|Home | About | Journals | Submit | Contact Us | Français|
Malignant rhabdoid tumors (MRT) are rare, lethal tumors of childhood that most commonly occur in the kidney and brain. MRT 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 miRNA sequencing (miRNA-Seq), and histone modification profiling to characterize extra-cranial MRT. Our analyses revealed gene expression and methylation sub-groups and focused on dysregulated pathways, including those involved in neural crest development.
Malignant rhabdoid tumors (MRT) are aggressive pediatric solid tumors with a median age at diagnosis of 11 months (Weeks et al., 1989). MRT may originate from cells in the neural crest lineage (Blatt et al., 1986; Fischer et al., 1989; Ota et al., 1993; Sugimoto et al., 1999) and can occur throughout the body, but they are detected frequently in kidneys (rhabdoid tumors of the kidney; RTK) and brain (Atypical Teratoid Rhabdoid Tumors; AT/RT). Although considered rare (Brennan et al., 2013), the clinical burden of MRT is considerable, with extra-cranial MRT in infants accounting for 18% of renal tumors, 14% of soft tissue tumors and 9% of liver tumors (Brennan et al., 2013). Overall 4-year survival is only 23.2% (Tomlinson, 2005) and thus more effective treatment options are needed.
SMARCB1 loss drives MRT (Biegel et al., 2002; Versteege et al., 1998). In rare cases, SMARCA4 loss is the driver (Schneppenheim et al., 2010). Whole genome sequence analysis of MRT has not been reported, but previous exome (Lee et al., 2012) and single nucleotide polymorphism (SNP) array studies (Hasselblatt et al., 2013; Jackson et al., 2009; McKenna et al., 2008) noted a paucity of somatic mutations in MRT genomes, compatible with the notion that MRT progression is driven predominantly by SMARCB1 loss. Despite this nearly ubiquitous driver event, studies have alluded to clinical heterogeneity in MRT, as there are a few long-term survivors (Ammerlaan et al., 2008; Hirth et al., 2003; Ravindra et al., 2002) and correlations of patient outcome with age at diagnosis and with tumor stage have been reported (Tekautz et al., 2005; Tomlinson, 2005).
SMARCB1 and SMARCA4 are core subunits of the chromatin-remodeling SWItch/Sucrose Non-Fermentable (SWI/SNF) complex, a highly conserved global transcription regulator that can recruit transcription factors to target genes (Kia et al., 2008) or modulate target gene expression by altering nucleosome position (Tolstorukov et al., 2013). Loss of SMARCB1 has been reported in other neoplasms such as epithelioid sarcomas (Sullivan et al., 2013) and schwannomatosis (Hadfield et al., 2008). Previous studies of MRT samples and cell lines described the consequences of SMARCB1 loss, which included dysregulated G0–G1 cell cycle transition (Betz et al., 2002; Zhang et al., 2002); aberrant activation of the sonic hedgehog (Jagani et al., 2010) and WNT/β-catenin signaling pathways (Mora-Blanco et al., 2014); and dysregulated expression of genes involved in embryonic stem cell self-renewal (Wilson et al., 2010) and neural or neural crest development (Gadd et al., 2010).
MRT molecular heterogeneity has also been reported. Torchia et al. (2015) reported two gene-expression sub-groups within AT/RT and correlation of these with survival characteristics. However, the existence of molecularly distinguishable sub-groups has not been described in extra-cranial MRT, and the extent to which the molecular signatures of AT/RT and extra-cranial MRT overlap is not fully understood, with inconsistent reports in the literature (Grupenmacher et al., 2013; Pomeroy et al., 2002).
As part of the Therapeutically Applicable Research To Generate Effective Treatments (TARGET) initiative (http://ocg.cancer.gov/programs/target), we aimed to provide comprehensive molecular profiles and integrative analyses of extra-cranial MRT.
To facilitate comparisons across the spectrum of MRT, we applied whole genome, RNA-Seq, miRNA-Seq, ChIP-Seq and genome-wide DNA methylation analyses (e.g., Roadmap Epigenomics Consortium, 2015) to extra-cranial MRT cases, MRT cell lines (G401, KP-MRT-RY, KP-MRT-NS) and human embryonic stem cell (hESC) lines (H7, H9, H14; Table S1).
