SEARCH

SEARCH BY CITATION

Abstract

  1. Top of page
  2. Abstract
  3. Materials and Methods
  4. Results
  5. Discussion
  6. Acknowledgements
  7. References
  8. Supporting Information

To identify new tumor-suppressor gene candidates relevant for human hepatocarcinogenesis, we performed genome-wide methylation profiling and vertical integration with array-based comparative genomic hybridization (aCGH), as well as expression data from a cohort of well-characterized human hepatocellular carcinomas (HCCs). Bisulfite-converted DNAs from 63 HCCs and 10 healthy control livers were analyzed for the methylation status of more than 14,000 genes. After defining the differentially methylated genes in HCCs, we integrated their DNA copy-number alterations as determined by aCGH data and correlated them with gene expression to identify genes potentially silenced by promoter hypermethylation. Aberrant methylation of candidates was further confirmed by pyrosequencing, and methylation dependency of silencing was determined by 5-aza-2′-deoxycytidine (5-aza-dC) treatment. Methylation profiling revealed 2,226 CpG sites that showed methylation differences between healthy control livers and HCCs. Of these, 537 CpG sites were hypermethylated in the tumor DNA, whereas 1,689 sites showed promoter hypomethylation. The hypermethylated set was enriched for genes known to be inactivated by the polycomb repressive complex 2, whereas the group of hypomethylated genes was enriched for imprinted genes. We identified three genes matching all of our selection criteria for a tumor-suppressor gene (period homolog 3 [PER3], insulin-like growth-factor–binding protein, acid labile subunit [IGFALS], and protein Z). PER3 was down-regulated in human HCCs, compared to peritumorous and healthy liver tissues. 5-aza-dC treatment restored PER3 expression in HCC cell lines, indicating that promoter hypermethylation was indeed responsible for gene silencing. Additionally, functional analysis supported a tumor-suppressive function for PER3 and IGFALS in vitro. Conclusion: The present study illustrates that vertical integration of methylation data with high-resolution genomic and transcriptomic data facilitates the identification of new tumor-suppressor gene candidates in human HCC. (HEPATOLOGY 2012;56:1817–1827)

Hepatocellular carcinoma (HCC) is the fifth-most frequent cancer worldwide and has a poor prognosis.1 Various etiologies have been linked to HCC development, most of which cause chronic liver damage and finally lead to liver cirrhosis. The most prevalent etiological factors are 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 lack viral hepatitis, alcoholic history, or other defined causes, such as genetic hemochromatosis or α1-antitrypsin deficiency, and these so-called cryptogenic HCCs have been shown to frequently evolve from nonalcoholic steatohepatitis. Although 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.

Human hepatocarcinogenesis is considered a stepwise process in which genetic and epigenetic alterations lead to the activation of oncogenes and the inactivation of tumor-suppressor genes (TSGs). In contrast to genetic alterations, epigenetic changes that include aberrant methylation and histone modification do not alter the genetic information, but affect the efficacy of messenger RNA (mRNA) transcription. Altered DNA methylation pattern belongs to the hallmarks of cancer.3 Although altered methylation has been initially assumed as a silencing mechanism for TSGs, developmental programs, and imprinting,4, 5 it is also crucial for maintaining cell identity and fate.6, 7 Aberrant hypermethylation of promoter-associated CpG islands has been observed in cancer and affects genes that are involved in main cellular processes, such as apoptosis, cell adhesion, DNA repair, and proliferation.8

In the past, DNA methylation analyses have been carried out mainly by locus-specific techniques after bisulfite conversion of unmethylated cytosines.9, 10

In this study, we performed a genome-wide methylation profiling of 63 HCC samples of well-defined etiologies that had been previously characterized for genomic aberrations by array-based comparative genomic hybridization (aCGH).11 We used the vertical integration of epigenomic, genomic, and expression data as a strategy for the identification of TSG candidates in human hepatocarcinogenesis and characterized the candidate genes, period homolog 3 (PER3), insulin-like growth-factor–binding protein, acid labile subunit (IGFALS), and protein Z (PROZ) in cell culture.

Materials and Methods

  1. Top of page
  2. Abstract
  3. Materials and Methods
  4. Results
  5. Discussion
  6. Acknowledgements
  7. References
  8. Supporting Information

Tumor Material and Patient Characteristics.

