Etiology-dependent molecular mechanisms in human hepatocarcinogenesis

Authors


  • Potential conflict of interest: Nothing to report.

Abstract

Hepatocellular carcinoma (HCC) is one of the most common cancers worldwide and is characterized by aggressive tumor behavior coupled with poor prognosis. Various etiologies have been linked to HCC development, most prominently chronic hepatitis B and C virus infections as well as chronic alcohol consumption. In approximately 10% of HCCs, the etiology remains cryptic; however, recent epidemiological data suggest that most of these cryptogenic HCCs develop due to nonalcoholic steatohepatitis. To identify etiology-dependent DNA copy number aberrations and genes relevant to hepatocarcinogenesis, we performed array-based comparative genomic hybridization of 63 HCCs of well-defined etiology and 4 HCC cell lines followed by gene expression profiling and functional analyses of candidate genes. For a 10-megabase chromosome region on 8q24, we observed etiology-dependent copy number gains and MYC overexpression in viral and alcohol-related HCCs, resulting in up-regulation of MYC target genes. Cryptogenic HCCs showed neither 8q24 gains, nor MYC overexpression, nor target gene activation, suggesting that tumors of this etiology develop by way of a distinct MYC-independent pathomechanism. Furthermore, we detected several etiology-independent small chromosome aberrations, including amplification of MDM4 on 1q32.1 and frequent gains of EEF1A2 on 20q13.33. Both genes were overexpressed in approximately half the HCCs examined, and gene silencing reduced cell viability as well as proliferation and increased apoptosis rates in HCC cell lines. Conclusion: Our findings suggest that MDM4 and EEF1A2 act as etiology-independent oncogenes in a significant percentage of HCCs. (HEPATOLOGY 2008.)

Hepatocellular carcinoma (HCC) is the most frequent type of liver cancer and is the third leading cause of cancer-related death worldwide.1 Various etiologies have been linked to HCC development, the most relevant being chronic hepatitis B virus (HBV) and hepatitis C virus (HCV) infections, chronic alcohol consumption, and, in certain geographical areas, aflatoxin B1 food contamination.2 Approximately 10% of HCC patients show no signs of hepatitis virus infection, alcoholic history, or other defined causes, such as genetic hemochromatosis or α1-antitrypsin deficiency. Most of these so-called cryptogenic HCCs evolve from nonalcoholic steatohepatitis (NASH).3 Whereas generation of reactive oxygen species has been suggested to drive hepatocarcinogenesis in HCCs of alcoholic or cryptogenic etiology, viral-associated mechanisms are complex and involve both host and viral factors.

Genomic instability is a common feature of human HCC. Using conventional and array-based comparative genomic hybridization (aCGH), frequent DNA copy number gains at 1q, 6p, 8q, 17q, and 20q and losses at 1p, 4q, 8p, 13q, 16q, and 17p have been identified as reviewed elsewhere.4 Although target genes such as RB1 (13q14) and TP53 (17p13) have been identified and validated, the driving genes for most commonly altered loci remain unknown. Several CGH studies tried to identify chromosome aberrations that correlate specifically with HCC etiology; however, many failed to uncover significant differences in DNA copy number alteration or candidate genes, possibly due to underrepresentation of samples from certain etiologic groups or due to the low resolution of the platforms used.4 Our recent meta-analysis of 785 HCCs identified significant correlations of deletions of 4q, 13q, and 16q with HBV etiology that may contribute to the functional loss of tumor suppressors in HBV-induced HCCs.5 However, because these analyses were based on conventional CGH, their resolution was limited, and many potential chromosome aberrations probably remained undetected.

In this study, we performed aCGH of 67 HCC samples of well-defined and equally distributed etiologies aiming at the identification of etiology-dependent and -independent DNA copy number alterations and gene expression profiling of pooled total RNAs to identify specific target genes. Validated candidate genes were further characterized by knockdown experiments in cell cultures.

Abbreviations

aCGH, array-based comparative genomic hybridization; HCC, hepatocellular carcinoma; HBV, hepatitis B virus; HCV, hepatitis C virus; IRS, immunoreactive score; PCR, polymerase chain reaction; QRT-PCR, quantitative reverse-transcription PCR; RT-PCR, reverse-transcription PCR; SAM, significance analysis of microarrays; SEM, standard error of the mean.

Patients and Methods

Tumor Material and Patient Characteristics.

