SEARCH

SEARCH BY CITATION

Keywords:

  • hepatocellular carcinoma;
  • methylation;
  • CpG island;
  • liver cirrhosis

Abstract

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

Abberrant DNA methylation is one of the hallmarks of cancerogenesis. Our study aims to delineate differential DNA methylation in cirrhosis and hepatic cancerogenesis. Patterns of methylation of 27,578 individual CpG loci in 12 hepatocellular carcinomas (HCCs), 15 cirrhotic controls and 12 normal liver samples were investigated using an array-based technology. A supervised principal component analysis (PCA) revealed 167 hypomethylated loci and 100 hypermethylated loci in cirrhosis and HCC as compared to normal controls. Thus, these loci show a “cirrhotic” methylation pattern that is maintained in HCC. In pairwise supervised PCAs between normal liver, cirrhosis and HCC, eight loci were significantly changed in all analyses differentiating the three groups (p < 0.0001). Of these, five loci showed highest methylation levels in HCC and lowest in control tissue (LOC55908, CELSR1, CRMP1, GNRH2, ALOX12 and ANGPTL7), whereas two loci showed the opposite direction of change (SPRR3 and TNFSF15). Genes hypermethylated between normal liver to cirrhosis, which maintain this methylation pattern during the development of HCC, are depleted for CpG islands, high CpG content promoters and polycomb repressive complex 2 (PRC2) targets in embryonic stem cells. In contrast, genes selectively hypermethylated in HCC as compared to nonmalignant samples showed an enrichment of CpG islands, high CpG content promoters and PRC2 target genes (p < 0.0001). Cirrhosis and HCC show distinct patterns of differential methylation with regards to promoter structure, PRC2 targets and CpG islands.

Hepatocellular carcinoma (HCC) is the most common primary liver malignancy and globally the third most common cause of cancer mortality.1, 2 Approximately one million new cases are diagnosed globally each year. The highest incidences are observed in sub-Saharan Africa and Eastern Asia, where hepatitis B virus and hepatitis C virus infections are endemic. HCC incidence is rising also due to an increase in incidence of alcoholic cirrhosis and nonalcoholic steatohepatitis.3, 4 As for many other tumors, the development of HCC is a multistep process characterized by the accumulation of genetic and epigenetic alterations leading to the activation of oncogenes and inactivation or loss of tumor suppressor genes. Genomic alterations as part of the somatic evolution of the cancer genome are common in human cancer in general and also in HCC. Frequent DNA copy number gains at chromosomes 1q, 6p, 8q, 17q and 20q and losses at 1p, 4q, 8p, 13q, 16q and 17p have been identified.5, 6 The driving genes for most of the copy number alterations remain to date unknown and may include MYC, RB1 and TP53 and depending on the type of the underlying liver disease.7, 8 In addition, altered microRNA and gene expression patterns have been associated with clinical prognosis of HCC.9, 10

Table 1. Overview of the samples used in our study
inline image

Altered DNA methylation patterns are one of the hallmarks of cancer.11, 12 Virtually all cancers are associated with aberrant DNA methylation, including lymphomas, colorectal, prostate and brain cancers.13–15 Initially, altered methylation has been shown to be one of the silencing mechanisms for tumor-suppressor-like genes16 but has subsequently been shown to be a phenomenon encompassing a wide array of gene types.15, 17 Among the methylation targets, polycomb repressive complexes (PRC1 and PRC2) regulate key developmental genes and play an important role in differentiation and the maintenance of cell fates.18–20 Polycomb target gene methylation has been reported as a specific pattern of de novo methylation in cancer,21, 22 although the stability and mechanistic interaction of this gene set in evolving cancer cells remain to be clarified.23

Locus-specific and global alterations of methylation have been reported in HCC. Global hypomethylation of liver tumors has first been observed in a mouse model24 and subsequently in the analysis of human HCC and surrounding liver tissue.25 On the other hand, association of global hypermethylation of CpG islands with progression of HCC was observed in the same study.25 A number of locus-specific studies have looked at the methylation status of certain tumor-suppessor genes: Among these, de novo methylation of p16INK4,26–28SOCS129, 30 and RASSF1A.28, 31 have been reported and recently reviewed.32, 33 In a study using 15 hepatoma cell lines after treatment with decitabine, epigenetic silencing through hypermethylation of scavenger receptor class A, member 5 (SCARA5) was observed and confirmed in a panel of human HCC samples.34

Here, we have performed an array-based methylation analysis on 27,578 methylation sites in 13 primary HCC samples in comparison to nontumorous cirrhotic tissue (n = 17) and normal liver (n = 12). By this comprehensive approach, we observed distinct patterns of DNA methylation between normal liver, cirrhotic controls and HCC that may shed further light on hepatic carcinogenesis.

