A conditional transposon-based insertional mutagenesis screen for genes associated with mouse hepatocellular carcinoma

Abstract

We describe a system that permits conditional mobilization of a Sleeping Beauty (SB) transposase allele by Cre recombinase to induce cancer specifically in a tissue of interest. To demonstrate its potential for developing tissue-specific models of cancer in mice, we limit SB transposition to the liver by placing Cre expression under the control of an albumin enhancer/promoter sequence and screen for hepatocellular carcinoma (HCC)-associated genes. From 8,060 nonredundant insertions cloned from 68 tumor nodules and comparative analysis with data from human HCC samples, we identify 19 loci strongly implicated in causing HCC. These encode genes, such as EGFR and MET, previously associated with HCC and others, such as UBE2H, that are potential new targets for treating this neoplasm. Our system, which could be modified to drive transposon-based insertional mutagenesis wherever tissue-specific Cre expression is possible, promises to enhance understanding of cancer genomes and identify new targets for therapeutic development.

Full text

A conditional transposon-based insertional mutagenesis
screen for genes associated with mouse hepatocellular
carcinoma
Vincent W Keng1,2, Augusto Villanueva3, Derek Y Chiang4,5, Adam J Dupuy6, Barbara J Ryan1,2, Ilze Matise1,
Kevin A T Silverstein1,7, Aaron Sarver1,7, Timothy K Starr1,2, Keiko Akagi8, Lino Tessarollo8, Lara S Collier9,
Scott Powers10, Scott W Lowe10, Nancy A Jenkins11, Neal G Copeland11, Josep M Llovet3,12,13 &
David A Largaespada1,2
We describe a system that permits conditional mobilization of a Sleeping Beauty (SB) transposase allele by Cre recombinase
to induce cancer specifically in a tissue of interest. To demonstrate its potential for developing tissue-specific models of cancer
in mice, we limit SB transposition to the liver by placing Cre expression under the control of an albumin enhancer/promoter
sequence and screen for hepatocellular carcinoma (HCC)–associated genes. From 8,060 nonredundant insertions cloned from
68 tumor nodules and comparative analysis with data from human HCC samples, we identify 19 loci strongly implicated in
causing HCC. These encode genes, such as EGFR and MET, previously associated with HCC and others, such as UBE2H, that
are potential new targets for treating this neoplasm. Our system, which could be modified to drive transposon-based insertional
mutagenesis wherever tissue-specific Cre expression is possible, promises to enhance understanding of cancer genomes and
identify new targets for therapeutic development.
Transposon-tagged mutagenesis has proven invaluable for functional
genomics in organisms such as Drosophila melanogaster1,2,butsimilar
progress in mammalian systems has been retarded by the long delay in
identifying transposons, such as Sleeping Beauty (SB), which are active
in mouse cells3. Although the low frequency of SB transposition in the
mouse germ line4–6 had suggested that the frequency of SB mobiliza-
tion in somatic tissues was too low to induce cancer, SB transposons
have been mobilized in somatic cells at frequencies high enough to
induce cancer in wild-type mice7and accelerate the formation of
tumors in p19Arf-deficient mice8. Nonetheless, it was not possible to
model specific tumor types when expressing SB from a ubiquitous
promoter. Whereas SB transposition in p19Arf-deficient mice acceler-
ated the formation of tumors normally observed in these mice8,in
wild-type mice, it resulted in the formation of aggressive hematopoietic
tumors that killed the animals by 4 months of age7. Mice did not live
long enough to develop other types of tumors such as solid tumors.
To address this need, we sought to develop a conditional SB
transposition system to screen for genes associated with different
types of cancer. For this, we first integrated the SB transposase
allele SB11 carrying a loxP-flanked (‘floxed’)-stop (lsl) cassette into
the mouse Rosa26 locus, which encodes a ubiquitously expressed
nonessential gene9. Genes inserted into the Rosa26 locus are expressed
in most tissues and not subject to epigenetic silencing normally
observed with transgenes9. Expression of the transposase knock-in
(Rosa26-lsl-SB11), which is normally blocked owing to the presence of
the floxed-stop cassette, can be reactivated in any target tissue using a
tissue-specific Cre recombinase to drive the transposition of the T2/
onc mutagenic transposon7,8. The T2/onc vector contains sequences
that can both cause misexpression of an oncogene and inactivate a
tumorsuppressorgene(Supplementary Fig. 1a online).
HCC is the third leading cause of cancer-related deaths globally10,
with potential curative treatment available for o30% of patients at the
time of diagnosis11. HCC is prevalent worldwide, with differences in
rates of its incidence reflecting regional diversity mostly related to the
geographic distribution of viral hepatitis11. A greater prevalence in
males, relative to females, has been explained by preliminary molecular
data12. Mutations in the TP53 gene are commonly found in HCC13,14.
The presence of many unexplained recurrent chromosomal abnormal-
ities and the identification of mRNA expression–based subsets of HCC
suggest the presence of unidentified genetic drivers of this disease15.
Received 16 December 2008; accepted 22 January 2009; published online 22 February 2009; doi:10.1038/nbt.1526
1Masonic Cancer Center, 2Center for Genome Engineering, University of Minnesota, Minneapolis, Minnesota 55455, USA. 3BCLC Group-Liver Unit, HCC Translational
Research Laboratory, IDIBAPS, CIBERehd, Hospital Clı
´nic, Barcelona 08036, Spain. 4Department of Medical Oncology and Center for Cancer Genome Discovery, Dana-
Farber Cancer Institute, Boston, Massachusetts 02115, USA. 5Cancer Program, The Broad Institute of Harvard and MIT, Cambridge, Massachusetts 02142, USA.
6Department of Anatomy and Cell Biology, University of Iowa, Iowa City, Iowa 52242, USA, 7Biostatistics and Informatics, University of Minnesota, Minneapolis, Minnesota
55455, USA. 8National Cancer Institute, Frederick, Maryland 21702, USA. 9Division of Pharmaceutical Sciences, School of Pharmacy, University of Wisconsin, Madison,
Wisconsin 53705, USA. 10Cold Spring Harbor Laboratory, Cold Spring Harbor, New York 11724, USA. 11Institute of Molecular and Cellular Biology, Singapore 138673,
Singapore. 12Mount Sinai Liver Cancer Program, Mount Sinai School of Medicine, New York, New York 10029, USA. 13Institucio
´Catalana de Recerca i Estudis Avanc¸ats,
Barcelona 08010, Spain. Correspondence should be addressed to D.A.L. (larga002@umn.edu).
264 VOLUME 27 NUMBER 3 MARCH 2009 NATURE BIOTECHNOLOGY
ARTICLES
© 2009 Nature America, Inc. All rights reserved.

We used a hepatocyte-specific albumin-Cre (Alb-Cre) transgene to
activate transposase expression in mice, specifically in the liver16.As
mutations in TP53 are the most frequently described mutations in
human HCC, we included a conditional dominant negative Trp53
transgene17 in a second construct, named p53-lsl-R270H (Supplemen-
tary Fig. 1a). We bred triple transgenic (Rosa26-lsl-SB11;T2/onc;Alb-
Cre) and quadruple transgenic (Rosa26-lsl-SB11;T2/onc;Alb-Cre; p53-
lsl-R270H) mice and monitored the onset of liver tumorigenesis in
both sets of lines (Supplementary Fig. 1b). When combined with
high-throughput sequencing, our conditional forward genetics screen
identified 19 genes potentially associated with oncogenesis in the liver
and prioritized epidermal growth factor receptor (Egfr)andubiquitin-
conjugating enzyme E2H (Ube2h) for experimental validation. Egfr was
the most frequently mutated gene in our screen and was especially
prevalent as a truncated form lacking the C-terminal half of the gene.
Moreover, initial human comparative studies suggest a nonsignificant
trend to higher tumor recurrence and poorer survival rates associated
with higher expression levels of UBE2H. This information enhances
insight into the genetic mechanisms associated with HCC and may
facilitate development of more effective therapies.
RESULTS
Hepatocyte-specific transposition and tumorigenesis
To demonstrate that the conditional transposase knock-in is activated
exclusively in the liver, we used an anti-SB transposase (SB) antibody
for immunohistochemical analyses of mice carrying both Alb-Cre and
Rosa26-lsl-SB11 transgenes (Fig. 1a). To confirm that transposition is
occurring in the livers of experimental transgenic animals, we used
excision PCR8to demonstrate amplicon excision (Supplementary
Fig. 2a online) when experimental and control animals from both
sexes were initially euthanized B100 d after birth. No visible lesions
were seen in any organs at this stage (data not shown).
Preneoplastic liver nodules were first detected in male triple (non-
predisposed genetic background) and quadruple (predisposed to HCC
by expression of a dominant-negative Trp53 allele) transgenic animals
B160 d after birth. Nodules in the quadruple transgenic animals were
larger and more numerous than those from triple transgenic animals
(Supplementary Fig. 2b). Double and triple transgenic mice carrying
all possible combinations of the four transgenes were also generated to
provide control cohorts. Throughout our studies, we saw no evidence
of tumorigenesis in control male littermates euthanized at a similar
age (data not shown).
