Prognostic Significance of Promoter DNA Methylation in Patients with Childhood Neuroblastoma

Neuroblastoma is an embryonal malignancy that accounts for 8% to 10% of all childhood cancers and is characterized by a diversity of clinical behaviors ranging from spontaneous regression to rapid and fatal tumor progression (1, 2). In recent years, several genetic changes have been identified in neuroblastoma tumors that are relevant to clinical progression, allowing individual tumors to be classified into distinct subsets. Prognostic markers, such as age at diagnosis, clinical stage, amplification of the MYCN oncogene, DNA ploidy, and molecular defects, such as allelic loss of chromosome 1p and 11q are used for risk stratification and treatment assignment. The most prominent of these prognostic markers is MYCN, an oncogene that is amplified in approximately 20% to 25% of all neuroblastoma cases and is strongly associated with advanced-stage disease (3). However, a significant number of patients with no MYCN amplification also show poor prognosis (1). Therefore, additional prognostic markers are needed to further define patient risk groups, particularly in patients without MYCN amplification.

More recently, it has become clear that the biology of neuroblastoma is determined not only by the genetic profile but also by the epigenetic profile of the tumor. DNA methylation is a well-characterized epigenetic mechanism and is an essential biochemical process that regulates gene transcription and normal cell development. DNA methylation silences gene expression through the addition of methyl groups to cytosine residues within CpG-rich sequences, known as CpG islands, present in the promoter region of genes. The availability of methyl groups for DNA methylation is dependent on folate status, as folate is a key source of S-adenosyl methionine, a universal methyl donor (4). Folate is also essential for DNA synthesis in rapidly growing cells such as that observed in fetal development as well as cancer. Hence, the bioavailability of folate may also enhance the growth of preexisting tumor cells (5). In fact, studies have implicated folate deficiency in several pathologic diseases, including cancer (6). One of the mechanisms by which folate deficiency can promote carcinogenesis is by reduced availability of one-carbon groups required for methylation reactions, which may lead to a decrease in levels of genomic methylation or DNA hypomethylation and concomitant promoter hypermethylation of specific genes (7–11). For example, hypomethylation of DNA at specific sites within the proto-oncogenes c-MYC, FOS, and HRAS has been observed in the livers of rats fed with methyl-deficient diet (12), whereas other investigations have shown that rats with folate or methyl deficiency induce site-specific methylation within the p53 tumor-suppressor gene, and the methylation of this gene was associated with reduced expression of p53 (13). Thus, the interplay between folate and DNA methylation has an important role in normal cell development as well as tumorigenesis.

Aberrant DNA methylation at promoter CpG islands is widely accepted as a common event in a variety of human cancers including neuroblastoma (14). Indeed, a growing list of aberrantly methylated genes has been described in neuroblastoma in the past decade, suggesting a role for DNA methylation in the tumorigenesis of neuroblastoma.

In this study, the methylation status of 48 candidate genes previously shown to be the targets of aberrant methylation in embryonal tumors, such as those involved in cell-cycle regulation, apoptosis, and cell differentiation, were determined using a quantitative DNA methylation detection method. Genes involved in the folate-metabolizing pathway were also included because of their role in regulating the intracellular pools of folate. We then examined the association between levels of DNA methylation in these genes and the risk of relapse or death in patients with neuroblastoma to identify additional prognostic markers for clinical progression.

Archival DNA was available for 131 children diagnosed with neuroblastoma in Australia and New Zealand. Treatment and clinical data including age at diagnosis, sex, neuroblastoma stage, relapse/death, and MYCN status were obtained from medical records. Patients were diagnosed between 1985 and 2000 and the median follow-up time was 3 years and 2 months. All children were treated using standard protocols according to their tumor stage as previously described (15). Event-free survival (EFS) was defined as the time from diagnosis to relapse or death within 5 years from diagnosis. The study was approved by institutional ethics committees, and informed consent was obtained for patients enrolled in the study. DNA extraction was conducted using QIAamp DNA Mini Kit (Qiagen, Inc.) according to manufacturer's instructions. DNA was eluted in 50 μL of elution buffer.