Material and Methods

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

Samples, patients and phenotypes

Matching samples of HCCs and tumor-free cirrhotic liver were obtained intraoperatively during resective liver surgery or transplantation at the surgery departments in Kiel and Berlin (Table 1). The tumor cell content of all HCC samples included in our study is above 90% as determined by pathology. Cirrhotic control tissue was obtained at a distance of at least 4 cm from the nearest tumor manifestation. Histological evaluation of the cirrhotic tissue from the same block of material documented lack of tumor invasion in all cases. Normal liver tissue was obtained surgically from patients undergoing major abdominal surgery. The study was approved by the ethics committees of the Medical Faculty of the Christian-Albrechts University, Kiel and the Charité University Hospital Berlin (reference numbers D 425/07 and D 411/10). Genomic DNA isolated from the HCC cell line Huh7 was used to test reproducibility of DNA methylation analyses.

DNA methylation analysis

DNA samples were checked for integrity by agarose gel electrophoresis. Bisulfite conversion of the DNA was performed using the “Zymo EZ DNA Methylation Kit” (Zymo Research, Orange, CA) according to the manufacturer's instructions with the modifications described in the Infinium Assay Methylation Protocol Guide (Illumina, San Diego, CA). All further analysis steps were performed according to the “Infinium II Assay Lab Setup and Procedures” and the “Infinium Assay Methylation Protocol Guide.” The processed DNA samples were hybridized to the HumanMethylation27 DNA Analysis Bead-Chip (Illumina, San Diego, CA). This array was developed to assay 27,578 CpG sites selected from more than 14,000 genes. Raw hybridization signals were processed using BeadStudio software (version 3.1.3.0, Illumina) applying the default settings.

DNA methylation data analysis

BeadStudio software (Illumina) was used to determine the gene call rate of individual hybridizations and to verify the reproducibility of analyses performed in triplicate using genomic DNA isolated from Huh7 cells.13 Subsequent principal component (PCA) and herarchical cluster analyses were performed using Qlucore's Omics Explorer 2.1 (Version 2.1(25); Qlucore, Lund, Sweden) using DNA methylation values (average β values) obtained from the BeadStudio analysis. Genes were considered being differentially methylated between two data sets if the false discovery rate (FDR, t-test) was below q < 0.01. Where stated in the text, we included only loci in the analyses, which showed a minimal difference of 0.2 between the methylation values of the samples to be compared. This analysis approach was used to identify differentially methylated CpG loci in noncirrhotic liver tissue versus HCC, noncirrhotic liver tissue versus cirrhosis as well as cirrhosis and HCC. We classified the differentially methylated genes as to whether the alteration has been found either in cirrhosis or HCC only or in both, and whether the genes were hypermethylated or hypomethylated. To avoid inaccurate classification driven by minor differences, we included only loci with a minimum β-value difference of 0.2. The means of the methylation values of the resulting loci were further analyzed using Prism software (ver. 4.02; GraphPad Software, San Diego, CA). Where indicated, average β values have been compared by the Kruskal–Wallis test. Odds ratios were calculated using Prism.

Analysis of enrichment for polycomb repressor complex 2 (PRC2) target genes and promoter classes in differentially methylated genes

Proportions of PRC2 target genes and promoter classes in the groups of genes differentially methylated and all genes present on the HumanMethylation27 Bead Chip were compared using the χ2 test (two-sided; Prism, ver. 4.02). A genome-wide mapping of PRC2 genes in embryonic stem cells is available as Supplemental material in the study by Lee et al.35 To analyze whether promoter regions of differentially methylated genes showed different CpG compositions, we used a previously described classification into promoters with high (HCP), intermediate (ICP), low (LCP) and mixed CpG content.13, 36–38 Information whether CpG loci investigated by the array were located in a CpG island was provided by Illumina.

