Main

Hepatoblastoma is the most common malignant liver tumor in childhood. Data from Surveillance Epidemiology and End Results (SEER) estimate an incidence of 0.5–1.5 cases per million children per year, which accounts for 79% of hepatic neoplasia in children younger than 15-y and for 1% of all cancers in this age group (1).

Few prognostic factors are available for hepatoblastoma. The most important features appear to be low alpha-fetoprotein level, pure fetal histology, gross resectability of the tumor, and the presence of metastases in the lung and small cell undifferentiated tumor type. In fact, complete surgical resection remains the most important intervention required for long-term survival (2,3).

Although some chromosomal alterations have been found in hepatoblastoma (1q, 2q, 7q, 8, 17q, and 20 amplifications), no etiological relationship or prognostic value could be attributed to these findings. The majority of tumors from patients with hepatoblastoma do not show any detectable chromosomal anomaly (2,4).

DNA methylation, catalyzed by DNA methyltransferases, involves the addition of a methyl group to the carbon 5 position of the cytosine ring in CpG dinucleotides (5). This is associated with several changes in chromatin structure, including the regulation of histone methylation and acetylation and the recruitment of proteins to the methylated sites, which usually leads to the obstruction of the promoter, hindering gene transcription and subsequently silencing the gene (6). There is increasing evidence that in addition to genetic aberrations, the epigenetic processes play a major role in carcinogenesis. Aberrant methylation (hypermethylation) of gene promoter regions is the most widely studied epigenetic abnormality in human malignancies. This epigenetic event acts as an alternative to mutations and deletions to disrupt tumor suppressor gene function (7).

Aberrant promoter methylation has been described for several genes in various malignant diseases, and each tumor type may have its own distinct pattern of methylation (8,9). It was reported previously that gene hypermethylation may occur in patients with hepatoblastoma. Harada et al. (10), analyzing aberrant promoter hypermethylation in pediatric tumors and cell lines, observed RASSF1A hypermethylation in 19% of hepatoblastoma specimens. Shim et al. (11) showed that 50% of 24 patients with hepatoblastoma had hypermethylated CDKN2A. Nagai et al. (12) found that CpG islands in promoter region of SOCS1 gene were hypermethylated in seven of 15 hepatoblastoma samples examined. It is important to note that all of these studies focused on few genes and were conducted using conventional methylation-specific PCR (MSP) technique.

Therefore, the aims of this study were a) to determine the promoter methylation status of 25 genes in hepatoblastoma using the real-time quantitative MSP (QMSP) technology, which is more sensitive than conventional MSP, b) to compare the methylation status with clinical characteristics of the patients, and c) to evaluate if this kind of molecular analysis could be useful as a prognostic marker.

MATERIALS AND METHODS

Patients, sample collection, and DNA preparation.

The study involved 20 patients with hepatoblastoma treated between 1984 and 2005 at Hospital do Cancer A. C. Camargo, São Paulo, Brazil. A majority of the patients (95%) were submitted to a neoadjuvant chemotherapy scheme. Formalin-fixed and paraffin-embedded (FFPE) tumor samples were obtained from pathology archives. Tissue sections were stained with hematoxylin/eosin according to standard procedures and all cases were revised and the diagnosis was confirmed. Five FFPE specimens of fetal liver from miscarriages (25–36 wk) were included in the study as normal controls.

The study was approved by the Ethics Committee of the Hospital do Câncer A. C. Camargo and performed in accordance with the ethical guidelines of the 1975 Declaration of Helsinki.

Microdissection was applied to paraffin-embedded sections to enrich tumor cell content to >90%. After microdissection, genomic DNA was extract from 10 tissue sections using the DNA Mini Kit (Qiagen, Hilden, Germany), following the manufacturer's recommendations.

Bisulfite treatment.