The panel of 48 candidate genes examined in this study was selected on the basis of previous reports of aberrant methylation in cancer or significant associations with the risk and outcome of cancer, particularly in neuroblastoma. Essential genes involved in the folate-metabolizing pathway were also included. Candidate genes were selected using PubMeth (http://www.pubmeth.org), a publicly accessible cancer-methylation database that contains a comprehensive overview of published information relating to genes previously reported to be methylated in various cancer types (16).

Candidate genes examined in the current study are listed in Table 1. Probes selected for the assay were located within CpG islands, which were identified through the University of California Santa Cruz Genome Browser Website (http://genome.ucsc.edu/) and were defined by: (i) GC content of 50% or greater, (ii) CpG island length greater than 200 bp, and (iii) the ratio of observed to expected CpG greater than 0.6. Where no CpG island was identified for a specific gene, CpG sites within 500 bp of the transcriptional start site or promoter region were considered. CpG sites with previously identified polymorphisms listed in public accessible database, dbSNP, National Center for Biotechnology Information (http://www.ncbi.nlm.nih.gov/projects/SNP/), were excluded from our investigations.

For each sample, 1 μg of genomic DNA was modified by sodium bisulfite using the EZ DNA Methylation Kit (Zymo Research Corporation) according to manufacturer's instructions. Methylation status of all samples was analyzed simultaneously using a customized GoldenGate Veracode DNA methylation assay (Illumina), according to manufacturer's instructions. Bisulfite-treated DNA was probed at 96 individual CpG loci within the 48 candidate genes (2 CpG sites per gene). Fluorescence levels of hybridized samples were detected using an Illumina BeadXpress platform (Illumina).

The results of the GoldenGate Veracode DNA methylation assay were validated using combined bisulfite restriction analysis (CoBRA) of a selected subgroup of genes in a subset of 48 primary neuroblastoma samples randomly selected from the original cohort of 131 patients. Briefly, 50 ng of bisulfite-modified DNA was amplified by PCR using 1× Amplitaq Gold buffer (Applied Biosystems), 0.5 U AmpliTaq Gold (Applied Biosystems), 1.5 mmol/L MgCl₂, 0.25 mmol/L dNTP, and 1 μmol/L of forward and reverse primers in a total reaction volume of 20 μL. Seminested PCR was conducted subsequently using 1 μL of the initial PCR reaction with the same conditions but with 0.4 μmol/L of forward and reverse primers and 1 U of AmpliTaq Gold. CoBRA primer sequences and annealing temperatures are listed in Supplementary Table S1. Amplified products were subjected to TaqI or BstUI digestion for the recognition of TCGA or CGCG sites for 2 hours at 65°C or 60°C, respectively, and resolved by gel electrophoresis.

As only 2 CpG sites were investigated for each candidate gene using the GoldenGate Assay, bisulfite sequencing of clones was used to confirm that the surrounding CpG sites were also methylated and to examine the level of methylation heterogeneity, which has been previously reported across a range of tumors (17). Methylation status of folate hydrolase (FOLH1), myogenic differentiation-1 (MYOD1), and thrombospondin-1 (THBS1) were confirmed in neuroblastoma cell lines, such as IMR-32 and NBL-S (American Type Culture Collection), as well as a representative subset of patient samples. Primers were designed to amplify the region encompassing the CpG site(s) interrogated by the GoldenGate Assay using bisulfite PCR (see Supplementary Table S2). The PCR products were ligated into the pCR2.1-TOPO vector (Invitrogen), according to manufacturer's instructions. Up to 12 individual colonies were chosen for colony PCR using the primers listed in Supplementary Table S2. PCR products were then sequenced to ascertain the methylation status of individual alleles.

