DNA methylation in hepatocellular carcinoma

Abstract

As for many other tumors, development of hepatocellular carcinoma (HCC) must be understood as a multistep process with accumulation of genetic and epigenetic alterations in regulatory genes, leading to activation of oncogenes and inactivation or loss of tumor suppressor genes (TSG). In the last decades, in addition to genetic alterations, epigenetic inactivation of (tumor suppressor) genes by promoter hypermethylation has been recognized as an important and alternative mechanism in tumorigenesis. In HCC, aberrant methylation of promoter sequences occurs not only in advanced tumors, it has been also observed in premalignant conditions just as chronic viral hepatitis B or C and cirrhotic liver. This review discusses the epigenetic alterations in hepatocellular carcinoma focusing DNA methylation.

Key Words: Hepatocellular carcinoma, DNA methylation, Histone modification, Tumor suppressor genes

INTRODUCTION

Hepatocellular carcinoma (HCC) is one of the most common cancer worldwide. It shows a wide geographical variation with low incidence areas in North America and Europe and high incidence areas in Africa and Asia. 70%-80% of hepatocellular carcinoma occurs in cirrhotic liver. In high incidence areas, such as Asia and Africa, HCC is strongly associated with chronic viral hepatitis B and C and liver cirrhosis. Nutritional factors, toxins and metabolic diseases contribute also to hepatocarcinogenesis[12].

As for many other tumors, development of HCC is due to a multistep process with accumulation of genetic and epigenetic alterations in regulatory genes, leading to activation of oncogenes and inactivation or loss of tumor suppressor genes (TSG).

In the last three decades, cancer has been understood as a summary of altered genetic and epigenetic events. The epigenetic pathway is, in contrast to genetic events, a reversible alteration and characterized by three main mechanisms: (1) DNA hypermethylation leading to inactivation, (2) DNA hypomethylation causing genomic instability, (3) histone modifications affecting chromatin conformation.

These processes, especially aberrant DNA methylation and histone modifications, are closely linked with each other by a protein complex of transcript activators and repressors and alter mRNA transcript expression of affected genes[3].

Characteristically, DNA methylation does not change the genetic information. It just alters the readability of the DNA and results in inactivation of genes by subsequent mRNA transcript repression.

In humans and other mammals, CpG island methylation is an important physiological mechanism. The inactivated X-chromosome of females, silenced alleles of imprinted genes or inserted viral genes and repeat elements are inactivated through promoter methylation[45].

DNA HYPERMETHYLATION

Promoter hypermethylation

Hypermethylation of CpG islands in promoter sequences is associated with silencing of tumor suppressor genes and tumor-related genes by subsequent downregulation of mRNA transcript expression. Epigenetic silenced genes are involved in important molecular pathways of carcinogenesis e.g., cell cycle regulation, apoptosis, DNA repair or cell adhesion.

According to other types of malignant tumors, in hepatocellular carcinomas, aberrant methylation of several TSG and tumor-related genes such as RASSF1A, hMLH1 or SOCS1 was frequently observed[6].

CpG island hypermethylation is not only seen in HCC tumor tissue. Even in premalignant conditions such as dysplastic nodules or cirrhotic liver, promoter methylation of several kinds of TSG, e.g. E-cadherin, GSTP1 or 6^Ink4a was demonstrated (Table 1 shows promoter methylation of different tumor-related genes in HCC).

Table 1 Methylation in hepatocellular carcinoma.

