Genome-Wide Molecular Profiles of HCV-Induced
Dysplasia and Hepatocellular Carcinoma
Elisa Wurmbach,1Ying-bei Chen,1Greg Khitrov,1Weijia Zhang,1Sasan Roayaie,1Myron Schwartz,1Isabel Fiel,1
Swan Thung,1Vincenzo Mazzaferro,2Jordi Bruix,3Erwin Bottinger,1Scott Friedman,1
Samuel Waxman,1and Josep M. Llovet1,3
is the main etiology underlying this cancer’s accelerating incidence. To characterize the
HCC, the gene expression profiles of 75 tissue samples were analyzed representing the
stepwise carcinogenic process from preneoplastic lesions (cirrhosis and dysplasia) to HCC,
infection. We identified gene signatures that accurately reflect the pathological progression
of disease at each stage. Eight genes distinguish between control and cirrhosis, 24 between
cirrhosis and dysplasia, 93 between dysplasia and early HCC, and 9 between early and
novel molecular tissue markers for early HCC diagnosis, specifically induction of abnormal
spindle-like, microcephaly-associated protein, hyaluronan-mediated motility receptor, pri-
mase 1, erythropoietin, and neuregulin 1. In addition, pathway analysis revealed dysregu-
lation of the Notch and Toll-like receptor pathways in cirrhosis, followed by deregulation of
several components of the Jak/STAT pathway in early carcinogenesis, then upregulation of
genes involved in DNA replication and repair and cell cycle in late cancerous stages. Con-
clusion: These findings provide a comprehensive molecular portrait of genomic changes in
progressive HCV-related HCC. (HEPATOLOGY 2007;45:938-947.)
is increasing in Europe and the United States2and is
currently the leading cause of death among individuals
who have cirrhosis.3Hepatitis C virus (HCV) infection is
the main risk factor in western countries and Japan,2,3
with one third developing HCC over their lifetime.
No clear sequence of genetic events can yet explain the
been proposed in which external stimuli induce genetic
plasia over 20 to 30 years.5Global gene expression
esis of HCC and explore its heterogeneous origins.7
Attempts to study risk factors (e.g., HBV or HCV
infection) and alcohol-induced HCC8-10or different
stages of the disease,11-13recurrence,14,15and survival16,17
have been inconsistent. To reduce this complexity, we
focused exclusively on gene expression in eight patholog-
ical stages of HCV-induced hepatocarcinogenesis and
Potential conflict of interest: Nothing to report.
common cancer worldwide, with 626,000 new
Abbreviations: HCC, hepatocellular carcinoma; PAM, prediction analysis of
microarray; qPCR, quantitative real-time reverse-transcriptase PCR; SAM, signif-
icance analysis of microarray.
From the1Mount Sinai Liver Cancer Program (Division of Liver Disease, Di-
vision of Hematology/Oncology, Department of Medicine; Recanati-Miller Trans-
plantation Institute; Department of Pathology), Mount Sinai School of Medicine,
New York, NY; the2National Cancer Institute, Milan, Italy; and the3BCLC
Group, IDIBAPS, Liver Unit, Hospital Clinic, Barcelona, Spain.
Received October 9, 2006; accepted November 30, 2006.
Supported by the Samuel Waxman Cancer Research Foundation, the National
Institutes of Health (grant DK37340 to S. F.), the Bendheim Family Trust (to
S. F.), the Instituto de Salud Carlos III (FIS 05/1285 to J. B.) and the Italian
Association for Cancer Research (to V. M.).
Josep M. Llovet is Professor of Research at Institut Catala ` de Recerca Avancada
(ICREA, IDIBAPS, Hospital Clı ´nic).
Address reprint requests to: Elisa Wurmbach, Ph.D., Division of Hematology/
Oncology, Department of Medicine, Mount Sinai School of Medicine, Madison
edu. Josep M. Llovet, M.D., Division of Liver Diseases, Department of Medicine,
York, NY 10029. E-mail: firstname.lastname@example.org; fax: 212-241-2138/212-
Copyright © 2007 by the American Association for the Study of Liver Diseases.
Published online in Wiley InterScience (www.interscience.wiley.com).
compared their differential expression to that of normal
Early detection of HCC followed by appropriate treat-
ment can decrease tumor-related deaths.3A variety of
genomic studies have identified potential biomarkers for
detection of early HCC,18such as HSP70,19GPC3,20
TERT, STK15, and PLA2.12We recently identified a
3-gene signature (GPC3, XLKD1, BIRC5) capable of
on the analysis of 56 potential candidate genes.21The
search for new and more accurate markers must be trans-
lated into the clinical setting, where they can advance
Patients and Methods
Samples were obtained from patients undergoing re-
section or liver transplantation at 3 university hospitals in
and Europe (Hospital Clı ´nic, Barcelona, Spain, and Na-
tional Cancer Institute, Milan, Italy). Laboratory tech-
niques were centralized in the laboratories of the
Divisions of Liver Disease and Hematology/Oncology
of Medicine. The research protocol was approved by the
respective institutional review boards, and informed con-
sent was obtained in all cases.
