Clustering of gene hypermethylation associated with clinical risk groups in neuroblastoma

Background: Neuroblastoma is the most common extracranial solid malignancy in infancy and childhood, but the biological factors involved in its development and progression are still unclear. Transcriptional silencing of tumor suppressor genes mediated by hypermethylation of promoter CpG islands is a hallmark of human tumors. We addressed the clinical relevance of promoter hypermethylation in neuroblastoma. Methods: We examined the methylation status of 45 candidate genes representative of many cellular pathways in 10 neuroblastoma cell lines and of 10 of these genes in 145 tumor samples (118 of them were primary neuroblastomas). We used Fisher's exact test to examine the association of CpG island methylation and clinical subgroups and Kaplan-Meier analysis to determine the association between methylation and survival in primary tumors. Cluster analysis was used to group cell lines and tumors by gene methylation status. Bonferroni-corrected statistical tests were two-sided. Results: Clustering of neuroblastoma cell lines on the basis of hypermethylation distinguished lines with MYCN amplification (a negative prognostic factor) from those without it (P = .012). Promoter hypermethylation of the developmental gene HOXA9 was associated with mortality in noninfant patients (P = .04) and in tumors lacking MYCN amplification (P = .023). Hypermethylation of the proapoptotic gene TMS1 and the cell cycle gene CCND2 was associated with stage 4-progressing tumors (P < .001), but the genes were never methylated in stage 4S tumors, which undergo spontaneous regression. Hypermethylation of the differentiation gene RARbeta2 was associated with patient survival (P = .032). Unsupervised hierarchical cluster analysis of all tumors based on methylation of the 10 genes separated several clinically relevant groups of tumors. Conclusions: Profiling the status of CpG island hypermethylation in human primary neuroblastomas may have clinicopathologic value.