CpG methylation of MGMT and hMLH1 promoter in hepatocellular carcinoma associated with hepatitis viral infection

Inactivations of DNA repair genes, O6-methylguanine-DNA methyltransferase (MGMT) and hMLH1, by promoter hypermethylation have been reported in several types of primary human neoplasia. This epigenetic inactivation mechanism remains elusive in hepatocellular carcinoma (HCC). To investigate the relation between the expression of MGMT and hMLH1 and the CpG methylation within their promoters in HCCs with or without hepatitis viral infection, we performed immunohistochemistry and urea/bisulphite sequencing on 46 HCCs, corresponding noncancerous tissues, and 20 normal liver tissues. MGMT- and hMLH1-negative HCCs were 60.9% (28 out of 46) and 21.8% (10 out of 46), respectively. HCCs lacking both proteins were 10.9% (five out of 46). The frequency and extent of CpG methylation in the MGMT promoter increased along with hepatitis viral infection and pathological progression. MGMT-negative tumours showed very frequent and widespread methylation in the promoter compared with MGMT-positive tumours. Half of the hMLH1-negative HCCs showed promoter hypermethylation. These data suggested that MGMT gene silencing in a subset of HCCs was likely caused by epigenetic alteration, such as promoter hypermethylation, and that the promoter hypermethylation silenced the hMLH1 gene in half of the hMLH1-negative tumours. A correlation between the promoter methylation status and viral infection, although it was weak, intimated that hepatitis viral infections could play a role in the CpG methylation of the MGMT promoter.

Inactivations of DNA repair genes, O 6 -methylguanine-DNA methyltransferase (MGMT) and hMLH1, by promoter hypermethylation have been reported in several types of primary human neoplasia. This epigenetic inactivation mechanism remains elusive in hepatocellular carcinoma (HCC). To investigate the relation between the expression of MGMT and hMLH1 and the CpG methylation within their promoters in HCCs with or without hepatitis viral infection, we performed immunohistochemistry and urea/bisulphite sequencing on 46 HCCs, corresponding noncancerous tissues, and 20 normal liver tissues. MGMT-and hMLH1-negative HCCs were 60.9% (28 out of 46) and 21.8% (10 out of 46), respectively. HCCs lacking both proteins were 10.9% (five out of 46). The frequency and extent of CpG methylation in the MGMT promoter increased along with hepatitis viral infection and pathological progression. MGMT-negative tumours showed very frequent and widespread methylation in the promoter compared with MGMT-positive tumours. Half of the hMLH1-negative HCCs showed promoter hypermethylation. These data suggested that MGMT gene silencing in a subset of HCCs was likely caused by epigenetic alteration, such as promoter hypermethylation, and that the promoter hypermethylation silenced the hMLH1 gene in half of the hMLH1-negative tumours. A correlation between the promoter methylation status and viral infection, although it was weak, intimated that hepatitis viral infections could play a role in the CpG methylation of the MGMT promoter. Hepatocellular carcinoma (HCC) is one of the most frequent human malignancies and a major cause of cancer-related deaths worldwide (Okuda et al, 1992). Most HCCs exhibit characteristics compatible with chronic hepatitis and cirrhosis caused by persistent infection of hepatitis B virus (HBV) and/or hepatitis C virus (HCV) (Okuda et al, 1992). Both chronic hepatitis and cirrhosis associated with viral infection are considered as precancerous conditions (Paradis et al, 1998). The process of chronic inflammation or cirrhosis initiates clonal expansion and facilitates regeneration of hepatocytes (Aihara et al, 1994). During neoplastic degenerative change, an accumulation of genetic alterations or epigenetic changes may occur. However, the molecular mechanisms of hepatocarcinogenesis associated with hepatitis viral infection have not been clarified.
O 6 -methylguanine-DNA methyltransferase (MGMT) is a DNA repair enzyme that plays an important role in the defence against the carcinogenic and cytotoxic effects of alkylating agents in cellular DNA (Pegg, 1990). Since ubiquitous and environmental alkylating agents such as N-nitroso compounds are principally metabolised and activated in mammalian hepatocytes, liver tissue is persistently exposed to activated alkylating agents (Gerson et al, 1986). It was demonstrated that Mgmt-targeted mice (Mgmt À/À ) treated with alkylating agents generated hepatocellular carcinoma (Iwakuma et al, 1997). Major and Collier (1998) suggested that MGMT protein activity decreased in chronic hepatitis, cirrhosis, and HCCs. We have recently reported that the MGMT expression was frequently lost in a variety of human tumours and was a significant prognostic factor (Matsukura et al, 2001). Since loss of MGMT expression was not commonly because of a genetic change, it has been suggested that another cause, such as epigenetic change, is involved (Esteller et al, 1999;Bhakat and Mitra, 2000).
Mismatch repair system (MMR) is an essential system by which cells correct errors in DNA replication during proliferation to maintain the fidelity of the genome (Lahue et al, 1989). One of the MMR genes, hMLH1, has been demonstrated to play a pivotal role in DNA MMR (Papadopoulos et al, 1994). In addition, the important association between MGMT and MMR in DNA repair was pointed out by Frayling (1999). Furthermore, it was shown that Mgmt À/À Mlh1 À/À (double knockout) mice  treated with alkylating agents exhibited high susceptibility to carcinoma (Kawate et al, 1998). It is, however, unclear whether tumours expressing neither of the genes exist in human HCCs. It has been proposed that aberrant DNA methylation of CpG islands in the promoter region is correlated with inactivation of tumour suppressor genes in human cancer. Esteller et al. (2001) have demonstrated the reduced expression of tumour suppressor genes such as p16, MGMT, and hMLH1 by promoter hypermethylation in several human neoplasias, and have suggested that this epigenetic change might be an early event in the pathogenesis of several human tumours. Recently, the correlation of p16 promoter hypermethylation with chronic hepatitis and cirrhosis associated with HBV or HCV infection has been reported (Kaneto et al, 2001).
Here, we report the existence of human HCCs lacking both MGMT and hMLH1 proteins, the relation between HCCs associated with hepatitis viral infection and detailed CpG methylation status of MGMT and hMLH1 promoter regions, and the specific CpG methylation pattern of MGMT-and hMLH1negative tumours.