Sample Characteristics. The characteristics of the 75
samples (48 patients) selected to assess the gene transcrip-
tional profiles are described in Supplementary Table 1.
Sixty-five samples were obtained from 38 patients with
HCV infection representing 13 samples from cirrhotic
HBV-positive markers, or a background of alcohol con-
sumption, nonalcoholic steatohepatitis, hemochromato-
sis, other causes of chronic liver disease or prior
locoregional treatment were excluded. Results were com-
pared with samples of normal tissue obtained from the
healthy livers of 10 patients undergoing resection for he-
patic hemangioma (1), focal nodular hyperplasia (3), ad-
enoma/cystadenoma (2), neuroendocrine tumor (1), and
living donor liver transplantation (1).
Sample Collection and Pathological Data. After
written informed consent was obtained, key clinicopath-
ological variables were recorded. Fresh tissue specimens
were collected in the operating room/pathology depart-
ment and processed within 1 hour to minimize degrada-
tion. Samples were split: one part was collected in liquid
nitrogen or RNA-later solution (Ambion, The Wood-
lands, TX), and stored at ?80°C until use, and the other
was paraffin-embedded for pathological examination. In
cases of liver transplantation, explanted livers were sec-
terms of size, color, texture, or degree of bulging were
examined microscopically. Tissue sampling was handled
using thin sections (4 ?m) of the target area, which was
microdissected under a scanning microscope for PCR
Pathological classification of the samples was estab-
lished by 2 expert pathologists (S. T. and I. F.) who re-
viewed each slide independently according to the
International Working Party.22Four pathological HCC
stages were defined among the 35 target samples: (i) very
early HCC (8 cases), which included well-differentiated
tumors ?2 cm in diameter with no vascular invasion/
which included tumors measuring ?2 cm with micro-
ferentiated tumors measuring 2-5 cm without vascular
invasion/satellites; or 2-3 well-differentiated nodules
measuring ?3 cm (size range: 3-45 mm); (iii) advanced
HCC (7 cases), which included poorly differentiated tu-
mors measuring ?2 cm with microvascular invasion/sat-
ellites or tumors measuring ?5 cm; and (iv) very
advanced HCC (10 cases), which included tumors with
macrovascular invasion or diffuse liver involvement.
Microarray Hybridization. RNA extraction was
performed as described previously.2128S/18S ratios mea-
sured with a Bioanalyzer (Agilent Technologies, Palo
Table 1). Five micrograms of purified total RNA was
labeled according to the manufacturer’s protocol (Af-
fymetrix, Santa Clara, CA). Fragmentation of the com-
plementary RNA to 50-200 bp was confirmed by the
Bioanalyzer. Fifteen micrograms of fragmented comple-
mentary RNA was hybridized on the human U133 plus
2.0 array (Affymetrix). To minimize variations from bio-
logical samples, we included in the microarray analysis
only samples with a present call between 38% and 47%,
and a 3? to 5? ratio for GAPDH of ?3.18 and for ACTB
of ?10.41 (Supplementary Table 1).
Microarray Data Analysis. The raw data were ana-
lyzed and processed in GeneTraffic (Stratagene, La Jolla,
CA). Data were normalized by applying the GC robust
multi-array average algorithm, and the baseline was cal-
culated using the geometric mean from the data of the 10
normal liver tissue samples. Hierarchical clustering was
used to detect overall expression patterns in specimens of
different stages. The microarray data were analyzed from
73 samples (2 samples were excluded for quality reasons).
Analysis included the significant analysis of microarray
(SAM) data and the predictor of microarray analysis
The significance level was chosen not to exceed 5% (P ?
0.05). Three cirrhotic samples from patients without
HEPATOLOGY, Vol. 45, No. 4, 2007WURMBACH ET AL. 939
HCC were excluded from validation tests, because no
difference was identified to the gene expression of cir-
rhotic tissue from patients with HCC.
Information about genes participating in known sig-
naling pathways were derived from Entrez Gene (http://
www.ncbi.nlm.nih.gov/entrez/ query.fcgi?db?gene) and
way.html) databases. To identify members of particular
This report has been written to conform to the Mini-
mum Information About a Microarray Experiment
guidelines.23The microarray data will be available at
www.ncbi.nlm.nih.gov/projects/geo (GSE 6764).
via quantitative real-time reverse-transcriptase PCR
(qPCR). Five micrograms of total purified DNase-I–
treated RNA was converted into complementary DNA
using oligo dT and Superscript III (Invitrogen, Carlsbad,
CA), followed by RNaseH (Invitrogen, Carlsbad, CA)
digestion. The complementary DNA was diluted 1/100
titative PCR (qPCR) assays were performed as de-
scribed,24with few modifications. The qPCR was
OR) and Platinum Taq (Invitrogen, Carlsbad, CA) on
the ABI Prism 7900 (Applied Biosystems, Foster City,
CA). Primers are listed in Supplementary Table 2. Am-
plicon size and reaction specificity were confirmed by
agarose gel electrophoresis and melting curve analysis.