Of six quadruple transgenic male experimental animals eutha-
nized between 101 and 223 d after birth, four (67%) had livers
with macroscopic preneoplastic nodules (Fig. 1b); a total of
67 nodules were isolated (Supplementary Table 1 online). In contrast,
we found evidence of neoplasms in only three (43%) of the
seven triple transgenic male animals euthanized between 105 and
289 d after birth and isolated 36 preneoplastic nodules from these
animals (Supplementary Table 1). Excision PCR assays confirmed
T
T
P
T
P
T
a
c
b
d
Figure 1 Accelerated tumorigenesis in p53-deficient livers compared with nonpredisposed livers of male mice. (a) Albumin-Cre (Alb-Cre) expression
efficiently deletes the floxed-stop (lsl) cassette within the Rosa26-lsl-SB11 transgene, permitting SB expression and subsequent somatic transposition. Left
panel, immunohistochemistry of (Alb-Cre; Rosa26-lsl-SB11) double-transgenic liver section treated without the primary anti-SB transposase (SB) antibody
(negative control). Right panel, immunohistochemistry of serial liver section treated with the primary anti-SB antibody. Sections were lightly stained with
hematoxylin after immunohistochemistry. Scale bar, 100 mm. (b) Liver from quadruple transgenic male mouse (159 d old), displaying many preneoplastic
nodules (scale bar in left panel, 0.5 cm). Middle (low magnification) and right (high magnification of boxed area from middle panel) panels show the
tumor histology of several adenomas using hematoxylin-eosin (HE) staining. These adenomas often compressed surrounding parenchyma. Cells within the
preneoplastic foci and adenomas were frequently vacuolated, containing distinct lipid vacuoles or clear cytoplasm (arrows). Nuclei were in the same size
or smaller than those in the normal hepatic parenchyma, and occasionally contained mitotic figures indicative of cell division. Adenomas were frequently
bordered by hepatocytes with markedly enlarged nuclei that were occasionally karyomegalic. T, tumor nodule; P, parenchymal liver cells; scale bars for
middle and right panels, 500 mmand100mm, respectively. (c) Liver from triple transgenic male mouse (330 d old) showing advanced tumor development.
Many large irregular nodules are well-vascularized (scale bar in left panel, 0.5 cm). Middle panel shows the HE histological section of one large neoplastic
nodule typical of hepatocellular adenoma consisting of variably vacuolated hepatocytes filled with lipid. Three arrows indicate the border between the
adenoma and nonneoplastic hepatic parenchyma (P), which is slightly compressed. Right panel shows high magnification of boxed area in the middle panel.
Note the enlarged nuclei of hepatocytes with moderate variation in nuclear size, prominent nucleoli, and mitotic figure (open arrow). T, tumor nodule;
P, parenchymal liver cells; scale bars in middle and right panels, 1,000 mmand50mm, respectively. (d) Liver with HCC (bottom, left panel) and lung
metastases (top, left panel) from triple transgenic male (440 d old; scale bar, 0.5 cm). HE staining of the liver (middle panel) and lung (right panel) show
advanced HCC in the liver and its metastasis into the lung. A partial HCC section reveals irregular trabeculae of neoplastic, diffusely necrotic hepatocytes
(black arrows) that are multifocally vacuolated. Trabeculae are separated by dilated sinusoids containing variable amounts of fibrin. The lung contains
multiple variably sized metastatic nodules of HCC (black arrows) that markedly compress the pulmonary parenchyma. Pulmonary alveoli are filled with many
foamy macrophages. Scale bars in middle and right panels, 100 mm.
NATURE BIOTECHNOLOGY VOLUME 27 NUMBER 3 MARCH 2009 265
ARTICLES
© 2009 Nature America, Inc. All rights reserved.

transposition in the livers of experimental animals that did not
produce tumors (Supplementary Fig. 2a).
Detailed histopathological analyses revealed that the livers of triple
and quadruple transgenic mice euthanized B150 d after birth
frequently contained preneoplastic foci of cellular alteration that
represents the earliest visible stage of neoplastic formation, with a
few adenomas (Fig. 1b). The liver of a 330-d-old triple transgenic
male mouse had multiple large well-vascularized tumors with micro-
scopic features of hepatic adenoma (Fig. 1c). Two older triple
transgenic male mice (440 and 460 d old) and a 432-d-old quadruple
transgenic male mouse displayed lung metastases as well as livers
with histopathological features of HCC (Fig. 1d and Supplementary
Tab l e 1 ). Immunohistochemical analysis of preneoplastic nodules
from all triple and quadruple transgenic livers stained positive for
SB, albumin and for the cellular proliferative marker, Ki67 (Fig. 2a).
This indicates that these nodules result from transposition events
originated from hepatocytes and have increased rates of proliferation.
The lung metastases also stained positive for SB, albumin and Ki67
using immunohistochemistry, indicating that they were derived from
the HCC (Fig. 2b).
Although RT-PCR revealed that the majority of preneoplastic
nodules expressed alpha-fetoprotein (Afp), a biomarker for human
HCC (Fig. 3), only a small subset of nodules expressed enough Afp to
enable detection by immunohistochemistry (data not shown).
RT-PCR also demonstrated the expression of secreted phosphoprotein
1(Spp1)—a gene associated with HCC metastasis18—in all preneo-
plastic nodules (Fig. 3d). Semi-quantitative RT-PCR demonstrated
upregulation of Ssp1 and Afp expression as liver tumorigenesis
progressed from adenoma to HCC (Supplementary Fig. 2c).
Immunohistochemical analyses for b-catenin levels demonstrated
increasing levels of expression as tumorigenesis progressed from
preneoplastic nodules to hepatic adenoma to HCC (Supplementary
Fig. 3 online). Mutations in the gene encoding b-catenin and its
elevated expression are also observed in human HCC19.Notably,triple
(n¼4) and quadruple (n¼4) transgenic female experimental
animals (euthanized 178–342 d after birth and 178–344 d after
birth, respectively) did not have any visible liver lesions (Supplemen-
tary Table 1). However, two female triple transgenic animals (512 and
575 d old) and one quadruple transgenic animal (432 d old) presented
livers with small preneoplastic nodules (Supplementar y Table 1). The
low frequency and prolonged latency of liver nodules in female
experimental animals mirrors the strong gender bias in the incidence
of HCC tumor seen in humans. Moreover, immunohistochemistry
revealed increased expression of Afp and the proliferative marker Ki67
in nontumorigenic liver sections from female mice (Supplementary
Fig. 4a online). Therefore, our conditional SB liver tumor model is
useful in elucidating genetic mechanisms for HCC tumorigenesis,
including lesions ranging from early hepatic adenomas to fully
developed HCC, including metastatic HCC.
Sequencing identifies common insertion sites in tumors
Supplementary Figure 1c online provides a flow chart for SB somatic
cell mutagenesis and barcode-assisted integration site amplification.
Briefly, T2/onc integration sites from 68 preneoplastic nodules (3 from
triple and 65 from quadruple transgenic animals) were cloned and
sequenced using barcoded primers and linker-mediated PCR. Subse-
quent pyrosequencing20 enables tens of thousands of T2/onc integra-
tion sites from a mixture of tumors to be characterized in a single
sequencing run (Supplementary Methods online). Pyrosequencing of
linker-mediated PCR products from these tumors generated over
140,000 individual sequences. Sequences containing o16 nucleotides
of genomic sequence were eliminated, leaving B106,000 sequences.
From these, 85,652 sequences were uniquely mapped at 95% identity
to the mouse genome. As SB has a tendency to ‘hop’ primarily within
the vicinity of the original site of integration6, we excluded insertions
that mapped to the transposon donor chromosome (chromosome
15). Further elimination of insertions that did not map to the
canonical TA insertion site required for SB integration21–23 left a
total of 68,782 sequences. We then combined all insertions that
mapped to the same TA dinucleotide and originated from the
same neoplastic nodule, leaving a final tally of the 8,060 non-
redundant insertions.
We next looked for regions in the genome that had more SB
insertions than predicted by random chance. These so-called common
insertions sites (CISs) are most likely to harbor disease-related genes.
SB Alb Ki67
Negative controlPrimary antibodyNegative controlPrimary antibody
SB Alb Ki67
a
b
T
P
T
P
P
T
M
P
M
P
M
P
Figure 2 Immunohistochemical analyses of liver adenomas. (a) Paraffin-
embedded liver tissue sections from triple or quadruple transgenic male
animals stained positive with antibodies against SB transposase (SB),
albumin (Alb) and the proliferative marker Ki67. Representative
immunohistochemical liver sections from a 160-d-old quadruple transgenic
male mouse are shown. Top panels, liver sections not treated with the
primary antibody (negative controls); bottom panels, serial liver sections
treated with primary antibody against the indicated protein; T, tumor nodule;
P, parenchymal cells; scale bars, 100 mm. (b) Immunohistochemical
analyses of the HCC-derived lung metastasis. Paraffin-embedded lung tissue
sections were stained with antibodies against SB, Alb and Ki67. Top panels,
lung sections not treated with the primary antibody (negative controls);
bottom panels, serial lung sections treated with the indicated primary
antibody; P, parenchymal lung cells; M, metastasis from HCC;
scale bars, 100 mm.
266 VOLUME 27 NUMBER 3 MARCH 2009 NATURE BIOTECHNOLOGY
ARTICLES
© 2009 Nature America, Inc. All rights reserved.

Based on Monte Carlo criteria for statistical significance (Supplemen-
tary Methods), we defined CISs as regions in the genome with six
insertions located within 130 kb of each other, five insertions within
65 kb or four insertions within 20 kb. In total, 30 CISs were identified
according to these criteria. Of these CISs, 11 appear to represent
background events resulting from false priming at a specific site. This
is because the T2/onc insertions either all begin at the same nucleo-
tide, occur in loci with no annotated genes or are present among CISs
defined by control insertion-site-mapping experiments using 3-week-
old transgenic-mouse tail DNA carrying both the T2/onc and Rosa26-
SB11 transgenes (Supplementary Methods). The final list of CISs
associated with mouse HCC is shown in Ta b l e 1 and the 8,060
nonredundant sites of insertion are provided in Supplementary Data
online. Notably, substantial overlap with this CIS list was seen in
another set of liver tumors induced by a villin-Cre transgene (data not
shown), further attesting to the significance of these genes for HCC.
