EpCAM and A-Fetoprotein Expression Defines Novel ... · Ly74, M1S2, M4S1, MIC18, MK-1, TROP1, and hEGP-2; ref. 17) is highly expressed in many human cancers with an epithelial origin

EpCAM and A-Fetoprotein Expression Defines Novel Prognostic

Subtypes of Hepatocellular Carcinoma

Taro Yamashita,1Marshonna Forgues,

1Wei Wang,

1Jin Woo Kim,

1Qinghai Ye,

4Huliang Jia,

4

Anuradha Budhu,1Krista A. Zanetti,

1,3Yidong Chen,

2Lun-Xiu Qin,

4

Zhao-You Tang,4and Xin Wei Wang

1

1Liver Carcinogenesis Section, Laboratory of Human Carcinogenesis, 2Genetics Branch, Center for Cancer Research, and 3CancerPrevention Fellowship Program, Division of Cancer Prevention, National Cancer Institute, Bethesda, Maryland;and 4Liver Cancer Institute and Zhongshan Hospital, Fudan University, Shanghai, China

Abstract

The heterogeneous nature of hepatocellular carcinoma (HCC)and the lack of appropriate biomarkers have hamperedpatient prognosis and treatment stratification. Recently, wehave identified that a hepatic stem cell marker, epithelial celladhesion molecule (EpCAM), may serve as an early biomarkerof HCC because its expression is highly elevated in premalig-nant hepatic tissues and in a subset of HCC. In this study, weaimed to identify novel HCC subtypes that resemble certainstages of liver lineages by searching for EpCAM-coexpressedgenes. A unique signature of EpCAM-positive HCCs wasidentified by cDNA microarray analysis of 40 HCC cases andvalidated by oligonucleotide microarray analysis of 238independent HCC cases, which was further confirmed byimmunohistochemical analysis of an additional 101 HCCcases. EpCAM-positive HCC displayed a distinct molecularsignature with features of hepatic progenitor cells includingthe presence of known stem/progenitor markers such ascytokeratin 19, c-Kit, EpCAM, and activated Wnt-B-cateninsignaling, whereas EpCAM-negative HCC displayed genes withfeatures of mature hepatocytes. Moreover, EpCAM-positiveand EpCAM-negative HCC could be further subclassified intofour groups with prognostic implication by determining thelevel of A-fetoprotein (AFP). These four subtypes displayeddistinct gene expression patterns with features resemblingcertain stages of hepatic lineages. Taken together, weproposed an easy classification system defined by EpCAMand AFP to reveal HCC subtypes similar to hepatic cellmaturation lineages, which may enable prognostic stratifica-tion and assessment of HCC patients with adjuvant therapyand provide new insights into the potential cellular originof HCC and its activated molecular pathways. [Cancer Res2008;68(5):1451–61]

Introduction

Cancer is heterogeneous in terms of its biological behaviorand response to treatment (1, 2). Because heterogeneity can

compromise treatment options, classification of cancer accordingto its biological nature is required to provide the best qualifiedtherapy for each cancer patient. Biomarker-based cancer classifi-cation has become clinically relevant in breast cancer (humanepidermal growth factor receptor-2 expression and trastuzumab/Herceptin response) and lung cancer (epidermal growth factorreceptor mutation and gefitinib/Iressa response; refs. 3–5). Thus,identification of potential biomarkers and therapeutic targets incancer to provide personalized therapy is of great interest (6).

Hepatocellular carcinoma (HCC) is the third leading cause ofcancer death worldwide with a dismal outcome (7). Several stagingsystems are currently available for HCC classification, which arebased on tumor morphology, tumor number and size, vascularinvasion, metastasis status, serum a-fetoprotein (AFP) levels, andhepatic reserve markers (8). Although these systems can gradepatients based on prognosis, they are unable to predict pharma-cologic responses to chemotherapeutic agents, largely due to a lackof specific biomarkers and mechanistic targets. To overcome theselimitations, new technologies based on genome-wide screeninghave been applied to HCC classification (9–14). For example,primary HCC tissues with a propensity to metastasize or recur havea significantly different gene expression profile when comparedwith relapse-free HCC tissues (9, 10, 15). In addition, geneexpression profiling of rat fetal hepatoblasts and HCC hasidentified a novel subtype that may have features of hepaticprogenitor cells (HPC; ref. 13). These studies indicate that profilingthe transcriptional characteristics (transcriptome) of HCC canprovide insight into the cellular origin of a tumor and uncover HCCsubtypes.

Recently, we have identified TACSTD1 , which encodes a pan-carcinoma antigen epithelial cell adhesion molecule (EpCAM), tobe an early biomarker of HCC because its expression is highlyelevated in premalignant hepatic tissues and in a subset of HCC(16). EpCAM (also known as CO17-1A, EGP, EGP40, GA733-2, KSA,Ly74, M1S2, M4S1, MIC18, MK-1, TROP1, and hEGP-2; ref. 17) ishighly expressed in many human cancers with an epithelial origin(18). The function of EpCAM and the regulation of its expressionare largely unknown (19). In the adult liver, hepatocytes arenegative and bile duct epithelium is positive for EpCAM expression.However, in the embryonic liver, the majority of hepatocytesexpress EpCAM (20). In the cirrhotic liver, EpCAM is expressed inproliferating bile ductules that are thought to be derived from HPC(20). Recent studies indicate that both hepatic stem cells and fetalhepatoblast cells (a HPC lineage evolved from hepatic stem cells)express EpCAM (21). Of note, f35% of HCC cases express EpCAM(16, 20, 22). Likewise, AFP, a known HCC prognostic factor, isexpressed in embryonic liver and is silent in adult liver (23), butf60% of HCC patients have elevated serum AFP (24). A subclass of

Note: Supplementary data for this article are available at Cancer Research Online(http://cancerres.aacrjournals.org/).

Current address for W. Wang: Center for Biologics Evaluation and Research, Foodand Drug Administration, Bethesda, MD 20892. Current address for J.W. Kim: Centerfor Human Genomics, Wake Forest University, Winston-Salem, NC 27157.

Requests for reprints: Xin Wei Wang, Laboratory of Human Carcinogenesis,Center for Cancer Research, National Cancer Institute, 37 Convent Drive, Room 3044A,MSC 4258, Bethesda, MD 20892-4258. Phone: 301-496-2099; Fax: 301-496-0497; E-mail:[email protected].

I2008 American Association for Cancer Research.doi:10.1158/0008-5472.CAN-07-6013

www.aacrjournals.org 1451 Cancer Res 2008; 68: (5). March 1, 2008

Research Article

Research. on March 22, 2020. © 2008 American Association for Cancercancerres.aacrjournals.org Downloaded from

http://cancerres.aacrjournals.org/

AFP-positive (AFP+) HCCs have a unique gene expression signatureand a poor survival outcome (10, 13). Taken together, these findingssuggest that the status of EpCAM and AFP in HCC may reflectunique HCC subtypes and thus could serve as unique biomarkersto stratify HCC patients.