Methylation intensity data were evaluated using GenomeStudio software (Illumina). Background intensity derived from built-in negative controls was subtracted from each methylation data point to minimize intra-assay variation. Methylation levels were quantified by the beta value (β), defined as the ratio of fluorescent signal from the methylated allele to the sum of the fluorescent signals of both methylated and unmethylated allele. The β-value represented a continuous measure of DNA-methylation levels in each sample, ranging from 0 in the case of completely unmethylated sites to 1 in completely methylated sites. The average β-value was derived from 30 replicate methylation measurements for each sample.

Statistical analyses were conducted using STATA version 10 (StataCorp). To see whether methylation levels differed between clinical groups, patients were grouped into distinct clinical groups, such as those with MYCN-amplified versus nonamplified tumor, those older than 18 months versus 18 months or younger, or those with stage IV versus stages I, II, III, and IVS of tumor. Because the β-value is a continuous measure of DNA methylation, the median β-value of each group was compared using Mann–Whitney U tests. Comparisons between groups with a median difference, |▵β|, more than 0.17 and P-values of less than 0.05 were considered significant (18).

For survival analyses, samples with β-values of 0.25 or less were designated as unmethylated, whereas samples with β-values of more than 0.25 were considered methylated (19, 20). Cumulative EFS was computed by the Kaplan–Meier method and compared between subgroups using log-rank tests to determine the association between methylation of specific genes and EFS. A Cox proportional hazards model was used to examine the influence of hypermethylation of specific genes as well as established prognostic factors (MYCN amplification, neuroblastoma stage, and age at diagnosis) on EFS.

Clinical characteristics of the primary neuroblastoma samples are shown in Supplementary Table S3. Approximately 37% of patients were of ages 18 months or younger at diagnosis, with a median age of 18.2 months (range: birth to 13 years and 6 months). More than 40% of patients were diagnosed with stage IV neuroblastoma, and 17% of tumors exhibited amplification of MYCN. As with previous studies, an increased risk of relapse or death was associated with MYCN amplification [HR = 4.93, 95% confidence interval (CI) = 2.78–8.75, P < 0.001], stage IV disease (HR = 3.96, 95% CI = 2.53–6.17, P < 0.001), and being older than 18 months at diagnosis (HR = 1.85, 95% CI = 1.00–3.39, P = 0.048), whereas sex was not predictive of outcome (HR = 0.75, 95% CI = 0.44–1.29, P = 0.298; refs. 15, 21).

DNA methylation profiles for replicate samples analyzed on separate plates displayed highly correlated β-values (Spearman correlation coefficient; r ≥ 0.99). Results from GoldenGate assays were validated using CoBRA in a subset of genes (IGFBP3, MTHFR, PYCARD, RASSF1, SFN, SLC19A1, and ZMYND10). The frequencies of DNA methylation were concordant (90%) between the GoldenGate and CoBRA assays (Supplementary Table S4). As shown in Table 1, CASP8, CDH1, CDKN2A, COL1A2, DNAJC15, FOLH1, MGMT, neurotrophic tyrosine kinase receptor 1 or TrkA (NTRK1), NTRK3, PYCARD, RARB, RASSF1A, S100A6, SFN, TERT, TIMP3, TNFRSF10A, and tumor necrosis factor receptor superfamily, member 10D (TNFRSF10D) were found to be methylated in more than 50% of primary neuroblastoma samples. Ninety-eight percent of primary tumors showed hypermethylation (β-value >0.75) of SFN and NTRK1. Other genes including CCND2, CDKN2B, COMT, DAPK1, DHFR, GSTP1, HIC1, HOXA9, HS3ST2, IGFBP3, LATS1, LATS2, human Lim-homeobox 9 (LHX9), MTHFR, MTR, MTRR, MYC, MYCN, MYOD1, NTRK2, RB1, S100A10, SCGB3A1, SLC19A1, SOCS1, SST, THBS1, TNFRSF10C, WIF1, and ZMYND10 were hypermethylated in less than 50% of tumor samples. Overall, a total of 15 of the 48 genes examined were unmethylated in more than 90% of the samples examined (Table 1).

