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 ≥0.8): 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. 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 ≥0.8): 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. Hepatocellular carcinoma (HCC) is the third leading cause of cancer-related death in the world.1Llovet J.M. Burroughs A. Bruix J. Hepatocellular carcinoma.Lancet. 2003; 362: 1907-1917Abstract Full Text Full Text PDF PubMed Scopus (3858) Google Scholar Its incidence is increasing in Europe and the United States.2Bosch F.X. Ribes J. Diaz M. Cleries R. Primary liver cancer: worldwide incidence and trends.Gastroenterology. 2004; 127: S5-S16Abstract Full Text Full Text PDF PubMed Scopus (2209) Google Scholar 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.1Llovet J.M. Burroughs A. Bruix J. Hepatocellular carcinoma.Lancet. 2003; 362: 1907-1917Abstract Full Text Full Text PDF PubMed Scopus (3858) Google Scholar, 3Sangiovanni 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: 1005-1014Abstract Full Text Full Text PDF PubMed Scopus (525) Google Scholar Simultaneously, however, an increasing number of small nodules of ∼2 cm are detected, which are difficult to characterize by radiologic or pathologic examination.4Ikai 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: 796-802Crossref PubMed Scopus (386) Google Scholar, 5Bolondi L. Gaiani S. Celli N. Golfieri R. Grigioni W.F. Leoni S. Venturi A.M. Piscaglia F. Characterization of small nodules in cirrhosis by assessment of vascularity: the problem of hypovascular hepatocellular carcinoma.Hepatology. 2005; 42: 27-34Crossref PubMed Scopus (370) Google Scholar, 6Kojiro M. Focus on dysplastic nodules and early hepatocellular carcinoma: an Eastern point of view.Liver Transpl. 2004; 10: S3-S8Crossref PubMed Scopus (152) Google Scholar Distinguishing dysplastic nodules from early tumors is an unresolved challenge, even among expert hepatopathologists.6Kojiro M. Focus on dysplastic nodules and early hepatocellular carcinoma: an Eastern point of view.Liver Transpl. 2004; 10: S3-S8Crossref PubMed Scopus (152) Google Scholar Immunostaining with CD34 and α-fetoprotein (AFP) has significant diagnostic limitations.7Park Y.N. Yang C.P. Fernandez G.J. Cubukcu O. Thung S.N. Theise N.D. Neoangiogenesis and sinusoidal “capillarization” in dysplastic nodules of the liver.Am J Surg Pathol. 1998; 22: 656-662Crossref PubMed Scopus (258) Google Scholar 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.8Bruix J. Sherman M. Llovet J.M. 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: 421-430Abstract Full Text Full Text PDF PubMed Scopus (3841) Google Scholar, 9Burrel M. Llovet J.M. Ayuso C. Iglesias C. Sala M. Miquel R. Caralt T. Ayuso J.R. Sole M. Sanchez M. Bru C. Bruix J. Barcelona Clinic Liver Cancer GroupMRI angiography is superior to helical CT for detection of HCC prior to liver transplantation: an explant correlation.Hepatology. 2003; 38: 1034-1042Crossref PubMed Google Scholar, 10Lencioni R. Cioni D. Crocetti L. Della Pina C. Bartolozzi C. Magnetic resonance imaging of liver tumors.J Hepatol. 2004; 40: 162-171Abstract Full Text Full Text PDF PubMed Scopus (48) Google Scholar Finally, serum biomarkers such as AFP, des-γ-carboxyprothrombin (DGCP), and AFP-L3 fraction are not accurate for the early diagnosis of HCC.11Bruix J. Sherman M. Management of hepatocellular carcinoma.Hepatology. 2005; 42: 1208-1236Crossref PubMed Scopus (5209) Google Scholar, 12Marrero J.A. Lok A.S. Newer markers for hepatocellular carcinoma.Gastroenterology. 2004; 127: S113-S119Abstract Full Text Full Text PDF PubMed Scopus (176) Google Scholar 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.12Marrero J.A. Lok A.S. Newer markers for hepatocellular carcinoma.Gastroenterology. 2004; 127: S113-S119Abstract Full Text Full Text PDF PubMed Scopus (176) Google Scholar 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),13Chuma 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: 198-207Crossref PubMed Scopus (283) Google Scholar glypican-3 (GPC3),14Capurro M. Wanless I.R. Sherman M. Deboer G. Shi W. Miyoshi E. Filmus J. Glypican-3: a novel serum and histochemical marker for hepatocellular carcinoma.Gastroenterology. 2003; 125: 89-97Abstract Full Text Full Text PDF PubMed Scopus (772) Google Scholar, 15Nakatsura 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: 16-25Crossref PubMed Scopus (375) Google Scholar, 16Hippo Y. Watanabe K. Watanabe A. Midorikawa Y. Yamamoto S. Ihara S. Tokita S. Iwanari H. Ito Y. Nakano K. Nezu J. Tsunoda H. 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: 2418-2423Crossref PubMed Scopus (265) Google Scholar telomerase reverse transcriptase (TERT),17Miura 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: 3205-3209Crossref PubMed Scopus (70) Google Scholar serine/threonine kinase 15 (STK6), and phospholipase A2 (PLAG12B).18Smith M.W. Yue Z.N. Geiss G.K. Sadovnikova N.Y. Carter V.S. Boix L. Lazaro C.A. Rosenberg G.B. Bumgarner R.E. Fausto N. Bruix J. Katze M.G. Identification of novel tumor markers in hepatitis C virus-associated hepatocellular carcinoma.Cancer Res. 2003; 63: 859-864PubMed Google Scholar A molecular index including a 13-gene set has also been proposed (including TERT, TOP2A, and PDGFRA).19Paradis 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: 733-741Abstract Full Text Full Text PDF PubMed Scopus (97) Google Scholar More recently, a microarray-generated signature of 120 genes was reported to discriminate between dysplastic nodules and HCC in hepatitis B virus (HBV) patients.20Nam S.W. Park J.Y. Ramasamy A. Shevade S. Islam A. Long P.M. Park C.K. Park S.E. Kim S.Y. Lee S.H. Park W.S. Yoo N.J. Liu E.T. Miller L.D. Lee J.Y. Molecular changes from dysplastic nodule to hepatocellular carcinoma through gene expression profiling.Hepatology. 