In this study, we determined whether an EpCAM-coexpressedsignature in HCC could be used to reveal HCC subtypes. Using geneexpression data from three independent HCC cohorts analyzed bycDNA microarray, oligonucleotide microarray, and immunohisto-chemistry, we successfully classified HCCs into EpCAM+ andEpCAM� groups. We revealed that EpCAM+ HCC had a uniquegene expression signature. Functional network analyses indicatedthat genes associated with protein synthesis and cellular develop-ment, including Wnt/h-catenin signaling, were activated, whereasmature hepatocyte–specific genes were inactivated in EpCAM+

HCCs. Immunohistochemistry analysis confirmed that EpCAM+

HCCs had an elevated expression of HPC markers [e.g., c-Kit,cytokeratin 19 (CK19), and EpCAM] and an activation of h-catenin,which suggests that EpCAM+ HCCs are related to a HPC-likephenotype with an activated Wnt-h-catenin signaling pathway.Noticeably, EpCAM+ and EpCAM� HCC could be furthersubclassified into four subtypes based on patients’ serum level ofAFP. Each of these four subtypes had a unique expression patternwith features resembling various stages of hepatic lineages, such ashepatic stem cell–like HCC, bile duct epithelium–like HCC,hepatocytic progenitor–like HCC, and mature hepatocyte–likeHCC. Importantly, hepatic stem cell–like and hepatocytic progen-itor–like HCC subtypes had a poor prognosis. We suggest that sucha convenient classification system based on the expression ofEpCAM and AFP can enable subgrouping of HCCs that are linkedto HCC prognosis and unique molecular features. This system mayallow for accurate assessment of outcome and the development ofpersonalized molecular targeted therapy in HCC patients.

Materials and Methods

Clinical specimens and RNA extraction. HCC samples were obtained

with informed consent from hepatitis B virus (HBV)–positive Chinesepatients who underwent radical resection at the Liver Cancer Institute and

Zhongshan Hospital (Fudan University, Shanghai, China) and the study was

approved by the Institutional Review Board of the Liver Cancer Institute.The 40 HCC cases (cohort 1) used in the cDNA microarray analysis were

previously described (9). In addition, we used our in-house oligonucleotide

microarray data of 238 independent HBV-positive HCC cases (cohort 2) for

validation of the EpCAM-coexpressed gene signature. Total RNA wasextracted using TRIzol (Invitrogen) according to the manufacturer’s

instructions. A total of 101 formalin-fixed and paraffin-embedded HCC

samples (cohort 3) were used for immunohistochemical analysis, 7 of which

were a part of cohort 1. Among the remaining 94 samples, 69 were also usedin a recent study (12). All HCC samples were histologically confirmed by two

independent pathologists and no fibrolamellar HCC was included.

Quantitative reverse transcription-PCR. TACSTD1, BAMBI, DKK1,UGT2B7 , and APOC1 expression were measured in quadruplicate in 32HCC and corresponding adjacent nontumor liver samples, for which a

sufficient amount and quality of RNA was available using the Applied

Biosystems 7700 Sequence Detection System (Applied Biosystems). Probesused for the analyses were TACSTD1 , Hs00158980_m1; BAMBI ,

Hs00180818_m1; DKK1 , Hs00183740_m1; UGT2B7 , Hs00426592_m1; APOC1 ,

Hs00155790_m1; and 18S , Hs99999901_s1 (Applied Biosystems). All

procedures were done according to the manufacturer’s instructions.Immunohistochemical analysis. Immunohistochemistry was done

with Envision+ kits (DAKO) according to the manufacturer’s instructions.

The primary antibodies were anti-p53 monoclonal antibody clone DO-7

(DAKO), anti–h-catenin monoclonal antibody clone 14 (BD Transduction

Laboratories), anti-CK19 monoclonal antibody clone RCK108 (DAKO),anti–cyclooxygenase 2 (Cox-2) monoclonal antibody (Cayman Chemical),

anti–c-Kit polyclonal antibodies (DAKO), and anti-EpCAM monoclonal

antibody clone VU-1D9 (Oncogene Research Products). The staining area

and intensities were evaluated in each sample and graded from 0 to 3(0, 0–5%; 1, 5–25%; 2, 25–50%; 3, >50%) and 0 to 2 (0, negative; 1, weak; 2,

strong), respectively. The sum of the area and intensity scores of each

marker (immunohistochemistry score) were calculated and used for

principal component analysis. Samples with >5% positive staining in agiven area for a particular antibody were considered as positive cases (most

of EpCAM+ cases had >25% positive cells). For indirect immunofluorescence

assay and confocal analysis, the primary antibodies anti-CK19 monoclonal

antibody clone RCK108, anti-EpCAM monoclonal antibody clone VU-1D9,and anti-AFP rabbit polyclonal antibody (DAKO) and secondary antibodies

Alexa 488 FITC-conjugated antimouse/antirabbit IgG or Alexa 568 Texas

red–conjugated antimouse/antirabbit IgG (Molecular Probes) were used.Confocal fluorescence microscopic analysis was done essentially as

previously described (25).

Microarray studies and statistical analysis. The cDNA microarray

(9) and the oligonucleotide microarray data sets are publicly available at theNational Center for Biotechnology Information Gene Expression Omnibus

database (GEO accession nos. GSE364 and GSE5975, respectively). The BRB-

ArrayTools software (version 3.3) was used for class comparison and

prediction analyses, as previously described (9, 12). Hierarchical clusteringanalysis was done with the GENESIS software (version 1.6) developed by

Alexander Sturn (IBMT-TUG). Multidimensional scaling analysis was done

with the Partek Genomics Suite software. The interaction networks ofEpCAM coexpressed genes were generated using Ingenuity Pathways

Analysis (version 3.1, Ingenuity Systems). The significance of gene

enrichment with a particular biologically relevant function was determined

by a one-sided Fisher’s exact test.The association of EpCAM expression and clinicopathologic character-

istics was examined with either Mann-Whitney U tests or m2 tests. The

comparison of gene expression data between two groups was examined by

Mann-Whitney U tests or among four groups by the Kruskal-Wallis test.Correlation of gene expression data was examined by the Spearman

correlation coefficient. The above analyses were done with GraphPad Prism

software 4.0 (GraphPad Software). Univariate and multivariate logisticregression analyses were done using STATA software 9.0 (STATACorp LP).

In both analyses, EpCAM expression was the independent binary variable

(EpCAM+ and EpCAM�), whereas expression of CK19, c-Kit, nuclear

h-catenin, Cox-2, and p53 were covariates in the model. The covariates werecategorized into two groups, 0 and 1, based on whether the sum of the area

and intensity scores of the marker for immunohistochemistry equaled 0 or

1 to 5, respectively. The Kaplan-Meier survival analysis with m2 test was done

to compare patient survival using Excel-based WinSTAT software 2001.1.

