1 2
4 _ _ 1 1
4 6 _ _ 1
5 _ _ _ 1
6 _ 11 19 58
6 1 _ 5 2
6 4 _ _ 1
6 5 _ 2 _
6 7 _ 2 _
6 12 _ _ 1
6 14 _ _ 1
6 18 1 _ _
7 _ _ 2 2
7 6 _ _ 1
3 _ _ _ 1
8 _ _ _ 1
2 4 9
Signature # Number of samples
no dominant sig
Supplemental Tables
Table S3b. Related to Figure 3. Significantly inactivated genes in periampullary MSS tumors
gene smg qinactivating
Mutstotal Muts chisq.fdr
geometric
mean
SMAD4 0 13 31 0.0001 0
KRAS 0 0 88 0.31 0
APC 6.25E-11 42 46 3.00E-60 1.37E-35
CDKN2A 8.38E-12 17 23 4.64E-17 1.97E-14
ELF3 1.59E-11 14 16 3.24E-17 2.27E-14
TP53 2.79E-12 41 104 4.93E-16 3.71E-14
ARID2 1.47E-06 12 15 7.23E-13 1.03E-09
SOX9 4.90E-06 9 9 1.57E-12 2.77E-09
ARID1A 1.61E-04 11 16 3.10E-09 7.06E-07
TGFBR2 1.61E-04 7 9 3.23E-06 2.28E-05
PIK3CA 1.73E-09 0 25 0.31 2.31E-05
PBRM1 1 7 8 1.82E-07 0.0004
ACVR1B 1.14E-05 7 15 0.02 0.0005
CTNNB1 6.98E-07 1 18 0.79 0.0007
PTEN 1.21E-05 3 10 0.31 0.002
ATM 1.15E-02 7 13 0.003 0.006
RECQL4 3.89E-01 6 8 0.0001 0.007
KDM6A 1 4 4 0.0005 0.02
AXIN1 2.93E-02 4 6 0.07 0.05
KIAA0040 2.01E-02 2 3 0.31 0.08
FBXW7 2.93E-02 4 11 0.31 0.10
TCEAL4 8.38E-02 0 7 0.79 0.26
Related to Figure 3 and uploaded as an Excel file:
Table S3a. MutSig generated smg list
Table S3c. Related to Figure 3. Missense variant expression
Pathway Gene name Total
with INDEL,
nonsense, or
splice
mutations
with
Missense
mutations
with both
mutation type
with
expressed
Missense
mutation
CTNNB1 3 0 3 0 3
APC2 1 0 1 0 0
FBXW7 3 2 2 1 2
FZD10 1 0 1 0 0
KRAS 17 0 17 0 16
NF1 4 1 3 0 3
PIK3CA 6 0 6 0 4
PIK3R1 3 2 1 0 1
PTEN 3 1 2 0 1
ERBB2 3 0 3 0 2
GNAS 2 0 2 0 1
TGFBR2 4 4 1 1 1
ACVR1B 2 0 2 0 1
SMAD4 7 2 6 1 6
TP53 18 9 10 1 9
ATM 4 3 2 1 2
CDKN2A 5 4 1 0 1
ARID2 4 3 2 1 2
SMARCA4 3 1 2 0 2
Sample Total and % expression: 67 57 85%
Without low cell. samples 65 57 88%
Sample with low cellularity
swi/snf
Sample number
WNT
signaling
RTK
signaling
TGFB
signaling
TP53
signaling
Table S3d. Related to Figure 3. Percentage of ELF3 inactivating mutations in different
tumor type
Table S5. Related to Figure 5. Fusion
Tumor type (studiesa) Count % Count %
Periampullary 160 17 10.6 14 8.75b
Bladder (BGI, MSKCC, TCGA, TCGA pub) 570 60 10.5 28 4.91b
Stomach (Pfizer, TCGA pub) 388 15 3.9 8 2.06
Cholangiocarcinoma (JHU, NCCS) 55 2 3.6 2 3.64
Cervical (TCGA) 194 7 3.6 2 1.03
pRCC (TCGA) 111 3 2.7 1 0.9
Colorectal (TCGA, TCGA pub, Genentech) 767 20 2.6 15 1.96
Lung adeno (Broad, TCGA PUB) 411 8 1.9 3 0.73
Melanoma (Broad, TCGA, Yale) 539 6 1.1 1 0.18
Esophagus (Broad, TCGA) 319 3 0.9 1 0.31
Pancreasc
457 3 0.7 1 0.22
Breast (BCCRC, TCGA, TCGA pub) 1,625 6 0.4 1 0.06
a data compiled by MSKCC cBio Portal http://www.cbioportal.org/
b significantly mutated
c Grimmond, unpublished
