A Molecular Signature to Discriminate Dysplastic Nodules From Early Hepatocellular Carcinoma in HCV Cirrhosis

GASTROENTEROLOGY 2006;131:1758 –1767

A Molecular Signature to Discriminate Dysplastic Nodules From Early Hepatocellular Carcinoma in HCV Cirrhosis JOSEP M. LLOVET,*,‡ YINGBEI CHEN,* ELISA WURMBACH,* SASAN ROAYAIE,* M. ISABEL FIEL,* MYRON SCHWARTZ,* SWAN N. THUNG,* GREGORY KHITROV,* WEIJIA ZHANG,* AUGUSTO VILLANUEVA,* CARLO BATTISTON,§ VINCENZO MAZZAFERRO,§ JORDI BRUIX,‡ SAMUEL WAXMAN,* and SCOTT L. FRIEDMAN* *Mount Sinai Liver Cancer Program (Divisions of Liver Disease and Hematology/Oncology, Department of Medicine, Recanati Miller Transplantation Institute, Department of Surgery; Department of Pathology), Mount Sinai School of Medicine, New York; ‡BCLC Group, IDIBAPS, Liver Unit, Hospital Clinic, Barcelona, Spain; and §National Cancer Institute, Milan, Italy

CLINICAL–LIVER, PANCREAS, AND BILIARY TRACT

Background & Aims: Small liver nodules ⬃2 cm are difficult to characterize by radiologic or pathologic examination. Our aim was to identify a molecular signature to diagnose early hepatocellular carcinoma (HCC). Methods: The transcriptional profiles of 55 candidate genes were assessed by quantitative real-time reverse-transcription polymerase chain reaction (RT-PCR) in 17 dysplastic nodules (diameter, 10 mm) and 20 early HCC (diameter, 18 mm) from HCV cirrhotic patients undergoing resection/transplantation and 10 nontumoral cirrhotic tissues and 10 normal liver tissues. Candidate genes were confirmed by quantitative RT-PCR in 20 advanced HCCs and by immunohistochemistry in 75 samples and validated in an independent set of 29 samples (dysplastic nodules [10] and small HCC [19; diameter, 20 mm]). Results: Twelve genes were significantly, differentially expressed in early HCCs compared with dysplastic nodules (⬎2-fold change; area under the receiver operating characteristic curve ⱖ0.8): this included TERT, GPC3, gankyrin, survivin, TOP2A, LYVE1, E-cadherin, IGFBP3, PDGFRA, TGFA, cyclin D1, and HGF. Logistic regression analysis identified a 3-gene set including GPC3 (18-fold increase in HCC, P ⫽ .01), LYVE1 (12-fold decrease in HCC, P ⫽ .0001), and survivin (2.2-fold increase in HCC, P ⫽ .02), which had a discriminative accuracy of 94%. The validity of the gene signature was confirmed in a prospective testing set. GPC3 immunostaining was positive in all HCCs and negative in dysplastic nodules (22/22 vs 0/14, respectively, P ⬍ .001). Nuclear staining for survivin was positive in 12 of 13 advanced HCC cases and in 1 of 9 early tumors. Conclusions: Molecular data based on gene transcriptional profiles of a 3-gene set allow a reliable diagnosis of early HCC. Immunostaining of GPC3 confirms the diagnosis of HCC.

H

epatocellular carcinoma (HCC) is the third leading cause of cancer-related death in the world.1 Its incidence is increasing in Europe and the United States.2 As a result of screening programs, early HCC diagnosis is feasible in 30%– 60% of cases in the West, enabling the application of curative treatments.1,3 Simultaneously, however, an increasing number of small nodules of ⬃2 cm are detected, which are difficult to characterize by radiologic or pathologic examination.4 – 6 Distinguishing dysplastic nodules from early tumors is an unresolved challenge, even among expert hepatopathologists.6 Immunostaining with CD34 and ␣-fetoprotein (AFP) has significant diagnostic limitations.7 Nonetheless, pathology is considered the gold standard of diagnosis. Noninvasive radiologic

criteria have been developed by using state-of-the-art radiologic techniques but are mostly reliable in tumors exceeding 2 cm in diameter.8 –10 Finally, serum biomarkers such as AFP, des-␥carboxyprothrombin (DGCP), and AFP-L3 fraction are not accurate for the early diagnosis of HCC.11,12 Tissue markers should be able to distinguish early HCC from other entities and, eventually, should be further tested as serum markers for surveillance purposes, as defined by the Early Detection Research Network of the National Cancer Institute.12 Genome-wide DNA microarray or quantitative real-time reverse-transcriptase polymerase chain reaction (RT-PCR) studies have attempted to identify markers of early HCC, such as heat shock protein 70 (HSP70),13 glypican-3 (GPC3),14 –16 telomerase reverse transcriptase (TERT),17 serine/threonine kinase 15 (STK6), and phospholipase A2 (PLAG12B).18 A molecular index including a 13-gene set has also been proposed (including TERT, TOP2A, and PDGFRA).19 More recently, a microarraygenerated signature of 120 genes was reported to discriminate between dysplastic nodules and HCC in hepatitis B virus (HBV) patients.20 Proteomic studies in tissue have not identified informative HCC markers so far.21 However, direct comparisons regarding gene expression in dysplastic nodules and early HCC are lacking in HCV patients. Overall, none of the reported genes or signatures is accepted as a molecular diagnosis of HCC.8,11 Distinction between preneoplastic nodules and early tumors has critical implications according to the guidelines of HCC management in Europe and the United States.8,11 Dysplastic lesions should be followed by regular imaging studies because one third of them will develop a malignant phenotype.22–24 Conversely, early tumors are treated by curative procedures such as resection, transplantation, and percutaneous ablation.1,25 Thus, there is an urgent need to identify better tools to characterize these lesions. Otherwise, the cost-effectiveness of the recall policies applied within surveillance programs will be significantly undermined. Abbreviations used in this paper: BIRC5, survivin; Ct, threshold cycle; GPC3, glypican-3; HBV, hepatitis B virus; HCV, hepatitis C virus; HCC, hepatocellular carcinoma; HGDN, high-grade dysplastic nodules; HSP70, heat shock protein 70; LGDN, low-grade dysplatic nodule; PLAG12B, phospholipase A2; ROC, receiver operating characteristic; STK6, serine/threonine kinase 15; TERT, telomerase reverse transcriptase. © 2006 by the AGA Institute 0016-5085/06/$32.00 doi:10.1053/j.gastro.2006.09.014

GENE SIGNATURE FOR EARLY HCC

1759

CLINICAL–LIVER, PANCREAS, AND BILIARY TRACT

December 2006

Figure 1. Flow chart summarizing the samples tested. We analyzed the gene expression of 55 genes in a training set consisting of 37 HCV-related samples (dysplastic nodules [17] and HCC [20]). The gene signature generated was then tested for consistency in 2 different groups of samples: (a) 20 advanced HCCs (by real-time RT-PCR), (b) 75 samples (control [3], cirrhotic [36], preneoplastic [14], and HCC [22]) (by immunohistochemistry) in HCV patients. Afterwards, we validate the data (real-time RT-PCR) in a testing set of 29 samples in HCV patients (dysplastic nodules [10] and small HCC [19]).