Results

Gene signature of EpCAM+ HCC. To identify EpCAM-coex-pressed genes in EpCAM+ HCC, we first analyzed a cDNA microarraydata set of 40 primary HCC tissues (cohort 1; clinicopathologiccharacteristics of the patients are available in SupplementaryTable S1; ref. 9). Fourteen (35%) HCC cases were considered EpCAM+

because they had z2-fold increase in the level of TACSTD1 in tumortissues compared with nontumor tissues. EpCAM-positive expres-sion was confirmed both by quantitative reverse transcription-PCR(RT-PCR) of 31 HCC samples with available RNA and byimmunohistochemistry of 16 HCC samples with available forma-lin-fixed tissues. Using the given criteria for defining positive andnegative cases, 26 of 31 (84%) cases from quantitative RT-PCR-arrayor 14 of 16 (88%) cases from immunohistochemistry arraycomparison were concordant. Further analysis revealed a significantpositive correlation between quantitative RT-PCR or immunohisto-chemistry and cDNA microarray results (r = 0.75, P < 0.0001 orr = 0.73, P = 0.0012, respectively; Supplementary Fig. S1).

Cancer Research

Cancer Res 2008; 68: (5). March 1, 2008 1452 www.aacrjournals.org



To search for differentially expressed genes between EpCAM+

and EpCAM� HCCs, we applied a class comparison analysis withunivariate t tests and a global permutation test of the class labels(�1,000) using a supervised strategy with BRB-ArrayTools (9, 12,

16). A comparison of EpCAM+ and EpCAM� HCCs yielded a totalof 71 differentially expressed genes (P < 0.005) with a z2-folddifference between the two classes (Fig. 1A). To further validateEpCAM-coexpressed genes, we tested the 71 significant genes in an

Figure 1. Gene signatures of EpCAM+ HCC. A, hierarchical cluster analysis of HCC. A dendrogram of the two-way hierarchical cluster analysis of gene expressiondata from 40 HCC cases (cohort 1) using TACSTD1 and 70 EpCAM-coregulated genes (P < 0.005, 2-fold cutoff, 1,000 univariate permutation tests) is shown.Each cell in the matrix represents the expression level of a gene in an individual sample. Red and green cells depict high and low expression levels, respectively, asindicated by the scale bar. B, hierarchical cluster analysis of HCC. An independent data set consisting of 238 HCC cases was analyzed using EpCAM-coregulatedgenes. EpCAM+ and EpCAM� HCC cases are clearly separated into two major branches. C, class prediction analysis of HCC. Six different class prediction algorithmswere used to classify 238 HCC cases using EpCAM-coregulated genes with 10-fold cross-validation. All classifier models predicted the EpCAM+ HCC class with82% to 97% accuracy with statistical significance (P < 0.001). CCP, compound covariate predictor; LDA, diagonal linear discriminant analysis; 1NN, 1-nearestneighbor; 3NN, 3-nearest neighbors; NC, nearest centroid; SVM, support vector machines.

Prognostic Subtypes of Hepatocellular Carcinoma




independent oligonucleotide microarray data set consisting of 238additional HCC cases (cohort 2; Supplementary Table S2). Amongthe 71 significant genes, 59 overlapped in this array platform andwere thus selected for further analysis. Hierarchical cluster analysisof these 59 genes resulted in a separation of two major subtypesamong the 238 HCC cases, one containing mainly EpCAM+ HCCsand the other containing mostly EpCAM� HCCs (Fig. 1B). Tofurther test if these 59 genes could be used as a signature to predictHCC subtypes based on EpCAM expression, we applied six differentmultivariate class prediction algorithms with 10-fold cross-validation and 1,000 random permutations (Fig. 1C). Theseanalyses resulted in a statistically significant prediction of EpCAM+

and EpCAM� HCC cases with accuracy ranging from 82% to 97%

(P < 0.001). Thus, we concluded that most of the 70 genes identifiedby microarray were associated with EpCAM expression.Functional networks activated in EpCAM+ HCC. To explore

the functional relation of EpCAM coexpressed genes, we carriedout an interaction network analysis. Among the 52 up-regulatedgenes, 35 mapped to seven relevant interaction networks withstatistical significance (P < 0.01; Supplementary Fig. S2). The topfive most statistically significant functions of the identifiedinteraction networks of up-regulated or down-regulated genes areincluded in Supplementary Table S3. We then removed redundantnodes that were not directly connected with EpCAM-coexpressedgenes. The final set of 22 EpCAM coexpressed genes (red labels)seemed to be functionally linked to the signaling networks of nine

Figure 2. Interaction network and functional analysis of EpCAM coregulated genes. Twenty-two up-regulated genes (red nodes ; A) and seven down-regulated genes(green nodes ; B ) in EpCAM+ HCC were clustered and are shown with their potential regulators (white nodes ; A and B ). Abbreviations of Edge Labels, Nodes, andEdges used in the analysis are indicated in Supplementary Fig. S2. C, quantitative RT-PCR analysis of representative genes involved in the functional networks(BAMBI, DKK1, UGT2B7 , and APOC1 ). Relative fold increases or decreases in HCCs compared with adjacent nontumor liver tissues are indicated in 32 HCC cases(cohort 1) with statistical significance (Mann-Whitney U test). D, scatter plot analysis of quantitative RT-PCR data. Gene expression levels of BAMBI and DKK1were positively correlated with those of TACSTD1 , whereas gene expression levels of UGT2B7 were negatively correlated with those of TACSTD1 in HCC withstatistical significance (Spearman’s correlation coefficient).

Cancer Research




genes (CALD1, CTNNB1, EGF, Ins1, MYCN, MYOD1, PPARA, TGFB1 ,and TP53 ; Fig. 2A). Similarly, among the 18 down-regulated genes,7 (green labels) seemed to be functionally linked to the signalingnetworks of five genes (NFE2L2, UGT1A6, TCF1, HNF4A , and PLG ;Fig. 2B). It seemed that genes associated with organ developmentand protein synthesis were selectively activated in EpCAM+ HCC,possibly through interactions of h-catenin (CTNNB1), transforminggrowth factor h1, insulin, epidermal growth factor, and p53

signaling. In contrast, a decrease of mature hepatocyte functionssuch as lipid metabolism and drug metabolism occurred inEpCAM+ HCC, possibly through inactivation of hepatic nuclearfactors and the Nrf2 signaling cascade.