All PCR reactions were run in triplicate. The raw data
were analyzed after subtracting the background and
setting the threshold to obtain the Ct value using
SDS2.2 (Applied Biosystems, Foster City, CA) in Ex-
cel; the median Ct was taken from triplicate reactions
and normalized to RPL41 and SFRS4. Results were
expressed as fold changes.
Gene Expression Patterns Confirm Pathological
Classification. We compared the gene expression pro-
files of 75 tissue samples using the genome-wide platform
of the expression profiles and their correlation with the
was performed using the Pearson correlation distance
(Fig. 1). Two main clusters were formed, one that in-
cluded 34/35 HCC samples (Fig. 1, in red) and one in-
cluding all 30 preneoplastic lesions. The normal samples
remained outside of these 2 main clusters. Only 1 neo-
plastic sample (ve-NY26) was misclassified (Fig. 1). This
very early HCC sample clustered together with dysplastic
nodules obtained from the same patient. Within the can-
cer cluster, the very early, early, and very advanced stages
of HCC formed subclusters; the precancerous cluster was
split into cirrhotic and dysplastic subclusters. Different
cluster algorithms (Manhattan- and Pearson-centered
and absolute) led to similar results (data not shown) sup-
porting the conclusion that the gene expression profiles
robustly reflected the histological classification.
Unique Molecular Markers of Individual Stages of
HCV-Induced HCC. To identify reliable molecular
markers of HCV-induced HCC stages, we applied two
statistical tests: SAM, which detect genes significantly
a subset of genes that best characterize each group on the
basis of “nearest shrunken centroids” (the more different
them). For the analyses, 5 groups were considered: con-
Fig. 1. Hierarchical clustering of 75 tissue samples included in the HCV-induced HCC study using the Pearson correlation distance. Tissue samples
are indicated on top (see Supplementary Table 1 for characteristics). The colored bars mark clusters: green, precancer; red, HCC; blue, controls.
Subclusters are marked in yellow, orange, red, green, and turquoise.
940 WURMBACH ET AL.HEPATOLOGY, April 2007
Very early and early HCC, as well as advanced and very
advanced HCC, were combined to enhance the statistical
power of the comparison of microarray data.
The differential expression of all genes found by SAM
the number of genes that were significantly differentially
expressed is summarized in Table 1. The PAM analysis
was optimized by testing different thresholds, an ap-
a robust prediction (Supplementary Fig. 2). The number
of genes found by PAM is displayed in Table 1. To define
the optimal set of genes able to delineate hepatocarcino-
genesis, we selected genes identified by both tests. Figure
samples. Roughly 50% of these genes were down-regu-
lated in HCC. For the up-regulated genes in HCC, num-
ber and magnitude was greater for advanced stages. A
few genes showed upregulation in dysplasia. Based on
the expression of these identified genes (Fig. 2A), PAM
was used for each of the 72 samples to determine the
probability of being assigned to one of the pathological
stages of hepatocarcinogenesis. As can be seen in Fig.
2B, 64 of 72 samples (89%) had a probability of almost
1 (100% fit), whereas the remaining had a suboptimal
probability and thus did not fit the classification. The
results were comparable to those obtained in the orig-
inal PAM analysis (compare Supplementary Fig. 2 with
Table 1. Genes Deregulated in Progression of HCV-Related HCC
GroupSAMPAM (Threshold) No. of Genes in the Signature
Control (n ? 10) vs. cirrhosis (n ? 13)
Cirrhosis vs. dysplasia (n ? 17)
Dysplasia vs. early HCC (n ? 18)
Early HCC vs. advanced HCC (n ? 17)
Dysplasia vs. HCC (n ? 35)
NOTE. The number of genes identified via microarray analysis defining all step-by-step carcinogenic stages from preneoplastic lesions to advanced HCC detected by
statistical tests is shown. The combination of SAM and PAM analyses identified the gene signatures. Thirty-one genes were detected in more than one comparison
(187 ? 31 ? 156). One gene was significantly changed between cirrhosis, dysplasia, and early HCC, and 30 genes were consistently identified in the comparisons
between dysplasia and HCC and dysplasia and early HCC.