2005; 42: 809-818Crossref PubMed Scopus (150) Google Scholar Proteomic studies in tissue have not identified informative HCC markers so far.21Paradis V. Degos F. Dargere D. Pham N. Belghiti J. Degott C. Janeau J.L. 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: 40-47Crossref PubMed Scopus (225) Google Scholar 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.8Bruix J. Sherman M. Llovet J.M. 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: 421-430Abstract Full Text Full Text PDF PubMed Scopus (3841) Google Scholar, 11Bruix J. Sherman M. Management of hepatocellular carcinoma.Hepatology. 2005; 42: 1208-1236Crossref PubMed Scopus (5209) Google Scholar Distinction between preneoplastic nodules and early tumors has critical implications according to the guidelines of HCC management in Europe and the United States.8Bruix J. Sherman M. Llovet J.M. 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: 421-430Abstract Full Text Full Text PDF PubMed Scopus (3841) Google Scholar, 11Bruix J. Sherman M. Management of hepatocellular carcinoma.Hepatology. 2005; 42: 1208-1236Crossref PubMed Scopus (5209) Google Scholar Dysplastic lesions should be followed by regular imaging studies because one third of them will develop a malignant phenotype.22Borzio M. Fargion S. Borzio F. Fracanzani A.L. Croce A.M. 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: 208-214Abstract Full Text Full Text PDF PubMed Scopus (172) Google Scholar, 23Terasaki S. Kaneko S. Kobayashi K. Nonomura A. Nakanuma Y. Histological features predicting malignant transformation of nonmalignant hepatocellular nodules: a prospective study.Gastroenterology. 1998; 115: 1216-1222Abstract Full Text Full Text PDF PubMed Scopus (113) Google Scholar, 24Kojiro M. Roskams T. Early hepatocellular carcinoma and dysplastic nodules.Semin Liv Dis. 2005; 25: 133-142Crossref PubMed Scopus (233) Google Scholar Conversely, early tumors are treated by curative procedures such as resection, transplantation, and percutaneous ablation.1Llovet J.M. Burroughs A. Bruix J. Hepatocellular carcinoma.Lancet. 2003; 362: 1907-1917Abstract Full Text Full Text PDF PubMed Scopus (3858) Google Scholar, 25Llovet J.M. Schwartz M. Mazzaferro V. Resection and liver transplantation for hepatocellular carcinoma.Semin Liver Dis. 2005; 25: 181-200Crossref PubMed Scopus (775) Google Scholar 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. 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). 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. 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 survivin 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). 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.26International Working PartyTerminology of nodular hepatocellular lesions.Hepatology. 1995; 17: 27-35Google Scholar 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.27Hytiroglou P. Morphological changes of early human hepatocarcinogenesis.Semin Liver Dis. 2004; 24: 65-75Crossref PubMed Scopus (72) Google Scholar 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 ≤2 cm in diameter without vascular invasion or satellites. (2) Early HCC: HCC ≤2 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.24Kojiro M. Roskams T. Early hepatocellular carcinoma and dysplastic nodules.Semin Liv Dis. 2005; 25: 133-142Crossref PubMed Scopus (233) Google Scholar This so-called stromal invasion according to Kojiro and Roskams was helpful in distinguishing very early HCC from HGDN.24Kojiro M. Roskams T. Early hepatocellular carcinoma and dysplastic nodules.Semin Liv Dis. 2005; 25: 133-142Crossref PubMed Scopus (233) Google Scholar Tissue sampling was handled by using thin sections (4 μmol/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. 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 RNase-free 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 μg 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.28Aerts J.L. Gonzales M.I. Topalian S.L. Selection of appropriate control genes to assess expression of tumor antigens using real-time RT-PCR.Biotechniques. 2004; 36: 84-91Crossref PubMed Google Scholar, 29Gong Y. Cui L. Minuk G.Y. Comparison of glyceraldehyde-3-phosphate dehydrogenase and 28s-ribosomal RNA gene expression in human hepatocellular carcinoma.Hepatology. 1997; 26: 803Crossref PubMed Google Scholar 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. 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 avidin-biotin peroxidase complex system (Dako, Carpinteria, CA). Sections 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). 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 ≥0.8 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 consistently 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.30Saeed A.I. 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: 374-378Crossref PubMed Scopus (4069) Google Scholar All other calculations were done by the SPSS package (SPSS 12.0, Inc., Chicago, IL). 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 ≥0.8, 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