Villin is expressed in the microvilli of brush border epithelium lining
of the gut and renal tubes in vertebrates. Importantly, the specific
insertion sites associated with individual preneoplastic nodules during
early tumorigenesis differed for each nodule, indicating that each
nodule is a unique clone. In general, each preneoplasic nodule was
characterized by a unique set of T2/onc insertions. Certain genes, such
as Egfr, were reproducibly mutated by insertion mutations in nodules
from the same mouse. However, these insertions are not in identical
TNFEGFR
MET
Transphosphorylate
PI3K
PAK 4
NFIB
TAO K 3
HIF1A
SFI1
UBE2H
QKI
JNK
AKTMAP2K4
20 Kb
(X2) (X3) (X4)
(X5) (X2)
(X13)
(X9)
(X4)
(X2)
(X3)
200 bp
Ex24 Ex25
SA
SA
polyA
polyA
MSCV SD
Wild-type (713 bp)
Various sizes depending
on insertion site
B6 H2OTumor DNA samples
**
MW
Afp
Egfr
Truncated Egfr
SB
Actb
Tumor samples
NRAS liver tumor
Normal SB liver
Alb
Ssp1
Adv. tumors
Erbb2
MW
Tail
Liver
Lung 1
Lung 2
Tail
Liver 1
Lung
Nodule
Liver 2
B6
H2O
ATR M71 ATR M81
Ex
T2/onc/Egfr
Gapdh
SA polyA MSCV SD
SA polyA
a
c
de
b
Figure 3 Frequent mutagenic transposon
insertions into epidermal growth factor
receptor (Egfr)andEGFR interacting
genes associated with HCC. (a) Ingenuity
pathway analysis using 17 human
homologs of the CIS genes obtained from
the liver cancer screen. Of the 17 genes
entered, 10 genes were referenced and
displayed in the network function
pathways associated with post-translational modification, cancer and tumor morphology. The EGFR signaling pathway shows interactions with JNK, TNF
and PI3K/AKT regulatory pathways. CIS genes are in black and other genes in this network are in blue. (b) Representation of insertions of the mutagenic
transposon (T2/onc) into intron 24 of Egfr. Red triangles, inverted repeats/direct repeats (IR/DR) transposon flanking sequences; SA, splice acceptor; polyA,
polyadenalytion signal; MSCV, long terminal repeat of the murine stem cell virus; SD, splice donor; open arrowhead, sense-orientated insertion of the T2/onc
relative to the Egfr gene; arrowhead, anti-sense orientated insertion of the T2/onc relative to the Egfr gene; arrows, endogenous and vector primers used for
the Egfr PCR genotyping shown in c. Numbers in parentheses indicate the frequency of transposon insertions at each particular site from different liver
preneoplastic nodules. (c) Confirmation of transposon insertions in intron 24 of Egfr. PCR genotyping was performed using genomic DNA isolated from
individual tumor nodules. A subset of these samples was subjected to Egfr PCR genotyping using endogenous and vector primers. Gel electrophoresis shows
the endogenous Egfr (713 bp) band (open arrowhead), with transposon-integrated bands of varying sizes, depending on the insertion site within intron 24.
Except for two insertion sites (asterisks) missed by pyrosequencing, all amplicons corresponded with the pyrosequencing data. MW, 100-bp molecular
standard; B6, C57BL/6 tail genomic DNA; H2O, double-distilled water (negative control). (d) RT-PCR analyses of tumor nodules. All neoplastic nodules were
positive for SB transposase (SB) and albumin (Alb) transcripts, indicating transposition and that nodules were derived from hepatocytes, respectively. Most
tumor nodules were positive for alpha-fetoprotein (Afp) transcripts, a clinical marker for HCC, and secreted phosphoprotein 1 (Ssp1), which is overexpressed
in various cancers including HCC. Nodules taken from a 330-d-old triple transgenic male mouse with advanced tumors (shown in Fig. 1c) were strongly
positive for Afp and Ssp1. All tumor nodules tested were positive for endogenous Egfr and for truncated-Egfr transcripts. NRAS liver tumor, HCC control
taken from a tumorigenic liver overexpressing NRAS G12V oncogene25;SB normal liver, normal liver taken from a SB-expressing mouse; beta-actin (Actb),
control to show equal loading of mRNA used for RT-PCR. RT-negative controls were also performed for each sample and no visible bands were seen for any
of the markers tested (data not shown). (e) Confirmation of transposition events and transposon insertions in intron 24 of Egfr for HCCs and lung metastases.
PCR genotyping was performed with genomic DNA isolated from the tails, livers and lung metastases of two triple transgenic male mice (ATR M71, 440 d
old and ATR M81, 460 d old). Top panel, excision PCR assays (Ex) for transposition events in the lung metastases and HCCs (open arrowhead). No excision
was detected in the tails of the triple transgenic male mice. Middle panel, PCR genotyping using only the endogenous Egfr forward and T/JB3 primers
(T2/onc/Egfr) to confirm transposon insertion in intron 24 of Egfr for the lung metastases and HCCs. Gel electrophoresis demonstrates the transposon-
integrated band (arrowhead) for both the lung metastases (lung) and HCCs (liver), but not in their tails. Gapdh, control to show equal use of genomic DNA
template (100 ng) in PCR reactions. Nodule, a liver tumor nodule from a different animal was used to compare different transposon insertion sites; MW,
100-bp molecular standard; B6, C57BL/6 tail genomic DNA; H2O, double-distilled water (negative control).
NATURE BIOTECHNOLOGY VOLUME 27 NUMBER 3 MARCH 2009 267
ARTICLES
© 2009 Nature America, Inc. All rights reserved.

TA dinucleotides, with a few exceptions. We therefore conclude that
each preneoplastic nodule was derived from an independent event
resulting from random transposon insertional mutagenesis events. In
some cases, identical Egfr gene insertions did occur in separate
nodules, but as all other insertions were different in those samples,
we concluded that the identical TA dinculeotide insertions into Egfr
had occurred by chance owing to the strong selective pressure for
insertions in intron 24 (see below). We previously observed T2/onc
insertions into identical TA dinucleotides in Braf and Notch1 in
independent tumors in situations of strong positive selection for
insertion into a specific part of an oncogene7,8. In contrast, our
lung metastasis analysis, described below, demonstrates that clonal
relationships can be detected between primary tumors and metastatic
derivatives because identical T2/onc insertions occur in individual
metastasis samples and a primary liver HCC tumor taken from the
same mouse.
Pathway analysis of select CIS genes
Ingenuity Pathways Analysis (IPA) is a software application that
enables network and functional analyses of gene sets of interest
based on a repository of molecular interactions, regulatory events
and gene-to-phenotype associations culled from the life sciences
literature. The application can determine cellular and disease pheno-
types most significant to a set of genes and can build molecular
networks based on literature findings and pathways. Therefore, we
used IPA to obtain a better understanding of the possible pathways
and interactions between CIS genes. Of the 17 CIS genes analyzed, the
three most significant signaling or disease functional annotations are
post-translational modification (P¼4.61E-09), cancer (P¼8.09E-06)
and tumor morphology (P¼8.09E-06) (Supplementary Table 2
online). The CIS list includes homologs of several human genes that
have been implicated in tumor formation and apoptosis of tumor cell
lines: EGFR,HIF1A,MAP2K4,MET,PAK 4,VRK2,TRPM7 and
TAOK3. IPA identified two network pathways overrepresented by
human homologs of CIS genes. The first network includes two
transcription factors (NFIB and HIF1A) and the second pathway
involves genes that interact with TNF. The combined pathways from
IPA are summarized in Figure 3a.
Frequent transposon insertions occur in Egfr
Transposon insertions in Egfr were detected in 85% (n¼58) of
preneoplastic liver nodules isolated from experimental animals. These
transposon insertions were most frequently detected in intron 24 of
Egfr (Ta b l e 1 and Fig. 3b) and in the antisense orientation, suggesting
truncation of the gene product. Three-primer PCR genotyping using
endogenous Egfr and transposon primers performed with genomic
DNA isolated from individual tumor nodules confirmed the presence
of transposon vectors in this locus (Fig. 3c). RT-PCR also confirmed
the presence of the predicted truncated Egfr transcript in these
preneoplastic nodules (Fig. 3d).
As Egfr insertions were also identified in preneoplastic nodules
taken from a triple transgenic mouse, Egfr mutations also appear to
contribute to tumorigenesis in a nonpredisposed genetic background.
The insertions in this animal are predicted to result in a truncated
Egfr protein (about 984 amino acids) containing the majority of the
kinase domain but lacking the C-terminal domain. Indeed, this
truncated Egfr was detected by western blot analysis in the liver
tumors of older experimental triple transgenic male mice (Supple-
mentary Fig. 4b).