Sixty-three human HCCs and 4 HCC cell lines [Hep3B (HBV-related), HuH7, PLC/PRF/5 (HBV-related), and HepG2] were analyzed via aCGH. The HCCs included 39 liver resections and 21 explant liver specimens. The median age was 57 years (range, 16-78), and the male/female ratio was 4:1. All diagnoses were confirmed via histological re-evaluation, and use of the samples was approved by the local ethics committees of the Universities of Heidelberg and Bern. Etiology was determined on the basis of serology, histology, and molecular tissue analysis via HBV-nested polymerase chain reaction (PCR) (for primer sequences, see Supplementary Table 3) and HCV reverse-transcription PCR (RT-PCR) as described.6 The HCCs comprised 14 tumors of HCV, 13 of each alcoholic and cryptogenic etiology, 11 HBV-induced HCCs, 3 tumors associated with hemochromatosis, 4 tumors with HCV/HBV coinfection, and 1 tumor associated with α1-antitrypsin deficiency. Four HBx-postive tumors without known chronic HBV infections were excluded from etiology-dependent analyses. In cryptogenic HCCs, diabetes mellitus type II was confirmed in 4 patients and impaired fasting glucose levels were observed in 3 patients.

Cell Lines and Culture Conditions.

HepG2 was cultured in Roswell Park Memorial Institute (RPMI) 1640 medium (PAA, Pasching, Austria) supplemented with 10% fetal bovine serum (Sigma, Deisenhofen, Germany) and 1% penicillin–streptomycin (10 mg/mL, Sigma). HuH7 and PLC/PRF/5 cells were grown in Dulbecco's modified Eagle's medium (PAA) and Hep3B cells in MEM medium (PAA) with the same supplements. All cell lines were cultured at 37°C (5% CO2) and passaged every 3 to 4 days.

Microarray Analyses.

Extraction of high molecular weight DNA and RNA from frozen HCC samples was performed via cesium chloride ultracentrifugation.7 Genomic DNA from blood of healthy donors was isolated using the Blood and Cell Culture Kit (Qiagen, Hilden, Germany). RNA quality and concentration were tested with the Agilent 2100 bioanalyzer (Agilent Technologies, Palo Alto, CA).

aCGH, or matrix CGH,8 was performed as reported previously.9 For gene expression profiling, 2 μg of pooled total RNA were amplified and hybridized as described,10 and data were analyzed using packages of the Bioconductor project11 implemented in the in-house developed ChipYard framework for microarray data analysis (http://www.dkfz.de/genetics/ChipYard/). Data were filtered according to signal/background ratio (>3.0 for aCGH; >2.0 for expression profiling), mean/median spot intensity (<0.3), and replicate standard deviation (<0.25) and normalized by print-tip loess. For gene expression profiling, additional between-array scale normalization was performed.12

To identify regions of similar genomic status within the aCGH data, we applied the segmentation software GLAD.13 Imbalances with log2 ratios of less than −1.0 were scored as putative homozygous deletions, because this threshold corresponds to an average copy number of less than 1, suggesting the presence of at least a subpopulation of cells with homozygous deletions. Gains with log2 ratios higher than 1.0 were scored as amplifications. Significance analysis of microarrays (SAM) was performed for categorical variables implementing a permutation-based Pearson's chi-squared statistic for each BAC clone.14 A false discovery rate of <0.5% was used as level of significance. For SAM, copy number ratios were discretized for each experiment using thresholds of ±2 standard deviations of the balanced clones. Chromosomal mapping information was based on Ensembl (v42). All data are available at the NCBI Gene Expression Omnibus database (http://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?token=hrsflskaayigkro&acc=GSE8351).

Quantitative RT-PCR.

One microgram of total RNA from tumors and normal-liver pool (n = 6) were treated with DNaseI (Invitrogen, Karlsruhe, Germany), reverse transcribed with SuperscriptII (Invitrogen, Karlsruhe, Germany), and analyzed using the 7900RT Fast Real-Time PCR System (Applied Biosystems, Foster City, CA) with Absolute SYBR Green ROX Mix (ABgene, Epsom, UK). Calculation of efficacy, normalization, and relative quantification versus reference genes (GUSB, PMMS1) were done according to published algorithms.15 All primer sequences are listed in Supplementary Table 3. P values were calculated by Student t test with the assumption that the mean of the normalized data distributed around zero.

Western Blots.

Protein lysates (20 μg) were separated in sodium dodecyl sulfate–polyacrylamide gel electrophoresis (6%-10%) using a Minigel apparatus (Bio-Rad, Munich, Germany) and plotted with a semidry transfer cell (Bio-Rad). Membranes were washed twice with Tris-buffered saline containing 0.1% Tween-20. Immobilized proteins were incubated with primary antibodies [monoclonal mouse anti-human beta-actin, 1:15,000 (MP Biomedicals, Heidelberg, Germany); monoclonal mouse anti-human EEF1A2, 1:5000 (Upstate Biotechnology Inc., Waltham, MA); polyclonal rabbit anti-human MDM4, 1:8000 (Aviva Systems Biology, San Diego, CA)] and horseradish peroxidase–linked anti-mouse or rabbit secondary antibodies (1:2000) (Cell Signaling Technology, Beverly, MA). Immunoblots were visualized using ECL plus (GE Healthcare, Munich, Germany).

