Epigenetic studies in rhabdomyosarcoma

Division Of Basic Sciences - Nci

Investigators

Linked publications & trials

Abstract

During the last year, we continued our studies of the DNA methylation patterns in mouse models of rhabdomyosarcoma (RMS). In particular, my collaborator Dr. Charles Keller (Children's Cancer Therapy Development Institute) used Cre-LoxP methodology to generate a series of mouse strains in which one or more driver mutations (Pax3-Foxo1 [P3F] fusion and deletions of Tp53, Rb1 or Ptch) is introduced into one of four myogenic lineages (as defined by activity of the Pax3, Pax7, Myf5 or Myf6 expression regulatory elements). These mice containing a germline Tp53 deletion with or without other germline driver mutations frequently develop soft tissue tumors resembling RMS or related sarcomas. The DNA methylation patterns of tumors from these mouse studies were then profiled to explore the biological basis for the DNA methylation differences previously described in comparisons of human fusion-negative (FN) and fusion-positive (FP) RMS tumors. In particular, we hypothesized that these DNA methylation differences may be due to differences in the genetic changes (driver mutation) that occur in these tumors or due to differences in the underlying cell of origin (lineage) of these tumors. For these studies, we selected 31 mouse tumors corresponding to a variety of lineages and driver mutations, and then utilized the Infinium Mouse Methylation BeadChip (Illumina) to analyze the genome-wide methylation patterns in DNA isolated from these tumors. We performed unsupervised analysis of the resulting DNA methylation patterns by hierarchical clustering and t-Distributed Stochastic Neighbor Embedding (t-SNE). Both methodologies identified two distinct clusters. In one cluster, the driver was always P3F and the targeted lineages were those defined by the Pax3, Myf5 or Myf6 expression elements. In the other cluster, the driver/lineage was either P3F in a Pax7-associated lineage or other driver mutations (Tp53, Rb1 or Ptch deletions without P3F) in any of the four lineages. Previous published studies demonstrated that P3F targeted into the Pax7-associated lineage has very low expression whereas P3F targeted to the other three lineages has much higher expression, and thus the two dominant clusters can be defined by high P3F vs low or no P3F expression. This finding supports the premise that high expression of the P3F driver mutation is a major determinant of the DNA methylation pattern. However, the finding that the Pax7-associated lineage is always associated with the opposite DNA methylation pattern, regardless of whether the driver mutation is P3F, indicates that lineage also is a major determinant of DNA methylation pattern. In addition, further analysis of the cluster with low/no P3F expression demonstrates two distinct subclusters. Although both subclusters contain tumors in which a Tp53 or PTC1 mutation is targeted to the Pax3 or Myf6 lineage, one of these subclusters contains all tumors in which a Tp53 or PTC1 mutation is targeted to the Pax7 lineage and the other subcluster contains all tumors in which a Tp53 or PTC1 mutation is targed to the Myf5 lineage. These DNA methylation differences further support the premise that the lineage also makes a major contribution to the DNA methylation pattern. As the presence of a highly expressed P3F gene is associated with a distinct DNA methylation in both human and mouse tumors, we investigated whether there are genes with common DNA methylation changes in human and mouse tumors. For promoter methylation changes, there were 41 genes with promoter hypermethylation in FP compared to FN tumors (6.9% of hypermethylated genes in mouse tumors and 7.4% in human tumors) and 108 genes with promoter hypomethylation in FP compared to FN tumors (16.3% of hypomethylated genes in mouse tumors and 6.9% in human tumors). Similarly, for body methylation changes, there were 143 genes with body hypermethylation in FP compared to FN tumors (15.4% of hypermethylated genes in mouse tumors and 17.3% in human tumors) and 283 genes with body hypomethylation in FP compared to FN tumors (25.0% in mouse tumors and 16.6% in human tumors). There was a significant association (P0.01, Fisher test) for all four comparisons of mouse and human tumors, and thus it appears that that there are numerous commonalities between the DNA methylation patterns in mouse and human RMS tumors. Despite these commonalities, comparison of mouse and human tumors by unsupervised analyses of the DNA methylation data demonstrated distinct species-specific clustering of these tumors, and thus the differences between the two species appears to be more dominant than any similarities associated with fusion status. Finally, we used both RNA expression and DNA methylation databases to identify genes demonstrating changes in DNA methylation and RNA expression in fusion-positive compared to fusion-negative tumors in both human and mouse. Our findings reveal 44 genes that fit these criteria; these genes break down into the following categories: 5 promoter hypermethylated/underexpressed, 2 promoter hypermethylated/overexpressed, 3 promoter hypomethylated/underexpressed, 1 promoter hypomethylated/overexpressed, 5 body hypermethylated/underexpressed, 4 body hypermethylated / overexpressed, 14 body hypomethylated/underexpressed, 10 body hypomethylated/overexpressed. Therefore only a subset of the DNA methylation changes are associated with changes into gene expression; these findings support the premise that many of these DNA methylation changes are associated with lineage-specific chromatin structure.