Lung metastases derived from HCC
Analysis of genomic DNA taken from metastases of two triple
transgenic male mice also demonstrated transposon insertion in
intron 24 of Egfr, indicating that they were derived from the HCCs
(Fig. 3e). Thirty-two additional lung metastatic nodules were isolated
from a 432-d-old quadruple transgenic male. Insertion sites from
these metastatic nodules were compared to three individual HCC
nodules taken from the same animal to identify a clonal relationship
between primary liver tumors and metastases, and between metas-
tases. One of the liver HCCs (HCC3) seemed to share a common
ancestor with a second HCC (HCC2) as both have identical Egfr gene
insertions, which are distinct from the Egfr insertion in HCC1
Table 1 Common insertion sites for HCC-associated genes
Gene Chr Position Range nNodules Mouse
Egfr 11 16765887-16872714 107 kb 69 58 5
Novel EST gene cluster 9 3000138-3038047 38 kb 20 17 5
Sfi1 11 3046719-3136227 90 kb 13 13 4
Zbtb20 16 43349510-43460987 111 kb 8 7 4
ENSMUSESTG0000001569 10 52995507-53074648 79 kb 7 5 2
Nfib 4 82058117-82133086 75 kb 7 7 4
Tao k3 5 117614813-117701538 87 kb 7 6 3
Slc25a13 6 6047524-6159681 112 kb 7 7 2
Qk 17 10379929-10457807 78 kb 6 6 3
Rnf13 y3 57552266-57663120 111 kb 6 6 3
Met 6 17449763-17545224 95 kb 6 6 3
March1 8 68422058-68551102 129 kb 6 5 2
Psd3 8 70451840-70580359 129 kb 6 6 3
Map2k4 11 65524193-65586089 62 kb 5 5 4
Trpm7 2 126659349-126720778 61 kb 5 5 3
Ube2hy6 30181012-30207531 27 kb 5 4 3
Vrk2y11 26373044-26373912 869 bp 4 3 3
Hif1a 12 75021346-75031073 10 kb 4 4 2
Pak4y7 29367702-29371179 3 kb 4 4 3
Chr, chromosome; Range, chromosomal position of transposon insertions; n, frequency of transposon insertions; y, genes that did not have any transposon insertions from liver tumors
generated with the villin-Cre transgenic mice used in the gastrointestinal cancer study (data not shown). Nodules, number of preneoplastic nodules from which the CIS was
determined; Mouse, number of mice from which the nodules were isolated. Position based on the Ensembl NCBI m37 April 2007 mouse assembly.
268 VOLUME 27 NUMBER 3 MARCH 2009 NATURE BIOTECHNOLOGY
ARTICLES
© 2009 Nature America, Inc. All rights reserved.

(Fig. 4a). Most of the metastases share four additional insertions with
HCC2, indicating that the metastases share a common ancestor with
HCC2. Three additional insertion mutations were found in most of
the metastases (Fig. 4a). From the phylogenetic tree generated from
the insertion sites (Fig. 4b and Supplementary Methods), primary
liver tumor HCC2 and all lung metastases have the closest common
ancestor, suggesting that the lung metastases are actually derived from
liver tumor HCC2. These preliminary data suggest that SB-induced
tumorigenesis allows one to derive clonal relationships between
primary and metastatic derivatives, and to discover metastases-specific
insertion mutations that may drive this biological process.
Comparison with human hepatocellular carcinoma samples
Representative oligonucleotide microarray analysis (ROMA) of 100
human HCCs showed that increases or decreases in copy numbers of
17 human homologs of our CIS genes have been associated with
human HCC (Supplementary Table 3 online). We predict the effects
of transposon insertions on CIS gene expression in Supplementary
Tab l e 3 and Supplementary Methods. Genes with distinct copy
number gains identified in human HCC samples (n¼100), homologs
of which are also disrupted in our mouse model, include EGFR,
SLC25A13,MET and UBE2H. Genes with distinct copy number losses
in human HCC samples, homologs of which were also identified as
mouse CIS genes in our analysis, include MARCH1,PSD3,MAP2K4
and NFIB.
We also analyzed another cohort of 132 human samples spanning
the whole spectrum of human hepatocarcinogenesis: normal liver
(n¼10), cirrhotic liver (n¼13), low-grade dysplastic nodules
(n¼10), high-grade dysplastic nodules (n¼8) and HCC
(n¼91). Fifteen of the CIS genes were analyzed by combined single
nucleotide polymorphism (SNP) and gene expression arrays. The
most appealing candidates for clinical correlations were selected
based on recurrent gene copy number changes, and correlated gene
expression changes were compared with control samples (Supple-
mentary Methods). Of the 15 genes, only three— MAP2K4,QKI and
UBE2H—satisfy these criteria. MAP2K4 and QKI have losses of DNA
copy numbers with reduced mRNA levels, whereas UBE2H has DNA
copy number gains with a substantial increase in mRNA levels
(Supplementary Fig. 5a online). Associations between MAP2K4,
QKI and UBE2H expression and clinicopathological variables were
analyzed in 82 hepatitis C–related HCC patients treated with liver
resection (Supplementary Methods). Although, owing to the small
sample population, these genes did not display a significant difference
in outcome measured by tumor recurrence or survival, high expres-
sion levels of UBE2H displayed a nonsignificant trend toward lower
survival rates (P¼0.09) compared with low expression levels
(Supplementary Fig. 5c). Studies involving the tyrosine kinase
receptors EGFR and MET, both located on chromosome 7, recently
showed that copy number gains of this chromosome are frequently
associated with HCC and define a molecular class of HCC patients24
(Supplementary Fig. 5b).
Functional validation of two CIS genes
As UBE2H seemed a strong candidate HCC oncogene, we used a cell
proliferation assay to test its oncogenic potential. AML12 cells (adult
mouse hepatocyte cell line, transgenic for human TGFA) stably
transfected with a Ube2h expression vector have a higher proliferative
rate than normal untransfected cells or AML12 cells transfected with
an empty vector (Supplementary Fig. 6 online).
We used the Fah-deficient mouse model25 to test whether the
truncated form of EGFR could contribute to neoplastic growth
in vivo. In this assay, a test transgene is codelivered with an Fah
expression vector to allow selective repopulation of genetically trans-
formed hepatocytes in vivo under conditions that would normally
kill hepatocytes. Two vectors were generated: one (pT2/FAHIL) co-
expresses Fah and firefly luciferase, whereas the other (pT2/PGK-
Truncated EGFR) expresses a truncated form of EGFR (exon 1 to
exon 24) only (Fig. 5a). We used tail-vein hydrodynamic injection26
to administer the vectors to Fah-deficient mice that express the
SB11 transposase knocked into the Rosa26 locus (Fah/SB11 mice).
Upon withdrawal of NTBC (nitisinone, Orfadin), the mice under-
went liver repopulation, as evidenced by stable weight gain and
increasing luciferase expression (Fig. 5b). When a mouse injected
with both pT2/FAHIL and pT2/PGK-Truncated EGFR was eutha-
nized 43 d after injection, several patches of hyperplastic nodules
were visible in the liver (Fig. 5c). RT-PCR revealed that these
nodules express Fah and the truncated form of EGFR (Fig. 5d).
Although immunohistochemistry confirmed the inability to detect
EGFR in normal Fah-deficient liver (Fig. 5e), it also confirmed
that induced hyperplastic liver nodules co-express Fah and EGFR
Chrom 11 - 16809602 (Egfr)
Chrom 13 - 3539376 (Gdi2)
Chrom 14 - 41925151 (Mat1a)
Chrom 4 - 109135602 (Rnf11)
Chrom 7 - 13719020 (V1rk1)
Chrom 10 - 36683084 (Hdac2)
Chrom 19 - 40609911 (Sorbs1)
Chrom 5 - 90663715 (Ankrd17)
HCC Lung metastases
ab
HCC1
HCC3
HCC2
Lu16
Lu10
Lu13
Lu21
Lu11
Lu12
Lu14
Lu22
Lu19
Lu28
Lu24
Lu18
Lu27
Lu29
Lu17
Lu9
Lu26
Lu23
Lu25
Lu15
Lu6
Lu20
Lu5
Lu30
Lu31
Lu7
Lu8
Lu4
Lu1
Lu2
Lu3
Lu32
Lu32
Lu6
Lu3
Lu4
Lu1
Lu22
Lu2
Lu11
Lu13
Lu10
Lu24
Lu20
Lu19
Lu28
Lu29
Lu17
Lu26
Lu21
Lu27
Lu18
Lu9
Lu7
Lu8
Lu31
Lu30
Lu25
Lu23
Lu15
Lu5
Lu14
Lu12
HCC2
HCC3
HCC1
Lu16
Figure 4 SB-induced tumorigenesis reveals clonal relationships between primary and metastatic
derivatives. (a) The clonal relationship between lung metastases and HCC samples. The heat-map
was generated by mapping insertion sites from 32 lung metastases nodules and 3 HCC nodules
taken from the same mouse (ATRP M232). Importantly, three additional insertion mutations were
common in most of the metastases, indicating genes potentially involved in metastasis. Red,
insertion detected at the indicated locus; black, no insertion detected at the indicated locus.
(b) Phylogenic tree generated from the insertion sites of the lung metastases and HCCs.
NATURE BIOTECHNOLOGY VOLUME 27 NUMBER 3 MARCH 2009 269
ARTICLES
© 2009 Nature America, Inc. All rights reserved.

(Fig. 5f,g). Notably, adjacent liver tissue, which appeared healthy,
was negative for both transcripts (Fig. 5d).
DISCUSSION
The recent development of target-based therapeutics for treating
cancer has sparked a worldwide effort to identify all of the genes
and signaling pathways that cause it. But despite the potential of
transposon-based insertional mutagenesis for identifying cancer genes,
it has been impossible to control transposition in a manner that allows
different types of cancer to be modeled. We used a conditional SB
allele and a hepatocyte-specific Cre recombinase to screen for HCC-
associated genes in mice. As expected, quadruple transgenic mice
displayed more numerous and larger tumor nodules than triple
transgenic animals, as a result of the Trp53 mutant background,
IRES LucFahPGK
IR/DR IR/DR
pA
Truncated EGFR (Ex 1 to Ex 24)PGK
IR/DR IR/DR
pA
pT2/PGK-
Truncated EGFR
pT2/PGK-FAHIL
Fah /SB11 M84 nodule 1
EGFR
Ex24/polyA
Fah
Actb
RT (+) RT (–)
Fah /SB11 M84 nodule 2
Fah /SB11 M84 nodule 3
Fah /SB11 M84 normal liver
Normal liver
Fah /SB11 M84 nodule 1
Fah /SB11 M84 nodule 2
Fah /SB11 M84 nodule 3
Fah /SB11 M84 normal liver
Normal liver
HE EGFR Negative control
Fah EGFR
Ne
g
ative control
HE
Ne
g
ative control
1.5
1.0
0.5
×106
a
d
fg
e
bc
Figure 5 The Fah-deficient mouse model validates the oncogenic potential of truncated
EGFR.(a) Vectors used for tail vein hydrodynamic injection. pT2/PGK-Truncated EGFR,
truncated EGFR cDNA (exon 1 to exon 24) under the control of the phosphoglycerate
kinase (PGK) promoter and flanked by SB inverted repeat/direct repeat (IR/DR) recognition
sequences essential for transposition. pT2/PGK-FAHIL, Fah cDNA under the control of the
PGK promoter and fused with an IRES-luciferase (Luc) reporter gene, flanked by SB IR/
DRs. (b) Luciferase activity in Fah/SB11 M84 taken 15 d after injection with pT2/PGK-
Truncated EGFR and pT2/PGK-FAHIL. Exposure time was 5 s. (c) Examination of the
abdominal cavity of animal Fah/SB11 M84 revealed many patches of small hyperplastic
liver nodules (arrowheads). These nodules were carefully removed for RNA extraction and
subsequent RT-PCR analyses. Adjacent normal liver tissue was analyzed for comparison.
