pnas201202490 5934..5941


Sleeping Beauty mutagenesis reveals cooperating
mutations and pathways in pancreatic adenocarcinoma
Karen M. Manna,1, Jerrold M. Warda, Christopher Chin Kuan Yewa, Anne Kovochichb, David W. Dawsonb,
Michael A. Blackc, Benjamin T. Brettd, Todd E. Sheetzd,e,f, Adam J. Dupuyg, Australian Pancreatic Cancer Genome
Initiativeh,2, David K. Changi,j,k, Andrew V. Biankini,j,k, Nicola Waddelll, Karin S. Kassahnl, Sean M. Grimmondl,
Alistair G. Rustm, David J. Adamsm, Nancy A. Jenkinsa,1, and Neal G. Copelanda,1,3

aDivision of Genetics and Genomics, Institute of Molecular and Cell Biology, Singapore 138673; bDepartment of Pathology and Laboratory Medicine and
Jonsson Comprehensive Cancer Center, David Geffen School of Medicine at University of California, Los Angeles, CA 90095; cDepartment of Biochemistry,
University of Otago, Dunedin, 9016, New Zealand; dCenter for Bioinformatics and Computational Biology, University of Iowa, Iowa City, IA 52242;
eDepartment of Biomedical Engineering, University of Iowa, Iowa City, IA 52242; fDepartment of Ophthalmology and Visual Sciences, Carver College of
Medicine, University of Iowa, Iowa City, IA 52242; gDepartment of Anatomy and Cell Biology, Carver College of Medicine, University of Iowa, Iowa City, IA
52242; hAustralian Pancreatic Cancer Genome Initiative; iCancer Research Program, Garvan Institute of Medical Research, Darlinghurst, Sydney, New South
Wales 2010, Australia; jDepartment of Surgery, Bankstown Hospital, Bankstown, Sydney, New South Wales 2200, Australia; kSouth Western Sydney Clinical
School, Faculty of Medicine, University of New South Wales, Liverpool, New South Wales 2170, Australia; lQueensland Centre for Medical Genomics, Institute
for Molecular Bioscience, University of Queensland, Brisbane, Queensland 4072, Australia; and mExperimental Cancer Genetics, Wellcome Trust Sanger
Institute, Hinxton, Cambridge CB10 1HH, United Kingdom

This contribution is part of the special series of Inaugural Articles by members of the National Academy of Sciences elected in 2009.

Contributed by Neal G. Copeland, February 15, 2012 (sent for review November 28, 2011)

Pancreatic cancer is one of the most deadly cancers affecting the
Western world. Because the disease is highly metastatic and difficult
to diagnosis until late stages, the 5-y survival rate is around 5%. The
identification of molecular cancer drivers is critical for furthering our
understanding of the disease and development of improved di-
agnostic tools and therapeutics. We have conducted a mutagenic
screen using Sleeping Beauty (SB) in mice to identify new candidate
cancer genes in pancreatic cancer. By combining SB with an onco-
genic Kras allele, we observed highly metastatic pancreatic adeno-
carcinomas. Using two independent statistical methods to identify
loci commonly mutated by SB in these tumors, we identified 681 loci
that comprise 543 candidate cancer genes (CCGs); 75 of these CCGs,
including Mll3 and Ptk2, have known mutations in human pancreatic
cancer. We identified point mutations in human pancreatic patient
samples for another 11 CCGs, including Acvr2a and Map2k4. Impor-
tantly, 10% of the CCGs are involved in chromatin remodeling, in-
cluding Arid4b, Kdm6a, and Nsd3, and all SB tumors have at least
one mutated gene involved in this process; 20 CCGs, including
Ctnnd1, Fbxo11, and Vgll4, are also significantly associated with poor
patient survival. SB mutagenesis provides a rich resource of muta-
tions in potential cancer drivers for cross-comparative analyses with
ongoing sequencing efforts in human pancreatic adenocarcinoma.

Pancreatic ductal adenocarcinoma is one of the most deadlyforms of cancer. In the United States, the rate of mortality is
nearly equal to the rate of new diagnoses (1). Early disease de-
tection is rare, and of those patients diagnosed with early-stage
disease, only 20% are candidates for surgical resection. Approx-
imately 50% of patients develop highly metastatic disease, for
which current treatment regimens provide little increase in lon-
gevity (2). Clearly, there is a need for better biomarkers of disease
and the identification of new therapeutic targets, particularly for
metastatic disease.
Human pancreatic cancer develops from preinvasive neoplasias,

typically intraepithelial neoplasias (PanINs), although intraductal
papillary mucinous neoplasia and mucinous cystic neoplasia can
also give rise to adenocarcinoma (3). The majority of pancreatic
adenocarcinoma is of ductal origin and contains a large desmo-
plastic component thought to promote tumorigenesis by modu-
lating the tumor microenvironment. Oncogenic KRAS mutations
are found in 90% of human tumors, and they appear in early
PanIN (4). The accumulation of additional inactivating driver
mutations in P16/CDKN2A, TP53, and SMAD4 occurs with high
frequency in later-stage PanIN, and these mutations are likely
required for tumor progression to invasive adenocarcinoma.

Recent high-throughput sequencing efforts have shown that
the majority of somatic point mutations in primary pancreatic
adenocarcinoma are missense mutations and that most occur at
low frequency (5). Pancreatic cancers exhibit a very unstable ge-
nome, and through whole-genome sequencing efforts, a specific
type of genomic instability, termed fold-back inversion, has been
elucidated (6). It is clear from these data that characterization of
a large number of tumors is required to fully capture the molec-
ular heterogeneity of the disease and identify the specific muta-
tions and signaling pathways that drive disease progression.
Pancreatic cancer has been successfully modeled in the mouse

by driving expression of oncogenic Kras in the pancreas either
alone or combined with inactivating alleles of homologs of known
human pancreatic cancer drivers (7–11), including Smad4 and
Trp53. All of these models show progressive neoplastic lesions,
termed mPanIN (12), and develop locally invasive, and in some
cases, metastatic ductal adenocarcinoma. Although these models
have furthered our understanding of pancreatic cancer biology,
new mutations that drive cancer development have not been
uncovered in these efforts.
We have taken an independent approach to identify genes

commonly mutated in a mouse model of pancreatic cancer. We
performed an insertional mutagenesis screen using the inducible
Sleeping Beauty (SB) transposon system (13) in combination with
an oncogenic Kras pancreatic cancer model. This approach is a
powerful means to identify mutations that cooperate with onco-
genic Kras to drive tumorigenesis in the mouse, because each
mutation is mapped using a transposon tag. Mice with SB in-
sertional mutagenesis exhibit all stages of mPanIN lesions. SB
decreases the latency of pancreatic tumorigenesis compared
with oncogenic Kras alone and increases both the severity and

Author contributions: K.M.M., N.A.J., and N.G.C. designed research; K.M.M., A.K., D.W.D.,
A.P.G.I., D.K.C., N.W., and K.S.K. performed research; B.T.B., T.E.S., A.J.D., A.G.R., and D.J.A.
contributed new reagents/analytic tools; K.M.M., J.M.W., C.C.K.Y., D.W.D., M.A.B., B.T.B.,
A.P.G.I., D.K.C., A.V.B., N.W., K.S.K., S.M.G., and A.G.R. analyzed data; and K.M.M. and
N.G.C. wrote the paper.

The authors declare no conflict of interest.

Freely available online through the PNAS open access option.
1Present address: Cancer Biology Program, Methodist Hospital Research Institute, Hous-
ton, TX 77030.