We produced whole genome sequence landscapes for 40 MRT cases (Table S1), analyzing sequence data from pairs of tumor and matched normal samples to reveal single nucleotide variants (SNVs), insertions and deletion mutations (InDels), and structural alterations e.g., copy number alterations (CNA) and loss of heterozygosity (LOH). Our analyses confirmed inactivating mutations, copy number losses, or somatic LOH affecting SMARCB1 in 39 of 40 cases. The single case lacking a SMARCB1 alteration harbored somatic LOH and a germ-line deletion of one base in SMARCA4 (Figure 1).
As expected (Lawrence et al., 2013; Lee et al., 2012), somatic SNV rates in MRT were low (median of 612.5 per case; 0.231 mutations / Mb; Figure 1) with 97.1% of mutations in non-coding regions. We observed a correlation between age at diagnosis and mutation rate (Figure S1A). The case PADYZI exhibited an increased proportion of T>G transversions (Figures 1, S1B) and, together with the case PAUEKW, had the highest number of somatic mutations (Figures 1, S1C). Mutations in DNA repair genes were not observed in either case. The PADYZI mutation spectrum was most strongly correlated to signatures 9 and 17 from Alexandrov et al. (2013), characterized by T>G mutations (Pearson correlation coefficients = 0.607 and 0.962, respectively). The mutation spectrum of PAUEKW was most strongly correlated with signatures 1A and 1B (Pearson correlation coefficients = 0.877 and 0.925, respectively), characterized by C>T mutations in the context of NpCpG trinucleotides that are presumed to arise from spontaneous deamination of 5-methyl-cytosine (Alexandrov et al., 2013).
We observed a total of 204 non-synonymous (NS) amino acid substitutions (median = 5 / case, range 1 – 17). Besides SMARCB1, DSCAM was the only recurrent target of somatic NS alteration, with two cases exhibiting heterozygous somatic NS substitutions (p.Val424Ile and p.Ser1354Thr; Figure 1). While DSCAM alterations may be passenger mutations, somatic DSCAM alterations were detected at frequencies of ~10% or more in prostate, uterine, colorectal and lung cancers in the Cancer Genome Atlas (TCGA) data sets as reported via cBioPortal (Cerami et al., 2012; Gao et al., 2013).
We extended our analysis to introns and regions 2 kb upstream of transcriptional start sites, looking for recurrent somatically mutated genes associated with dysregulated gene expression. We identified SPECC1L intron mutations in 6 cases and KCNJ3 intron mutations in 8 cases (Figures S1D, S1E). The difference of mean gene expression values between cases with mutations in these genes and non-mutated cases was significant (BH-corrected Wilcoxon p values = 1.82e-10 and 2.30e-08 for SPECC1L and KCNJ3, respectively), with SPECC1L mutations associated with decreased SPECC1L expression (log2 fold change (FC) = −1.45; Figure S1F) and KCNJ3 mutations associated with increased KCNJ3 expression (log2 FC = 1.20; Figure S1G).
Eleven loci were affected by recurrent somatic copy number deletions, and four loci were affected by recurrent somatic gains (Figure S1H). 9 of 15 loci with CNA were on chromosome 22, on which SMARCB1 resides (Figure S1I). Genes within gains did not exhibit significant expression increases, but 115 genes within deletions exhibited significant expression decreases compared to cases without copy number deletions (BH-adjusted Wilcoxon p value < 0.05; Table S2). All but 2 chromosome 22 deletions involved SMARCB1 (Figure 1), and these were focal deletions affecting NF2 and LIF. Regions 5q34 (5 cases); 5q31.3 – 5q32 (2 cases) and 7q35 (2 cases) exhibited copy number (CN) losses (Figure S1H). We observed significant expression decreases of TCERG1, LARS, YIPF5 and RBM27 in the 5q31.3 – 5q32 region (BH-adjusted Wilcoxon p value < 0.05) compared to cases retaining this region.