Sixty-three human HCCs were analyzed for genome-wide methylation changes. The HCCs included 38 liver resections and 22 explant liver specimen; median age at surgery was 57 years (range, 16-78), and the male/female ratio was 4:1. All diagnoses were confirmed by histological reevaluation, and use of the samples was approved by the local ethics committee. From 3 patients, two HCC nodules were included that previously showed different aCGH, indicating independent tumor development. Etiology was determined as previously described.11 The underlying etiologies were HBV (n = 11), HCV (n = 15), HCV/HBV coinfection (n = 1), alcohol (n = 14), cryptogenic (n = 15), genetic hemochromatosis (n = 2), and two HBV X protein–positive tumors without chronic HBV infection. The patients' characteristics are shown in Supporting Table 1.

Illumina Infinium Methylation Assay.

The Infinium HumanMethylation27 BeadChip (v1.2; Illumina, San Diego, CA) was used to obtain genome-wide DNA methylation profiles across 27,578 CpG dinucleotides located in a region of 1 kb around the transcription start site of 14,495 genes. Genomic DNA (gDNA) was isolated as previously described,11 and high-molecular-weight DNA was subjected to bisulfite conversion using the EZ DNA Methylation Kit (Zymo Research, Irvine, CA), according to the manufacturer's instruction, for use with the Infinium bead array platform. Using 500 ng of the bisulfite-converted gDNA, converted and unconverted (i.e., methylated) sites were interrogated simultaneously by two probes, one designed against the methylated site and one against the unmethylated site, followed by a single base extension of differentially labeled fluorescent nucleotides used for detection. The methylation of the individual CpG sites was quantified by the ratio signal from a methylated probe relative to the sum of both methylated and unmethylated probes. This value (β) varies continuously from 0 (unmethylated) to 1 (fully methylated) and was generated by Illumina Genome Studio software (v1.0). Complete methylation data are available online (http://livercancer.de/index.php?page=supplementary-material).

Pyrosequencing.

The methylation status of PER3 was validated by pyrosequencing using the Pyromark Q24 System (Hs_PER3_01_PM PyroMark CpG Assay; QIAGEN, Hilden, Germany), according to the manufacturer's protocol. Methylation values based on Pyromark Q24 software (2.0; QIAGEN) were compared to the array data of the CpG sites of interest (HumanGRCh37; Chr.1p36.23: 7,845,070).

Reverse Transcription and Polymerase Chain Reaction.

RNA was isolated from 100 mg of snap-frozen tissue using the RNeasy Mini-Kit (QIAGEN), according to the manufacturer's instructions. One microgram of total RNA from tumors and healthy liver tissues (n = 6) were reverse transcribed with the RevertAid H minus Reverse Transcriptase (Fermentas, St. Leon-Rot, Germany) and analyzed using the ABI Prism 7300 Real-Time PCR System (Sequence Detection Software v1.2.2; Applied Biosystems, Foster City, CA) with Absolute SYBR Green ROX Mix (ABgene, Epsom, United Kingdom). Calculations of efficacy, normalization, and relative quantification versus 18s ribosomal RNA were done according to published algorithms.12 The primer sequences are listed in Supporting Table 2.

DNA Microarray Hybridization and Analysis.

Quality and integrity of the total RNA was controlled using an Agilent Technologies 2100 Bioanalyzer (Agilent Technologies, Waldbronn, Germany). Two hundred nanograms of total RNA were applied for Cy3-labeling reaction using the one-color Quick Amp Labeling protocol (Agilent Technologies). Labeled complementary RNA was hybridized to Agilent human 8 × 60 K microarrays at 68°C for 16 hours and scanned using the Agilent DNA Microarray Scanner. Expression values were calculated by the software package, Feature Extraction 10.5.1.1. Complete data are available online (http://livercancer.de/index.php?page=supplementary-material).

Western Blotting.

Tissues were homogenized using the Precellys tissue homogenizer (PeqLab Biotechnology, Erlangen, Germany) and 1× lysis buffer (Cell Signaling Technology, Danvers, MA), supplemented with 1 μM of proteinase inhibitor (Serva, Heidelberg, Germany) and 1× PhosSTOP (Roche, Mannheim, Germany).

Protein lysates (100 μg) were separated by dodium dodecyl sulfate/polyacrylamide gel electrophoresis (8%-12%) using a Minigel apparatus (Bio-Rad, Munich, Germany) and blotted using a semidry transfer cell (Bio-Rad). Polyvinylidene difluoride membranes were washed twice with Tris-buffered saline containing 0.1% Tween 20. Immobilized proteins were incubated with primary antibodies (Abs) (Supporting Table 3) and horseradish-peroxidase–linked antimouse or rabbit secondary Abs (1:2,000; Cell Signaling Technology). Immunoblottings were visualized using ECL plus (GE Healthcare, Munich, Germany).

