Proteomic profiling reveals the prognostic value of adenomatous polyposis coli–end-binding protein 1 in hepatocellular carcinoma


  • Potential conflict of interest: Nothing to report.


Histological differentiation is a major pathological parameter associated with poor prognosis in patients with hepatocellular carcinoma (HCC) and the molecular signature underlying HCC differentiation may involve key proteins potentially affecting the malignant characters of HCC. To develop prognostic biomarkers for HCC, we examined the global protein expression profiles of 45 surgically resected tissues, including 27 HCCs with different degree of histological differentiation, 11 adjacent nontumor tissues, and seven normal liver tissues. Unsupervised classification grouped the 45 samples according to their histological classification based on the protein expression profiles created by laser microdissection and two-dimensional difference gel electrophoresis (2D-DIGE). Statistical analysis and mass spectrometry identified 26 proteins with differential expression, of which 14 were functionally linked to c-Myc, AP-1, HIF1A, hepatocyte nuclear factor 4 alpha, or the Ras superfamily (RhoA, CDC42, and Rac1). Among the proteins identified, we focused on APC-binding protein EB1 (EB1) because it was dominantly expressed in poorly differentiated HCCs, which generally correlate with the poor prognosis in patients with HCC. In addition, EB1 is controlled by c-Myc, RhoA, and CDC42, which have all been linked to HCC malignancy. Immunohistochemistry in a further 145 HCC cases revealed that EB1 significantly correlated with the degree of histological differentiation (P < 0.001), and univariate and multivariate analyses indicated that EB1 is an independent prognostic factor for recurrence (hazard ratio, 2.740; 95% confidence interval, 1.771–4.239; P < 0.001) and survival (hazard ratio, 2.256; 95% confidence interval, 1.337–3.807; P = 0.002) of patients with HCC after curative surgery. Conclusion: Proteomic profiling revealed the molecular signature behind the progression of HCC, and the prognostic value of EB1 in HCC. (HEPATOLOGY 2008;48:1851-1863.)

Hepatocellular carcinoma (HCC) is one of the most common and aggressive malignancies world-wide and is the third leading cause of cancer death.1 HCC is a major health problem with high prevalence in Asia and Africa,2, 3 and recent studies indicated that the incidence of HCC has increased substantially in the United States and the United Kingdom over the last decades.4, 5 The prognosis for patients with HCC presently remains dismal, and novel diagnostic and therapeutic modalities, or improvement of existing therapeutic strategies, have long been required to improve the clinical outcome of patients with HCC.

Histological differentiation is a hallmark of malignant potential of HCC; patients with poorly-differentiated tumors tend to have worse prognosis than those with well-differentiated tumors.6 Therefore, the molecular background of histological differentiation may involve prognostic biomarker candidates, which may lead to novel diagnostic and therapeutic modalities. However, although several factors regulating histological differentiation have been reported,7, 8 proteins that underlie HCC differentiation and correlate with HCC prognosis are presently unclear.

Recently, the advent of novel technologies linked with the Human Genome Database enabled global protein expression studies, namely proteomics.9 The proteome is the functional translation of the genome, directly regulating cancer behavior, and is thus a rich source for identifying biomarkers and therapeutic targets. Proteomic studies have identified the proteins whose expression correlates with early recurrence of HCC.10, 11 The proteins implicated in early recurrence may have a clinical utility in predicting poor prognosis. Recently, we identified the proteins associated with histological differentiation in esophageal cancer using two-dimensional difference gel electrophoresis (2D-DIGE).12 As the identified proteins included those associated with malignant attributes of tumor cells such as lymph node metastasis, this approach is also worth being applied to the study of HCC.

We conducted a proteomic study on HCC tissues with varying degrees of histological differentiation, as well as adjacent nontumor tissues and normal liver tissues, and captured a molecular signature that underlies HCC differentiation and affects the malignant potential of HCC. We found that expression of APC-binding protein EB1 (EB1) was specific to moderately-differentiated and poorly-differentiated HCCs, and revealed the prognostic value of EB1 expression, employing immunohistochemistry on additional HCC cases.