Gene	Location	Function	Methylation frequency (%)	Ref.
p16INK4a	9q21	CDK inhibitor	17-83	[10,11]
p14ARF	9q21	MDM2 inhibitor	25-30	[7,12]
CASP8	2q33	Apoptosis	72	[17]
TMS1/ASC	16p11.2	Apoptosis	80	[23]
E-Cadherin	16q22.1	Cell adhesion	33-67	[28-30]
M-Cadherin	16q24.1	Cell adhesion	55	[34]
H-Cadherin	16q24.2-3	Cell adhesion	21	[17]
TIMP3	22q12	MMP inhibitor	13-19	[37-39]
hMLH1	3p21.3	Mismatch repair	18-44	[47-50]
hMSH2	2p21-22	Mismatch repair	68	[47,49,50]
hMSH3	5q11-12	Mismatch repair	75	[47,49,50]
MGMT	10q26	DNA repair	22-39	[53-55]
GSTP1	11q13	Detoxification	41-58	[53,58-61]
SOCS-1	16p13.13	Cytokine inhibitor	60	[67]
SOCS-3	17q25.3	Cytokine inhibitor	33	[68]
RASSF1A	3p21.3	Apoptosis	54-95	[71,75,76]
BLU	3p21.3	Unknown	20	[71]
SEMA3B	3p21.3	Apoptosis	80	[71]
FHIT	3p14.2	histidine triad protein	71	[86]

Proliferation and apoptosis

One of the most important pathways affected in HCC are the Rb (Retinoblastoma) gene and INK4a-ARF pathway[7]. The INK4a-ARF locus is coding two cell-cycle regulatory proteins, p16^INK4a and p14^ARF, acting through the Rb-CDK4 and p53 pathways. p16^INK4a binds to cyclin-dependent protein kinase 4 (CDK4) and inhibits the ability of CDK4 to interact with cyclin D1. p14^ARF prevents the p53 degradation through its binding to MDM-2 and induces cell cycle arrest[89].

p16^INK4a is one of the most altered tumor suppressor gene in human cancer. In HCC, loss of p16^INK4a is mainly caused by aberrant promoter methylation, whereas deletions and mutations of this gene locus are infrequently seen. CpG island promoter methylation was reported from 55% to 73%, but aberrant methylation occurred also in non-cancerous liver tissue with cirrhosis in 29% or chronic hepatitis B and C up to 23%[1011]. Compared to p16^INK4a methylation, p14^ARF promoter methylation was observed less frequently in 8% to 20% of HCC. It was demonstrated that inactivation of p14^ARF is due to homozygous deletions. No correlation was found between p53 mutations and promoter methylation of p16^INK4a or p14^ARF[712].

Caspase 8 (CASP8) is a key apoptotic gene that is involved in death receptor and the mitochondrial pathways and acts as initiator CASP[13]. CASP8 is silenced by aberrant hypermethylation of its promoter in childhood neuroblastomas[1415]. In HCC, CASP8 aberrant promoter methylation was reported by Yu et al[16] with a frequency up to 72%.

TMS1/ASC, another proapoptotic gene, functions as a negative regulator of nuclear factor kappaB (NF-κB) and blocks transcription of survival signals[17]. First, TSM1 was identified as a target of methylation-induced silencing in cell lines overexpressing DNMT1. Epigenetic inactivation of TMS1 was demonstrated in human glioblastomas, ovarian cancer, human melanoma, colorectal carcinomas or in lung cancer and breast cancer[18–22]. In HCC, TMS1 promoter methylation was observed in 80%[23].

In hepatocellular carcinoma cell lines, restoration of TMS1 transcript was induced by demethylating agent 5’-AZA and trichostatin, a histone deacetylase inhibitor. Furthermore, in these cell lines TMS1 DNA methylation was associated with histone H3 lysine 9 hypoacetylation and trimethylation[24].

Cell adhesion and invasion

E-cadherin (CDH1): E-cadherin, a member of calcium-mediated membrane glycoproteins, is expressed in all epithelial cells acting as an adhesion molecule. Inactivation of E-cadherin induces loss of adherens junctions and impairment of cell adhesiveness and cell proliferation signalling pathways. In tumours, reduction of E-cadherin expression results in tumour progression, cell invasion and formation of metastasis[2526].

Downregulation of E-cadherin, caused by genetic and epigenetic mechanism, is a frequent event in most type of epithelial carcinomas. In poorly-differentiated breast and gastric cancer, somatic mutations of E-cadherin are frequently found. Further, in all cases of familial gastric cancer, loss of E-cadherin is mainly caused by germline mutations[27]. In other types of tumors, mutations are infrequent events and repression of E-cadherin is mainly caused by aberrant promoter methylation.