Bisulfite treatment of DNA converts unmethylated cytosines to uracil, but the methylated ones remain as cytosines. Sodium-bisulfite conversion of 2 μg of DNA was performed by the modification of a previously described method (13). In brief, 2 μg of DNA from each sample was denatured in 0.2 M NaOH for 20 min at 50°C (in a total volume of 20 μL). The denatured DNA was diluted in 500 μL of a freshly made bisulfite solution (2.5 M sodium metabisulfite, 125 mM hydroquinone, 350 mM sodium chloride, pH 5.0) and incubated for 3 h at 70°C in the dark. Bisulfite-modified DNA was purified using the Wizard DNA Clean-Up System (Promega, San Luis Obispo, CA) according to the manufacture's instructions and eluted in 45 μL of water at 80°C. After treatment with NaOH (final concentration 0.3 M) for 10 min at room temperature, the treated DNA was precipitated by the addition of 75 μL of ammonium acetate (5M), 2.5 volumes of ethanol, and 2 μL of glycogen (5 mg/mL). Each resulting DNA pellet was washed with 70% ethanol, dried, dissolved in 110 μL of water, and stored at −80°C.

Gene selection.

A total of 25 genes were selected for QMSP-based examination of methylation abnormalities. All the genes evaluated in this study showed tumor suppressor activities and their silencing could contribute to tumorigenesis process. CCNA1, CCND2, CDKN2A, CDKN2B, P14ARF, RB1, and SOCS1 are involved in cell cycle control, CDH1 in cell adhesion, PTGS2 in the regulation of inflammatory response, ESR1, APC, DAPK, RASSF1A, RARβ, TIMP3, and THBS1 in signal transduction processes, GSTP1 in cell detoxification, MGMT and MLH1 in DNA repair, CALCA and MT1G in cell-cell signaling processes, and SFRP1 and SCGB3A1 in cell differentiation and proliferation. The function of AIM1 is not yet well understood. MINT31 is a chromosomal region frequently methylated in human cancers. It has been shown that these genes are affected by aberrant promoter methylation in association with transcription silencing in different types of human malignancies.

Quantitative methylation-specific PCR analyses.

Bisulfite-modified DNA was used as a template in fluorogenic QMSP assays carried out in a final volume of 25 μL in 96-well plates in a ABI Prism SDS 7000 (Applied Biosystems, Foster City, CA). PCR was performed in separate wells for each primer/probe set and each sample was run in triplicate. The final reaction mixture contained 3 μL of bisulfite-modified DNA, 1.2 μM of forward and reverse primers, 200 nM probe, 0.5 U of platinum Taq polymerase (Invitrogen, Frederick, MD), 200 μM dNTPs, 16.6 mM ammonium sulfate, 67 mM Trizma, 6.7 mM magnesium chloride (2.5 mM for CDKN2A), 10 mM mercaptoethanol, 0.1% DMSO, and 1 × ROX dye (Invitrogen, Frederick, MD). PCR was conducted with the following conditions: 95°C for 2 min, followed by 45 cycles at 95°C for 15 s and 60°C for 40 s.

Each plate included patient DNA samples, multiple water blanks, a negative control (normal leukocyte DNA), and serial dilutions (90–0.009 ng) of a positive control for constructing the calibration curves on each plate. Leukocyte DNA from a healthy individual was methylated in vitro with SssI methyltransferase (New England Biolabs Inc., Beverly, MA) to generate completely methylated DNA at all CpG and used as positive control.

Primers and probes were designed to specifically amplify the promoter regions of the 25 genes of interest and the internal control gene, ACTB (Table 1) (14–23). The percent of methylated reference (PMR) was calculated according to previously reported equation (methylated target-gene quantification in the sample/ACTB quantification in the sample)/(methylated target-gene quantification in the reference/ACTB quantification in the reference) (20). The reference sample consisted of SssI methyltransferase-treated DNA used for the standard curve and was amplified as a reaction control in all PCR plates.

Table 1 Primers and probes used in the QMSP assays
Full size table

Cases were scored as positive if a PMR value of ≥0.05% was obtained. This cutoff was chosen for being clinically relevant and also to exclude very low-level background readings that can occur in some samples.

Statistical analysis.

SPSS 10.0 (Statistical Package for Social Science) for Windows was used for all statistical analyses. Descriptive statistics were used to summarize study data. Statistical significance was defined as a two-tailed p value ≤0.05. Comparisons between clinical-demographic variables and methylation patterns were performed using the χ2 test or Fisher's exact test.

Survival curves were estimated using the Kaplan-Meier method. Survival data were censored for patients alive at the last observation. The log-rank test was used to compare survival outcomes. The univariate Cox regression model was used to evaluate the methylation level influence on the overall survival (OS).

RESULTS

Patient characteristics and clinical predictors.