2D-DIGE, two-dimensional difference gel electrophoresis; AP-1, activator protein 1; APC, adenomatous polyposis coli; C/EBP beta, CCAAT/enhancer-binding protein, beta subunit; CSA, catalyzed signal amplification; Cy, cyanine; EB1, APC-binding protein EB1; HCC, hepatocellular carcinoma; HIF1A, hypoxia-inducible factor 1, alpha subunit; mDia2, mammalian Diaphanous-related formin; PAGE, polyacrylamide gel electrophoresis; Rac, ras-related C3 botulinum toxin substrate, Ras, ras sarcoma oncoprotein; RhoA, ras homology gene family, member A; SDS, sodium dodecyl sulfate; TNM, tumor-node-metastasis; XML, extensible markup language.

Materials and Methods

Detailed procedures are available in the Supporting Information.

Patients and Tissue Samples.

A total of 45 surgically resected tissues were included in this study. The tissue samples were divided into five groups according to their histological classification: seven normal liver tissues, 11 nontumor tissues adjacent to tumors, six well-differentiated HCCs, 14 moderately-differentiated HCCs, and seven poorly-differentiated HCCs (Supporting Table 1). The tissues were obtained at the National Cancer Center Hospital; the HCC and adjacent nontumor tissues were from patients with HCC who underwent initial hepatic resection between June 2005 and April 2006, and the normal liver tissues were from patients who underwent hepatic resection for metastatic liver tumor from colorectal cancer in the same period. Different histological areas of HCC were obtained from identical tumor tissues of three cases whose tumors showed histological heterogeneity. The clinicopathological features of the patients are listed in Supporting Table 1. For the EB1 expression study, we examined an additional 145 patients with HCC who underwent initial surgical resection between February 1992 and December 2000 at the National Cancer Center Hospital. None of the patients of this study received any preoperative therapy. Tumors were classified according to the World Health Organization classification13 and the International Union against Cancer tumor-node-metastasis (TNM) classification.14 The ethical review board of the National Cancer Center approved this project.

Laser Microdissection.

Specific populations of cells were recovered by laser microdissection according to our previous reports15, 16 (Fig. 1A). In brief, 10-μm-thick frozen sections were created from tumor tissues and stained with hematoxylin. The cells were recovered under microscopic observation with the assistance of a ultraviolet laser (MMI CellCut; Molecular Machines & Industries, Glattbrugg, Switzerland). A 1-mm2 tissue area (∼3,000 cells) was recovered for one gel. The recovered cells were lysed in urea lysis buffer containing 6 M urea, 2 M thiourea, 3% {3-[(3-cholamidopropyl)dimethylammonio]-1-propanesulfonate}, and 1% Triton X-100, and were stored at −80°C until use.

Figure 1.

Protein expression profiling using laser microdissection and 2D-DIGE with high sensitive fluorescent dyes. (A) Specific populations of cells were recovered using laser under microscopic observation. (B) The extracted proteins were labeled with fluorescent dyes and separated by two-dimensional polyacrylamide gel electrophoresis (2D-PAGE). (C) Evaluation of the reproducibility of 2D-DIGE by scatter graphs.

2D-DIGE and Image Analysis.

The 2D-DIGE was performed as described.15, 16 In brief, a common internal control sample was created by mixing a small portion of all protein samples used in this study, and labeled with cyanine 3 (Cy3) fluorescent dye (CyDye DIGE Fluor saturation dye; GE Healthcare Biosciences, Uppsala, Sweden). Individual samples were labeled with cyanine 5 (Cy5) fluorescent dye (CyDye DIGE Fluor saturation dye; GE Healthcare Biosciences). These differently-labeled protein samples were mixed together and separated according to their isoelectric point and molecular weight. The first dimension separation was achieved using a 24-cm-length immobiline gel (IPG, pI 4-7; GE Healthcare Biosciences) and Multiphor II (GE Healthcare Biosciences), while the second-dimension separation used a homemade gradient gel with GiantGelRunner (Biocraft, Tokyo, Japan), with a separation distance of 36 cm. The gels were scanned using a laser scanner (Typhoon Trio; GE Healthcare Biosciences) at the appropriate wavelength for Cy3 or Cy5. For all protein spots, the Cy5 intensity was normalized with the Cy3 intensity in the same gel using the DeCyder software (version 5.0; GE Healthcare Biosciences), so that gel-to-gel variations were canceled out (Fig. 1B). We monitored the reproducibility of our system by running the same sample twice (case 24; Supporting Table 1). The scatter-plot demonstrated that the intensity value of 98% of protein spots was scattered within a two-fold difference, and that the correlation coefficient was 0.9017, showing the high reproducibility of the profiling method used (Fig. 1C). The spot intensity data were exported from the DeCyder software as extensible markup language (XML) format files, which are amenable to data analysis.