The present study aimed to identify a molecular diagnosis of early HCC. For that purpose, we tested the transcriptional profiles of 55 genes—previously implicated either as biomarkers of HCC or contributing to hepatocarcinogenesis—in 37 samples (17 dysplastic nodules and 20 early HCCs) of patients with hepatitis C virus (HCV) infection. A molecular diagnosis based on a 3-gene signature obtained was then confirmed in 20 advanced cases by real-time RT-PCR, and by immunohistochemistry in 75 samples, and was externally validated in an independent set of 29 samples (10 dysplastic nodules and 19 small HCC).

Materials and Methods Patients and Samples Samples were obtained from patients undergoing resection or liver transplantation in 3 university hospitals: 1 from the United States (Mount Sinai Hospital, NY) and 2 from Europe (Hospital Clínic, Barcelona, Spain; and National Cancer Institute, Milan, Italy). Laboratory techniques have been centralized in the laboratories of the Division of Liver Diseases, Hematology/Oncology, and the Center of Life Sciences of the Mount Sinai School of Medicine. The research protocol was approved by the institutional review boards of the 3 institutions, and informed consent was obtained in all cases.

Characteristics of the Samples A total of 106 fresh-frozen samples from HCV-positive patients were used in the study (Figure 1). Seventy-seven samples were used to generate the gene signature (17 cases of dysplasia and 20 early tumors, 10 controls and 10 cirrhotic

tissues) and confirm it (20 cases of advanced HCC), and 29 samples were used to validate the gene signature in an independent set. Seventy-five of these samples were used for the immunostaining analysis. Supplementary Table 1 (see Supplementary Table 1 online at www.gastrojournal.org) summarizes the characteristics of the 20 early HCC samples from patients in the training set. All patients presented with HCV-induced well-differentiated or moderately differentiated HCC, with a median tumor size of 18 mm [14 cases less than 20 mm; range, 8 – 45 mm]. Two cases showed presence of vascular invasion and/or satellite lesions at the pathologic examination. Patients with HBV-positive markers or a background of alcohol consumption, nonalcoholic steatohepatitis, hemochromatosis, or other causes of chronic liver disease were excluded. Lesions previously treated by percutaneous ablation or chemoembolization/lipiodolization were also excluded. The gene transcriptional profiles of these tumors were compared with 17 dysplastic nodules—10 low-grade dysplastic nodules (median size: 8.5 mm [range, 6 –12 mm]) and 7 high-grade dysplastic nodules (median size: 8.5 mm [range, 7–15 mm])— obtained from patients undergoing liver transplantation. Results were compared with 10 nontumoral cirrhotic tissues from HCC patients and 10 samples of normal tissue obtained from the healthy liver of patients undergoing resection for hepatic hemangioma (3), focal nodular hyperplasia (3), adenoma/cystadenoma (2), neuroendocrine tumor (1), and living donor liver transplantation (1). The messenger RNA (mRNA) expression profiles of the candidate genes in the training set were tested in 20 samples of advanced HCC to confirm their consistent dysregulation. In addition, the protein expression status of glypican 3 and sur-

1760

LLOVET ET AL

vivin was tested by immunohistochemistry in 75 samples (control [3], cirrhotic [36], preneoplastic [14], and HCC [22]). Finally, the gene signature was validated in an independent set of 29 samples: 10 dysplastic nodules and 19 small HCC (mean size, 2 ⫾ 0.6 cm; range, 0.9 –3 cm).

Sample Collection and Pathologic Data

CLINICAL–LIVER, PANCREAS, AND BILIARY TRACT

Once written informed consent was obtained, the main clinical and pathologic variables of the patients were recorded. Fresh tissue specimens were collected in the operating room/ pathology department and processed within 1 hour to minimize the alteration of gene expression because of ischemia. Samples were split in two. One part of each specimen was collected in either liquid nitrogen or RNA-later solution (Ambion Corp, The Woodlands, TX) and stored at ⫺80°C until use, and the other half was formalin fixed and paraffin embedded for morphologic examination and immunostaining analysis. Pathologic examination was considered the gold standard. Two expert pathologists reviewed each slide independently, then reached an agreement on the diagnosis of the lesions (S.T. and I.F.). Serial sectioning of the specimen at 5-mm intervals was performed on the grossing table. Nodules were classified as either low-grade dysplastic nodules (LGDN), high-grade dysplastic nodules (HGDN), or HCC according to the terminology of the International Working Party.26 A dysplastic nodule (DN) was identified as a distinct nodular lesion that varied from the surrounding liver parenchyma, with different size (minimum diameter, 6 mm), color, texture, and bulge from the cut surface.27 The distinction between low- and high-grade dysplastic nodules was based on the histologic features, in particular the degree of cytologic and architectural atypia. The hepatocytes in LGDN appeared normal or showed minimal nuclear atypia and slightly increased nucleus to cytoplasmic (N:C) ratio. Mitotic figures were absent. HGDN was identified if it demonstrated cytologic or structural atypia, or both, but insufficient for the diagnosis of HCC. The cytologic atypia may be diffuse or focal and characterized by nuclear hyperchromasia, nuclear contour irregularities, cytoplasmic basophilia or clear cell change, high N:C ratio, and occasional mitotic figures. Architecturally, the cell plates are thickened up to 3 cells, with occasional foci of pseudoglandular formation and nodule-in-nodule configuration. In addition, 2 pathologic stages were defined among the 20 target HCC samples: (1) Very early HCC was defined as well-differentiated tumors ⱕ2 cm in diameter without vascular invasion or satellites. (2) Early HCC: HCC ⱕ2 cm with microscopic vascular invasion/satellites or 2- to 5-cm well/moderately differentiated HCC without vascular invasion/satellites or 2 or 3 nodules ⬍3 cm well-differentiated. Histologically, early HCC was usually well differentiated, the trabeculae were usually greater than 3 cells thick, and pseudoglandular forms were often prominent. Invasion of portal tracts within and at the periphery of the lesion was occasionally present.24 This so-called stromal invasion according to Kojiro and Roskams was helpful in distinguishing very early HCC from HGDN.24 Tissue sampling was handled by using thin sections (4 ␮mol/L) of the target area, which was microdissected under a scanning microscope for PCR studies. The key genes were further tested in 20 samples of patients with advanced HCC, including 10 samples of patients with macroscopic vascular invasion/diffuse HCC.