The patient clinical characteristics are summarized in the Table 2. The age at diagnosis ranges from 0.36 to 83.68 mo (mean of 23.89 mo). Boys (65%) were more affected. Time of symptoms ranges from 0 to 12 mo (mean of 3 mo).

Table 2 Clinical characteristics of patients with hepatoblastoma at diagnosis
Full size table

Eight-five percent of patients had The International Childhood Liver Tumour Strategy Group (SIOPEL) PRETEXT III and IV disease (24). Metastasis at diagnosis was detectable in six of the 20 patients (30%) and the lung was the most frequent site. Alpha-fetoprotein level at diagnosis was obtained from 16 patients and two of them showed <100 ng/dL and died of disease. In 19 of the 20 patients (95%), the treatment protocol was based on neoadjuvant chemotherapy followed by tumor resection and adjuvant chemotherapy. The chemotherapy schema was cisplatin/carboplatin/doxorubicin in 50% of the cases, whereas cisplatin/doxorubicin was used in 45% of patients. Only one (5%) patient was submitted to cisplatin monotherapy. According to Radiologic WHO criteria, 60% of the cases exhibited partial response to the chemotherapy treatment, whereas 20% showed stable disease. Surgical resectability included partial hepatectomy (70%), segmentectomy (10%), and liver transplantation (20%). Mixed histologic subtype was the most frequent (75%). The fetal component was present in 85% of the cases.

Quantitative methylation-specific PCR in hepatoblastoma.

The first approach was to verify the methylation pattern of the 25 target genes in 10 tumor samples (pilot group). There was no methylation of AIM1, CCNA1, CDKN2A, DAPK, MGMT, MINT31, MLH1, P14ARF, RB1, and THBS1 in the samples tested and hypermethylation of CALCA (20%), CDKN2B (20%), ESR1 (20%), GSTP1 (10%), RARβ (30%), SFRP1 (20%), and TIMP3 (20%) were rare events. Conversely, APC (60%), CCND2 (50%), CDH1 (50%), MT1G (70%), PTGS2 (70%), RASSF1A (70%), SCGB3A1 (40%), and SOCS1 (60%) were frequently methylated (Table 3). Those genes that showed a high methylation frequency in hepatoblastoma (namely APC, CCND2, CDH1, SCGB3A1, MT1G, PTGS2, RASSF1A, and SOCS1) were tested in 10 additional cases and five normal controls.

Table 3 Promoter methylation frequency for the 25 genes analyzed in hepatoblastoma samples and in the normal control group (n = 5), pilot group (n = 10), and total group (n = 20)
Full size table

We found that RASSF1A was methylated in 80% of all analyzed cases (16 of 20), PTGS2 and MT1G in 55% (11 of 20), SOCS1 in 40% (8 of 20), APC, CCND2, and CDH1 in 30% (6 of 20), and SCGB3A1 in 25% (5 of 20) (Table 3 and Fig. 1). We also observed that 95% of neoplastic samples presented methylation for at least one of these eight genes.

Figure 1
figure 1

Patterns of hypermethylation observed for the eight selected genes in hepatoblastoma (cases) and fetal normal liver (control). X axis, proportion of methylated cases/tested cases for each sample type. Y axis, quantity of hypermethylation as log (PMR + 1) where PMR was calculated by the equation (methylated target gene in the sample/ACTB in the sample)/(methylated target gene in the reference/ACTB in the reference) and the reference was a in vitro methylated DNA sample. (A) APC, (B) CCND2, (C) CDH1, (D) PTGS2, (E) SCGB3A1, (F) MT1G, (G) RASSF1A, (H) SOCS1.

Full size image

Although CCND2, SCGB3A1, and PTGS2 genes are methylated in hepatoblastoma, they are also methylated in the control fetal normal liver samples and, for this reason, are not good candidates for tumor markers. Moreover, the methylation of APC, CDH1, MT1G, RASSF1A, and SOCS1 seems to be specific to tumor samples because no hypermethylation for these genes was detected in the control samples (Table 3 and Fig. 1).

Methylation levels and clinical-pathologic correlations.

The methylation pattern of APC, CDH1, MT1G, RASSF1A, and SOCS1 was analyzed for potential correlations with clinical characteristics of patients with hepatoblastoma, including age, histologic subtype, PRETEXT stage, SIOPEL risk group (25), and presence of metastasis at diagnosis. No significant correlation was observed between these clinical features and methylation status of the genes tested.