In according to CC, mutations of E-cadherin are rare events in HCC. Reduced or loss of E-cadherin expression is mainly caused by aberrant CpG island methylation with a detectable frequency from 33% to 67%[28–30]. Wei et al[31] described, that loss of E-cadherin was closely associated with loss of heterozygosity (LOH) of E-cadherin and CpG hypermethylation. In precancerous conditions just as dysplastic nodules or liver tissue with chronic hepatitis or cirrhosis, aberrant E-cadherin methylation was detected in 8% and 46%, respectively[2930].

Other factors, except epigenetic inactivation or mutations, leading to inactivation of E-cadherin include transcriptional repression by binding of transcriptional factors, e.g. the repressors Snail or Sip-1 to CDH1-E box elements[3233].

But not only E-cadherin, as a member of cadherin genes, is epigenetically altered in HCC. Yamada et al[34] reported the highest methylation frequency of M-cadherin with a frequency to 55% among seven elucidated cadherin genes. Methylation-induced silencing of H-Cadherin (CDH13) was observed by Yu et al[17], reaching 21%.

TIMP-3: Tissue inhibitor of metalloproteinase-3 (TIMP-3) leads to inhibition of cell migration and angiogenesis. In human carcinoma cell lines, overexpression of TIMP-3 suppresses cell growth and induces apoptosis by stabilization of TNF-alpha receptors on the cell surface[3536]. A TIMP-3 downregulation was demonstrated in different kinds of tumors, mostly mediated by CpG island promoter methylation[6].

In HCC, TIMP-3 methylation is an infrequent event reaching 13% to 19%. No methylation was found in normal liver tissue. Lü et al[37] demonstrated TIMP-3 methylation in 25% of hepatocellular cancer emboli in portal veins. The aberrant promoter methylation is accompanied by loss or reduced TIMP-3 mRNA and protein expression[37–39].

TFPI-2:TFPI-2 is a Kunitz-type serine protease inhibitor that represses cellular invasion in several kinds of tumors, e.g. in lung cancer or pancreas carcinomas, by suppressing plasmin-mediated activation of MMP-1 and MMP-3 or inhibition of plasmin and trypsin activity[40–43].

Wong et al[44] observed TFPI-2 downregulation with reduced or loss mRNA transcript expression in HCC with a frequency of 90%. In 47% of the observed HCC, aberrant CpG methylation was seen, but not in normal liver tissue. In HCC cell lines with epigenetically induced silencing, a TFPI-2 mRNA transcript re-expression was induced by combined treatment with the demethylating agent 5'-AZA-DC and the histone deacetylase inhibitor TSA.

DNA repair

Mismatch repair system: Defects in DNA repair mechanisms may result in accumulation of mutations and genomic instability. The mismatch repair system (MMR) is one of the most important DNA repair mechanisms correcting errors in DNA replication. Defects of the MMR leading to microsatellite instabilities (MSI) have been observed in approximately 15% of sporadic colorectal and gastric carcinomas[4546]. Promoter methylation of MMR genes in HCC occurred with a frequency of 5% to 13% for hMLH1, 68% for hMSH2 and 75% for hMSH3. A high methylation frequency of hMSH2 and hMSH3 was observed in HCC corresponding non neoplastic liver tissue, especially in cirrhotic liver tissue, reaching 55 % and 70%, but not in normal liver tissue. No correlation was found neither to viral hepatitis nor to MSI status and DNA methylation of analyzed MMR genes[47–50].

MGMT (O6-methylguanine DNA methyltransferase): O6-methylguanine DNA methyltransferase (MGMT) is another important DNA repair gene with the highest activity in the liver[51]. MGMT protects cells from DNA damage caused by mutagenic and cytotoxic agents leading to alkylation at O6-guanine. Loss or reduced MGMT expression due to CpG islands methylation was detected in several kinds of human cancers[52]. In HCC, aberrant methylation occurred with a frequency of 22% to 39%, whereas the MGMT promoter shows higher methylation levels in chronic viral hepatitis associated HCC[53–55]. Interestingly, Su et al[53] reported that MGMT promoter methylation occurred to a similar extent in non neoplastic liver tissue compared to HCC.