2A complete list of the APGI can be found in the SI Appendix.
3To whom correspondence should be addressed. E-mail: ncopeland@tmhs.org.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.
1073/pnas.1202490109/-/DCSupplemental.

5934–5941 | PNAS | April 17, 2012 | vol. 109 | no. 16 www.pnas.org/cgi/doi/10.1073/pnas.1202490109

D
o
w

n
lo

a
d
e
d
 a

t 
C

a
rn

e
g
ie

 M
e
llo

n
 U

n
iv

e
rs

ity
 o

n
 A

p
ri
l 5

, 
2
0
2
1
 

http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1202490109/-/DCSupplemental/sapp.docx
mailto:ncopeland@tmhs.org
http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1202490109/-/DCSupplemental
http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1202490109/-/DCSupplemental
www.pnas.org/cgi/doi/10.1073/pnas.1202490109


multiplicity of ductal adenocarcinomas. Multiple metastases are
also observed in a subset of animals.
Using the transposon tag, we identify 543 candidate cancer

genes (CCGs) that show significant enrichment for transposon
insertions in SB-driven pancreatic cancer. Several signaling
pathways and cellular processes are enriched in these CCGs,
most significantly TGF-β signaling and chromatin modification.
We find 20 genes not previously associated with pancreatic
cancer that are significantly associated with poor patient out-
come. We also identified point mutations in human homologs of
11 SB CCGs by exome sequencing of patient samples. SB mu-
tagenesis is an important tool for the discovery of potential
molecular drivers in pancreatic cancer and complements current
human sequencing efforts by providing an enriched dataset for
cooperating mutations involved in pancreatic adenocarcinoma.

Results
SB Drives Pancreatic Tumorigenesis in the Mouse. To identify genes
that cooperate with activated Kras in pancreatic tumorigenesis, we
performed a forward genetic screen in mice carrying a conditional
oncogenic KrasG12D allele (LSL-KrasG12D) (14) using SB trans-
poson-mediated insertional mutagenesis (13). Mice carrying the
LSL-KrasG12D allele were crossed to mice carrying a pancreatic-
specific Pdx1-Cre driver (15) to remove the floxed stop cassette
(LSL) and activate expression of oncogenic KrasG12D in the pan-
creas. These mice were then crossed to a compound transgenic
line containing up to 350 copies of a mutagenic SB transposon
from a single donor site in the genome, and an inducible SB floxed
stop transposase allele knocked into the ubiquitously expressed
Rosa26 locus (Rosa26-LSL-SB11) (13). Because of the presence
of the floxed stop cassette, the transposase is not expressed unless
Cre recombinase protein is present. We used two transposon
transgenic lines, T2Onc2 and T2Onc3, located on different
chromosomes to obtain insertion data from the entire genome.
Typically, data from a transposon donor chromosome are dis-
regarded because of computational limitations in discerning local
hopping events from selected transposition events. T2Onc2 car-
ries a strong promoter derived from the murine stem cell virus,
whereas T2Onc3 carries the ubiquitously expressed CAG pro-
moter to activate expression of proto-oncogenes. Both trans-
posons also carry splice acceptor sites in both orientations and
a bidirectional polyA signal and they can, therefore, inactivate the
expression of tumor suppressor genes when integrated within the
coding region in either orientation.

Five different cohorts of mice carrying different combinations
of alleles were bred and aged for tumor formation (Fig. 1). We
observed a significant decrease in tumor-free survival when on-
cogenic KrasG12D was combined with T2Onc3 but not T2Onc2
(Fig. 1), although the tumors were more histologically advanced
with both T2Onc2 and T2Onc3 mutagenesis compared with
KrasG12D alone (Fig. 2). Gross inspection of the pancreas at
necropsy revealed pancreatic masses that often involved neigh-
boring organs (Fig. 2A). Metastatic nodules on the liver (Fig. 2B)
and lymph nodes (Fig. 2C) were also often seen. The majority of
the animals with oncogenic KrasG12D developed early mPanIN
lesions as previously reported (8) (Fig. 2 D and E). Six animals
with oncogenic KrasG12D alone also developed noninvasive early
adenocarcinoma (Fig. S1 A–C), whereas two animals developed
spontaneous adenocarcinoma; 19 animals from the KrasG12D;
T2Onc2 and KrasG12D;T2Onc3 cohorts developed adenocarci-
noma, whereas 14 of these animals developed highly invasive
metastatic adenocarcinoma (Fig. 2F). Most of the KrasG12D;
T2Onc3 animals developed multiple primary pancreatic tumors
(two to three on average), and all had metastases to multiple sites
associated with the human disease, including lung (Fig. 2G)
and lymph nodes (Fig. 2 H and I). Three animals from the SB
cohorts lacking oncogenic KrasG12D also developed invasive
adenocarcinoma.
Tumors from all cohorts were histologically characterized for

markers commonly associated with ductal adenocarcinoma. Early
mPanIN lesions showed high mucin content, which was shown by
Alcian blue staining, that was reduced in late-stage mPanIN
lesions and largely absent in adenocarcinomas (Fig. S1D). Mucin
accumulates in human PanIN and may serve as an early bio-
marker of disease (16). Fibrosis was also observed in many of the
adenocarcinomas, and these adenocarcinomas stained positive
for both smooth muscle actin and Masson’s Trichome for

Fig. 1. SB mutagenesis drives pancreatic tumorigenesis in the mouse. Five
cohorts of mice were aged and monitored for pancreatic tumor de-
velopment. All five cohorts carried the Pdx1-Cre driver to activate oncogenic
KrasG12D and SB transposase in the pancreas. KrasG12D was required for the
high frequency induction of pancreatic tumors. Animals with oncogenic
KrasG12D and T2Onc3 died significantly earlier than animals in the other four
cohorts (P < 0.001). n is the number of animals aged for tumors.

Cecum Mass

Kidney

PanIN1

PanIN3

PanIN2

A B C

D E F

G H I

Fig. 2. Histopathological classification of pancreatic lesions. SB mutagenesis
generated pancreatic lesions at all stages of tumor progression. (A) Mice
exhibited pancreatic masses that often involved neighboring organs such as
the kidney and cecum. Metastatic nodules on the liver (B) and lymph nodes
(C) were also often visible on gross inspection. A spectrum of mPanIN lesions
was also observed with oncogenic KrasG12D, including early mPanIN1B
lesions (D) showing enlarged cell volume and high mucin content in addition
to mPanIN2 and mPanIN3 lesions (E). Nineteen animals from the T2Onc2 and
T2Onc3 cohorts developed ductal pancreatic tumors. Fourteen animals de-
veloped multiple invasive ductal adenocarcinomas (F). Several animals also
developed multiple metastatic lesions to the lung (G) and lymph nodes
(H and I).

Mann et al. PNAS | April 17, 2012 | vol. 109 | no. 16 | 5935

M
ED

IC
A
L
SC

IE
N
C
ES

IN
A
U
G
U
R
A
L
A
R
TI
C
LE

D
o
w

n
lo

a
d
e
d
 a

t 
C

a
rn

e
g
ie

 M
e
llo

n
 U

n
iv

e
rs

ity
 o

n
 A

p
ri
l 5

, 
2
0
2
1
 

http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1202490109/-/DCSupplemental/pnas.201202490SI.pdf?targetid=nameddest=SF1
http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1202490109/-/DCSupplemental/pnas.201202490SI.pdf?targetid=nameddest=SF1


collagen (Fig. S1 E and F). The presence of fibrotic foci is
a hallmark of pancreatic cancer, and high levels of smooth muscle
actin were recently correlated with poor survival of patients with
advanced pancreatic ductal adenocarcinomas (17). Pdx1 was also
reexpressed in both adenocarcinomas and metastases (Fig. S1 G–
I), and these lesions were of ductal origin, which was shown by
cytokeratin-19 staining (Fig. S1 J–L).