To further validate the gene signatures associated with EpCAMexpression, we carried out quantitative RT-PCR of severalrepresentative genes in the pathway. We selected the up-regulatedBAMBI and DKK1 (representing classic Wnt-h-catenin signaling

Figure 3. Immunohistochemical analyses of EpCAM+ and EpCAM�

HCC cases. A, representative photomicrographs of EpCAM+ HCC andEpCAM� HCC stained with anti-EpCAM, anti–h-catenin, anti-CK19, anti–c-Kit,anti-p53, and anti–Cox-2 antibodies (bar, 50 Am). B, summary ofimmunohistochemical analysis of EpCAM+ and EpCAM� HCC. The frequencyof expression of CK19, c-Kit, and nuclear h-catenin was significantly higherin EpCAM+ HCC when compared with EpCAM� HCC. C, principalcomponent analysis of immunohistochemistry expression data. Theimmunohistochemistry score of each marker was calculated as describedin Materials and Methods. A clear separation between EpCAM+ and EpCAM�

HCC is visualized.





targets refs. 26, 27) and the down-regulated UGT2B7 and APOC1genes (markers of mature hepatocytes). Consistent with the arraydata, the expression levels of BAMBI and DKK1 were significantlyincreased and the expression levels of UGT2B7 and APOC1were significantly decreased in EpCAM+ HCC (Fig. 2C). Therewas a positive correlation between BAMBI and TACSTD1 (r = 0.73,P < 0.0001) or DKK1 and TACSTD1 (r = 0.72, P < 0.0001) and aninverse correlation between UGT2B7 and TACSTD1 (r = �0.61,P = 0.0002) or APOC1 and TACSTD1 (r = �0.34, P = 0.055;Fig. 2D).

Validation of functional networks activated in EpCAM+

HCC by immunohistochemical analysis. The above datasuggest that EpCAM+ HCC may potentially reflect an HPCorigin, which includes an elevated expression of HPC markerssuch as activated Wnt-h-catenin signaling, a key player duringembryogenesis. To further validate our findings, we carried outimmunohistochemical analysis on an additional 101 paraffin-embedded HCC tissues (cohort 3; Supplementary Table S4). Wealso selected h-catenin and p53 because they were identified astwo molecular nodes in EpCAM+ HCC by interaction network

Figure 4. Molecular features of HCCsubtypes defined by EpCAM and AFPexpression. A, immunofluorescenceof a cirrhotic liver section stained withanti-EpCAM (green ), anti-AFP (red), and4¶,6-diamidino-2-phenylindole (DAPI ; blue ).EpCAM/AFP double-positive cells were onlydetected in periportal area (left ). Fourdistinct cell types were detected based onthe expression of EpCAM and AFP (right ;1–4 ). Dotted lines, cirrhotic nodules. B,multidimensional scaling analysis ofmicroarray data of 40 HCC samples(cohort 1) and 238 HCC samples (cohort 2)based on the expression of 71EpCAM-coregulated genes. Axes, first threeprincipal components of these genes.Orange, type A HCC; yellow, type B HCC;light blue, type C HCC; blue, type D HCC.C, expression patterns of EpCAMcoregulated genes and well-known HPCmarkers in each HCC subtype. Geneexpression levels of KRT19, KRT7 , andTACSTD1 (EpCAM; stem/epithelialmarker), AFP and DLK1, PROM1 (CD133;stem/hepatoblast marker), BAMBI andDKK1 (Wnt/h-catenin signal marker), andCYP3A4 (mature hepatocyte marker) weremeasured in 238 HCC cases analyzed byoligonucleotide microarray. Each HCCsubtype represents a distinct molecularportrait of EpCAM coregulated genes andHPC marker expression.

Cancer Research




analysis (Fig. 2A), CK19 and c-Kit as markers for HPC, and Cox-2as a control.

Representative staining of each marker on serial sections ofEpCAM+ and EpCAM� HCC is shown in Fig. 3A . The typicalexpression pattern of each marker in adjacent nontumor livertissues is depicted in Supplementary Fig. S3. In EpCAM+ HCC,there was strong positive cytoplasmic and nuclear immunoreac-tivity to h-catenin, accompanied by immunoreactivity to CK19 andc-Kit (Fig. 3A). In contrast, in EpCAM� HCC, there was only weakimmunoreactivity to h-catenin in the cell membrane, and neitherCK19 nor c-Kit expression was detected. The frequency ofexpression of CK19, c-Kit, and nuclear h-catenin was significantlyhigher in EpCAM+ HCC when compared with EpCAM� HCC(P = 0.0007, P < 0.0001, and P = 0.0007, respectively; Fig. 3B). Nosignificant difference in the frequency of p53 and Cox-2 expressionwas observed between EpCAM+ and EpCAM� HCC. Most of theEpCAM+ cases had >25% positively stained cells whereas most ofthe CK19- or c-Kit-positive cases had f5% to 10% positivelystained cells. To explore the association of EpCAM expression withthe expression of other markers in HCC, we carried out a logisticregression analysis using EpCAM expression as the independentbinary variable (Supplementary Table S5). Both univariate andmultivariate analyses revealed that CK19, c-Kit, and h-cateninexpression were strongly associated with an increased odds ofhaving EpCAM expression in HCC tissues, whereas p53 and Cox-2expression were not associated with EpCAM expression. Multi-dimensional scaling based on EpCAM, h-catenin, CK19, and c-Kitexpression revealed a clear separation between EpCAM+ andEpCAM� HCC (Fig. 3C), indicating a robust and simple HCCsubtype classification system based on EpCAM expression andassociated HPC markers.HCC subtypes defined by EpCAM and AFP and human liver

cell lineages. The studies above indicate that EpCAM+ andEpCAM� HCCs are distinct tumor types with a unique activationof certain molecular signaling pathways and stem cell markers.These results suggest that difference of EpCAM expression inHCC may reflect the difference of cellular origins in liver lineages.However, when we examined the expression of EpCAM inadjacent noncancerous liver tissues, it was noted that EpCAMwas abundantly expressed in both small bile ductules and acluster of hepatocytes near the portal triad where suspectedhepatic stem cell niches were located (Supplementary Fig. S3). Itis possible that EpCAM expression may be confined to earlystages of hepatic and/or biliary lineages. Recent findings bySchmelzer et al. (21, 28) indicate that another known HPCmarker, AFP, is differentially expressed between hepatic stem cellsand hepatoblasts. Consistently, in the cirrhotic liver tissues,whereas some hepatocytes were EpCAM+ and AFP+, otherhepatocytes were only EpCAM+ or AFP+ (Fig. 4A, left). ManyEpCAM+ AFP� cells had a morphology resembling ductular cells(image 1) whereas EpCAM+ AFP+ cells resembled small hepato-cyte-like cells (image 2). Although EpCAM� AFP+ cells (image 3)and EpCAM� AFP� cells (image 4) bore a resemblance to maturehepatocytes, EpCAM� AFP+ cells were always detected in theperiportal area. These results are consistent with the findings bySchmelzer et al. (21, 28), suggesting that cells with differingEpCAM and AFP status reflect particular human liver celllineages. Thus, we hypothesized that additional HCC subtypesmay exist within EpCAM+ HCCs that resemble small bile ductulesor HPC/stem cells and that EpCAM+ or EpCAM� HCC can befurther subgrouped based on the expression of AFP.