Data Analysis.

The numerical data in the XML files were imported to Expressionist software (GeneData, Basel, Switzerland) for scatter-plotting, hierarchical clustering, and principal component analysis. The Kruskal-Wallis test and Bonferroni adjustment were used to identify the protein spots that were differentially expressed in the five tissue groups examined.

Mass Spectrometric Protein Identification.

The proteins corresponding to the protein spots were identified by mass spectrometry according to our report.12 Cy5-labeled proteins separated by 2D-polyacrylamide gel electrophoresis (PAGE) were recovered in gel plugs and digested with modified trypsin (Promega, Madison, WI). The trypsin digests were subjected to liquid chromatography, coupled with tandem mass spectrometry equipped with a nanoelectrospray ion source (Paradigm MS4 dual solvent delivery system; Michrom BioResources Inc., Auburn, CA) for microflow high-performance liquid chromatography (HPLC), an HTS PAL auto sampler (CTC Analytics, Zwingen, Switzerland), and a Finnigan LTQ linear ion trap mass spectrometer (ThermoElectron Co., San Jose, CA) equipped with a nanoelectrospray ion source (AMR Inc., Tokyo, Japan). The Mascot software (version 2.1; Matrix Science, London, UK) was used to search for the mass of the peptide ion peaks against the SWISS-PROT database (Homo sapiens; 12867 sequence in Sprot_47.8 fasta file). Proteins with a Mascot score of 35 or more were used for protein identification. When multiple proteins were identified in a single spot, the proteins with the highest number of peptides were considered as those corresponding to the spot.

Western Blotting.

Protein samples were separated by sodium dodecyl sulfate (SDS)-PAGE and subsequently blotted on a nitrocellulose membrane. Immunoblot analysis was performed using the antibodies EB1 (1:200; Santa Cruz Biotechnology, Santa Cruz, CA), proliferating cell nuclear antigen (1:5,000; BD Transduction Laboratories, San Jose, CA), heat shock protein 90 (1;1,000; BD Transduction Laboratories), arginase-I (1;1,000; BD Transduction Laboratories), actin (1;2;000; Abcam, Cambridge, UK), horseradish peroxidase–conjugated secondary antibodies (1:1,000; GE Healthcare Biosciences), and enhanced chemiluminescence (ECL; GE Healthcare Biosciences).


Immunohistochemical staining for EB1 was performed on formalin-fixed, paraffin-embedded tissue sections using the CSA II system (DAKO, Glostrup, Denmark) following the manufacturer's instructions. For antigen retrieval, the sections were autoclaved in 10 mM citrate buffer (pH 6.0) at 121°C for 10 minutes. We used rabbit anti-EB1 polyclonal antibodies (sc-15347; Santa Cruz Biotechnology) at a dilution of 1:500. Staining was assessed by two independent observers in a blinded fashion for clinical data. The bile duct epithelium served as an internal control of positive staining. If more than 50% of tumor cells were positively stained, the tumor was judged as EB1-positive. Staining evaluation was done at the dominant differentiation area of the tumor if the tumor had areas with varying degrees of differentiation.

Pathway Analysis of Expression Data.

The pathway analysis of the protein expression pattern was performed using the MetaCore software (GeneGo Inc., St. Joseph, MI). MetaCore identifies networks based on a manually curated database containing known molecular interactions, functions, and disease interrelationships, using proteome data sets. The pathways were identified by the probability that a random set of proteins the same size as the input list would give rise to a particular mapping by chance.

Statistical Analysis.