Sequence Analysis of Transposon Insertion Sites Identifies Pancreatic
CCGs. To identify genes that contribute to pancreatic tumor de-
velopment in oncogenic KrasG12D mice, we PCR-amplified and
sequenced the transposon insertion sites from 21 SB-induced
pancreatic tumors isolated from 19 animals; 9 tumors were from
T2Onc2 animals, and 12 tumors were from T2Onc3 animals
(Materials and Methods). High-throughput sequencing using the
Roche 454 Titanium platform identified 19,927 unique insertion
sites that mapped to the mouse genome. These insertion sites
were then analyzed using the Gaussian kernel convolution (GKC)
method to identify common insertion sites (CISs) (SI Materials
and Methods) (18). CISs are regions in the genome that contain
more transposon insertions than expected by chance and are most
likely to contain a pancreatic cancer gene. The GKC method used
30-, 50-, 75-, 120-, and 240-K kernel widths. The outputs for each
convolution were merged, and insertions contained within the
smallest kernel were used to define the CIS (Dataset S1). The
merged GKC output identified 133 CISs after correcting for
multiple testing (P < 0.05). The 133 CIS loci identified 136 CCGs

in addition to one Riken clone (1700081L11Rik), one genome
marker (D17Wsu92e), and one predicted gene (Gm1568).
Mapping of the 19,927 unique SB insertions on a Circos plot

(19), which visualizes genome data on a circular layout, showed
that the transposon insertions were well-distributed across the
genome, with about one-half of the insertions located in the plus
strand (Fig. 3, orange lines) and one-half in the minus strand (Fig.
3, purple lines). The GKC CIS peaks (Fig. 3, black lines) were also
fairly well-distributed across the mouse autosomes with the ex-
ception of chromosome 1, which was devoid of CISs (Fig. 3). This
finding may reflect the fact that the T2Onc2 donor concatamer is
located on chromosome 1, and all insertions on this chromosome
in T2Onc2 tumors were disregarded in the statistical analysis for
CISs because of the local hopping phenomenon.
The location and directionality of the transposon insertions can

be used to predict whether each CCG may function as an onco-
gene or tumor suppressor (TS) gene. Transposon insertions tend
to occur near the 5′ end of an oncogene in the same transcrip-
tional orientation, where they promote gene expression from the
strong promoter within the transposon. In contrast, insertions in
TS genes tend to be distributed across the coding region with little
orientation bias, consistent with gene inactivation. By mapping the
orientation of the transposon with respect to the orientation of the
mutated gene, the majority of pancreatic tumor CCGs are pre-
dicted to function as TS genes (Datasets S1 and S2). This finding is
similar to what has been observed for CCGs identified in trans-
poson screens performed in other types of solid tumors (20–22),

Fig. 3. Circos map of pancreatic cancer candidate cancer genes identified by the GKC method. Transposon insertions in the plus (orange lines) and minus
(purple lines) strands show genome-wide coverage of mutagenesis. GKC CCGs are illustrated on the outer perimeter of the plot with their exact location
denoted by a black line. Genes listed in red are mutated in human pancreatic cancer. The blue lines in the center connect bolded GKC CCGs that significantly
co-occur in tumors (Fisher exact test, P < 0.0003).

5936 | www.pnas.org/cgi/doi/10.1073/pnas.1202490109 Mann et al.

D
o
w

n
lo

a
d
e
d
 a

t 
C

a
rn

e
g
ie

 M
e
llo

n
 U

n
iv

e
rs

ity
 o

n
 A

p
ri
l 5

, 
2
0
2
1
 

http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1202490109/-/DCSupplemental/pnas.201202490SI.pdf?targetid=nameddest=SF1
http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1202490109/-/DCSupplemental/pnas.201202490SI.pdf?targetid=nameddest=SF1
http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1202490109/-/DCSupplemental/pnas.201202490SI.pdf?targetid=nameddest=SF1
http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1202490109/-/DCSupplemental/pnas.201202490SI.pdf?targetid=nameddest=SF1
http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1202490109/-/DCSupplemental/pnas.201202490SI.pdf?targetid=nameddest=STXT
http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1202490109/-/DCSupplemental/pnas.201202490SI.pdf?targetid=nameddest=STXT
http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1202490109/-/DCSupplemental/sd01.xlsx
http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1202490109/-/DCSupplemental/sd01.xlsx
http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1202490109/-/DCSupplemental/sd02.xlsx
www.pnas.org/cgi/doi/10.1073/pnas.1202490109


but it is in contrast to what is observed in hematopoietic tumors,
where the majority of CCGs seem to function as oncogenes (23).
Eighteen pancreatic GKC CCGs (14%) have also been pre-

viously shown to be mutated in human pancreatic cancer by exon
resequencing or whole-genome sequencing (Dataset S2), and
they are listed in red on the Circos plot in Fig. 3 (5, 6). These
genes are involved in chromatin remodeling (Crebbp, Ino80,
Mecom, Mll5, Mll3, Kdm2a, and Myst3), cell adhesion (Ptk2 and
Ptprk), protein transport (Xpo7), and transcriptional regulation
(Crebbp, Mllt10, Smad4, and Zfp148). We also found significant
enrichment for pancreatic GKC CCGs in regions containing
genomic rearrangements in human pancreatic tumors (n = 6,
P = 5.74E-03) (Materials and Methods) (6).
In addition, we found significant enrichment of GKC CCGs in

the genes listed in the Cancer Gene Census (n = 19, P = 4.47E-
11) (Materials and Methods and Dataset S2). The Cancer Gene
Census is an ongoing effort to catalog those genes for which
mutations have been causally implicated in cancer (24). Many
CCGs are also mutated in other forms of human cancer, which
was shown by exon resequencing or whole-genome sequencing,
particularly in breast (23%) (25) and ovarian (50%) cancer
(Dataset S2) (26). The degree of overlap with ovarian cancer is
likely caused by the large sample size and sequence coverage of
the recently annotated ovarian cancer genome by the Cancer
Genome Atlas project (26).

Cooperating GKC CCGs in SB-Induced Pancreatic Tumors. To identify
GKC CCGs that might cooperate in SB-induced pancreatic tu-
morigenesis, we performed pairwise comparisons using the Fisher
exact test to look for significantly comutated loci in the pancreatic
cancer genome. We identified 12 pairs of genes with P < 0.0003
from 862 pairwise comparisons (Dataset S3). These co-occurring
relationships are shown in the Circos plot in Fig. 3 by the blue lines
in the center connecting co-occurring GKC CCGs (bold font).
More than one-half of these genes are involved in regulating cell
growth and differentiation (Atf7, Chka, and Pten), apoptosis (Rere,
Ctbp2, and Shb), or cell migration and invasion (Abi1, Ankrd28,
Dstn, Fndc3b, Galnt7, Pdpk1, Pten, Shb, and Socs5).

