Identification of novel immunohistochemical tumor markers for primary hepatocellular carcinoma; clathrin heavy chain and formiminotransferase cyclodeaminase


  • Potential conflict of interest: Nothing to report.


Early diagnosis of hepatocellular carcinoma (HCC) greatly improves its prognosis. However, the distinction between benign and malignant tumors is often difficult, and novel immunohistochemical markers are necessary. Using agarose two-dimensional fluorescence difference gel electrophoresis, we analyzed HCC tissues from 10 patients. The fluorescence volumes of 48 spots increased and 79 spots decreased in tumor tissues compared with adjacent nontumor tissue, and 83 proteins were identified by mass spectrometry. Immunoblot confirmed that the expression of clathrin heavy chain (CHC) and Ku86 significantly increased, whereas formiminotransferase cyclodeaminase (FTCD), rhodanese, and vinculin decreased in tumor. The protein expression in tumor and nontumor tissues was further evaluated by immunostaining. Interestingly, CHC and FTCD expression was strikingly different between tumor and nontumor tissues. The sensitivity and specificity of individual markers or a combination for the detection of HCC were 51.8% and 95.6% for CHC, 61.4% and 98.5% for FTCD, and 80.7% and 94.1% for CHC+FTCD, respectively. Strikingly, the sensitivity and specificity increased to 86.7% and 95.6% when glypican-3, another potential biomarker for HCC, was used with FTCD. Moreover, CHC and FTCD were useful to distinguish early HCC from benign tumors such as regenerative nodule or focal nodular hyperplasia, because the sensitivity and specificity of the markers are 41.2% and 77.8% for CHC, 44.4% and 80.0% for FTCD, which is comparable with those of glypican-3 (33.3% and 100%). The sensitivity significantly increased by combination of these markers, 72.2% for CHC+FTCD, and 61.1% for CHC+glypican-3 and FTCD+glypican-3, as 44.4% of glypican-3 negative early HCC were able to be detected by either CHC or FTCD staining. Conclusion: Immunostaining of CHC and FTCD could make substantial contributions to the early diagnosis of HCC. (HEPATOLOGY 2008.)

Primary hepatocellular carcinoma (HCC) is a major health problem worldwide.1, 2 It is known that HCC develops from a chronic inflammatory liver disease due to hepatitis B virus and hepatitis C virus infection; therefore, HCC shows especially high prevalence in Asia and Africa, where the rate of hepatitis C virus infection is high.3 In Japan, HCC has been ranked as the third most common cancer causing death.4 Screening tests are serological and radiological. Alpha-fetoprotein, lens culinaris agglutinin-reactive fraction of alpha-fetoprotein, and serum protein induced by vitamin K absence or antagonist-II are the most commonly used diagnostic markers for HCC, although their sensitivity and specificity are not high enough and are inadequate for identifying early stage HCC.5, 6 The radiological test most widely used for surveillance is ultrasonography. Although ultrasound is able to detect small nodules of smaller than 2 cm, biopsy of these lesions is recommended for the diagnosis of HCC if the vascular profile on dynamic imaging is not characteristic of HCC.7 Such small masses range from benign nodules to malignant HCCs, and it is difficult, even for experienced pathologists, to distinguish dysplasia and well-differentiated HCC, especially when the lesion is small; therefore, development of new immunocytochemical markers is needed to diagnose early HCC.

Recently, the human genome project has been completed, and the genome database published. Moreover, high-throughput analysis of proteins has become possible by the development of tandem mass spectrometry technology. The breakthrough of this proteome technology enabled comparative studies of comprehensive protein expression and the identification of protein. As for HCC, proteome analysis using two-dimensional electrophoresis (2-DE), two-dimensional fluorescence difference gel electrophoresis (2D-DIGE), and liquid chromatography have recently been reported.8–10 Although a number of proteins have been identified as candidate markers for HCC,11, 12 none have been applied in the clinical setting; therefore, a more comprehensive and sophisticated approach is mandatory to find novel proteins associated with HCC. Oh-ishi et al.13 developed agarose 2-DE, which uses agarose gel in the first dimension. This method not only allows for large-scale quantitative comparisons of protein expression but is also able to resolve high-molecular-mass proteins larger than 150 kDa that are difficult to resolve with immobilized pH gradient (IPGs). We have previously identified several novel proteins with altered expression in primary colorectal cancer and esophageal cancer using agarose 2-DE or agarose 2D-DIGE.14, 15 These techniques appear to have advantages of adequate sensitivity, high reproducibility, and a wide dynamic range.

In this study, we aimed to identify novel biomarkers useful for the diagnosis of HCC. For that purpose, we compared protein expressions between HCC and adjacent nontumor tissues using the agarose 2D-DIGE method. Differentially expressed proteins were validated by immunoblot or immunostaining and were further evaluated for their potential as novel immunohistochemical markers.


2-DE, two dimensional electrophoresis; 2D-DIGE, two dimensional fluorescence difference gel electrophoresis; CHC, clathrin heavy chain; eHCC, early hepatocellular carcinoma; FNH, focal nodular hyperplasia nodules; FTCD, formiminotransferase cyclodeaminase; HCC, primary hepatocellular carcinoma; Ku86, 82-kDa ATP-dependent DNA helicase II; LRN, large regenerative nodule; mRNA, messenger RNA.

