Comprehensive genomic analysis of rhabdomyosarcoma reveals a landscape of alterations affecting a common genetic axis in fusion-positive and fusion-negative tumors

Authors

Jack F. Shern

Li Chen

Juliann Chmielecki

Jun S. Wei

Rajesh Patidar

Mara Rosenberg

Lauren Ambrogio

Daniel Auclair

Jianjun Wang

Young K. Song

Catherine Tolman

Laura Hurd

Hongling Liao

Shile Zhang

Dominik Bogen

Andrew S. Brohl

Sivasish Sindiri

Daniel Catchpoole

Thomas Badgett

Gad Getz

Jaume Mora

James R. Anderson

Stephen X. Skapek

Frederic G. Barr

Matthew Meyerson

Douglas S. Hawkins

Javed Khan

Doi

10.1158/2159-8290.CD-13-0639

PMID: 24436047 · DOI: 10.1158/2159-8290.CD-13-0639 · Journal: Cancer Discovery (2014)

TL;DR

Shern et al. performed whole-genome (n=44), whole-exome (n=103), whole-transcriptome (n=80), and high-resolution SNP-array analyses on a total of 147 rhabdomyosarcoma (RMS) tumor/normal pairs in a collaboration between the NCI, Children’s Oncology Group, and the Broad Institute. They show that RMS partitions cleanly by PAX-fusion status rather than by classical ARMS/ERMS histology: PAX-fusion-positive (PFP) tumors carry few additional somatic mutations (mean 6.4 non-synonymous mutations/tumor) and rely on amplifications (MYCN, CDK4, MIR17HG) or imprinted-locus loss, whereas PAX-fusion-negative (PFN) tumors accumulate ~17.8 non-synonymous mutations/tumor (P=2×10⁻⁴) and concentrate hits in an FGFR4/[[RAS]]/PIK3CA axis. Beyond the previously known RMS drivers (NRAS, KRAS, HRAS, FGFR4, PIK3CA, CTNNB1, FGFR4), the authors nominate novel recurrent mutations in FBXW7 and BCOR. Receptor tyrosine kinase/RAS/PIK3CA-axis alterations affect 93% (41/44) of WGS tumors, framing this axis as the dominant therapeutic target in RMS.

Cohort & data

  • 147 tumor/normal pairs total assembled from the Pediatric Oncology Branch (NCI), Children’s Oncology Group, The Tumour Bank at The Children’s Hospital at Westmead (NSW, Australia), and Hospital Sant Joan de Deu de Barcelona — see PMID:24436047.
  • Cancer types: Rhabdomyosarcoma (RMS), partitioned into Alveolar RMS (ARMS) and Embryonal RMS (ERMS); analysis re-classifies tumors by PAX-fusion status (PFP vs PFN).
  • Dataset: rms_nih_2014. All sequence data uploaded to dbGaP.
  • Assays / methods:
    • 44 tumors → whole-genome sequencing (Complete Genomics platform, mean 105× depth, 97% genome coverage).
    • 103 additional tumors → whole-exome sequencing (SOLiD 4 with Agilent SureSelect 37.8 Mb bait on 90 paired; Illumina HiSeq with SureSelect v2 on 13 paired) processed with MuTect / MutSig for the Illumina arm and GATK UnifiedGenotyper for the SOLiD arm.
    • 80 tumors → polyA RNA-seq on Illumina HiSeq2000 (100 bp paired-end), mapped with TopHat2; fusions via tophat-fusion and deFuse.
    • All tumors → Illumina Omni 2.5M or 5M SNP arrays (analogous to affymetrix-snp6 workflow).
    • Variant annotation via ANNOVAR and Oncotator; reference build hg19.
  • Verification: SOLiD exome on 30/44 WGS samples, custom AmpliSeq Cancer Panel on Ion Torrent PGM (sensitivity 84%, specificity 93%), and Sanger sequencing for figure/table calls.