GASTROENTEROLOGY Vol. 131, No. 6

Quantitative Real-Time RT-PCR RNA isolation, quality, and cDNA synthesis. We collected 40 mg to 1000 mg of tissue from each lesion. Fresh tissue specimens collected were saturated in RNA-later-ICE reagent (Ambion) and quickly ground under liquid nitrogen to maintain the RNA integrity and enhance the yield. The resulting tissue powder was homogenized in Trizol reagent (Invitrogen, Carlsbad, CA) with a polytron homogenizer. Total RNA was extracted from tissue homogenates according to the manufacturer’s instructions and was additionally digested with RNasefree DNase and purified with RNeasy columns (Qiagen, Valencia, CA). The purity of RNA samples was assessed by measuring the OD260/OD280 ratios on a NanoDrop ND-1000 spectrometer (NanoDrop, Wilmington, DE), resulting in a ratio of 2.00:2.08 in all cases. The quality and integrity of RNA was measured by a bioanalyzer (Agilent, Palo Alto, CA). Complementary DNA was synthesized from 5 ␮g purified total RNA derived from each sample using SuperScript III reverse transcriptase (Invitrogen) according to the manufacturers’ instructions. Real-Time RT-PCR. Expression of mRNA for genes of interest was measured by the Taqman real-time PCR method using an ABI PRISM 7900HT Sequence Detection System (Applied Biosystems, Foster City, CA). The probe and primer set for each gene was derived from Taqman Gene Expression Assays (Applied Biosystems). The real-time reactions were set up as triplicates for each gene in 384-well plates and run at the default PCR thermal cycling conditions: 50°C, 2 minutes; 95°C, 10 minutes; 40 cycles of 95°C, 15 seconds; and 60°C, 1 minute. Median threshold cycle (Ct) value from the triplicates was used in all the calculations. Normalization and genes tested. Fifty-five genes were selected from a thorough review of published studies and included potential molecular biomarkers or genes involved in hepatocarcinogenesis (see Supplementary Table 2 online at www.gastrojournal.org). Ribosomal RNA (18S) was chosen for normalization.28,29 To ensure the validity of using 18S to calculate the relative expression fold change, the 55 genes were tested together with the assay for 18S gene at 5 dilutions (2-fold series) of randomly selected HCC cDNA samples. All genes showed slope values (Ct vs log concentration blot) within a slope18S ⫾ 0.1. Significant results were validated using SYBR green.

Immunohistochemistry Formalin-fixed, paraffin-embedded tissue sections were baked at 55°C overnight, deparaffinized in xylene, and rehydrated in a graded series of ethanol solutions. Antigen retrieval was performed by immersing the slides in 10 mmol/L citrate buffer, pH 6.0, and heating them in a microwave at power level 10 for 3 minutes, followed by power level 7 for 10 minutes. To reduce background staining, the sections were incubated in 10% normal serum from the species in which the secondary antibody was raised. The optimal dilutions of the primary antibodies (monoclonal anti-GPC3 1:50, BioMosaics (Burlington, VT); rabbit antisurvivin 1:250, Abcam) were applied to the sections overnight at 4°C. After washing in phosphate-buffered saline (PBS), sections were incubated with the biotinylated secondary antibodies for 30 minutes at 37°C. Endogenous peroxide was blocked by immersing the slides in 3% hydrogen peroxide for 15 minutes. The antibody binding was detected with the avidinbiotin peroxidase complex system (Dako, Carpinteria, CA). Sec-

December 2006

GENE SIGNATURE FOR EARLY HCC

1761

Genes Up-regulated (n ⫽ 5) Telomerase reverse transcriptase: TERT Glypican-3: GPC3 Gankyrin: PSMD10 Survivin: BIRC5 Topoisomerase II: TOP2A Down-regulated (n ⫽ 7) Lymphatic vessel endothelial hyaluronan 1 (LYVE1): XLKD1 E-cadherin: CDH1 Insulin-like growth factor binding protein-3: IGFBP3 Platelet-derived growth factor receptor-A: PDGFRA Transforming growth factor-␣: TGF-␣ Cyclin D1: CCND1 Hepatocyte growth factor: HGF

Dysplastic (n ⫽ 17)

Early HCC (n ⫽ 20)

AUC

P value

3.5 (0.1⫺36) 2 (0.4⫺24) 1.1 (0.5⫺2.3) 1.5 (0.7⫺6.2) 2.1 (0.9⫺7.3)

38 (4.7⫺382) 36.6 (0.3⫺578) 2.6 (0.1⫺4.7) 3.3 (0.8⫺23.5) 6.5 (0.5⫺59)

0.92 0.84 0.82 0.80 0.85

.0001 .001 .0001 .002 .0001

0.6 (0.1⫺1.8) 2.2 (1.1⫺5) 1.7 (0.6⫺3) 1.8 (0.5⫺6.3) 1.2 (0.5⫺2.3) 2.8 (1.3⫺6.1) 1.3 (0.6⫺2.4)

0.05 (0.01⫺0.3) 0.8 (0.2⫺3.9) 0.2 (0.02⫺1.7) 0.4 (0.03⫺6.1) 0.3 (0.06⫺1.3) 1.6 (0.6⫺13) 0.3 (0.04⫺2.1)

0.90 0.84 0.85 0.85 0.86 0.91 0.90

.0001 .0001 .0001 .0001 .0001 .0001 .0001

NOTE. Gene expression is presented as fold change considering gene expression in normal tissue ⫽ 1. Results are presented as median (95% CI). All genes showed significant dysregulation by Mann–Whitney test, ⬎2-fold change in HCC compared with dysplasia, and AUC ⱖ0.80. AUC, area under the ROC curve.

tions were then counterstained with hematoxylin, dehydrated in a graded series of alcohol and xylene, and coverslipped. The variables measured were as follows: (1) Determination of immunostaining intensity (score 0 –3⫹; 0, negative; 1, weak; 2, moderate; 3, strong); (2) staining pattern (focal, diffuse); (3) subcellular localization (membrane, cytoplasm, or nucleus).

Statistical Analysis Results are expressed as mean ⫾ SD for continuous variables with normal distribution and median (95% confidence interval [CI]) for the other continuous and categorical variables. All the RT-PCR calculations were analyzed by using the expression of each gene in a given sample (Ct) normalized by the level of 18S in the sample (Ct⫺Ct18S ⫽ dCt) and further adjusted by the gene expression in the control group (ddCt). Results are expressed as fold changes (log 2 scale), considering the gene expression of the control group as 1. Comparisons between groups were done by the nonparametric Mann–Whitney test for continuous variables and the Fisher exact test for comparison of proportions. The area under the receiving operating curves (AUC) was assessed for all the genes to discriminate dysplastic nodules and early cancer. Correlations were calculated with the nonparametric Spearman’s coefficient. Strategy for selecting the best model: Genes significantly dysregulated in HCC in comparison to dysplastic nodules ⬎2-fold change by the Mann–Whitney test and Fisher exact test and showing an AUC ⱖ0.8 were included in a multivariate forward stepwise logistic regression analysis to determine the independent predictors of early HCC. In addition, ROC curves were used to establish the best cut-off to categorize each gene for the regression analysis. The diagnostic accuracy of the gene signatures proposed was calculated by sensitivity, specificity, positive and negative predictive values, and likelihood ratio, considering early HCC as the disease. The likelihood ratio for a positive result is the ratio of the chance of a positive result in a cancer sample to the chance of a positive result in the dysplastic sample. The molecular signatures identified were obtained from the analysis of 2 groups of genes: (1) including 12 genes significantly and con-

sistently up- or down-regulated in HCC and (2) including only the 5 genes significantly up-regulated in early HCC. A gene dendogram was obtained by hierarchical clustering of expression data by samples and genes using average linkage and Pearson correlation distance by using the TIGR-MEV program.30 All other calculations were done by the SPSS package (SPSS 12.0, Inc., Chicago, IL).