GSTP1 (Glutathione S-transferase P1): The detoxi-ficating glutathione S-transferase P1 (GSTP1) gene protects cells from cytotoxic and carcinogenic influences in due to inactivation of electrophilic carcinogens by conjugation with glutathione. Promoter methylation of GSTP1 is best analyzed in prostate cancer. GSTP1 methylation is an early event in prostatic carcinogenesis, because in high-grade prostatic intraepithelial neoplasia loss of GSTP1 expression is caused by DNA methylation. Many other tumor types including breast cancer and cholangiocarcinoma showed a GSTP1 hypermethylated promoter[5657]. In HCC, methylation of the GSTP1 gene occurred in 41% to 85%[5358–61]. Zhang elucidated GSTP1 methylation in HCC in presence of environmental chemical carcinogens. A significant correlation was observed with higher aflatoxin B₁ (AFB₁)-DNA adducts in tumor tissue in contrast to tumor tissue without or lower levels of AFB₁-DNA adducts. However, no association was found between GSTP1 methylation and polycyclic aromatic hydrocarbon-DNA adducts[59]. So far, aberrant methylation of GSTP1 is not only detectable in tumor tissue, Wang et al[62] observed a GSTP1 hypermethylation in serum of HCC patients.

Suppressors of cytokine signaling (SOCS): Suppressors of cytokine signaling 1 and 3 (SOCS-1 and SOCS-3) are intracellular proteins that act as negative regulators of Janus kinase (JAK) and signal transducer and activators of signaling pathways (STAT). The JAK/STAT signaling pathway plays an important role in cell growth and differentiation or immune reaction mediated by cytokines. Cytokines activate JAK's by binding to membrane receptors that leads to phosphorylation of STAT's and activates target genes. SOCS1 and SOCS-3 bind direct and indirect to JAK's and inhibit the phosphorylation of STAT's and activation of target genes[6364].

Aberrant methylation of SOCS-1 and SOCS-3 promoter sequence has been reported in several kinds of human cancer. SOCS-1 and SOCS-3 CpG island hypermethylation is an early event in human carcinogenesis. Recently, we have shown methylation-induced downregulation of SOCS-1 and SOCS-3 in precursor lesions of Barrett’s adenocarcinomas and precursor lesions of squamous carcinomas of head and neck[6566]. In HCC, aberrant promoter methylation of SOCS-1 and SOCS-3 occurred with a frequency of 60% and 33%, respectively[6768]. Methylation of the SOCS-1 gene was detected in HCV-induced chronic hepatitis and liver cirrhosis, reaching 45%, whereas the methylation frequency increased with fibrosis stage with the highest proportion in liver cirrhosis[6970].

Methylation hot spot 3p

The short arm of human chromosome 3 belongs to regional methylation hot spots in addition to chromosomal locus 11p and 17p. Alterations of the genetic information on chromosome 3 are one of the most frequent and earliest steps in the carcinogenesis of several types of tumors. LOH of chromosome 3p occurred in about 30% of hepatocellular carcinomas[71].

In different kinds of human cancer, epigenetic inactivation via promoter methylation of several genes located on 3p, including RASSF1A on 3p21.3, hMLH1 at 3p21.3, RARβ2 at 3p24.2, was shown.

One of the most frequent observed and most epigenetically inactivated genes of 3p is RASSF1A, a multifunctional tumor suppressor gene that protects cells from genomic instability and transformation by stabilizing the microtubules[7273]. An aberrant promoter methylation was detected in about 50% of malignant tumors. In renal cell carcinoma and small cell lung cancer, the highest prevalence was observed, reaching about 91%[74]. In HCC, hypermethylation occurred in approximately 54% to 95%, whereas HBV-associated HCC showed higher levels of RASSF1A methylation compared to HCC without risk factors. RASSF1A methylation occurs not only in HCC, methylation is even observed in non-neoplastic precancerous conditions like cirrhotic liver and chronic hepatitis[71].

Semaphorin 3B (SEMA3B) and BLU are two other putative tumour suppressor genes located on 3p21.3, whereas the function of BLU still remains unclear. In lung cancer, BLU overexpression inhibits tumor colony formation efficiency. Qiu et al[77] reported that BLU might function as an environmental stress-responsive gene, regulated by E2F, at least in nasopharyngeal carcinomas. However, BLU methylation is a rare event in human cancer. We detected BLU promoter methylation in about 20% of our examined HCC[7178].