We carried out multidimensional scaling analysis based on theexpression of EpCAM-coexpressed genes in both HCC cohort 1and cohort 2 (Fig. 4B) stratified by EpCAM and AFP status, whichresulted in four subgroups: A, B, C, and D. When a well-definedserum AFP cutoff (10, 13, 29) was added to the stratification,EpCAM+ (types A and B) and EpCAM� (types C and D) HCC werelargely separated by EpCAM-coexpressed genes in three dimen-sional space, consistent with the results of hierarchical clusteringusing the same gene set (Fig. 1A and B). Noticeably, the AFP-high(type B and C) and AFP-low (type A and D) HCC subtypes werealso separated in both cohorts, indicating that each HCC subtypehad a distinct expression pattern. Furthermore, we used sixdifferent multivariate Class Prediction algorithms with 10-foldcross-validation and 1,000 random permutations described aboveto compare type A with type B, as well as type C with type D.These analyses showed that type A was significantly differentfrom type B and type C was significantly different from typeD, based on the expression of EpCAM-coexpressed genes(Supplementary Table S6).

Next, we closely examined the expression pattern of well-known HPC and mature hepatocyte markers in each HCC subtypeusing cohort 2 (Fig. 4C ; refs. 21, 30). Stem/biliary epithelialmarkers KRT19 (CK19) and KRT7 (CK7) were more abundantlyexpressed in EpCAM+ AFP� HCC, whereas stem/hepatoblastmarkers DLK1 and PROM1 (CD133) as well as Wnt/h-cateninsignaling target genes DKK1 and BAMBI were more abundantlyexpressed in EpCAM+ AFP+ HCC. In contrast, the classic maturehepatocyte marker CYP3A4 was weakly expressed in EpCAM+

AFP+ HCC and mostly expressed in EpCAM� AFP� HCC. Takentogether, it seemed that there are distinct differences in thetranscriptional characteristics of these four HCC subtypesstratified by EpCAM and AFP status, which may correlate withcertain stages of human liver lineages.Prognosis of HCC subtypes defined by EpCAM and AFP. To

investigate the clinical outcomes of each HCC subtypes identifiedabove, we carried out Kaplan-Meier survival analysis of HCC casesin cohort 1 (n = 40), cohort 2 (n = 238), and cohort 3 (n = 94; sevencases used for immunohistochemistry were excluded becausethey were included in cohort 1). When Kaplan-Meier analysiswas done according to these classifications, EpCAM+ AFP+ (type B)and EpCAM� AFP+ (type C) HCC correlated with poor prognosis,whereas EpCAM� AFP� (type D) HCC correlated with anintermediate prognosis (Fig. 5A). Interestingly, EpCAM+ AFP�

(type A) HCC correlated with a good prognosis, and these resultswere validated in all three independent cohorts with statistical orborderline significance (P = 0.06, P = 0.05, and P = 0.0004 in cohorts1, 2, and 3, respectively).

Next, we investigated the clinical characteristics of each HCCsubtype in cohort 2 and cohort 3 (Table 1). Serum AFP values, onsetages, frequencies of portal vein invasion, and tumor-node-metastasis (TNM) stages were different among four HCC subtypeswith statistical significance (P < 0.05). EpCAM+ AFP+ HCCdeveloped in young patients with advanced TNM stages and portalvein invasion, whereas EpCAM� AFP� HCC developed in moreelder patients with early TNM stages. EpCAM+ AFP� HCC alsodeveloped in young patients but was associated with early TNMstages and low frequencies of portal vein invasion. These datasuggested that HCC subtypes similar to certain maturation lineagesrepresent distinct prognosis and clinical stages, which maycorrelate with various cell proliferative and invasive tumor cellsproperties in each HCC subtype.





Discussion

Prognostic variability and resistance against therapeutic agentsin cancer are likely governed by the heterogeneity of activated geneexpression patterns and molecular pathways, which may be relatedto differences in cellular origin (2, 31). The concept that some HCCmay originate from a HPC has recently emerged (32, 33). The liver,an organ capable of self-regenerating, contains several cell typeswith longevity (30). Experimental evidence indicates that hepaticstem cells, rich in fetal and neonatal livers, can give rise tohepatoblasts, which further differentiate toward either the biliarylineage or hepatocytic progenitor cells and then mature hepato-cytes (21, 30, 34). Recent studies indicate that whereas certainHPCs express EpCAM and hepatoblasts express EpCAM and AFP,mature hepatocytes are negative for both markers (21, 28).Consistently, human fetal liver multipotent progenitor cells andrat hepatic progenitor/oval cells express EpCAM but not AFP(35, 36). It should be noted that biliary epithelial cells or malignantcells of biliary lineage are EpCAM+ but AFP�. In this study, werevealed that HCC can be stratified by EpCAM and AFP status intofour subtypes that may reflect different hepatic lineages and

prognosis. Pathway analysis indicates that the underlying molec-ular biological activities differ significantly between the two poorprognostic HCC subtypes (i.e., EpCAM+ AFP+ and EpCAM� AFP+).Whereas activation of Wnt-h-catenin signaling is mainly associatedwith EpCAM+ AFP+ HCC, mature hepatocyte–specific genes arehighly expressed in EpCAM� AFP+ HCC. These findings suggestthat these two types of aggressive HCC are distinct and thusdifferent therapeutic interventions should be considered. It shouldbe emphasized that such an easy classification has beenreproduced in three independent cohorts that have been analyzedby different technical platforms. Importantly, this classification canbe reproduced in another cohort with a mixed etiologicbackground including hepatitis C cases, indicating a commonapplicability of this approach.5 It should be noted that theclassification system is only applicable at the present time forsurgically eligible HCC patients because only f20% of HCCpatients are currently qualified for resection. It would be

Figure 5. A, prognostic outcomes of HCC subtypes. Kaplan-Meier survival analyses of four different types of HCCs stratified by EpCAM and AFP status in cohort1 analyzed by cDNA microarray (left ), in cohort 2 analyzed by oligo microarray (middle ), and in cohort 3 analyzed by immunohistochemistry (right ). AFP+ andAFP� HCCs were defined as those with >300 or V300 ng/mL of serum AFP, respectively. The statistical P value was generated by the Cox-Mantel log-rank test. Thesurvival curve of each HCC subtype is indicated as red (EpCAM+AFP�), blue (EpCAM+AFP+), green (EpCAM�AFP+), or black (EpCAM�AFP�). B, a modeldepicting the parallels between proposed HCC subtypes and normal human liver cell lineage. HCC subtypes and liver lineages are characterized by expression ofEpCAM and AFP. Red and blue letters show positive and negative expression, respectively. Different clinical outcomes associated with each HCC subtype areindicated. h-Catenin signaling is activated in HpSC-HCC, implying the utility of Wnt/h-catenin signaling inhibitors for eradication of this HCC subtype.