Verification of gene fusions confirmed 18 somatic and 8 germ-line fusion events (Table S2). 22 of the 26 fusions involved chromosome 22 (Figure S1J), with 7 of these fusions arising as a consequence of SMARCB1 deletion. SPECC1L was a fusion partner in 4 of these 7 cases. Although these fusions may be passenger events, SPECC1L alterations have been linked to phenotypes in human patients (Saadi et al., 2011) and animal models (Gfrerer et al., 2014; Saadi et al., 2011) that are consistent with disruption of neural crest cell migration.
MRT are poorly differentiated cancers and the cell of origin is unknown, complicating the selection of samples for comparative analyses. MicroRNA (miRNA) profiles exhibit superior accuracy compared to RNA expression profiles in distinguishing tissue types of both malignant and normal cells (Calin and Croce, 2006; Lu et al., 2005; Rosenfeld et al., 2008) and have been used to identify tissues of origin of poorly differentiated tumors (Ferracin et al., 2011). We thus analyzed miRNA-Seq data from 66 primary MRTs (Table S1), detecting 535 mature miRNAs in MRT cases. Using these, we performed unsupervised clustering analysis to compare MRT miRNA profiles to TCGA and other miRNA profiles from 11819 cases representing 37 cancer types and 27 normal tissue types. MRT cases segregated into two groups (Figures 2, S2; Table S3). The larger group (n=57) clustered with normal cerebellum samples (“MD_NORM”, “GBM_NORM”) and the TCGA category that contains paragangliomas and pheochromocytomas (“PCPG_TUM”), which are cancers of neural crest origin (Bolande, 1974). The smaller group (n=9) clustered with synovial sarcomas, which exhibit SWI/SNF dysregulation (Kadoch and Crabtree, 2013) and may be of neural origin (Ishibe et al., 2008).
We identified candidate miRNA targets, applying an approach (Lim et al., 2015) we developed to identify targets with at least one miRNA binding site predicted by both the miRanda (John et al., 2004) and TargetScan (Friedman et al., 2009) algorithms, and where the miRNA:mRNA pair showed anti-correlated gene expression across 40 MRT samples (Table S1). GO analysis (Huang et al., 2009a; 2009b) of targets with the most significant candidate miRNA:mRNA interactions (Spearman correlation rho <−0.65) indicated that candidate miRNA targets included those involved in neuron development and differentiation (e.g., VANGL2, ULK1, NGFR; Table S3), compatible with the notion that these processes are targets of miRNA-mediated dysregulation in MRT. We also observed that SMARCE1 was a putative target of miR-200a, compatible with the notion that miRNA dysregulation in MRT might influence expression of SWI/SNF complex members (Figure S3A).
NMF consensus clustering (Gaujoux and Seoighe, 2010) of miRNA-Seq profiles (n=66) identified an optimum of two sub-groups (Figures 3, S3B, S3C). 17 miRNAs were over-expressed in miRNA sub-group 1, and 12 were over-expressed in miRNA sub-group 2 (Figure 3; log2 fold change >1, FDR <0.05). All 5 members of the EMT-repressing miR-200 family (miR-200a, miR-200b, miR-200c, miR-141, miR-429; Feng et al., 2014) were expressed at lower levels in sub-group 1 compared to sub-group 2. The lower expression levels of the miR-200 family in sub-group 1 are compatible with the notion that EMT may be more active in sub-group 1 relative to sub-group 2. Our observation is compatible with previous reports of EMT regulation by SWI/SNF members (Huang et al., 2008; Sánchez-Tilló et al., 2010).