Tissue Microarrays and Immunohistochemistry.

A tissue microarray (TMA) containing tissue from healthy livers (n = 20), nontumorous liver tissue of HCC patients (n = 66), and HCCs (n = 76; Supporting Table 4) was constructed as previously described,13 and immunohistochemistry (IHC) was performed on 5-μm sections. PER3 (Ab dilution 1:100; Acris Antibodies GmbH, Herford, Germany) antigen was retrieved using citrate buffer (pH 6.1; Dako, Glostrup, Denmark). For detection, the EnVision method (Dako) was used. Counterstaining was performed using hemalum. Staining was assessed using the immunoreactive score, as described previously11: 0, absent; 1-4, weak; 5-8, moderate; 9-12, strong expression.

Vector Design.

A Gateway Cloning system (Invitrogen, Darmstadt, Germany) was used for the expression of PER3 (CV029774.1) and PROZ (BC074906.2; Center for Cancer Systems Biology, The ORFeome Collaboration, CCSB 51a; Harvard Medical School, Boston, MA) in HCC cells. An entry vector (pDONR223) containing the target gene was used to generate the expression construct by homologous recombination with a pDEST27 vector. Recombination was carried out according to the manufacturer's instructions. The inserts of the vectors were validated by sequencing. A pCMV-SPORT6 vector containing IGFALS (BC025681) was used as supplied (Open Biosystems, Huntsville, AL). Cell transfection was performed using the FuGENE HD (Promega, Mannheim, Germany) transfection reagent, according to manufacturer's protocol. Cells were harvested 48 hours after transfection.

Cell Lines, 5′-Aza-2′-Deoxycytidine Treatment, Transfection, and Functional Analyses.

HuH7, PLC/PRF/5, and SNU387 cells were cultured in Dulbecco's modified Eagle's medium and RPMI medium, respectively, supplemented with 10% fetal bovine serum (PAA Laboratories, Pasching, Austria) and 1% penicillin/streptomycin (10 mg/mL; PAA) at 37°C (5% CO2) and passaged every 3-4 days. Cells were plated on 6-cm dishes 24 hours before treatment with 10 μM of 5-aza-2′-deoxycytidine (5-aza-dc; Sigma-Aldrich, St. Louis, MO) or dimethyl sulfoxide as a control. Media and chemicals were changed every 24 hours, and plates were harvested after 96 hours of treatment. All transfections were performed using oligofectamine (Invitrogen), according to the manufacturer's protocol. The short interfering RNA (siRNA) sequences are listed in Supporting Table 2. The final siRNA concentration used was 30 nM (Eurofins MWG Operon, Ebersberg, Germany). Cell viability (tetrazolium assay), apoptosis (fluorescence-activated cell-sorting [FACS] assay), and migration (two-dimensional scratch assay) were determined as described previously.11, 14 Clonogenicity was analyzed 7 days after the seeding of 15K cells.11, 14 For all cell-based assays, results were obtained from six replicate wells in three independent experiments.

Data Quality Control and Statistical Analyses.

CpG-specific methylation patterns revealed non-normal, highly variable distributions that motivated the use of nonparametric statistics. The present study had 90% power to detect a methylation difference of 0.35. This was determined by multiplying the sample size for a two-sample t test (two-sided, α = 0.05, six tumors per healthy sample, common standard deviation of 0.3) by the asymptotic relative efficiency for Wilcoxon's signed-rank test under normality (0.995). Genome-wide methylation data were first filtered according to the following criteria: a β value between 0 and 1, a detection P value (estimated by Illumina) below 0.01, and a positive median absolute deviation of single CpG methylation values. CpG-site methylation differences between HCC and healthy tissue were tested by Wilcoxon's rank-sum tests and quantified by median differences with 95% confidence intervals (CIs). Candidate sites were considered differentially methylated if the false discovery rate (FDR) was below 0.5% (FDR q value <0.005). P values for internal validation of methylation differences relied on Wilcoxon's signed-rank test. The correlation between gene expression and promoter methylation was tested by Wilcoxon's signed-rank tests and measured by Spearman's rank correlations. P values <0.05 were considered statistically significant. Statistical analyses were implemented using the R package (R v2.10.1; http://www.r-project.org), Bioconductor,15 and SPSS 19.0 (SPSS, Inc., Chicago, IL). Pathway analyses, based on the Kyoto Encyclopedia of Genes and Genomes (KEGG), and cytoband analyses were carried out using WEBGestalt software (http://bioinfo.vanderbilt.edu/webgestalt/).