Materials and Methods

The following details can be found in the Supplementary Information 1: protein extraction, fluorescent dye (CyDye) labeling, agarose 2D-DIGE, enzymatic in-gel digestion of proteins, identification of proteins, and quantification of messenger RNA (mRNA).

Human Tissue Samples.

Ten HCC tissues were obtained at resection in the Department of General Surgery, Chiba University Hospital. The clinical features of these 10 cases are summarized in Table 1. Written informed consent was obtained from each patient before surgery. Excised samples were obtained within 1 hour after the operation from the tumor and adjacent non-tumor tissue. All excised tissues were immediately placed in liquid nitrogen and stored at −80°C until analysis.

Table 1. Clinical Features of Patients with HCC
No.AgeSexVirusSize (mm)Adjacent TissueAJCC Stage
  1. HCV, hepatitis C virus; LC, liver cirrhosis; CH, chronic hepatitis.

169Male70 × 70NormalIII
265MaleHCV60 × 45LCIII
376Male55 × 45CHIII
480MaleHCV30 × 38LCII
558Female45 × 40LCII
661MaleHCV35 × 32CHII
765MaleHCV25 × 16LCII
875MaleHCV25 × 23CHI
975MaleHCV25 × 20LCII
1079MaleHCV110 × 90CHIII


Protein extracts were separated by electrophoresis on 10% to 20% polyacrylamide gradient gel. Proteins were transferred to polyvinylidene fluoride membranes (Millipore, Bedford, MA) in a tank transfer apparatus (Bio-Rad, Hercules, CA), and the membranes were blocked with 5% skim milk in phosphate-buffered saline. Anti-clathrin heavy chain (CHC) mouse monoclonal antibody (BD Biosciences Pharmingen) diluted 1:4000, anti-82 kDa adenosine triphosphate-dependent DNA helicase II (Ku86) mouse monoclonal antibody (COSMO BIO Co., Ltd, Tokyo, Japan) diluted 1:4000, anti-vinculin mouse monoclonal antibody (Upstate Biotechnologies, NY) diluted 1:8000, anti-formiminotransferase cyclodeaminase (FTCD) rabbit polyclonal antibody (Abcam, Cambridge, UK) diluted 1:2000, and anti-thiosulfate sulfurtransferase (rhodanese) rabbit polyclonal antibody (Santa Cruz Biotechnology Inc., Santa Cruz, CA) diluted 1:1000 in blocking buffer were used as primary antibodies. Goat anti-mouse immunoglobulin G (IgG) horseradish peroxidase (Bio-Rad Laboratories, Hercules, CA) diluted 1:3000, and rabbit antigoat IgG horseradish peroxidase (Cappel, West Chester, PA) diluted 1:500 in blocking buffer were used as secondary antibodies. Antigens on the membrane were detected with enhanced chemiluminescence detection reagents (GE Healthcare).


From 20 HCC specimens (five well-differentiated, 10 moderately differentiated, and five poorly differentiated HCC), paraffin-embedded blocks of tumor and adjacent nontumor tissue were collected in the Department of General Surgery, Chiba University Hospital. Four-μm sections from paraffin tissue were fixed on slide glasses. In addition, tissue arrays (CA3, CSN3, CS3; SuperBio-Chips, Seoul, Korea) were used for immunohistochemistry, which contained 83 tumor (14 well differentiated, 40 moderately differentiated, 11 poorly differentiated, and 18 unclassified HCC) and 68 nontumor liver tissues. Two adenoma specimens were obtained from the Division of Clinical Investigation, National Hospital Organization, Chiba Medical Center. Three large regenerative nodules (LRN), five focal nodular hyperplasia nodules (FNH), and 18 early HCC (eHCC) specimens were obtained from the Institute of Gastroenterology, Tokyo Women's Medical University Hospital. Tissues were deparaffinized in xylene and rehydrated by reducing the concentration of ethanol (100%, 100%, and 70%, 5 minutes each). Antigents were unmasked with microwave irradiation for 5 minutes in pH 6.0 citric buffer three times. Primary antibodies were diluted as follows. Anti-CHC antibody diluted 1:200, anti-FTCD antibody diluted 1:200, anti-rhodanese antibody diluted 1:100, and anti-Glypican-3 antibody (Biomosaics, Burlington, VT) diluted 1:100 in blocking buffer. EnVision + system (DAKO Japan, Kyoto, Japan) was used to visualize tissue antigens. Tissue sections were counterstained with hematoxylin for 1 minute. Protein expression was scored as negative (0), weak (1), moderate (2), and strong (3). Two pathologists evaluated immunohistochemical staining of the samples. The results of the evaluation agreed in 96.0% of cases. When the results were discordant, the judgment was made by the other investigator.


Identification of Altered Expressed Proteins in Human HCC Tissue.