Tissue Microarrays and Immunohistochemistry.

A tissue microarray containing 216 samples (normal liver, nontumorous liver tissue of HCC patients, and HCCs) was constructed as described previously,16 and immunohistochemistry was performed on 5-μm sections. Antigens of beta-catenin, EEF1A2, M30, and MIB-1 were retrieved using citrate buffer (pH 6.1) (Dako, Glostrup, Denmark) and detected using the following primary antibodies: monoclonal mouse anti-human beta-catenin (1:500) (BD Biosciences, Heidelberg, Germany), monoclonal mouse eEF1A2 (1:1500) (Upstate Biotechnology), monoclonal mouse M30 (1:10) (Roche, Mannheim, Germany), and monoclonal mouse MIB-1 (Dako). Visualization was done using the labeled streptavidin biotin or the EnVision method (Dako). Counterstaining was performed using hemalum. Staining was assessed using the following immunoreactive scores (IRS)17: 0, absent; 1–4, weak; 5–8, moderate; 9–12, strong expression. For analysis of proliferation (MIB-1) and apoptosis (M30), the percentage of positive tumor cells was calculated. Statistical analyses were performed by Spearman's correlation or Mann-Whitney U test using SPSS 14.0 software (SPSS, Chicago, IL). A P value of <0.05 was considered statistically significant.

Knockdown Experiments.

For inhibition of EEF1A2 and MDM4 gene expression, HepG2 and Hep3B cells were transiently transfected with 5 nmol of small interfering RNAs (Supplementary Table 3) (MWG Biotech, Ebersberg, Germany) using Oligofectamine Reagent (Invitrogen) and were analyzed 72 hours after transfection.

Cell viability and proliferation rates were determined using the MTT assay18 and the Cell Proliferation ELISA Biotrak System, Version 2 (GE Healthcare), respectively. Apoptosis rates were quantified by fluorescence-activated cell sorting analysis of propidium iodide–stained nuclei19 using CellQuest software on a FACScan flow cytometer (Becton Dickinson). For all cell-based assays, results were obtained from 6 replicate wells in 3 independent experiments.

Results

aCGH of HCC.

In this study, 63 HCCs and 4 HCC cell lines were analyzed for DNA copy number alterations via aCGH (Supplementary Tables 1 and 2). We detected several recurrent novel as well as previously described HCC-typical chromosome aberrations (Fig. 1). The most frequent gains were at 1q (70.6%), 6p (41.2%), 8q (48.5%), 17q (30.9%), 19p (30.9%), and 20q (42.6%), and losses at 1p (30.9%), 4q (45.6%), 8p (51.5%), 13q (35.3%), 16q (30.9%), and 17p (36.8%). We also identified 36 loci exhibiting small amplifications or putative homozygous deletions, 7 of which were altered in more than 1 tumor (Table 1). Recurrent amplifications were found at 1q23.2-23.3, 8q24.23-24.3, 11q13.2, 11q13.3, and 14q12. A putative homozygous deletion was detected at 8p23.1 in 2 tumors. One single amplicon at 1q32.1 consisting of 3 BAC clones and spanning a region of 350 kb contained the oncogene MDM4. The recurrently amplified region at 8q24.23-24.3 was 6.82 Mb in size and harbored the genes GLI4 and PTK2, among others. Further amplicons were detected at 10p11.21-q11.21, 17q21.33, and 18p11.32-11.31 as well as a homozygous deletion at 12p13.33. By defining minimally overlapping regions of all sample data, we were able to fine-map the size of the commonly gained region at 20q13 to a region of 2.5 Mb at 20q13.3, containing the potential oncogene EEF1A2, and to narrow the size of the previously reported homozygously deleted region at 4q34.3-qter20 to 8.64 Mb.

Figure 1.

Chromosome alterations in 63 HCCs and 4 cell lines analyzed via aCGH. Frequency data of BAC clones are ordered according to chromosome position. Upper gray bars indicate the frequency of gains and lower black bars the frequency of losses. Chromosome boundaries are indicated by dashed lines.