We used RNA-Seq to analyze gene expression in 40 MRT cases and 3 hESC lines (Table S1), the latter selected based on studies indicating similarities between MRT and hESC in the expression of stem-cell associated genes (Wilson et al., 2010) and embryonic stem cell markers (Deisch et al., 2011; Venneti et al., 2011). Informed by our miRNA analyses (Figure 2), we obtained RNA-Seq data from 4 fetal cerebellum samples to use as additional normal comparators. 2713 and 4502 genes were differentially expressed between MRT cases and hESC lines and between MRT cases and fetal cerebellum samples, respectively (FDR <0.05, log2FC >1; Table S4). Identification of genes common to both comparisons resulted in 398 genes with increased abundance in MRT vs both comparators and 615 genes with decreased abundance in MRT vs both comparators. In MRT, the most significantly over-expressed genes were linked to immune function (e.g., IGKC, IGKJ5) and embryonic development (e.g., MGP, LUM; Figure 4A; Table S4), in agreement with previously reported associations of SWI/SNF loss with immune (Agalioti et al., 2000; Cui et al., 2004; Golding et al., 1999; Morshead et al., 2003) and developmental dysregulation (Li, 2002). The most significantly under-expressed genes have been linked to aspects of embryonic development (e.g., ZIC3, SOX3) and neuron function (e.g., GABRA3, CADPS; Figure 4B; Table S4).
We used the top 25% most variably expressed protein-coding genes (n=3179) to assess whether mRNA expression supported the existence of distinct subgroups. NMF consensus clustering revealed an optimum of 2 subgroups (Figures 4C, S4A, S4B), a number generally supported by hierarchical clustering and principal component analysis (PCA; Figures 4D, ,4E).4E). Organ site was significantly associated with sub-group 1 (Fisher’s exact test p value = 0.04), which included all 6 extra-renal MRT cases from liver and soft tissues (Figure 4C; Table S1).
To address gene expression differences between sub-groups, we identified 880 genes differentially expressed (FDR < 0.05) between sub-groups 1 and 2 (Figure 4F; Table S4). In sub-group 1, the 20 most significantly over-expressed genes (log2 FC >2.5) were immunoglobulins, including those noted above, and genes linked to BMP-signaling (e.g., BMP4, SOSTDC1) and differentiation (e.g., DLK1, MEOX2). In sub-group 2, the 20 most significantly over-expressed genes (log2 FC >6) were linked to cell adhesion and migration (e.g., PCDH18, SMOC2), WNT signaling (e.g., WNT5A, HIC1) and differentiation (TCF21, MEIS1). GO analysis indicated significant enrichment of neural crest development and neural crest differentiation terms (BH-adjusted p value = 0.049; Table S4), consistent with the MRT cell of origin deriving from the neural crest lineage (Blatt et al., 1986; Fischer et al., 1989; Ota et al., 1993; Sugimoto et al., 1999). Further, the enriched categories spanned several stages of neural crest development (Simões-Costa and Bronner, 2015; Figure 4C). We also noted enhanced representation of annotated tumor suppressors and oncogenes among differentially expressed genes (hypergeometic p value = 1.03e-39; Table S4).
Our study notes the possibility of gene expression sub-groups in extra-cranial MRT, although gene expression sub-groups were identified in AT/RT (Torchia et al., 2015). To assess whether our MRT sub-groups were more similar to AT/RT or to RTK, we used data from Grupenmacher et al. (2013), which analyzed gene expression differences between AT/RT and RTK. We compared the genes differentially expressed between MRT sub-groups to 29 genes reported by Grupenmacher et al. (2013) as over-expressed in AT/RT relative to RTK (Table S4). Of these, 11 were differentially expressed between MRT subgroups 1 and 2 (FDR < 0.05). 10 of the 11 genes were over-expressed in subgroup 1 (Table S4), including genes linked to neuronal differentiation (e.g., BEX1, FAM5C/BRINP3; Coyle et al., 2011; Terashima et al., 2010; Vilar et al., 2006) or neural crest differentiation (e.g., ENC1; Rabadán et al., 2013). Conversely, of the 92 genes reported (Grupenmacher et al., 2013) as over-expressed in RTK relative to AT/RT, 21 were differentially expressed between sub-group 1 and 2, and all were over-expressed in sub-group 2 (one-sided Fisher’s exact p value = 1.193e-13; Table S4). These results are compatible with the notion that MRT sub-group 1 may be more similar to AT/RT, and MRT sub-group 2 may be more similar to RTK.