To search for novel biomarkers useful for the diagnosis of HCC, we used the agarose 2D-DIGE method to explore proteins differently expressed between HCC and adjacent nontumor tissues. Each nontumor sample was labeled with Cy3, each cancer sample was labeled with Cy5, and pooling aliquots were labeled with Cy2, respectively. These labeled proteins were mixed and separated in the same 2D gel (Fig. 1A). Protein spots that were increased or decreased in tumor tissues were displayed as red or green, respectively. These spots were detected and quantitated with DeCyder imaging analysis software, and then statistical analysis was performed across the 10 gels. The fluorescence volumes of 48 spots increased and 79 spots decreased in cancer tissues compared with adjacent nontumor tissue (Student t test, P < 0.05). To identify the proteins, 500 μg whole-cell lysates of HCC or nontumor tissues (Table 1; cases 1 and 2) were separated by conventional agarose 2-DE, and proteins were visualized by Coomassie blue staining (Fig. 1B). We carefully compared the DIGE image with Coomassie blue staining gels and picked altered protein spots manually. A total of 101 (83 proteins) of 127 spots were identified by mass spectrometry (Tables 2 and 3). The expression of these identified proteins was differentially expressed in most of the 2D-DIGE gel (Tables 2 and 3). Although many have previously been reported as differentially expressed proteins in HCC, which we were able to reproduce using a proteomic approach, a few were further tested for their clinical use. Moreover, most down-regulated proteins were related to detoxification and metabolism, which probably reflect liver dysfunction accompanying the development of HCC. Thus, we made an attempt to find proteins that could be potential diagnostic markers for HCC.

Figure 1.

Proteome analysis of HCC tissues using agarose 2D-DIGE and agarose 2-DE. Whole-cell lysates were prepared from matched samples of tumor tissue, adjacent nontumor tissue, and pooling aliquots (internal control). (A) Increased protein spots in tumor tissues are displayed in red (Cy5), and decreased protein spots in tumor tissues are displayed in green (Cy3). (B) Conventional agarose 2-DE patterns were visualized by Coomassie Blue staining. Protein spots cut from this gel were identified by mass spectrometry and are shown in red circles (up-regulated in HCC) or black circles (down-regulated in HCC). 2D-DIGE, two-dimensional fluorescence difference gel electrophoresis.

Table 2. Protein Expression in HCC and Adjacent Nontumorous Tissue
Protein Increased in Tumor Tissue
NoDatabase Accession No.Protein NameAverage MassHomogeneity Rate (%)T-testScoreCoverage (%)Fold IncreaseReferences*
  • *

    The references details can be found in Supplementary Information 2.

  • Previously reported to be up-regulated in HCC.

  • Previously reported to be down-regulated in HCC.

T1gi-2506872Fibronectin precursor262,586800.03573.83.21.53(1)
T2gi-4758012Clathrin heavy chain 1191,59589<0.00142.23.12.26 
T3gi-19913410Major vault protein99,248100<0.00182.48.61.73(2)
T4gi-2804273Alpha actinin 4102,250780.00888.610.31.36 
T5gi-4507677Tumor rejection antigen (gp96)92,450900.022167.420.21.67(3)
T6gi-6005942Valosin-containing protein89,247900.028128.218.41.49(4)
T7gi-34304590Heat shock 90kDa protein 1 beta83,1941000.00251.97.02.13(5)
T8gi-1086394582-kDa ATP-dependent DNA helicase II (Ku86)82,6861000.02053.08.03.04 
T9gi-4506077Protein kinase C substrate 80K-H isoform 159,27878<0.00152.510.71.69 
T10gi-862457Enoyl-CoA hydratase82,888800.02143.06.82.09(6)
T11gi-4557385Complement component 3 precursor187,027800.01488.15.31.71(7)
T12gi-4389275Albumin complex with myristic/tri-liodobenzoicacid66,0171000.001127.017.11.43 
T14gi-12937960kDa Heat shock protein, mitochondrial precursor60,998830.004139.421.21.88(5)
T15gi-576554Antithrombin III variant52,673750.04130.48.41.53(9)
T16gi-4757900Calreticulin precursor48,1231000.01783.315.41.52(10)
T17gi-2506774Keratin, type II cytoskeletal 8 (Cytokeratin 8)53,623880.008150.232.52.29 
T18gi-4504505Hydroxysteroid (17-beta) dehydrogenase 479,6681000.01557.28.32.09(11)
T19gi-24497583Aldo-ketoreductase family 1, member C336,835900.038114.325.71.86(12)
T20gi-4504447Heterogeneous nuclear ribonucleoprotein A2/B1 isoform A235,987880.00840.612.31.44 
T21gi-21735621Mitochondrial malate dehydrogenase precursor35,485880.03753.318.71.28(13)
T22gi-503176511-Beta-hydroxysteroid dehydrogenase 132,382880.03721.34.21.28 
T23gi-30584583Homo sapiens tyrosine 3-monooxygenase29,25090<0.00193.537.22.16 
Table 3. Protein Expression in HCC and Adjacent Nontumorous Tissue
NoDatabase Accession No.Protein NameAverage MassHomogeneity Rate (%)T-testScoreCoverage (%)Fold DecreaseReferences*
  • *

    The references can be found in Supplementary Information 2.