The correlation between EB1 expression and clinicopathological features was evaluated by the Fisher exact test for categorical variables and the Mann-Whitney U test for continuous variables. The time to recurrence and overall survival were calculated from the first resection of the primary tumor to the first radiological evidence of recurrence or to death, respectively. All time-to-event end points were computed by the Kaplan-Meier method.17 Patients dying without recurrence were censored in determining recurrence. Potential prognostic factors were identified by univariate analysis using the log-rank test. Independent prognostic factors were evaluated using a Cox proportional hazards regression model and a stepwise selection procedure. P differences <0.05 were considered to be significant. Statistical analyses were performed using the SPSS statistical package (SPSS, Chicago, IL).


Proteomic Profiling of HCC.

To examine the overall features of the proteome, we performed unsupervised classification using the intensity values of 3,319 protein spots that were observed in more than 80% of the protein expression profiles of the common internal control sample. Results of hierarchical clustering were associated with histological grouping: the seven normal liver tissues, 11 adjacent nontumor tissues, six well-differentiated HCCs, and one moderately-differentiated HCC were grouped together, while the 13 moderately-differentiated HCCs and seven poorly-differentiated HCCs were clustered together forming a separate group (Fig. 2A). Principal component analysis also showed similar results; normal and adjacent nontumor tissues were grouped together while well-differentiated tumors were segregated from the group of moderately-differentiated or poorly-differentiated tumors (Fig. 2B). These observations suggested that the overall features of the proteome may reflect the major histological patterns.

Figure 2.

Proteomic classification of tissue samples and identification of proteins with different expression levels between the sample groups. (A) Hierarchical clustering and (B) principal component analysis grouped the tissue samples based on the intensity of 3319 protein spots. (C) Heat map of the 41 selected protein spots is shown. The results of protein identification are demonstrated in the right side of (C).

To identify the proteins that are differentially expressed in the five tissue groups examined, we performed a Kruskal-Wallis test and applied Bonferroni adjustment. We selected the protein spots such that the Bonferroni adjusted P value was <0.01 and the expression ratio between groups with the greatest difference was at least three times or more. Consequently, we found 41 protein spots meeting this criterion (Supporting Fig. 1). The expression pattern of these selected 41 protein spots in all tissue samples is shown in Fig. 2C. Using hierarchical clustering, we found that the protein spots were subdivided into clusters A and B, based on whether their intensity was upregulated or downregulated in the group of moderately-differentiated HCCs and poorly-differentiated HCCs (Fig. 2C).

Protein Identification and Network Analysis.

Mass spectrometric study resulted in the identification of 26 unique proteins corresponding to the 41 protein spots (Fig. 2C, right side; Table 1; Supporting Table 2). Functional classification according to Gene Ontology ( demonstrated that a large proportion of the identified proteins are involved in amino acid metabolism, oxidoreduction, and lipid metabolism (Fig. 3A; Table 1). The proteins corresponding to the protein spots in clusters A and B were classified according to their known function (Fig. 3B,C; Table 1). Proteins in clusters A included ones involved in cell proliferation, protein folding, and cytoskeletal/structural proteins. Proteins in cluster B included ones involved in amino acid metabolism, oxidoreduction, and lipid metabolism, all of which maintain normal hepatic functions. Western blotting results were consistent with the 2D-DIGE results, validating the differential expression of the identified proteins (Supporting Fig. 2A,B).

Table 1. A List of Identified Proteins
Accession No.*Identified Protein*LocusSpot No.P Value*pI (cal)MW (cal) (D)Protein ScorePeptide MatchesSequence Coverage (%)
  • *

    Proteins in Supporting Fig. 3 are shown as bold.

  • *

    Accession numbers of proteins and protein name were derived from Swiss-Prot and NCBI nonredundant databases.

  • Protein function was categorized by accessing Gene Ontology database ( and literature curation.

  • Gene locus was determined according to NCBI database.

  • ‡Spot numbers refer to those in Figure 2C and Supplemental Fig. 1.

  • §Bonferroni adjusted P value.

  • Theoretical isoelectric point and molecular weight obtained from Swiss-Prot and the ExPASy database (

  • Mascot score for the identified proteins based on the peptide ions score (P < 0.05) (

  • **

    Cluster A and B are shown in Fig. 2C.