Non silent
mutation
Inactivating
mutationNumber
cases
Gene 1 Gene 2 SampleGene
chr 1
Genomic
break pos 1
Gene
chr 2
Genomic
break pos 2orf
read
throughprobability
SLC45A3 ELK4 AMP-2618 1 205630989 1 205593019 Y Y 0.90
FAM227B MET AMP-2117 15 49688526 7 116371722 Y N 0.76
Related to Figure 4 and uploaded as an Excel file:
Table S4a. Percentage mutation and chromosomal alteration
Table S4b. Percentage alteration (mutation and CNV combined) and statistical value (chi
square test)
Table S6a. Related to Figure 6a. Arm-level copy number alterations by anatomical site
ARM DUOAC AMPAC CAC
1p del + +
1q amp + +
3p del + + +
3q amp +
4p-q del + + +
5p amp +
5q del + + +
6p amp +
6p del +
6q del + + +
7p-q amp + + +
8p del + +
8q amp + + +
9p-q del + + +
13q amp + + +
14p-q del +
17q del + + +
18p-q del + + +
20p-q amp + + +
21p-q del + + +
22p-q del + + +
Related to Figure 4 and uploaded as an Excel file:
Table S6b. GISTIC analysis of CNA from Illumina Human Omni 2.5 SNP array
Supplemental Acknowledgements
Australian Pancreatic Cancer Genome Initiative: We would like to thank C. Axford, M.-A. Brancato, S. Rowe, M.
Thomas, S. Simpson and G. Hammond for central coordination of the Australian Pancreatic Cancer Genome
Initiative, data management and quality control; M. Martyn-Smith, L. Braatvedt, H. Tang, V. Papangelis and M.
Beilin for biospecimen acquisition; and Deborah Gwynne for support at the Queensland Centre for Medical
Genomics. We also thank M. Hodgins, M. Debeljak and D. Trusty for technical assistance at Johns Hopkins
University. N. Sperandio and D. Filippini for technical assistance at Verona University. Robert L. Sutherland –
author deceased during study.
Supplemental Experimental Procedures
Immunohistochemistry Staining
Primary antibodies and dilutions were: MUC1 1:200 (Ma695, Novocastra) and CDX2 1:100 (CDX2-88, Biocare
Medical, Concord, CA, USA). All slides were processed using an automated staining system – Leica BondIII
autostainer (Leica Microsystems, Wetzlar, Germany) according to the manufacturer’s instructions using the
manufacturer’s retrieval solutions. For MUC1 and CDX2 heat induced epitope retrieval was performed for 30
minutes in the alkaline retrieval solution ER2 (AR9640, Vision BioSystems.
DNA isolation
APGI and TUD: Nucleic acids were extracted using the Qiagen Allprep® Kit (Qiagen) in accordance with the
manufacturer’s instructions with purification of DNA and RNA from the same sample. BCM: Blood was collected
and isolated using the PaxGene collection and isolation system (Qiagen). Tumors were flashed frozen and DNA
isolated with the Puregene isolation kit (Qiagen). MDA: DNA was isolated using TRIZOL and QIAmp spin column
(Qiagen).
