We determined the methylation status by the MethyLight assay of ten genes in both normal and malignant liver tissues. We have identified three genes that were frequently methylated (in approximately 50% of patients) in normal liver tissues. Five additional genes were significantly more frequently methylated in HCC tissues compared to normal liver tissues. Finally, we observed that methylation of CDKN2A was frequent in HCV, but not HBV-associated HCC. These data suggest that epigenetic alterations play an important role in hepatocarcinogenesis, with different pathways affected in HBV or HCV associated HCCs.
In analyzing our DNA hypermethylation results, we dichotomized semi-quantitative MethyLight data in two different ways: as positive for any hypermethylation at PMR > 0% and as positive for high levels of hypermethylation at PMR ≥ 4% (data not shown). Earlier studies have reported that PMR > 4% is associated with loss of gene expression and best discriminates between normal and malignant or premalignant tissues (Eads et al., 2000
; Ogino et al., 2006
). However, other researchers have used qualitative PCR methods that report samples as positive with any observable hypermethylation signal (Virmani et al., 2003
), while other researchers have used three PMR intervals in their analyses, with no methylation at PMR = 0, low methylation at PMR < median for a particular gene, and high methylation at PMR > median for that gene (Tsou et al., 2005
). We have previously confirmed that low levels of hypermethylated gene are present in samples with 0% < PMR < 4% (unpublished data), but the biological significance of this level of methylation is unclear. In this present study, we performed univariate and multivariate analyses using PMR > 0% as the criterion for hypermethylation positivity. Our basic conclusions remained the same, regardless of the PMR cutoff chosen.
Previous studies examining methylation of tumor suppressor genes in normal liver tissues have reported conflicting results, with most studies reporting no or low methylation of several tumor suppressor genes in normal liver tissues (Lee et al., 2003
), but several studies reporting high frequency of DNA methylation of these genes in normal liver tissues. For example, Lehmann et al reported high frequency of DNA methylation of APC, CCND2, GSTP1, RASSF1A, SOCS-1 in normal liver biopsies obtained from organ donors, although the methylation levels were lower than in HCC samples (Lehmann et al., 2005
; Lehmann et al., 2007
). Harder et al reported 100% methylation of APC but no methylation of GSTP1 in 16 normal liver tissues (Harder et al., 2008
). Although Nishida et al identified a group of seven genes (HIC-1, CASP8, SOCS-1, RASSF1A, p16, and APC) that was frequently methylated in normal liver tissues, their normal liver tissues were obtained either from colon cancer patients who had hepatic metastasis, or from patients who suffered from focal nodular hyperplasia, hepatic hemangioma, or hepatic adenoma (Nishida et al., 2008
). In the present study, we observed high frequency of methylation of APC (48%), CCND2 (56%), and GSTP1 (52%) in normal liver tissues, even in the absence of HBV or HCV infection. Similar to studies by Lehmann et al reported for these three genes (Lehmann et al., 2005
), we reported here that methylation levels in normal liver tissues were lower than in liver cancer tissues. Among samples that were methylated for the specific gene, the average PMR in cancerous tissues was significantly higher for APC (p=0.01) and GSTP1 (p=0.0054) than in normal liver tissues. These methylation changes present in normal tissues might represent the consequence of accumulated environmental exposures, including both cytotoxic and carcinogenic chemicals being detoxified in liver. Indeed, Zhang et al reported that methylation of GSTP1, the key enzyme involved in detoxifying aflatoxin B1, was correlated with the aflatoxin-DNA adduct levels in the HCC tissues (Zhang et al., 2005
). However, in the present study we did not see a trend of increasing methylation with age among normal liver tissues, and aflatoxin levels were not available. Alternatively, these methylation changes could represent tissue-dependent differentially methylated genes, which have begun to be elucidated by several genome-wide profiling analyses on normal tissues (Kitamura et al., 2007
; Nagase and Ghosh, 2008
; Weber et al., 2007
Previous studies also tried to identify specific DNA methylation patterns associated with various risk factors in hepatocarcinogenesis, including viral infections, aflatoxin exposure and alcohol consumption. Methylation of CTGF, RARB, E-cadherin and p73 was more frequently seen in HBV-associated HCCs than in HCV-associated HCCs in several studies (Chiba et al., 2005
; Yang et al., 2003
), while RUNX3, APC, SOCS-1 and p14 were preferentially methylated in HCV-HCC (Mori et al., 2005
; Yang et al., 2003
). Several recent studies also linked environmental exposures to specific DNA methylation patterns. High frequencies of p16, GSTP1, MGMT and RASSF1 methylation were significantly associated with high level of AFB1-DNA adducts in HCC tumors (Zhang et al., 2002
; Zhang et al., 2005
; Zhang et al., 2003
; Zhang et al., 2006
). CDKN2A methylation has been shown to be present in early stages of HBV-associated hepatocarcinogenesis, not only in high frequency in HCCs, but also was in cirrhotic nodules (CNs) and dysplastic nodules (DNs), known precursor lesions of HCCs (Shim et al., 2003
). Further, p16 methylation has been shown to preferentially occur in liver tissues with HBV infections compared to liver tissues without HBV infection (Jicai et al., 2006
). Although a few studies did not detect differences of p16 methylation between HBV-HCCs and HCV-HCCs (Fukai et al., 2005
; Kaneto et al., 2001
), several studies consistently observed higher frequency (although not significant) of p16 methylation in HCV-HCCs than HBV-HCCs (Katoh et al., 2006
; Li et al., 2004
; Narimatsu et al., 2004
). In the present study, we did not detect any CDKN2A methylation in normal liver tissues, but methylation of CDKN2A was significantly higher in HCV-HCCs than in HBV-HCCs. In addition, we observed higher frequencies of methylation of RASSF1, SFRP1 and HOXA9 in HBV-HCCs than in HCV-HCCs, but these differences did not attain statistical significance.
Although our retrospective case-control study suggests that different pathways are preferentially inactivated epigenetically in HCCs caused by different etiologies, our current study has several limitations. Although our samples are all from patients enrolled at either University of Washington Medical Center or Harborview Medical Center, 64% HCV-HCC patients were Caucasians, while 75% of HBV-HCC patients were Asian. As virus type was strongly correlated with race, we can not rule out the possibility that the methylation differences we observed with respect to virus type are not confounded by the different genetic backgrounds in the HBV vs. HCV-associated cancers. Further, our sample size was small, especially for the group of HBV-HCC patients. This is due to the low frequency of HBV infection in developed countries. Our conclusions need to be further confirmed in larger retrospective studies in developing countries where both HBV and HCV infections are prevalent, where the effects of race, age and gender can be adequately adjusted for. Future prospective studies need to address these limitations to definitively delineate mechanistic pathways of hepatocarcinogenesis in relationship with environmental risk factors.
Hepatocellular carcinoma is one of the most common malignancies worldwide with very poor prognosis. Elucidating the underlying molecular mechanisms in relationship with various etiological agents is a necessary step for the development of novel and effective treatments that will ultimately improve patients’ survival.