5 Snorri Thorgeirsson, personal communication.

Cancer Research




interesting to determine whether serum EpCAM and AFP can beused as a prognostic classifier. In addition, a majority of the casesanalyzed in this study are still HBV+. A further validation study isclearly required to reveal the utility of this classification system inHCC with various etiologic backgrounds. Moreover, it is interestingto determine in the future study whether EpCAM can serve as abiomarker to detect early lesions including dysplastic nodules.

Based on the unique gene expression patterns, which may berelated to different liver cell lineages (30), we propose the followingnomenclatures for the four subtypes described above: EpCAM+

AFP+ HCC as hepatic stem cell–like HCC (HpSC-HCC), EpCAM+ AFP�

HCC as bile duct epithelium–like HCC (BDE-HCC), EpCAM� AFP+

HCC as hepatocytic progenitor–like HCC (HP-HCC), and EpCAM�

AFP� HCC as mature hepatocyte–like HCC (MH-HCC) (Fig. 5B).We have shown that several known HPC markers such as CK19 andc-Kit are more abundantly expressed in HpSC-HCC whereas maturehepatocyte–specific genes such as CYP3A4 are more abundantly

expressed in mature hepatocyte–like HCC. Moreover, HpSC-HCC haselevated Wnt-h-catenin signaling, implying that activation of thispathway may be a critical event in the molecular pathogenesis of thisHCC subtype. Interestingly, bile duct epithelium–like HCC correlatedwith a good prognosis and the molecular details should be clarified infuture studies. Taken together, we suggest that these four HCCsubtypes, classified by EpCAM and AFP, represent different molecularportraits of HCC, which may reflect different prognosis and specificactivated pathways depending on tumor cell origin.

Several studies have discussed HCC with putative HPC origins(33, 37). In particular, a novel poor prognostic HCC subtype thatshares gene expression patterns with fetal hepatoblasts wasrecently identified (13). CK19, CK7, and vimentin were moreabundantly expressed in the poorly prognostic hepatoblastomasubtype than in the hepatocyte subtype of HCC, which maycorrespond to our HpSC-HCC and MH-HCC, respectively. Althoughconceptually similar, several differences clearly exist between our

Table 1. Clinicopathologic characteristics of HCC subtypes defined by EpCAM and AFP

HCC subtypes EpCAM+ AFP� [A] EpCAM+ AFP+ [B] EpCAM� AFP+ [C] EpCAM� AFP� [D] P*

Cohort 2 (n = 238) n = 35 n = 60 n = 47 n = 96

Age (y, mean F SD) 48.0 F 8.6 46.0 F 10.7 53.1 F 10.9 52.9 F 10.5 0.0005

Sex (male/female) 30/5 50/10 40/7 87/9 0.56Liver cirrhosis

c(yes/no/no data) 33/2/— 56/4/— 44/3/— 88/7/1 0.95

AFP [ng/mL; median (25–75%)] 24.1 (7.5–95.6) 1,706 (865–5,915) 1,635 (794–2,908) 11.8 (4.0–48.6) <0.0001

Histologic gradeb

1–2 15 14 15 41

2–3 18 44 30 48

3–4 1 2 2 5

No data 1 0 0 2 0.24Tumor size (cm, mean F SD) 3.9 F 2.4 5.1 F 3.0 4.4 F 2.3 4.4 F 3.0 0.19

Multinodular (yes/no) 6/29 16/44 16/31 15/81 0.06

Portal vein invasion (yes/no/no data) 0/35/— 11/49/— 7/39/1 9/87/— 0.05

TNM classificationI 17 24 19 46

II 15 22 15 42

III 3 14 13 8 0.044

Cohort 3 (n = 101) n = 17 n = 22 n = 22 n = 31

Age (y, mean F SD) 49.4 F 10.8 44.6 F 7.5 46.1 F 9.7 54.1 F 11.3 0.014

Sex (male/female) 16/0 19/3 20/2 29/2 0.47Liver cirrhosis

c(yes/no) 15/2 21/1 19/3 30/1 0.44

AFP [ng/mL; median (25–75%)] 50 (10.5–133) 1,476 (844–3,090) 1,423 (1009–2,122) 12 (5–75) <0.0001

Histologic gradeb

1–2 6 12 10 202–3 9 7 12 11

3–4 2 3 0 0 0.087

Tumor size (cm, mean F SD) 5.4 F 2.3 7.3 F 3.6 8.0 F 4.5 6.8 F 3.8 0.26

Multinodular (yes/no) 17/0 19/3 17/5 25/6 0.21Portal vein invasion (yes/no) 6/11 15/7 17/5 9/22 0.001

TNM classification

I 7 4 4 9II 6 7 1 10

III 4 11 17 12 0.025

*Statistical significance was calculated by m2 test or Kruskal-Wallis test.cChild-Pugh score was also used to evaluate the degree of cirrhosis and patient outcomes and no significant difference among four HCC subtypes could

be found (data not shown).bEdmondson-Steiner histologic grades.





study and that by Lee et al. For example, Lee et al. have shown thatactivation of activator protein 1 (AP-1) transcription factors mayplay a key role in the development of the hepatoblastoma subtype.In contrast, we have found that HpSC-HCC is associated withactivation of Wnt-h-catenin signaling and the Myc transcriptionfactor. These differences may be due to mutations in differentmolecular nodes caused by different etiologic factors acting withinthe same pathway. Importantly, however, a functional merge of theAP-1, Myc, and Wnt-h-catenin signaling molecular networks isclearly evident in many published studies and highlights theimportance of identifying a common molecular node to achievetargeted therapies. It is also possible that heterogeneity, a feature ofcancer stem cells (38), may exist in these types of tumors becauseboth hepatoblastoma subtype HCC and HpSC-HCC are consideredto be evolved from hepatic stem cells. Further detailed analyses ofthese HCC subtypes may provide molecular insight into livercancer stem cells.