The levels of methylation observed for the 48 gene promoters were analyzed in patient samples based on tumor stage, age at diagnosis, and MYCN-amplification status. A median level of methylation was determined for each patient characteristic, with for example, the median level of methylation within MYCN-amplified patients as compared with the median level of methylation in nonamplified patients. Associations that were statistically significant at a probability level of more than 0.05 are summarized in Table 2. Patients diagnosed at age more than 18 months had significantly higher levels of methylation of DNAJC15, NTRK1, and TNFRSF10D genes, as compared with children diagnosed at age 18 months or less (Fig. 1A; P < 0.01). A similar result was also observed for the methylation levels of the DNAJC15, NTRK1, and PYCARD genes in MYCN-amplified samples in comparison with nonamplified samples (Fig. 1B; P < 0.001). Median levels of promoter methylation observed in the remaining gene promoters did not seem to differ based on the individual patient characteristics examined (P > 0.05).

Overall, patients with promoter hypermethylation of FOLH1, LHX9, MYOD1, and THBS1 displayed significantly lower EFS as compared with those without methylation (log-rank test; P < 0.004; Fig. 2). In patients lacking MYCN amplification, hypermethylation of FOLH1 and MYOD1 was significantly associated with poor outcome as compared with those without methylation (log-rank test, P ≤ 0.01; Fig. 2). As shown in Table 3, univariate analysis showed that patients with overall high levels of methylation or hypermethylation of FOLH1, LHX9, MYOD1, or THBS1 had a significantly increased risk of relapse or death. In multivariate analyses, associations remained significant for all genes except LHX9 after adjusting for MYCN amplification status, age, and stage of disease at diagnosis (Table 3).

Bisulfite sequencing of 2 neuroblastoma cell lines and a subset of primary neuroblastoma samples previously examined in the GoldenGate assay confirmed methylation of FOLH1, MYOD1, and THBS1 within the promoter region or CpG island (Supplementary Fig. S2). We also examined whether hypermethylation of 1 or more genes was a stronger predictor of outcome in our patient cohort. We focused on FOLH1, MYOD1, and THBS1 as these genes were strongly associated with poorer clinical outcome in the multivariate analysis. As shown in Table 4, the risk of relapse or death was more than 2 times higher in patients displaying hypermethylation of at least 2 of these genes after adjusting for MYCN status, stage, and age at diagnosis (HR: 2.72, 95% CI = 1.55–4.78, P = 0.001).

In this study, we used the GoldenGate Veracode methylation assay to assess levels of promoter DNA methylation of 48 genes in 131 patients with neuroblastoma and evaluated the potential clinical significance of associations between promoter gene methylation, established prognostic risk factors, and risk of relapse or death. We observed higher levels of promoter methylation of DNAJC15, NTRK1, and TNFRSF10D in older patients, and higher levels of promoter methylation of DNAJC15, NTRK1, and PYCARD in patients with MYCN-amplified tumors. Our investigations also showed that promoter hypermethylation of FOLH1, MYOD1, and THBS1 were independent predictors of outcome after adjusting for MYCN amplification, age at diagnosis, and tumor stage. Moreover, more than 30% of patients displayed promoter hypermethylation in at least 2 of these genes and were more than 2 times more likely to progress than those who did not display promoter hypermethylation after adjusting for known prognostic factors.

As with previous studies, CASP8, CDKN2A, CDH1, PYCARD, RASSF1A, SFN, and TIMP3 were found to be hypermethylated in 68% to 100% of primary neuroblastoma samples (22), whereas gene promoters that were not previously investigated for methylation levels in neuroblastoma but shown to be methylated in other pediatric tumors (23–26) such as COL1A2, DNAJC15, NTRK1, NTRK3, RARB, S100A6, and TERT were also found to be hypermethylated in 70% to 100% of neuroblastoma samples. Genes previously reported to be methylated in adult tumors such as CCND2, COMT, DAPK1, RB1, and WIF1 were not found to be hypermethylated in any of the 131 neuroblastoma tumors examined, suggesting that levels of promoter methylation in pediatric and adults tumors differ.