Gene ontology analysis

The Panther tool for “Gene Expression Data Analysis” has been used to identify biological process terms significantly enriched in groups of genes differentially methylated in controls, cirrhosis and HCC (htpp://www.pantherdb.org). This tool uses binomial statistics to compare classifications of multiple clusters of lists to a reference list to statistically determine over- or under-representation of categories.39 A Bonferroni correction for multiple testing has been applied. A list of genes present on the HumanMethylation27 Bead Chip acted as reference list (as provided by Illumina).

Results

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

Reproducibility and quality assessment of the BeadArray methylation analysis

We quantified the methylation status of 27,568 methylation sites in 39 tissue samples using the BeadArray technology.40 The complete dataset is provided in Supporting Information S1. To assess the technical robustness, one cell line (Huh7) was run in triplicate. The reproducibility of the method was demonstrated by the calculation of pairwise correlation coefficients (Spearman rank correlation measure R2), which showed a median value of 0.963 (Supporting Information Fig. S1), which is in line with earlier studies using the BeadArray technology.13 The median call rate on the methylation arrays was 99.99% (range 97.33–100.00%) indicating robust technical performance of the hybridization (Supporting Information Fig. S2). Therefore, no sample or CpG locus was excluded from further analysis. High reproducibility of the array and good correlation with results obtained by MSP, bisulfite pyrosequencing and bisulfite sequencing has been shown before by us and others.13

Differential methylation of loci in cirrhosis as compared to normal controls is maintained in HCC

To identify CpG loci differentially methylated between normal noncirrhotic liver tissue on the one hand and diseased liver tissue (combined cirrhosis and HCC) on the other hand, we first performed a supervised principal component analysis (PCA) by applying a threshold for the FDR of q < 0.01. By this approach, we intended to identify all significant changes from a normal to a diseased liver without excluding genes with minor but significant changes in their DNA methylation pattern. This analysis resulted in 267 differentially methylated CpG loci corresponding to 96 genes hypermethylated and 157 genes hypomethylated in cirrhosis and HCC. Both a herachical cluster analysis (Fig. 1a) and the corresponding supervised PCA (Fig. 1b) from these 267 CpG loci showed a clear separation of noncirrhotic controls and HCC. While in general the cirrhosis samples were on the same major branch of the tree like the noncirrhotic controls (Fig. 1) and are consequently more similar to the controls than to the HCC samples, the methylation status of the affected loci changed gradually from the transition from normal to cirrhosis and further to a malignant phenotype. This supervised analysis suggests that a set of epigenetic alterations acquired during cirrhosis is maintained in HCC.

thumbnail image

Figure 1. Hierarchical cluster analysis (Weighted Average, a) and supervised Principal Component Analysis (PCA, b) of DNA methylation data. Only CpG loci differentially methylated between noncirrhotic controls and non-normal liver tissue (HCC and cirrhosis) were included (q < 0.01, t-test). Normal control samples are indicated by blue squares above the cluster analysis (a) or blue spheres in the PCA (b). Accordingly, cirrhosis samples are indicated in green and HCC samples in red. In the heatmap green color indicates low, black intermediate and red high DNA methylation values (a).

Download figure to PowerPoint

Genes differentially methylated in HCC, cirrhosis and normal liver tissue

To focus on CpG loci showing a clinically and diagnostically more relevant absolute difference between noncirrhotic controls, cirrhosis or HCC, we included only those loci in the further analyses, which showed a minimal difference in their mean β values of 0.2 (corresponding to ∼20% difference in methylation) and applied a FDR of q < 0.01 between samples sets to be compared. We performed supervised PCA to differentiate HCC, cirrhosis and normal liver tissue based on their individual DNA methylation patterns. Applying the conditions described above unraveled 1276 CpG loci corresponding to 1069 genes differentially methylated in noncirrhotic control tissue and HCC (Fig. 2a and Supporting Information Fig. S3). Of these, the majority of 998 GpG loci were hypomethylated, where 278 loci were hypermethylated in HCC as compared to noncirrhotic liver. Gene ontology analysis demonstrated the enrichment of genes involved in immunity, communication, signal transduction and cell motility (Supporting Information Table S2).