SNP Array Assays
In brief, 200 ng of genomic DNA was denatured with NaOH, followed by isothermal whole genome amplification at
37oC for 16-24 hours. The amplified DNA was enzymatically fragmented and hybridized to the BeadChip for 16-24
hours at 48oC (with 8 samples processed in parallel for each BeadChip). After a series of washing steps to remove
unhybridized and non-specifically hybridized DNA fragments, allele-specific single-base extension reactions were
performed to incorporate labeled nucleotides into the bead-bound primers. A multi-layer staining process was
conducted to amplify signals from the labeled extended primers, and then the coated beads were imaged with the
Illumina iScan system.
Sequencing design
A total of 160 samples were used in the study. Whole exome (WES) and custom design targeted capture sequencing
methods were both used in the discovery phase based on the tumor purity or the time the samples were received.
WES was performed on 152 samples and targeted capture on 8 samples. Validation of the WES identified driver
mutations was done with targeted capture and the targeted capture discovery was validated with ultra-deep
sequencing.
Whole Exome Sequencing
Library Preparation
DNA samples were constructed into Illumina paired-end pre-capture libraries according to the manufacturer’s
protocol (Illumina Multiplexing_SamplePrep_Guide_1005361_D) with modifications as described in the BCM-
HGSC Illumina Barcoded Paired-End Capture Library Preparation protocol. Libraries were prepared using Beckman
robotic workstations (Biomek NXp and FXp models). The complete protocol and oligonucleotide sequences are
accessible from the HGSC website (https://hgsc.bcm.edu/sites/default/files/documents/Illumina_Barcoded_Paired-
End_Capture_Library_Preparation.pdf). Briefly, 1 ug of DNA in 100ul volume was sheared into fragments of
approximately 300-400 base pairs in a Covaris plate with E210 system (Covaris, Inc. Woburn, MA) followed by
end-repair, A-tailing and ligation of the Illumina multiplexing PE adaptors. Pre-capture Ligation Mediated-PCR
(LM-PCR) was performed for 7 cycles of amplification using the Phusion High-Fidelity PCR Master Mix (NEB,
Cat. no. M0531L). Universal primer IMUX-P1.0 and a pre-capture barcoded primer IBC were used in the PCR
amplification. In total, a set of 12 such barcoded primers were used on these samples. Purification was performed
with Agencourt AMPure XP beads after enzymatic reactions. Following the final XP beads purification,
quantification and size distribution of the pre-capture LM-PCR product was determined using the LabChip GX
electrophoresis system (PerkinElmer).
Exome Capture
For exome capture, four pre-capture libraries were pooled together (~300 ng/sample, 1.2 ug/pool). These pooled
libraries were then hybridized in solution to the HGSC VCRome 2.1 design (Bainbridge et al., 2011) (42Mb,
NimbleGen) according to the manufacturer’s protocol NimbleGen SeqCap EZ Exome Library SR User’s Guide
(Version 2.2) with minor revisions. Human COT1 DNA and full-length Illumina adaptor-specific blocking
oligonucleotides were added into the hybridization to block repetitive genomic sequences and the adaptor sequences.
Post-capture LM-PCR amplification was performed using the Phusion High-Fidelity PCR Master Mix with 14
cycles of amplification. After the final AMPure XP bead purification, quantity and size of the capture library was
analyzed using the Agilent Bioanalyzer 2100 DNA Chip 7500. The efficiency of the capture was evaluated by
performing a qPCR-based quality check on the four standard NimbleGen internal controls. Successful enrichment of
the capture libraries was estimated to range from a 6 to 9 of ΔCt value over the non-enriched samples.
Sequencing
Library templates were prepared for sequencing using Illumina’s cBot cluster generation system with TruSeq PE
Cluster Generation Kits (Cat. no. PE-401-3001). Briefly, these libraries were denatured with sodium hydroxide and
diluted to 6-9 pM in hybridization buffer in order to achieve a load density of ~800K clusters/mm2. Each library
pool was loaded in a single lane of a HiSeq flow cell, and each lane was spiked with 2% phiX control library for run
quality control. The sample libraries then underwent bridge amplification to form clonal clusters, followed by
hybridization with the sequencing primer. Sequencing runs were performed in paired-end mode using the Illumina
HiSeq 2000 platform. Using the TruSeq SBS Kits (Cat. no. FC-401-3001), sequencing-by-synthesis reactions were
extended for 101 cycles from each end, with an additional 7 cycles for the index read. Sequencing runs generated
approximately 300-400 million successful reads on each lane of a flow cell, yielding 7-13 Gb per sample. With these
sequencing yields, samples achieved an average of 92% of the targeted exome bases covered to a depth of 20X or
greater.