Key findings

  • Two RMS genotypes by PAX-fusion status. PFP tumors (n=50: 35 PAX3-FOXO1 + 15 PAX7-FOXO1) and PFN tumors define the dominant axis of molecular classification; the genotype outperforms histology (ARMS vs ERMS) for capturing biology — 7 fusion-negative ARMS clustered with ERMS by expression and mutation profile.
  • Cryptic PAX rearrangements. Three “fusion-negative” ARMS tumors harbored alternative PAX rearrangements: RMS235 and RMS2031 carried PAX3-NCOA1 fusions, and RMS2046 carried a novel PAX3-INO80D fusion (in-frame on RNA-seq) in a region of massive 2q rearrangement.
  • Mutation burden differs by genotype. Mean non-synonymous somatic mutations per tumor: PFN 17.8 vs PFP 6.4 (P=2×10⁻⁴). PFN tumors also showed greater aneuploidy (P=1×10⁻⁵). One PFP tumor (RMS224, 3-month-old) had no protein-coding somatic alterations besides the PAX3-FOXO1 fusion and copy-neutral 11p LOH. Mutation count scales with age of diagnosis with a steeper slope in PFN tumors.
  • Structural variation. WGS identified 553 somatic SVs affecting 419 genes; 48 recurrently affected. Beyond PAX3/7 fusions, recurrently SV-affected genes included MIR17HG, CNR1, CDKN2A, tyrosine kinase signaling genes (ERBB4, RPTOR, FRS2, CACNA1A), and muscle-development genes (NRG1, FOXP2). Of 296 between-gene DNA junctions, 19% produced fusion transcripts; PAX fusions accounted for one-third of fusion transcripts and no additional recurrent fusions were detected.
  • Recurrent point mutations. 542 somatic mutations altered 495 genes in the WGS discovery set (40 recurrent). The significant gene list (Table 1) included previously known RMS drivers HRAS, KRAS, NRAS, FGFR4, PIK3CA, NF1, plus novel recurrent hits in FBXW7 and BCOR.
  • RAS/PI3K/RTK axis activation in PFN tumors. PFN frequencies: NRAS 11.7%, KRAS 6.4%, HRAS 4.3% (all at codons 12/13/61, no RAS mutations in PFP). NF1 PFN 5.3% mutated + 9% with 17q11.2 LOH. One BRAF V600E (RMSS013). PIK3CA PFN 7.4% at oncogenic codons Q546 (helical) or H1047 (kinase); two tumors (RMS2028, RMS217) had concurrent PIK3CA + RAS mutations. One PFP ARMS tumor (RMS244) also carried PIK3CA. PIK3CD damaging mutation in RMS2107; homozygous PTEN deletion in RMS2117. RTK mutations: FGFR4 (PFP 0%, PFN 9.6%), PDGFRA 1.4%, ERBB2 1.4%.
  • Cell-cycle / chromatin drivers. FBXW7 mutated in 7.4% of PFN tumors, all at conserved WD40 arginine residues (R387P, R441G, R367P); no PFP mutations. CTNNB1 altered in 3% of PFN at oncogenic S33 (n=1) and T41 (n=2). TP53 mutated exclusively in PFN (5.3%) with 12% of all tumors carrying 17p13.1 LOH; one germline pathogenic TP53 R248 (patient RMS212). Low-frequency mitotic checkpoint hits: BUB1B 1.4%, FOXM1 1.4%, CCND1 1%, CCND2 1%. BCOR (Xp11.4) altered in 7% of all RMS (9 PFN: 7 mutations + 2 focal homozygous deletions; 1 PFP indel).
  • Expressed mutations enrich for drivers. RNA-seq of 80 tumors showed 58% of DNA-level somatic mutations had RNA evidence (vs ~36% reported in adult breast/lymphoma). Median expressed mutations per tumor: PFN 9 vs PFP 2.5. 33 genes recurrently harbored expressed mutations including PTPN11 (0 PFP vs 2 PFN), ATM (1 PFP vs 1 PFN), ZNF350 (BRCA1-interacting; 1 PFP vs 1 PFN), TRPC4AP (MYCN-interacting; 0 PFP vs 2 PFN). Singleton expressed mutations were discovered in FOXO1 and ARID1A — not previously reported in RMS.
  • Copy-number alterations.
    • 11p15.5 LOH (encompassing imprinted IGF2) in 50% of tumors (16 PFP / 59 PFN).
    • IGF1R focal amplification in 2.7% (1 PFP / 3 PFN); one IGF2 3′UTR somatic indel (RMS2037).
    • 12q13-q14 amplification (containing CDK4, 25 genes total) in 9.7%, predominantly PFP (12 PFP / 1 PFN).
    • 12q15 amplification (containing FRS2 and MDM2) in 9%, predominantly PFN (9 PFN / 1 PFP).
    • 2p24 MYCN amplification in 5%, predominantly PFP (8 PFP / 1 PFN).
    • PAX7-FOXO1 fusion gene amplification in 12/15 PAX7-FOXO1 tumors.
    • 13q31-32 MIR17HG amplification exclusively in PFP (4.5%).
    • Homozygous CDKN2A deletion in 3%; 9p21.3 LOH in 9% (1 PFP / 13 PFN) — lower than the historically reported 25%.
    • Recurrent chromosome 8 gain in 46% of PFN; recurrent gains of chromosomes 2, 7, 11, 13; recurrent losses of 1p, 9, 16.
  • Pathway integration. Reactome over-representation on 2,119 somatically altered genes flagged FGFR signaling as the most significantly altered pathway (P=4.6×10⁻⁵, 29/112 candidate genes); 88% (22/25) of WGS PFN samples had pathway hits. No canonical pathway was significantly enriched when PFP-only altered genes (435) were analyzed separately.
  • Convergent genetic axis. GSEA showed that 444 genes >4-fold differentially expressed in a PAX3-FOXO1-expressing fibroblast model (cell line 7250_PF) — and confirmed in published PAX3-FOXO1 ChIP-seq targets and a transgenic mouse forelimb/somite model — were enriched in the 2,119 somatically altered genes from this cohort (P=3×10⁻³), with stronger enrichment in PFN (P=7×10⁻³) than PFP (P=0.08) mutated gene sets. 116 leading-edge genes including FGFR4 were shared across all three PAX-fusion model systems.
  • Combined axis frequency. Alteration of the receptor tyrosine kinase / RAS / PIK3CA axis was found in 93% (41/44) of WGS-surveyed tumors.

