Content uploaded by Jürgen Borlak
Author content
All content in this area was uploaded by Jürgen Borlak
Content may be subject to copyright.
ORIGINAL PAPERS
Epidermal growth factor-induced hepatocellular carcinoma: gene
expression profiles in precursor lesions, early stage and solitary tumours
Ju
¨
rgen Borlak*
,1,4
, Tatiana Meier
1,4
, Roman Halter
1,4
, Reinhard Spanel
2
and Katharina Spanel-Borowski
3
1
Department of Pharmacology and Molecular Medicine, Fraunhofer Institute of Toxicology and Experimental Medicine, Nikolai-
Fuchsstr. 1, 30625 Hannover, Germany;
2
Institute of Pathology, Viersen, Germany;
3
Institute of Anatomy, University of Leipzig,
Germany
Epidermal growth factor is an important mitogen for
hepatocytes. Its overexpression promotes hepatocellular
carcinogenesis. To identify the network of genes regulated
through EGF, we investigated the liver transcriptome
during various stages of hepatocarcinogenesis in EGF2B
transgenic mice. Targeted overexpression of IgEGF
induced distinct hepatocellular lesions and eventually solid
tumours at the age of 6–8 months, as evidenced by
histopathology. We used the murine MG U74Av2
oligonucleotide microarrays to identify transcript signa-
tures in 12 tumours of small (n ¼ 5, pooled), medium
(n ¼ 4) and large sizes (n ¼ 3), and compared the findings
with three nontumorous transgenic livers and four control
livers. Global gene expression analysis at successive stages
of carcinogenesis revealed hallmarks linked to tumour size.
A comparison of gene expression profiles of nontumorous
transgenic liver versus control liver provided insight into
the initial events predisposing liver cells to malignant
transformation, and we found overexpression of c-fos, eps-
15, TGIF, IGFBP1, Alcam, ets-2 and repression of Gas-1
as distinct events. Further, when gene expression profiles of
small manifested tumours were compared with nontumor-
ous transgenic liver, additional changes were obvious and
included overexpression of junB, Id-1, minopontin, villin,
claudin-7, RR M2, p34cdc2, cyclinD1 and cyclinB1 among
others. These genes are therefore strongly associated with
tumour formation. Our study provided new information on
the tumour stage-dependent network of EGF-regulated
genes, and we identified candidate genes linked to tumori-
genes and progression of disease.
Oncogene (2005) 24, 1809–1819. doi:10.1038/sj.onc.1208196
Published online 24 January 2005
Keywords: HCC; EGF; transgenic mice; tumour stages;
gene expression profiling
Introduction
It is estimated that about 350 000 new cases of
hepatocellular carcinoma (HCC) arise per year (Schafer
and Sorrell, 1999). The most prominent risk factors are
chronic hepatitis B and C virus infection, Aflatoxin B1
exposure and alcohol-related cirrhosis. As of today, the
precise molecular mechanism in the onset and progres-
sion of disease remains uncertain.
Overexpression of liver mitogens may be an important
mechanism of disease. EGF and TGFa are potent
mitogens for hepatocytes (Wang et al., 1999; Rescan
et al., 2001). Signalling of these mitogens is through
binding to members of the EGF-receptor family. Their
expression is unregulated in HCC and this supports
autocrine growth stimulation of hepatoma cells (Yama-
guchi et al., 1995; Chung et al., 2000). EGF also plays an
important role in hepatocyte morphology (Rescan et al.,
2001). Overexpression of EGF might be an important
step towards development of liver cancer and is
suspected to play a particular role in spontaneous liver
tumour development (Ostrowski et al., 2000). There is
suspicion that EGF plays a role in Helicobacter
hepaticus-induced chronic hepatitis with progression to
hepatocellular cancer (Ramljak et al., 1998). Previous
investigations demonstrated targeted overexpression of
a secretable form of EGF (IgEGF) to result in multiple
highly malignant HCCs, with 100% fatalities around
7–8 months after birth (To
¨
njes et al., 1995). This
transgenic mouse line therefore mimics effectively the
consequence of altered EGF signalling via the EGF
receptor. We used this mouse model to identify the
network of EGF-regulated genes at various stages of
tumour development and in solid tumours. Global gene
expression analysis at successive stages of carcinogenesis
holds promise for an identification of master genes for
the onset and progression of disease. Thus, we aimed to
identify candidate genes associated with early stages of
tumorigenesis and with developed HCCs. Overall, this
study aimed for a better understanding of the network
of EGF-regulated genes in liver carcinogenesis.
Results
Histopathology of liver tumours
Macroscopically, all EGF-overexpressing animals devel-
oped tumours at the age of 6–9 months. A total of six
Received 20 February 2004; revised 28 July 2004; accepted 9 August
2004; published online 24 January 2005
*Correspondence: J Borlak; E-mail: Borlak@item.fraunhofer.de
4
These authors contributed equally to this work
Oncogene (2005) 24, 1809–1819
&
2005 Nature Publishing Group
All rights reserved 0950-9232/05 $30.00
www.nature.com/onc
tumour-bearing EGF2B mice as well as tumour-free
parenchyma and liver from wild-type animals were
investigated. The HCCs ranged from less to highly
differentiated. As a rule, small HCCs were adenoma-like
tumours with well-developed hepatocytes, whereas
larger HCC showed a lot of cellular dedifferentiation
and enhanced nuclear atypia (Figure 1).
There were precursor lesions in the tumour-free liver
tissue as well. Basically all of the trabecular parenchyma
showed nuclear atypia with enlarged nuclei, mainly
increased numbers of large polyploid nuclei, anisocar-
yosis and some polymorphism, defined as large-cell
dysplasia (LCD). This LCD seemed to merge into multi-
centric nodule formation, usually with small cell changes
and various degrees of atypia. According to the actually
proposed terminology of nodular hepatocellular lesions
in human pathology (Hepatology 1995, International
Working Party), these nodules can be defined as
dysplastic foci (DF) and dysplastic nodules (DN).