GKC CCGs Are Enriched for Signaling Pathways and Processes Involved
in Cancer. Similar to what has been reported for human tumors,
pathway analysis of the pancreatic CCGs using Ingenuity Pathway
Analysis (IPA) (27) suggests that the GKC CCGs function in a
limited number of canonical signaling pathways, with many func-
tioning in the same signaling pathway (Table 1) (5, 26, 28). Several
of these pathways are important for human pancreatic cancer, in-
cluding TGF-β, ERK/MAPK, and Wnt/β-catenin signaling path-
ways. The PI3K/Akt signaling axis is also significantly enriched for
transposon-induced mutations that predict pathway activation. This
finding is consistent with the known relationship between activated
RAS, Akt, and tumor maintenance (29, 30). In addition, GKC
CCGs are enriched for genes involved in Ephrin receptor signaling,

Table 1. Signaling pathways enriched for CCGs

Signaling pathways
and cellular
processes

Percent
tumor

Number CCG
(GKC method) P value

Number CCG
(gCIS method) P value

Representative
genes

Analysis
platform

Molecular mechanisms
of cancer

100 13 2.25E-06 35 8.79E-07 Crebbp, Gsk3b, Mll3, Pten,
and Nsd1

IPA

GO:0016568−
chromatin
modification

100 14 3.52E-05 34 3.26E-08 Kdm6a, Mll3, Mll5, Myst3,
and Nsd1

DAVID

TGF-β 90 7 3.46E-06 16 1.5E-06 Acvr2a, Crebbp, Smad2,
Smad4, and Smurf2

IPA

RAR activation 90 9 5.72E-06 21 1.27E-05 Ncor1, Nsd1, Pdpk1, and Pik3r1 IPA
HGF signaling 90 7 1.12E-05 21 7.69E-09 Atf2, Dock1, Grb2, Ptk2,

Ptpn11, and Rap1a
IPA

ERK/MAPK 90 8 8.89E-05 19 2.84E-04 Grb2, Mapk1, Pak1, and Ptk2 IPA
Tight junction
signaling

90 8 1.25E-03 17 9.91E-05 Ash1l, Ctnna1, Inadl, Magi1,
and Magi2

KEGG

GO:0003682−
chromatin
binding

90 9 6.6E-03 22 3.04E-05 Adnp, Arid4b, Gata6, Mbd1,
and Ncoa1

DAVID

PI3K/Akt 86 7 3.72E-05 15 2.14E-04 Mapk1, Pdpki, Pik3ca, Pten,
and Rps6kb1

IPA

Integrin 86 8 1.48E-04 23 1.39E-05 Arhgap5, Itgb1, Ptk2, and
Rapgef1

IPA

Ephrin receptor 86 7 3.7E-04 21 1.51E-05 Abi1, Cdc42, Epha6, Gnaq,
and Gna14

IPA

Rac Signaling 86 5 1.25E-03 15 5.29E-05 Cdc42, Iqgap1, Iqgap2,
Map3k1, and Pak1

IPA

Formation of filaments 81 10 6.75E-05 32 4.41E-05 Akap13, Atxn2, Ctnnd1,
and Fhit

IPA

Wnt/β-catenin 76 5 1.02E-02 18 3.17E-04 Csnk1d, Gnaq, Gsk3β, EP300,
and Ppp2r5e

IPA

Adherens junction 76 5 1.6E-02 15 7.42E-06 Actn1, Cdh1, Crebbp, Ctnna1,
and Ctnnd1

KEGG

Analysis of CCGs using IPA, DAVID, and KEGG revealed several canonical signaling pathways and processes enriched for CCGs from SB-driven pancreatic
tumors. Percent tumor is the proportion of tumors that had a mutation in at least one CCG found within the pathway; 48 GKC CCGs and 174 gCCGs were
significantly enriched in canonical pathways identified by IPA. P values for pathway enrichment were adjusted for multiple testing using the Benjamini–
Hochberg method for control of the false discovery rate.

Mann et al. PNAS | April 17, 2012 | vol. 109 | no. 16 | 5937

M
ED

IC
A
L
SC

IE
N
C
ES

IN
A
U
G
U
R
A
L
A
R
TI
C
LE

D
o
w

n
lo

a
d
e
d
 a

t 
C

a
rn

e
g
ie

 M
e
llo

n
 U

n
iv

e
rs

ity
 o

n
 A

p
ri
l 5

, 
2
0
2
1
 

http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1202490109/-/DCSupplemental/sd02.xlsx
http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1202490109/-/DCSupplemental/sd02.xlsx
http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1202490109/-/DCSupplemental/sd02.xlsx
http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1202490109/-/DCSupplemental/sd03.xls


which is important for cytoskeletal reorganization and cell pro-
liferation, human growth factor (HGF) signaling, which mediates
cell growth through growth factor signaling, and retinoic acid
receptor (RAR) signaling, which is involved in regulating cell
growth and invasion.
We also used the Database for Annotation, Visualization and

Integrated Discovery (DAVID) (31, 32) to identify cellular pro-
cesses that are enriched in the pancreatic CCGs. DAVID is
a functional annotation tool used to interrogate gene sets using
over 40 annotation categories, including gene ontology terms,
protein–protein interactions, and biological pathways. Chroma-
tin modification was one of the most significantly enriched pro-
cesses for CCGs (Table 1). Indeed, all tumors had a mutation in
at least one gene encoding a histone-modifying enzyme. Several
of these genes, including Kdm6a (33), Mll3 (34), and Mll5 (35),
are mutated in several human cancers, pointing to a complex role
for epigenetics in human cancer and in particular, pancreatic
cancer. Filament assembly and specifically actin remodeling are
also significantly enriched for GKC CCGs. Genes involved in
adherens junctions and tight junctions identified using the Kyoto
Encyclopedia of Genes and Genomes (KEGG) were also fre-
quently mutated in SB-induced pancreatic cancers. These pro-
cesses play important roles in cell invasion and metastasis.

Gene centric CISs and Their Potential Role in Pancreatic Cancer. The
GKC method identifies CISs by looking for regions in the genome
of a given kernel size in which SB insertions occur at a higher
frequency than predicted by chance. We reasoned, however, that
because the vast majority of CISs in pancreatic tumors encode TS
genes and transposon insertions in TS genes are often randomly
distributed across the coding region, the GKC method might miss
a number of important CISs. We therefore decided to look for
CISs by a second gene centric method that does not rely on pat-
terns of clustered transposon insertions (36). This approach con-
siders all uniquely mappable TA dinucleotides in the coding region
of each RefSeq gene as a fraction of all uniquely mappable TA
dinucleotides in the mouse genome. The population frequency of
transposon insertions in each RefSeq is then calculated (after fil-
tering insertions on transposon donor chromosomes). χ2 analysis is
used to calculate the probability of the transposon mutating a TA
in a given gene with a higher frequency than predicted by chance.
Using this approach, we identified 671 gene centric CISs

(gCISs; P < 0.05) that contained three or more intragenic inser-
tions in at least three independent tumors (Dataset S4). Three
tumors is the minimal number of tumors required to identify a
CIS by the GKC method, and they represent 15% of our SB tu-
mor cohort. The gCIS method identified 653 candidate cancer
genes, 14 Riken clones, and 4 genomic markers; 92% of the GKC
CISs were also identified by the gCIS computational method.
We found significant enrichment of gene centric CCGs (gCCGs)
among the genes listed in the Cancer Gene Census (n = 34, P =
1.67E-12) (Dataset S5). Importantly, we also found significant
enrichment for gCCGs in the genes mutated in human pancre-
atic cancer (n = 53, P = 5.11E-09) or located near sites of ge-
nomic rearrangement in human pancreatic cancer (n = 25,
P = 1.80E-21).
When we looked for canonical signaling pathways and pro-

cesses using gCCGs, we found 402 additional genes involved in
pathways and processes enriched for pancreatic CCGs that were
identified by the GKC method. Rac signaling, involved in actin
polymerization as well as HGF, ephrin receptor, integrin, and
Wnt/β-catenin signaling pathways, showed significant enrichment
for gCCGs (Table 1). In addition, genes involved in tight junc-
tions and adherens junctions were significantly enriched for
gCCGs as were genes involved in chromatin modification. These
data indicate that the gCIS method identifies additional candi-
date cancer genes important for pancreatic adenocarcinoma in
our model.

