Genome-Wide Profiles of Extra-cranial Malignant Rhabdoid Tumors Reveal Heterogeneity and Dysregulated Developmental Pathways

Abstract

Malignant rhabdoid tumors (MRTs) are rare lethal tumors of childhood that most commonly occur in the kidney and brain. MRTs are driven by SMARCB1 loss, but the molecular consequences of SMARCB1 loss in extra-cranial tumors have not been comprehensively described and genomic resources for analyses of extra-cranial MRT are limited. To provide such data, we used whole-genome sequencing, whole-genome bisulfite sequencing, whole transcriptome (RNA-seq) and microRNA sequencing (miRNA-seq), and histone modification profiling to characterize extra-cranial MRTs. Our analyses revealed gene expression and methylation subgroups and focused on dysregulated pathways, including those involved in neural crest development.

Figure 1. Mutations identified in 40 MRT cases

The figure shows selected alterations detected using genome analysis (top), the number and types of somatic SNVs (middle; median of 612.5 mutations is indicated by a dashed line), and clinical characteristics of 40 MRT cases (bottom). Each column represents a sample. See also Figure S1 and Tables S1 and S2.

Figure 2. Clustering miRNA expression profiles of MRT with other tumor and normal tissue types

MRT samples are represented individually, while non-MRT tumor and normal tissue types are represented by the medians of miRNA expression of all samples of that type. Orange: MRT cases clustered with normal cerebellum samples (“MD_NORM” and “GBM_NORM”) and pheochromocytomas and paragangliomas (“PCPG_TUM”). Green: MRT cases clustered with synovial sarcomas (“SARC”). See also Figure S2 and Table S3.

Figure 3. NMF clustering of miRNA expression profiles

NMF consensus heat map (top), clinical characteristics of MRT cases in each group (middle), and miRNAs that were differentially expressed between subgroup 1 and 2 (n=29; log₂ FC >1, FDR <0.05) (bottom). Pan-Cancer miRNA cluster: See Figure 2. The labels, “Cerebellum” and “Synovial sarcoma”, indicate MRT cases that clustered with normal cerebellum and pheochromocytomas and paragangliomas, and with synovial sarcomas, respectively. The silhouette width represents the robustness of the clustering solution of two groups by NMF. Each column represents a sample. See also Figure S3 and Table S3.

Figure 4. Analysis of mRNA expression profiles

(A, B) Bar graphs show adjusted p values and numbers of genes (in brackets) mapped to top enriched GO terms for over-expressed genes (A) and under-expressed genes (B) in MRT compared to fetal cerebellum and hESC (gray dashed line at p value = 0.05). (C) NMF consensus heat map and clinical characteristics of 40 MRT samples (top), a heat map showing over-expressed genes in sub-group 1 and 2, which were enriched for pathways (* indicates pathways with BH-corrected enrichment p value < 0.05; pathways are assigned to genes as indicated by the black and grey bars on the right of the heat map) (middle), and mutations, differential expression and differential CGI promoter methylation in genes that regulate various stages of neural crest development (Simões-Costa and Bronner, 2015), including neural plate border formation, neural crest specification and migration, and differentiation of neural crest cells into various cell types including cardiac neural crest (“Card”), myoblast (“Myo”) and sympathetic neurons (“SN”) (bottom). Each column represents a sample. (D, E) Hierarchical clustering (D) and principal component analysis (E) of 40 MRT samples. Blue and red colors indicate samples in NMF-derived sub-group 1 and 2, respectively. (F) 880 genes differentially expressed between sub-group 1 and sub-group 2 (FDR < 0.05). Blue shades indicate over-expressed genes in sub-group 1, while red shades indicate over-expressed genes in sub-group 2. Lighter shades indicate genes detected at FDR threshold at 0.05, while the darker shades indicate genes detected at FDR threshold at 0.01. Labels indicate the 20 most differentially expressed genes. See also Figure S4 and Table S4.

Figure 5. Analysis of DNA methylation profiles

(A) Correlation between unsupervised clustering of promoter CpG island (CGI) methylation levels and patient age at diagnosis (1000 tests based on all promoter CGIs; approximately unbiased p value = 0.03). The heat map displays promoter CGIs with at least 15% difference in methylation (row means) between the two sub-groups. (B) Box plots show the distribution of CpG methylation levels outside CGIs (left) and the average methylation levels within CGIs (right) for MRT cases in subgroups A (n=10) and B (n=30), MRT cell lines (n=3), NPC (n=4) and hESC (n=3). The box indicates the interquartile range (IQR), with the line within the box indicating the median level. The lines below and above the box indicate values within 1.5 IQR from the first and the third quartile, respectively. (C) Pathway analysis of genes with hypermethylated promoter CGIs in sub-group A compared to sub-group B (BH-adjusted p value < 0.05). The bar graph shows adjusted p values and numbers of genes (in brackets) mapped to top enriched terms from Swiss-Prot Protein Information Resource (“SP”), InterPro (“I”), protein annotation from the Simple Modular Architecture Research Tool (SMART; “SM”), UniProt sequence annotation (“UP”), and biological processes and molecular function from Gene Ontology (“GO-B” and “GO-M”, respectively; gray dashed line at p value = 0.05). (D) Heat map shows tumor suppressor promoter CGIs with methylation gain in MRT sub-groups compared to hESC (BH-adjusted Welch’s t-test p value < 0.05). See also Table S5.

Figure 6. Comparison of global H3K27me3 marks between MRT and SMARCB1-intact samples

(A) Clustering of H3K27me3 promoter density of MRT and normal cell types. Note that the case PAKLYZ that clustered with normal cell types expressed both SMARCB1 and SMARCA4 (mRNA 48.8 and 37.7 RPKM, respectively). (B) Pathway analysis of gene promoters with lower H3K27me3 levels in MRT compared to normal samples. The bar graph shows the adjusted p values and numbers of genes (in brackets) mapped to top enriched terms from InterPro (“I”) and biological processes from Gene Ontology (“GO-B”) (gray dashed line at p value = 0.05). See also Figure S5 and Table S6.

Figure 7. MRT-specific super-enhancers

(A) Genome-wide H3K27ac profiles were used to define enhancer (right bars in lighter shade) and super-enhancer states (left bars in darker shades) in MRT, fetal brain and hESC cell types. Bar colors indicate different sample types: Fetal brain (blue; “BrainHu04”), hES (orange; “NAAEMA”, “NAAEMB”, “NAAEMC”), and MRT (red). (B) Significantly enriched functional terms for MRT-specific super-enhancer-associated genes (BH-adjusted p value < 0.05). Bar graph shows p values and numbers of genes (in brackets) mapped to top enriched terms from InterPro using DAVID (gray dashed line at p value = 0.05). (C) UCSC browser screen shot showing H3K27ac peaks at the HOXC locus in MRT samples (black), hESCs (orange) and in fetal brain (light blue; “BrainHu04”). The Y-axis scale represents read density. (D, E) The distribution of expression levels of HOTAIR (D) and all HOXC gene family members (E) in MRT, hESC (“ES Cell”), and normal cerebellum tissues from adult (“Adult Cere”) and fetus (“Fetal Cere”). The box indicates the IQR, with the line within the box indicating the median level. The lines below and above the box indicate values within 1.5 IQR from the first and the third quartile, respectively. See also Figure S6 and Table S7.