(d) RT-PCR analyses of the liver nodules and adjacent normal tissue. Liver hyperplastic nodules expressed both Fah and the truncated form of EGFR,
whereas the adjacent normal tissue was negative for both transcripts. RT (+), first strand cDNA synthesis with reverse transcriptase added; RT (–), first strand
cDNA synthesis without reverse transcriptase. (e) Normal histology of Fah-deficient liver (hematoxylin-eosin stain, HE) and inability to detect EGFR by
immunohistochemical staining. EGFR, treated with EGFR primary antibody; negative control, serial section not treated with the indicated primary antibody.
Scale bars, 100 mm. (f) Histology of liver hyperplastic nodules induced by truncated form of EGFR using HE staining. Top panel, the capsular surface of the
liver was irregularly nodular (arrow), but overall hepatic architecture was preserved with regularly spaced central veins and portal tracts. Scale bar, 500 mm.
Bottom panel, a portion of hepatic lobule containing variably sized hepatocytes with two cytomegalic and karyomegalic hepatocytes in the center, oneof
which is binucleated (arrows). Occasional hepatocytes have vacuolated cytoplasm. Hepatic cords are not evident due to cellular crowding. Scale bar,50mm.
(g) Representative hyperplastic nodule (enclosed within dashed circular line) within hepatic parenchyma comprising closely packed sheets of variably sized
hepatocytes, including a karyomegalic cell. Note the mild compression of the surrounding hepatic parenchyma. Scattered neutrophils and lymphocytes, and
mild extramedullary hematopoiesis suggested a low degree of inflammation. Immunohistochemical analyses of serial liver sections treated with the indicated
primary antibody confirmed the co-expression of Fah and EGFR in liver nodules. Most of the hepatocytes within the hyperplastic nodule (enclosed within
dashed circular line) expressed Fah. Hepatocytes within the hyperplastic nodule (enclosed within dashed circular line) and within surrounding parenchyma
stained weakly for EGFR. EGFR staining is also prominent in the cytoplasmic membranes of cells bordering sinusoids. Negative control, serial sections not
treated with the indicated primary antibody. Scale bars, 100 mm.
270 VOLUME 27 NUMBER 3 MARCH 2009 NATURE BIOTECHNOLOGY
ARTICLES
© 2009 Nature America, Inc. All rights reserved.

which predisposes animals to cancer. Our conditional SB liver-tumor
model is useful in elucidating genetic mechanisms for all stages of
HCC tumorigenesis—from early hepatic adenoma to fully developed
HCC, including metastasis.
Pyrosequencing technology facilitates the use of transposons for
cancer gene identification by enabling amplification and sequencing of
tens of thousands of SB insertion sites from a mixture of tumors in a
single sequencing run. From 8,060 nonredundant insertions subse-
quently cloned from 68 tumor nodules, we identify 19 loci that seem
strongly implicated in HCC. Our list of genes with multiple examples
of CISs includes several homologs of human genes (e.g., EGFR,
HIF1A,MAP2K4,MET,PAK4,VRK2,TRPM7 and TAOK 3)that
have been implicated in tumor formation and apoptosis of tumor
cell lines. IPA identified two network pathways overrepresented by
these homologs of CIS genes. The first network includes two tran-
scription factors, NFIB and HIF1A, which are capable of transducing
EGFR-initiated phosphorylation-signaling cascades. HIF1A has also
been suggested to play a role in tumor vascularization27. The second
pathway involves genes that interact with TNF.TNF can induce
tyrosine phosphorylation and internalization of EGFR, playing a
critical role in NF-kB activation28.NF-kB, in turn, plays an important
role in regulating apoptosis during liver tumorigenesis29.
Transposon insertions in mouse Egfr that cause truncations in the
C-terminal half of the gene product were common in SB-induced liver
tumors. Deletions of the C-terminal domain of human EGFR (966–
1006) have been shown to increase both autokinase activity and
transforming ability in vitro and in vivo30. Internal deletions in the
C terminus of EGFR have also been detected in naturally occurring
EGFR mutants displaying tumorigenic properties30–32,probably
resulting in constitutively active forms of the protein owing to the
destabilization of the inactive EGFR monomeric complex33.Ithas
been suggested that truncated Egfr can form a heterodimer with Erbb2
and transphosphorylate the tyrosine sites34. Tyrosine-phosphorylated
Erbb2 could then lead to the activation of other signaling pathways by
different mechanisms and may play a role in HCC tumorigenesis35.
Besides HCC, EGFR overexpression has also been associated with
humanbreastandgutcancers
36–39.EGFR is overexpressed in 15–40%
of human HCCs and EGF signaling is activated in B50% of human
HCCs39,40. Although extra copies of EGFR were seen in 17 out of 38
(45%) HCC tumors, increased expression did not correlate with the
increase in EGFR copy number36. Recent findings suggest that EGF
signaling could even be related to HCC development based on
significant differences in EGF genotype prevalence according to the
risk of developing HCC41. In addition, use of erlotinib (Tarceva) to
specifically target EGFR has shown interesting preliminary results in
phase 2 clinical trials in human HCC15,42.
The genes EGFR,SLC25A13,MET and UBE2H identified using CISs
from our mouse model all showed distinct increases in copy number
in human HCC samples. EGFR and MET are known proto-oncogenes,
whereas SLC25A13 and UBE2H may have novel oncogenic activities in
HCC. MET encodes the tyrosine kinase receptor for HGF and is
overexpressed in HCC43. Although our algorithm predicted a Met
gene disruption in our SB-induced tumors (Supplementary Table 3),
we suspect that these insertions actually activate the oncogenic activity
of Met; five of six insertions could produce a kinase domain–
containing truncated protein or activate the gene by enhancer inser-
tion. It is also possible that loss of function of Met contributes to
tumor development as Met knockout mice are more prone to
developing liver tumors44. Genes with distinct copy number losses
in human HCC samples whose homologs were also identified as CIS
genes by our mouse model, include MARCH1,PSD3,MAP2K4 and
NFIB.MAP2K4 has been identified as a putative tumor-suppressor
gene in human solid tumors of breast, prostate and pancreas, and may
have a similar function in the liver45–47. Although PSD3 and MARCH1
have not been shown to be involved in cancer, based on data presented
here, they may have tumor-suppressor activity in HCC. Interestingly,
the transcription factor NFIB is known to be upregulated in hepatitis-
induced HCC48. Another interesting finding is that a large number of
the CIS genes have human homologs that map to chromosome 7,
which has copy number amplifications in 415% of human
HCCs49,50. Moreover, when another cohort of 132 human samples
spanning the whole spectrum of human hepatocarcinogenesis was
compared with 15 human homologs of the CIS genes by combined
SNP and gene expression arrays, preliminary results indicated a
nonsignificant trend to higher tumor recurrence and poorer survival
rates associated with higher expression levels of UBE2H.Our
validation experiments and human comparative studies suggest a
role for UBE2H in liver tumorigenesis. Furthermore, validation
experiments confirmed the contribution of truncated EFGR to neo-
plastic growth in vivo.
A molecular classification of HCC based on gene copy number
alteration and expression profiling was recently proposed24. The five
classes, based on hierarchical clustering of gene expression data, are
b-catenin, proliferation- and interferon-related neoplasms, a novel
class of neoplasm defined by polysomy of chromosome 7 and an
unannotated category. Although we did not recover recurrent inser-
tions in Ctnnb1 or homologs of any of the several human genes known
to be implicated in HCC, we did observe an increase in b-catenin
protein expression. We plan to use mRNA microarray profiling of SB-
induced HCC to clarify whether our system models one or more of
the non-CTNNB1 subclasses of HCC. Regardless, based on our
comparison of CIS genes to gene copy number and expression changes
in human HCC, it appears that homologs of three of the genes on our
list—UBE2H, QKI and MAP2K4—are strong candidates for driving
HCC. Moreover, homologs of several of the other CIS genes—
including MET,EGFR and HIF1A—have been studied specifically in
the context of human HCC and are likely to play a role in the
development of this disease. Taken together, this indicates that the SB
screen yields a high fraction of relevant events in human HCC.
These studies, combined with others showing that conditional
transposon-based insertional mutagenesis can be used to model
solid tumors in other organ sites such as brain and gastrointestinal
tract (data not shown), define a powerful new method for dissecting
the cancer genome and for developing better treatments for
cancer. Future research directions include using this technology for
further validation of both the HCC- and metastasis-associated genes
identified here.
METHODS
Generation of transgenic animals. Alb-Cre transgenic animals were pur-
chased from Jackson Laboratory16. They were initially bred with T2/onc
homozygotes to obtain doubly transgenic animals carrying both Alb-Cre
and T2/onc. The T2/onc transgenic line with the donor concatemer on
chromosome 15, generated as previously described8, was used in this study.