Table 1. Small Gains/Amplifications and Putative Homozygous Deletions
Small gains/amplifications (log2 ratio > 1)
CytobandStartEndSize (Mb)Log2 ratioCandidate gene(s)CountSample ID
1q23.2-23.3RP11-190A12RP11-122G181.71.1 2HCC 58, 62
1q24.2-24.3RP11-349E20RP4-598F21.362.1BAT2D1, KIAA10961HCC 15
1q32.1RP11-739N20RP11-430C70.351.4MDM41HCC 29
4q11-12RP11-365H22RP11-120K161.281.0 1cell line HuH7
6p25.1RP11-174B19CTC-471F30.181.0 1HCC 51
7q31.1-31.2RP11-563O5RP5-866N182.682.8MET1HCC 11
8q24.22-23RP6-98A24RP11-343P92.531.05WISP13HCC 18, 29, 32
8q24.23-24.3RP11-172M18RP5-1056B246.860.98PTK2, GLI43HCC 18, 29, 32
10p11.21-q11.21RP11-22G4RP11-124O118.41.9FZD81HCC 58
11p14.2-14.1RP11-430L3RP11-587D211.531.0 1HCC 15
11q12.3-13.1RP11-838I13RP11-697H90.281.0 1cell line HuH7
11q13.2RP11-755F10RP11-715F101.421.5 2cell lines Hep3B, HuH7
11q13.3RP11-554A11RP11-211G230.53.6MYEOV1HCC 41
11q13.3RP11-300I6RP11-300I60.162.6CCND1, FGF193HCC 15, 41, 48
13q13.2-13.3RP11-266E6RP11-131P103.961.2CCNA11cell line PLC
14q12RP11-419C10RP11-187E132.342.5ARHGAP52cell linesHuH7 HepG2
17q11.2RP1-66C13RP11-138P220.262.1NOS2A (INOS)1HCC 48
17q11.2RP11-68I3RP11-403E90.791.8EFCAB51HCC 48
17q12RP11-104J23RP11-445F121.061.9 1HCC 48
17q12RP11-19G24RP11-115K30.792.1TCF21cell line PLC
17q12-21.1RP11-62N23RP11-94L150.253.36ERBB21HCC 48
17q21.33RP5-875H18RP11-893F20.331.1COL1A11HCC 9
18p11.32-11.31RP11-419P8RP11-193E150.631.0KNTC2 (HEC1)1HCC 48
19p13.11RP11-165I10RP11-165I100.181.6 1cell line PLC
19p12CTD-2561J22CTD-2561J220.221.3 1HCC 18
19q12RP11-345J21CTC-448F20.442.8CCNE11HCC 5
20q13.2RP4-700G13RP5-1075G210.141.3 1cell line PLC
Putative homozyous deletions (log2 ratio < −1)
4q34.3-qterRP11-226A18RP11-45F238.64−1.2FAT, CASP3, p33ING2, CLDN221cell line PLC
6q27RP1-182D15RP5-1086L220.57−1.0DLL11HCC 39
8p23.1RP11-826L17RP11-161B10.25−1.2 2HCC 5, 58
11q14.1-14.2RP11-90K17RP11-878E110.37−1.1 1HCC 28
12p13.33RP11-21K20RP5-1096D140.36−1.2 1HCC 19
13q14.13-14.2RP11-351K3RP11-214O112.24−1.4 1cell line PLC
13q14.3-q22.2RP11-40A8RP11-332E325.1−1.1 1cell line Hep3B
22q13.31RP5-1163J1CTD-3035N51.57−1.0CELSR11HCC 12
Xq21.31RP4-542O23RP13-166C102.62−1.6 1cell line HuH7

Gain of 8q Is Absent in HCCs of Cryptogenic Etiology.

To identify recurrent etiology-dependent aberrations, we compared the aCGH results of 4 etiologic groups (HBV, HCV, alcoholic, and cryptogenic). When comparing HBV- and HCV-associated HCCs, we observed major differences in the frequencies of gains at 1q32.1 (HBV, 15.4%; HCV, 57.1%), 7q22.1 (HBV, 0%; HCV, 42.9%), and 10q26.3-qter (HBV, 7.7%; HCV, 50.0%), and losses at 4q34.3-qter (HBV, 76.9%; HCV, 28.6%), 9p24.3 (HBV, 69.2%; HCV, 21.4%), and 14q21.2-32.33 (HBV, 38.5%; HCV, 0.0%) (Table 2). HCCs derived from alcoholic and cryptogenic cirrhosis displayed differences in DNA copy number for gains of 1q31.1 (alcoholic, 69.2%; cryptogenic, 30.8%), 8q11.21 (alcoholic, 53.8%; cryptogenic, 7.7%), 8q24.12-21 (alcoholic, 76.9%; cryptogenic, 0.0%), and 10q26.2 (alcoholic, 46.2%; cryptogenic, 7.7%) and losses of 13q14.13-q14.3 (alcoholic, 21.3%; cryptogenic, 61.5%), 17p11.2 (alcoholic, 15.4%; cryptogenic, 53.8%), 18q21.1-21.32 (alcoholic, 7.7%; cryptogenic, 53.8%), 22q12.3 (alcoholic, 0.0%; cryptogenic, 38.5%), and 22q13.2-13.31 (alcoholic, 0.0%; cryptogenic, 38.5%).

