Genome-wide profiles of extra-cranial malignant rhabdoid tumors reveal heterogeneity and dysregulated developmental pathways

Authors

Hye-Jung E. Chun

Emilia L. Lim

Alireza Heravi-Moussavi

Saeed Saberi

Karen L. Mungall

Mikhail Bilenky

Annaick Carles

Elizabeth J. Perlman

Martin Hirst

Marco A. Marra

Doi

10.1016/j.ccell.2016.02.009

PMID: 26977886 · DOI: 10.1016/j.ccell.2016.02.009 · Journal: Cancer Cell (2016)

TL;DR

Chun et al. produced an integrative genome-wide reference of 40 extra-cranial malignant rhabdoid tumors (MRT, mostly pediatric kidney and soft-tissue cases banked through the TARGET initiative and the Children’s Oncology Group), combining whole-genome sequencing, whole-genome bisulfite sequencing, RNA-Seq, miRNA-Seq and H3K27me3 / H3K27ac ChIP-Seq. They confirm that SMARCB1 loss is the dominant driver in 39/40 cases (the remaining case is driven by SMARCA4 loss), and show that despite this near-uniform driver, MRT split into reproducible miRNA, mRNA and methylation sub-groups with distinct developmental and epigenetic signatures, including loss of H3K27me3 and gain of MRT-specific super-enhancers at HOX clusters. miRNA clustering links extra-cranial MRT to neural-crest-derived tissues.

Cohort & data

  • 40 MRT tumor/normal pairs analyzed by whole-genome sequencing (median tumor content 88.0% by APOLLOH); 66 primary MRTs profiled by miRNA-Seq; 40 MRT cases plus 3 hESC lines profiled by RNA-Seq with 4 fetal cerebellum normals; 40 MRT cases, 3 MRT cell lines (G401, KP-MRT-RY, KP-MRT-NS) and 3 hESC lines (H7, H9, H14) plus 4 NPCs profiled by whole-genome bisulfite sequencing; 35 primary MRTs, 3 MRT cell lines, 2 hESC lines, 1 normal kidney, 2 fetal brains and 2 NPCs profiled by H3K27me3 ChIP-Seq; 10 MRT samples + 3 cell lines + 3 hESC + 1 fetal brain by H3K27ac ChIP-Seq (PMID:26977886).
  • 67 primary extra-cranial MRT cases overall (58 kidney, 7 soft tissue, 2 liver) and 40 matched normals (16 adjacent kidney, 24 peripheral blood) banked by the COG Biopathology Center under the National Wilms Tumor Study Group 5 / COG AREN03B2 protocols (PMID:26977886).
  • Cancer types: rhabdoid tumor of the kidney (MRT) and extra-renal / liver MRT (MRTL).
  • Dataset: mrt_bcgsc_2016. Sequencing reads and analyzed files are deposited in NCBI dbGaP (phs000470) with additional matrices on the TARGET data portal (PMID:26977886).
  • Assays: whole-genome-seq, whole-genome-bisulfite-seq, rna-seq, mirna-seq, chip-seq (H3K27me3, H3K27ac); sub-group discovery used consensus-hierarchical-clustering / NMF.