Results

  1. Top of page
  2. Abstract
  3. Materials and Methods
  4. Results
  5. Discussion
  6. Acknowledgements
  7. References
  8. Supporting Information

Genome-wide Methylation Profiling of Human HCC.

Genome-wide methylation profiles were obtained from 63 HCC samples and 10 healthy liver controls. A total of 12,008 CpG sites fulfilled the quality-control criteria, and of these, a total of 2,226 CpG sites were differentially methylated (q value <0.005) in tumors, compared to healthy liver samples. Among them, 537 CpG sites were hypermethylated and 1,689 CpG sites were hypomethylated in tumors. Top hypermethylated CpG sites are shown in Table 1, whereas Supporting Table 5 shows all CpG sites with an absolute median methylation difference >0.35. A gene-ontology analysis of hypermethylated genes revealed an enrichment of genes that are either involved in metabolic processes or that are known to be commonly altered in cancer (Supporting Table 6). Additionally, the set of hypermethylated genes comprised 20 of 125 genes that had been previously described as silenced by the Polycomb Repressive Complex 2 (PCR2; Supporting Table 7).16 CpG sites with a median hypomethylation of at least −0.40 in HCCs are shown in Table 2. Among the hypomethylated genes (q value <0.005), we observed an enrichment for certain chromosomal regions (1q: FDR q value = 1e-06; 11p15 FDR q value = 7e-05; 12p13: FDR q value = 6e-07; 19q13: 1e-07; 20p13: FDR q value = 0.0038; 21q22: FDR q value = 8e-10). Furthermore, genes that have been shown to be imprinted were frequently hypomethylated in human HCCs, compared to healthy livers (Supporting Table 8).

Table 1. List of the Top Hypermethylated Genes With a Median Methylation Difference ≥0.45 in HCCs, Compared to Healthy Samples
CpG siteMedian Difference95% CIFDR q ValueGeneNameChromosome
cg090536800.620.510.69<0.001UTF1Undifferentiated embryonic cell transcription factor 110q26
cg257208040.600.510.68<0.001TLX3T-cell leukemia homeobox 35q35.1
cg265214040.600.300.72<0.001HOXA9Homeobox A97p15.2
cg040347670.590.330.68<0.001GRASPGRP1 (general receptor for phosphoinositides 1)- associated scaffold protein12q13.13
cg228819140.570.300.68<0.001NID2Nidogen 2 (osteonidogen)14q22.1
cg188159430.540.360.64<0.001FOXE3Forkhead box E31p32
cg223751920.540.320.65<0.001IGF1RInsulin-like growth factor 1 receptor15q26.3
cg080978820.520.390.66<0.001POU4F1POU class 4 homeobox 113q31.1
cg024401770.520.400.62<0.001ZNF702Zinc finger protein 702, pseudogene19q13.41
cg185361480.520.240.65<0.001TBX4T-box 417q21-q22
cg047973230.510.220.68<0.001SOCS2Suppressor of cytokine signaling 212q
cg233917850.510.370.63<0.001DNM3Dynamin 31q24.1
cg086687900.500.340.65<0.001ZNF154Zinc finger protein 15419q13.4
cg151916480.500.210.59<0.001SALL3Sal-like 3 (Drosophila)18q23
cg138014160.500.330.61<0.001AKR1B1Aldo-keto reductase family 1, member B1 (aldose reductase)7q35
cg063772780.500.200.63<0.001RUNX3Runt-related transcription factor 31p36
cg090997440.490.320.70<0.001CDKN2ACyclin-dependent kinase inhibitor 2A (p16)9p21
cg004894010.490.300.62<0.001FLT4fms-related tyrosine kinase 45q34-q35
cg062918670.470.390.55<0.001HTR75-Hydroxytryptamine receptor 710q21-q24
cg027555250.470.290.58<0.001NETO2Neuropilin (NRP) and tolloid (TLL)-like 216q11.2
cg075331480.470.200.61<0.001TRIM58Tripartite motif containing 581q44
cg092600890.470.330.62<0.001NKX6-2NK6 homeobox 210q26.3
cg056848910.470.320.58<0.001DAB2IPDAB2 interacting protein9q33.1-q33.3
cg218708840.460.320.54<0.001GPR25G-protein-coupled receptor 251q32.1
cg155202790.460.280.58<0.001HOXD8Homeobox D82q31.1
cg154336310.450.280.57<0.001IRX2Iroquois homeobox 25p15.33
Table 2. List of the Top Hypomethylated Genes With a Median Methylation Difference at Least −0.40 in HCC, Compared to Healthy Samples
CpG siteMedian Difference95% CIFDR q ValueGeneNameChromosome
cg06806711−0.46−0.53−0.36<0.001MS4A1Membrane-spanning 4-domains, subfamily A, member 111q12-q13.1
cg25856811−0.46−0.54−0.35<0.001SPRR3Membrane-spanning 4-domains, subfamily A, member 111q12-q13.1
cg09120035−0.46−0.55−0.32<0.001CYP11B1Cytochrome P450, family 11, subfamily B, polypeptide 18q21-q22
cg06627364−0.45−0.51−0.36<0.001MGC4677Long intergenic nonprotein coding RNA 1522p11.2
cg04505023−0.44−0.55−0.30<0.001SPRR1ASmall proline-rich protein 1A1q21-q22
cg17725968−0.43−0.54−0.25<0.001PDHA2Pyruvate dehydrogenase (lipoamide) alpha 24q22-q23
cg11009736−0.43−0.51−0.34<0.001MARCOMacrophage receptor with collagenous structure2q14.2
cg15320474−0.43−0.53−0.32<0.001UBDUbiquitin D6p21.3
cg18780284−0.42−0.51−0.33<0.001SPRR1BSmall proline-rich protein 1B1q21-q22
cg23595927−0.41−0.52−0.27<0.001MYL5Myosin, light chain 5, regulatory4p16
cg18675600−0.41−0.49−0.26<0.001PTP4A3Protein tyrosine phosphatase type IVA, member 38q24.3
cg08878744−0.41−0.49−0.29<0.001LCE1BLate cornified envelope 1B1q21.3
cg07592353−0.40−0.49−0.25<0.001GABRA6Gamma-aminobutyric acid (GABA) A receptor, alpha 65q34
cg08763351−0.40−0.52−0.28<0.001SPRR4Small proline-rich protein 41q21.3
cg10501065−0.40−0.50−0.26<0.001IGF2ASInsulin-like growth factor 2 antisense (nonprotein coding)11p15.5