  • Previously reported to be down-regulated in HCC.

N1gi-24657579VCL protein (VINCULIN)116,7181000.01558.36.01.81 
N2gi-1709947Pyruvate carboxylase, mitochondrial precursor129,533700.007137.310.81.70(14)
N3gi-4938304Lysine-ketoglutarate reductase102,06490<0.00162.86.91.55 
N4gi-19353009Similar to elongation factor 2b57,4551000.00830.86.21.39 
N5gi-8659555Aconitase 198,3181000.008151.518.81.39 
N7gi-40789249Aspartyl-tRNA synthetase 2 (mitochondrial)73,4981000.01136.09.02.13 
N8gi-12655193Phosphoenolpyruvate carboxykinase 2 (mitochondrial)70,635750.025189.420.11.86 
N9gi-11761629Fibrinogen, alpha chain isoform alpha preproprotein69,695800.00386.917.92.07 
N11gi-4758312Electron-transferring-flavoprotein dehydrogenase68,489890.00482.612.31.48 
N12gi-4557645Heterogeneous nuclear ribonucleoprotein L isoform a60,1691000.00741.710.41.74 
N13gi-20149621Hypothetical protein LOC2600758,8921000.001168.635.12.29 
N15gi-11140815Formiminotransferase cyclodeaminase58,8711000.004158.620.32.26 
N16gi-7431380Uridine diphosphoglucose dehydrogenase55,0401000.03231.07.71.31 
N17gi-4507813UDP-glucose dehydrogenase54,9711000.03250.012.41.31 
N19gi-25108887Aldehyde dehydrogenase family 7 member A155,348780.00325.26.01.71 
N20gi-4885281Glutamate dehydrogenase 161,379100<0.001181.026.81.42 
N21gi-13027638UDP-glucose pyrophosphorylase 2 isoform a56,947100<0.001119.623.42.37 
N22gi-7705688Leucine aminopeptidase56,031100<0.00194.017.42.44 
N23gi-28949044Human mitochondrial aldehyde dehydrogenase54,4261000.023101.315.21.49(16)
N24gi-20151189Human glutamate dehydrogenase-apo form55,990100<0.001181.026.81.74 
N25gi-16306550Selenium binding protein 152,3391000.01090.018.81.42 
N26gi-22547189Serine hydroxymethyl transferase 1 (soluble) isoform 248,978890.01089.722.12.30 
N27gi-4503481Eukaryotic translation elongation factor 1 gamma50,1001000.00130.95.82.23 
N28gi-6730018Human L-arginine:glycine amidinotransferase44,6251000.001163.726.52.23 
N29gi-50317513-Hydroxy-3-methylglutaryl-coenzyme A synthase 256,581800.007170.122.11.66 
N30gi-19743875Fumarate hydratase precursor54,61989<0.001121.027.22.31 
N31gi-16878083Enolase 346,88489<0.00149.312.42.31 
N32gi-4557888Keratin 1848,010860.037134.918.21.41(17)
N33gi-16950633Argininosuccinate synthetase46,482100<0.00184.416.93.58(18)
N34gi-4530461Betaine-homocysteine methyltransferase44,980100<0.001139.033.45.19 
N35gi-28178832Isocitrate dehydrogenase 2 (NADP+), mitochondrial50,891900.001110.027.02.09 
N36gi-4557587Fumaryl acetoacetate hydrolase (fumary lacetoacetase)46,3261000.001106.524.12.17(19)
N37gi-7110715SEC14-like 246,127890.03883.317.61.69 
N38gi-12804931Acetyl-coenzyme A acyltransferase 241,9061000.001119.429.42.27 
N39gi-7542837Medium-chain acyl-CoA dehydrogenase46,57088<0.00188.618.41.71 
N40gi-4557237Acetyl-coenzyme A acetyltransferase 1 precursor45,18178<0.001138.435.31.90 
N41gi-4501853Acetyl-coenzyme A acyltransferase 144,273710.04656.79.31.68 
N42gi-4504067Aspartate aminotransferase 146,2291000.01895.721.72.00 
N44gi-4501929Class I alcohol dehydrogenase, alpha subunit39,84088<0.001121.223.04.12 
N45gi-494091Chain A, alcohol dehydrogenase (beta-1 isoenzyme)39,6061000.01984.820.43.67 
N46gi-20530221BLOCK 2537,747800.02930.018.82.34 
N47gi-13096743Chain A, human gamma-2 alcohol dehydrogenase39,676100<0.001106.620.94.21 
N48gi-25777615Proteasome 26S non-ATPase subunit 737,007800.00621.47.31.86(21)
N49gi-1352403Fructose-1,6-bisphosphatase 136,810100<0.001150.443.92.28 
N50gi-4507155Sorbitol dehydrogenase38,293900.00569.221.92.06 
N51gi-113611Fructose-bisphosphate aldolase B (liver-type aldolase)39,4551000.003137.928.22.59 
N52gi-1705823Aldo-ketoreductase family 1 member C437,0971000.045110.026.61.38(21)
N53gi-688031Aryl sulfotransferase ST1A334,191890.00330.411.82.63 
N54gi-8815565Alcohol/hydroxysteroid sulfotransferase33,747890.00329.38.62.63 
N55gi-9506741Glycine N-methyltransferase32,724750.00442.413.62.28 
N56gi-12654663Esterase D/formylglutathione hydrolase31,502750.00456.522.12.28 
N57gi-9087220Sulfotransferase 1A1 (arylsulfotransferase 1)34,1791000.00984.028.62.01 
N58gi-17402865Thiosulfate sulfurtransferase (rhodanese)33,410100<0.00199.528.21.58 
N59gi-4503607Electron transfer flavoprotein, alpha polypeptide35,06190<0.001120.843.91.72 
N60gi-45033012,4-Dienoyl CoA reductase 1 precursor36,049890.00491.524.21.76 