Simultaneously, transgenic animals heterozygous for Rosa26-lsl-SB11 and
p53-lsl-R270H (purchased from NCI, Frederick Mouse Repository) were
interbred to obtain doubly transgenic animals. The two doubly trans-
genic lines were finally interbred to generate the required triple (Alb-Cre/
T2/onc/Rosa26-lsl-SB11), quadruple (Alb-Cre/T2/onc/Rosa26-lsl-SB11/p53-
lsl-R270H) and control animals of various transgene combinations. The
genetic background of these animals was mixed, allowing for a diverse
genetic population analysis.
NATURE BIOTECHNOLOGY VOLUME 27 NUMBER 3 MARCH 2009 271
ARTICLES
© 2009 Nature America, Inc. All rights reserved.

PCR genotyping. Identification of the various genotypes from both adult
transgenic animal and pups was performed as follows. First, genomic DNA was
isolated from tail clippings using standard proteinase-K treatment, phenol-
chloroform extraction and ethanol precipitation. Genomic DNA was then
dissolved in sterile TE (10 mM Tris-HCl (pH 7.5), 1 mM EDTA (pH 8)) and
quantified using a Nanodrop spectrophotometer. PCR genotyping was per-
formed using 100 ng of diluted genomic DNA as template. PCR primers used
for Alb-Cre were forward 5¢-CACACTGAAATGCTCAAATGGGAGA-3¢and
reverse 5¢-GGCAAATTTTGGTGTACGGTCAGTA-3¢(amplicon 456 bp); T2/
onc forward 5¢-CGCTTCTCGCTTCTGTTCGC-3¢and reverse 5¢-CCACCCC
CAGCATTCTAGTT-3¢(amplicon 264 bp); Rosa26-lsl-SB11 were Rosa26 wild-
type forward 5¢-CTGTTTTGGAGGCAGGAA-3¢,Rosa26 wild-type reverse 5¢-
CCCCAGATGACTACCTATCCTCCC-3¢,SB reverse 5¢-CTAAAAGGCCTATCA
CAAAC-3¢(Rosa26 wild-type and Rosa26-lsl-SB11 amplicons are 420 bp and
266 bp, respectively); p53-lsl-R270H were p53 wild-type forward 5¢-TTACA
CATCCAGCCTCTGTGG-3¢,p53 wild-type reverse 5¢-CTTGGAGACATAGC
CACACTG-3¢,p53-lsl-R270H conditional forward 5¢-AGCTAGCCACCA
TGGCTTGAGTAAGTCTGCA-3¢(p53 wild-type and p53-lsl-R270H condi-
tional allele amplicons are 170 bp and 270 bp, respectively); glyceraldehyde-
3-phosphate dehydrogenase (Gapdh)wereforward5¢-GGAGCCAAACGGGT
CATCATCTC-3¢and reverse 5¢-GAGGGGCCATCCACAGTCTTCT-3¢(ampli-
con 233 bp). PCR conditions for Taq polymerase (CLP) were used according to
the manufacturer’s instructions with an initial denaturing step of 94 1Cfor
5min;35cyclesofdenaturingat941C for 1 min, annealing at 55 1Cfor1min
and extension at 72 1C for 1 min; followed by a final extension at 72 1Cfor
7 min. PCR products were separated on a 2% agarose gel and genotype
determined by the absence or presence of expected amplicons.
Liver tumor analysis. The whole liver was carefully removed from the
euthanized animal, washed and placed in cold PBS. The number of surface
liver tumor nodules was counted for all liver lobes. All reasonably sized tumor
nodules (42 mm in diameter) were carefully removed from the liver lobes
using fine forceps and placed in fresh cold PBS. These separated nodules were
then halved using a sterile razor blade and split into samples for DNA and RNA
extraction. Tissue samples for RNA were stored at –80 1C in RNAlater (Sigma)
to prevent RNase contamination and degradation. Histological sections were
taken only for larger tumor nodules (42mmindiameter),inadditiontothe
samples for DNA and RNA extraction. DNA extraction was done as previously
described in the PCR genotyping section. Extraction of RNA was done using
the Trizol reagent (Invitrogen) using protocols described by the manufacturer.
Formalin fixed-paraffin embedded sections from various tissues were sectioned
at 5 mm using a standard microtome (Leica), mounted and heat-fixed onto
glass slides. Tissue section slides were either processed and stained with
hematoxylin-eosin (HE) using standard protocols, or used for immunohisto-
chemistry as described in the next section.
Immunohistochemistry. Formalin-fixed, paraffin-embedded sections from
various tissues were sectioned at 5 mm, mounted and heat-fixed onto glass
slides to be used for immunohistochemical analyses. Briefly, the glass section
slides were dewaxed and rehydrated through a gradual decrease in ethanol
concentration. The antigen epitopes on the tissue sections were then unmasked
using a commercially available unmasking solution (Vector Laboratories)
according to the manufacturer’s instructions. The tissue section slides were
then treated with 3% hydrogen peroxide to remove any endogenous
peroxidases. Blocking was performed at 4 1C using a M.O.M. mouse
immunoglobulin-blocking reagent (Vector Laboratories) in a humidified
chamber for several hours. The sections were then incubated overnight at
41C in a humidified chamber using various primary antibodies: SB transposase
(1:100) (R&D Systems), Alb (1:200) (Abcam), Afp (1:100) (GeneTex), Ki67
(1:200) (Novocastra), b-catenin (1:500) (BD) and Fah (1:250) (AbboMax).
After primary incubation, sections were washed thoroughly in PBS before
incubating with horseradish peroxidase-secondary antibody raised against the
primary antibody initially used. After thorough washes with PBS, the sections
were treated with freshly prepared DAB substrate (Vector Laboratories) and
allowed to develop adequate signal before stopping the reaction in water. For
EGFR immunohistochemistry, EGFr Kit (Clone 31G7) (Zymed Laboratories,
Invitrogen) was used according to the manufacturer’s instructions, except for
the following modification. An additional overnight blocking step using
the M.O.M. mouse immunoglobulin-blocking reagent was incorporated
after proteinase K treatment to reduce background staining. Finally, sections
were then lightly counter-stained with hematoxylin, dehydrated through
gradual increase in ethanol concentration, cleared in Citrosol and mounted
in Permount (Fisher).
Pyrosequencing. Protocol for amplicon sequencing using the GS20 Flex
pyrosequencing machine was as previously described by Roche. Briefly,
100 ng of genomic DNA isolated from individual tumors was digested with
either BfaIorNlaIII, for left or right transposon IR/DR, respectively. A small
volume of this enzyme digest was used for splinkerette linker attachment using
the appropriate linker. To make the BfaI linker, the following oligonucleotide
sequences were annealed together using standard protocols, top strand 5¢-
GTAATACGACTCACTATAGGGCTCCGCTTAAGGGAC-3¢and bottom strand
5¢-TAGTCCCTTAAGCGGAG-3¢. As for the NlaIII linker, the following oligo-
nucleotide sequences were annealed together using standard protocols,
topstrand5¢-GTAATACGACTCACTATAGGGCTCCGCTTAAGGGACCATG-
3¢andbottomstrand5¢- GTCCCTTAAGCGGAGCC-3¢. Linker ligations were
performed overnight at 16 1C using T4 DNA ligase (New England Biolabs). The
ligation reaction was cleaned using MinElute 96-well plates (Qiagen) in a
vacuum manifold and resuspended in 40 ml of sterile double-distilled water
(DDW). This resuspended solution was then digested with either BamHI or
XhoI, for left or right transposon IR/DR, respectively. A small volume was then
used for primary PCR using the following primers. Left IR/DR primer (BfaI),
5¢-CTGGAATTTTCCAAGCTGTTTAAAGGCACAGTCAAC-3¢; right IR/DR
primer (NlaIII), 5¢-GCTTGTGGAAGGCTACTCGAAATGTTTGACCC-3¢and
common splinkerette primer was used for both IR/DRs, 5¢-GTAATACGACTC
ACTATAGGGC-3¢. PCR conditions for Ta q polymerase (CLP) were used
according to the manufacturer’s instructions of an initial denaturing step of
94 1C for 5 min; 30 cycles of denaturing at 94 1C for 30 s, annealing at 60 1Cfor
30 s and extension at 72 1C for 1.5 min; followed by a final extension at 72 1C
for 5 min. One microliter of the diluted first PCR product sample (1:75) was
used as a template for the secondary PCR under the following conditions.
Nested versions of the above primers carrying the required fusion sequences for
GS20 Flex pyrosequencing (Fusion A and Fusion B), as well as a unique 10-bp
barcode recognition sequence for each tumor sample. Primers were designed as
such that the nested transposon primer have the Fusion A and barcode attached
(Fusion A – barcode – nested primer) and the nested linker primer has the
Fusion B sequence attached (linker nested – Fusion B). PCR conditions for Taq
polymerase (Roche FastStart High Fidelity) were used according to the
manufacturer’s instructions of an initial denaturing step of 94 1C for 5 min;
35 cycles of denaturing at 94 1C for 30 s, annealing at 60 1Cfor30sand
extension at 72 1C for 1.5 min; followed by a final extension at 72 1C for 5 min.
After the secondary PCR, the reaction was purified using MinElute 96-well
plates in a vacuum manifold and resuspended in 30 mlofsterileTE.The
amount of DNA in each PCR sample was quantified using the QuantIT
picogreen kit (Invitrogen) and the samples were diluted to a final concentration
of 2 105molecules/ml for pyrosequencing.
Selection criteria for common insertion sites (CISs). See Supplementary
Methods.
Ingenuity Pathways Analysis (IPA). Ingenuity Systems at http://www.
ingenuity.com/.