To identify methylation differences related to tumor etiology, we carried out additional analyses comparing the methylation of four etiological subgroups (HBV, HCV, alcoholic, and cryptogenic) with healthy samples. Eighty-one CpG sites showed specific methylation differences in HBV-induced HCCs, whereas HCV-induced HCCs showed 198 exclusive differentially methylated CpG units (Fig. 1). Supporting Table 9 lists those 65 genes that are altered in every etiological subgroup.

thumbnail image

Figure 1. Venn diagram demonstrating etiology-dependent methylation changes in human HCCs.

Download figure to PowerPoint

Narrowing in on TSG Candidates Through Vertical Integration of Epigenomic, Genomic, and Transcriptomic Profiling.

TSGs are important gatekeepers that protect against somatic evolution of cancer. According to Knudson's hypothesis, the inactivation of TSGs requires the inactivation of both alleles. The flow chart in Fig. 2 describes the strategy for the identification of new potential TSGs. To prioritize the 537 potential TSGs, we first considered the genomic alterations as determined previously by aCGH.11 For this purpose, we selected all chromosomal regions showing small losses (<5 Mb) of genomic information and that were present in at least 10% of cases. These included 139 chromosomal regions. After integrating hypermethylation with genomic regions that showed genomic losses, 17 candidate genes remained (Table 3). We subsequently considered gene expression as an additional selection layer. Finally, PER3, PROZ, and IGFALS remained as genes that showed an inverse correlation between gene expression and promoter methylation, indicating that promoter hypermethylation was responsible for their silencing in human HCCs (Table 3).3

thumbnail image

Figure 2. Flow chart illustrating the strategy used for the identification of tumor-suppressor gene candidates.

Download figure to PowerPoint

thumbnail image

Figure 3. Immunochemistry for PER3 on tissue microarray. Healthy livers (A-C), nontumorous liver tissues of HCC patients (D-F), and primary HCCs (G-I).

Download figure to PowerPoint

Table 3. Genes With Hypermethylated CpG Sites That Also Show Genomic Loss According to Previous Analysis15
inline image

Validation of PER3 as a TSG Candidate in Human HCC.