Transcript profiling
Abundant expression of transgenic EGF in liver and
tumours of transgenic mice was confirmed by RT–PCR,
as depicted in Figure 2. As transcript profiling of
juvenile tumours is of considerable value for an
identification of priming factors in tumour development,
we divided tumours into three groups according to their
size. We investigated gene expression profiles in devel-
oping liver carcinomas and compared global expression
profiles of wild-type animals with liver tumours from
small to large size, as well as macroscopically non-
tumorous livers of tumour-bearing animals. In all, 4175
genes and ESTs were commonly expressed in all eight
tumour samples and 4149 genes were expressed in n ¼ 4
normal liver samples, the expression of many genes
being increased or decreased during the process of liver
carcinogenesis (Table 1). To determine significant gene
expression changes, we performed T-test analysis
between control liver and sets of tumours, taking only
changes into account, which were detected in all tumour
samples or in all control livers for identification of
overexpressed or repressed genes accordingly. Further,
we performed ranking analyses (see Materials and
methods) that enabled a more stringent comparison
according to the consistency of gene expression changes
(Table 1). Notably, gene expression profiles were
significantly changed in transgenic nontumorous liver,
but the number of deregulated genes was further
increased in tumours. In all, 109 of the upregulated
and 55 of the repressed genes show a signal intensity
X70, a fold change X3orp3, a P-value p0.05 and
100% concordance of expression changes in pairwise
comparative analyses in at least one group of the
tumours, when compared with normal liver. A total of
59 of the overexpressed and five of the repressed genes
selected by their importance and their possible functions
are listed in Table 2a and b.
Some genes known to be important for cancer biology
(e.g. cyclinD1 and PDGFa) with a lower FC are also
included. For the pool of small tumours, T-test analysis
could not be carried out. Instead, the inclusion criterion
was 100% concordance of expression changes, when
compared to n ¼ 4 normal liver samples. Table 2
allows a direct comparison of gene expression profiles
of different tumour sizes, from EGF-overexpressing
Figure 1 Histology of tumours and precursor lesions in EGF2B transgenic mice. Normal liver tissue of nontransgenic controls with
normal polyploidal variation of hepatocytic nuclear size (central vein: cv) (a). LCD in tumour-free parenchyma of EGF2B mice
merging with initial perivenous DF (arrows) and DN (arrowheads): Small-cell nodular proliferations (b). Well-differentiated,
adenoma-like small HCC: irregular, mainly bilayered trabeculae, slight polymorphism and lipid vacuolization (c). Multilayered
trabeculae of larger HCC (d). Same magnifications. Staining: H (a, b); PAS (c); H and E (d)
Early tumour stage in EGF-induced HCC
J Borlak et al
1810
Oncogene
tumour-free transgenic liver with LCD to large mani-
fested tumours. We found certain genes to be equally
deregulated in displastic transgenic liver as in tumours
(e.g. c-fos, eps-15, EGR-1, TGIF, IGFBP1, Alcam),
while others were dramatically deregulated at the onset
of small HCC (e.g. p34cdc2, Id1, junB, minopontin,
claudin7). Besides, we identified 23 genes that were
uniquely expressed in all tumours, but not in controls
Table 1 Number of genes and ESTs expressed in EGF-induced liver tumours and normal liver
Genes or ESTs EGF-transgenic liver
(three samples)
All tumours
(eight samples)
Normal liver
(four samples)
Detected in all samples within a group 4494 4175 4149
Not detected in all samples within a group 6569 5768 6655
Upregulated in liver tumours according to T-test (P-valuep0.05); FCX2 140 265 —
Upregulated in liver tumours according to comparison ranking (100%
concordance in comparative analyses); FCX2
67 89 —
Downregulated in liver tumours according to T-test (P-valuep0,05); FCp2 52 130 —
Downregulated in liver tumours according to comparison ranking (100%
concordance in comparative analyses); FCp2
23 59 —
Uniquely expressed either in all tumours or normal liver, P-value in T-test
p0.05, 100% concordance in comparative analyses; FCX2/p2
23 1
Figure 2 (a) Structure of the transgene. albP, murine albumin promoter; Ig-S, Ig-signal sequence; I, Intron sequence; EGF, synthetic
EGF; SVA, SV40 poly-A signal. (b) PCR analysis of Ig/EGF from tail biopsies to identify transgenic mice. Lane 1: nontransgenic mice;
lanes 2–5: transgenic mice; M: molecular weight standard; lane 6: amplified fragment of the transgene was digested with EcoRI to
obtain fragments of 210 and 107 bp. (c) RT–PCR analysis of the housekeeping gene (b-actin) and the transgene. Lanes 1–3:
nontransgenic liver, lanes 4–6: macroscopic nontumorous liver of transgenic mice, lane 7: pool of small tumours; lanes 8–10: tumours
of medium size; lanes 11 and 12: tumours of large size; lane 13: negative control (water). (d) RT–PCR of selected genes (lanes represent
samples as described in (c))
Early tumour stage in EGF-induced HCC
J Borlak et al
1811
Oncogene
Table 2 Gene expression signatures in EGF-induced mouse liver tumours: (a) upregulated genes; (b) downregulated genes