Levels of promoter methylation of DNAJC15, NTRK1, and TNFRSF10D were significantly higher in older patients at diagnosis (P < 0.01), whereas higher levels of promoter methylation of DNAJC15, NTRK1, and PYCARD were observed in patients with MYCN amplification (P < 0.001). Previous studies have shown that the transcriptional silencing of DNAJC15, also known as methylation-controlled J (MCJ), is epigenetically regulated by methylation (25). Hypermethylation of this gene has also been observed in pediatric brain tumors and Wilms' tumors and in ovarian cancers that displayed chemotherapeutic resistance (25, 27, 28). Although these cancers are biologically different from neuroblastoma, amplification of the proto-oncogenes, such as c-myc, MYCN, or L-myc have been observed in a small proportion of these tumors (29–31), suggesting a possible interaction between proto-oncogenes and methylation that may contribute to the tumorigenesis of these cancers.

Our finding that hypermethylation of the NTRK1 promoter was positively associated with MYCN amplification is also consistent with previous reports showing that expression of NTRK1 is negatively correlated with MYCN amplification and that NTRK1 gene expression is associated with favorable neuroblastoma tumors that regress or differentiate (32, 33). Hypermethylation of the proapoptotic gene PYCARD (PYD and CARD domain-containing protein, also known as TMS1) has also been previously reported to be associated with MYCN amplification and advanced-stage neuroblastoma (34). However, no associations were observed between the levels of methylation of NTRK1 or PYCARD and the clinical outcome in our study. While the reasons for discordant results are not clear, they are likely due to differences in patient cohorts, variation in methods to detect methylation, and disparities in the regions of NTRK1 and PYCARD analyzed.

Although the biologic significance of TNFRSF10D in carcinogenesis is unclear, previous studies have shown methylation of TNFRSF10D to be associated with reduced EFS and overall survival in patients with neuroblastoma independent of MYCN amplification (35, 36). Despite the absence of this association in our study, higher levels of TNFRSF10D methylation seen in older patients provided some evidence that TNFRSF10D methylation may have a role in influencing the clinical outcome of older neuroblastoma patients.

We identified 3 genes that displayed promoter hypermethylation and independently predicted an increased risk of relapse or death. FOLH1 encodes a protein that hydrolyses natural food folates from a polyglutamated state to a monoglutamated form before absorption can occur (37). Hypermethylation of FOLH1 has been shown to be correlated with chromosome 17q gain, a genetic abnormality often observed in neuroblastoma, as well as weakly associated with an increased risk of death (38). Although FOLH1 is not directly involved in one-carbon folate metabolism, studies have reported that polymorphisms in the FOLH1 gene can result in impaired intestinal absorption of dietary folates, leading to low blood-folate levels and hyperhomocysteinemia (37, 39). Hence, methylation-mediated inactivation of FOLH1 may provide an alternative mechanism for impaired folate absorption, and further studies examining the impact of FOLH1-promoter methylation in patients with neuroblastoma are warranted.

Higher levels of promoter methylation in MYOD1 and THBS1 also independently predicted an increased risk or relapse or death in our cohort. MYOD1 encodes for a transcription factor that shares homology to the MYC family of genes, such as c-myc, which is exclusively expressed in fetal- or adult-skeletal muscle (40), whereas THBS1 is an inhibitor of angiogenesis and has previously been shown to be hypermethylated and silenced in primary neuroblastoma tumor samples and cell lines (41, 42). De novo methylation of the MYOD1 CpG islands has been observed during the establishment of immortal cell lines, suggesting that silencing of MYOD1 via promoter hypermethylation may lead to immortalization and oncogenic transformation (43). Although MYOD1 has been reported to be transcriptionally inactive in neuroblastoma (44), to our knowledge, promoter methylation of MYOD1 has not yet been examined in this malignancy. Despite in vitro studies showing restoration of THBS1 gene expression in neuroblastoma cells following treatment with a demethylating agent (45), clinical studies have not been able to detect any association between methylation levels of THBS1 and survival in patients with neuroblastoma (42, 46).