Our gene expression profiling, network pathway analysis, andimmunohistochemical analysis reveal a close correlation betweenWnt-h-catenin signaling and EpCAM+ HCC. Wnt-h-catenin signal-ing plays a pivotal role in embryogenesis and stem cellmaintenance (39) and is known to be activated during liverdevelopment (40, 41). Importantly, Wnt-h-catenin signaling is alsosignificantly activated in HCC (42). Furthermore, h-cateninmutations have been identified in a majority of hepatoblastoma,the pediatric tumor believed to originate from hepatoblasts (43).Although the importance of this signaling pathway in hepatocarci-nogenesis is well established, a method to measure its activityremains controversial. Current assays measure nuclear or cyto-plasmic h-catenin or frizzled-7 receptor overexpression as markersof Wnt-h-catenin activation (42–44). However, different technical

platforms and measurement of different targets result in adifferential proportion of HCCs with activated Wnt-h-cateninsignaling. In addition, the prognostic value of Wnt-h-catenin signalactivation is contentious and divisible, which may reflect thetentative criteria of Wnt-h-catenin signaling activation (45–49).Our novel HCC classification model based on EpCAM and AFPexpression clearly stratified HCC subtypes with features of certainliver lineages, and Wnt-h-catenin signaling was more likelyactivated in HpSC-HCC with poor prognosis. Encouragingly, wehave recently shown that EpCAM is a direct transcriptional targetof the Wnt-h-catenin signaling pathway, further emphasizing thefunctional link between these two molecular nodes (50). Further-more, either silencing of EpCAM expression by RNA interference orspecific inhibition of h-catenin resulted in cell killing of EpCAM+

HCC cell lines (50). It is plausible that both EpCAM and h-cateninmay serve as unique molecular targets for HpSC-HCC. Furtherstudies using animal models would be needed to establish theutility of such stratification and targeted therapy approach.

Acknowledgments

Received 10/29/2007; revised 12/17/2007; accepted 12/27/2007.Grant support: Intramural Research Program of the U.S. National Cancer Institute,

China National Natural Science Foundation for Distinguished Young Scholars grant30325041 and Key Program Project grant 30430720, China National ‘‘863’’ R&D High-Tech Key Project grant 2002BA711A02-4, State Key Basic Research Program of Chinagrant G1998051210, and Key Program Project of the Shanghai Science and TechnologyCommittee grant 04JC14028 (H.L. Jia, Q.H. Ye, L-X. Qin, and Z-Y. Tang).

The costs of publication of this article were defrayed in part by the payment of pagecharges. This article must therefore be hereby marked advertisement in accordancewith 18 U.S.C. Section 1734 solely to indicate this fact.

We thank Curtis C. Harris and Snorri S. Thorgeirsson for useful comments, ZhihuaGeng for technical assistance, Jennifer Wang for artworks, and NIH Fellows EditorialBoard, Dorothea Dudek, and Karen MacPherson for editorial assistance.

Cancer Research


References

1. Vogelstein B, Kinzler KW. Cancer genes and thepathways they control. Nat Med 2004;10:789–99.2. Hanahan D, Weinberg RA. The hallmarks of cancer.

Cell 2000;100:57–70.3. Slamon DJ, Leyland-Jones B, Shak S, et al. Use of

chemotherapy plus a monoclonal antibody againstHER2 for metastatic breast cancer that overexpressesHER2. N Engl J Med 2001;344:783–92.4. Paez JG, Janne PA, Lee JC, et al. EGFR mutations in

lung cancer: correlation with clinical response togefitinib therapy. Science 2004;304:1497–500.5. Lynch TJ, Bell DW, Sordella R, et al. Activating

mutations in the epidermal growth factor receptorunderlying responsiveness of non-small-cell lung cancerto gefitinib. N Engl J Med 2004;350:2129–39.6. Ludwig JA, Weinstein JN. Biomarkers in cancer

staging, prognosis and treatment selection. Nat RevCancer 2005;5:845–56.7. Parkin DM, Bray F, Ferlay J, Pisani P. Global cancer

statistics, 2002. CA Cancer J Clin 2005;55:74–108.8. Sala M, Forner A, Varela M, Bruix J. Prognostic

prediction in patients with hepatocellular carcinoma.Semin Liver Dis 2005;25:171–80.9. Ye QH, Qin LX, Forgues M, et al. Predicting hepatitis B

virus-positive metastatic hepatocellular carcinomasusing gene expression profiling and supervised machinelearning. Nat Med 2003;9:416–23.10. Lee JS, Chu IS, Heo J, et al. Classification and

prediction of survival in hepatocellular carcinoma bygene expression profiling. Hepatology 2004;40:667–76.11. Ding SJ, Li Y, Tan YX, et al. From proteomic analysis

to clinical significance: overexpression of cytokeratin 19correlates with hepatocellular carcinoma metastasis.Mol Cell Proteomics 2004;3:73–81.

12. Budhu A, Forgues M, Ye QH, et al. Prediction ofvenous metastases, recurrence and prognosis in hepa-tocellular carcinoma based on a unique immuneresponse signature of the liver microenvironment.Cancer Cell 2006;10:99–111.13. Lee JS, Heo J, Libbrecht L, et al. A novel prognostic

subtype of human hepatocellular carcinoma derivedfrom hepatic progenitor cells. Nat Med 2006;12:410–6.14. Wurmbach E, Chen YB, Khitrov G, et al. Genome-

wide molecular profiles of HCV-induced dysplasia andhepatocellular carcinoma. Hepatology 2007;45:938–47.15. Iizuka N, Oka M, Yamada-Okabe H, et al. Oligonu-

cleotide microarray for prediction of early intrahepaticrecurrence of hepatocellular carcinoma after curativeresection. Lancet 2003;361:923–9.16. Kim JW, Ye Q, Forgues M, et al. Cancer-associated

molecular signature in the tissue samples of patientswith cirrhosis. Hepatology 2004;39:518–27.17. Herlyn M, Steplewski Z, Herlyn D, Koprowski H.

Colorectal carcinoma-specific antigen: detection bymeans of monoclonal antibodies. Proc Natl Acad SciU S A 1979;76:1438–42.18. Balzar M, Winter MJ, de Boer CJ, Litvinov SV. The

biology of the 17-1A antigen (Ep-CAM). J Mol Med 1999;77:699–712.19. Winter MJ, Nagelkerken B, Mertens AE, Rees-Bakker

HA, Briaire-de Bruijn IH, Litvinov SV. Expression ofEp-CAM shifts the state of cadherin-mediated adhesionsfrom strong to weak. Exp Cell Res 2003;285:50–8.20. de Boer CJ, van Krieken JH, Janssen-van Rhijn CM,

Litvinov SV. Expression of Ep-CAM in normal, regener-ating, metaplastic, and neoplastic liver. J Pathol 1999;188:201–6.21. Schmelzer E, Wauthier E, Reid LM. The phenotypes

of pluripotent human hepatic progenitors. Stem Cells2006;24:1852–8.

22. Ruck P, Wichert G, Handgretinger R, Kaiserling E.Ep-CAM in malignant liver tumours. J Pathol 2000;191:102–3.23. Takayasu K, Arii S, Ikai I et al. Prospective cohort

study of transarterial chemoembolization for unresect-able hepatocellular carcinoma in 8510 patients. Gastro-enterology 2006;131:461–9.24. Taketa K. a-Fetoprotein: reevaluation in hepatology.