Fig. 2. Molecular markers for HCV induced HCC. (A) Heat map of the gene signature, identified by SAM and PAM. The 72 samples are represented
in columns (c, control; ci, cirrhosis; dn, dysplasia; ve/e, early HCC; a/aa, advanced HCC) and the expression pattern (red, up-regulation; green,
downregulation; black, no change) of the genes is shown in rows. (B) PAM analysis, using the genes displayed in (A) to calculate the probability of
each sample being assigned to one of the five stages of HCV-HCC as predicted by each sample’s gene signature (c, control; ci, cirrhosis; dn, dysplasia;
e, early HCC; a, advanced HCC). The x-axis represents the sample as indicated at the top of the figure. The y-axis represents the calculated probability,
the sum of which must be 1 for each sample.
HEPATOLOGY, Vol. 45, No. 4, 2007 WURMBACH ET AL.941
In summary, 8 key genes can differentiate between
control and cirrhotic samples, 24 between cirrhosis and
dysplasia, 53 between dysplasia and HCC, and 93 be-
tween dysplasia and early tumors, whereas 9 genes were
found to define the progression of HCC (Table 1 and
Supplementary Table 3). Multiclass analysis did not led
to such informative results (data not shown).
Gene markers for a specific disease stage should be
unique and expressed at elevated levels. Eight genes were
chosen fitting these criteria for confirmation via qPCR:
one gene for cirrhosis (CLDN10), three genes for dyspla-
sia (GREM2, EPO, and NRG1); and four genes as po-
were indeed differentially expressed (Fig. 3). Student t
HCC, as well as between dysplasia and early HCC, re-
sulted in P values below 0.05 for the differential expres-
molecular markers for different stages of HCC.
Significantly Regulated Genes With Known Biolog-
ical Functions. To examine the SAM deregulated genes
(Table 1 and Supplementary Fig. 1) in the context of the
initiation and progression of HCC, we grouped the genes
according to their biological function using the informa-
gov/entrez/query.fcgi?db?gene) and KEGG (http://
www.genome.jp/kegg/pathway.html) databases. Table 2
shows the number of genes significantly deregulated with
at least 2 genes per function to emphasize the functions
with most genes differentially expressed.
In cirrhotic liver samples, genes with functions in the
immune response and cell adhesion were upregulated,
and genes involved in metabolism showed varying
changes. These results reflect the transition from normal
to cirrhosis, where the liver function becomes impaired
and extracellular matrix deposition increases.25Only a
few genes were significantly down-regulated between cir-
rhosis and dysplasia, most likely because of the heteroge-
neity of the dysplastic nodules (Supplementary Fig. 1B).
In contrast, the onset of HCC was associated with
major changes. Genes whose products had functions in
cell cycle, protein biosynthesis and RNA processing, cell
division, DNA replication, protein modification, ubiq-
uitin cycle, or chromatin modulation were up-regulated.
receptor interactions, Ca signaling, the Jak/STAT path-
way, and blood coagulation were down-regulated. Fur-
characterized by additional up-regulation of genes in-
volved in cell cycle regulation.
Pathways Affected in HCV-Induced HCC. To gain
more insight into the types of pathways that might be
affected during HCV-induced HCC, we chose 15 major
pathways of signaling, processing of genetic information,
and cell growth and death (Supplementary Table 4 and
Fig. 3. Quantitative PCR of 8 molecular markers: (A) CLDN10; (B) GREM2; (C) EPO; (D) NRG1; (E) ASPM; (F) PRIM1; (G) HMMR; and (H) IRAK1.
Quantitative PCR was performed on 72 tissue samples. P values for relevant stage transitions are indicated below each panel. Medians of normalized
fold changes, with standard deviations, were plotted for each disease stage.
942 WURMBACH ET AL.HEPATOLOGY, April 2007
Cirrhosis showed strong up-regulation of JAG1, a li-
gand of the Notch receptor, and participants of the Toll-
like receptor pathway such as STAT1 and CXCL9-11,
IFN-?–inducible genes.26Dysplasia was characterized by
up-regulation of genes involved in cytokine–cytokine re-
ceptor interactions and the Jak/STAT pathway (EPO,
EPOR, CISH, STAT3, and SOCS3) as well as cell cycle
regulation (GADD45B, GADD45G, and GADD45A).
Early HCCs were marked by striking changes in all 15
pathways—in particular, the down-regulation of several
components of the Toll-like receptor pathway (IFNAR1,
TLR4, FOS, and CD14), tumor suppressors of the Jak/
STAT pathway (EPO, EPOR, SPRY2, and SOCS2) as
well as participants of the transforming growth factor ?
pathway (BMPR2, ID2, THBS1, and DCN), and the
insulin-signaling pathway (FBP1, PCK1, PCK2, GYS2,
FOXO1A, and SOCS2). Up-regulation was found for
components of the wnt pathway (DKK1, FZD6, FZD7,
Table 2. Numbers of Significantly Regulated Genes with Known Biological Functions Detected via SAM Analysis
HCC OnsetHCC Progression
Early (ve/e) vs.