Results Gene Expression Profiles of Dysplastic Nodules and Early HCC Selection of the significant genes. Twelve genes were significantly, differentially expressed in early HCC compared with dysplastic nodules: 5 genes were up-regulated in cancer including TERT, GPC3, gankyrin (PSMD10), survivin (BIRC5), and TOP2A, and 7 were down-regulated including LYVE1 (XLKD1), E-cadherin (CDH1), IGFBP3, PDGFRA, TGFA, cyclin D1 (CCND1), and HGF (Table 1). Differential expression of all 12 genes was associated with an area under the ROC ⱖ0.8, and more than 2-fold change (either up- or down-regulation). Among the up-regulated genes, the median increase of GPC3 was 18-fold, TERT 10.8-fold, and survivin 2.2-fold increase in early HCC compared with dysplastic nodules. Among the down-regulated genes, LYVE1 was decreased 12-fold in early HCC compared with dysplastic nodules, IGFBP3 8.5-fold, and E-cadherin 2.8-fold. A dendrogram heat map graph was generated that displays a hierarchical clustering of these 12 genes and 37 samples according to the transcriptional profiles obtained by real time RT-PCR (see Supplementary Figure 1 online at www.gastrojournal.org). By using the 12-gene set, all early HCCs were properly classified, and only 1 dysplastic nodule was misclassified. Gene signatures and accuracy of the models. To optimize the selection of the most informative set of genes, we used logistic regression analysis categorizing the genes according to the best cut-off as determined by ROC curves. Several models were obtained depending on whether the analysis included the regression analysis with the 12 deregulated genes or only the 5

CLINICAL–LIVER, PANCREAS, AND BILIARY TRACT

Table 1. Genes Significantly Dysregulated in Early HCC Compared With Dysplatic Nodules

1762

LLOVET ET AL

GASTROENTEROLOGY Vol. 131, No. 6

Table 2. Accuracy of the Models Early HCC (n ⫽ 20)

Dysplastic nodule (n ⫽ 17)

Early HCC No HCC

19 1

1 16

Early HCC No HCC

18 2

0 17

Early HCC No HCC

20 0

3 14

Early HCC No HCC

19 1

2 15

Models tested Combination of 3 de-regulated genes: 1. LYVE1, glypican-3, survivin Predicted: 2. LYVE1, survivin, E-cadherin Predicted: Combination of 2 up-regulated genes: 1. Glypican-3, survivin Predicted: CLINICAL–LIVER, PANCREAS, AND BILIARY TRACT

2. Glypican-3, TERT Predicted:

Accuracy Overall

S/Sp

PPV/NPV

94.60%

95%/94%

95%/94%

94.60%

90%/100%

100%/89%

LR 16

N/Aa

92%

100%/82%

86%/100%

5.8

92%

95%/88%

90%/93%

8.6

S, Sensitivity; Sp, specificity; PPV, positive predictive value; NPV, negative predictive value; LR, likelihood ratio for a positive result. aLR cannot be calculated, zero denominator.

genes up-regulated in cancer (Table 2). Two 3-gene sets showed an accuracy of 94% in discriminating early HCC from dysplastic nodules. The best model includes LYVE1 (P ⫽ .0001), GPC3 (P ⫽ .0001), and survivin (P ⫽ .001), with a sensitivity of 95%, specificity of 94%, positive predictive value of 95%, negative predictive value of 94%, and likelihood ratio for a positive result of 16. When applying this model, only 2 samples were misclassified: 1 dysplastic nodule and 1 early HCC (NY24, NY6, respectively; Figure 2). Transcriptional profiles of these genes along with the ROC curves are shown in Figure 3A and B. The other 3-gene set model included LYVE1, survivin, and E-cadherin (accuracy, 94%; sensitivity, 90%), whereas 2 models only combining genes up-regulated (GPC3-survivin or GPC3-TERT) showed an accuracy of 92%, although the latter presented a better likelihood ratio (5.8 vs 8.5, respectively).

Gene expression profiles of the 5 relevant genes in cirrhotic tissue and in advanced HCC. The gene transcriptional profiles of the 5 genes involved in the selected models were further tested in a set of 10 cirrhotic tissues and in 20 patients with HCC at more advanced stages of the disease (10 cases with macroscopic vascular invasion/diffuse hepatic disease). As shown in Table 3, all 5 genes displayed a consistent trend of up-regulation (GPC3, TERT, survivin) or down-regulation (LYVE1, E-cadherin) at advanced stages of the neoplasm (see Supplementary Figure 2 online at www.gastrojournal.org). Validation of the gene signature in an independent sample set. We tested the expression profile of the 5 critical genes (LYVE-1, TERT, BIRC5, GPC3, CDH1) in a validation set of dysplastic nodules (10) and small HCC (19). As shown in Table 4, the dysregulation of the genes was consistent with the training set. In summary, 3 genes were significantly up-regulated in HCC in comparison with dysplastic nodules: TERT was 10.9fold change up-regulated in HCC, GPC3 was 9.29-fold change up-regulated, and BIRC5 was 5-fold change up-regulated. Conversely, 2 genes were significantly down-regulated in HCC compared with dysplastic nodules: LYVE-1 was down-regulated to 1/14 of expression in HCC, whereas E-cadherin was reduced to 1/4 level of expression in HCC compared with dysplastic nodules.

Immunohistochemistry Analysis

Figure 2. Graphic displaying the observed groups and predicted probabilities using the best model of combination of 3 genes (LYVE1, GPC3, and BRIC5) with an overall accuracy of 94%. Y-axis shows number of samples, and x-axis shows the percentage of certainty of classification of a given sample (0%, dysplastic nodule; 100%, early HCC). By using this model, only 2 samples were misclassified (arrows, NY24 and NY6).