Targeted capture Sequencing (Custom Design Capture)
Illumina Library Preparation & Sequence Capture
Deep sequencing
DNA samples were constructed into Illumina paired-end pre-capture libraries according to the manufacturer’s
protocol (Illumina Multiplexing_SamplePrep_Guide_1005361_D) with the below described modifications. Libraries
were prepared using Beckman robotic workstations (Biomek FXp model).
Briefly, 0.5 ug of DNA in 70 ul volume was sheared into fragments of approximately 200-300 base pairs in a
Covaris plate with E210 system (Covaris, Inc. Woburn, MA) followed by end-repair (NEBNext End-Repair
Module; Cat. No. E6050L), A-tailing (NEBNext® dA-Tailing Module; Cat. No. E6053L) and ligation of the
Illumina multiplexing PE adaptors with barcode sequences using the ExpressLink™ T4 DNA Ligase (A custom
product from LifeTech). In total, a set of 12 such barcodes were used on these samples. Pre-capture Ligation
Mediated-PCR (LM-PCR) was performed for 6-8 cycles using the Library Amplification Readymix containing
KAPA HiFi DNA Polymerase (Kapa Biosystems, Inc., Cat # KK2612). Universal primer IMUX-P1.0 and IMUX-
P3.0 were used in the PCR amplification. Purification was performed with Agencourt AMPure XP beads after
enzymatic reactions. Following the final XP beads purification, quantification and size distribution of the pre-
capture LM-PCR product was determined using the LabChip GX electrophoresis system (PerkinElmer) and gel
analysis using AlphaView SA Version 3.4 software.
For the hybridization step, 12 such six pre-capture libraries were pooled together (~125 ng/sample for a total 1.5 ug
per pool). These pooled libraries were then hybridized in solution to the custom capture design (255.48 Kb) as
described earlier for exome library enrichment with the exception that for post-capture LM-PCR amplification,
KAPA HiFi DNA Polymerase (Kapa Biosystems, Inc., Cat # KK2612) with 12 cycles of amplification. Post
enrichment steps were similar to the exome library enrichment.
The custom capture was designed to cover the following genes and regions. (SMG are indicated in red):
Exonic regions
Gene Chr
Gene Chr
Gene Chr
APC chr5
ARID1A chr1
ATXN1 chr6
CTNNB1 chr3 ARID2 chr12
ATRX chrX
SOX9 chr17 ARID1B chr6
DAXX chr6
AXIN1 chr16 SMARCA4 chr19
MEN1 chr11
FBXW7 chr4 SMARCA2 chr9
MTOR chr1
FZD10 chr12 PBRM1 chr3
SETD2 chr3
TGFBR2 chr3 MAP2K1 chr15
TSC2 chr16
ACVR1B chr12 MAP2K4 chr17
APC2 chr19
ELF3 chr1 FBN2 chr5
CDKN1B chr12
SMAD4 chr18 FBN3 chr19
CPA1 chr7
SMAD3 chr15 KDM6A chrX
DCC chr18
KRAS chr12 MLL3 chr7
DMKN chr19
NF1 chr17 RNF43 chr17
ERBB3 chr12
PIK3CA chr3 SF3B1 chr2
ERBB4 chr2
PIK3R1 chr5 ALK chr2
MAGEA3 chrX
PTEN chr10 BRCA1 chr17
MAGEA6 chrX
ERBB2 chr17 BRCA2 chr13
ROBO3 chr11
GNAS chr20 BAP1 chr3
IRF2BP2 chr1
TP53 chr17 IDH1 chr2
CDC27 chr17
ATM chr11 IDH2 chr15
CDKN2A chr9 NRAS chr1
Selected regions
Gene Chr Region covered
ACVR2A chr2 homopolymer (chr2:148683686-148683693)
AIM2 chr1 homopolymer (chr1:159032487-159032496)
ATR chr3 homopolymer (chr3:142274740-142274749; chr3:142217556-142217563)
MLH3 chr14 homopolymer (chr14:75514604-75514612)
MSH3 chr5 homopolymer (chr5:79970915-79970922)
MSH6 chr2 homopolymer (chr2:48030640-48030647; chr2:48027196-48027202)
PMS2 chr7 homopolymer (chr7:6027157-6027164)
POLE chr12 exonuclease (chr12:133257855-133248899)
Ultra-deep sequencing (single molecule reconstruction)
DNA (90-500 ng) was sheared into 200-300 base pairs followed by end-repair and A-tailing similar to the above
described deep sequencing libraries but during the ligation step, six different barcodes were ligated to each sample.