Previous reports have linked SWI/SNF complex action to CpG methylation (e.g., Kia et al., 2008). We thus sought to explore the relationship between SMARCB1 loss and genome-wide DNA methylation patterns using whole genome bisulfite sequencing analyses of 40 MRT cases, 3 MRT cell lines and 3 hESC lines (Table S1). We obtained 4 neural progenitor cell samples (NPC; Roadmap Epigenomics Consortium, 2015) to use as normal comparators. Unsupervised clustering of promoter CpG island (CGI) methylation revealed two MRT subgroups (Figure 5A). Sub-group A exhibited higher overall promoter CGI methylation levels compared to sub-group B (Welch’s t-test p value = 4.629e-16), a feature that was conserved outside CGIs (Welch’s t-test p value = 2.2e-16; Figure 5B). MRT cell lines had the highest degree of promoter CGI methylation and clustered with sub-group A. In contrast, sub-group B MRT clustered with hESC and had lower overall methylation levels both within and outside CGIs. MRT promoter-associated CGIs in both sub-groups were hypermethylated compared to NPCs and embryonic stem cells, while non-CGI regions were hypomethylated (Figure 5B), consistent with patterns generally observed in cancer epigenomes (Feinberg and Vogelstein, 1983; Toyota et al., 1999). Correlating methylation sub-groups to clinical patient data revealed enrichment for patients > 1 year old in sub-group A (Figure 5A; one-sided Fisher’s exact p value = 0.0013). The only patient younger than one year in sub-group A was 354 days old. Pathway analysis (Huang et al., 2009a; 2009b) of promoters that gained methylation in sub-group A showed significant enrichment in homeobox terms (BH-adjusted Welch’s t test p value < 0.05; Figure 5C).
Relative to hESC, we noted a significant increase in the number of tumor suppressor gene (TSG) promoters that gained methylation compared to TSG promoters that lost methylation (one-sided Fisher’s exact p value = 0.02041; Figure 5D; Table S5). These included TSGs (annotated in the TSGene database; Zhao et al., 2013), reported as methylated in other cancer types, including RASSF1 (Dammann et al., 2003), IRX1 (Guo et al., 2010), TWIST2 (Thathia et al., 2012), DLEC1 (Gebhard et al., 2006; Qiu et al., 2008; Wang et al., 2012; Ying et al., 2009; Zhang et al., 2010), and TBX5 genes (Yu et al., 2010).
SMARCB1 loss has been associated with gains in global levels of H3K27me3 in extra-cranial MRT and AT/RT models (Wilson et al., 2010). We thus assessed whether promoter H3K27me3 density could discriminate MRT samples from normal neuronal, progenitor and kidney cell types.
To assess MRT promoter H3K27me3 density, we performed H3K27me3 ChIP-Seq analysis of 35 primary MRT samples, 3 MRT cell lines, 2 hESC lines and 1 normal kidney sample (Figure S5; Table S1). We also obtained H3K27me3 ChIP-Seq data from 2 normal fetal brain samples and 2 NPC samples from the Roadmap Epigenomics Consortium (2015; Table S1). We defined gene promoters as regions +/− 2 kb of transcription start sites, and identified promoters with differential H3K27me3 densities in MRT samples compared to normal samples. Hierarchical clustering of the 200 promoters most differentially methylated between MRT and normal comparators revealed three gene clusters. One exhibited increased H3K27me3 density relative to normal comparators (Figure 6A) while the other two clusters exhibited decreased H3K27me3 density. Pathway analysis (Huang et al., 2009a; 2009b) of the genes associated with the 42 promoters with increased H3K27me3 density did not reveal a statistically significant enrichment of any functional category. In contrast, analysis of genes associated with the 158 promoters exhibiting decreased H3K27me3 density revealed statistically significant enrichment in homeobox, HOX and homeodomain terms (FDRs ranged from 4.17e-44 to 1.67e-37; Figure 6B, Table S6). Other enriched terms were related to pattern formation, morphogenesis and organogenesis, and also appeared to be driven primarily by homeobox genes. Enrichment of terms related to DNA binding, transcription and transcription regulation was also noted (Table S6).
Consistent with the observation that homeobox genes had differentially methylated promoters, H3K27me3 analyses supported the notion that epigenetic reprogramming of homeobox genes may be associated with SMARCB1 loss. However, despite nearly universal SMARCB1 loss, this reprogramming is heterogeneous within our MRT cohort.
