In recent studies, we examined DNA methylation in a new cohort of 48 RMS tumors (21 FP and 27 FN) on the Illumina HumanMethylation450 BeadChip platform. Unsupervised clustering analysis using the most variable probes (top 1%) in the 48 RMS tumors revealed that DNA methylation patterns in these tumors segregated into two distinct clusters; one cluster contains all 21 FP cases along with 2 FN cases and a second cluster contains 25 of the 27 FN cases. This study thus confirms our previous findings in a separate cohort of 37 RMS cases analyzed with the Infinium HumanMethylation27 BeadChip platform. Beyond the initial separation of FP and FN cases, our current DNA methylation studies with the 48 RMS case cohort identified two major subgroups within the FP cluster and two major subgroups within the FN cluster. To further examine the significance of these novel methylation-defined subgroups, we compared the DNA methylation profiles in this new cohort of 48 RMS tumors with mutation, gene fusion, and copy number data for these cases. In the FP cluster, the two subgroups are associated with fusion subtypes such that one subgroup is enriched for PAX3-FOXO1-positive cases and the second subgroup is enriched for PAX7-FOXO1-positive cases. In the FN cluster, the two subgroups are associated with RAS mutation status such that one subgroup is enriched for FN cases with RAS gene mutations and the other subgroup is enriched for FN cases without RAS gene mutations. We then validated these associations of methylation-defined subgroups with fusion status in FP RMS and with RAS mutation status in FN RMS in a second independent cohort of 21 FP (12 PAX3-FOXO1-positive and 9 PAX7-FOXO1-positive) and 17 FN RMS tumors. Further comparison of the DNA methylation differences in these subgroups with RNA expression data identified several relevant genes that were differentially expressed between the subgroups in a methylation-dependent manner. To further analyze the DNA methylation differences between FP and FN RMS tumors, the CpG sites that are differentially methylated between FP and FN tumors were localized with respect to genomic transcriptional organization and were found to be differentially distributed with respect to specific genomic features. In general, the differentially methylated sites were enriched in the intergenic regions and depleted in the promoter and 3' UTR regions. Furthermore, hypomethylated probes were enriched in FP tumors in the intergenic region, whereas hypermethylated probes were enriched in FP tumors in the gene body, 3' UTR and intergenic regions. In our new larger cohort of cases, there was also a significant difference in the distribution of PAX3-FOXO1 binding sites between genes with and without differential methylation. Integrative analysis of PAX3-FOXO1 binding sites, promoter methylation and gene expression demonstrated that genes with PAX3-FOXO1 binding sites tended to be more highly expressed in FP tumors (compared to FN tumors) than genes without P3F binding sites regardless of promoter methylation status. Promoter hypomethylation is most highly associated with enhanced expression among genes with PAX3-FOXO1 binding sites; however, the group of highly expressed genes with PAX3-FOXO1 binding sites and promoter hypomethylation is small in number compared to the much larger group of highly expressed genes with PAX3-FOXO1 binding sites but without promoter hypomethylation. Though unsupervised clustering analysis indicated that FP tumors and cell lines cluster as do the FN tumors and cell lines, a principal component analysis clarified these relationships by showing that the two groups of cell lines are located at a considerable distance from the two tumor subtypes. Analysis of the most varied probes in the tumors indicated that the vast majority of these probes are hypermethylated in all RMS cell lines. An assessment of DNA methylation in xenografts derived from these RMS cells lines reveals methylation patterns similar to the cell lines. In contrast, the DNA methylation patterns in patient-derived xenografts (that have not been subjected to cell culture) are generally more similar to patterns in the primary human tumors. These analyses indicate that RMS cell lines and associated xenografts do not faithfully recapitulate the DNA methylation patterns that characterize primary tumors, and further indicate that patient-derived xenografts may have greater utility in the study of epigenetic alterations present in RMS primary human tumors.
