To extend our studies of DNA methylation in RMS tumors, we used the HM450 array to assess genome-wide DNA methylation patterns in a discovery cohort of 21 FP and 17 FN tumors. Unsupervised analyses confirmed our previous finding of a close association of DNA methylation pattern and fusion status. In addition to the DNA methylation differences between the FP and FN categories, our DNA methylation analysis also revealed two major subsets within the FP cluster, and two major subsets within the FN cluster. Our evaluation revealed that the PAX3-FOXO1 and PAX7-FOXO1 fusions were each enriched in one of the two subsets in the FP category. Similarly, RAS mutant and RAS wild-type cases were each enriched in one of the two subsets in the FN category. We validated these findings in an additional cohort of 21 FP and 27 FN tumors. As was found in the discovery cohort, there were two major clusters of FP cases and two major clusters of FN cases. Comparison of the PAX3-FOXO1 and PAX7-FOXO1 fusion status in the FP cases confirmed a significant association between the gene fusion subtype and DNA methylation pattern in this validation cohort. In addition, comparison of the RAS mutation status in the FN cases also confirmed a statistically significant association between RAS mutation status and DNA methylation pattern. To further explore DNA methylation differences between the PAX3-FOXO1 and PAX7-FOXO1 subtypes, we combined the discovery and validation cohorts. Comparison of these two fusion subsets by a supervised bioinformatic approach revealed greater than 17-fold more hypermethylated than hypomethylated probes in PAX3-FOXO1- compared with PAX7-FOXO1-positive tumors. One notable gene with significantly more promoter hypermethylation in PAX3-FOXO1- than PAX7-FOXO1-positive tumors is CDKN1C, a putative tumor suppressor gene that is localized within a region of 11p15.5 allelic loss in RMS and other tumors. RNA expression studies demonstrated significantly less expression of CDKN1C mRNA in PAX3-FOXO1- compared to PAX7-FOXO1-positive tumors, and subsequent immunohistochemistry studies on confirmed a significant difference in CDKN1C protein expression between PAX3-FOXO1- and PAX7-FOXO1-positive tumors. To further investigate DNA methylation differences within the FN category, we combined FN RMS tumors from the discovery and validation cohorts. A supervised analysis of DNA methylation in these FN cases revealed 1.5-fold more hypermethylated than hypomethylated CpG sites in RAS mutant compared to RAS wild-type FN RMS tumors. Using this collection of differentially methylated CpG sites, a subsequent clustering analysis of the combined FN cohorts and normal skeletal muscle samples again identified two main subgroups. The normal muscle samples and most RAS wild-type cases clustered in one subgroup and all RAS mutant cases along with additional RAS wild-type cases clustered in the second subgroup. Examination of the DNA methylation pattern within the latter RAS mutant-enriched subgroup revealed two smaller clusters, one cluster consisting of most RAS mutant tumors and no RAS wild-type tumors, and a second cluster consisting of a mixture of RAS mutant and RAS wild-type tumors. Further analysis demonstrated that mutations of two additional RAS pathway genes (NF1 and SOS1) were present at a high frequency in the RAS wild-type tumors in this latter subgroup, thus supporting the hypothesis the DNA methylation differences between the two main FN subsets are associated with a larger set of mutations involving the RAS signaling pathway. Based on this premise, a supervised approach comparing FN tumors with mutant RAS pathway genes versus wild-type RAS pathway genes identified 16-fold more hypomethylated than hypermethylated CpG sites in FN RMS tumors with a mutant RAS pathway compared to a wild-type RAS pathway. One notable gene with significantly more promoter hypomethylation in mutant RAS pathway tumors is ALDH1A3, which was previously reported as a potential marker for cancer stem cells in embryonal RMS. RNA expression studies demonstrated significantly more ALDH1A3 mRNA expression in RAS pathway mutant versus wild-type tumors. To determine which if any RMS model systems recapitulate the DNA methylation patterns found in human primary RMS tumors, we compared genome-wide methylation patterns in primary RMS tumors to long-term RMS cell lines, xenografts derived from these long-term cell lines (CDXs) and patient-derived xenografts (PDXs). Though hierarchical clustering again demonstrates a FP cluster and a FN cluster, each of these main clusters is subdivided into two smaller clusters, one containing nearly all cell lines and CDXs and one containing nearly all PDXs and primary tumors. Furthermore, the heat maps reveal that the vast majority of analyzed CpG sites are hypermethylated in cell lines and CDXs in contrast to the wider distribution of hypo- and hypermethylation found in primary tumors and PDXs. The finding of two PDXs clustering with cell lines and CDXs suggests that PDXs may lose the native methylation pattern under some unknown circumstances. Further studies showing no statistical differences in overall methylation levels between PDXs and primary tumors in contrast to the significantly higher overall methylation levels in cell lines and CDXs confirms that PDXs are the optimal experimental model to study the biological significance of DNA methylation patterns in RMS tumors. To extend this conclusion, we assessed the stability of the DNA methylation patterns when short-term cultures (1-2 months) were established from the PDX tumors. An unsupervised analysis comparing RMS PDX tumors and short-term cultures with a set of long-term RMS cell lines demonstrated that the overall DNA methylation pattern was maintained during short-term culture for all FN short term cultures and most FP short term cultures. These findings thus provide the possibility of using a short-term culture to perform experimental manipulations of PDXs before returning these cells to the in vivo environment.