Egfr PCR genotyping. PCR genotyping was used to confirm the presence of the
T2/onc transposon insertion in intron 24 of the Egfr gene. Briefly, genomic
DNA was isolated from individual tumor nodules using protocols already
described in the PCR genotyping section. PCR genotyping was performed
using 100 ng of diluted genomic DNA as template. PCR primers used for Egfr
intron 24 were forward, 5¢-TACATGGTCAAAATCTCTCCAATAGGTC-3¢and
reverse, 5¢-ATTAGAAAGGGCAACGAAGCTTGC-3¢, with an expected ampli-
con of 713 bp. A third primer specific for the IR/DR-R (T/JB3) of the T2/onc
transposon vector was also included, 5¢-AGGGAATTTTTACTAGGATTAA
ATGTCAGG-3¢. PCR conditions were as described previously in the PCR
genotyping section. The amplicon sizes varied depending on the position of the
T2/onc transposon vector insertion site.
272 VOLUME 27 NUMBER 3 MARCH 2009 NATURE BIOTECHNOLOGY
ARTICLES
© 2009 Nature America, Inc. All rights reserved.

When the T2/onc/Egfr amplicon is expected to overlap the endogenous Egfr
product, a PCR genotyping using only the T/JB3 and Egfr intron 24 forward
primers is used instead with the same PCR conditions.
RT-PCR. Extraction of RNA from tumor nodules was done using the Trizol
reagent using protocols described by the manufacturer. First strand cDNA
synthesis was performed using the Transcriptor First Strand cDNA Synthesis
Kit (Roche) as described by the manufacturer using 1 mg total RNA as template.
Both reactions using with (RT+) and without (RT–) the reverse transcriptase
were performed for all the samples. Subsequent PCR was performed using 1 ml
of the cDNAas template withvarious primer pairs. Primer sequences for alpha-
fetoprotein (Afp) were forward 5¢-CCTGTGAACTCTGGTATCAG-3¢and
reverse 5¢-GCTCACACCAAAGCGTCAAC-3¢(amplicon 410 bp); secreted
phosphoprotein 1 (Ssp1) forward 5¢-CTTTCACTCCAATCGTCCCTAC-3¢and
reverse 5¢-GCTCTCTTTGGAATGCTCAAGT-3¢(amplicon 305 bp); Sleeping
Beauty (SB) transposase forward 5¢-ATGGGAAAATCAAAAGAAATCAGCC-3¢
and reverse 5¢-CGCACCAAAGTACGTTCATCTCTA-3¢(amplicon 221 bp);
albumin (Alb) forward 5¢-CCCCACTAGCCTCTGGCAAAAT-3¢and reverse
5¢-CTTAAACCGATGGGCGATCTCACT-3¢(amplicon 127 bp); epidermal
growth factor receptor (Egfr) forward 5¢-GATAGATGCTGATAGCCGCCC
AAAG-3¢and reverse 5¢-TCATGCTCCAATAAACTCACTGCTT-3¢(amplicon
772 bp); truncated-Egfr forward (same forward primer used for Egfr)and
reverse (specific for the T2/onc SV40-polyA) 5¢-TGCTTTATTTGTGA
AATTTGTGATGCTATTG-3¢(amplicon 320 bp); receptor tyrosine-protein
kinase erbb2 (Erbb2) forward 5¢-CCCAGATCTCCACTGGCTCC-3¢and reverse
5¢-TTCAGGGTTCTCCACAGCACC-3¢(amplicon 376 bp); beta-actin (Actb)
forward 5¢-GTGACGAGGCCCAGAGCA AGAG-3¢and reverse 5¢-AGGG
GCCGGACTCATCGTACTC-3¢(amplicon 938 bp); neomycin (Neo) forward
5¢-ATGATTGAACAAGATGGATTGCACG-3¢and reverse 5¢-AAGGTGAGATG
ACAGGAGATCCTG-3¢(amplicon 321 bp); ubiquitin-conjugating enzyme E2H
(Ube2h) forward 5¢-CTGAGCGGACCCCACGGGAC-3¢and reverse 5¢-CAG
CAACTGGGGCAGGAAGG-3¢(amplicon 505 bp); fumarylacetoacetate hydro-
lase (Fah) forward 5¢-ATGAGCTTTATTCCAGTGGCC-3¢and reverse 5¢-
ACCACAATGGAGGAAGCTCG-3¢(amplicon 503 bp); truncated EGFR for-
ward 5¢-GACCCCCAGCGCTACCTTGTCATTCAG-3¢and reverse (specific for
the rabbit b-globin polyA) 5¢-GCCACACCAGCCACCACCTTCTG-3¢(ampli-
con 140 bp). PCR conditions are similar to PCR genotyping described
previously except 25 to 30 cycles were performed to avoid amplicon saturation.
Representational oligonucleotide microarray analysis (ROMA). Microarray
analysis was performed on human HCC samples as previously described51.
Cell proliferation assay. AML12 (CRL-2254) was obtained from America Type
Culture Collection (ATCC) and maintained according to the recommended
culture conditions. An expression vector for Ube2h (MC200579) mouse cDNA
was obtained from Origene. The empty vector (pcDNA) purchased from
Invitrogen, was used as a negative control. Cell transfections were performed
using Lipofectamine LTX (Invitrogen) with PLUS (Invitrogen) according to the
manufacturer’s recommendation. Transfected cell lines were grown in medium
containing neomycin (0.5 mg/ml) for 2 weeks to select for stable cell popula-
tions. Stable cell populations for each expression vector were obtained from
three individual transfections. Cell proliferation rate of the stable cell popula-
tions was determined using the CellTiter 96 Aqueous One Solution Cell
Proliferation Assay (Promega) according to the manufacturer’s protocols.
Hydrodynamic injection. Hydrodynamic injections were performed as pre-
viously described25. Briefly, fumarylacetoacetate hydrolase (Fah)-null mice
carrying the Rosa26-SB11 transgene were generated. Truncated-EGFR (exon 1
to exon 24) was PCR amplified from pBabe-Puro-LTR-EGFR (a kind gift from
Heidi Gruelich) using the following primers: exon 1 forward 5¢-ATGC
GACCCTCCGGGACGGC-3¢and exon 24 reverse 5¢-CTGAATGACAAGG
TAGCGCTGGGGGTC-3¢was placed under the control of a phosphoglycerate
kinase (PGK) promoter and cloned into the pT2 vector containing the SB
flanking IR/DR recognition sequences to obtain pT2/PGK-Truncated EGFR.
Two other constructs were also prepared: pT2/PGK-FAHIL, vector containing
the Fah and luciferase gene under the control of the PGK promoter52. Twenty
micrograms of each construct was hydrodynamically injected into 6-week-old
Fah-null/Rosa26-SB11 male mice (Fah/SB11) using previously established
conditions26. These mice are normally maintained with 7.5 mg/ml 2-(2-nitro-
4-trifluoromethylbenzoyl)-1,3-cyclohexanedione (NTBC) drinking water but
replaced with normal drinking water after hydrodynamic injection of transpo-
son vectors. These experimental animals were observed for weight changes and
luciferase activity as previously described25.
Note: Supplementary information is available on the Nature Biotechnology website.
ACKNOWLEDGMENTS
The authors wish to thank Christine E. Nelson, Stefanie S. Breitbarth, Michelle
K. Gleason and Geoff Hart for their excellent technical support; Jason B. Bell
for performing the hydrodynamic injections; and Heidi Gruelich, Dana Farber
Cancer Institute, for her kind gift pBabe-Puro-LTR-EGFR. We are also
grateful to the Minnesota Supercomputing Institute for providing extensive
computational resources (hardware and systems administration support)
used to carry out the sequence analysis. A.V. is supported by a Sheila Sherlock
fellowship from the European Association for the Study of the Liver. L.S.C. is
supported by a 1K01CA122183-01 grant from the National Cancer Institute.
N.G.C., N.A.J. and L.T. are supported by the Department of Health and Human
Services, National Institutes of Health and the National Cancer Institute. J.M.L.
is supported by the US National Institute of Diabetes and Digestive and Kidney
Diseases (1R01DK076986-01), Spanish National Institute of Health (SAF-2007-
61898) and Samuel Waxman Cancer Research Foundation. D.A.L. is supported
by U01 CA84221 and R01 CA113636 grants from the National Cancer Institute.
COMPETING INTERESTS STATEMENT
The authors declare competing financial interests: details accompany the full-text
HTML version of the paper at http://www.nature.com/naturebiotechnology/
Published online at http://www.nature.com/naturebiotechnology/
Reprints and permissions information is available online at http://npg.nature.com/
reprintsandpermissions/
1. Spradling, A.C. et al. Gene disruptions using P transposable elements: an integral
component of the Drosophila genome project. Proc. Natl. Acad. Sci. USA 92,
10824–10830 (1995).
2. Bellen, H.J. et al. P-element-mediated enhancer detection: a versatile method to study
development in Drosophila.Genes Dev. 3, 1288–1300 (1989).
3. Ivics, Z., Hackett, P.B., Plasterk, R.H. & Izsvak, Z. Molecular reconstruction of Sleeping
Beauty, a Tc1-like transposon from fish, and its transposition in human cells. Cell 91,
501–510 (1997).
4. Dupuy, A.J.,Fritz, S. & Largaespada, D.A. Transposition and gene disruption in the male
germline of the mouse. Genesis 30, 82–88 (2001).
5. Horie, K. et al. Efficient chromosomal transposition of a Tc1/mariner-like transposon
Sleeping Beauty in mice. Proc. Natl. Acad. Sci. USA 98, 9191–9196 (2001).
6. Keng, V.W. et al. Region-specific saturation germline mutagenesis in mice using the
Sleeping Beauty transposon system. Nat. Methods 2, 763–769 (2005).