Gene ACC Gene description Transgenic nontumorous
liver
Tumours
Small size (o2 mm)
(Pool)
Medium size (5 mm) Large size (10 mm)
FC P-value
(%)
FC %
a
FC P-value
(%)
FC P-value
(%)
(a) Upregulated in EGF-induced liver tumours
b
Upregulated in transgenic liver and tumours
Unknown function
Trefoil factor 3 D38410 Mucin-associated polypeptide 105.8 0.009 164.3 100 117.8 0.156 (75) 50.1 0.372 (66)
H19mRNA X58196 Post-transcriptional regulator,
overexpressed in HCC
24.84 0.317 164.8 100 136.1 0.046 40.55 0.269
CD63 antigen D16432 Tetraspanin, late endosome
marker
11.33 0.009 34.98 100 27.65 0.018 50.35 0.005
G7e U69488 Viral envelope-like protein 7.65 0.052 (83) 34.39 100 19.91 0.005 23.96 0.007
Fibrinogen-like protein 2 M16238 Unknown 6.7 0.081 5.01 100 6.52 0 8.26 0.006
Growth promotion
PDGFa M29464 Platelet-derived growth factor a,
induced in HCC
2.05 0.011 (41) 2.77 100 2.51 0.022 (87) 2.44 0 (83)
TGFa M92420 Transforming growth factor a,
induced in HCC
1.69 0.027 (41) 1.93 100 2.07 0.026 (68) 3.62 0.036
eps-15 L21768 Substrate of the EGF-receptor,
transforming capacity
2.82 0.02 2.89 100 2.48 0.005 (50) 3.44 0.012
FBJ osteosarcoma
oncogene (c-fos)
V00727 Oncogene, signal transduction 2.82 0 4.09 100 2.9 0.048 5.64 0.18
Insulin-like growth
factor-binding protein I
(IGFBP1)
X81579 Regulation of cell growth 18.8 0.186 (75) 21.18 100 20.83 0.032 30.48 0.005
TGIF X89749 Inhibitor of TGFb signalling 2.79 0.107 2.56 100 2.6 0.011 3.62 0.067
Transcription factor
ets-2 J04103 Transcription factor, induced in
HCC
3.94 0.102 (91) 3.44 100 3.61 0.002 4.25 0
Angiogenesis
Early growth response
(EGR-1)
M28845 Zinc-finger encoding gene,
regulates tumour angiogenesis
10.9 0.103 (91) 9.06 100 6.65 0.007 (93) 8.26 0.021
Adhesion
Alcam L25274 Adhesion molecule, involved in
tumour development
3.84 0.006 5.18 100 3.89 0.032 4.69 0.008
Defense response
TIS21 M64292 Negative control of cell growth 6.46 0.111 (66) 5.64 100 5.18 0.093 (25) 5.92 0.002 (83)
B-cell translocation gene 3
(ANA)
D83745 Antiproliferative protein 7.17 0.093 8.48 100 10.15 0.001 22.42 0.074
BLNK AF068182 Central linker protein in B-cell
activation
2.25 0 3.65 100 3.48 0.036 4.54 0.011
Metabolism
Stearoyl-CoA desaturase 2 M26270 Fatty acid biosynthesis;
upregulated in HCC
12.13 0.062 33.4 100 34.91 0.062 28.62 0.01
Lipoprotein lipase M63335 Fatty acid degradation 27.02 0.033 22.25 100 20.03 0.019 23.78 0.05
Early tumour stage in EGF-induced HCC
J Borlak et al
1812
Oncogene
Table 2 (Continued )
Gene ACC Gene description Transgenic nontumorous
liver
Tumours
Small size (o2 mm)
(Pool)
Medium size (5 mm) Large size (10 mm)
FC P-value
(%)
FC %
a
FC P-value
(%)
FC P-value
(%)
Retinol-binding protein I X60367 Vitamin A, E binding 7.87 0.005 13.29 100 10.03 0.001 9.99 0.009
Extracellular matrix
Nidogen1 L17324 Basement membrane compo-
nent, cell–matrix interaction, cell
adhesion
1.95 0.002 (100) 3.05 100 2.26 0 (100) 3.66 0.018 (100)
Miscellaneous
Phospholipid scramblase 1
(TRA1)
D78354 Potential role in growth factor
signalling pathways; associated
with leukomogenesis
3.61 0.083 4.87 100 4.86 0.003 5.65 0.016
Overexpressed at the onset of small tumours
Cell cycle promotion
CyclinB1 X64713 Cell cycle control Absent 2.85 100 1.92 0.007 (68) 3.07 0.036
CyclinD1 AI849928 Cell cycle control 1.19 (100) 2.79 100 2.88 0.019 (87) 2.68 0.119 (83)
Cell division cycle control
protein 2a (p34 cdc2)
M38724 Mitosis-specific phosphory-
lation of cytoskeletal protein
Absent 5.74 100 4.82 0.026 (0) 6.77 0.05 (66)
Ki67 X82786 Tumour cell proliferation
marker
1.28 0.004 (66) 3.88 100 2.64 0.008 3.55 0.074 (83)
Transcription factors
Inhibitor of DNA binding
1 (Id-1)
M31885 Helix–loop–helix protein;
inactivates p16/pRB pathway in
prostate cancer
Absent 4.41 100 1.57 0.020 (12) 2.98 0.013 (66)
junB U20735 Transcription factor,
proto-oncogene
Absent 3.38 100 1.97 0.232 (0) 2.77 0.091 (58)
Ets transcription factor
(ELF3)
AF016294 Member of ETS family,
involved in cancer
1.84 0.202 (16) 3.19 100 2.88 0.011 (25) 2.22 0.047 (25)
Signalling
Cell adhesion kinase L57509 CAK receptor kinase Absent 7.31 100 5.03 0.006 (25) 9.2 0.013
Cytoskeleton
villin M98454 Matrix protein in microtubuli;
upregulated HCC
Absent 8.75 100 1.45 0.166 (25) 1.87 0.178 (66)
Angiogenesis
Calpactin I (annexin II) M14044 Calcium-dependent
phospholipid binding;
upregulated in proliferating
hepatocytes; role in angiogenesis
1.92 0.019 (75) 7.86 100 7.6 0.018 19.17 0.07
Lymphotoxin b U16985 TNF family cytokine, lymph
node development; may initiate
angiogenesis
Absent 23.58 100 5.73 0.01 (75) 13.19 0.276
Adhesion
claudin-7 AF087825 Tight junction adhesion
protein, activates processing of
pro-matrix metalloproteinase-2
Absent 4.69 100 Absent Absent
Cell death
Cell death factor CIDE-A AF041376 DNA fragmentation, activation
of apoptosis
Absent 5.92 100 3.94 0.068 (50) 4.48 0.225 (33)
Early tumour stage in EGF-induced HCC
J Borlak et al
1813
Oncogene
Table 2 (Continued )
Gene ACC Gene description Transgenic nontumorous
liver
Tumours
Small size (o2 mm)
(Pool)
Medium size (5 mm) Large size (10 mm)
FC P-value
(%)
FC %
a
FC P-value
(%)
FC P-value
(%)
Invasion, metastasis
Minopontin (osteopontin) X13986 Secreted adhesive glycoprotein,
upregulated in HCC