Proteins in cluster A**         
Cell proliferation         
 P12004Proliferating cell nuclear antigen20pter-p1227941.93E-034.572909294210
 Q15691APC-binding protein EB120q11.1-q11.2333607.64E-035.0230020189420.6
 P51858Hepatoma-derived growth factorIq21-q2329384.54E-034.726886200421.7
Protein folding         
 P14625Heat shock protein 90 kDa beta member 112q24.2–q24.316962.89E-034.769269626756.5
 Q9UHV9Prefoldin subunit 21q23.347305.86E-036.216695186322.7
Cytoskeletal/structural protein         
Signal transduction         
 P06702Protein S100-A91q2146805.02E-035.711329181226.3
 P61923Customer subunit zeta-112q13.2–q13.345308.44E-034.6920242117211.9
Proteins in cluster B**         
Amino acid metabolism         
 Q00266S-pdenosylmethionlne synthetase isoform type-110q2220041.26E-035.864419012428.4
 Q14749Glycine N-methyltransferase6p1226967.85E-036.5833046158314.3
 P327544-hydroxyphenylpyruvate dioxygenase12q24-qter23767.33E-036.5449465391025.5
 NP_036335Glyoxylate reductase/hydroxypyruvate reductase9q1225209.29E-037.0136045312619.5
 O95154Aflatoxin B1 aldehyde reductase member 31p35.1-p36.2328845.26E-036.6737582412623.6
 P78417Glutathione transferase omega-110q25.138242.73E-036.2327833202418.3
Lipid metabolism         
 P54868Hydroxymethylglutaryl-CoA synthase, mitochondrial precursor1p13-p1217091.60E-038.45711317436.5
 P45954Short/branched chain specific acyl-CoA dehydrogenase, mitochondrial precursor10q26.1324288.62E-036.5347797469821.8
 P16219Short-chain specific acyl-CoA dehydrogenase, mitochondrial precursor12q22-qter25261.01E-038.13446116831635.7
 P30084Enoyl-CoA hydratase, mitochondrial precursor10q26.2-q26.332947.45E-038.3431823242316.9
 P05062Fructose-bisphosphate aldolase B9q21.3-q22.221875.79E-0383996119539.6
 P09467Fructose-1,6-bisphosphatase 19q22.326618.97E-036.6370594861136.5
Signal transduction         
 P52566Rho GDP-dissociation Inhibitor 212p12.340461.49E-035.122900136215
 P30039MAWD-binding protein10pter-q25.332845.68E-036.063205011828.7
Figure 3.

Functional classification and network analysis of the identified proteins. (A) Functional classification of all proteins, (B) proteins in cluster (A), and (C) proteins in cluster (B). Clusters (A) and (B) are shown in Fig. 2. (D) EB1 is controlled by c-Myc, RhoA, and CDC42, which have all been linked to HCC malignancy. The differently colored nodes represent transcription factors (red shape), Ras-superfamilies (light blue shapes), and other proteins (blue shapes). The green line indicates a positive effect and the gray indicates an unspecified effect.

We explored the biological significance of the altered protein expression patterns by classifying the associated proteins within the context of functional pathways and networks using MetaCore, and we found that 14 of the 26 identified proteins could be functionally linked (Supporting Fig. 3). The transcription factors in this network included c-Myc, hepatocyte nuclear factor 4 alpha, AP-1, HIF1A, and C/EBP beta, which connected five, five, three, two, and one genes, respectively. Other proteins responsible for the regulation of multiple proteins included Ras superfamily proteins such as RhoA, CDC42, and Rac1.

Clinical Significances of EB1 Expression in HCC.