Genes & alterations

  • PAX3 / PAX7 / FOXO1 — defining t(2;13) and t(1;13) translocations produce PAX3-FOXO1 (n=35) and PAX7-FOXO1 (n=15) fusions; PAX7-FOXO1 fusion gene amplified in 12/15 PAX7-FOXO1 tumors.
  • PAX3–NCOA1 — recurrent alternative fusion in RMS235 and RMS2031 (fusion-negative ARMS by RT-PCR; intra-chromosomal rearrangement).
  • PAX3–INO80D — novel fusion in RMS2046, in-frame on RNA-seq; partner is a subunit of the ATP-dependent chromatin-remodeling complex.
  • NRAS, KRAS, HRAS — codon 12/13/61 oncogenic hotspots; predominantly ERMS/PFN (PFN frequencies 11.7% / 6.4% / 4.3%); no RAS mutations in PFP.
  • NF1 — PFN 5.3% mutated + 9% 17q11.2 LOH; positions as RAS-axis tumor suppressor.
  • BRAF V600E — one PFN tumor (RMSS013).
  • PIK3CA — Q546 (helical) and H1047 (kinase) oncogenic codons; PFN 7.4%, plus one PFP ARMS (RMS244); two tumors had concurrent RAS + PIK3CA mutations.
  • PIK3CD — predicted damaging mutation in RMS2107.
  • PTEN — homozygous deletion in RMS2117.
  • FGFR4 — recurrently mutated (PFP 0%, PFN 9.6%); shared leading-edge PAX-FOXO1 downstream target; pathway flagged by Reactome.
  • PDGFRA — recurrent mutation, 1.4%.
  • ERBB2 — recurrent mutation, 1.4%; ERBB4 recurrently SV-affected.
  • FBXW7 — novel recurrent driver in 7.4% of PFN tumors, exclusively at WD40-domain arginines R387P / R441G / R367P.
  • BCOR — novel recurrent driver in 7% of all RMS cases (Xp11.4): 7 PFN missense + 2 PFN focal homozygous deletions + 1 PFP indel.
  • CTNNB1 — oncogenic S33/T41 mutations in 3% of PFN, including one fusion-negative ARMS.
  • TP53 — PFN-exclusive somatic mutation (5.3%) + 12% LOH at 17p13.1; germline pathogenic R248 in RMS212.
  • BUB1B, FOXM1, CCND1, CCND2 — low-frequency (1–1.4%) mitotic-checkpoint / cell-cycle hits.
  • PTPN11 — 2 expressed mutations, both PFN.
  • ATM — 2 expressed mutations (1 PFP, 1 PFN).
  • ZNF350 (BRCA1-interacting), TRPC4AP (MYCN-interacting) — recurrently expressed mutations (2 each).
  • ARID1A, FOXO1 — singleton expressed mutations, not previously reported in RMS.
  • IGF1R — focal amplification in 2.7% of cases.
  • IGF2 — paternally imprinted gene within 11p15.5 LOH region (50% of tumors); one 3′UTR somatic indel.
  • CDK4 — within 12q13-q14 amplicon (9.7%, PFP-skewed).
  • MDM2, FRS2 — within 12q15 amplicon (9%, PFN-skewed).
  • MYCN — 2p24 amplification in 5%, PFP-skewed.
  • CDKN2A — homozygous deletion 3%; 9p21.3 LOH in 9% (PFN-skewed).
  • MIR17HG — 13q31-32 amplification exclusively in PFP (4.5%) and recurrently SV-affected.
  • MYOD1, MET, CNR1 — altered PFN genes that are known downstream targets of PAX3–FOXO1 (P=1.54×10⁻³).
  • RPTOR, CACNA1A, NRG1, FOXP2 — additional recurrently SV-affected genes (tyrosine kinase signaling and muscle development).