1.26 0.422 (0) 18.22 100 2.01 0.593 (25) 3.39 0.067
Ribonucleotide reductase
M2 subunit (RR M2)
M14223 DNA synthesis and repair, role
in cancer/metastases
1.35 0.046 (0) 4.49 100 2.8 0.024 (75) 2.06 0.001 (8)
Lymphocyte antigen 6
complex (Ly6d)
X63782 GPI-anchored protein,
correlates with malignancy
of mouse tumours
Absent 10.67 100 17.29 0.041 47.42 0.177
Proteolysis, peptidolysis
CarboxypeptidaseE X61232 Abundantly expressed in
hepatoma and HCC
0.6 0.253 (58) 3.71 100 5.02 0.195 52.71 0.162
Metabolism
Nonallelic mRNA for
pancreatic a-amylase
isoenzyme (pCEPa12)
X02578 Overexpressed in lung cancers 1.67 0.150 (60) 63.51 100 47.27 0.287 (75) 85.87 0.14
Transport
Rab3D M89777 Small GTPase, regulatory role
in vesicular transport
Absent 14.73 100 8.16 0.002 13.87 0.056
Solute carrier family 7 AJ012754 Cationic amino-acid
transporter
Absent 6.05 100 7.03 0.001 5.13 0.001 (66)
Immune response
Toll-like receptor 6
(TLR6)
AB02088 Activation of Nf-kB and c-Jun
N-terminal kinase (JNK),
immune response
2.22 0.002 (58) 7.19 100 3.75 0.013 (81) 3.67 0.015
Miscellaneous
Endothelial monocyte
activated polypeptideI
(EMAP)
U41341 Tumour-derived cytokine 3.45 0.048 (33) 28.63 100 17.84 0.04 59.61 0.082
Other genes upregulated in tumours
Signalling
Ect2 oncogene L11316 Signal transduction,
Rho-specific exchange factor
2.19 0.088 (58) 3.96 100 2.83 0.01 (93) 3.55 0.088 (83)
A6 related protein Y17808 Mouse homolog of human
protein kinase
2.4 0.144 (50) 6.42 100 6.98 0.013 14 0.013
Transcription factor
LRG-21 U19118 Transcription factor 3.1 0.179 (33) 4.65 100 4.05 0.119 5.49 0.015
Angiogenesis, invasion, metastasis
RhoC X80638 Small GTPase, involved in
angiogenesis and metastasis
2.17 0.026 (66) 4.11 100 3.11 0.01 7.45 0
Plasminogen activator
inhibitor-1 (PAI-1)
M33960 Serine protease inhibitor
involved in angiogenesis,
invasion, metastasis
4.43 0.187 (66) 1.21 0 3.35 0.273 (25) 12 0.01
Serpinb6 U25844 Serine protease inhibitor 3,
regulation of tumour
progression, inflammation
and cell death
1.29 0.077 (41) 1.81 100 2.34 0.082 (81) 8.06 0.007
Early tumour stage in EGF-induced HCC
J Borlak et al
1814
Oncogene
Table 2 (Continued )
Gene ACC Gene description Transgenic nontumorous
liver
Tumours
Small size (o2 mm)
(Pool)
Medium size (5 mm) Large size (10 mm)
FC P-value
(%)
FC %
a
FC P-value
(%)
FC P-value
(%)
Extracellular matrix
Procollagen type IV a 1 M15832 Basement membrane-associated
protein, cell adhesion, upregu-
lated in human HCC
1.66 0.083 (58) 3.27 100 3.94 0.006 (100) 3.81 0.204 (100)
Cytoskeleton
Cytokeratin endoA X15662 Intermediate filament protein,
specific to carcinoma
2.01 0.203 (66) 3.14 100 2.4 0 (100) 4.48 0.081 (100)
Cytoceratin endoB M22832 Intermediate filament protein,
specific to carcinoma
2.39 0.098 (83) 3.83 100 2.73 0 (100) 4.74 0.041 (100)
Miscellaneous
AE binding protein AF053943 Wound healing 3.94 0.169 (33) 8.43 100 7.81 0 16.5 0.032
Reduced expression 3
(REX-3)
AF05134 Downregulated by retinoic acid-
induced growth inhibition in
murine teratocarcinoma cells
1.46 0.031 9.03 100 4.45 0 14.42 0.035
Carbonic anhydrase 2 M25944 Role in facilitating transport of
CO
2
, regulated by micro-
environmental hypoxia
2.22 0.013 (41) 5.26 100 8.0 0.122 (81) 22.79 0.226 (66)
MT-ACT48 AJ238894 Mitochondrial long-chain
acyl-CoA thioesterase
(eps-15 partner)
1.36 0.134 (16) 2.05 75 1.66 0.003 (62) 3.4 0.001
ww domain-binding 5 U92454 Binding domain for proline-rich
sequences of various structural,
regulatory and
signalling proteins
1.49 0.034 4.5 100 3.68 0 (57) 5.18 0.036
rbm3 AB016424 Cold shock protein 2.24 0.009 (75) 5.36 100 3.47 0.029 (93) 7.36 0.109
(b) Downregulated in EGF-induced liver tumours
c
Cyt. P450 retinoic acid Y12657 RA oxidation 3.3 0.003 (91) 5.43 100 4.23 0.001 5.05 0.004
Growth arrest-specific 1
(GAS-1)
X65128 Cell proliferation inhibitor 2.42 0.005 (75) 3.98 100 4.81 0.002 6.69 0.002
Tyrosinase-related protein-2
(tyrp-2)
X63349 Melanine biosynthesis 2.77 0.059 (58) 13.27 75 11.3 0.002 (93) 24.85 0.018
Cytochrome P450 2f2 M77497 Naphthalene detoxification 1.36 0.285 (8) 10.51 100 7.24 0.006 12.79 0.006
SULT-X1 AF02604 Sulphotransferase-related
protein
1.7 0.089 (50) 16.27 100 8.67 0.01 16.27 0.009
ACC: accession number; FC: fold change; P-value: P-value in T-test; %: Concordance (%) of change calls in the pairwise comparisons (each tumour compared to each normal liver sample) by
which genes were up- or downregulated. If not indicated, concordance ¼ 100%.
a
Instead of P-value for the pool of small tumours, the concordance (%) is shown, because for a pool no T-test
analyses could be carried out.
b
The genes had in at least one group of the tumours an FCX3, a P-value in T-test o0.05, a signal intensity >70 and were upregulated in 100% of pairwise analyses,
compared with normal liver of control mice. Grey-coloured genes were absent in the control tissues. Grey-coloured rows indicate genes expressed in all tumours and not in controls.
c
The genes had
in at least one group of tumours an FC o3, a P-value in T-test o0.05, were downregulated in pairwise analyses when compared to normal liver of control mice and a signal intensity >70 in
control liver. Grey-coloured rows are genes, which were expressed in all normal tissues but not in tumours.