Up Down UpDownUp Down UpDownUpDown
Toll-like receptor pathway
Abbreviations: C, control; CI, cirrhosis; DN, dysplasia; TGF-?, transforming growth factor ?.
*Function/pathways were included if containing at least 2 deregulated genes.
HEPATOLOGY, Vol. 45, No. 4, 2007 WURMBACH ET AL. 943
Finally, progression of HCC was characterized by
strong up-regulation of genes involved in cell prolifera-
tion, DNA repair and replication (PRIM1, PRIM2),
(ASPM, PTTG1, CCNB1, CDKN2C, and CDKN2A).
Cell Cycle Regulation. Cell cycle regulation was
the most widely affected pathway in HCC, and corre-
lated best with the progression of cancer. CDC14B,
MAD1L1, GADD45G, GADD45B, GADD45A, HDAC6,
CDKN1A, and CCND1 were up-regulated in dys-
plastic nodules followed by down-regulation in early
HCC. Confirmatory qPCR data were obtained for
GADD45A, GADD45B, and HDAC6 on all tissue
samples (Fig. 4A). In HCC, 56 genes showed progres-
sive increase during the transition from very early to
very advanced stages (Fig. 4B and Supplementary Fig.
3). The up-regulation of 4 of these genes (CCNB1,
CDKN2C, CDKN2A, and PTTG1) was confirmed via
qPCR (Fig. 4A).
Some of these cell cycle deregulated genes have inhib-
itory functions, which is consistent with our data:
GADD45 (down-regulated in HCC) inhibits CCNB
inhibits CCND (not significantly regulated in HCC);
CDKN1A (not significantly regulated in HCC) inhibits
CCNE and CCNA (both up-regulated in HCC).
Through analysis of gene expression profiles of 75
samples, representing the entire spectrum of progression
of HCV-HCC, we have identified a set of genes defining
a molecular portrait of progressive stages of hepatocarci-
nogenesis. This set discriminates between each of the
stages from normal, cirrhotic, and dysplastic to early and
advanced HCC. Moreover, a detailed analysis yielded
new biomarkers that may be used for early detection of
Fig. 4. Cell cycle involved in HCV-related HCC. (A) Quantitative PCR of GADD45A, GADD45B, HDAC6, CCNB1, CDKN2C, CDKN2A, and PTTG1
performed on 72 tissue samples. P values for relevant stage transitions are indicated below each panel. Medians of normalized fold changes, with
standard deviations, were plotted for each disease stage. (B) Heat map of 56 genes involved in cell cycle regulation (corresponding to a region from
Supplementary Fig. 3, enlarged).
944WURMBACH ET AL.HEPATOLOGY, April 2007
sion pattern of a subgroup of lesions in either HBV-in-
duced HCC, or from patients with mixed etiologies
(HBV, HCV, hemochromatosis, alcohol, chronic cho-
lestasis, and others),9-16,27-29we have instead comprehen-
sively analyzed each of the stages of HCV-induced HCC,
the main etiology underlying the increasing HCC inci-
dence in the United States and Europe.2,3In contrast to
some studies that compared tumor tissues with the sur-
rounding tissue,9,12we used normal liver tissues as a base-
line. This approach enabled us to detect differential
expression between normal liver and all stages involved in
the hepatocarcinogenic process, including cirrhosis.
Screening programs in the western countries and in
in 30%-60% of cases,3,30and new biomarkers are ur-
gently needed.18,30Moreover, serum biomarkers such as
AFP, des-gamma-carboxyprothrombin, and AFP-L3
fraction are not reliable for early diagnosis.31
ing HSP70, GPC3, STK15, and PLA2.12,19,20A molecular
index including a 13-gene set has also been proposed (in-
cluding TERT, TOP2A, CXCL12, and IGF2).32More re-
cently, a microarray-generated signature of 120 genes
reportedly discriminates between dysplastic nodules and
HCC in HBV patients.13Recently, we reported a 3-gene
tic lesions from early tumors via qPCR and immunohisto-
chemistry.21These 3 genes, GPC3, XLKD1, and BIRC5,
were confirmed by microarray analysis to be significantly
tures (Supplementary Table 3). Despite differences in study
design between our report and those examining different or
ing tissues, many of the dysregulated genes were found in
common with prior studies (GPC3, XLKD1, BIRC5,
TOP2A, PTTG1, BUB1, DKK1, CCNB1, HMMR,
CDKN2C, MCM7, GALNT109,11,13,15,21,27,29). However,
we have extended this analysis further by validating those
genes that were significantly up-regulated in specific disease
stages with qPCR. This approach identified 3 genes that
were elevated in dysplasia (GREM2, EPO and NRG1), 7
that were elevated in HCC (PRIM1, ASPM, HMMR,
PTTG1, CCNB1, CDKN2C, and CDKN2A), and 1 that
Notably, the significance of these differential expressions, in
particular during the transition from preneoplastic to early
HCC was very high (Figs. 3, 4). Four of these genes
(HMMR, PTTG1, CCNB1, and CDKN2C) have been
linked previously to HCC,9,11,15,27,29and the receptor of
EPO was downregulated in HCC.28PRIM1, ASPM, and
HMMR were associated with other neoplasms, but none of
these genes has been tested yet as a serum or tissue marker.