Among the identified proteins, EB1 is controlled by c-Myc, RhoA and CDC42, which have all been linked to HCC malignancy in previous reports18–21 (Fig. 3D). For this reason, we further examined the relationship of EB1 with certain clinicopathological parameters in an additional 145 HCC cases that were not included in the proteomic study, employing immunohistochemistry. Immunohistochemical staining for EB1 was observed in the cytoplasm of tumor cells, inflammatory cells, and bile duct epithelium, while hepatocytes in nontumor areas showed no immunostaining (Fig. 4). The EB1-positive and EB1-negative tumor tissues had significantly different histological differentiation, alpha-fetoprotein expression, TNM stage, tumor size, portal vein invasion status, and intrahepatic metastasis status (P < 0.01; Table 2). As these parameters have been correlated with the clinical outcome of patients with HCC, we further investigated the correlation of EB1 with prognostic data. As shown in the Kaplan-Meier survival curve (Fig. 5), patients with EB1-positive HCC tumors had significantly worse prognosis than those with EB1-negative HCC tumors, in terms of both overall survival rate (P < 0.0001) and cumulative recurrence rate (P < 0.0001). Univariate and multivariate analyses revealed that EB1 is an independent prognostic factor for overall survival (hazard ratio, 2.256; 95% confidence interval, 1.337-3.807; P = 0.002) and recurrence (hazard ratio, 2.740; 95% confidence interval, 1.771-4.239; P < 0.001) along with other established clinicopathological parameters such as liver cirrhosis, portal vein invasion, tumor number, and intrahepatic metastasis (Table 3).

Figure 4.

Expression of EB1 in HCC tissues with different histological differentiation. (A,B) Well-differentiated, (C,D) moderately-differentiated, (E,F) and poorly-differentiated HCCs were examined. (A,C,E) Hematoxylin and eosin–stained tissues; (B,D,F) tissues stained with anti-EB1 antibody. Note that EB1 expression correlated with the degree of histological differentiation. Nontumor liver and HCC are indicated by (N) and (T), respectively.

Table 2. Correlations Between Clinicopathological Features and EB1 Expression
VariableEB1 Positive (Number of Cases)EB1 Negative (Number of Cases)Correlation (EB1) P Value*
  • Bold indicates significant values. HBV, hepatitis B virus; HCV, hepatitis C virus; AFP, alpha-fetoprotein

  • *

    Fisher exact test for categorical variables and Mann-Whitney U test for continuous variables.

  • Expressed as median (range).

Age61 (48–80)65 (26–83)0.161
Gender  0.825
Virus infection status  0.302
Child-Pugh Classification  1.000
Liver cirrhosis  0.502
AFP (ng/mL)  <0.001
 343.5 (3–27170)20.3 (1–9994) 
TNM Stage  <0.001
 I or II2089 
 III or IV2016 
Tumor number  1.000
Tumor size (mm)  0.008
 45 (13–155)30 (6–185) 
Differentiation  <0.001
 Well differentiated024 
 Moderately differentiated973 
 Poorly differentiated318 
Portal vein invasion  <0.001
Intrahepatic metastasis  <0.001
Figure 5.

Correlation of EB1 expression with clinical outcome of HCC patients after curative resection. The patients with HCC tissues showing EB1 expression had poorer prognosis in terms of (A) overall survival and (B) overall recurrence.

Table 3. Univariate and Multivariate Analyses of Prognostic Factors for Patients with HCC
VariablenUnivariate Analysis
5-Years (%)P Value5-Years (%)P Value
  1. Bold indicates significant values. *Two groups were divided by the median. HBV, hepatitis B virus; HCV, hepatitis C virus; AFP, alpha-fetoprotein; Beta, regression coefficient; SE, standard error; HR, hazard ratio; CI, confidence interval.