Tissue specimens
A total of 46, HCCs and adjacent noncancerous liver tissues (mean age 63.8 years; 35 males and 11 females; seven HBV positives, 33 HCV positives, one HBV/HCV positive and five HBV/HCV negatives) and 20 normal liver (mean age 59.9 years; 14 males and six females) tissue specimens were obtained surgically and frozen at À801C. All specimens were subjected to pathological diagnosis. DNAs of all tumour samples were extracted from pathologically obvious cancerous regions in the resected liver. HBV and HCV infections were diagnosed serologically with HBs antigen (LPIA-200; Diatron Laboratories, Tokyo, Japan) and anti-HCV antibody (Immunocheck-HCV Ab; International Reagent, Kobe, Japan), respectively. Informed consent was obtained from all patients.

Anti-MGMT antibody
Polyclonal rabbit antibodies against human MGMT protein were prepared using TrpE fusion protein, as described (Nakabeppu and Nathans, 1991). Escherichia coli BL21 (DE3) carrying pET3d:TrpE-hMGMT-1 that encodes the TrpE polypeptide fused to a region of MGMT (residues 1 -45) at the C terminus was used to produce each fusion protein (Studier et al, 1990), and polyclonal antibodies against the fusion protein were raised in rabbits. The serum was initially applied to a TrpE-hMGMT-1-coupled column; their bound materials were eluted at pH 2.3 and dialysed against 10 mM Tris-HCl (pH 7.4) and 150 mM NaCl. To increase antibody specificity, the eluted fraction was applied to an affinity column with TrpE-mMGMT-1, in which a corresponding region of mouse MGMT (residues 1 -58) was fused to TrpE (Kawate et al, 1995), as a ligand. Then the bound fraction was eluted and dialysed. This fraction was used as an anti-MGMT antibody. The specificity of the antibody has been reported previously (Matsukura et al, 2001).

Immunohistochemistry
Immunohistochemical studies for MGMT (Matsukura et al, 2001) and hMLH1 (Wang et al, 2001) were performed as described previously. In the present study, mouse monoclonal antibody against hMLH1 protein (clone G168-728; PharMingen, San Diego, CA, USA) (Wang et al, 2001) was used. Positive staining was identified by the presence of brown staining in the nucleus and/or cytoplasm. MGMT and hMLH1 expression were evaluated as positive if the distribution of stained cells was more than 10% of cancer cells. The expression status of MGMT and hMLH1 was assessed by two pathologists without a knowledge of the clinicopathological features of the case or the clinical outcome.

Urea/bisulphite modification of DNA and PCR amplification
The urea/bisulphite treatment of genomic DNA was performed as described by Paulin et al (1998). The modified DNA was resuspended in 20 ml of water and immediately subjected to PCR or stored at -201C. The whole CpG island in the MGMT promoter region was amplified by nested PCR (Figure 2A). Primer sequences and PCR conditions are presented in Table 1. The first round of amplification was performed with 50 ng of the bisulphite-treated DNA. Then 1/1000th of the first PCR product was subjected to the second round PCR. The size of the nested PCR product was 835 bp. The CpG island of hMLH1 was also amplified by nested PCR (Figure 2B). The size of the nested PCR product was 815 bp.