Validation of Differentially Expressed Protein Between Tumor and Nontumor Tissues.

Although 2-DE is a powerful technique, multiple proteins may be included in one spot, leading to misinterpretation of the results. Therefore, to confirm the difference of protein expression between tumor and nontumor tissues, validation using other methods is essential. Thus, immunoblot analyses of several proteins with commercially available antibodies were performed to confirm the differential protein expression in tumor tissues. CHC and Ku86 were up-regulated, whereas FTCD, rhodanese, and vinculin were down-regulated in most tumor tissues (Fig. 2). It is interesting to note that a ladder of smaller bands below full-length vinculin was observed and one of the bands around 60 kDa, which might be cleaved products of vinculin, was stronger in nontumor tissues than in tumor tissues.

Figure 2.

Immunoblot analysis of differential protein expression in tumor tissues. Total protein lysates prepared from nine matched samples of tumor (T) and adjacent nontumor tissue (N) were separated by electrophoresis on 10% to 20% polyacrylamide gradient gel, and immunoblotted with anti-clathrin heavy chain (CHC) antibody, anti-82 kDa adenosine triphosphate–dependent DNA helicase II (Ku86) antibody, anti-vinculin antibody, anti-formiminotransferase cyclodeaminase (FTCD) antibody, anti-rhodanese antibody, and anti–β-actin antibody (loading control). The intensity of each band was measured with NIH Image, and these proteins levels between tumor and nontumor tissue, normalized with β-actin, were calculated. The expressions of CHC and Ku86 were increased in tumor tissues, whereas vinculin, FTCD, and rhodanese were decreased in tumor tissues.

Quantification of mRNA Levels.

Differentially expressed proteins are commonly regulated at the transcriptional level or through translational and posttranslational modifications. To explore the mechanisms leading to the changes of protein expression, we examined the mRNA level of the proteins by quantitative reverse transcription polymerase chain reaction. The mRNA levels of FTCD, rhodanese, and vinculin were decreased in most tumor tissues, consistent with the changes of protein expression. In contrast, CHC and Ku86 mRNA levels did not correlate with their protein expression levels (Fig. 3); therefore, overexpression of CHC and Ku86 in tumor tissues does not occur at the transcriptional level.

Figure 3.

Quantification of mRNA levels in tumor tissues. Total RNAs were prepared from nine matched samples of tumor (T) and adjacent nontumor tissue (N), and real-time quantitative reverse transcription polymerase chain reaction of CHC (clathrin heavy chain), Ku86 (anti–82-kDa ATP-dependent DNA helicase II), vinculin, FTCD (formiminotransferase cyclodeaminase), and rhodanese mRNA was performed using a LightCycler. These mRNA levels were normalized by β-actin level.

Immunohistochemical Analysis.

Although there was no bias in the cellularity of tumor and adjacent nontumor tissues, whole tissue sections included nonhepatic parenchymal cells, and the altered protein expression in our 2-DE analysis may emanate from such nonhepatocyte components. Thus, the differential protein expression in HCC was also validated by immunohistochemistry to examine the localization of identified proteins. Paraffin-embedded tumor tissue and adjacent nontumor tissues of all 20 cases were stained with antibodies that were used in immunoblot analysis (Fig. 4). CHC has been reported to localize in the plasma membrane and the cytoplasmic face of intracellular organelles. Although no staining of CHC was observed in nontumor tissues, tumor cells showed scattered staining in the cytoplasm and plasma membrane (Fig. 4A). Bile duct, endothelial cell, and Kupffer cells were also positively stained. FTCD showed strong and uniform staining in the cytoplasm of nontumor tissue compared with faint staining in the cytoplasm of tumor cells (Fig. 4B). Rhodanese showed a mixture of scattered and strong staining in the cytoplasm of nontumor tissue, whereas tumor tissue was scarcely stained (Fig. 4C). These results confirmed the differential expression of proteins between tumor and nontumor tissues.

Figure 4.