thumbnail image

Figure 2. Supervised principal component analyses based on CpG loci differentially methylated (q < 0.01, t-test) either between HCC and noncirrhotic control tissue (a), HCC and cirrhotic tissue (b) or cirrhotic and noncirrhotic control tissue (c). Only CpG loci showing a difference of at least 0.2 (Δβ > 0.2) in their DNA methylation values (avgerage β) between the tissue samples compared were included in the analysis. HCC: red spheres, cirrhotic tissue: green spheres, noncirrhotic control tissue: blue spheres.

Download figure to PowerPoint

Furthermore, differential methylation analysis of data obtained from cirrhotic liver tissue and HCC resulted in 378 CpG loci hypermethylated and 1050 loci hypomethylated in HCC as compared to cirrhotic liver, corresponding to 307 and 829 genes, respectively (Fig. 2b and Supporting Information Fig. S4). These genes were significantly enriched for biological processes such as cell structure, signaling, motility and immunity (Supporting Information Table S3).

Finally, a supervised PCA analysis separating normal liver from cirrhotic tissue identified 247 differentially methylated CpG loci, of these 118 were hypermethylated and 129 were hypomethylated in cirrhosis of the liver (Fig. 1c and Supporting Information Fig. S5) as compared to normal liver. These genes contribute to nucleic acid metabolism and signal transduction (Supporting Information Table S4). In addition, we performed analyses to identify etiology-specific DNA methylation patterns by comparing the patterns in cirrhotic livers and tumors by pathogenesis. However, the results failed to reach a robust significance level (FDR > 0.05).

Classification of CpG loci aberrantly methylated in cirrhosis or HCC

We next classified the differentially methylated genes identified according to their methylation status in normal controls, cirrhosis and HCC. This resulted in eight different groups, which correlated to genes, which were hypomethylated (Fig. 3a) or hypermethylated (Fig. 3b) in HCC only, genes which were epigentically altered already after the first transition from normal liver to cirrhosis but were unaffected from the further transition to HCC (Figs. 3c and 3d), genes that were significantly hypermethylated (Fig. 3e) or hypomethylated (Fig. 3f) in cirrhosis but were methylated to a comparable extend in normal liver and HCC and genes, which were increasingly hypomethylated or hypermethylated (Fig. 4) after both transitions. A complete list of loci is available from the Supporting Information S1.

thumbnail image

Figure 3. Features of the different DNA methylation groups. Genes significantly epigenetically altered (q < 0.01, Δβ > 0.2) were classified whether they were found hypomethylated (a) or hypermethylated (b) in HCC only, hypomethylated (c) or hypermethylated (d) in both cirrhosis and HCC, or hypermethylated (e) or hypomethylated (f) in cirrhosis only (line graphs). Bar plots show percentage of loci located in CpG islands (CpG islands), genes containing either promoters with high CpG (HCP), intermediate CpG (ICP), low CpG (LCP) or mixed CpG (mixed) content as well as the percentage of genes being target genes of the polycomb repressor complex 2 in embryonic stem cells (PRC2 targets) or being imprinted (imprinted). White bars: percentage of genes present on the array, black bars: percentage of genes in the specified group of epigenetically altered genes. This analysis demonstrates that genes de novo methylated in HCC only are characterized by promoters containing CpG islands and mostly high CpG contents. Furthermore, these genes are enriched for PRC2 target genes. This hold also true for genes found hypermethylated in HCC compared to cirrhosis, even if these genes are not differentially methylated between noncirrhotic controls and HCC. In contrast, genes found hypomethylated in HCC or differentially methylated between normal controls and both cirrhosis and HCC are depleted for genes with high CpG content promoters and CpG islands.

Download figure to PowerPoint

thumbnail image

Figure 4. Scatter plot of DNA methylation values (avgerage β) of CpG loci found significantly hyper- or hypomethylated between normal liver controls and cirrhosis as well as between cirrhosis and HCC (p < 0.0001, Kruskal–Wallis test; Δβ > 0.2). Medians with interquartile ranges are shown.