Cloning of PCR product and sequencing
The amplicons were cloned into the pSTBlue-1 Acceptor TM Vector (Novagen Inc., Madison, WI, USA) to transform competent JM109 cells (Wako Pure Chemical Industries, Ltd, Osaka, Japan). The plasmid DNAs were amplified by colony PCR reactions (Table 1). A total of 10 of the colony PCR products from each sample were sequenced.

Statistical analysis
Differences among groups (normal liver tissues vs chronic hepatitis, liver cirrhosis and HCC, or MGMT-negative vs -positive HCCs) were tested by one-way analysis of variance (ANOVA). Probability levels of o0.05 were considered statistically significant.

Immunohistochemistry of MGMT and hMLH1
Immunohistochemical analysis was performed on normal liver, noncancerous tissues, and HCCs to examine whether MGMT and hMLH1 were expressed. Expression of these proteins was detected in cells from normal liver, as well as in chronic hepatitis and liver cirrhosis, including hepatocytes, bile duct cells, vascular endothelial cells, smooth muscle, and so on (data not shown). We used bile duct cells as an internal positive control within the sections because these cells in cancerous regions also commonly expressed MGMT and hMLH1 ( Figure 1B, D). The signals of MGMT-positive tumour cells were as intense as those of the bile duct cells ( Figure 1A), while those of MGMT-negative tumour cells were virtually undetectable ( Figure 1B). As for hMLH1, signal intensity similar to those of bile duct cells was detected in hMLH1-positive tumour cells ( Figure 1C), but not in hMLH1-negative tumours ( Figure 1D). MGMT-and hMLH1negative tumours were 60.9% (28 out of 46) and 21.8% (10 out of 46), respectively. Tumours lacking both proteins were 10.9% (5 out of 46) of all tumours. The results of immunohistochemistry and the characteristics of patients with HCC are shown in Table 2.

Methylation analysis of MGMT promoter region in HCCs, noncancerous, and normal liver tissues
Next, we performed urea/bisulphite sequencing to investigate the methylation status of MGMT and hMLH1 promoter region in HCCs, noncancerous tissues, and normal liver. Prior to the study of primary samples, to confirm that the nested PCR we used could accurately reflect the methylation status of genomic DNA, we performed PCR using DNA solutions containing methylated and unmethylated genomic DNAs with different ratios. The methylated and unmethylated DNAs were extracted from an MGMT-deficient cell line, SW48 (Aquilina et al, 1998), in which the MGMT promoter was fully methylated and an MGMT-proficient cell line, HepG2 (Fritz and Kaina, 1992), in which the promoter was not methylated at all. We found that the ratio of methylated to unmethylated clones obtained from the nested PCR -cloningsequencing was consistent with the theoretical ratio (data not shown). The CpG island of MGMT including 97 CpG sites is shown in Figure 2A. We examined the detailed methylation status of all CpG sites by urea/bisulphite DNA sequencing in 46 HCCs ( Figure 3A), corresponding noncancerous tissues ( Figure 3B), and 20 normal liver tissues ( Figure 3C). The methylation was somewhat observed at the 5 0 border of the CpG island irrespective of cancerous or noncancerous samples. These methylated CpGs in MGMT-positive tumours, noncancerous tissues except for 4NC and 5NC, and normal tissues never extended beyond the first nine CpGs. However, a subset of MGMT-negative HCCs, 1C -15C, showed that the methylation extended in the 3 0 direction beyond the first nine CpGs. Tumour 1C showed all 97 hypermethylated CpG sites in the island (data not shown). Other MGMT-negative tumours, 16C -28C, showed a similar extent of methylation to MGMTpositive tumours and normal tissues.
From the point of view of hepatitis viral infection, the methylation frequency at particular CpG sites in chronic hepatitis, liver cirrhosis, and HCC was higher than that in normal liver with statistical significance ( Figure 4A). The frequency at those sites increased along with pathological progression. Furthermore, the methylation also extended towards downstream along with pathological progression ( Figure 4A). These data suggested that hepatitis viral infection might be involved in methylation of the MGMT promoter region. As for MGMT expression, MGMTnegative tumours showed high frequency and widespread methylation compared with MGMT-positive tumours and normal liver ( Figure 4B). There were significant statistical differences in the methylation frequency at many CpG sites between MGMT-negative and -positive HCCs. This result suggested that hypermethylation of the MGMT promoter might play some role in the lack of MGMT expression. Although we did not use the microdissection method, our results might be of less estimation because of the existence of nontumour DNA. If it is removed, further significant differences could be obtained.