Key findings

  • SMARCB1 is near-universal: inactivating mutations, copy-number losses or somatic LOH at SMARCB1 were detected in 39/40 cases; the single SMARCB1-intact case carried somatic LOH plus a one-base germ-line SMARCA4 deletion (PMID:26977886).
  • Quiet genomes outside the driver: median 612.5 somatic SNVs per case (0.231 mutations/Mb), with 97.1% of mutations in non-coding regions, consistent with prior MRT exome reports. 204 non-synonymous substitutions in total (median 5/case, range 1–17) (PMID:26977886).
  • DSCAM is the only recurrent non-SMARCB1 NS target: two cases harbored heterozygous somatic non-synonymous DSCAM substitutions (p.Val424Ile, p.Ser1354Thr); flagged as possibly passenger (PMID:26977886).
  • Recurrent intronic mutations linked to expression: SPECC1L intron mutations in 6 cases were associated with decreased SPECC1L expression (log2 FC = −1.45; BH-Wilcoxon p = 1.82e-10), and KCNJ3 intron mutations in 8 cases were associated with increased KCNJ3 expression (log2 FC = 1.20; p = 2.30e-08) (PMID:26977886).
  • Chr22 dominates structural variation: 9/15 recurrent CNA loci and 22/26 verified gene fusions involve chromosome 22; 7 of those fusions arise as direct consequences of SMARCB1 deletion, with SPECC1L as fusion partner in 4/7. Two chromosome-22 deletions did not include SMARCB1 and instead focally deleted NF2 and LIF (PMID:26977886).
  • Two mutational signatures stand out: case PADYZI was dominated by T>G transversions (Pearson correlation to Alexandrov signatures 9 and 17 of 0.607 and 0.962); case PAUEKW matched signatures 1A/1B (correlations 0.877 and 0.925); neither carried DNA-repair gene mutations (PMID:26977886).
  • miRNA sub-groups link MRT to neural-crest lineages: unsupervised clustering of 535 miRNAs against TCGA / normal references (11,819 cases, 37 cancer types, 27 normal tissues) split 66 MRTs into a larger group (n=57) clustering with normal cerebellum and TCGA pheochromocytoma/paraganglioma, and a smaller group (n=9) clustering with synovial sarcomas. NMF on miRNA-Seq independently identified 2 sub-groups in which all 5 miR-200 family members (miR-200a/b/c, miR-141, miR-429) were lower in sub-group 1, suggesting more active EMT in sub-group 1 (PMID:26977886).
  • Putative miRNA targets enriched for neuron-development genes (VANGL2, ULK1, NGFR); SMARCE1 is a putative miR-200a target (PMID:26977886).
  • mRNA sub-groups recapitulate AT/RT vs RTK split: NMF on the top-25% most-variable protein-coding genes (n=3,179) yielded 2 sub-groups (n=40); organ site was associated with sub-group 1 (all 6 extra-renal cases; Fisher’s exact p=0.04). Sub-group 1 over-expressed immunoglobulins, BMP4, SOSTDC1, DLK1, MEOX2; sub-group 2 over-expressed PCDH18, SMOC2, WNT5A, HIC1, TCF21, MEIS1. Comparison with Grupenmacher 2013 markers showed sub-group 1 is more AT/RT-like (10/11 of the AT/RT-up genes shared) and sub-group 2 is more RTK-like (21/21 of the RTK-up shared genes; one-sided Fisher p = 1.193e-13) (PMID:26977886).
  • Most over- and under-expressed genes vs hESC + cerebellum: top induced include immune (IGKC, IGKJ5) and developmental (MGP, LUM) genes; top repressed include developmental (ZIC3, SOX3) and neuronal (GABRA3, CADPS) genes (PMID:26977886).
  • Two methylation sub-groups correlate with age: WGBS-based promoter CGI clustering produced sub-group A (higher CGI methylation; Welch t-test p = 4.629e-16) enriched for patients >1 year (one-sided Fisher p = 0.0013, with the only sub-group-A patient under one year being 354 days old) and sub-group B (clustering with hESC). MRT cell lines clustered with sub-group A. Sub-group A promoters that gained methylation were enriched for homeobox terms; tumor-suppressor promoters (RASSF1, IRX1, TWIST2, DLEC1, TBX5) preferentially gained methylation versus hESC (one-sided Fisher p = 0.02041) (PMID:26977886).
  • Loss of H3K27me3 at homeobox promoters: H3K27me3 ChIP-Seq on 35 MRTs vs normals identified 158 promoters with decreased H3K27me3 strongly enriched for homeobox / HOX / homeodomain terms (FDR 4.17e-44 to 1.67e-37); 42 promoters with increased H3K27me3 had no significant enrichment (PMID:26977886).
  • MRT-specific super-enhancers blanket HOX and histone clusters: 136 super-enhancers (present in ≥50% of MRT and absent in fetal brain + hESC) annotate 197 genes enriched for DNA-binding/homeobox terms (FDR < 0.0001), driven by HOXA, HOXB, HOXC, HISTH1 and HISTH2 clusters; super-enhancers also fall on FLT3LG and STAT3. The HOXC super-enhancer covers the HOTAIR lncRNA, paralleling and extending prior reports of HOTAIR over-expression in AT/RT (PMID:26977886).

Genes & alterations

  • SMARCB1 — biallelic inactivation in 39/40 extra-cranial MRT (mutations, deletions, somatic LOH); the dominant driver. Loss is associated with downstream loss of H3K27me3 at homeobox promoters and gain of MRT-specific super-enhancers at HOX clusters (PMID:26977886).
  • SMARCA4 — sole driver in the one SMARCB1-intact case, via somatic LOH plus a germ-line one-base deletion (PMID:26977886).
  • DSCAM — recurrent (2/40) heterozygous somatic missense substitutions (p.Val424Ile, p.Ser1354Thr); only non-SMARCB1 recurrently NS-mutated gene in the cohort (PMID:26977886).
  • SPECC1L — intronic mutations in 6/40 cases associated with reduced expression; partner in 4/7 SMARCB1-deletion-driven fusions; alterations consistent with disrupted neural-crest cell migration (PMID:26977886).
  • KCNJ3 — intronic mutations in 8/40 cases associated with increased expression (PMID:26977886).
  • NF2, LIF — focally deleted by the only chromosome-22 deletions that spare SMARCB1 (PMID:26977886).
  • SMARCE1 — putative miR-200a target, suggesting miRNA-mediated dysregulation of SWI/SNF members (PMID:26977886).
  • HOTAIR — covered by an MRT-specific super-enhancer in the HOXC cluster and over-expressed in MRT vs normals, paralleling AT/RT (PMID:26977886).
  • BMP4, DLK1, MEOX2 — among the most over-expressed genes in mRNA sub-group 1 (BMP signaling / differentiation) (PMID:26977886).
  • WNT5A, PCDH18, TCF21, MEIS1 — among the most over-expressed genes in mRNA sub-group 2 (cell adhesion/migration, WNT, differentiation) (PMID:26977886).
  • ZIC3, SOX3 — among the most under-expressed developmental genes vs hESC + fetal cerebellum (PMID:26977886).
  • FLT3LG, STAT3 — flagged within MRT-specific super-enhancers; both have prior roles in stem-cell self-renewal and differentiation (PMID:26977886).
  • Tumor-suppressor promoters with sub-group-A hypermethylation: RASSF1, IRX1, TWIST2, DLEC1, TBX5 (PMID:26977886).