Download figure to PowerPoint

Genes epigenetically altered either in cirrhosis or HCC or both are characterized by specific promoter features

We next analyzed whether genes hypomethylated or hypermethylated in cirrhosis or HCC or both as compared to normal liver were characterized by a specific CpG composition of their promoter sequences. Our data demonstrated that genes which were hypermethylated in HCC either relative to all nonmalignant samples (p < 0.0001, OR = 7.6511; Fig. 3b) or compared to cirrhosis only (p = 0.0347, OR = 3.1949; Fig. 3f) were significantly enriched for CpG islands, whereas genes hypomethylated in HCC (p < 0.0001, OR = 0.1369; Fig. 3a) or similarily but aberrantly methylated in both, cirrhosis and HCC, were significantly depleted for CpG islands (p = 0.0018, OR = 0.3183; Fig. 3c and p < 0.0001, OR = 0.0857; Fig. 3d).

Furthermore, genes found hypermethylated in HCC only were significantly enriched for those genes characterized by promoters containing a HCP (p = 0.0031, OR =1.4734; Fig. 3b) and depleted for genes with LCPs (p < 0.0001, OR = 0.2061; Fig. 3b). In contrast, genes hypomethylated in HCC only (p < 0.0001, OR = 5.9773; Fig. 3a) or aberrantly methylated in cirrhosis and HCC (p < 0.0001, OR = 4.3103; Fig. 3c and 3d) were enriched for LCP and depleted for HCPs (p < 0.0001, OR = 0.0796; Fig. 3a, p = 0.0487, OR = 0.4468; Fig. 3c, and p < 0.0032, OR = 0.3632; Fig. 3d). By trend, genes hypermethylated in cirrhosis exclusively showed a similar pattern like gene hypomethylated in cirrhosis and HCC (Fig. 3e).

Genes hypermethylated in cirrhosis as compared to HCC were enriched for PRC2 target genes in stem cells

It has been previously described for several tumor entities that genes hypermethylated in cancer cells are enriched for those genes, which are repressed by PRC2 (polycomb repressive complex 2) in embryonic stem cells.35 Therefore, we further investigated whether the groups of genes described above were enriched for those genes. As shown in Figure 3a, genes hypomethylated in HCC as compared to nonmalignant samples were significantly depleted for PRC2 targets [p < 0.0001 (χ2 test), OR = 0.2589], whereas genes hypermethylated in HCC only (p < 0.0001, OR =5.3996; Fig. 3b) or hypermethylated in HCC as compared to cirrhosis only (p = 0.0015, OR = 3.5702; Fig. 3f) were significantly enriched for PRC2 target genes in embryonic stem cells. Genes represented by other groups did not show significant changes in the composition of the PRC2 target genes.

These data demonstrate that hypermethylation of CpG islands, HCPs and PRC2 target genes are characteristics of HCC, whereas hypermethylation of genes in the cirrhotic liver is restricted to LCPs and non-CpG island promoters.

Imprinted genes are epigenetically altered in HCC

Remarkably, genes found aberrantly methylated in HCC as compared to nonmalignant samples were significantly enriched for imprinted genes, independently whether the genes were hypomethylatedor hypermethylated in HCC as compared to cirrhosis and normal liver (p < 0.0103; OR = 1.9980; Fig. 3a and p < 0.0291; OR = 2.5536; Fig. 3b).

Genes classified into different groups are enriched for different gene ontology terms

A gene ontology analysis was performed to reveal the contribution of genes found hypomethylated or hypermetyhlated in either HCC exclusively or in both HCC and cirrhosis to distinct pathways or biological processes. Genes found hypermethylated in HCC as compared to nonmalignant samples were significantly enriched for 20 cellular pathways, including several signaling pathways, whereas genes found hypomethylated were significantly enriched for five pathways (Supporting Information Table S7). Furthermore, specifically genes found hypermethylated in HCC as compared to nonmalignant samples are involved in several developmental and differentiation processes (Supporting Information Table S8). Interestingly, the group of genes aberrantly methylated in both, HCC and cirrhosis as compared to normal liver tissue were only enriched for three pathways. More details are available in the Supporting Information Tables S5 and S6.