The immunostaining analysis was designed to assess the in situ protein expression of the up-regulated genes comprising the best molecular signature of early HCC (GPC3, survivin). The analysis was performed in 36 paired samples of nontumoral cirrhotic tissue and 14 dysplastic nodules, 22 HCC samples (9 early HCC and 13 advanced HCC), and 3 healthy controls (see Supplementary Table 3 online at www.gastrojournal.org). GPC3 immunostaining was positive in all HCC cases and was negative in all dysplastic nodules (22/22 vs 0/14, respectively, P ⬍ .001) and normal controls. Figure 4 (and see Supplementary Figure 3 online at www.gastrojournal.org) displays examples of GPC3-negative staining of cirrhotic tissue and

GENE SIGNATURE FOR EARLY HCC

1763

CLINICAL–LIVER, PANCREAS, AND BILIARY TRACT

December 2006

Figure 3. (A) Gene expression profiles of the 3 most informative genes comparing dysplastic nodules (n ⫽ 17) and early HCC (n ⫽ 20). Results are expressed as fold-change. Boxes reflect median gene expression (25–75 percentile). (B) Area under the ROC curve considering HCC as disease: 2 genes were up-regulated, GPC3 (AUC ⫽ 0.84) and survivin (AUC ⫽ 0.8), and 1 was down-regulated, LYVE1 (AUC ⫽ 0.9). dysplastic nodules and positive staining for early HCC. The number of cells stained varied from focal areas (focal pattern) to all cells (diffuse pattern). There were several patterns of GPC3 staining: (1) Diffuse staining of the cytoplasm—sometimes accompanied by membranous staining—was more often seen in advanced poorly differentiated HCCs. Nuclear staining was visible only in 2 advanced HCCs but was also identified in infiltrating inflammatory cells within the tumor. (2) Perinuclear distribution was more frequently noted in the early cases than in advanced cases (9/9 cases vs 4/13 cases, respectively). (3) In pseudoglandular HCCs, staining was observed along the apical surface. Stronger staining intensity along canalicular membrane was also seen. A weak focal staining was detected in 7 of

27 nontumoral cirrhotic tissues (see Supplementary Figure 4 online at www.gastrojournal.org). Overall, there was a significant correlation between the gene expression of GPC3 and the immunostaining status and intensity (Spearman’s correlation: 0.8, P ⫽ .0001). Cytoplasmic survivin staining was negative in the controls and positive in cirrhosis (31 out of 36), dysplastic nodules (13 out of 14), and early HCC (7 out of 9) (see Supplementary Figure 5 online at www.gastrojournal.org). Significant differences were observed in subcellular colocalization: nuclear survivin staining was positive in 12 of 13 advanced cases compared with 1 of 9 early cases and none of the dysplastic and cirrhotic tissues (P ⫽ .001). The number of positive cells ranged from 1

1764

LLOVET ET AL

GASTROENTEROLOGY Vol. 131, No. 6

Table 3. Gene Transcriptional Profiles of the 5 Key Genes Selected in the Models Tested in Cirrhosis, Dysplastic Nodules, and Early and Advanced HCC Genes

Cirrhosis (n ⫽ 10)

Dysplastic (n ⫽ 17)

Early HCC (n ⫽ 20)

Advanced HCC (n ⫽ 20)

Glypican-3 (GPC3) TERT Survivin (BIRC5) LYVE1 (XLKD1) E-cadherin (CDH1)

14.8 (5.1⫺118) 1.8 (0.7⫺13) 2.6 (1.5⫺4.2) 0.9 (0.4⫺1.5) 3.9 (2⫺8)

2 (0.4⫺24) 3.5 (0.1⫺36) 1.5 (0.7⫺6.2) 0.6 (0.1⫺1.8) 2.2 (1.1⫺5)

36.6 (0.3⫺578) 38 (4.7⫺382) 3.3 (0.8⫺23.5) 0.05 (0.01⫺0.3) 0.8 (0.2⫺3.9)

412 (1⫺2364) 187 (0.8⫺2277) 24 (1.6⫺122) 0.08 (0.01⫺0.32) 0.8 (0.3⫺3.2)

NOTE. Gene expression is presented as fold changes considering gene expression in normal tissue ⫽ 1. Results are presented as median (95% CI). All comparisons are significant (P ⬍ .05, except for cirrhosis vs dysplastic [TERT, XLKD1], and for early vs advanced [XLKD1, CDH1]).

CLINICAL–LIVER, PANCREAS, AND BILIARY TRACT

to ⬎10 per 20⫻ field. Tumor infiltrating inflammatory cells occasionally displayed nuclear staining for survivin. There was a significant correlation between the gene expression of survivin and the subcellular colocalization (positive nuclear staining) (Spearman’s correlation: 0.73, P ⫽ .0001).

Discussion The wide implementation of surveillance programs in cirrhotic patients in North America and Europe is leading to the detection of small liver nodules of ⱕ2 cm in size by which a definitive diagnosis is often difficult to establish. This clinical problem will increase in the next 10 years because of the American Association for the Study of Liver Diseases and the European Association for the Study of the Liver recommendations mandating surveillance by ultrasonography every 6 months in high-risk populations.8,11 The mean size of the nodules detected is decreasing, resulting in complex recall policies to determine their nature. In cirrhotic livers, only half of nodules of less than 1 cm are ultimately malignant.31–33 Pathologic differentiation of preneoplastic lesions, particularly HGDN and early HCC, is difficult, even for expert hepatopathologists.6 New advances in imaging techniques enable the characterization of small nodules as HCC in few cases but require sophisticated radiologic equipment and expertise.5 The present study is the first attempt to devise a molecular model for the diagnosis of small early HCC in HCV patients that is technically simple and applicable in clinical practice. The 3-gene set signature with highest accuracy includes GPC3, survivin, and LYVE1 as the key genes to differentiate dysplastic nodules from early HCC by real-time RT-PCR. The accuracy of the model was 94%, and the individual genes showed ⬎2-fold change and an area under the ROC of ⱖ0.8. The 3-gene set was validated in an independent set of 29

samples. In addition, the study also devises 2 alternative gene sets focusing on up-regulated genes that include either GPC3-survivin or GPC3-TERT. To date, the solitary genes and molecular indexes that have been proposed as markers of HCC were derived from the comparison of gene expression between cirrhotic tissue and cancer, generally at advanced stages, leading to signatures unable to resolve the diagnostic problem. In contrast, we directly compared the 2 clinically conflicting entities. GPC3 is a heparin sulfate proteoglycan previously reported to be up-regulated in HCC in comparison with preneoplastic lesions and cirrhotic tissues at the mRNA34,35 and protein levels.14,36 Recent studies suggest that GPC3 promotes the growth of HCC by stimulating the canonical Wnt pathway.37 Transcriptional profiles of GPC3 were increased 18-fold in early HCC compared with dysplastic nodules (⬃10-fold change in the validation set), showing an area under the ROC of 0.84 for HCC diagnosis. Additionally there is a 38-fold increase and 412-fold increase in GPC3 mRNA in early and advanced HCC, respectively, compared with normal tissue. In the immunohistochemical study, GPC3 was very specific for HCC, in concordance with recent studies,36 showing a significant correlation between gene expression and the staining intensity. Unlike previous studies, however, we successfully used a commercially available antibody (monoclonal anti-GPC3 1:50, Zymed Laboratories, South San Francisco, CA). Thus, GPC3 is a useful tissue marker both at the mRNA and the protein level. The fact that GPC3 was up-regulated in nontumoral cirrhotic tissue (both at mRNA level and protein level—weak focal positive immunostaining in 7/27 cirrhotic tissues) should be taken into account when using it as a serum marker, as recently proposed.14 –16 Survivin is a member of the inhibitor of apoptosis proteins (IAP) family. This molecule is actively suppressed by p53 and has been functionally positioned downstream of