Age*  0.7909 0.6517
 <647653.8 ± 6.2 77.8 ± 5.2 
 ≥646948.8 ± 6.3 73.7 ± 5.5 
Gender  0.9721 0.9862
 Female3351.6 ± 9.2 78.1 ± 7.5 
 Male11251.3 ± 5.1 75.2 ± 4.4 
Virus infection status  0.3794 0.0827
 HBV4653.1 ± 8.0 60.0 ± 7.8 
 HCV7754.3 ± 6.1 81.1 ± 4.6 
 Both1136.4 ± 14.5 87.9 ± 11.0 
 None1140.0 ± 15.5 81.8 ± 11.6 
Child-Pugh Classification  0.0407 0.1315
 A13452.9 ± 4.6 74.7 ± 4.0 
 B1134.1 ± 15.0 100.0 ± 0.0 
Liver cirrhosis  0.0277 0.6586
 Absence10957.6 ± 5.1 74.6 ± 4.4 
 Presence3635.2 ± 8.1 79.0 ± 7.4 
AFP*  0.0532 0.0210
 <27.27060.4 ± 6.2 67.1 ± 6.1 
 ≥27.27341.4 ± 6.2 85.8 ± 4.3 
TNM Stage  <0.0001 <0.0001
 I or II10961.9 ± 5.0 69.6 ± 4.7 
 III or IV3616.2 ± 7.1 94.3 ± 3.9 
Tumor number  0.0856 0.0238
 Single11255.9 ± 5.0 70.9 ± 4.6 
 Multiple3336.8 ± 9.0 90.9 ± 5.0 
 Tumor size (mm)*  0.0627 0.0178
 <356662.1 ± 6.3 72.9 ± 5.8 
 ≥357942.5 ± 6.0 78.2 ± 4.9 
Differentiation  <0.0001 <0.0001
 Well differentiated2473.9 ± 9.2 76.3 ± 9.2 
 Moderately differentiated8259.6 ± 5.8 66.2 ± 5.6 
 Poorly differentiated3916.3 ± 7.3 95.9 ± 3.8 
Portal vein invasion  <0.0001 0.0001
 Absence8767.7 ± 5.3 68.4 ± 5.3 
 Presence5824.7 ± 6.4 86.6 ± 4.8 
Intrahepatic metastasis  <0.0001 <0.0001
 Absence11260.7 ± 4.9 70.6 ± 4.6 
 Presence3319.0 ± 7.4 93.7 ± 4.3 
EB1 expression  <0.0001 <0.0001
 Negative10562.1 ± 5.0 69.3 ± 4.8 
 Positive4022.0 ± 7.4 92.2 ± 4.7 
 Multivariate Analysis
BetaSEHR95% CIP value
 Liver cirrhosis0.9220.2582.5151.517–4.170<0.001
 Portal vein invasion0.6700.3011.9551.083–3.5280.026
 Intrahepatic metastasis0.6990.2892.0121.143–3.5440.015
 EB1 expression0.8140.2672.2561.337–3.8070.002
 Tumor number0.5940.2131.8111.192–2.7520.005
 Intrahepatic metastasis1.0300.2272.8021.795–4.375<0.001
 EB1 expression1.0080.2232.7401.771–4.239<0.001


The need for improvement of the management of HCC has led to a strong demand for the development of novel prognostic biomarkers for HCC. Global genomic and transcriptomic expression studies have been conducted to detect such prognostic molecular biomarkers for HCC. For example, using array-based comparative genomic hybridization analysis, chromosomal loss on 17p13.3 and gain on 8q11 were shown to have significant effects on patient outcome.22 Using complementary DNA microarray technology, osteopontin was identified as a critical player in HCC metastasis,23 and AP-1 transcription factors were shown to have key roles in the development of a poor-prognosis subtype of HCC.24 In contrast, few such studies have been performed using a proteomic approach.11 Here we present the results of such a proteomic study, and propose EB1 as a prognostic biomarker for HCC.

In this study, we examined HCC tissues classified according to their histological differentiation. As the degree of histological differentiation is a hallmark of the malignant potential of HCC, the proteomic background of HCC differentiation may involve key proteins for HCC progression. Unsupervised classification of tissues based on their protein expression profiles without any a priori assumptions was associated with their histological presentation, indicating that the overall features of the proteome may reflect the major histological differences between tissues. We subsequently identified 26 proteins that showed the most variable expression between the groups with different histology.

Functional classification demonstrated that proteins associated with cell proliferation, protein folding, and cytoskeletal structure were increased, and proteins associated with amino acid metabolism, oxidoreduction, and lipid metabolism were reduced during HCC progression. These findings suggest that the proteins with increased expression during pathological progression have a different functional tendency compared to those with reduced expression. The different proteomic aberrations observed in the varying stages of cancer progression, as reflected in the different histology groups, are unlikely to represent random events. In contrast to the functional classification of the proteins, the chromosomal localization of the genes corresponding to the identified proteins, as identified by searching the National Center for Biotechnology Information database (Table 1), did not have an obvious tendency. These observations suggest that proteomic studies may provide unique information not generated by genomic studies.