Hepatology 1990;12:1420–32.25. Wang W, Budhu A, Forgues M, Wang XW. Temporal

and spatial control of nucleophosmin by the Ran-Crm1complex in centrosome duplication. Nat Cell Biol 2005;7:823–30.26. Chamorro MN, Schwartz DR, Vonica A, Brivanlou AH,

Cho KR, Varmus HE. FGF-20 and DKK1 are transcrip-tional targets of h-catenin and FGF-20 is implicated incancer and development. EMBO J 2005;24:73–84.27. Sekiya T, Adachi S, Kohu K et al. Identification of

BMP and activin membrane-bound inhibitor (BAMBI),an inhibitor of transforming growth factor-h signaling,as a target of the h-catenin pathway in colorectal tumorcells. J Biol Chem 2004;279:6840–6.28. Schmelzer E, Zhang L, Bruce A, et al. Human hepatic

stem cells from fetal and postnatal donors. J Exp Med2007;204:1973–87.29. Johnson Pj. The role of serum a-fetoprotein estima-

tion in the diagnosis and management of hepatocellularcarcinoma. Clin Liver Dis 2001;5:145–59.30. Fausto N. Liver regeneration and repair: hepatocytes,

progenitor cells, and stem cells. Hepatology 2004;39:1477–87.31. Alizadeh AA, Eisen MB, Davis RE, et al. Distinct types

of diffuse large B-cell lymphoma identified by geneexpression profiling. Nature 2000;403:503–11.32. Thorgeirsson SS, Grisham JW. Hepatic stem cells.

Semin Liver Dis 2003;23:301.





33. Libbrecht L, Desmet V, Van Damme B, Roskams T.The immunohistochemical phenotype of dysplastic fociin human liver: correlation with putative progenitorcells. J Hepatol 2000;33:76–84.34. Roskams T. Liver stem cells and their implication in

hepatocellular and cholangiocarcinoma. Oncogene 2006;25:3818–22.35. Dan YY, Riehle KJ, Lazaro C, et al. Isolation of

multipotent progenitor cells from human fetal livercapable of differentiating into liver and mesenchymallineages. Proc Natl Acad Sci U S A 2006;103:9912–7.36. Yovchev MI, Grozdanov PN, Joseph B, Gupta S,

Dabeva MD. Novel hepatic progenitor cell surface mar-kers in the adult rat liver. Hepatology 2007;45:139–49.37. Robrechts C, De Vos R, Van den HM, et al. Primary

liver tumour of intermediate (hepatocyte-bile ductcell) phenotype: a progenitor cell tumour? Liver 1998;18:288–93.38. Morrison SJ, Kimble J. Asymmetric and symmetric

stem-cell divisions in development and cancer. Nature2006;441:1068–74.39. Reya T, Clevers H. Wnt signalling in stem cells and

cancer. Nature 2005;434:843–50.

40. Micsenyi A, Tan X, Sneddon T, Luo JH, MichalopoulosGK, Monga SP. h-Catenin is temporally regulated duringnormal liver development. Gastroenterology 2004;126:1134–46.41. Monga SP, Monga HK, Tan X, Mule K, Pediaditakis P,

Michalopoulos GK. h-Catenin antisense studies inembryonic liver cultures: role in proliferation, apoptosis,and lineage specification. Gastroenterology 2003;124:202–16.42. Branda M, Wands JR. Signal transduction cascades

and hepatitis B and C related hepatocellular carcinoma.Hepatology 2006;43:891–902.43. Giles RH, van Es JH, Clevers H. Caught up in a Wnt

storm: Wnt signaling in cancer. Biochim Biophys Acta2003;1653:1–24.44. Merle P, de la MS, Kim M, et al. Functional

consequences of frizzled-7 receptor overexpression inhuman hepatocellular carcinoma. Gastroenterology2004;127:1110–22.45. Inagawa S, Itabashi M, Adachi S, et al. Expression and

prognostic roles of h-catenin in hepatocellular carcino-ma: correlation with tumor progression and postoper-ative survival. Clin Cancer Res 2002;8:450–6.

46. Mao TL, Chu JS, Jeng YM, Lai PL, Hsu HC. Expressionof mutant nuclear h-catenin correlates with non-invasive hepatocellular carcinoma, absence of portalvein spread, and good prognosis. J Pathol 2001;193:95–101.47. Hsu HC, Jeng YM, Mao TL, Chu JS, Lai PL, Peng SY.h-Catenin mutations are associated with a subset oflow-stage hepatocellular carcinoma negative for hepa-titis B virus and with favorable prognosis. Am J Pathol2000;157:763–70.48. Endo K, Ueda T, Ueyama J, Ohta T, Terada T. Im-

munoreactive E-cadherin, a-catenin, h-catenin, andg-catenin proteins in hepatocellular carcinoma: rela-tionships with tumor grade, clinicopathologic parame-ters, and patients’ survival. Hum Pathol 2000;31:558–65.49. Wong CM, Fan ST, Ng IO. h-Catenin mutation and

overexpression in hepatocellular carcinoma: clinico-pathologic and prognostic significance. Cancer 2001;92:136–45.50. Yamashita T, Budhu A, Forgues M, Wang XW.

Activation of hepatic stem cell marker EpCAM byWnt-h-catenin signaling in hepatocellular carcinoma.Cancer Res 2007;67:10831–9.



2008;68:1451-1461. Cancer Res Taro Yamashita, Marshonna Forgues, Wei Wang, et al. Prognostic Subtypes of Hepatocellular Carcinoma

-Fetoprotein Expression Defines NovelαEpCAM and

Updated version

http://cancerres.aacrjournals.org/content/68/5/1451

Access the most recent version of this article at:

Material

Supplementary

http://cancerres.aacrjournals.org/content/suppl/2008/03/03/68.5.1451.DC1

Access the most recent supplemental material at:

Cited articles

http://cancerres.aacrjournals.org/content/68/5/1451.full#ref-list-1

This article cites 50 articles, 9 of which you can access for free at:

Citing articles

http://cancerres.aacrjournals.org/content/68/5/1451.full#related-urls

This article has been cited by 39 HighWire-hosted articles. Access the articles at:

E-mail alerts related to this article or journal.Sign up to receive free email-alerts

Subscriptions

Reprints and

[email protected] at

To order reprints of this article or to subscribe to the journal, contact the AACR Publications

Permissions

Rightslink site. (CCC)Click on "Request Permissions" which will take you to the Copyright Clearance Center's

.http://cancerres.aacrjournals.org/content/68/5/1451To request permission to re-use all or part of this article, use this link



http://cancerres.aacrjournals.org/content/suppl/2008/03/03/68.5.1451.DC1

http://cancerres.aacrjournals.org/content/68/5/1451.full#ref-list-1

http://cancerres.aacrjournals.org/content/68/5/1451.full#related-urls

http://cancerres.aacrjournals.org/cgi/alerts

mailto:[email protected]



EpCAM and A-Fetoprotein Expression Defines Novel ... · Ly74, M1S2, M4S1, MIC18, MK-1, TROP1, and hEGP-2; ref. 17) is highly expressed in many human cancers with an epithelial origin

Documents