Table 4. Validation of the 5 Genes in an Independent Set of 29 Samples Genes

Dysplastic (n ⫽ 10)

Early HCC (n ⫽ 19)

AUC

P valuesa

Fold changes in the training set (median)

Glypican-3: GPC3 Survivin: BIRC5 LYVE1: XLKD1 TERT E-cadherin (CDH1)

1 (0.3⫺3.1) 1 (0.5⫺1.9) 1 (0.3⫺2.9) 1 (0.7⫺1.35) 1 (0.6⫺1.5)

9.29 (0.7⫺115) 5.05 (1.4⫺18) 0.07 (0.01⫺0.4) 10.9 (2.1⫺56) 0.25 (0.09⫺0.7)

0.77 0.84 0.88 0.74 0.87

.014 .001 .0001 .0001 .0001

⫻18 (increase) ⫻2.2 (increase) ⫻12 (decrease) ⫻10.8 (increase) ⫻2.8 (decrease)

NOTE. Gene expression profiles by Real Time RT-PCR. Gene expression is presented as median fold changes considering gene expression in dysplastic nodules as ⫽ 1 (95% CI). AUC, Area under the ROC curve. aP values: Mann–Whitney U test.

GENE SIGNATURE FOR EARLY HCC

1765

Figure 4. (A and B) Immunostaining for GPC3, counterstained with hematoxylin. (A) GPC3 in 0.8 cm HCC (right) and negative staining in the cirrhotic nodule (left) (original magnification, 100⫻). (B) Higher magnification showing diffuse cytoplasmic staining for GPC3 in tumor cells (original magnification, 400⫻).

several signaling pathways.38 Survivin has been implicated in colorectal, non–small-cell lung, and melanoma carcinogenesis.38 In HCC, it has been involved in cell proliferation and blocking apoptosis.39 Survivin expression was increased 3.3fold in early HCC and 24-fold in advanced HCC, compared with normal tissue. In our model, survivin added information to the gene set of GPC3-LYVE1 (2.2-fold change; AUC, 0.8), a fact that was confirmed in the validation set (5-fold change; AUC, 0.84). In addition, we identified a correlation between the level of mRNA expression and subcellular (nuclear) localization of the protein. A potential limitation of this marker is that 3 alternatively spliced transcripts have also been described (survivin-DeltaEx3, survivin 2B, and survivin 3B).40 LYVE1 is a hyaluronan receptor expressed by endothelial cells of normal lymphatic vessels but not by blood vessels. LYVE-1 is reportedly down-regulated within some solid tumors, such as breast, lung, and endometrial cancer, as a result of the destruction of the lymph vessels, whereas its expression is conserved in the tumor periphery.41– 43 Expression of LYVE1 was previously reported to be down-regulated in HCC at the mRNA42 and protein levels.41 Our data suggest a clear and progressive down-regulation of LYVE1 from cirrhosis to HCC. In fact, LYVE1 was 1.6-fold decreased in dysplastic nodules but 20-fold decreased in early HCC (12-fold change difference; AUC, 0.9). TERT and E-cadherin were also informative in our models. Activation of TERT is well-documented in early stages of HCC, and it is thought to be required for telomere stabilization and tumor progression.44 TERT was clearly up-regulated in early HCC (10.8-fold increase compared with dysplastic nodules) and showed an exponential increase in advanced HCC (187-fold increase). One limitation of this marker is the low amount of transcript in early tumors (cycle 32-34), consistent with the absence of signal/call detected in microarray studies.45 Finally, E-cadherin displayed a consistent down-regulation in early tumors, compared with dysplastic nodules (2.8-fold decrease, confirmed in the validation set ⬃4-fold change decrease). This protein, which is implicated in the Wnt canonical signaling

pathway, is reportedly down-regulated in cancer.46 Our data also suggest that other biomarkers such as HSP70, STK6, PLA2G13, FLT-3, and AFP were not useful to differentiate the 2 entities. The novelty of our investigation relies on the identification of a 3-gene set for the differential diagnosis of small nodules (median size of HCC was of 18 to 20 mm in both sets and of dysplastic nodules ⬃10 mm) in HCV patients. The combined mRNA analysis provides an accurate, simple, and objective diagnosis of the nature of the lesion, applicable in routine clinical use. For that purpose, we used standard commercially available PCR reagents enabling the reproducibility of the results. In addition, we performed a careful pathologic examination, dissecting the target lesions from the surrounding tissue and thus enabling the translation in the clinical practice through core biopsies. We proved that the data generated at early stages are consistent with changes also observed at advanced stages and that the results were further validated in a testing set of samples. These advantages make the current investigation unique compared with the signatures reported to date, either using microarray analysis18,20 or real-time RT-PCR.19 Smith et al proposed a 50gene signature to discriminate early HCC and cirrhosis18; Nam et al reported a 120-gene signature in HBV patients to differentiate dysplastic nodules and HCC.20 None of our 3 genes was included in their “early stage” signature, but glypican 3 was included in the “late stage” signature to discriminate different stages of the neoplasm. The lack of interface between these 2 datasets should be further explored, and the differences in methodology (qRT-PCR vs DNA microarrays) and etiology (HCV vs HBV) should be kept in mind. Finally, Paradis et al reported the first molecular index generated by RT-PCR. In this latter study, the training and testing samples included smaller numbers of dysplastic nodules/small tumors than reported herein, and 13 genes were required to obtain adequate diagnostic accuracies.19 In parallel to the clinical validation of our gene set, the search for new and more precise biomarkers must continue. Genome-wide microarrays and tissue proteomics are the most powerful tech-

CLINICAL–LIVER, PANCREAS, AND BILIARY TRACT

December 2006

1766

LLOVET ET AL

GASTROENTEROLOGY Vol. 131, No. 6

nologies and should be thoroughly tested in adequate target lesions and patient populations.47

Supplementary Data Supplementary data associated with this article can be found, in the online version, at doi:10.1053/j.gastro.2006.09.014.

17.