These barcodes are 9-bp in length and a total of 48 such barcodes were used in this study. The post ligated DNA was
then purified with Agencourt AMPure XP beads, quantified using Qubit® 2.0 Fluorometer and 10 ng of this DNA
was amplified using KAPA HiFi DNA Polymerase (Kapa Biosystems, Inc., Cat # KK2612) for 22 cycles.
For the hybridization step, 8 such pre-capture libraries were pooled together (~188 ng/sample for a total 1.5 ug per
pool). These pooled libraries were then hybridized in solution to the custom capture design (255.48 Kb as described
earlier for exome library enrichment with the exception that for post-capture LM-PCR amplification, KAPA HiFi
DNA Polymerase (Kapa Biosystems, Inc., Cat # KK2612) with 12 cycles of amplification. Post enrichment steps
were similar to the exome library enrichment.
Sequencing Statistical analysis
Sequencing data were analyzed to assure pertinent representation. The whole exome sequencing (WES) as well as
the custom design targeted capture (CDC) met high quality standard.
Mutation analysis
The distribution of mutations identified among the 160 WES and targeted capture samples (see Sequencing design)
is listed in following table.
Data analysis
Microsatellite instability coefficient
MSI is characterized by the high frequency of INDEL mutation in homopolymer regions. We identified 9 genes with
such mutations in periampullary tumors (see Table below). To calculate a MSI coefficient, the insertion or deletion
at a site were counted in the normal and the tumor of each sample; the allelic fraction of each event was calculated
and the normal value subtracted from the tumor value. The sum from the allelic fraction of all genes represented the
homopolymer instability coefficient (see figure S1a).
Total MB Total ReadsBase 20+
cov
Avg
Coverage
Normal 9,810 96,666,088 93.4% 123
Tumor 9,645 95,026,884 92.0% 120
Normal 3,736 36,994,079 99.98 3,872
Tumor 3,677 36,410,698 99.96 3,748
Average value
CDC: 160
cases
WES: 152
cases
Patient
numberTotal
Mis
sense
Non
senseINDEL
Splice
SiteSilent
Non
coding
RNA
AMPAC 92 6767 4236 325 452 104 1647 3
MSI AMPAC 6 7388 3868 208 1356 177 1760 19
CAC 42 2379 1568 104 115 32 558 2
MSI CAC 2 5244 3044 167 593 98 1335 7
DUOAC 12 1122 710 61 39 17 295 0
MSI DUOAC 6 5974 2852 153 1445 241 1263 20
Total 160 28,874 16,278 1,018 4,000 669 6,858 51
Gene Homopolymer Frameshift location
MLH3 T 9 Chr 14: 75514604-75514612
MSH3 A 8 Chr 5: 79970915-79970922
MSH6 C 8 Chr 2: 48030640-48030647
ACVR2A A 8 Chr 2: 148683686-148683693
AIM2 A 10 Chr 1: 159032487-159032496
ATR A 10 Chr 3: 142274740-142274749
MLL3 T 9 Chr 7: 151874148-151874156
RNF43 C 7 Chr 17: 56435161-56435167
TGFBR2 A 10 Chr 3: 30691872-30691881