Because genes involved in the circadian rhythms have been implicated in tumorigenesis, we wanted to independently validate the methylation and expression changes detected for PER3. An excellent correlation existed between the values obtained using the Infinium array and the pyrosequencing approach, demonstrating that measurements of PER3 methylation were highly reproducible (Supporting Fig. 1A) (Spearman's rho: 0.93; P < 0.001). Additionally, the silencing of PER3 mRNA in human HCCs was confirmed by real-time reverse-transcriptase polymerase chain reaction (Spearman's rho: 0.78; P < 0.001; data not shown).

PER3 was down-regulated in three of six HCC cell lines analyzed (Fig. 4A). To confirm that promoter hypermethylation was responsible for gene silencing, we treated Hep3B, HuH7, and HepG2 cells that showed PER3 down-regulation with 5-aza-dC, an inhibitor of DNA methyltransferase 1, which restored PER3 expresssion in all three cell lines (shown representatively for HuH7 cells in Supporting Fig. 1B), demonstrating that the PER3 down-regulation was indeed the result of promoter hypermethylation.

thumbnail image

Figure 4. Functional analyses after PER3 reexpression of HuH7 cells. (A) Relative PER3 expression in various HCC cell lines. (B) Reexpression of PER3 significantly reduces clonogenicity and (C) cell viability in HuH7 cells with methylated PER3 promoter, whereas siRNA-mediated silencing in PER3-expressing SNU387 cells has a vice versa effect. (D) FACS analysis demonstrates increased apoptosis after reexpression in HuH7 cells, compared to mock-transfected cells, whereas apoptosis is diminished after PER3 knockdown in SNU387 cells. (E) Detection of PER3, phosphorylated CHEK2, as well as caspase-3 and PARP cleavage after reexpression in HuH7 cels. (F) Migration is neither significantly altered after PER3 reexpression in HuH7 (left panel) nor siRNA-mediated silencing in SNU387 cells.

Download figure to PowerPoint

PER3 Expression Is Down-regulated in Human HCC.

Next, we determined whether PER3 protein expression was down-regulated in human HCCs. We performed IHC using TMA (Fig. 3). PER3 expression was detectable in all healthy liver tissues (n = 20). Low PER3 expression was observed in 60% of healthy livers, whereas 35% displayed a moderate expression, and 5% showed high PER3 expression. In nontumorous liver tissues of HCC patients (n = 66), 3% displayed no detectable PER3 signal at all, whereas most samples showed either weak (61%), moderate (27%), or high (9%) expression of PER3. Of the HCCs (n = 76), 22% did not show any, 58% showed weak, 17% showed moderate, and only 3% showed strong PER3 staining. Statistical analysis revealed significantly reduced PER3 expression in HCCs, compared to nontumorous liver tissues of HCC patients as well as compared to healthy liver samples (both P < 0.01). Additionally, we observed a significantly lower PER3 expression in HCCs with vascular invasion, compared to HCCs without vascular invasion (P < 0.05). No significant associations were found with gender, etiology, tumor size, and International Union Against Cancer stage (P > 0.05).

PER3 Is a TSG in Human HCC.

To test whether the newly identified TSG candidate was indeed functional, we transiently reexpressed PER3 in HuH7 cells. This reexpression reduced clonogenicity (0.52 ± 0.03; P < 0.01; Fig. 4B) and overall cell viability, compared to mock-transfected cells (0.43 ± 0.04; P < 0.01; Fig. 4C), whereas migration was not significantly affected (0.89 ± 0.04; P > 0.05; Fig. 4F). Reduced cell viability was the result of apoptosis, as shown by FACS analysis (2.39-fold ± 0.002, compared to mock-transfected cells; Fig. 4D) and cleavage of caspase-3 and poly(ADP-ribose)-polymerase (PARP) (Fig. 4E). This effect was associated with the phosphorylation of CHK2 checkpoint homolog (CHEK2). In contrast, siRNA-mediated silencing of PER3 in SNU387 cells increased cell viability (1.27 ± 0.04; P < 0.01; Fig. 4C) and decreased apoptosis rate, compared to mock-transfected cells (0.41 ± 0.03; P < 0.05; Fig. 4D).

Functional Characterization of the Potential TSGs, IGFALS and PROZ, in Human HCC.