Table 2. Etiology-Dependent Chromosome Aberrations in HCC*
CytobandAlterationStartEndSize (Mb)Candidate gene(s)Frequencies
HBVHCVAlcoholCrypto-genic
  1. For each genomic alteration, the etiologic group with the highest frequency is highlighted in boldface. All loci altered with a difference in frequency of at least 30% between 2 etiologic groups are listed. Identification of candidate genes is based on the etiology-specific gene expression data. The only significant (P < 0.01) alteration is highlighted in gray.

1p35.3-35.2LossRP5-1092A3RP11-2010143.0 46.214.323.121.4
1p21.2-21.1LossRP11-202K23RP11-446M50.9COL11A138.57.126.930.8
1q31.1GainRP11-166A4RP11-768E161.9CFHR414.350.069.230.8
1q32.1GainRP11-109H10RP11-104D30.4 15.457.153.830.8
3q26.2-26.31GainRP11-816J6RP11-44A12.9TLOC130.80.07.77.7
3q28GainRP11-466I16RP11-293N10.1 30.80.07.77.7
4q34.3-qterLossRP11-624A4RP11-45F2312.0FAT76.928.638.553.8
5p15.33-p12GainRP11-117B23RP11-28I944.3CDH10, CDH623.128.646.212.1
5q13.3-14.1LossRP11-206N2RP11-2J211.2 30.80.07.77.7
6q24.3-qterLossRP11-497D6RP1-191N2123.5ACAT2, MLLT423.135.761.530.8
7q11.22-qterGainRP11-358M3RP5-1058P1931.0HGF, CYP3A430.033.415.112.2
8p23.3-p12LossRP11-338B22RP11-11N932.6XKR676.942.935.833.9
8q11.21GainRP11-738G5RP11-46G171.2 23.157.153.87.7
8q24.12-21GainRP11-775B15RP11-17E1610.1MYC, TATDN146.264.376.90.0
9p24.3LossGS1-41L13GS1-77L230.2 69.221.415.415.4
9p21.3LossRP11-149I2RP11-149I20.0CDKN2B46.221.430.823.1
10q26.2LossRP11-310M21RP11-540F80.2 7.70.046.27.7
10q26.3-qterGainRP11-140A10XX-2136c481.5 7.750.030.823.1
13q14.13-q14.3LossRP11-139H14RP11-40A85.2P2RY5. SETDB230.835.723.161.5
14q21.2-32.33LossRP11-356O9RP11-417P2468.2LTBP238.50.023.115.4
16q22.1-22.2LossRP11-419C5RP11-499D31.2NQO125.650.07.738.5
17p11.2LossRP11-524F11RP11-744K174.4ALDH3A1, MFAP418.211.015.453.8
18q21.1-21.32LossRP11-171C21RP11-520K1811.1ONECUT223.17.17.753.8
19p13.11GainRP11-837J10RP11-837J100.2 46.214.330.823.1
20q13.33GainRP1-138B7CTB-81F122.5CSE1L61.528.653.830.8
22q12.3GainLL22NC03-22D1RP5-824I191.1TOM138.57.17.77.7
22q12.3LossRP11-1056I22RP11-813F130.3 7.70.00.038.5
22q13.2-13.31LossRP5-979N1RP1-37M34.6PRR57.73.20.038.5

Statistical significance of the identified etiology-dependent chromosome aberrations was evaluated via SAM. Comparing cryptogenic and alcohol-derived HCCs, a 10.1-Mb region at 8q24.12-21 covered by 45 significantly altered BAC clones was identified (false discovery rate <0.5%; P < 0.01). This region, which harbors the oncogene MYC, is always balanced in cryptogenic HCCs (100%) but frequently gained in alcohol-related (76.9%) as well as other HCCs. SAM analyses of the clinical parameters grading, UICC (International Union Against Cancer) stage, hemangiosis carcinomatosa, and tumor size did not reveal any significant correlations with chromosome aberrations, with the sole exception of gain of 1q21.2-q44 occurring more often in large (>5 cm) than in small (<5 cm) tumors (P < 0.01).

MYC Is the Potential Target Gene of the 8q Gain in HCC.