CCGs in SB-Driven Pancreatic Cancer Predict Human Pancreatic Cancer
Patient Survival. We next wanted to assess whether any of the GKC
CCGs that we identified in our transposon screen had a significant
impact on human pancreatic cancer. We used expression array
data from human pancreatic tumors available from the National
Center for Biotechnology Information Gene Expression Omnibus
database (37, 38) (GSE21501) for which patient survival data were
available (39). We considered a CCG as being validated if at least
one of the probe sets on the array had a significant association with
patient survival (based on Cox proportional hazards regression)
after P value adjustment for multiple testing (SI Materials and
Methods). Based on these criteria, 20 GKC CCGs were signifi-
cantly associated with patient survival in this dataset. Fig. S2 shows
the survival curves generated by dividing the samples into high
and low expression groups at different percentiles; 14 of 20 genes
showed a positive correlation between the predicted mutagenic
effect of SB transposition in mouse pancreatic tumors (i.e., gene
activation or inactivation) and gene expression levels associated
with poor survival in the human tumors. The probability of ob-
taining this result by chance was 0.03. This work reports CRKRS,
PARD6G, PUM2, SOCS5, and ZFX as candidate oncogenes and
AKAP13, CTNND1, DYRK1A, FBXO11, GSK3B, MYO1D,
VGLL4, ZC3H7A, and ZFAND3 as candidate tumor suppressors
in human pancreatic cancer.
We then interrogated a tissue microarray comprised of three-

replicate 1-mM cores from 150 patients with pancreatic adeno-
carcinoma (40) for levels of protein encoded by frequently mu-
tated CCGs found in enriched signaling pathways. For two
proteins, CTNND1 and GNAQ, we observed a decrease or ab-
sence of staining in tumors with intact staining in adjacent benign
ducts or acini present within the same core (Fig. 4). CTNND1 is
membrane-bound and plays a role in adherens junctions by
regulating E-cadherin (41); it has also been shown to regulate
WNT signaling (42). GNAQ is a cytoplasmic guaninine–nucle-
otide binding protein that signals to RhoA and is implicated in
cell proliferation and migration (43). Low staining for CTNND1
(P = 0.023) was significantly associated with poor survival (Fig.
4) using a univariate Cox analysis of overall survival (SI Materials
and Methods), whereas low staining for GNAQ showed a bor-
derline significant association with poor survival (P = 0.051).
Analysis of CTNND1 staining in a second independent pancre-
atic cancer tissue microarray of 142 patient samples (44, 45)
confirmed that decreased CTNND1 is significantly associated
with poor survival in pancreatic cancer patients who underwent
operative resection (Fig. S3). These data also substantiate that
disruption of genes participating in cell–cell interactions and
regulation of cell migration in SB-driven pancreatic tumors have
biological relevance to disease.

Exome Sequencing from Human Pancreatic Cancer Patients Uncovers
Damaging Point Mutations. Finally, we investigated the coding
mutation status of the 136 GKC CCGs in human pancreatic cancer
patient samples. We queried the Australian Pancreatic Cancer
Genome Initiative exome sequence mutation discovery database,
which was created from a collection of 60 pancreatic cancer patient
DNAs with matched normal controls. Surprisingly, 68 GKC CISs
exhibited coding mutations in at least one patient. We validated
nonsilent coding mutations in nine human homologs of the mouse
GKC CISs in these samples using targeted PCR and Ion Torrent
sequencing at >1,000-fold depth of coverage. These homologs are
ACVR2A, AFF4, AP1G1, CREBBP, MEIS2, MKLN1, MLL3,
SMAD4, and THSD7A (Dataset S6). In addition, mutations in
CTNNA3 and MAP2K4, two human homologs of SB CCGs iden-
tified using the gCIS method, were also validated. Loss of SMAD4
is a well-documented driver in human pancreatic cancer and
mouse models. The work by Jones et al. (5) previously identified
mutations in MLL3, a member of the mixed lineage leukemia gene
family, in both their exome discovery and validation screens for

5938 | www.pnas.org/cgi/doi/10.1073/pnas.1202490109 Mann et al.

D
o
w

n
lo

a
d
e
d
 a

t 
C

a
rn

e
g
ie

 M
e
llo

n
 U

n
iv

e
rs

ity
 o

n
 A

p
ri
l 5

, 
2
0
2
1
 

http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1202490109/-/DCSupplemental/sd04.xlsx
http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1202490109/-/DCSupplemental/sd05.xlsx
http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1202490109/-/DCSupplemental/pnas.201202490SI.pdf?targetid=nameddest=STXT
http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1202490109/-/DCSupplemental/pnas.201202490SI.pdf?targetid=nameddest=STXT
http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1202490109/-/DCSupplemental/pnas.201202490SI.pdf?targetid=nameddest=SF2
http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1202490109/-/DCSupplemental/pnas.201202490SI.pdf?targetid=nameddest=STXT
http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1202490109/-/DCSupplemental/pnas.201202490SI.pdf?targetid=nameddest=STXT
http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1202490109/-/DCSupplemental/pnas.201202490SI.pdf?targetid=nameddest=SF3
http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1202490109/-/DCSupplemental/sd06.xlsx
www.pnas.org/cgi/doi/10.1073/pnas.1202490109


pancreatic cancer mutations. CREBBP, which encodes an acety-
lase that targets histones and nonhistones, and CTNNA3, an
ortholog of CTNND1 also involved in cell adhesion, were also
identified in their discovery screen.
Seven of the validated GKC CISs are unique. The majority of

the mutations in these genes are missense, although nonsense
mutations were identified in two genes—ACVR2A, which enc-
odes a type II receptor of the TGF-β superfamily that also par-
ticipates in integrin signaling, and MAP2K4, which is involved in
JUN kinase activation. Interestingly, all genes with validated
mutations are predicted to be potential tumor suppressors by our
screen. We also validated mutations in orthologs of four GKC
CISs that encode proteins with similar biological function. AFF2
is an ortholog of the GKC CIS AFF4, which encodes proteins
involved in both splicing and transcriptional control. DOCK2 and
DOCK3 are orthologs of the GKC CIS DOCK1 (DOCK180),
which is involved in Rac signaling. THSD7B, mutated in the
discovery screen by Jones et al. (5) and an ortholog of THSD7A,
is also mutated in the Australian Pancreatic Cancer Genome
Initiative cohort. Given the heterogeneity of mutations in pan-
creatic cancer, it is encouraging that many of our SB CCGs have
validated mutations in a cohort of only 60 patients. Additional
experimentation is required to determine the impact of these
mutations on disease development and prognosis.