PRIM1 plays an essential role in cell proliferation, and am-
bladder cancer, breast cancer, and osteosarcoma.33PRIM1
ASPM may regulate neural stem cell proliferation and/or
edly up-regulated in ovarian cancer.34Finally, HMMR, an
oncogene, encodes a cell surface receptor that mediates mo-
tility of many cell types and acts downstream of Ras.35
Current studies have been unable to define either the
critical hits and sequence of events leading to HCC,4,5or
its cellular origins.6Once established, the canonical wnt
pathway is activated in at least one-third of HCCs,4,5,36-38
cases,4,5activation of the Ras/MAPK pathway has been
tribute to this neoplasm, making it difficult to establish
which of these are critical.4,5,36
The present study describes dysregulated genes
throughout all stages of HCV-induced HCC. This ap-
proach, unique in its design, detected novel genes dys-
regulated in the early stages of hepatocarcinogenesis,
which may play an important role in the progression of
immune response (including IFN-?–inducible genes
such as CXCL9-11), cell adhesion, and metabolism were
differentially expressed, in agreement with published da-
ta.25,26Conversely, the strong up-regulation of JAG1 in
prostate cancer and cervical tumors.40In addition, several
components of the Jak/STAT and the Toll-like receptor
pathway were deregulated in precancerous tissues and
early HCC. STAT1 was up-regulated in cirrhosis, whose
gene product can be bound by the HCV core protein and
plays an important role in antiviral immunity.41Several
components of the Jak/STAT pathway—particularly
EPO, EPOR, CISH, STAT3, and SOCS3—were also
up-regulated in dysplasia and may yield new insights into
well as of SOCS2 and of SPRY2 in HCC is consistent
becomes activated in early HCC.39
Another novel finding is the increase in GREM2 mes-
senger RNA in dysplasia. GREM2 is a BMP antagonist
www.genome.jp/kegg/pathway.html), several compo-
nents of which were also down-regulated in HCC, in
accordance with published data.43Components of the
HEPATOLOGY, Vol. 45, No. 4, 2007 WURMBACH ET AL. 945
wnt/beta-catenin pathway, such as FZD7 and FZD6,
were up-regulated at early stages of HCC. FZD7, a wnt
receptor, was overexpressed in 90% of HBV-induced
On the other hand, DKK1, a wnt inhibitor, was also
up-regulated in HCC, similar to another HCC microar-
PTCH, a critical receptor of the Hedgehog pathway, was
up-regulated, suggesting a potential role in early carcino-
Many microarray studies confirm the cell cycle dereg-
to be the most affected pathway, involving 56 genes, with
novel alterations present in dysplasia (up-regulation of
product of MAD1L1 has an antiproliferative function in
that it suppresses TERT,46which is up-regulated in
HCC.21,47MAD1L1 is decreased in cancer.48This gene
product also affects the coordination of ribosome biogen-
esis and cell growth, which is consistent with our data.
In HCC, GADD45 is decreased,49allowing the in-
crease of CCNB1, which promotes the G2/M transition
and is elevated in cancers,50including HCC. In addition,
CDKN2C and CDKN2A messenger RNAs were signifi-
cantly up-regulated in HCC. Increased CDKN2C mes-
senger RNA has been detected repeatedly in HCC via
microarray analysis,9,12whereas CDKN2A is up-regu-
lated in pancreatic tumors. The products are likely tumor
suppressors, because both inhibit the CDK4 kinase
gene), which associates with CCND1.
In conclusion, we identified gene signatures that accu-
rately distinguish the pathological stages of HCC and
have uncovered potential molecular markers for early di-
agnosis. Furthermore, several components of signaling
pathways are dysregulated in preneoplastic stages and are
further altered during the transition to HCC—in partic-
ular, components of the Jak/STAT signaling pathway.
The progression of HCC was best characterized by pro-
gressive up-regulation of 56 cell cycle genes. Combined,
these data provide a basis for optimizing the selection of
molecular targets for HCC therapies.
of the manuscript and to Liliana Ossowski for helpful
Many thanks to Andreas Jenny,
1. Parkin DM, Bray F, Ferlay J, Pisani P. Global cancer statistics, 2002. CA
Cancer J Clin 2005;55:74-108.
2. El-Serag HB, Mason AC. Rising incidence of hepatocellular carcinoma in
the United States. N Engl J Med 1999;340:745-750.