7. Dupuy, A.J., Akagi, K., Largaespada, D.A., Copeland, N.G. & Jenkins, N.A. Mammalian
mutagenesis using a highly mobilesomatic Sleeping Beauty transposon system. Nature
436, 221–226 (2005).
8. Collier, L.S., Carlson, C.M., Ravimohan, S., Dupuy, A.J. & Largaespada, D.A. Cancer
gene discovery in solid tumours using transposon-based somatic mutagenesis in the
mouse. Nature 436, 272–276 (2005).
9. Zambrowicz, B.P. et al. Disruption of overlapping transcripts in the ROSA beta
geo 26 gene trap strain leads to widespread expression of beta-galactosidase in
mouse embryos and hematopoietic cells. Proc. Natl. Acad. Sci. USA 94,
3789–3794 (1997).
10. Parkin, D.M., Bray, F., Ferlay, J. & Pisani, P. Global cancer statistics, 2002. CA Cancer
J. Clin. 55, 74–108 (2005).
11. Llovet, J.M., Burroughs, A. & Bruix, J. Hepatocellular carcinoma. Lancet 362,
1907–1917 (2003).
12. Naugler, W.E. et al. Gender disparity in liver cancer due to sex differences in MyD88-
dependent IL-6 production. Science 317, 121–124 (2007).
13. Bressac, B., Kew, M., Wands, J. & Ozturk, M. Selective G to T mutations of p53 gene in
hepatocellular carcinoma from southern Africa. Nature 350, 429–431 (1991).
14. Buendia, M.A. Genetics of hepatocellular carcinoma. Semin. Cancer Biol. 10,
185–200 (2000).
15. Llovet, J.M. & Bruix, J. Molecular targeted therapies in hepatocellular carcinoma.
Hepatology 48, 1312–1327 (2008).
16. Postic, C. et al. Dual roles for glucokinase in glucose homeostasis as determined by
liver and pancreatic beta cell-specific gene knock-outs using Cre recombinase. J. Biol.
Chem. 274, 305–315 (1999).
17. de Vries, A. et al. Targeted point mutations of p53 lead to dominant-negative
inhibition of wild-type p53 function. Proc. Natl. Acad. Sci. USA 99, 2948–2953
(2002).
18. Rohde, F. et al. Expression of osteopontin, a target gene of de-regulated Wnt
signaling, predicts survival in colon cancer. Int. J. Cancer 121, 1717–1723
(2007).
NATURE BIOTECHNOLOGY VOLUME 27 NUMBER 3 MARCH 2009 273
ARTICLES
© 2009 Nature America, Inc. All rights reserved.

19. Calvisi, D.F., Factor, V.M.,Loi, R. & Thorgeirsson, S.S. Activation of beta-catenin during
hepatocarcinogenesis in transgenic mouse models: relationship to phenotype and
tumor grade. Cancer Res. 61, 2085–2091 (2001).
20. Margulies, M. et al. Genome sequencing in microfabricated high-density picolitre
reactors. Nature 437, 376–380 (2005).
21. van Luenen, H.G., Colloms, S.D. & Plasterk, R.H. The mechanism of transposition of
Tc3 i n C. elegans.Cell 79, 293–301 (1994).
22. Vos, J.C., De Baere, I. & Plasterk, R.H. Transposase is the only nematode protein
required for in vitro transposition of Tc1. Genes Dev. 10, 755–761 (1996).
23. Lampe, D.J., Churchill, M.E. & Robertson, H.M. A purified mariner transposase is
sufficient to mediate transposition in vitro. EMBO J. 15, 5470–5479 (1996).
24. Chiang, D.Y. et al. Focal gains of VEGFA and molecular classification of hepatocellular
carcinoma. Cancer Res. 68, 6779–6788 (2008).
25. Wangensteen, K.J. et al. A facile method for somatic, lifelong manipulation of multiple
genes in the mouse liver. Hepatology 47, 1714–1724 (2008 ).
26. Bell, J.B . et al. Preferential delivery of the Sleeping Beauty transposon system to livers
of mice by hydrodynamic injection. Nat. Protocols 2, 3153–3165 (2007).
27. Lee, T.K. et al. Regulation of angiogenesis by Id-1 through hypoxia-inducible factor-
1alpha-mediated vascular endothelial growth factor up-regulation in hepatocellular
carcinoma. Clin. Cancer Res. 12, 6910–6919 (2006).
28. Hirota, K., Murata, M., Itoh, T., Yodoi, J. & Fukuda, K. Redox-sensitive transactivation
of epidermal growth fac tor receptor by tumor necr osis factor confers the NF -kappa B
activation. J. Biol. Chem. 276, 25953–25958 (2001).
29. Arsura, M. et al. Transientactivation of NF-kappaB through a TAK1/IKK kinase pathway
by TGF-beta1 inhibits AP-1/SMAD signaling and apoptosis: implications in liver tumor
formation. Oncogene 22, 412–425 (2003).
30. Chang, C.M. et al. A minor tyrosine phosphorylation site located within the CAIN
domain plays a critical role in regulating tissue-specific transformation by erbB kinase.
J. Virol. 69, 1172–1180 (1995).
31. Boerner, J.L., Danielsen, A. & Maihle, N.J. Ligand-independent oncogenic signaling by
the epidermal growth factor receptor: v-ErbB as a paradigm. Exp. Cell Res. 284,
111–121 (2003).
32. Frederick, L., Wang, X.Y., Eley, G. & James, C.D. Diversity and frequency of epidermal
growth factor receptor mutations in human glioblastomas. Cancer Res. 60,
1383–1387 (2000).
33. Landau, M., Fleishman, S.J. & Ben-Tal, N. A putative mechanism for downregulation of
the catalytic activity of the EGF receptor via direct contact between its kinase and
C-terminal domains. Structure 12, 2265–2275 (2 004).
34. Sasaoka, T. et al. Involvement of ErbB2 in the signaling pathway leading to cell cycle
progression from a truncated epidermal growth factor receptor lacking the C-terminal
autophosphorylation sites. J. Biol. Chem. 271, 8338–8344 (1996).
35. Bekaii-Saab, T., Williams, N., Plass, C., Calero, M.V. & Eng, C. A novel mutation in the
tyrosine kinase domain of ERBB2 in hepatocellular carcinoma. BMC Cancer 6,278
(2006).
36. Buckley, A.F., Burgart, L.J., Sahai, V. & Kakar, S. Epidermal growth factor receptor
expression and gene copy number in conventional hepatocellular carcinoma. Am. J.
Clin. Pathol. 129, 245–251 (2008).
37. Rajkumar, T. & Gullick, W.J.The type I growth factor receptors in human breast cancer.
Breast Cancer Res. Treat. 29, 3–9 (1994).
38. Lopes, L.F. & Bacchi, C.E. EGFR and gastrointestinal stromal tumor: an immunohis-
tochemical and FISH study of 82 cases. Mod. Pathol. 20, 990–994 (2007).
39. Villanueva, A. et al. Pivotal role of mTOR signaling in hepatocellular carcinoma.
Gastroenterology (2008).
40. Villanueva, A., Newell, P., Chiang, D.Y., Friedman, S.L. & Llovet, J.M. Genomics and
signaling pathways in hepatocellular carcinoma. Semin. Liver Dis. 27,55–76
(2007).
41. Tanabe, K.K. et al. Epidermal growth factor gene functional polymorphism and the risk
of hepatocellular carcinoma in patients with cirrhosis. J. Am. Med. Assoc. 299,53–60
(2008).
42. Philip, P.A. et al. Phase II study of erlotinib (OSI-774) in patients with advanced
hepatocellular cancer. J. Clin. Oncol. 23, 6657–6663 (2005).
43. Salvi, A. et al. In vitro c-met inhibition by antisense RNA and plasmid-based RNAi
down-modulates migration and invasion of hepatocellular carcinoma cells. Int. J.
Oncol. 31, 451–460 (2007).
44. Takami, T. et al. Loss of hepatocyte growth factor/c-Met signaling pathway accelerates
early stages of N-nitrosodiethylamine induced hepatocarcinogenesis. Cancer Res. 67,
9844–9851 (2007).
45. Su, G.H., Song, J.J., Repasky, E.A., Schutte, M. & Kern, S.E. Mutation rate of
MAP2K4/MKK4 in breast carcinoma. Hum. Mutat. 19,81(2002).
46. Xin, W. et al. MAP2K4/MKK4 expression in pancreatic cancer: genetic validation of
immunohistochemistry and relationship to disease course. Clin. Cancer Res. 10,
8516–8520 (2004).
47. Nakayama, K. et al. Homozygous deletion of MKK4 in ovarian serous carcinoma.
Cancer Biol. Ther. 5, 630–634 (2006).
48. Iizuka, N. et al. Differential gene expression in distinct virologic types of hepato-
cellular carcinoma: association with liver cirrhosis. Oncogene 22, 3007–3014
(2003).
49. Moinzadeh, P., Breuhahn, K., Stutzer, H. & Schirmacher, P. Chromosome alterations in
human hepatocellular carcinomas correlate with aetiology and histological
grade–results of an explorative CGH meta-analysis. Br. J. Cancer 92, 935–941
(2005).
50. Zimmermann, U. et al. Chromosomal aberrations in hepatocellular carcinomas: rela-
tionship with pathological features. Hepatology 26, 1492–1498 (1997).
51. Zender, L. et al. Identification and validation of oncogenes in liver cancer using an
integrative oncogenomic approach. Cell 125, 1253–1267 (2006).
52. Wilber, A. et al. Messenger RNA as a source of transposase for sleeping beauty
transposon-mediated correction of hereditary tyrosinemia type I. Mol. Ther. 15,
1280–1287 (2007).
274 VOLUME 27 NUMBER 3 MARCH 2009 NATURE BIOTECHNOLOGY
ARTICLES
© 2009 Nature America, Inc. All rights reserved.