Early tumour stage in EGF-induced HCC
J Borlak et al
1815
Oncogene
(Table 1), and these included Ets transcription factor,
calpactin, stearoyl-CoA desaturase 2 and Rab3D among
others (Table 2). In contrast, Gas1 and tyrosinase-
related protein-2 were detected in all normal liver
samples, but not in tumours (Tables 1, 2).
Notably, among genes which were upregulated in
tumours (Table 2a), we found several genes with
an inferred role in tumour development such as
growth factors TGFa, some components of EGF/
TGFa-mediated signalling pathway (c-fos, eps-15, MT-
Act48), EGR-1, cell division cycle control protein 2a,
proto-oncogene junB, IGFBP1, cyclinB1, cyclinD1 and
Id-1, the latter two interfering with the Rb pathway.
Additionally, genes that are associated in human cancers
with metastasis and angiogenesis like rhoC, PAI-1 and
calpactin 1 (annexin II) were upregulated. Further, we
observed overexpression of some genes coding for
components of the cytoskeletal network (villin, cytoker-
atin endoA and B) and basal membrane (collagen IVa1,
nidogen), and proteins that are involved in survival
factor signalling pathway like IGFBP1 and TLR6. The
most highly overexpressed genes in all tumours included
H19 mRNA, trefoil factor 3, stearoyl-CoA desaturase 2,
lipoprotein lipase, nonallelic mRNA for pancreatic
a-amylase isoenzyme, CD63 antigen, G7e – their
contribution in carcinogenesis may be of great impor-
tance, but requires further study. The expression of
certain genes was obviously enhanced as part of a
defence response to the free development, for example,
cell death factor CIDE-A, TIS21 and ANA (BTG3). In
strong contrast, most of the repressed genes code for
proteins important for liver function, including drug
binding, detoxification and metabolism (see Table 2b).
We thus provide evidence for loss of metabolic
competence in liver tumour tissue.
The expression of a selected number of genes was
additionally analysed by semiquantitative RT–PCR
(Figure 2d). There was good agreement between
microarrays and RT–PCR experiments, for example,
no expression of G7e and Trefoil factor 3 in controls
but strong expression of TGIF and LRG-21 in large
tumours and strong expression of Trefoil factor 3 and
G7e in small tumours. In the case of calpactin, the level
of induction determined by RT–PCR and the micro-
array differed, though an identical trend was observed,
for example, induction. Presumably, the different
methods of gene expression analysis (see Materials and
methods) and the algorithm applied for data analysis of
microarray experiments produced different estimate of
induction levels.
Discussion
Liver pathology
Morphological phenotyping enabled new insights into
the process of hepatocellular carcinogenesis. Particu-
larly, the onset and development of malignant tumours
were followed. Notably, no benign tumours were found.
We assume LCD to be the precursor lesion being at the
edge of malignant change. No inflammatory process
was observed and no cirrhotic conversion of liver tissue
was noted. This contrasts with microscopic lesions
frequently seen in humans, where cirrhotically altered
organs are the main precursor of carcinogenesis
(Kubicka et al., 2000). An important finding of our
study was the 100% incidence of malignant tumour
within 6–8 months after birth, and we suggest the
following sequence of events: diffuse LCD merges into
multiple DF and DN, with local growth towards HCC.
Transcript profiling
The transgenic mouse line EGF2B develops HCCs
(To
¨
njes et al., 1995) as a consequence of overexpression
of a secretable form of EGF, which is known to be a
strong mitogen for liver cells. There is cumulative
evidence for EGF or TGFa overexpression to be
necessary, but not sufficient in inducing carcinogenesis
in mice (Sandgren et al., 1993; Wu et al., 1994; To
¨
njes
et al., 1995). As shown in our study and by others,
undue exposure to EGF predisposes liver tissue to
cancer. Large-scale expression analysis enabled initial
changes to be studied. We compared the expression
profiles of tumour-free liver of transgenic with normal
liver of control mice. In transgenic displastic liver,
certain genes were upregulated at the same level as in
tumours. These genes include eps-15, a substrate of
EGF-R with transforming capacity (Alvarez et al.,
1995), and c-fos, a transcription factor of the EGF
signalling pathway known to be induced by EGF and
other growth factors (Dey et al., 1991). Therefore,
transcriptional activation of the EGF/TGFa signalling
pathway in the liver and tumours of EGF2B transgenic
mice is a prominent feature. Similarly, the transcrip-
tion factor EGR-1, which is co-regulated with c-fos
(Chavrier et al., 1989), was also overexpressed. Recently,
EGR-1 was shown to play an important role in tumour
angiogenesis and growth (Fahmy et al., 2003). Remark-
ably, we observed a strong upregulation of insulin-like
growth factor-binding protein 1 (IGFBP1) (Table 2a),
a hepatocyte-derived and secreted protein, which is
required for liver regeneration. The recent report of Leu
et al. (2003) provided evidence for IGFBP1 to function
as a critical hepatic survival factor in the liver by
reducing the level of proapoptotic signals. Therefore,
overexpression of IGFBP1 may lead to enhanced
survival of tumorous liver cells.
Further, the role of BLNK (Table 2a), which is
involved in activation of nuclear factor NF-kB (Tan
et al., 2001) and Toll-like receptor, which activates both
NF-kB and c-jun N-terminal kinase (JNK) (Takeuchi
et al., 1999), is difficult to comprehend, because these
genes are mainly induced in immune-competent cells
after stimulation. On the other hand, their expression
in the liver would indicate an imbalance of the IGF/
IGFBP system of EGF2B transgenic mice and support
the survival of tumorous liver cells. As no infiltration of
the tumour tissue by immune cells was observed,
enhanced expression of these genes may support
a function beyond immune-competent cells. There is
Early tumour stage in EGF-induced HCC
J Borlak et al
1816
Oncogene
a clear need to clarify the role of these genes in HCC
formation.