SEMA3B, a member of the Semaphorin family, suppresses tumor formation in lung cancer and induces apoptosis. It has been demonstrated that SEMA3B induced apoptosis is antagonized by VEGF¹⁶⁵ in due to an interaction with NP-1 receptor[7980].

Aberrant methylation of SEMA3B was detected in lung cancer and gliomas[8182]. We reported a high prevalence of SEMA3B methylation in HCC, reaching 80%. In contrast, the tumor surrounding non-neoplastic liver exhibited an unmethylated SEMA3B promoter. Further, RASSF1A and SEMA3B expression was restored by treatment with the demethylating drug 5-AZA-C in HCC cell lines, suggesting that promoter hypermethylation is responsible for silencing transcript expression[717281].

The fragile histidine triad (FHIT) gene, located to 3p14.2, embraces FRA3b, the most actively fragile site in humans[8384]. Functional and structural alterations of FHIT were identified in several kinds of human cancer. Methylation induced silencing was described in lung and breast cancers[85]. In HCC, promoter methylation of FHIT is a frequent and early event. Sun et al[86] observed FHIT hypermethylation with a frequency of 71% in HCC, 64% in non neoplastic liver tissue and 14 % in normal liver.

CpG island methylator phenotype (CIMP)

Carcinomas with high rates of accumulated aberrant promoter methylation of tumor-related genes are characterized as CIMP⁺ (CpG island methylator phenotype). CIMP⁺ was first described for colorectal and gastric cancer by Toyota and Issa in 1999[8788]. Shen et al[89] reported that CIMP⁺ is associated with environmental exposures in HCC. HCC from patients without precancerous conditions or risk factors, respectively, showed significantly lower levels of methylation than HCC arising from patients with chronic hepatitis B and C or patients with cirrhosis.

CIMP positive HCC (tumours with five genes that are concordantly methylated) showed a significantly association with methylation of the TSG p14, p15, p16 ER, RASSF1A or WT1 and elevated serum alpha-fetoprotein (AFP) levels. Further, CIMP⁺ was commonly seen in HCC with increased serum AFP levels[9091].

DNA-methyltransferases (DNMT)

DNA hypermethylation is catalyzed by the family of DNA methyltransferases (DNMT) including DNMT1, DNMT3a and DNMT3b. DNMT1 is required for maintenance of DNA methylation whereas DNMT3a and DNMT3b function as de novo DNA methyltransferases[92–94]. DNMT2 was former described as DNMT because of its strong similarity with m5C methyltransferases of pro- and eukaryotes. But it was recently shown that DNMT2 does not methylate DNA. It’s the first described RNA cytosine methyltransferase that methylate position 38 in Aspartic acid transfer RNA[95].

In human cancer just as in HCC, an upregulation of DNMT activity is seen in contrast to global hypomethylation. Park et al[96] described a significantly overexpression of DNMT1 and DNMT3b in HCC compared to non-neoplastic liver tissue. DNMT3a showed similar or higher expression levels. Saito et al[97] observed higher expression levels of all three DNMTs in HCC and cirrhotic liver than in normal liver. Increased DNMT1 and DNMT3a expression was also reported in dysplastic nodules[98].

According to other tumors, no correlation was seen between DNMT upregulation and promoter hypermethylation-induced inactivation of tumor-related genes. The certain mechanisms of DNMT upregulation remains still unclear, but it is suggested that aberrant DNMT activity, especially of DNMT1, is due to rapid proliferation of cancer cells because DNMT1 binds to proliferating cell nuclear antigen (PCNA)[99–101].

DNA hypomethylation on pericentromeric satellite regions results in chromosomal instability. During hepatocarcinogenesis, DNA hypomethylation of these regions was reported in HCC and precancerous conditions. The splice variant of DNMT3b, DNMT3b4 that may lack DNA methyltransferase activity is associated with DNA hypomethylation on pericentromeric satellite regions. Saito et al[102] reported that overexpresssion of DNMT3b4 was seen in cirrhotic liver, chronic hepatitis and HCC whereas increased DNMT3b4 levels correlated with DNA hypomethylation on pericentromeric satellite regions.