Only eight CpG loci showed significant alterations in their methylation pattern in all three PCA analyses (Fig. 4). Five loci showed highest methylation levels in HCC and lowest in control tissue (LOC55908, CELSR1, CRMP1, GNRH2, ALOX12 and ANGPTL7) while two loci showed the opposite direction of change (SPRR3 and TNFSF15). These eight loci differentiated between normal tissue of the liver, cirrhosis and HCC (p < 0.0001, Kruskal–Wallis test). We verified the array-based results by bisulfite pyrosequencing of five genes confirming the data obtained with the array as shown in Supporting Information Figure 6.

Discussion

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

By assaying 27,578 CpG sites in the genome, we studied the patterns of altered methylation between normal liver, cirrhosis and HCC. In this analysis, distinct patterns of altered methylation arise that may contribute to a better understanding of hepatic carcinogenesis:

In relation to previous gene-based candidate studies, we confirm the hypermethylation of CDKN2A (p16INK4)26–28 in HCC that showed a methylation pattern compatible with Figure 3b. The pattern of methylation of SOCS1 that had been identified as a target of de novo methylation in HCC previously,29, 30 corresponded to Figure 3f, that is, was hypomethylated in cirrhosis samples only and thus showed a higher methylation level between HCC and the surrounding cirrhotic liver. In comparison to normal liver, however, the methylation level was unaltered (Supporting Information Table S1). No differential methylation of RASSF1A28, 31 was observed in our study. All these locus-specific confirmatory findings have to be viewed with some caution, because the exact CpG islands may differ between the methylation sites assayed between the different studies.

Cirrhosis represents a premalignant liver condition, as the majority of HCC s arise in the context of a liver cirrhosis.41 In line with this paradigm, a substantial number of genes (as depicted in Fig. 1) show an altered methylation pattern between normal and cirrhotic liver that is either progressing or maintained in HCC (Figs. 3c and 3d). Eight CpG loci showed significant alterations in their methylation patterns in all three PCA analyses (Fig. 4). Thus, these loci reflected a pattern that in fact differentiated statistically between normal tissue of the liver, cirrhosis and HCC (p < 0.0001, Kruskal–Wallis test). Of these, five loci showed highest methylation levels in HCC and lowest in control tissue (LOC55908, CELSR1, CRMP1, GNRH2, ALOX12 and ANGPTL7) while two loci showed the opposite direction of change (SPRR3 and TNFSF15).

Additionally, we compared our results with data from a previous study by Gao and coworkers.42 Interestingly, after excluding all loci from Gao's study not present on the HumanMethylation27 Bead Chip, we found a significant overlap between the group of genes hypermethylated in HCC in our study and in HCC in Gao's study as well as in the group of genes becoming hypermethylated during progression from Gao's study (p < 0.001). We found no significant overlap between other groups. However, one has to keep in mind that Gao et al. and we have probably investigated a different set of CpGs in the respective genes due to the different technology used.

To more specifically understand the differential methylation biology of fibrogenesis and carcinogenesis in the liver, we mined the differentially methylated genes according to alterations in cirrhosis versus HCC. The resulting groups of genes were analyzed for the enrichment of genes containing promoters with high or low CpG content and the presence of CpG islands, target genes of the polycomb repressor complex 2 (PRC2) in embryonic stem cells (ESC) and their contribution to distinct biological processes and pathways.