Further, elevated expression level of TGIF, an
inhibitor of antigrowth factor TGFb-responsive trans-
cription (Melhuish et al., 2001) and transcription factor
ets-2 among others (Table 2a), can also contribute to
the initial transformation of hepatocytes in EGF-
overexpressing mice.
Overexpression of an adhesion molecule Alcam, a
member of the immunoglobulin superfamily, may
contribute to the invasive capabilities of transformed
liver cells (Choi et al., 2000). Remarkably, two genes
involved in lipid metabolism, for example, lipoprotein
lipase and stearoyl-CoA desaturase-2, were uniquely
expressed or strongly induced in liver and tumours of
transgenic mice. This suggests that important functions
of the coded genes in lipid and fatty acid metabolism of
tumour cells most probably contribute to cell membrane
synthesis.
A comparison of the expression profiles of EGF-
overexpressing tumour-free transgenic liver with those
observed in manifested small liver tumours allows an
identification of candidate genes additionally required
for malignant transformation. We observed proto-
oncogene junB, cell division cycle control protein
p34cdc2, cyclinD1, cyclinB1, Id-1, minopontin, villin,
claudin-7, ribonucleotide reductase (RR M2), Ly6d and
cell adhesion kinase, which were not changed in
transgenic liver (mostly not detected both in control
and transgenic liver), but were dramatically induced in
the pool of small tumours (Table 2a). CyclinD1 forms a
complex with cdk4 to inactivate Rb by phosphorylation
and Id-1 was shown to inactivate the p16/pRB pathway
by preventing the deactivation of cyclin/cdk complexes
in human prostate cancer (Ouyang et al., 2002). Thus,
the overexpression of cyclinD1 and Id1 in liver cells
can interfere with pRb pathway, leading to exaggerated
cell division signalling. We observed overexpression of
minopontin (osteopontin), RRM2 and Ly6d in small
tumours of EGF2B mice. Overexpression of the latter
genes was observed in human cancers and correlated
with invasiveness and metastatic potential (Chen et al.,
2000; Witz, 2000; Gotoh et al., 2002).
Many regulated genes showed permanent increase
or decrease in their expression level during carcino-
genesis. Importantly, we failed to detect Gas-1 in all
tumour samples, while its expression was evident in all
controls and a low expression in two of three
nontumorous transgenic livers (see Table 2a). This
protein is of importance in growth suppression and
it was suggested that pRb and/or p53 play an active
role in mediating the growth-suppressor effect of Gas-1
(Del Sal et al., 1994, 1995; Evdokiou and Cowled, 1998).
It appears that loss of growth control through GAS-1
may be a necessary event in the multi-step neoplastic
transformation.
Additionally, we observed a significant increase in the
transcript level of the small GTPase rhoC (Table 2a),
and of Ect2 (Table 2a), the guanine nucleotide exchange
factor for Rho GTPases (Tatsumoto et al., 1999), which
plays a critical role in Rho activation (Kimura et al.,
2000). RhoC is involved in controlling cell motility and
focal adhesion, and was recently demonstrated to be
associated with vascular invasion in human HCC
(Okabe et al., 2001) and may play a role in metastasis
of human melanoma (Clark et al., 2000).
Transcript profiling of large-size tumours evidenced
induction of PAI-1, serpin b6, calpactin (annexin II),
carboxypeptidase E and EMAP. Their specific role in
tumour progression still needs to be delineated.
It is apparent that the EGF-transgenic mouse model
is valuable for the study of HCC. Indeed, the tumours in
these animals share known features with those pre-
viously observed in humans following EGF induction or
by malignant transformation, and our analyses revealed
novel candidate genes associated with tumorigenesis.
This included Rab 3D, cell adhesion kinase, trefoil
factor 3, A6-related protein, LRG-21, cold shock
protein cbm-3, AE-binding protein, ww domain-binding
protein, Tra1, fibrinogen-like protein and tyrosinase-
related protein-2, but their specific role in liver
carcinogenesis needs to be elucidated.
In conclusion, we report tumour size-dependent gene
expression in EGF-induced HCCs. We observed en-
hanced expression of villin, cell death factor CIDE-A,
claudin 7 and junB in specifically small tumours. We
further observed induction of autocrine growth with
increased expression of TGFa, PDGFa and eps-15,
the latter being a substrate for the EGF receptor with
transforming capacity. In all tumours, induction of c-fos
and egr-1 was significant, as was the induction of the
survival factor IGFBP1, which provided tumours with
an important advantage. In large tumours, RhoC
activation was linked to vascular invasion. Finally, loss
of sensitivity to antigrowth signals could be traced back
to induction of TGIF, Id-1 and cyclinD1 to interfere
with the pRb control of cell division, whereas the
repression of the tumour suppressor gene Gas-1 allowed
for proliferation invasion and metastatic growth. In
future, promoter analyses of deregulated genes is needed
to identify the molecular rules of promoter activation
and the transcription factors acting in concert in
malignancies of the liver. Most certainly, the EGF
transgenic mouse model contributes towards a mole-
cular understanding of liver carcinogenesis
Materials and methods
Maintenance of the transgenic mouse line
The EGF2B transgenic line was described earlier by To
¨
njes
et al. (1995). Transgenic mice were maintained as hemizygotes
in the CD2F1-(DBA/2 Balb/c) background. PCR was
carried out with Platinum PCRSuperMix (InVitrogen). An-
nealing temperature and the number of cycles are indicated in
brackets after each primer pair. The transgene was verified by
PCR of DNA extracted from tail biopsies (Hogan et al., 1994)
and the following forward primer (fp) and reverse primer
(rp) pair was used for a transgene-specific amplification:
forward primer: 5
0
-CTAGGCCAAGGGCCTTGGGGGCTC
TTGCAG-3
0
; reverse primer: 5
0
-CATGCGTATTTGTCCAG
AGCTTCGATGTA-3
0
(611C, 32 cycles).