3. Llovet JM, Burroughs A, Bruix J. Hepatocellular carcinoma. Lancet 2003;
4. Bruix J, Boix L, Sala M, Llovet JM. Focus on hepatocellular carcinoma.
Cancer Cell 2004;5:215-219.
5. Thorgeirsson SS, Grisham JW. Molecular pathogenesis of human hepato-
cellular carcinoma. Nat Genet 2002;31:339-346.
6. Farazi PA, DePinho RA. Hepatocellular carcinoma pathogenesis: from
genes to environment. Nat Rev Cancer 2006;6:674-687.
7. Thorgeirsson SS, Lee JS, Grisham JW. Molecular prognostication of liver
cancer: end of the beginning. J Hepatol 2006;44:798-805.
8. Delpuech O, Trabut JB, Carnot F, Feuillard J, Brechot C, Kremsdorf D.
Identification, using cDNA macroarray analysis, of distinct gene expres-
sion profiles associated with pathological and virological features of hepa-
tocellular carcinoma. Oncogene 2002;21:2926-2937.
9. Okabe H, Satoh S, Kato T, Kitahara O, Yanagawa R, Yamaoka Y, et al.
Genome-wide analysis of gene expression in human hepatocellular carci-
nomas using cDNA microarray: identification of genes involved in viral
carcinogenesis and tumor progression. Cancer Res 2001;61:2129-2137.
10. Iizuka N, Oka M, Yamada-Okabe H, Mori N, Tamesa T, Okada T, et al.
Comparison of gene expression profiles between hepatitis B virus- and
croarray data on the basis of a supervised learning method. Cancer Res
11. Ye QH, Qin LX, Forgues M, He P, Kim JW, Peng AC, et al. Predicting
hepatitis B virus-positive metastatic hepatocellular carcinomas using gene
expression profiling and supervised machine learning. Nat Med 2003;9:
Identification of novel tumor markers in hepatitis C virus-associated hep-
atocellular carcinoma. Cancer Res 2003;63:859-864.
13. Nam SW, Park JY, Ramasamy A, Shevade S, Islam A, Long PM, et al.
Molecular changes from dysplastic nodule to hepatocellular carcinoma
through gene expression profiling. HEPATOLOGY 2005;42:809-818.
14. Kurokawa Y, Matoba R, Takemasa I, Nagano H, Dono K, Nakamori S, et
al. Molecular-based prediction of early recurrence in hepatocellular carci-
noma. J Hepatol 2004;41:284-291.
15. Iizuka N, Oka M, Yamada-Okabe H, Nishida M, Maeda Y, Mori N, et al.
Oligonucleotide microarray for prediction of early intrahepatic recurrence
16. Lee JS, Chu IS, Heo J, Calvisi DF, Sun Z, Roskams T, et al. Classification
and prediction of survival in hepatocellular carcinoma by gene expression
profiling. HEPATOLOGY 2004;40:667-676.
17. Lee JS, Heo J, Libbrecht L, Chu IS, Kaposi-Novak P, Calvisi DF, et al. A
hepatic progenitor cells. Nat Med 2006;12:410-416.
18. Marrero JA, Lok AS. Newer markers for hepatocellular carcinoma. Gas-
19. Chuma M, Sakamoto M, Yamazaki K, Ohta T, Ohki M, Asaka M, et al.
Expression profiling in multistage hepatocarcinogenesis: identification of
20. Capurro M, Wanless IR, Sherman M, Deboer G, Shi W, Miyoshi E, et al.
Glypican-3: a novel serum and histochemical marker for hepatocellular
carcinoma. Gastroenterology 2003;125:89-97.
21. Llovet JM, Chen Y, Wurmbach E, Roayaie S, Fiel MI, Schwartz M, et al.
A molecular signature to discriminate dysplastic nodules from early hepa-
tocellular carcinoma in HCV cirrhosis. Gastroenterology 2006;131:1758-
22. International Working Party. Terminology of nodular hepatocellular le-
sions. International Working Party. HEPATOLOGY 1995;22:983-993.
23. Brazma A, Hingamp P, Quackenbush J, Sherlock G, Spellman P, Stoeck-
ert C, et al. Minimum information about a microarray experiment (MI-
AME)-toward standards for microarray data. Nat Genet 2001;29:365-
24. Wurmbach E, Yuen T, Sealfon SC. Focused microarray analysis. Methods
946 WURMBACH ET AL.HEPATOLOGY, April 2007
25. Albanis E, Friedman SL. Antifibrotic agents for liver disease. Am J Trans-
26. Bieche I, Asselah T, Laurendeau I, Vidaud D, Degot C, Paradis V, et al.
Molecular profiling of early stage liver fibrosis in patients with chronic
hepatitis C virus infection. Virology 2005;332:130-144.
27. Patil MA, Chua MS, Pan KH, Lin R, Lih CJ, Cheung ST, et al. An
integrated data analysis approach to characterize genes highly expressed in
hepatocellular carcinoma. Oncogene 2005;24:3737-3747.