To identify the potential target gene of the 8q gain, we performed microarray-based gene expression analysis (Fig. 2A) comparing a complementary DNA pool of 5 alcohol-induced HCCs (with 8q gains) with a complementary DNA pool from 4 cryptogenic HCCs (with balanced 8q). Within the 10.1-Mb region on 8q24.12-21, we identified several genes that were overexpressed in alcohol-related HCCs (Supplementary Table 4), of which MYC was the most highly up-regulated (18.4-fold), by far exceeding the effect expected from gene dosage alone. Genome-wide, MYC was among the 30 most up-regulated genes in the pooled HCCs of alcoholic history. By performing quantitative RT-PCR (QRT-PCR) on a set of 20 HCCs, we found MYC to be significantly (P < 0.01) overexpressed in HCCs associated with HBV, HCV, and alcohol abuse, but not in cryptogenic HCCs (Fig. 2B). Ingenuity Pathway Analysis (Ingenuity Systems Inc., Redwood City, CA) of gene expression data revealed up-regulation of several MYC target genes or genes involved in MYC signaling in alcoholic (data not shown), but not cryptogenic HCCs. In cryptogenic HCCs, the genome-wide most highly up-regulated gene was FGF19 at 11q13.1 with a log2 ratio of 9.1, representing a more than 500-fold higher expression relative to the alcoholic HCC pool (Supplementary Table 4).

Figure 2.

Etiology-dependent DNA copy number alteration and gene expression on chromosome 8. Data are log2 transformed and plotted as difference ratios (alcoholic minus cryptogenic). Gray bars indicate gene expression data, and the black line indicates the moving average (30 periods) of DNA copy number data. The black box indicates the minimally altered region defined by all cases of the aCGH analysis. Expression microarray data were generated with pooled complementary RNAs; DNA copy number data represent the median values of the individual hybridizations. QRT-PCR gene expression ratios are normalized to normal liver and plotted as log2 transformed ratios. (A) Difference of log2 ratios of alcohol-related HCCs with gains of 8q minus cryptogenic HCCs with balanced chromosome 8q. (B) Relative MYC messenger RNA expression assessed via QRT-PCR in cryptogenic HCCs versus HCCs of other etiologies.

MDM4 and EEF1A2 Gene Expression Is Up-Regulated in HCC.

On the basis of our aCGH data, we considered several HCC candidate genes, including MDM4 on 1q32.1; CASP3, FAT, p33ING and CLDN22 on 4q34.3-qter; PTK2 and GLI4 on 8q24.12-qter; and EEF1A2 on 20q13.33 (Table 1). Measuring expression of these candidates via QRT-PCR in 20 HCC and 4 HCC cell line samples, we found messenger RNA overexpression of MDM4 (P < 0.01) and EEF1A2 (P < 0.01) in HCCs compared with pooled (n = 6) control liver (Fig. 3A), whereas no differences (P < 0.05) could be detected for MDM4 and EEF1A2 expression between the etiologic groups. This was confirmed by western blot analysis, which revealed up-regulation of MDM4 in 11 of 24 and EEF1A2 in 14 of 24 samples, distributed over all etiologic groups, and comparatively weak expression in normal liver (Fig. 3B). The remaining candidates were not differentially expressed in HCC (data not shown).

Figure 3.

Overexpression of MDM4 and EEF1A2 in HCC. (A) MDM4 and EEF1A2 gene expression in HCC measured by QRT-PCR. Expression data are log2 transformed and plotted as ratios relative to expression in normal liver. (B) Exemplary western blot analyses of MDM4 and EEF1A2 proteins in primary HCCs (3 of each etiologic group analyzed) and normal liver.

Immunohistochemistry on tissue microarrays (Fig. 4) showed EEF1A2 immunostaining to be absent or very weak in normal liver tissue [n = 27; mean IRS, 1.31; standard error of the mean (SEM), 0.308]. In nontumorous liver tissues of HCC patients (n = 84; mean IRS, 2.61; SEM, 0.243), only 21% had no detectable EEF1A2 signal, whereas most samples showed either weak (58%), moderate (17%), or strong (1%) (IRS = 9) expression of EEF1A2. Of the HCCs (n = 100; mean, 5.28; SEM, 0.409), 17% did not show any, 29% weak, 24% moderate, and 30% strong EEF1A2 staining. Approximately one-third (9 of 30) of the latter tumors received the highest possible score (IRS = 12). Statistical analysis revealed increased EEF1A2 expression in nontumorous liver tissues of HCC patients compared with normal liver (P < 0.01) and in HCCs compared with normal liver as well as with nontumorous liver tissue of HCC patients (P < 0.01). Thus, there was an increase of EEF1A2 protein expression from normal liver over nontumorous liver tissue of HCC patients to HCC. No significant correlations (P < 0.05) with sex, etiology, tumor size, blood vessel invasion, and UICC stage were detected. Nevertheless, we observed a moderate correlation between EEF1A2 expression and proliferation (Ki67 index, ρ = 0.40; P < 0.01) and a weak correlation with apoptosis rate (M30, ρ = 0.17; P = 0.016), respectively. Additionally, we observed a weak correlation of EEF1A2 expression and nuclear beta-catenin accumulation (ρ = 0.15; P = 0.027) (data not shown). MDM4 protein expression analysis on tissue microarrays could not be performed due to significant staining artifacts on paraffin-embedded tissues.