Methylation analysis of hMLH1 promoter region in HCCs, noncancerous, and normal liver tissues
The CpG island of hMLH1 including 67 CpG sites is shown in Figure 2B. All of the CpG sites were also analysed by urea/ bisulphite sequencing. In all, 10 hMLH1-negative HCCs were detected by immunohistochemistry, and five of these showed promoter hypermethylation ( Figure 5). The methylation patterns of these tumours were particular, that is, the middle portion of the CpG island was methylated and the methylation frequency of each CpG site in the region was approximately 60%. There was no methylated CpG in hMLH1-positive HCCs, noncancerous regions, and normal liver tissues (data not shown).

DISCUSSION
To date, it is thought that inactivation of the MGMT gene is because of epigenetic changes such as DNA methylation (Esteller et al, 1999(Esteller et al, , 2001 or some other unknown mechanisms because the genetic change is uncommon. Although the methylation of MGMT and hMLH1 promoter was analysed previously in HCCs, the method used was mostly methylation-specific PCR (MSP) (Esteller et al, 2001). No study has identified critical sites by a survey of detailed methylation of the promoter regions of these genes. This is the first study to examine the detailed CpG methylation status of MGMT and hMLH1 promoter regions in HCCs, their adjacent tissues, and normal liver tissues.
We compared MGMT-negative and -positive HCCs in order to determine the critical CpG site for the MGMT silencing. A subset of   (Harris et al, 1994) has not been reported in liver tumours by MSP (Esteller et al, 2001), and we also could not find any methylated CpG in the enhancer, except for sample 1C. It was reported that a single-site methylation upstream of the p53 promoter, not in the enhancer (Lozano and Levine, 1991), reduced its expression during hepatocarcinogenesis in vivo (Pogribny et al, 2000). Therefore, it is possible that the methylation of specific CpG site(s) in the MGMT promoter, aside from the enhancer, could result in downregulation of MGMT gene expression in HCCs. Since we found that some MGMT-negative tumours did not show the very frequent and widespread methylation status, not only DNA methylation but also histone deacetylation, chromatin remodelling, post-transcriptional, and post-translational inactivation might be correlated with the MGMT deficiency. MGMT is converted to an inactive form after removing the methyl group from O 6 -methylguanine (Ishibashi et al, 1994). The inactivated MGMT is not degraded but remains in an immunoreactive state in normal cells (Liu et al, 2001), whereas it is degraded rapidly via the ubiquitin proteolytic pathway in tumour cells (Srivenugopal et al, 1996;Liu et al, 2001). In the light of this, some tumours that correspond to HCCs, 16C -28C, might be regarded as MGMTnegative because of the rapid degradation of the inactive form, although the gene might have been transcribed. Our careful analysis revealed that both the frequency and extent of MGMT methylation increased along with pathological progression. Although little is known of the link between hepatitis viral infections and the methylation machinery of endogenous genes, our findings intimated that CpG methylation of the MGMT gene was associated with hepatitis viral infections. The levels of DNA methyltransferase mRNA in chronic hepatitis, cirrhosis, and HCCs associated with hepatitis viral infections increased compared with those in normal liver tissues (Sun et al, 1997). Human immunodeficiency virus induced the methylation of interferon g (IFNG) through increased DNA methyltransferase activity (Mikovits et al, 1998). Therefore, CpG methylation of the MGMT promoter could possibly be caused by elevated DNA methyltransferase activity in hepatocarcinogenesis. In contrast to the MGMT, there was no methylation in hMLH1-positive HCCs, noncancerous, and normal liver tissues with or without hepatitis viral infection (data not shown), suggesting no relation between hMLH1 promoter methylation and viral infections. The methylation mechanism for hMLH1 should be different from that for MGMT. In the present study, the frequency of hMLH1 promoter hypermethylation in HCCs was 10.9% (five out of 46) -half of hMLH1-negative HCCs -and consistent with the previous finding, 10.0% (two out of 20) (Esteller et al, 2001). There was no promoter methylation in the remaining five hMLH1-negative HCCs. In colon cancer, gene mutations were found in hMLH1-negative tumours without promoter hypermethylation (Cunningham et al, 1998). Although we did not investigate the gene mutations in those HCCs, genetic alterations would be involved in hMLH1-negative HCCs without promoter hypermethylation. Figure 5 Detailed CpG (No. 14 -61) methylation status of all analysed 67 CpGs within the hMLH1 promoter region in 10 hMLH1-negative HCCs. Each circle graph represents the percentage of methylated clones (number of methylated clones/10 analysed clones Â 100).