The immediate goal of this project was to investigate differences in DNA methylation between fusion-positive and fusion-negative rhabdomyosarcoma (RMS) tumors. In preparation for this analysis, the Infinium HumanMethylation27 BeadChip database was edited to remove probes that localized to the X and Y chromosomes or repetitive regions, probes containing polymorphic sites, probes with ambiguous genomic locations or probes without detectable signal. Application of unsupervised analysis algorithms to this edited database demonstrated two distinct clusters, one containing the fusion-positive RMS cases and one containing the fusion-negative RMS cases. This differential methylation was further characterized by an increased incidence of hypomethylated sites in the fusion-positive compared to the fusion-negative tumors. Among probes that are differentially methylated between the two RMS subtypes, there were 3-fold more hypomethylated than hypermethylated sites in the fusion-positive subset. This differential methylation was particularly prominent in probes mapping to promoter regions and/or CpG islands. Since the initial analysis determined that methylation status was sufficient to classify the fusion status of an RMS tumor, we next investigated the number of probes that were needed for this classification. A training algorithm to find the minimum number of discriminating CpG sites determined that the methylation status of 11 probes is sufficient to classify the fusion-positive from the fusion-negative RMS tumors. In this group of 11 probes, seven were hypomethylated and four were hypermethylated in the fusion-positive compared to the fusion-negative cases. Pyrosequencing studies of eight of the associated genes independently confirmed that the CpG site queried on the array as well as other nearby CpG sites were differential methylated. To validate these findings, we assessed the methylation status in a collection of ten RMS cell lines (five fusion-positive and five fusion-negative). Unsupervised analysis revealed that these cell lines clustered according to fusion status and that the eleven probe signature was sufficient to correctly classify the fusion status in these 10 lines. In another set of analyses, we investigated the relationship between methylation and gene expression in these RMS tumors. In RMS cases with both methylation (HumanMethylation27) and expression (U133 Plus 2.0) data, there was an overall inverse relationship of mRNA expression and promoter methylation. However, among the 145 genes with both differential promoter methylation and differential expression, 57% showed an inverse correlation between methylation and expression whereas 43% showed a direct correlation. To determine whether the fusion transcription factors (PAX3-FOX01 and PAX7-FOXO1) impact on these issues, we analyzed the distribution of PAX3-FOXO1 binding sites among these differentially methylated genes. Though there was no significant difference in the distribution of PAX3-FOXO1 binding sites between genes with and without differential methylation, PAX3-FOXO1 binding sites are enriched among differentially methylated genes that are also differentially expressed.