For functional characterization, IGFALS and PROZ were also transiently expressed in vitro. IGFALS reexpression in HuH7 cells significantly reduced overall cell viability (0.58 ± 0.05; P < 0.01; Fig. 5A) and clonogenicity, compared to mock-transfected cells (0.67 ± 0.04; P < 0.01; Fig. 5B), which was associated with apoptosis induction (2.60-fold ± 0.01; Fig. 5C). Migration was not significantly affected (1.14 ± 0.06; P > 0.05; Fig. 5D).

thumbnail image

Figure 5. Functional analyses after IGFALS and PROZ expression in HCC cells. (A) Cell viability, (B) clonogenicity, (C) apoptosis, and (D) migration after expression of IGFALS in HuH7 and PROZ in PLC/PRF/S cells, respectively.

Download figure to PowerPoint

In contrast, expression of PROZ in PLC/PRF/5 cells increased cell viability (1.28 ± 0.02; P < 0.01; Fig. 5A) and clonogenicity (1.31 ± 0.04; P < 0.01; Fig. 5A), compared to mock-transfected cells, which was associated by a slightly decreased apoptosis rate (0.81-fold ± 0.04; P < 0.05; Fig. 5C). PROZ expression had no significant effect on migration of PLC/PRF/5 cells (0.87 ± 0.06; P > 0.05).

Discussion

  1. Top of page
  2. Abstract
  3. Materials and Methods
  4. Results
  5. Discussion
  6. Acknowledgements
  7. References
  8. Supporting Information

On the basis of genome-wide array-based profiling of a series of well-characterized human HCCs that had been previously analyzed by high-resolution aCGH,11 we detected etiology-dependent and -independent methylation changes in human HCC that may help to improve our understanding of human hepatocarcinogenesis.

Some of the genes showing aberrant methylation in our analysis have been addressed in single-locus–specific analyses, such as cyclin-dependent kinase inhibitor 2A,9 runt-related transcription factor 3,10 homeobox A9,17 DAB2 interacting protein,18 and suppressor of cytokine signaling 2.19 Additionally, we confirmed recent array-based methylation analysis that showed hypermethylation of dynamin 3, fms-related tyrosine kinase 4, forkhead box E3, G-protein-coupled receptor 25, GRP1-associated scaffold protein, homeobox D1, 5-hydroxytryptamine receptor 7, insulin-like growth factor 1 receptor, neuropilin and tolloid-like 2, nidogen 2, NK6 homeobox 2, POU class 4 homeobox 1, undifferentiated embryonic cell transcription factor 1, T-box 4 (TBX4), T-cell leukemia homeobox 3, tripartite motif containing 58, zinc finger protein 154, and zinc finger protein 702, pseudogene in human HCCs.20-24 However, none of these genes, except for TBX4, were considered top candidates in previous analyses,20 most likely the result of the relatively low number of cases analyzed.20-22

Activating mutations of β-catenin (catenin [cadherin-associated protein], beta; CTNNB1) have been frequently reported in human HCCs25 and have been associated with a chromosomal stable phenotype.26, 27 Recently, using locus-specific methylation analysis, Nishida et al. reported significantly higher methylation in CTNNB1-mutated HCCs, compared to other HCCs,28 a finding we could also demonstrate on a genome-wide scale (Supporting Fig. 2), suggesting that methylation profiling may significantly contribute to a comprehensive molecular classification of human hepatocarcinogenesis.

The evolutionary, highly conserved Polycomb group proteins promote gene repression through modification of chromatin structure and form multiple Polycomb Repressive Complexes (PRC) that exert intrinsic histone methyltransferase activity and maintain methylation of core histones.29, 30 PRC2 has been linked to both stem cell biology and cancer.31, 32 Polycomb target gene methylation has been reported to result in a specific stem-cell–like chromatin pattern through de novo methylation in cancer.16, 33 Here, we could validate that PRC2 target genes are prone to promoter hypermethylation in human HCC, as recently proposed by Ammerpohl et al.22 Polycomb group proteins may represent interesting translational targets, because the S-adenosylhomocysteine hydrolase inhibitor, 3-deazaneplanocin A, has been shown to selectively induce apoptosis in cancer cells through depletion of PRC2 components.34

When comparing the hypomethylated genes with previous array-based methylation-profiling approaches,21-24, 35, 36 two new genes showed loss of imprinting in (catenin [cadherin-associated protein], alpha and insulin growth factor 2 antisense [nonprotein coding]), and the long noncoding RNA, LINC00152, was detected as differentially hypomethylated during hepatocarcinogenesis. Furthermore, the recently described phenomenon that hypomethylated promoters form cluster across the genome (chromosomes 16, 17, 19, 20, 21, 22, and X)37 could be confirmed and refined (e.g., to regions 19q13, 20p13, and 21q22). Additionally, new clusters of hypomethylated DNA could be identified at 1q, 11p15, and 12p13.