Super-enhancers, defined by high regional H3K27ac density compared to typical enhancers, are important regulatory entities in cancer and are often associated with driver pathways specific to the cancer type from which they were derived (Hnisz et al., 2013). We performed H3K27ac ChIP-Seq on 10 primary MRT samples, 3 MRT cell lines and 3 hESC lines, using the latter for comparison to the MRT samples (Table S1). In addition, we obtained ChIP-Seq data from 1 normal fetal brain sample from the Roadmap Epigenomics Consortium (2015; Table S1) to use as a normal comparator. We analyzed H3K27ac peaks to define enhancers and ranked these based on the H3K27ac peak enrichment signals and their local density to identify super-enhancers (Figures 7A, S6) as described in previous studies (Hnisz et al., 2013; Whyte et al., 2013). To reveal candidate active regulatory regions unique to MRT, we identified super-enhancers that were in >= 50% of MRT samples and that were absent in normal fetal brain and hESC. This yielded 136 MRT-specific super-enhancers associated with 197 genes (Table S7). Analysis of these genes revealed enrichment of terms associated with DNA binding, homeobox and homeoprotein (FDR < 0.0001; Figure 7B; Table S7). These enrichments appeared to be driven by members of the HOXA, HOXB, HOXC, HISTH1 and HISTH2 gene clusters. Super-enhancers were also associated with FLT3LG and STAT3 (Bromberg et al., 1999; Lynch et al., 1997), both of which are involved in the regulation of cellular differentiation (Laouar et al., 2003) and self-renewal capacity in embryonic stem and neural crest cells (Ying et al., 2003).
The MRT specific super-enhancers blanketing the histone H2 gene cluster and the HOXC cluster (Figure 7C) were striking and in the latter case included the regulatory RNA HOTAIR, providing a possible explanation for the increase in HOTAIR expression previously reported in AT/RT (Chakravadhanula et al., 2014) that we also observed in MRT vs normal comparators (Figure 7D). We also observed high expression of HOXC gene members in MRT compared to normal cell types (Figure 7E). Enrichment of HOX gene clusters in super-enhancer regions is consistent with our observation of loss of H3K27me3 at these loci.
We present here a reference genomic landscape of extra-cranial MRT, and as a demonstration of their utility, use the data to implicate genes and pathways in MRT biology. Although it was clear that the dominant driver alteration in MRT is SMARCB1 loss, our data indicate that the effects of SMARCB1 loss on transcriptional regulation are not uniform across the MRT cases.
Clustering of miRNA-Seq data of MRT with those from other tumor and normal cell types revealed two sub-groups. The smaller sub-group clustered with synovial sarcomas, which exhibit SWI/SNF dysregulation (Kadoch and Crabtree, 2013) and may be of neural origin (Ishibe et al., 2008). The larger sub-group clustered with normal cerebella, pheochromocytomas and paragangliomas, both of which are cancers of neural crest origin (Bolande, 1974). Our results are thus compatible with the notion that extra-cranial MRT may be derived from neural crest cells. This notion was further reinforced by our RNA-Seq analyses, which revealed the differential expression of genes that regulate neural crest induction, neural differentiation and axon guidance.
Clustering of CGI promoter methylation revealed two sub-groups with distinct methylation profiles that correlated with age at diagnosis. The sub-group consisting primarily of patients older than one year exhibited gain of methylation at CGI promoters, particularly of homeobox genes and tumor suppressor genes, which is consistent with the notion that the impact of SMARCB1 loss on CGI methylation is heterogeneous.
The genome-wide profile of H3K27me3 density at gene promoters distinguished MRT from other SMARCB1-intact cell types. Loss of H3K27me3 marks was observed at promoters of homeobox genes, and MRT-specific super-enhancers were detected at the HOXA, HOXB and HOXC clusters, consistent with the notion that epigenetic reprogramming may promote expression of homeobox genes and contribute to dysregulated developmental processes in MRT.