Immunohistochemical analyses of differential protein expression in tumor tissues. From HCC specimens, paraffin-embedded blocks of tumor and adjacent nontumor tissue were collected. Four-micron sections from paraffin tissue were fixed on slide glasses. The primary antibody is equal to immunoblot analysis. EnVision + system was used to visualize tissue antigens, and tissue sections were counterstained with hematoxylin. (A) Although no staining of CHC (clathrin heavy chain) was observed in nontumor tissues, tumor cells showed scattered staining in the cytoplasm and plasma membrane. (B) FTCD (formiminotransferase cyclodeaminase) showed strong and uniform staining in the cytoplasm of nontumor tissue compared with faint staining in the cytoplasm of tumor cells. (C) Rhodanese showed a mixture of scattered and strong staining in the cytoplasm of nontumor tissue, whereas tumor tissue was scarcely stained.

Clinical Application.

Discrimination of well-differentiated HCC and nontumor tissues within a cirrhotic liver is often difficult even for experienced pathologists, and additional immunohistochemical markers are needed. Although the expression level of CHC and FTCD was strikingly different between tumor and nontumor tissues, analysis of 10 cases is not enough to consider CHC and FTCD as potential histological markers for HCC. Also the histology of nontumor tissues of the 10 cases was variable. To validate the usefulness of CHC and FTCD staining for the diagnosis of HCC, we obtained a commercial tissue array of HCC in which the degree of tumor differentiation and clinicopathological features had been proven (Table 4). The expression level of CHC and FTCD was scored as 0, 1, 2, or 3 by the staining intensity of the proteins. Most HCC tissues showed strong CHC expression (score 3) and negative to weak (score 0, 1) FTCD expression (43 of 83 cases and 51 of 83 cases). In contrast, most non-HCC tissues showed negative to moderate CHC expression (score 0, 1, 2) and moderate to strong FTCD (score 2, 3) expression (65 of 68 cases and 67 of 68 cases) (Table 5A). The sensitivity and specificity for the diagnosis of HCC using CHC expression level above were 51.8% and 95.6%, whereas those using FTCD expression level were 61.4% and 98.5%, respectively. If the combination of CHC and FTCD expression levels were used, the sensitivity and specificity for the diagnosis of HCC were 80.7% and 94.1%, respectively (Table 5B). Interestingly, CHC and FTCD expression level in tumor tissues correlates with tumor differentiation (well-differentiated HCC, 21.4%, 28.6%; moderately differentiated HCC, 52.5%, 15.0%; poorly differentiated HCC, 72.7%, 9.1%, respectively) (Table 5C). CHC and FTCD expression levels did not correlate with other clinicopathological features (age, sex, stage, and tumor size) (data not shown). These results indicated that immunostaining of CHC and FTCD could contribute to the pathological diagnosis of HCC.

Table 4. Histology of HCC and Non-HCC Tissues on Tissue Array
HCC tissueWell-differentiated HCC14
 Moderately differentiated HCC40
 Poorly differentiated HCC11
Non-HCC tissueChronic hepatitis8
 Dysplastic nodule1
 Nonspecific reactive change11
 Reactive hepatitis20
Table 5. Immunohistochemical Analysis From Tissue Array of HCC
 HCC (n = 83)Non-HCC (n = 68) Sensitivity (%)Specificity (%)
CHC = 3434036551.895.6
FTCD ≤ 1513216761.498.5
Glypican-3 ≥ 2523126662.797.1
CHC = 3 or FTCD ≤ 1671646480.794.1
CHC = 3 or Glypican-3 ≥ 2592466271.191.2
FTCD ≤ 1 or Glypican-3 ≥ 2721136586.795.6
33 (4.4%)3 (21.4%)21 (52.5%)8 (72.7%)45 (66.2%)4 (28.6%)6 (15.0%)1 (9.1%)
227 (39.7%)6 (42.9%)18 (45.0%)3 (27.3%)22 (32.4%)4 (28.6%)9 (22.5%)4 (36.4%)
131 (45.6%)4 (28.6%)1 (2.5%)0 (0%)1 (1.5%)4 (28.6%)16 (40.0%)4 (36.4%)
07 (10.3%)1 (7.1%)0 (0%)0 (0%)0 (0%)2 (14.3%)9 (22.5%)2 (18.2%)

Glypican-3 has been reported as a promising marker in the distinction between HCC and nonmalignant hepatocellular lesions.16, 17 Therefore, we compared the diagnostic value of CHC and FTCD for HCC with that of glypican-3 and also examined whether the combination of the three potential markers can improve the diagnostic accuracy of HCC. The sensitivity and specificity of glypican-3 were 62.7% and 97.1%, respectively, which were comparable with those of CHC or FTCD (Table 5B). Strikingly, the sensitivity and specificity increased to 86.7% and 95.6% when glypican-3 was used with FTCD. These results indicate that combination of the three markers greatly improves the diagnostic accuracy of HCC.