Discussion
We report an SB insertional mutagenesis screen to identify loci that
cooperate with oncogenic Kras to drive pancreatic adenocarcinoma
in the mouse. SB-driven pancreatic adenocarcinomas show all
stages of mPanIN that develop into early noninvasive adenocarci-
noma and finally, invasive, highly metastatic adenocarcinoma. Im-
portantly, these tumors also have a high desmoplastic component,
a predominant histopathic feature of human pancreatic cancer. The
T2Onc3 transposon proved to be more potent than T2Onc2 in
inducing metastatic tumors. The exact mechanism for the differ-
ence between the two transposons in driving pancreatic cancer is
unknown. One possibility is that the location of the transposon

concatemer donor influences the development of the tumor. The
T2Onc3 donor resides on chromosome 9, and pancreatic cancer
genes in close proximity may be frequently mutated in these tumors
because of local hopping.
Both T2Onc2 and T2Onc3 cohorts exhibited considerably

higher numbers of nonredundant insertions per tumor relative to
the number of transposons in the donor concatemer (fivefold
greater insertions for T2Onc2 and 20-fold greater insertions for
T2Onc3 on average). The large number of mapped insertions
(19,927 from 21 independent tumors) is likely because of the fact
that many insertions are background passengers as opposed to
driver mutations, and it also reflects the oligoclonality of these
tumors. Using both the GKC and the gCIS methods, we identified
681 statistically significant CISs that comprise 543 pancreatic
CCGs. The tumor heterogeneity makes it difficult to identify co-
occurrence of insertions in individual subclones. One way to begin
to address this issue is to use laser capture microdissection to
determine the transposon insertion profiles in tumor subclones.
Ultimately, it will take single-cell sequencing to understand the
degree of heterogeneity in pancreatic and other solid tumors.
The majority of the genes that we identified in our SB pancreatic

insertional mutagenesis screen (90%) are predicted to be disrupted
based on the orientation of the transposon with respect to the gene.
For most CISs, only one transposon insertion is observed per tu-
mor, although there are examples of tumors with multiple inser-
tions that contribute to a CIS. We believe that these singly mutated
genes do play a role in pancreatic cancer for several reasons. Many
of these TS genes function in known signaling pathways in cancer,
and they are found mutated in human cancer with a frequency
greater than predicted by chance. Indeed, 75 CCGs overlap with
genes mutated or deleted in human pancreatic adenocarcinoma
from published sequencing efforts. In addition, although solid
tumors induced by SB seem to select for loss of function mutations
(13, 45), SB-induced hematopoietic cancers seem to select for
oncogenes. This finding argues that there is something special
about solid tumors that selects for loss of function mutations. Some
of the CISs that we identified may represent haploinsufficient

Fig. 4. CTNND1 and GNAQ show absent or weak staining in human pancreatic tumors. Immunohistochemistry was performed on a human pancreatic tissue
microarray for (A) CTNND1 and (B) GNAQ. Arrows indicate tumors with weak or absent staining, whereas adjacent benign ducts or acini (indicated by white
arrowheads) verify the presence of intact staining in tissue sections. Weak or absent staining of (C) CTNND1 (log rank P = 0.022) and (D) GNAQ (log rank P =
0.051) is predictive of poor survival. Kaplan–Meier plots (C and D) of patient survival represent two patient groups dichotomized into the top two tertiles
(green) vs. bottom tertile (blue) using the histoscore for high and low staining.

Mann et al. PNAS | April 17, 2012 | vol. 109 | no. 16 | 5939

M
ED

IC
A
L
SC

IE
N
C
ES

IN
A
U
G
U
R
A
L
A
R
TI
C
LE

D
o
w

n
lo

a
d
e
d
 a

t 
C

a
rn

e
g
ie

 M
e
llo

n
 U

n
iv

e
rs

ity
 o

n
 A

p
ri
l 5

, 
2
0
2
1
 



tumor suppressors, a recent example being Fbxo11 in B-cell lym-
phoma (46). Alternatively, acquired point mutation, rearrange-
ment, or epigenetic silencing may be the preferred mechanism for
inactivating the second allele in solid tumors.
Analysis of the human pancreatic cancer genome by se-

quencing SNP and Comparative Genomic Hybridization (CGH)
arrays has shown that chromosome loss or gene deletions are
frequent events in human pancreatic cancer (5, 6, 47). Missense
mutations predominate the pancreatic genome mutational
landscape. Using an algorithm to predict the significance of
cancer-associated mutations (i.e., not passengers), the work by
Jones et al. (5) reported that 17% of the missense mutations
identified in their discovery set for human pancreatic cancer are
likely to contribute to tumorigenesis. Our SB-driven pancreatic
tumors provide an enriched dataset of frequently mutated genes
that include many of the genes mutated in the human discovery
set. These genes include Mecom (mutated in 57% of SB-driven
pancreatic tumors), Mll5, Ptprk, Zc3h7a (each mutated in 43% of
tumors), and Ino80 (mutated in 38% of tumors). We also find
transposon insertions in several genes identified in the work by
Campbell et al. (6) to be either deleted (PKT2, SLC12A8, and
XPO7) or involved in intra- (KDM2A and MYO1D) or in-
terchromosomal (MLLT10) rearrangements that are each mu-
tated in >25% of the SB pancreatic tumors. The high mutation
frequency of these genes and others in our SB screen provides
evidence that many of the CCGs may indeed be drivers of
pancreatic cancer. We also identified nonsilent mutations in
eleven CCGs in human pancreatic cancer patient samples from
the Australian Pancreatic Cancer Genome Initiative. We antic-
ipate that additional mutations in human homologs of SB CCGs
will be identified in human pancreatic cancer as more patient
samples are sequenced.
The most significantly mutated gene identified by both the

GKC and gCIS methods is Pten. Mutations in this tumor sup-
pressor gene have not been identified in pancreatic cancer, al-
though the chromosomal region in which it lies on 10q23 does
undergo loss of heterozygosity (48–50). Inactivation of Pten in the
mouse, recently shown to cooperate with oncogenic Kras to drive
pancreatic adenocarcinoma (51), may be the easiest way to de-
regulate Akt signaling in the mouse. Although TP53 mutations
are frequent in human pancreatic cancer, we do not detect Trp53
as a CIS in our screen. However, we do find mutations in
a number of genes in the Trp53 pathway, including Crebbp, Jmy,
and Mapk8. Usp9x, the second most significantly mutated gene in
our screen and predicted to be inactivated, has not been associ-
ated as a tumor suppressor with any cancer, although point
mutations have been identified in breast (52), lung (53), and
ovarian cancers (26). USP9x encodes a deubiquitinase that has
been shown to stabilize MCL1 in several tumor types (54). We
identify several genes in our screen involved in ubiquitin-medi-
ated protein degradation, a few of which (ARIH1, FBXO11,
UBR5, and USP24) have mutations in human cancer. Fbxo11 has
been shown to modulate TGF-β signaling in a mouse mutant (55)
and influence Trp53 neddylation (56). Ubr5, also known as EDD,
has a number of potential targets, including β-catenin (57).
Remarkably, 10% of the pancreatic CCGs that we identify

function in chromatin remodeling, and 100% of tumors had a
mutation in at least one of these genes. Mll3 is a component of the
Asc-2 containing (Ascom) complex involved in H3K4 trimethy-
lation, a histone modification associated with gene activation (58).
This complex also contains Kdm6a, suggesting that this H3K27
trimethyl-demethylase may also have a role in methylating H3K4.
Mutations were also found in Whsc1l1 (Nsd3), a methyltransferase
known to target both H3K4 and H3K27. Analyses of methylation
patterns and patient outcome in human pancreatic ductal adeno-

carcinoma have shown that low levels of histone methylation are
generally predictors of poor survival (40, 59). Recent findings from
exome sequencing of pancreatic neuroendocrine tumors show that
chromatin remodeling and specifically, inactivating mutations in
MEN1, a component of MLL1- and MLL2-H3K4 methyl-
transferase complexes, play important roles in this disease as well
(60). Identification of the genes regulated by these epigenetic
marks will be critical to understanding how epigenetic mechanisms
can modulate pancreatic cancer.
Our data support the observations made in the work by Jones