Early tumour stage in EGF-induced HCC
J Borlak et al
1817
Oncogene
Gene expression studies by RT–PCR
RT–PCR was employed to confirm expression of the transgene
and some selected genes in controls, tumours and nontumor-
ous liver of transgenic mice. The primer design was carried out
with the MacVectort 6.5.3 software and cross-reaction of
primers with other genes was excluded by comparison of the
sequence of interest with a data bank (Blast 2.0 US National
Centre for Biotechnology information). We also used UCSC
genome bioinformatics to design intron-spanning primers.
Total RNA was isolated with the Qiagen RNA purification kit
according to the manufacturer’s instructions. Reverse tran-
scription was carried out using Omniscript (Qiagen), Oligo-dT
primers (InVitrogen) and RNasin (Promega), followed by
PCR amplification (see above) with the following primer pairs:
EGF, fp: GCTGTGACGGTCCTTACAATG; rp: CAGTTC
CCACCACTTCAGGTC (611, 29 cycles). Calpactin, fp: GAG
CATCAAGAAAGAGGTCAAAGG; rp: TTCAGTCATCC
CCACCACACAG (651C, 28 cycles). LRG21, fp: AGATGAG
AGGAAAAGGAGGCGG; rp: GGGTGGAAAAGGAGGA
TTCAGTAAG (651C, 28 cycles). G7e, fp: GGTCTTTCACA
AGCAGTGCCTG; rp: AAACCAAGTTCCAATGGGGG
(571C, 32 cycles). TFF3, fp: GCAAATGTCAGAGTGGACT
GTGG; rp: GGCTGTGAGGTCTTTATTCTTCAGG (621C,
28 cycles). TGIF, fp: AACGCCTATCCCTCAGAGCAAG;
rp: GTCCAACTACGCAGGAATGAAATG (651C, 30 cycles).
b-Actin was used as a housekeeping gene, because its
expression was found to be unchanged in controls (nontrans-
genic), transgenic (nontumorous) and tumours of EGF2B mice
(microarray analyses) The following primer pair was used: fp:
GGCATTGTTACCAACTGGGACG; rp: CTCTTTGATGT
CACGCACGATTTC (651C, 25 cycles). PCR reaction pro-
ducts were separated on 1% agarose gels stained with ethidium
bromide and photographed on a transilluminator (Kodak
1544 CF). A semiquantitative measurement was carried out
using Kodak 1D software (v.3.5.3)
Histology
Tumour tissue and tumour-free tissue of transgenic animals, as
well as liver from control animals, were fixed in 4%
formaldehyde in PBS and embedded in paraffin by standard
operating procedures. Paraffin blocks were sectioned into 3–
5 mm thick slices and stained with haematoxylin and eosin (H
and E), haematoxylin only (H) and PAS for light microscopic
evaluation.
Sample collection and preparation
Mice were anaesthesized by an overdose of CO
2
at the age of
6.5–9 months. The thorax was opened by standard surgical
procedures and the liver was explanted and rinsed with PBS.
The tumours were inspected macroscopically and separated
from the liver. One group consisted of four analysed tumours
of about 5 mm in size (medium size) derived from n ¼ 4
animals, and a second group of three tumours of about 10–
15 mm (large size) were derived from n ¼ 3 animals. Five
tumours from n ¼ 3 animals were about 1 mm (small size) and
were pooled to improve the yield in RNA. Tumour-free tissue
was taken from the liver of (n ¼ 3) tumour-bearing transgenic
mice. Healthy liver from n ¼ 4 non transgenic CD2F1 mice of
about the same age were used as controls. Upon anatomical
preparation tissue was frozen immediately in liquid nitrogen.
RNA isolation and production of copy RNA
The cRNA samples were prepared following the Affymetrix
Gene Chip
s
Expression Analysis Technical Manual (Santa
Clara, CA, USA). Briefly, total RNA was isolated from frozen
tissues using QIAGEN’s RNeasy total RNA isolation
procedure. A second cleanup of isolated RNA was performed
using the same RNA isolation kit. In all, 10 mg of total
RNA was used for the synthesis of double-stranded cDNA
with Superscript II RT and other reagents from Invitrogen
Life Technologies. HPLC-purified T7-(dT)
24
(GenSet SA)
was used as a primer. After cleanup, double-stranded
cDNA was used for the synthesis of biotin-labelled cRNA
(Enzo
s
BioArray High Yield RNA Transcript Labeling Kit,
Affymetrix). cRNA purified with RNeasy spin columns from
Qiagen was cleaved into fragments of 35–200 bases by metal-
induced hydrolysis.
Array hybridization and scanning
A measure of 10 mg of biotinylated fragmented cRNA
was hybridized onto the Murine Genome U74Av2 Array
(MG-U74Av2). The array consists of 12 488 probe sets, that
represent RefSeq annotated sequences (B6000) in the Mouse
UniGene database, as well as B6000 EST clones.
The hybridized, washed and coloured arrays were scanned
using the Agilent GeneArray
s
Scanner. Scanned image files
were visually inspected for artifacts and then analysed, each
image being scaled to an all probe set intensity of 150 for
comparison between chips. The Affymetrix
s
Microarray Suite
(version 5.0) was used to control the fluidics station and the
scanner, to capture probe array data and to analyse
hybridization intensity data. Default parameters provided in
the Affymetrix data analysis software package were applied in
running of analyses.
Data analysis
The hybridization values for each gene probe presented on the
array with a set of 16 perfect and mismatch oligonucleotide
pairs were calculated within Affymetrix
s
Microarray Suite 5.0
Software, using the manufacturer’s statistical algorithm. The
results were reported as numeric expression values – signal
intensities and absolute information – detection calls ‘Present’
or ‘Absent’ produced by two independent algorithms. The
results of a single comparison analysis between two different
arrays were reported for each gene as signal logarithm ratio
(log
2
ratio) and a change call ‘Increase’ or ‘Decrease’. Multiple
data from replicate samples were evaluated and compared
using statistical analyses with the Affymetrix
s
Data Mining
Tool 3.0 (DMT). The average and standard deviation statistics
within Affymetrix
s
DMT was used to summarize the
expression level (the signal values) for each transcript across
the replicates. The unpaired one-sided T-test converting
P-value to a two-sided P-value was used to determine the
direction and significance of change in a transcript’s expression
level between sets of tumours (except for the pool of small
tumours), transgenic liver and normal livers, with the P-value
cutoff determined as 0.05. Besides, only those genes that were
detected (had call ‘Present’) in all samples of a tumour set or
transgenic liver for the upregulated genes and in all control
livers for the downregulated genes were taken into considera-
tion as differentially expressed. Fold-change values were
calculated as the ratio of the average expression levels for
each gene between two tissue sets. Comparison ranking
analysis was additionally employed to study the concordance
of gene expression changes in pairwise comparisons of tumour
samples with control livers. The results are shown as % of
‘Increase’ or ‘Decrease’ calls in individual comparisons, for
example, 16 analyses (four tumours versus four controls) for
Early tumour stage in EGF-induced HCC
J Borlak et al
1818
Oncogene
tumours of median size, 12 (three tumours versus four
controls) for tumours of large size and 12 analyses for
transgenic liver. The small tumours (B1 mm) were pooled
and compared with the four individual controls, resulting in
four comparisons.