The 5-y OS was 74.67%. There was no significant difference in OS by age, histologic subtype and PRETEXT stage. But, SIOPEL risk-based stratification (55.5% high risk versus 100% standard risk; p = 0.0217) and presence of metastasis at diagnosis (92.8% localized versus 33.3% metastatic; p = 0.0066) influenced the OS. Furthermore, the log-rank test comparing OSs showed that MT1G hypermethylation was also correlated with poor prognosis (p = 0.0431; Fig. 2).

Figure 2
figure 2

Kaplan-Meier survival curve according to MT1G hypermethylation. The 5-y overall survival for the nonmethylated group (solid line) was 90%, whereas the 5-y overall survival for the hypermethylated group (broken line) was 60% (p = 0.0431).

Full size image

In addition, a Cox regression univariate model analysis showed a significant positive association between the increase of MT1G methylation and bad prognosis for patients with hepatoblastoma (p = 0.033; Table 4).

Table 4 Cox regression model comparing methylation level and overall survival
Full size table

DISCUSSION

The analysis of the clinical and epidemiologic characteristics of the cases analyzed here showed that they are similar to those included in earlier studies (3).

Many studies have described the importance of DNA methylation in the extinction of tumor suppressor gene activity in many human cancers. Considerable variations exist in promoter methylation profiles of different cancers, such that individual tumor types have characteristic methylation profiles (9). To date, few studies evaluated gene hypermethylation in hepatoblastoma (10–12,26,27). All of these studies relied on conventional MSP. Therefore, this study is the first to use QMSP to examine the methylation profile of this pediatric hepatic malignancy. QMSP is a real-time PCR-based assay that is more sensitive (it is capable of detecting methylated alleles in a background of normal at a threshold of 1 of 1,000 to 1 of 10,000) and more specific than conventional MSP since, in addition to the two PCR primers, the fluorescent-labeled hybridization probe must anneal correctly (23). We used QMSP to test a panel of 25 genes that include eight genes already reported to be methylated in this malignancy by conventional MSP assay (APC, CDH1, CDKN2A, DAPK GSTP1, MGMT, RARβ e RASSF1A) and 17 genes never examined in this tumor (AIM1, CALCA, CCNA1, CCND2, CDKN2B, ESR1, MINT31, MLH1, MT1G, P14 ARF, PTGS2, RB1, SCGB3A1, SFRP1, SOCS1, THBS1, and TIMP3).

Because of restrictions in the available amount of DNA isolated from paraffin-embedded tumor samples, we first investigated the panel of 25 genes in 10 tumor cases. As a next step, we analyzed the eight most frequently methylated genes in a second series of cases (n = 10). According to the QMSP results, the genes most frequently affected by aberrant methylation in hepatoblastoma are APC, CCND2, CDH1, SCGB3A1, MT1G, PTGS2, RASSF1A, and SOCS1. However, the hypermethylation of CCND2, PTGS2, and SCGB3A1 is not specific (these genes were also found methylated in the normal fetal liver samples).

Except for MGMT, DAPK, and SOCS1, the methylation frequencies found for the genes evaluated in our study are higher than previously reported. This fact could be explained by the higher sensibility of the QMSP method in relation to conventional MSP. However, it is important to note that precautions should be taken when comparing results of different studies concerning aberrant promoter methylation. Although the same tumor type could be evaluated, different methodologies and the use of different primers may yield varied results.

RASSF1A, the most frequently hypermethylated gene in our samples (80%), encodes a protein similar to the RAS effector protein (Nore1). This protein seems to interact with the DNA repair protein XPA and inhibits cyclin D1 accumulation, stopping the cell cycle. There are evidences that the lost or altered expression of this gene is associated with lung, colorectal, breast, and hepatocarcinoma pathogenesis (28–30). Harada et al. (10), using conventional MSP to analyze 27 hepatoblastoma samples, described the hypermethylation of RASSF1A (19%), RARβ (4%), and GSTP1 (4%) and they could not detect methylation in CDKN2A, MGMT, APC, DAPK, CDH1, and CDH13. In another study, Sugawara et al., (27) using MSP and bisulfite-modified DNA sequencing, detected RASSF1A hypermethylation in 38.5% of 39 patients with hepatoblastoma. It is important to note that there is a relevant difference between the study of Sugawara et al. and ours. In the former, all samples were obtained from patients without neoadjuvant chemotherapy, whereas ours were collected after chemotherapy exposure. Therefore, one can ask if the chemotherapy effect on the tumor could be related to the difference in RASSF1A methylation profile observed in two studies (38.5 versus 80%). In fact, Koul et al. (31) related high frequency of RASSF1A hypermethylation in cisplatin-resistant germ cell tumors. Therefore, it is possible that 1) cisplatin treatment is able to stimulate RASSF1A hypermethylation or 2) the exposure to this drug allows the selection of chemo-resistant clones (making these clones more abundant in the selected specimen). If this second hypothesis is true, a high RASSF1A hypermethylation in hepatoblastoma samples before cisplatin-based therapy should indicate cisplatin resistance and could be useful as a drug-resistance predictor. This hypothesis remains to be confirmed by additional studies.