28. Ge X, Yamamoto S, Tsutsumi S, Midorikawa Y, Ihara S, Wang SM, et al.
Interpreting expression profiles of cancers by genome-wide survey of
breadth of expression in normal tissues. Genomics 2005;86:127-141.
genomics based identification of potential human hepatocarcinogenesis-
Res Commun 2005;330:489-497.
31. Bruix J, Sherman M, Llovet JM, Beaugrand M, Lencioni R, Burroughs
AK, et al. Clinical management of hepatocellular carcinoma. Conclusions
of the Barcelona-2000 EASL conference. European Association for the
Study of the Liver. J Hepatol 2001;35:421-430.
32. Paradis V, Bieche I, Dargere D, Laurendeau I, Laurent C, Bioulac Sage P,
et al. Molecular profiling of hepatocellular carcinomas (HCC) using a
large-scale real-time RT-PCR approach: determination of a molecular di-
agnostic index. Am J Pathol 2003;163:733-741.
et al. Amplifications of DNA primase 1 (PRIM1) in human osteosarcoma.
Genes Chromosomes Cancer 1999;26:62-69.
for a mitotic spindle protein. Hum Mol Genet 2005;14:2155-2165.
35. Hall CL, Yang B, Yang X, Zhang S, Turley M, Samuel S, et al. Overex-
pression of the hyaluronan receptor RHAMM is transforming and is also
required for H-ras transformation. Cell 1995;82:19-26.
new therapeutic targets. Semin Liver Dis 2005;25:212-225.
37. Laurent-Puig P, Legoix P, Bluteau O, Belghiti J, Franco D, Binot F, et al.
Genetic alterations associated with hepatocellular carcinomas define dis-
tinct pathways of hepatocarcinogenesis. Gastroenterology 2001;120:
38. Merle P, Kim M, Herrmann M, Gupte A, Lefrancois L, Califano S, et al.
Oncogenic role of the frizzled-7/beta-catenin pathway in hepatocellular
carcinoma. J Hepatol 2005;43:854-862.
39. Calvisi DF, Ladu S, Gorden A, Farina M, Conner EA, Lee JS, et al.
Ubiquitous activation of Ras and Jak/Stat pathways in human HCC. Gas-
al. JAGGED1 expression is associated with prostate cancer metastasis and
recurrence. Cancer Res 2004;64:6854-6857.
41. Lin W, Choe WH, Hiasa Y, Kamegaya Y, Blackard JT, Schmidt EV, et al.
Hepatitis C virus expression suppresses interferon signaling by degrading
STAT1. Gastroenterology 2005;128:1034-1041.
42. Fong CW, Chua MS, McKie AB, Ling SH, Mason V, Li R, et al.
Sprouty 2, an inhibitor of mitogen-activated protein kinase signaling,
is down-regulated in hepatocellular carcinoma. Cancer Res 2006;66:
43. Seidler DG, Goldoni S, Agnew C, Cardi C, Thakur ML, Owens RT, et
al. Decorin protein core inhibits in vivo cancer growth and metabolism
by hindering epidermal growth factor receptor function and triggering
apoptosis via caspase-3 activation. J Biol Chem 2006;281:26408-
44. Wirths O, Waha A, Weggen S, Schirmacher P, Kuhne T, Goodyer CG, et
signaling, in human hepatoblastomas and Wilms’ tumors. Lab Invest
45. Sicklick JK, Li YX, Jayaraman A, Kannangai R, Qi Y, Vivekanandan P, et
al. Dysregulation of the Hedgehog pathway in human hepatocarcinogen-
esis. Carcinogenesis 2006;27:748-757.
46. Lin SY, Elledge SJ. Multiple tumor suppressor pathways negatively regu-
late telomerase. Cell 2003;113:881-889.
47. Farazi PA, Glickman J, Jiang S, Yu A, Rudolph KL, DePinho RA. Differ-
ential impact of telomere dysfunction on initiation and progression of
hepatocellular carcinoma. Cancer Res 2003;63:5021-5027.
48. Kienitz A, Vogel C, Morales I, Muller R, Bastians H. Partial downregu-
lation of MAD1 causes spindle checkpoint inactivation and aneu-
ploidy, but does not confer resistance towards taxol. Oncogene 2005;
49. Gramantieri L, Chieco P, Giovannini C, Lacchini M, Trere D, Grazi GL,
et al. GADD45-alpha expression in cirrhosis and hepatocellular carci-
50. Zhao M, Kim YT, Yoon BS, Kim SW, Kang MH, Kim SH, et al. Expres-
sion profiling of cyclin B1 and D1 in cervical carcinoma. Exp Oncol
HEPATOLOGY, Vol. 45, No. 4, 2007 WURMBACH ET AL.947