Figure 4.

Immunohistochemical staining of EEF1A2 on a HCC tissue microarray; (A-C) Normal livers. (D-F) Nontumorous liver tissues of HCC patients. (G-I) Primary HCCs.

Inhibition of MDM4 and EEF1A2 Reduces Cell Viability and Proliferation and Increases Apoptosis in HCC Cell Lines.

To study the function of the newly identified candidates in HCC, we inhibited MDM4 and EEF1A2 expression in the cell lines HepG2 (wild-type TP53) and Hep3B (homozygous TP53 deletion) using small interfering RNAs. Cell viability, proliferation, and apoptosis were determined relative to untreated cells. Silencing of MDM4 in HepG2 and Hep3B cells resulted in a 74% (0.26 ± 0.01; P < 0.01) and 83% (0.17 ± 0.01; P < 0.01) reduction of cell viability, respectively. This effect was associated with a strong reduction of cell proliferation (0.10 ± 0.02 and 0.44 ± 0.05 in HepG2 and Hep3B cells, respectively; P < 0.01), but seemed to be mediated mainly by induction of apoptosis (9.5-fold ± 0.5 and 8.8-fold ± 0.7 in HepG2 and Hep3B cells, respectively; P < 0.01). Reduction of EEF1A2 expression resulted in an 83% (0.17 ± 0.01; P < 0.01) and 76% (0.23 ± 0.03; P < 0.01) decrease of cell viability in HepG2 and Hep3B cells, respectively. Regarding proliferation, silencing of EEF1A2 resulted in a 90% (0.10 ± 0.01; P < 0.01) decrease in HepG2 cells and a 39% (0.61 ± 0.04; P < 0.01) reduction in Hep3B cells, which was accompanied by 5.1-fold (SEM 0.3; P = 0.03) and 4.8-fold (SEM 0.5; P = 0.015) increased apoptosis rates in HepG2 and Hep3B cells, respectively (Fig. 5).

Figure 5.

Knockdown experiments of MDM4 and EEF1A2 in HepG2 and Hep3B cell lines. (A) BrdU proliferation assay with gene-specific and control siRNAs. Inhibition of MDM4 and EEF1A2 expression decreased HepG2 and Hep3B proliferation rates. (B) MTT cell viability assay using gene-specific and control siRNA. Inhibition of MDM4 and EEF1A2 expression decreased HepG2 and Hep3B cell viability. (C) PI-FACS analysis of cells transfected with gene-specific or control siRNAs. Inhibition of MDM4 and EEF1A2 expression increased HepG2 and Hep3B apoptosis rates. Asterisks indicate significant (P < 0.01) differences compared with untreated cells.

Discussion

On the basis of chromosome aberrations, we detected several etiology-specific changes, including those identified in our previous meta-analysis of conventional CGH data.5 Additionally, we could show that gains of 8q are etiology-related, being frequent in HCCs associated with HBV, HCV, and alcohol abuse but absent in cryptogenic HCCs in our analysis. Because gain of chromosome arm 8q is one of the most frequent genomic alterations in human HCCs (48.8% in our aCGH analysis and 46.6% in our previous CGH meta-analysis5), several putative oncogenes from this region have been suggested as contributors to hepatocarcinogenesis.20, 21 Gene expression analyses of alcohol-induced and cryptogenic HCCs showed overexpression of MYC and activation of its target genes in alcohol-induced HCCs, identifying MYC as the likely target gene of 8q gain. Activation of MYC signaling is observed in 30%-60% of primary HCCs22 and was suggested to be important in HBV genotype C–related HCCs.23 Although its role in HCC development remains unclear, MYC might play a role in the maintenance of putative HCC progenitor cells similar to the situation described in primitive neuroectodermal tumors.24 In light of the observed low MYC messenger RNA expression in cryptogenic HCCs, it seems unlikely that MYC contributes to the pathogenesis of these tumors.

In cryptogenic HCCs, we observed strong up-regulation of FGF19, suggesting its potential involvement in an alternative, MYC-independent pathomechanism. FGF19 is known to bind FGFR4, 1 of 4 high-affinity FGF receptors, and to activate FGF signaling, which was shown to be required for the sustained self-renewal and pluripotency of human embryonal stem cells.25 Furthermore, support for a role of FGF19 in HCC comes from studies showing that intraperitoneal injection of recombinant FGF19 causes increased hepatocellular proliferation in mice,26 transgenic mice ectopically expressing FGF19 in skeletal muscle develop HCCs,27 and FGF19-specific antibodies prevent formation of hepatocellular carcinomas in FGF19-transgenic mice.28 Of the FGF19-transgenic mice,27 44% showed activating CTNNB1 mutations, compatible with the frequency of nuclear beta-catenin accumulation we observed in cryptogenic HCCs in this series (3/10) (data not shown). In the majority of the cryptogenic samples (7/13) we analyzed, anamnestic and serological data suggested a diabetogenic metabolic status, which indicates that these HCCs indeed developed from nonalcoholic steatohepatitis, as recently proposed,3 and represent a homogeneous group.