APC protein is evolved with the CTNNB1 degradation and Wnt signaling pathway. Germ-line defects APC gene are related with familial adenomatous polyposis (FAP), Gardner and Turcot syndromes (32). As previously mentioned, patients with FAP have increased risk for development of hepatoblastoma (2). So, APC inactivation seems to be a mechanism associated with the hepatoblastoma pathogenesis and we found 30% of hypermethylation in this gene in our cases.

CDH1 encodes the adhesion glycoprotein cadherin. Mutations in this gene are associated with gastric, breast, colorectal, thyroid, and ovarian cancers. It is thought that the loss of function of cadherin may be related with cancer progression by increasing proliferation, invasion, and metastasis (33–35). Some studies have demonstrated the correlation between Wnt-pathway alterations and genesis and/or prognosis of hepatoblastoma (36,37). Considering that CDH1 is a member of the Wnt-pathway maybe its silencing by aberrant methylation could contribute to the genesis of hepatoblastoma. We detected 60% of tumors with hypermethylated CDH1.

SOCS1 encodes a JAK-ligand protein, which regulates the cytosine-dependent signal transduction in the JAK/STAT pathway. The absence of SOCS could stimulate the kinase activity of JAK, promoting cell proliferation (38). The association between hypermethylation and transcriptional silencing of SOCS1 has already been reported in hepatoblastoma (12). In this study, they analyzed 15 patients with hepatoblastoma and found 46.6% of SOCS1 hypermethylation. This frequency is similar to our findings (40%) and suggests a participation of SOCS1 in hepatoblastoma carcinogenesis.

MT1G encodes the metallothionein subtype 1G, a cysteine-rich residues protein that binds several heavy metals. Apparently, its biologic function is related to zinc and copper homeostasis and with cadmium toxicity protection. Jahroudi et al. (39) described a higher gene transcriptional activity of MT1G associated with heavy metals and dexamethasone exposure in cell line. Nagata et al. (40) compared the expression pattern of hepatoblastoma and normal liver tissue by microarray and found that MT1G was hypoexpressed in the tumor. Henrique et al. (41) demonstrated that MT1G hypermethylation was associated with advanced stages of prostate cancer. In addition, MT1G hypermethylation was not found in any normal prostate tissue in this study, which indicates that this molecular alteration should be associated with pathologic states. In our study, MT1G presented a high-specific hypermethylation frequency (55%) in the hepatoblastoma samples.

Another objective of this study was to investigate whether gene hypermethylation could be associated with clinical parameters, and particularly OS. Previous studies have failed to demonstrate association between methylation profile and clinical features of patients with hepatoblastoma (10–12,26,27). This study, for the first time, demonstrated a strong association between MT1G methylation level and poor outcome of patients with hepatoblastoma. According to our results, it seems that the higher the MT1G hypermethylation level in the tumor, the lower the patient OS.

In this study, despite the rarity of hepatoblastoma, we evaluate the hypermethylation pattern of 25 genes in 20 cases. We report for the first time that APC, CDH1, MT1G, and SOCS1 are frequent targets of aberrant methylation in this malignancy and also demonstrate that MT1G hypermethylation correlates significantly to bad prognosis of patients with hepatoblastoma.

Our results, together with other findings, demonstrate that aberrant promoter methylation is a frequent event also in hepatoblastoma and suggest that patients with hepatoblastoma may benefit from demethylating drug treatments.

In summary, this study represents the first quantitative evaluation of the methylation profile of hepatoblastoma and our data provide further evidence that the methylation pattern may be useful as a prognostic indicator in this malignancy.