Gene ontology analysis revealed that genes found hypermethylated in HCC only were significantly enriched for cellular pathways with known impact on cancerogenesis, including the heterotrimeric G-protein signaling pathway-Gi alpha and Gs alpha-mediated pathway (p < 0.001), the PI3 kinase pathway (p < 0.001), the Insulin/IGF pathway-protein kinase B signaling cascade (p < 0.001) or the TGF-β signaling pathway (p = 0.005). Genes found hypomethylated in HCC only were significantly depleted for the angiogenesis pathway (p = 0.004), the muscarinic acetylcholine receptor 1 and 3 signaling pathway (p = 0.0232), the TGF-β and the EGF receptor signaling pathway (p = 0.0448 and p = 0.0465, respectively). Further analysis showed that genes epigenetically altered in HCC are significantly enriched for genes with impact on development and differentiation, that is, system development (p < 0.001), mesoderm (p < 0.001) and ectoderm (p < 0.001) development, pattern specification processes (p < 0.001), embryonic (p < 0.001) or gut mesoderm (p < 0.001) development. Interestingly, genes hypomethylated in HCC were involved in several processes of the immune system (p < 0.001), which might contribute to tumor escape from the immune response. In contrast, genes found differentially methylated in cirrhosis and HCC as compared to normal liver were enriched for genes contributing to the circadian clock system (p = 0.0324), the TCA cycle (p = 0.0369) and general transcription by RNA polymerase I (p = 0.0329).

In conclusion, while genes aberrantly methylated exclusively in HCC contribute to signalling pathways and biological processes described to be altered in malignant tumors, genes aberrantly methylated in both, HCC and cirrhosis, as compared to normal liver are predominantly involved in metabolic pathways.

Furthermore, in these analyses, we observed that genes found hypermethylated or hypomethylated in HCC as compared to cirrhosis were enriched for HCPs and LCPs, respectively. This is in line with data on lymphomas, in which HCPs are predominately hypermethylated. In contrast, genes hypomethylated in lymphoma are predominantly characterized by LCPs.13, 37, 38 Furthermore, genes becoming hypermethylated in HCC were also found being enriched for CpG islands. This finding supports the widely accepted model of de novo methylation of CpG islands in cancer11, 43–45 and has previously already shown for the progression of HCC.25 Interestingly, genes that become hypermethylated during the first “transition” from normal liver to cirrhosis and that maintain this methylation pattern during the development of HCC are depleted for CpG islands, HCPs and PRC2 targets while they are enriched for LCPs. Therefore, genes consisting CpG islands, HCPs and genes being PRC2 targets in embryonic stem cells become predominantly hypermethylated in HCC only, therefore, being a characteristic of the malignant cells (Figs. 3b and 3f). Indeed, several authors reported that PRC2 target genes are enriched among the genes found hypermethylated in cancer.13, 21, 46, 47 In contrast, genes found hypermethylated also in cirrhosis were depleted for PRC2 target genes in embryonic stem cells. As the PRC2 complex plays an important role in maintaining the stem cell status in undifferentiated cells, hypermethylation of PRC2 targets in HCC supports the hypothesis, that aberrant DNA methylation might occur already in a cancer precursor cell, and the epigenetic changes found in tumor samples might represent a kind of “stemness memory” or early stages of tumorigenesis. However, the depletion of PRC2 targets in cirrhosis argues against a uniform transition from a normal hepatocyte over cirrhosis to malignancy. Probably, the characteristics of stemness in HCC are be retrieved from a progenitor cell. Hypothetically, HCC might develope only from a small number of cells present in the cirrhotic liver, which are characterized by specific stem cell characteristics, such as suggested for the oval cell population.48, 49

In summary, the simultaneous analysis of about equal numbers of normal liver, cirrhotic liver and HCC with a novel, global tool assessing CpG islands allowed the delineation of distinct methylation patterns in cirrhotic and malignantly transformed liver suggesting a specific methylation biology of these processes. The origin of methylation stemness in HCC cannot, however, be finally clarified by our study—to pinpoint the clonal origins, further studies using HCC arising in noncirrhotic livers and the analysis of microdissected oval cell subpopulations will be necessary.

References

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

Supporting Information

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

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

FilenameFormatSizeDescription
IJC_26136_sm_suppinfofig1.tif74KSupporting Information Figure 1
IJC_26136_sm_suppinfofig2.tif66KSupporting Information Figure 2
IJC_26136_sm_suppinfofig3.tif483KSupporting Information Figure 3
IJC_26136_sm_suppinfofig4.tif493KSupporting Information Figure 4
IJC_26136_sm_suppinfofig5.tif501KSupporting Information Figure 5
IJC_26136_sm_suppinfofig6.tif135KSupporting Information Figure 6
IJC_26136_sm_suppinfotable1.doc1343KSupporting Information Table

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.