It has recently been recommended to perform a biopsy to identify the features of malignancy when small hepatic masses are detected. As a result, a distinction among regenerative, dysplastic, and malignant hepatocellular nodules is needed on liver biopsy specimens. Therefore, we tested whether we can distinguish eHCC from benign tumors such as dysplastic and regenerative nodules. A total of 18 eHCC tissues and 10 benign tumor tissues (five FNH, three LRN, and two adenomas) were immunostained with CHC, FTCD, and glypican-3 antibodies (Table 6). Note that high-grade dysplastic nodules were included in eHCC because they have been considered as premalignant or malignant lesions by abnormally increased arteriolar and capillary supply.18 In contrast, low-grade dysplastic nodules were included in benign tumor. Seven eHCCs were distinguished from adjacent nontumor tissues by stronger staining of CHC, whereas one of FNH and LRN was weakly stained with CHC antibody (Fig. 5A, Table 6). Eight eHCCs showed weaker staining of FTCD than adjacent nontumor tissues (Fig. 5B, Table 6). In contrast, all of the FNH and LRN tissues were moderately stained, which is indistinguishable from their adjacent nontumor tissues. Two adenoma tissues showed weaker staining of FTCD than nontumor tissues. Six eHCCs and none of the benign tumors showed stronger staining of glypican-3 than adjacent nontumor tissues. The sensitivity and specificity of CHC, FTCD, and glypican-3 individually for detection of early HCC was 41.2% and 77.8% for CHC, 44.4% and 80.0% for FTCD, and 33.3% and 100% for glypican-3 (Table 6). The sensitivity of CHC or FTCD was better than that of glypican-3. Moreover, the sensitivity significantly increased by combination of these markers, 72.2% for CHC + FTCD, 61.1% for CHC + glypican-3 and FTCD + glypican-3. This is because 44% of glypican-3–negative eHCCs were able to be detected by either CHC or FTCD staining. These results support that CHC and FTCD are potential biomarkers for early detection of HCC.

Table 6. Immunohistochemical Analysis of CHC, FTCD, and Glypican 3 in Early HCC Tissues
 Early HCCBenign TumorSensitivity (%)Specificity (%)
T > N or T < NT = NT > N or T < NT = N
  1. T, tumor tissues; N, nontumor tissues.

Figure 5.

Immunohistochemical analyses of differential protein expression in early HCC tissues. Early HCC tissues were stained with hematoxylin-eosin (upper panel) or with anti-CHC (clathrin heavy chain)/FTCD (formiminotransferase cyclodeaminase) antibody (lower panel). Arrows indicate borders between early HCC and nontumor tissue. (A) Early HCCs were distinguished from adjacent nontumor tissues by stronger staining of CHC. (B) Early HCCs showed weaker staining of FTCD than adjacent nontumor tissues.


In this study, we compared protein expressions between HCC and adjacent nontumor tissues using a proteome method. A total of 83 proteins with altered expression were identified. Validation of the differentially expressed protein by immunoblot or immunostaining demonstrates that CHC, Ku86, FTCD, rhodanese, and vinculin showed striking differences between tumor and nontumor tissues. Evaluation of the staining intensity of CHC and FTCD enabled us not only to distinguish nontumor and tumor tissues with high accuracy but to discriminate eHCC and benign tumors such as dysplastic and regenerative nodules, which is challenging for expert pathologists. Moreover, CHC and FTCD were able to detect several glypican-3–negative eHCCs, which considerably improved the diagnostic accuracy of eHCC by combination of these markers.

In recent years, the incidence of HCC has been increasing in a number of countries, including Europe and the United States.19 As a result, considerable emphasis is now placed on the surveillance of HCC. Recent guidelines for HCC management recommend the combined use of alpha-fetoprotein and ultrasonography for HCC surveillance.7 When small hepatic masses of 1 to 2 cm within a cirrhotic liver are detected, these lesions should undergo biopsy if they do not exhibit typical radiological features of HCC. Accordingly, a distinction between benign and malignant tumor is demanded for pathologists in small biopsies, and further immunohistochemical markers with sufficient sensitivity and specificity are desired. Some markers that can distinguish HCC from dysplastic nodules in cirrhosis have recently been reported.17 The diagnostic yield of three putative HCC markers, HSP70, glypican 3, glutamine synthetase, was investigated; these were previously proposed by other researchers as promising markers for HCC. However, we identified two novel proteins, CHC and FTCD, by comprehensive proteome analysis, and they were found to be useful for the pathological diagnosis of HCC. Diagnostic values, such as the sensitivity and specificity of proteins for HCC, are comparable to glypican-3 in our analyses. More importantly, the sensitivity and specificity significantly increased when immunostaining of glypican-3 was used with that of CHC and FTCD. Thus, a combination of these markers is useful for screening of HCC.

Overexpression of CHC in HCC was confirmed by immunoblotting, and most HCC showed strong and scattered staining in the cytoplasm and plasma membranes. Although CHC overexpression has not been reported in any other primary human cancers, fusion of the CHC gene to other genes, such as ALK and TFE3, has been documented in large B-cell lymphoma, pediatric renal adenocarcinoma, and inflammatory myofibroblastic tumor.20–24 These results indicate that deregulated expression of CHC might play important roles for tumorigenesis. CHC is known to be localized in the plasma membrane and the cytoplasmic face of intracellular organelles in the plasma membrane, called coated vesicles and coated pits. These specialized organelles are involved in the intracellular trafficking of receptors and endocytosis of a variety of macromolecules.25 Recently, Royle et al.26 showed that clathrin stabilizes fibers of the mitotic spindle to aid the congression of chromosomes. Because deregulation of mitotic processes leads to chromosomal instability, known as marker of cancer, the importance of clathrin in normal mitosis may be relevant to understanding human cancers. We have previously shown that kinetochore proteins, CENP-A and CENP-H, are up-regulated in human primary colon cancer, and their overexpression induces aneuploidy.14, 27 Similarly, the up-regulation of CHC observed in this study might cause chromosome missegregation and lead to HCC development.

