In the current study we used quantitative bisulfite Pyrosequencing to assess promoter methylation levels of tumor suppressor genes known to be frequently hypermethylated in cancer. In 30/38 tumors we found significant hypermethylation in one or more of the following genes; BLU, CASP8, DCR2, CDH1, RASSF1A and RASSF2. Overall Z-scores for the TSGs assessed were significantly associated with adverse outcome. Furthermore, six of the 38 tumors conformed to the criteria for CIMP, i.e. CpG island methylator phenotype. The non-tumoral adrenal medullary material in this study represents the closest possible healthy analogue to the cells that comprise neuroblastomas. Neuroblastomas arise solely in neural crest-derived cells of sympathoadrenal lineage (
 and others), the very cells that would form the adrenal medulla (and abdominal sympathetic ganglia) in healthy individuals
. In this current study they were only used to confirm that TSG methylation is low in healthy tissue, and to determine suitable cut-off levels for different TSGs. The level of cut-off for hypermethylation in individual genes was set conservatively to exceed the methylation density observed in the reference adrenal medulla (Table
1). For genes where no methylation was observed the cut-off was set to 10% to avoid false positives resulting from background fluctuation. For the genes where methylation was detected in reference adrenal medulla (CASP8A2, DCR2, RASSF1A) the cut-off in tumors was set well above the level observed in the reference samples (Table
1). Other recent publications on DNA hypermethylation in neuroblastoma either do not utilize reference controls
[9, 10], or compare to “normal adrenal” and blood lymphocyte DNA
. We used the reference tissue only to assess presence or absence of hypermethylation; for the comparison between methylation levels and clinical/genetic phenotypes, we took into account Z-scores for all 38 tumors in the series, independent of whether they were classified as hypermethylated or not. Hence, methylation levels in reference adrenal medullary material were not included in any statistical calculation.
Although several recent studies have assessed methylation in genes with putative tumor suppressor properties
[9, 10], this current study is the first to employ a quantitative method, pyrosequencing, to assess promoter methylation in multiple TSGs. Our data partially corroborate the findings of Grau et al. and Hoebeeck et al., as we find abundant methylation in CASP8A and RASSF1A; however, we did not observe significant hypermethylation in the PTEN promoter as reported by Hoebeeck et al. We here acknowledge the sensitivity of methylation specific PCR (MSP) in detecting low levels of methylation. However, this highly sensitive, nonquantitative assay is known to produce false positive results
[35, 36]. Further, the biological significance of methylation detected by MSP may be limited, as the technique is capable of producing positive results down to a methylation level <1% (Rand et al., own observations). This corresponds to a very limited subset of cancerous cells in a tumor, and may in fact represent DNA methylation in contaminating cell types. This underscores the importance of employing quantitative methods when assessing DNA methylation – whenever possible combined with relevant reference samples for the sake of arbitration.
CASP8, defined as a tumor suppressor gene by Teitz et al., promotes apoptosis upon activation of the Fas apoptotic pathway through the Fas ligand
. There has been some debate concerning the localization of the CASP8 regulatory region. In 2000 Teitz et al. reported agreement between methylation of a CpG-rich region (defined by GenBank accession number AF210257, positions 536–856; Figure
4) and absent CASP8 expression in neuroblastomas. Also, the region was reported to be methylated almost exclusively in neuroblastomas with MYCN amplification. Teitz et al. indicated that this could signify that inactivation of the Fas apoptotic pathway is needed for the survival of neuroblastoma cells expressing high levels of MYCN. A contrasting view was presented by Banelli et al. who did not find correlation between CASP8 silencing and MYCN amplification, although higher frequencies of methylation were detected in MYCN-amplified cells
. They also argued that the CpG-rich intragenic region assessed by Teitz et al. is not a true regulatory region for CASP8. Instead, a region flanking exon 1 was proposed as the CASP8 promoter (Figure
4). In the current paper we have assessed the regulatory regions suggested in both publications, designated as CASP8 A1
 and CASP8 A2
4). Furthermore, we have compared the methylation levels in these sites in neuroblastomas to those of healthy reference tissues, which was not undertaken in the previous studies. The differences observed between methylation of the regions CASP8 A1 and CASP8 A2 in reference adrenal medullary DNA are striking: the region CASP8 A1, while abundantly methylated in neuroblastomas (mean methylation 16.2%; 21 tumors over cut-off), was devoid of methylation in reference adrenal medulla. In contrast, CASP8 A2 had a high degree of methylation in both tumors and references(9 tumors over cut-off; mean methylation in tumors 42% vs. 36% in reference samples). Thus, hypermethylation at CASP8 A1 was found as a better indicator of a pathologic condition than high methylation at the CASP8 A2 region. Furthermore, CASP8 A1 and CASP8 A2 both showed striking variations in methylation densities between individual CpGs (Figure
3). For CASP8 A1 a gradient with increasing methylation from CpG 1 to CpG 4 was noted, while for CASP8 A2 very high methylation was frequently recorded at CpG 2, 3 and 5 (Figure
3). These observations underline the importance of analyzing more than single CpGs as an indicator of methylation density.