et al. (5) that a limited number of canonical signaling pathways
are mutated in the majority of pancreatic tumors. We find sig-
nificant enrichment for SB CCGs in known signaling pathways
involved in pancreatic cancer, including TGF-β, Wnt/β-catenin,
and integrin signaling. Importantly, we identify genes involved in
these enriched pathways that have not been reported as mutated
in human pancreatic cancer. The mutational heterogeneity in SB
tumors may arise from the plasticity of pathways that permits
deregulation by a number of distinct genetic alterations. This
finding argues for a need to better understand pathway pertur-
bations in cancer at the cellular level rather than at the gene level.
SB insertional mutagenesis shows great power for identifying

candidate cancer genes with direct relevance to human cancers.
Our model of SB-driven pancreatic adenocarcinoma shows all
stages of pancreatic lesions leading up to adenocarcinoma and
metastasis. The ease of mapping transposon insertion sites within
the cancer genome makes it possible to interrogate the gene
perturbations that occur at each stage, which can potentially lead
to the identification of cancer drivers required for the initiation,
progression, and metastasis of pancreatic adenocarcinoma. The
significant overlap of our findings with sequencing data from
human tumors highlights the relevance of this pancreatic cancer
mouse model to human pancreatic adenocarcinoma. Data from
the SB-driven pancreatic tumors will undoubtedly aid in the
prioritization and validation of forthcoming human sequencing
data from pancreatic adenocarcinoma.

Materials and Methods
Mice. We used the following alleles to generate a mouse model of pancreatic
cancer: LSL-KrasG12D (14), Pdx1-Cre (15), T2Onc2 (6113) (23), T2Onc3 (12740),
and Rosa26-LSL-SB11 (61). The resulting cohorts were on a mixed B6.129
genetic background. All animals were monitored on a biweekly basis in
accordance with the Institutional Animal Care and Use Committee guide-
lines. Gross necropsies were performed, and one-half of the pancreas was
snap-frozen; the other one-half (retaining the duodenum) was fixed and
paraffin-embedded. All masses larger than 5 mm were recovered; one-half
was snap-frozen, and the other one-half was fixed and paraffin-embedded.

Cloning and Mapping Transposon Insertions Sites. Isolation of the transposon
insertion sites was performed using splinkerette PCR to produce barcoded
PCR products that were pooled and sequenced on the 454 GS-Titanium
sequencers (Roche) platform. Reads from sequenced tumors were mapped to
the mouse genome assembly National Center for Biotechnology Information
m37 and merged together to identify nonredundant SB insertion sites (SI
Materials and Methods).

ACKNOWLEDGMENTS. The authors thank Keith Rogers, Susan Rogers, and
the Institute for Molecular and Cell Biology Histopathology Core. We thank
Doug Melton for the Pdx1-Cre animals and Angela Chou, Anatomical Pa-
thologist at the Garvan Institute of Medical Research, for performing the
second blinded scoring for the tissue microarray. We also acknowledge Pear-
lyn Cheok, Nicole Lim, and Dorothy Chen for their help with tumor moni-
toring. This work was supported in part by the Biomedical Research Council,
Agency for Science, Technology, and Research, Singapore; the National
Health and Medical Research Council of Australia; the Queensland Govern-
ment; the Cancer Council New South Wales; the Cancer Institute New South
Wales; the Royal Australian College of Surgeons; the Australian Cancer Re-
search Foundation; the St. Vincent’s Clinic Foundation; the Avner Nahmani
Pancreatic Cancer Foundation; and the R. T. Hall Trust.

5940 | www.pnas.org/cgi/doi/10.1073/pnas.1202490109 Mann et al.

D
o
w

n
lo

a
d
e
d
 a

t 
C

a
rn

e
g
ie

 M
e
llo

n
 U

n
iv

e
rs

ity
 o

n
 A

p
ri
l 5

, 
2
0
2
1
 

http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1202490109/-/DCSupplemental/pnas.201202490SI.pdf?targetid=nameddest=STXT
http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1202490109/-/DCSupplemental/pnas.201202490SI.pdf?targetid=nameddest=STXT
www.pnas.org/cgi/doi/10.1073/pnas.1202490109


1. Jemal A, Siegel R, Xu J, Ward E (2010) Cancer statistics, 2010. CA Cancer J Clin 60:
277–300.

2. Stathis A, Moore MJ (2010) Advanced pancreatic carcinoma: Current treatment and
future challenges. Nat Rev Clin Oncol 7:163–172.

3. Maitra A, Fukushima N, Takaori K, Hruban RH (2005) Precursors to invasive pancreatic
cancer. Adv Anat Pathol 12:81–91.

4. Hansel DE, Kern SE, Hruban RH (2003) Molecular pathogenesis of pancreatic cancer.
Annu Rev Genomics Hum Genet 4:237–256.

5. Jones S, et al. (2008) Core signaling pathways in human pancreatic cancers revealed by
global genomic analyses. Science 321:1801–1806.

6. Campbell PJ, et al. (2010) The patterns and dynamics of genomic instability in met-
astatic pancreatic cancer. Nature 467:1109–1113.

7. Aguirre AJ, et al. (2003) Activated Kras and Ink4a/Arf deficiency cooperate to produce
metastatic pancreatic ductal adenocarcinoma. Genes Dev 17:3112–3126.

8. Hingorani SR, et al. (2003) Preinvasive and invasive ductal pancreatic cancer and its
early detection in the mouse. Cancer Cell 4:437–450.

9. Bardeesy N, et al. (2006) Both p16(Ink4a) and the p19(Arf)-p53 pathway constrain
progression of pancreatic adenocarcinoma in the mouse. Proc Natl Acad Sci USA 103:
5947–5952.

10. Bardeesy N, et al. (2006) Smad4 is dispensable for normal pancreas development yet
critical in progression and tumor biology of pancreas cancer. Genes Dev 20:
3130–3146.

11. Hingorani SR, et al. (2005) Trp53R172H and KrasG12D cooperate to promote chro-
mosomal instability and widely metastatic pancreatic ductal adenocarcinoma in mice.
Cancer Cell 7:469–483.

12. Hruban RH, et al. (2006) Pathology of genetically engineered mouse models of
pancreatic exocrine cancer: Consensus report and recommendations. Cancer Res 66:
95–106.

13. Copeland NG, Jenkins NA (2010) Harnessing transposons for cancer gene discovery.
Nat Rev Cancer 10:696–706.

14. Jackson EL, et al. (2001) Analysis of lung tumor initiation and progression using
conditional expression of oncogenic K-ras. Genes Dev 15:3243–3248.

15. Gu G, Dubauskaite J, Melton DA (2002) Direct evidence for the pancreatic lineage:
NGN3+ cells are islet progenitors and are distinct from duct progenitors. De-
velopment 129:2447–2457.

16. Gold DV, Karanjawala Z, Modrak DE, Goldenberg DM, Hruban RH (2007) PAM4-re-
active MUC1 is a biomarker for early pancreatic adenocarcinoma. Clin Cancer Res 13:
7380–7387.

17. Fujita H, et al. (2010) alpha-Smooth muscle actin expressing stroma promotes an ag-
gressive tumor biology in pancreatic ductal adenocarcinoma. Pancreas 39:1254–1262.

18. de Ridder J, Uren A, Kool J, Reinders M, Wessels L (2006) Detecting statistically sig-
nificant common insertion sites in retroviral insertional mutagenesis screens. PLoS
Comput Biol 2:e166.

19. Krzywinski M, et al. (2009) Circos: An information aesthetic for comparative ge-
nomics. Genome Res 19:1639–1645.

20. Starr TK, et al. (2009) A transposon-based genetic screen in mice identifies genes
altered in colorectal cancer. Science 323:1747–1750.