Acknowledgements
The excellent technical assistance of Mrs Edith Aretz, Ms Ines
Noack and Mr Albert Rast is gratefully acknowledged. We
thank the Lower Saxony Ministry of Culture and Sciences and
the Volkswagen foundation for providing a grant to J Borlak.
References
Alvarez CV, Shon KJ, Miloso M and Beguinot L. (1995).
J. Biol. Chem., 270, 16271–16276.
Chavrier P, Janssen-Timmen U, Mattei MG, Zerial M, Bravo
R and Charnay P. (1989). Mol. Cell. Biol., 9, 787–797.
Chen S, Zhou B, He F and Yen Y. (2000). Antisense Nucleic
Acid Drug Dev., 10, 111–116.
Choi S, Kobayashi M, Wang J, Habelhah H, Okada F,
Hamada J, Moriuchi T, Totsuka Y and Hosokawa M.
(2000). Clin. Exp. Metastasis, 18, 45–50.
Chung YH, Kim JA, Song BC, Lee GC, Koh MS, Lee YS,
Lee SG and Suh DJ. (2000). Cancer, 89, 977–982.
Clark EA, Golub TR, Lander ES and Hynes RO. (2000).
Nature, 406, 532–535.
Del Sal G, Collavin L, Ruaro ME, Edomi P, Saccone S, Valle
GD and Schneider C. (1994). Proc. Natl. Acad. Sci. USA, 91,
1848–1852.
Del Sal G, Ruaro EM, Utrera R, Cole CN, Levine AJ and
Schneider C. (1995). Mol. Cell. Biol., 15, 7152–7160.
Dey A, Nebert DW and Ozato K. (1991). DNA Cell Biol., 10,
537–544.
Evdokiou A and Cowled PA. (1998). Exp. Cell Res., 240,
359–367.
Gotoh M, Sakamoto M, Kanetaka K, Chuuma M and
Hirohashi S. (2002). Pathol. Int., 52, 19–24.
Hogan B, Beddingtin R, Costantini F and Lacy E. (1994).
Manipulating the Mouse Embryo: A Laboratory Manual.
Cold Spring Harbor Laboratory: Cold Spring Harbor, NY.
Fahmy RG, Dass CR, Sun LQ, Chesterman CN and
Khachigian LM. (2003). Nat. Med., 9, 1026–1032.
Kimura K, Tsuji T, Takada Y, Miki T and Narumiya S.
(2000). J. Biol. Chem., 275, 17233–17236.
Kubicka S, Rudolph KL, Hanke M, Tietze MK, Tillmann
HL, Trautwein C and Manns M. (2000). Liver, 20,
312–318.
Leu JI, Grissey MA and Taub R. (2003). J. Clin. Invest., 111,
129–139.
Melhuish TA, Gallo CM and Wotton D. (2001). J. Biol.
Chem., 276, 32109–32114.
Okabe H, Satoh S, Kato T, Kitahara O, Yanagawa R,
Yamaoka Y, Tsunoda T, Furukawa Y and Nakamura Y.
(2001). Cancer Res., 61, 2129–2137.
Ostrowski J, Woszczynski M, Kowalczyk P, Wocial T, Hennig
E, Trzeciak L, Janik P and Bomsztyk K. (2000). Br. J.
Cancer, 82, 1041–1050.
Ouyang XS, Wang X, Ling MT, Wong HL, Tsao SW and
Wong YC. (2002). Carcinogenesis, 23, 721–725.
Ramljak D, Jones AB, Diwan BA, Perantoni AO, Hochadel
JF and Anderson LM. (1998). Cancer Res., 58, 3590–3597.
Rescan C, Coutant A, Talarmin H, Theret N, Glaise D,
Guguen-Guillouzo C and Baffet G. (2001). Mol. Biol. Cell.,
12, 725–738.
Sandgren EP, Luetteke NC, Qiu TH, Palmiter RD, Brinster
RL and Lee DC. (1993). Mol. Cell. Biol., 13, 320–330.
Schafer DF and Sorrell MF. (1999). Lancet, 35, 1253–1257.
Takeuchi O, Kawai T, Sanjo H, Copeland NG, Gilbert
DJ, Jenkins NA, Takeda K and Akira S. (1999). Gene, 231,
59–65.
Tan JE, Wong SC, Gan SK, Xu S and Lam KP. (2001). J. Biol.
Chem., 276, 20055–20063.
Tatsumoto T, Xie X, Blumenthal R, Okamoto I and Miki T.
(1999). J. Cell. Biol., 147, 921–928.
To
¨
njes RR, Lohler J, Oı
´
Sullivan JF, Kay GF, Schmidt GH,
Dalemans W, Pavirani A and Paul D. (1995). Oncogene, 10,
755–758.
Wang Y, Ripperger J, Fey GH, Samols D, Kordula T, Wetzler
M, Van Etten RA and Baumann H. (1999). Hepatology, 30,
682–697.
Yamaguchi K, Carr BI and Nalesnik MA. (1995). J. Surg.
Oncol., 58, 240–245.
Witz IP. (2000). J. Cell. Biochem. Suppl., 34, 61–66.
Wu JC, Merlino G, Cveklova K, Mosinger Jr B and Fausto N.
(1994). Cancer Res., 54, 5964–5973.
Early tumour stage in EGF-induced HCC
J Borlak et al
1819
Oncogene