Clinical implications

  • No prospective treatment claims — this is a descriptive reference landscape, not a therapy or biomarker-validation study. The authors motivate the work by noting that overall 4-year MRT survival is only 23.2% and that more effective treatment options are needed (PMID:26977886).
  • Heterogeneity despite a single driver: the data argue against treating extra-cranial MRT as a molecularly uniform disease. miRNA, mRNA and methylation sub-groups exist, and the mRNA sub-groups recapitulate the AT/RT vs RTK distinction reported by Grupenmacher et al. (2013), suggesting sub-group-specific therapeutic hypotheses are warranted (PMID:26977886).
  • Methylation sub-group A correlates with age >1 year at diagnosis, raising the possibility that DNA-methylation status could refine MRT risk stratification beyond age and stage alone (PMID:26977886).
  • Epigenetic vulnerabilities: the authors highlight loss of H3K27me3 at homeobox promoters and HOX-cluster super-enhancers as candidate epigenetic dependencies in MRT (PMID:26977886).

Limitations & open questions

  • No drug or treatment outcomes: the manuscript presents no patient-level survival analyses tied to the molecular sub-groups beyond the age-at-diagnosis correlation; therapeutic actionability of the HOX/super-enhancer / EMT findings is hypothesis-generating only (PMID:26977886).
  • Cohort skews to renal MRT: 58/67 primary tumors are kidney, 7 soft tissue, 2 liver — limiting power for site-specific biology, particularly for MRTL/extra-renal MRT (PMID:26977886).
  • Cell of origin not directly proven: miRNA clustering with neural-crest-derived tissues (PCPG, normal cerebellum) is suggestive but indirect; the authors explicitly note that the MRT cell of origin remains unknown (PMID:26977886).
  • Recurrent intronic SPECC1L / KCNJ3 mutations: associated with expression changes but mechanism (regulatory disruption vs sequencing artifact in introns of frequently rearranged regions) is not resolved here (PMID:26977886).
  • Inconsistency with prior literature on MRT vs AT/RT overlap: the authors note the literature is inconsistent (Grupenmacher 2013 vs Pomeroy 2002); their data partially reconcile these by showing that extra-cranial MRT itself is heterogeneous along an AT/RT-like vs RTK-like axis (PMID:26977886).
  • Sample-size differences across assays (40 WGS, 66 miRNA-Seq, 40 RNA-Seq + 40 WGBS, 35 H3K27me3 ChIP-Seq, 10 H3K27ac ChIP-Seq) make integrated multi-omic per-case analyses uneven across the cohort (PMID:26977886).

Citations from this paper used in the wiki

  • “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.” — Results, MRT whole genome sequence landscape (p. 3).
  • “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).” — Results, MRT whole genome sequence landscape (p. 3–4).
  • “We identified SPECC1L intron mutations in 6 cases and KCNJ3 intron mutations in 8 cases … BH-corrected Wilcoxon p values = 1.82e-10 and 2.30e-08 for SPECC1L and KCNJ3, respectively.” — Results (p. 4).
  • “22 of the 26 fusions involved chromosome 22 … with 7 of these fusions arising as a consequence of SMARCB1 deletion. SPECC1L was a fusion partner in 4 of these 7 cases.” — Results (p. 4).
  • “MRT cases segregated into two groups … The larger group (n=57) clustered with normal cerebellum samples … and the TCGA category that contains paragangliomas and pheochromocytomas … The smaller group (n=9) clustered with synovial sarcomas.” — Results, miRNA analysis (p. 4–5).
  • “Sub-group A exhibited higher overall promoter CGI methylation levels compared to sub-group B (Welch’s t-test p value = 4.629e-16) … enrichment for patients > 1 year old in sub-group A (Figure 5A; one-sided Fisher’s exact p value = 0.0013).” — Results, methylation sub-groups (p. 7).
  • “This yielded 136 MRT-specific super-enhancers associated with 197 genes … 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.” — Results, super-enhancers (p. 8).
  • “Overall 4-year survival is only 23.2% (Tomlinson, 2005) and thus more effective treatment options are needed.” — Introduction (p. 2).

This page was processed by crosslinker on 2026-05-14.