Our approach to profile the methylation changes in HCCs and to integrate these data with preexisting aCGH and expression data revealed three TSG candidates (IGSALS, PER3, and PROZ), of which PER3 and IGFALS were validated as TSGs in human hepatocarcinogenesis.38 PER3 expression was significantly lower in HCCs with vascular invasion, a negative prognostic feature.39 The PER3 gene is located on chromosomal arm 1p36, which, in addition, showed genomic losses in 16% (139 of 871) of human HCCs.40 It belongs to the period gene family (PER) that controls circadian rhythms.41 The circadian clock is organized through a complex network of feedback loops that drive rhythmic expression patterns of core clock components in mammals.42 Furthermore, the PER family members—including PER3—have been implicated in cell-cycle control, DNA damage response, as well as tumor progression and recurrence.43-46 PER3 physically interacts with ataxia telangiectasia mutated and the checkpoint kinase, CHEK2, and silencing of PER3 impairs CHEK2 activation after DNA damage, whereas its overexpression results in apoptosis through induction of CHEK2. Thus, PER3 is likely to function as a checkpoint protein relevant for checkpoint activation and apoptosis.47 In line with these observations, our data demonstrate an induction of CHECK2 phosphorylation after PER3 reexpression, indicating that the protumorigenic PER3 function in human HCC is likely to be mediated by CHEK2 (Supporting Fig. 3).

Besides transcriptional activation resulting from loss of promoter-specific imprinting or reactivation of the fetal promoter pattern, dysregulation of insulin-like growth factor-II (IGF-II) signaling in HCC predominantly occurs at the level of IGF-II bioavailability. The majority of IGF-II circulates in the serum as a complex with the insulin-like growth-factor–binding protein (IGFBP)-3 or IGFBP-5 as well as an acid-labile subunit (IGFALS). The function of IGFALS is to prolong the half-life of the IGF-IGFBP-3/IGFBP-5 binary complexes.48 The down-regulation of IGFBPs may increase the IGF-II bioavailability in HCC.49 Our findings after IGFALS expression in HCC cells indicate that, in addition, the epigenetic silencing of IGFALS contributes to the dysregulation of IGF-II signaling in HCC.

Several hemostatic system components, including factor X, contribute to cancer progression. PROZ is a vitamin K–dependent factor that, in complex with the protein Z–dependent protease inhibitor (ZPI), inhibits activated factor X on phospholipid surfaces.50 Although PROZ matched our selection criteria for a TSG, the functional analysis did not support its tumor-suppressive function. In contrast, reexpression in PLC/PRF/5 cells even suggested a protumorigenic function in vitro. In line with our functional findings, PROZ expression has been observed in several human cancers, suggesting that the PROZ/ZPI complex might support the invasion and metastasis of tumor cells.50

In summary, we describe aberrant methylation profiles in human HCC and provide evidence that the integration of epigenetic alteration pattern is essential for a comprehensive classification of human hepatocarcinogenesis. Additionally, we show that the vertical integration of methylation data with high-resolution genomic and transcriptomic data allows for the identification of promising TSG candidates in human HCC. It highlights the potential for efficient epigenetic approaches for the prevention and therapy of human HCCs.

Acknowledgements

  1. Top of page
  2. Abstract
  3. Materials and Methods
  4. Results
  5. Discussion
  6. Acknowledgements
  7. References
  8. Supporting Information

The authors are grateful to Verena Kautz, Sarah Meßnard, and Eva Eiteneuer for their excellent technical assistance.

References

  1. Top of page
  2. Abstract
  3. Materials and Methods
  4. Results
  5. Discussion
  6. Acknowledgements
  7. References
  8. Supporting Information

Supporting Information

  1. Top of page
  2. Abstract
  3. Materials and Methods
  4. Results
  5. Discussion
  6. Acknowledgements
  7. References
  8. Supporting Information

Additional Supporting Information may be found in the online version of this article.

FilenameFormatSizeDescription
HEP_25870_sm_SuppFig1.tif2274KSupporting Information Figure 1.
HEP_25870_sm_SuppFig2.tif9303KSupporting Information Figure 2.
HEP_25870_sm_SuppFig3.tif13886KSupporting Information Figure 3.
HEP_25870_sm_SuppTables.doc458KSupporting Information Tables.

Please note: Wiley Blackwell is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.