Network analysis using literature mining revealed that the identified proteins were functionally linked to certain transcription factors and Ras superfamilies. The transcription factors that linked the identified proteins are known to be frequently activated in HCCs. For instance, c-Myc amplification has been frequently observed in HCCs and is associated with poor prognosis.18, 19 Liu et al.25 have reported that AP-1 is frequently activated at the early stages of HCC. HIF1A is also important in the progression of hepatocarcinogenesis.26 Other proteins responsible for the regulation of multiple proteins have also been reported to be correlated with HCC malignancy. Overexpression of RhoA20 is associated with poor prognosis in HCC, and Rac activation is associated with metastasis of HCC.27 Activation of CDC42 was involved in the metastasis of HCC cells.21 The proteins we identified may be downstream effectors of known key regulators of carcinogenesis and tumor progression. Although correlations between these molecules and HCC progression have been independently and separately reported, the present global protein expression study enabled a panoramic view of the molecular background of the progression of HCC.

Our study showed the prognostic value of EB1 expression in HCC. EB1 has been identified as a protein bound to the APC tumor suppressor gene product.28 In vitro wounding assays revealed that EB1 and APC promote cell migration, stabilizing microtubules in a coordinate manner.29 EB1 inhibits the ability of APC to bind to actin filaments, which may be required for maintenance of cell-cell adhesion.30 Recently, Wang et al.31 reported that in esophageal cancer, overexpression of EB1 promotes cell growth by activating the beta-catenin/T-cell factor pathway, and this pathway is often activated in HCC.32 Taken together, the interaction of EB1 and APC may play a key role in cytoskeleton organization, cell migration and proliferation, and its aberrant regulation could affect the malignant behavior of HCC, probably resulting in poor prognosis for HCC patients.

Network analysis has revealed that c-Myc, RhoA, and CDC42 regulate the expression and function of EB1. The expression of EB1 is controlled by c-Myc33 and c-Myc amplification has prognostic significance in HCC.18 EB1 functions downstream of RhoA and CDC42 by interacting with mDia2,29, 34, 35 and both RhoA and CDC42 have also been linked to HCC malignancy.20, 21 Taken together, we considered that EB1 may also be associated with poor prognosis.

Our findings can be of use in the search for biomarker identification and development. We studied the relationship of EB1 expression with clinical outcome of the HCC patients in additional 145 cases. Univariate and multivariate analyses revealed that EB1 is an independent prognostic factor for both recurrence and survival of HCC patients. Our results indicate that EB1 expression may be used as a novel prognostic biomarker of HCC. The correlation between EB1 expression and prognosis in HCC has not been examined or demonstrated previously.

The expression of EB1 was observed only in moderately or poorly differentiated HCC in this study (Table 2), and the prognostic utility of EB1 expression is therefore limited to patients with these tumors. The overall survival rate of patients with well-differentiated HCC was significantly higher than that of patients with poorly-differentiated or moderately-differentiated tumors (Table 3), a finding that is consistent with the report by Ariizumi et al.,36 in which the 5-year survival rate was 78.1%, 49.0%, and 37.4% for the well-differentiated, moderately-differentiated, and poorly-differentiated tumors, respectively. The evaluation of EB1 expression could therefore provide prognostic information for the subgroups of HCC, which could benefit from a more refined prognostic protocol.

Our list of identified proteins also included gene products reported as prognostic biomarker candidates (Fig. 2C). Expression of both hepatoma-derived growth factor37 and proliferating cell nuclear antigen38 have been correlated with poor differentiation, shorter survival periods, and higher incidence of recurrence in HCC patients. Overexpression of vimentin has been associated with the metastasis in HCC.39 The relation between HCC differentiation and poor prognosis may be explained by these proteins and EB1, and the combined use of these biomarker candidates may improve the diagnosis and prognostic performance.

In conclusion, through the present global protein expression study the molecular background of histological differentiation in HCC was revealed and EB1 was established as a prognostic biomarker for both recurrence and survival. Poor outcomes in HCC are mainly due to postsurgical tumor recurrence, but recent advances in adjuvant therapies have improved survival periods for patients with recurrence.40 The immunohistochemical examination of EB1 expression will help identify patients with high risk for recurrence, and close postoperative follow-up and additional treatment may improve the clinical outcome of these patients. Taken together, our results provide the possibility of novel strategies for HCC management.


The excellent technical support of Yukiko Fujie and Mina Fujishiro with the electrophoresis is greatly appreciated.