A long-standing dilemma in neuroblastoma research is the proposed association between CASP8 methylation and MYCN amplification
[9, 10, 28, 38]. Like Grau et al. we do not find a correlation between CASP8 A1 methylation and MYCN amplification
. In contrast, Hoebeeck et al., who also assessed the CASP8 A1 region, performed a meta-analysis including a total of 115 neuroblastomas that linked methylation of the CASP8 A1 region with MYCN-amplification
. However, the included studies utilized non-quantitative MSP
[9, 28, 38]. Our findings provide an epigenetic explanation to a previous study wherein loss of CASP8 protein expression was observed in a majority of neuroblastomas
. We further support, backed by quantitative epigenetic data, the observation from that study that no correlation exists between loss of CASP8 and adverse neuroblastoma features and outcomes (such as MYCN amplification and reduced survival)
The CpG Island Methylator Phenotype, CIMP, is characterized by concerted abnormal hypermethylation in CpG rich gene promoters. Such epigenetic remodeling could lead to the simultaneous inactivation of cellular functions that regulate growth, differentiation and apoptosis and thus contribute to neoplasia development and disease phenotype. Indeed, CIMP has been described in a number of cancers, including neuroblastomas, often in conjunction with unfavorable disease progression
[15, 16]. Abe et al. defined CIMP as simultaneous methylation in CpG islands of the PCDHB and PCDHA gene families, and the HLP, DKFZp451I127 and CYP26C1 genes, and found this genotype associated to MYCN amplification
[17, 18]. In the current study CIMP was observed in 6/38 tumors and in 7/7 cell lines. However, no significant correlations between CIMP and clinical/genetic features were observed. These results support that concerted promoter hypermethylation is an important facet of the neuroblastoma causality, however if hypermethylation of key TSGs are involved in fatal disease progression they would partly differ from those assessed in this study. By contrast, significantly higher mean TSG Z-scores were observed in tumors with poor outcome at follow-up, which further indicates that hypermethylation is a component of morbidity in neuroblastomas. Analysis of a subset of TSG promoters is likely to render a somewhat fragmental insight into the role of promoter hypermethylation in neuroblastoma development. More global approaches such as methylation arrays are likely to yield a more detailed account of the key genes which, by undergoing hypermethylation, significantly impact tumor progression.
LINE-1 is frequently hypomethylated in cancer
[24, 26], leading to its activation – which in turn causes genomic instability
. In the current study LINE-1 Z-scores were higher in stage 4 tumors as compared to tumors of stage 1, 2, 3, and 4 S, indicating global hypermethylation. The finding indicates that genomic instability in metastatic neuroblastomas is not caused by LINE-1 activation. Taking into consideration that DNA methylation is a disseminative event, this finding contributes to an emerging picture of aberrant epigenetic patterning in deleterious neuroblastomas.
Several TSGs showed prominent promoter hypermethylation in this study. The two most notably methylated TSG promoters were CASP8 A1 and RASSF1A. Furthermore BLU and DCR2 also exhibited high levels of methylation, and have been reported previously in association with CIMP
. Interestingly, serum levels of RASSF1A and DCR2 methylation were reported to have prognostic importance
[41, 42]. The epigenetic inactivation of genes with such diverse tumor suppressive functions as growth arrest (RASSF1A) and apoptosis (CASP8 A1 and DCR2) may be requisite for neuroblastoma tumorigenesis. In this regard it is interesting to observe that promoter hypermethylation in the same genes occurred in neuroblastoma cell lines, a phenomenon also observed by Hoebeeck et al.. In the current study the relative methylation patterns were similar between cell lines and primary tumors (Table
2): for TSGs in which high levels of methylation were observed in cell lines, methylation was abundant even amongst the tumor samples. An appealing prospect from this is that neuroblastoma cell lines may be scanned for putative methylation in TSGs, perhaps using array techniques, in order to identify the most relevant TSGs in neuroblastomas. In an approach by Carén et al. that utilized chemical de-methylation of cell lines in combination with expression arrays a number of genes were identified as differentially methylated in neuroblastomas. Interestingly, no bona fide TSGs were identified. Several studies now confirm the presence of gene hypermethylation in neuroblastoma. While this is an important observation the cause-effect relationship between tumorous neuroblastoma and gene hypermethylation should be further investigated to facilitate improved treatment approaches.