References

CLINICAL–LIVER, PANCREAS, AND BILIARY TRACT

1. Llovet JM, Burroughs A, Bruix J. Hepatocellular carcinoma. Lancet 2003;362:1907–1917. 2. Bosch FX, Ribes J, Diaz M, Cleries R. Primary liver cancer: worldwide incidence and trends. Gastroenterology 2004;127(Suppl 1):S5–S16. 3. Sangiovanni A, Del Ninno E, Fasani P, De Fazio C, Ronchi G, Romeo R, Morabito A, De Franchis R, Colombo M. Increased survival of cirrhotic patients with a hepatocellular carcinoma detected during surveillance. Gastroenterology 2004;126:1005– 1014. 4. Ikai I, Arii S, Kojiro M, Ichida T, Makuuchi M, Matsuyama Y, Nakanuma Y, Okita K, Omata M, Takayasu K, Yamaoka Y. Reevaluation of prognostic factors for survival after liver resection in patients with hepatocellular carcinoma in a Japanese nationwide survey. Cancer 2004;101:796 – 802. 5. Bolondi L, Gaiani S, Celli N, Golfieri R, Grigioni WF, Leoni S, Venturi AM, Piscaglia F. Characterization of small nodules in cirrhosis by assessment of vascularity: the problem of hypovascular hepatocellular carcinoma. Hepatology 2005;42:27–34. 6. Kojiro M. Focus on dysplastic nodules and early hepatocellular carcinoma: an Eastern point of view. Liver Transpl 2004;10(Suppl 1):S3–S8. 7. Park YN, Yang CP, Fernandez GJ, Cubukcu O, Thung SN, Theise ND. Neoangiogenesis and sinusoidal “capillarization” in dysplastic nodules of the liver. Am J Surg Pathol 1998;22:656 – 662. 8. Bruix J, Sherman M, Llovet JM, Beaugrand M, Lencioni R, Christensen E, Burroughs A, et al. Clinical management on hepatocellular carcinoma. Conclusions of the Barcelona-2000 EASL Conference. J Hepatol 2001;35:421– 430. 9. Burrel M, Llovet JM, Ayuso C, Iglesias C, Sala M, Miquel R, Caralt T, Ayuso JR, Sole M, Sanchez M, Bru C, Bruix J, and Barcelona Clinic Liver Cancer Group. MRI angiography is superior to helical CT for detection of HCC prior to liver transplantation: an explant correlation. Hepatology 2003;38:1034 –1042. 10. Lencioni R, Cioni D, Crocetti L, Della Pina C, Bartolozzi C. Magnetic resonance imaging of liver tumors. J Hepatol 2004;40: 162–171. 11. Bruix J, Sherman M. Management of hepatocellular carcinoma. Hepatology 2005;42:1208 –1236. 12. Marrero JA, Lok AS. Newer markers for hepatocellular carcinoma. Gastroenterology 2004;127(5 Suppl 1):S113–S119. 13. Chuma M, Sakamoto M, Yamazaki K, Ohta T, Ohki M, Asaka M, Hirohashi S. Expression profiling in multistage hepatocarcinogenesis: identification of HSP70 as a molecular marker of early hepatocellular carcinoma. Hepatology 2003;37:198 –207. 14. Capurro M, Wanless IR, Sherman M, Deboer G, Shi W, Miyoshi E, Filmus J. Glypican-3: a novel serum and histochemical marker for hepatocellular carcinoma. Gastroenterology 2003;125:89 –97. 15. Nakatsura T, Yoshitake Y, Senju S, Monji M, Komori H, Motomura Y, Hosaka S, Beppu T, Ishiko T, Kamohara H, Ashihara H, Katagiri T, Furukawa Y, Fujiyama S, Ogawa M, Nakamura Y, Nishimura Y. Glypican-3, overexpressed specifically in human hepatocellular carcinoma, is a novel tumor marker. Biochem Biophys Res Commun 2003;306:16 –25. 16. Hippo Y, Watanabe K, Watanabe A, Midorikawa Y, Yamamoto S, Ihara S, Tokita S, Iwanari H, Ito Y, Nakano K, Nezu J, Tsunoda H,

18.

19.

20.

21.

22.

23.

24. 25.

26. 27. 28.

29.

30.

31.

32.

Yoshino T, Ohizumi I, Tsuchiya M, Ohnishi S, Makuuchi M, Hamakubo T, Kodama T, Aburatani H. Identification of soluble NH2-terminal fragment of glypican-3 as a serological marker for early-stage hepatocellular carcinoma. Cancer Res 2004; 64:2418 –2423. Miura N, Maeda Y, Kanbe T, Yazama H, Takeda Y, Sato R, Tsukamoto T, Sato E, Marumoto A, Harada T, Sano A, Kishimoto Y, Hirooka Y, Murawaki Y, Hasegawa J, Shiota G. Serum human telomerase reverse transcriptase messenger RNA as a novel tumor marker for hepatocellular carcinoma. Clin Cancer Res 2005;11:3205–3209. Smith MW, Yue ZN, Geiss GK, Sadovnikova NY, Carter VS, Boix L, Lazaro CA, Rosenberg GB, Bumgarner RE, Fausto N, Bruix J, Katze MG. Identification of novel tumor markers in hepatitis C virus-associated hepatocellular carcinoma. Cancer Res 2003; 63:859 – 864. Paradis V, Bieche I, Dargere D, Laurendeau I, Laurent C, Bioulac Sage P, Degott C, Belghiti J, Vidaud M, Bedossa P Molecular profiling of hepatocellular carcinomas (HCC) using a large-scale real-time RT-PCR approach: determination of a molecular diagnostic index. Am J Pathol 2003;163:733–741. Nam SW, Park JY, Ramasamy A, Shevade S, Islam A, Long PM, Park CK, Park SE, Kim SY, Lee SH, Park WS, Yoo NJ, Liu ET, Miller LD, Lee JY. Molecular changes from dysplastic nodule to hepatocellular carcinoma through gene expression profiling. Hepatology 2005;42:809 – 818. Paradis V, Degos F, Dargere D, Pham N, Belghiti J, Degott C, Janeau JL, Bezeaud A, Delforge D, Cubizolles M, Laurendeau I, Bedossa P. Identification of a new marker of hepatocellular carcinoma by serum protein profiling of patients with chronic liver diseases. Hepatology 2005;41:40 – 47. Borzio M, Fargion S, Borzio F, Fracanzani AL, Croce AM, Stroffolini T, Oldani S, Cotichini R, Roncalli M. Impact of large regenerative, low-grade and high-grade dysplastic nodules in hepatocellular carcinoma development. J Hepatol 2003;39:208 –214. Terasaki S, Kaneko S, Kobayashi K, Nonomura A, Nakanuma Y. Histological features predicting malignant transformation of nonmalignant hepatocellular nodules: a prospective study. Gastroenterology 1998;115:1216 –1222. Kojiro M, Roskams T. Early hepatocellular carcinoma and dysplastic nodules. Semin Liv Dis 2005;25:133–142. Llovet JM, Schwartz M, Mazzaferro V. Resection and liver transplantation for hepatocellular carcinoma. Semin Liver Dis 2005; 25:181–200. International Working Party. Terminology of nodular hepatocellular lesions. Hepatology 1995;17:27–35. Hytiroglou P. Morphological changes of early human hepatocarcinogenesis. Semin Liver Dis 2004;24:65–75. Aerts JL, Gonzales MI, Topalian SL. Selection of appropriate control genes to assess expression of tumor antigens using real-time RT-PCR. Biotechniques 2004;36:84 –91. Gong Y, Cui L, Minuk GY. Comparison of glyceraldehyde-3-phosphate dehydrogenase and 28s-ribosomal RNA gene expression in human hepatocellular carcinoma. Hepatology 1997;26:803. Saeed AI, Sharov V, White J, Li J, Liang W, Bhagabati N, Braisted J, Klapa M, Currier T, Thiagarajan M, Sturn A, Snuffin M, Rezantsev A, Popov D, Ryltsov A, Kostukovich E, Borisovsky I, Liu Z, Vinsavich A, Trush V, Quackenbush J. TM4: a free, open-source system for microarray data management and analysis. Biotechniques 2003;34:374 –378. Bolondi L, Sofia S, Siringo S, Gaiani S, Casali A, Zironi G, Piscaglia F, Gramantieri L, Zanetti M, Sherman M. Surveillance programme of cirrhotic patients for early diagnosis and treatment of hepatocellular carcinoma: a cost-effectiveness analysis. Gut 2001;48:251–259. Tanaka Y, Sasaki Y, Katayama K, Hiramatsu N, Ito A, Murata H, Enomoto N, Oshita M, Mochizuki K, Tsujii M, Tsuji S, Kasahara A,