In univariate analysis, higher levels of methylation of LHX9 were found to be associated with an increased risk of relapse or death. However, this apparent association disappeared after adjusting for MYCN amplification. Nevertheless, the LHX9-gene promoter may be a potential target for demethylating agents in patients with MYCN-amplified tumors, particularly as it encodes for a transcription factor involved in the control of neuronal differentiation as well as brain development (47). Moreover, methylation-mediated silencing of LHX9 is frequently observed in pediatric malignant astrocytomas, the most common form of glioma (48). To our knowledge, analyses of promoter methylation in LHX9 have not been reported in patients with neuroblastoma. Hence, further studies are required to fully elucidate the function of LHX9-promoter methylation in MYCN-amplified neuroblastoma tumors.

Of the 3 genes identified, hypermethylation of at least 2 genes was associated with an increased risk of relapse or death in patients with neuroblastoma. These results suggest the coordinated methylation of several gene loci or a CpG island methylator phenotype (CIMP). Previous investigations have reported that methylation of the protocadherin-β (PCDHB) gene family, either alone or in combination with methylation of the hepatocyte growth factor-like protein (HLP) and cytochrome p450 (CYP26C1) genes, is a potential CIMP associated with poorer survival in patients with neuroblastoma (49, 50). Our investigations implicate additional genes that may provide an improved CIMP for predicting the outcome of neuroblastoma.

While previous reports have shown the presence of methylation-mediated silencing in neuroblastoma, the frequency of methylation has been shown to vary between different studies, possibly due to the different techniques used between studies as reviewed in ref. 51. The methylation detection method used in our study was both sensitive and quantitative, whereas several other techniques, such as methylation-specific PCR or CoBRA that are commonly used in research studies are nonquantitative or semiquantitative, and other quantitative methods, such as pyrosequencing or bisulfite sequencing can be costly and labor intensive. Hence, uniform methods or scoring systems need be established to improve comparison of results between laboratories. The Illumina GoldenGate assay provides a standardized method where specific primers and probes have been predesigned to interrogate CpG sites that are individually assigned with a unique identifying code and allow direct comparisons between laboratories. Therefore, this method has potential to be used in a clinical setting for prognostic evaluation of patients.

Gene associations found in the present study may contribute to improved prediction of clinical outcomes, especially in patients without MYCN amplification. Our study provides strong evidence to support the hypothesis that epigenetic changes in multiple genes have the capacity to alter the clinical phenotype of neuroblastoma and that the increasing number of methylated genes increases the risk of relapse or death. While further studies are required to delineate the full phenotypic consequences of DNA methylation in these and other gene promoters, our findings highlight the potential use of methylation profiling to provide additional prognostic information and detect new therapeutic targets for selected patient subsets. The establishment of a rapid standardized molecular approach to assess gene-promoter–methylation status of neuroblastoma tumors will be essential for the translation of these and other prognostic findings into the clinical setting.

No potential conflicts of interest were disclosed.

Conception and design: D.T. Lau, L.B. Hesson, L.J. Ashton

Development of methodology: L.B. Hesson, M. Haber, L.J. Ashton

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): D.T. Lau, L.J. Ashton

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): D.T. Lau, L.B. Hesson, G.M. Marshall, M. Haber, L.J. Ashton

Writing, review, and/or revision of the manuscript: D.T. Lau, L.B. Hesson, M.D. Norris, G.M. Marshall, M. Haber, L.J. Ashton

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): D.T. Lau

Study supervision: L.B. Hesson, L.J. Ashton

D.T. Lau was supported by the National Health and Medical Research Council Public Health Postgraduate Research Scholarship.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.