We further identified MDM4 and EEF1A2 as etiology-independent oncogenes in human hepatocarcinogenesis. Gains of 1q are the predominant early and etiology-independent genomic alterations in HCC.5 Amplifications of chromosome band 1q32 have been repeatedly detected via CGH analyses in other human tumors such as breast cancer and malignant gliomas,29 and the mouse double-minute 4 homolog gene (MDM4 or HDMX) was described as the main amplification target in gliomas.29 Mdm4 inhibits p53 transcriptional activator function by binding its N-terminal transactivation domain in vitro, and recent data from mouse models suggest that Mdm4 is a more potent inhibitor of Tp53 transactivation in vivo than Mdm2.30, 31 We found MDM4 to be activated on messenger RNA and protein levels in approximately 50% of the HCC samples analyzed. Knockdown of MDM4 in HCC cell lines reduced cell viability and proliferation and increased apoptosis. The effect of MDM4 inhibition on proliferation was more prominent in HepG2 (wild-type TP53) than in Hep3B cells (homozygous TP53 deletion)—suggesting that, alternatively to TP53 deletion and/or mutation, MDM4 overexpression might suppress p53 function in some HCCs, as previously reported for malignant gliomas.29 However, the protumorigenic function of MDM4 may additionally be mediated by TP53-independent mechanisms, as evidenced by the comparable effects of MDM4 knockdown on viability and proliferation in both TP53 wild-type and deficient cells. One recently reported example of such a mechanism is the limitation of HCC formation in response to telomere dysfunction.32

EEF1A2 maps to chromosome region 20q13.3, which showed copy number gain in almost half (43%) of the HCCs we analyzed. Frequent amplification of this region has also been described in various tumor types such as breast, colon, and ovary.33–35 EEF1A2 is a protein translation factor that is able to transform mammalian cells and is highly expressed in tumors of the ovary, breast, and lung.36–38 It has been described as an activator of Akt, induces filopodia production in rodent and human cell lines, and enhances cell invasion and migration in an Akt- and PI3K-dependent manner, demonstrating a role in controlling phosphatidylinositol signaling, actin remodeling, and cell motility.39, 40 More recently, EEF1A2 overexpression was reported in 2 HCCs with mutated PIK3CA and consequent activation of the Akt pathway.41 In the mouse fibroblast cell line NIH 3T3, eEF1A2 was found to interact with Prdx-I, reducing activation of caspases 3 and 8, increasing activation of Akt, and protecting cells against apoptotic death.42 We therefore speculate that activation of EEF1A2 might foster Akt signaling, which is seen in 40%-60% of primary HCCs.22 Although we identified EEF1A2 by frequent copy number gains of the 20q13.33 locus, overexpression was not solely dependent on the genomic status of this locus, suggesting alternative mechanisms to regulate gene expression. Because immunohistochemistry analysis demonstrated increased EEF1A2 protein levels in nontumorous, mostly cirrhotic liver tissue of HCC patients, EEF1A2 activation may be an early event during malignant transformation. As for MDM4, cell assays clearly showed the protumorigenic potential of EEF1A2.

Interestingly, knockdown of MDM4 and EEF1A2 produced qualitatively and quantitatively similar effects, except for a significant difference with respect to apoptosis induction. The cause for this similarity is not known; however, we speculate that some degree of coregulation might be achieved through interaction of EEF1A2 and MDM2/MDM4-mediated regulation of p53 via Akt.39 This might also be a potential cause for the differential effect of EEF1A2 knockdown on TP53 wild-type and deficient hepatoma cells.

In conclusion, using a combination of genomic, messenger RNA, and protein expression and functional analyses, we identified and validated MDM4 and EEF1A2 as oncogenes in primary HCC. Furthermore, we showed that MYC acts in an etiology-dependent fashion, probably playing an important role in HCCs with frequent gain of 8q (alcohol-induced and virally induced), but not in cryptogenic, potentially nonalcoholic steatohepatitis–induced, HCCs.

Acknowledgements

We thank Dr. Margarete Odenthal (Institute of Pathology, University of Cologne, Germany) for providing primer sequences for HBV DNA detection on FFPE-tissues and Dr. Christian Rudlowski (University Women's Hospital Bonn, Germany) for support, as well as Eva Eiteneuer for technical assistance. We also thank the Mapping Groups of the Wellcome Trust Sanger Institute for initial clone supply and verification.

Ancillary