33.

34.

35.

36.

37.

38.

39.

40. 41.

Tomoda K, Nakamura H, Hayashi N, Hori M. Probability of hepatocellular carcinoma of small hepatocellular nodules undetectable by computed tomography during arterial portography. Hepatology 2000;31:890 – 898. O’Malley ME, Takayama Y, Sherman M. Outcome of small (10 – 20 mm) arterial phase-enhancing nodules seen on triphasic liver CT in patients with cirrhosis or chronic liver disease. Am J Gastroenterol 2005;100:1523–1528. Zhu ZW, Friess H, Wang L, Abou-Shady M, Zimmermann A, Lander AD, Korc M, Kleeff J, Buchler MW. Enhanced glypican-3 expression differentiates the majority of hepatocellular carcinomas from benign hepatic disorders. Gut 2001;48:558 –564. Hsu HC, Cheng W, Lai PL. Cloning and expression of a developmentally regulated transcript MXR7 in hepatocellular carcinoma: biological significance and temporospatial distribution. Cancer Res 1997;57:5179 –5184. Yamauchi N, Watanabe A, Hishinuma M, Ohashi K, Midorikawa Y, Morishita Y, Niki T, Shibahara J, Mori M, Makuuchi M, Hippo Y, Kodama T, Iwanari H, Aburatani H, Fukayama M. The glypican 3 oncofetal protein is a promising diagnostic marker for hepatocellular carcinoma. Mod Pathol 2005;18:1591–1598. Capurro MI, Xiang YY, Lobe C, Filmus J. Glypican-3 promotes the growth of hepatocellular carcinoma by stimulating canonical Wnt signaling. Cancer Res 2005;65:6245– 6254. Kannangai R, Wang J, Liu QZ, Sahin F, Torbenson M. Survivin overexpression in hepatocellular carcinoma is associated with p53 dysregulation. Int J Gastrointest Cancer 2005;35:53– 60. Ito T, Shiraki K, Sugimoto K, Yamanaka T, Fujikawa K, Ito M, Takase K, Moriyama M, Kawano H, Hayashida M, Nakano T, Suzuki A. Survivin promotes cell proliferation in human hepatocellular carcinoma. Hepatology 2000;31:1080 –1085. Li F. Role of survivin and its splice variants in tumorigenesis. Br J Cancer 2005;92:212–216. Mouta Carreira C, Nasser SM, di Tomaso E, Padera TP, Boucher Y, Tomarev SI, Jain RK. LYVE-1 is not restricted to the lymph vessels: expression in normal liver blood sinusoids and downregulation in human liver cancer and cirrhosis. Cancer Res 2001; 61:8079 – 8084.

GENE SIGNATURE FOR EARLY HCC

1767

42. Colombat M, Paradis V, Bieche I, et al. Quantitative RT-PCR in cirrhotic nodules reveals gene expression changes associated with liver carcinogenesis. J Pathol 2003;201:260 –267. 43. Williams CS, Leek RD, Robson AM, Banerji S, Prevo R, Harris AL, Jackson DG. Absence of lymphangiogenesis and intratumoural lymph vessels in human metastatic breast cancer. J Pathol 2003;200:195–206. 44. Satyanarayana A, Manns MP, Rudolph KL. Telomeres and telomerase: a dual role in hepatocarcinogenesis. Hepatology 2004; 40:276 –283. 45. Wurmbach E, Chen Y, Khitrov G, Roayaie S, Fiel I, Schwartz M, Thung S, Battiston C, Mazzaferro V, Bruix J, Bottinger E, Friedman S, Waxman S, Llovet JM. Gene expression profiles in pre-neoplastic lesions and hepatocellular carcinoma in hepatitis C virus -cirrhotic patients. Proc AACR 2005;46:A836. 46. Wei Y, Van Nhieu JT, Prigent S, Srivatanakul P, Tiollais P, Buendia MA. Altered expression of E-cadherin in hepatocellular carcinoma: correlations with genetic alterations, ␤-catenin expression, and clinical features. Hepatology 2002;36:692–701. 47. Lossos IS, Czerwinski DK, Alizadeh AA, Wechser MA, Tibshirani R, Botstein D, Levy R. Prediction of survival in diffuse large-B-cell lymphoma based on the expression of six genes. N Engl J Med 2004;350:1828 –1837.

Received May 14, 2006. Accepted September 7, 2006. Address requests for reprints to: Josep M. Llovet, MD, Division of Liver Diseases, Mount Sinai School of Medicine, Madison Avenue 1425, 11F-70, Box 1123, New York, New York 10029. e-mail: [email protected]; fax: (212) 849-2574. Supported by the Samuel Waxman Cancer Research Foundation, a grant from AGAUR (2003BEAI00138, 2004BE00226, Generalitat de Catalunya; to J.M.L.) and Instituto de Salud Carlos III (PI02/0596, Fondo de Investigaciones Sanitarias 2002–2005; to J.M.L. and F.I.S. PI050150 to J.B.), the National Institute of Health DK37340 (to S.L.F.) and the Bendheim Family Trust (to S.L.F.), and the Italian Association for Cancer Research (AIRC; to V.M. and C.B.). Josep M. Llovet is Professor of Research at Institut Català de Recerca Avancada (ICREA, IDIBAPS, Hospital Clínic).

CLINICAL–LIVER, PANCREAS, AND BILIARY TRACT

December 2006