21. Starr TK, et al. (2011) A Sleeping Beauty transposon-mediated screen identifies mu-
rine susceptibility genes for adenomatous polyposis coli (Apc)-dependent intestinal
tumorigenesis. Proc Natl Acad Sci USA 108:5765–5770.

22. March HN, et al. (2011) Insertional mutagenesis identifies multiple networks of co-
operating genes driving intestinal tumorigenesis. Nat Genet 43:1202–1209.

23. Dupuy AJ, Akagi K, Largaespada DA, Copeland NG, Jenkins NA (2005) Mammalian
mutagenesis using a highly mobile somatic Sleeping Beauty transposon system. Na-
ture 436:221–226.

24. Futreal PA, et al. (2004) A census of human cancer genes. Nat Rev Cancer 4:177–183.
25. Wood LD, et al. (2007) The genomic landscapes of human breast and colorectal

cancers. Science 318:1108–1113.
26. Network TCGAR; Cancer Genome Atlas Research Network (2011) Integrated genomic

analyses of ovarian carcinoma. Nature 474:609–615.
27. IPA Ingenuity Systems. Available at http://www.ingenuity.com. Accessed August 17,

2011.
28. Sjöblom T, et al. (2006) The consensus coding sequences of human breast and co-

lorectal cancers. Science 314:268–274.
29. Lim KH, Counter CM (2005) Reduction in the requirement of oncogenic Ras signaling

to activation of PI3K/AKT pathway during tumor maintenance. Cancer Cell 8:381–392.
30. Singh A, et al. (2009) A gene expression signature associated with “K-Ras addiction”

reveals regulators of EMT and tumor cell survival. Cancer Cell 15:489–500.

31. Huang W, Sherman BT, Lempicki RA (2009) Systematic and integrative analysis of
large gene lists using DAVID bioinformatics resources. Nat Protoc 4:44–57.

32. Huang W, Sherman BT, Lempicki RA (2009) Bioinformatics enrichment tools: Paths
toward the comprehensive functional analysis of large gene lists. Nucleic Acids Res 37:
1–13.

33. van Haaften G, et al. (2009) Somatic mutations of the histone H3K27 demethylase
gene UTX in human cancer. Nat Genet 41:521–523.

34. Balakrishnan A, et al. (2007) Novel somatic and germline mutations in cancer candi-
date genes in glioblastoma, melanoma, and pancreatic carcinoma. Cancer Res 67:
3545–3550.

35. Damm F, et al. (2011) Prognostic importance of histone methyltransferase MLL5 ex-
pression in acute myeloid leukemia. J Clin Oncol 29:682–689.

36. Brett BT, et al. (2011) Novel molecular and computational methods improve the ac-
curacy of insertion site analysis in Sleeping Beauty-induced tumors. PLoS One 6:
e24668.

37. Edgar R, Domrachev M, Lash AE (2002) Gene Expression Omnibus: NCBI gene ex-
pression and hybridization array data repository. Nucleic Acids Res 30:207–210.

38. Barrett T, et al. (2011) NCBI GEO: Archive for functional genomics data sets—10 years
on. Nucleic Acids Res 39:D1005–D1010.

39. Stratford JK, et al. (2010) A six-gene signature predicts survival of patients with lo-
calized pancreatic ductal adenocarcinoma. PLoS Med 7:e1000307.

40. Manuyakorn A, et al. (2010) Cellular histone modification patterns predict prognosis
and treatment response in resectable pancreatic adenocarcinoma: Results from RTOG
9704. J Clin Oncol 28:1358–1365.

41. Ishiyama N, et al. (2010) Dynamic and static interactions between p120 catenin and E-
cadherin regulate the stability of cell-cell adhesion. Cell 141:117–128.

42. Casagolda D, et al. (2010) A p120-catenin-CK1epsilon complex regulates Wnt sig-
naling. J Cell Sci 123:2621–2631.

43. Lutz S, et al. (2007) Structure of Galphaq-p63RhoGEF-RhoA complex reveals a path-
way for the activation of RhoA by GPCRs. Science 318:1923–1927.

44. Chang DK, et al. (2009) Margin clearance and outcome in resected pancreatic cancer.
J Clin Oncol 27:2855–2862.

45. Biankin AV, et al. (2009) Expression of S100A2 calcium-binding protein predicts re-
sponse to pancreatectomy for pancreatic cancer. Gastroenterology 137:558–568.

46. Duan S, et al. (2012) FBXO11 targets BCL6 for degradation and is inactivated in dif-
fuse large B-cell lymphomas. Nature 481:90–93.

47. Harada T, et al. (2008) Genome-wide DNA copy number analysis in pancreatic cancer
using high-density single nucleotide polymorphism arrays. Oncogene 27:1951–1960.

48. Birnbaum DJ, et al. (2011) Genome profiling of pancreatic adenocarcinoma. Genes
Chromosomes Cancer 50:456–465.

49. Okami K, et al. (1998) Analysis of PTEN/MMAC1 alterations in aerodigestive tract
tumors. Cancer Res 58:509–511.

50. Sakurada A, et al. (1997) Infrequent genetic alterations of the PTEN/MMAC1 gene in
Japanese patients with primary cancers of the breast, lung, pancreas, kidney, and
ovary. Jpn J Cancer Res 88:1025–1028.

51. Hill R, et al. (2010) PTEN loss accelerates KrasG12D-induced pancreatic cancer de-
velopment. Cancer Res 70:7114–7124.

52. Kan Z, et al. (2010) Diverse somatic mutation patterns and pathway alterations in
human cancers. Nature 466:869–873.

53. Ding L, et al. (2008) Somatic mutations affect key pathways in lung adenocarcinoma.
Nature 455:1069–1075.

54. Schwickart M, et al. (2010) Deubiquitinase USP9X stabilizes MCL1 and promotes tu-
mour cell survival. Nature 463:103–107.

55. Tateossian H, et al. (2009) Regulation of TGF-beta signalling by Fbxo11, the gene
mutated in the Jeff otitis media mouse mutant. Pathogenetics 2:5.

56. Abida WM, Nikolaev A, Zhao W, Zhang W, Gu W (2007) FBXO11 promotes the
Neddylation of p53 and inhibits its transcriptional activity. J Biol Chem 282:
1797–1804.

57. Hay-Koren A, Caspi M, Zilberberg A, Rosin-Arbesfeld R (2011) The EDD E3 ubiquitin
ligase ubiquitinates and up-regulates beta-catenin. Mol Biol Cell 22:399–411.

58. Lee S, et al. (2006) Coactivator as a target gene specificity determinant for histone H3
lysine 4 methyltransferases. Proc Natl Acad Sci USA 103:15392–15397.

59. Wei Y, et al. (2008) Loss of trimethylation at lysine 27 of histone H3 is a predictor of
poor outcome in breast, ovarian, and pancreatic cancers. Mol Carcinog 47:701–706.

60. Jiao Y, et al. (2011) DAXX/ATRX, MEN1, and mTOR pathway genes are frequently
altered in pancreatic neuroendocrine tumors. Science 331:1199–1203.

61. Dupuy AJ, et al. (2009) A modified sleeping beauty transposon system that can be
used to model a wide variety of human cancers in mice. Cancer Res 69:8150–8156.

Mann et al. PNAS | April 17, 2012 | vol. 109 | no. 16 | 5941

M
ED

IC
A
L
SC

IE
N
C
ES

IN
A
U
G
U
R
A
L
A
R
TI
C
LE

D
o
w

n
lo

a
d
e
d
 a

t 
C

a
rn

e
g
ie

 M
e
llo

n
 U

n
iv

e
rs

ity
 o

n
 A

p
ri
l 5

, 
2
0
2
1
 

http://www.ingenuity.com