FTCD showed strong uniform staining in most nontumor tissue, whereas weak staining was observed in HCC. Interestingly, the intensity of FTCD staining in well-differentiated HCC tissues was more likely to be stronger than that in poorly differentiated HCC tissues, suggesting that the expression of FTCD might be involved in the dedifferentiation of tumor cells. FTCD was previously identified as a 58-kDa rat liver protein with the cytoplasmic surface of the Golgi apparatus in vivo.28 It is considered that FTCD is a liver-specific enzyme that controls folic acid metabolism.29 Although FTCD has also been recognized as a liver-specific antigen recognized by the sera of patients with autoimmune hepatitis,30 its involvement in carcinogenesis has not been reported. Thus, our observation is the first report that suggests that down-regulation of FTCD participates in liver carcinogenesis. Further studies are needed to elucidate the precise mechanism of FTCD down-regulation in HCC.

Ku86 is a DNA end-binding molecule that plays an important role in the process of DNA damage signaling and repair, which is thought to maintain genomic stability. Mice deficient in Ku86 showed a marked increase in chromosomal aberrations, and the loss of p53 and Ku86 promotes tumorigenesis, which suggests that Ku86 suppresses tumor development by maintaining the integrity of the genome.31 Furthermore, recent observation showed that haplo-insufficiency of Ku80 (=Ku86) in poly(ADP-ribose) polymerase-1 (PARP-1)−/− mice promotes the development of hepatocellular adenoma and hepatocellular carcinoma,32 suggesting that down-regulation of Ku80 is also important for liver carcinogenesis. In contrast, there are some examples in which the up-regulation of Ku86 is associated with tumor progression. Increased expression of Ku70 and Ku86 in a COX-2–dependent mechanism might be associated with hyperproliferation of gastric cancer cells.33 In addition, increased expression of Ku86 has been reported in B-cell chronic lymphocytic leukemia and in aggressive breast tumors.34, 35 More precise work is needed to examine the expression level of Ku86 in various tumors and to test whether overexpression of Ku86 is a cause or consequence of tumorigenesis.

Rhodanese (EC was originally identified as a mitochondrial matrix enzyme and was proposed to play a role in cyanide detoxification.36 Recently, it was demonstrated that H2S is a potent toxin normally present in the colonic lumen, which may play a role in ulcerative colitis, and rhodanese is the principal enzyme involved in H2S detoxification.37 In fact, the expression of rhodanese was focally lost in ulcerative colitis.38 Moreover, rhodanese was markedly reduced in advanced colon cancer.38 Given that chronic inflammation is an important underlying condition for tumor development,39 anti-inflammatory protein such as rhodanese might prevent tumor progression. Recent data have also expanded the concept that inflammation is a critical component of carcinogenesis. In this regard, down-regulation of rhodanese might be a cause of HCC development and could be a potential target for cancer therapy.

Vinculin has a crucial role in the maintenance and regulation of cell adhesion and migration. On recruitment to cell–cell and cell–matrix adherens-type junctions, vinculin becomes activated and mediates various protein–protein interactions that regulate the links between F-actin and the cadherin and integrin families of cell adhesion molecules.40 Because the loss of cell–cell and cell–matrix interaction is crucial for the development of tumors, down-regulation of vinculin might contribute to carcinogenesis. In fact, the expression of vinculin was repressed in lung carcinoma in surfactant protein C (SP-C)/c-raf transgenic mice.41 Overexpression of vinculin suppresses tumorigenicity in transformed cells,42 whereas cancer cells lacking vinculin enhance cell motility and are highly metastatic.43, 44 Our finding that vinculin was repressed in HCC further supports its tumor suppressor function. Interestingly, although full-length vinculin is 117 kDa, smaller molecular weight protein (the major one being 60 kDa) was observed and down-regulated in nontumor tissues. Several reports have shown proteolytic cleavage of vinculin. For example, vinculin is proteolyzed by calpain into at least three fragments (105, 95, 85 kDa45) during platelet aggregation. Conversely, alpha-actinin–vinculin interactions causes the conformational change of vinculin and generate an approximately 60-kDa fragment of vinculin by papain treatment46; therefore, it is necessary to confirm whether these low molecular proteins are degradation products of vinculin.

In summary, we identified several proteins that are useful to confirm the diagnosis of HCC. They could make significant contributions to the diagnosis of HCC and might also be potential therapeutic targets for cancer control and prevention. Further investigation is needed to uncover the mechanisms responsible for altered protein expressions in HCC.