DNA HYPOMETHYLATION

In human cancer, global DNA hypomethylation leads to genomic instability, affects repeated DNA sequences, tissue-specific genes and proto-oncogenes or causes loss of imprinting with a biallelic expression, just as in case of IGF2. Further, the level of DNA hypomethylation increases with tumor progression[103104]. In recent years, DNA hypomethylation was shown in several human cancer and some premalignant alterations, i.e. colorectal adenomas and carcinomas, adenocarcinoma of prostate, breast cancer or intestinal type of gastric carcinoma and hepatocellular carcinoma, respectively[105–108].

Lin et al[109] observed 5-methylcytosine (m5C) content in hepatocellular carcinogenesis by comparing hepatocellular carcinoma with non neoplastic liver, including cirrhotic livers. In all cancer tissues, 5-methylcytosine was significantly reduced. No difference of the m5C content was detected in cirrhotic and non cirrhotic liver tissue. The reduced 5mC level was associated with large tumor size and poorly histopathological grade.

It is suggested that (re-)activation of retroposons might be associated with global hypomethylation because approximately 90% of all m5C lies in these elements. An association between hypomethylation and transposon activation, especially of LINE-1 transposons, has been observed in human testicular carcinoma cell lines, urothelial carcinoma cell lines and teratocarcinoma cell lines[110–112]. But in HCC, an activation of transposable elements just as LINE-1 retrotransposons via hypomethylation could not be detected yet[109].

HISTONE MODIFICATION

Histone modifications are strongly associated with formation of the nucleosome structure and are closely linked to CpG island methylation by interacting with Methyl-CpG-binding proteins (MBD's) and DNA methyltransferases (DNMT's). Modifications including methylation, acetylation or phosphorylation of certain position of the histone tails. Whereas histone methylation is associated either to activation or to repression, histone hypoacetylation mediated by histone deacetylases leads mostly to DNA relaxation and subsequent accessibility for transcriptional factors with repression of transcription.

Lee et al[113] reported that HCC with low survival expressed higher levels of genes involved in histone modifications just as PTMA and SET, two proteins that are members of inhibitors of histoneacetyltransferases complex.

P73, a member of the TP53 family represses AFP expression during normal hepatic development by chromatin structure alterations. In hepatoma cells, transactivated p73 suppresses endogenous AFP transcription via reducing of acetylated histone H3 lysine 9 and increasing dimethylated histone H3 lysine 9[114].

CONCLUSION

In hepatocarcinogenesis, aberrant methylation of tumor related genes occurs not only in advanced tumour stages, it’s a frequent and early event. Promoter methylation of different kinds of tumor suppressor genes including p16, SOCS1 and SOCS3 or RASSF1A, has been demonstrated in premalignant conditions just as chronic hepatitis or liver cirrhosis. Moreover, the frequency of aberrant promoter methylation increases during the progression from precancerous lesion to HCC. In HBV or HCV-associated chronic hepatitis, methylation frequency of detected genes is significantly higher than in non-neoplastic non-viral liver tissue. Therefore, epigenetic changes in preneoplastic or early neoplastic stages may serve as indicator or “biomarker” for screening of patients with an increased risk for HCC.

Further, HCC is one of the most common causes of cancer death worldwide with a poor prognosis. Only few therapeutic interventions exist. It was demonstrated that re-expression of tumor suppressor genes that are epigenetically silenced is possible by using demethylating and histone modifying agents. In the next years, this might be a possible therapeutic approach analogue to other malignant diseases, e.g. myelodysplastic syndrome, but the used therapeutic agents that influence DNA hypermethylation are toxic and lead to genome wide alteration of the methylation pattern with possible activating of oncogenes or imprinted genes. Another possible aspect of chemotherapy might be to modulate the epigenetically involved pathways by using small molecules that are more specific. But further investigations in clinical trails are needed to prove and integrate epigenetic pathway modulating agents.