Taken together, our observations are compatible with the notion that the MRT cell of origin is related to neural crest cells, and that the consequences of SMARCB1 loss in MRT are heterogeneous at the epigenetic and gene expression levels.
67 primary extra-cranial MRT (58 from kidneys, 7 from soft tissues and 2 from liver) and 40 matched normal (16 adjacent kidney and 24 peripheral blood) samples were obtained from patients registered on the Children’s Oncology Group (COG) protocol. Included were pre-therapy tumor and normal DNA from peripheral blood or kidney from rhabdoid tumors, registered on the National Wilms Tumor Study Group 5 or on COG AREN03B2 banked by the COG Biopathology Center with parental informed consent. Studies were performed with the approval of the University of British Columbia - British Columbia Cancer Agency Research Ethics Board (REB number H09-02558). Three MRT cell lines (G401, KP-MRT-RY and KP-MRT-NS; Garvin et al., 1993; Katsumi et al., 2008; Sugimoto et al., 1999), and three human embryonic stem (hES) cell lines (H7, H9 and H14; Thomson et al., 1998) were used in our study. The complete sample names and other details are provided in Table S1. Details of sequencing library construction are in Supplemental Information. Tumor contents were estimated using APOLLOH (median = 88.0%; Ha et al., 2012).
Details are provided in Supplemental Information.
To facilitate comparisons between cranial and extra-cranial MRT and to study the genome-wide consequences of SMARCB1 loss, we performed comprehensive genomic analyses of pediatric extra-cranial MRT cases. Despite being almost uniformly driven by SMARCB1 loss, we observed molecular heterogeneity within extra-cranial MRT. Methylation sub-groups associated with patient age, and differentially expressed members of the EMT, BMP- and WNT-mediated developmental pathways were observed in gene expression sub-groups, as was dysregulation of genes involved in neural crest development. We present evidence for altered epigenetic states, noting the existence of MRT-specific super-enhancers and dysregulation of homeobox genes.
Chun et al. perform integrated molecular analyses of extra-cranial malignant rhabdoid tumors (MRT) and show that, although SMARCB1 loss drives nearly all MRTs, there are two sub-groups of MRT that are associated with patient age and differentially expressed genes.
We thank the Library Construction, Biospecimen, Sequencing and Bioinformatics teams at Canada’s Michael Smith Genome Sciences Centre for technical assistance, Dr. Karen Novik for project management, and Dr. Robert Camfield for reviewing the manuscript. M.A.M. acknowledges the BC Cancer Agency, the BC Cancer Foundation, the Canada Foundation for Innovation, the Canada Research Chairs program, Genome British Columbia and the Canadian Institutes of Health Research (CIHR-262119 and FDN-143288). This project has been funded in whole or in part with Federal Funds from the National Cancer Institute, National Institutes of Health, under Contract No. NO1-CO-12400. The content of this publication does not necessarily reflect the views or policies of the Department of Health and Human Services, nor does mention of trade names, commercial products, or organizations imply endorsement by the U.S. Government.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
Author ContributionsM.A.M., E.J.P., D.S.G. and M.H. conceived the study, led the experimental design and provided editorial input. H-J.E.C., E.L.L., A.H-M., S.S., M.H. and M.A.M. interpreted the data and designed the presentation of data. H-J.E.C., M.A.M., E.L.L., A.H-M. and M.H. wrote the paper. The Children’s Oncology Group provided primary tumor, normal and cell line samples. E.J.P. provided clinical data. M.D.T. provided cerebellum RNA-Seq data. H-J.E.C., E.L.L., A.H-M., S.S., K.L.M., J.Q.Q., M.B., A.C., K.T., I.S., D.L.P., K.Z., A.H., W.L., R.G., M.N., V.L. and E.P. contributed to analyses. N.T., T.W., E.C., Y.M., and S.J.M.J. provided bioinformatics support. Y-J.Z., J.E.S., A.J.M. and R.A.M. performed library construction and sequencing. All authors read and approved the final manuscript.
Data availability and accession number
Sequencing reads and analyzed data files are available through NCBI dbGaP (accession number phs000470). Additional data are available at http://target.nci.nih.gov/dataMatrix/TARGET_DataMatrix.html.