Clinical implications

  • Re-classification of RMS. Authors argue PAX-fusion status (PFP vs PFN) is a more accurate biological and prognostic stratification than ARMS/ERMS histology — consistent with prior clinical observations that fusion-negative ARMS behaves like ERMS. The discovery of cryptic PAX rearrangements in histologically fusion-negative ARMS (PAX3–NCOA1, PAX3–INO80D) has direct implications for therapeutic stratification: such patients may belong with PFP rather than PFN.
  • Targetable axis in 93% of tumors. FGFR4, IGF1R, PDGFRA, ERBB2/ERBB4, MET, MDM2, CDK4, and PIK3CA all have approved or late-stage inhibitors and are altered in this cohort, providing a “framework for genomics-directed therapies” that could be tested in clinical trials.
  • MEK inhibition as a RAS-pathway strategy. Authors highlight that the RAS pathway (counting FGFR4, RAS, NF1, PIK3CA) is mutationally activated in ≥45% of PFN tumors and cite the success of trametinib in NRAS-mutant melanoma plus preclinical RMS evidence as rationale for MEK1/2 inhibition in RMS.
  • Novel candidate targets. BCOR (chromatin repressor interacting with class I/II HDACs; previously implicated in AML, retinoblastoma, medulloblastoma) and FBXW7, ARID1A, ZNF350, TRPC4AP are flagged as targets requiring functional validation.
  • Prognostic CNAs. 12q13-q14 amplification (containing CDK4) is reaffirmed as a feature with worse overall survival in RMS independent of fusion status (citing prior literature).
  • Higher expressed-mutation rate in pediatric tumors. 58% of DNA-level somatic mutations were RNA-expressed (vs ~36% in adult breast/lymphoma cohorts), arguing that pediatric RMS may have a higher driver-to-passenger ratio and offer tractable therapeutic targets.

Limitations & open questions

  • Low overall mutation rate is consistent with other pediatric solid tumors (~0.1 protein-coding changes/Mb in PFP) but makes statistical detection of recurrently altered genes underpowered for low-frequency drivers.
  • PFP biology is dominated by transcriptional reprogramming from the PAX fusion, but PAX3-FOXO1 alone is insufficient for transformation in model systems — the cooperating event in most PFP tumors appears to be CNA (MYCN, CDK4, MIR17HG amplification; CDKN2A deletion; 11p15.5 LOH) rather than a point-mutation driver. Identifying the rare additional somatic drivers in PFP remains an open question.
  • Mechanism behind the age-vs-mutation slope difference between PFP and PFN (steeper in PFN) is unresolved: cell of origin, proliferation/apoptotic rate, or DNA-repair deficit are all proposed.
  • Functional validation outstanding for the novel FBXW7, BCOR, ARID1A, ZNF350, TRPC4AP mutations as drivers and therapeutic targets.
  • No directly actionable RAS inhibitor — authors lean on MEK1/2 inhibition (e.g. trametinib) as an effector-pathway strategy pending direct RAS-targeted agents.
  • CDKN2A LOH frequency (9%) is lower than the previously reported 25% — methodological differences (SNP-array resolution, cohort composition) are not fully reconciled.

Citations from this paper used in the wiki

  • Abstract & landscape framing: “we performed whole-genome, whole-exome and transcriptome sequencing to characterize the landscape of somatic alterations in 147 tumor/normal pairs… alteration of the receptor tyrosine kinase/RAS/PIK3CA axis affects 93% of cases” (Abstract).
  • Discovery cohort sizing & coverage: “A set of 44 RMS tumors with matched normal leukocyte DNA was sequenced with whole-genome paired-end sequencing (WGS)… mean depth of 105X… 97% of the genome… 103 additional tumors and their matched germlines… Eighty of the tumors were analyzed by whole-transcriptome sequencing” (Results, p.3).
  • Alternative PAX fusions: “Cases RMS235 and RMS2031 harbored a PAX3-NCOA1 fusion… a novel PAX fusion in a region of massive rearrangement of chromosome 2q in RMS2046… fusion of the N-terminus of PAX3 (first seven exons) and the C-terminus of INO80D” (Results, p.3).
  • Mutation burden contrast: “PFN tumors had significantly more verified somatic non-synonymous mutations per tumor than PFP tumors (17.8 and 6.4 respectively) (P=2×10⁻⁴)… PFN samples had an overall increase in aneuploidy (P=1×10⁻⁵)” (Results, p.4).
  • RAS/PI3K/RTK frequencies: “NRAS (PFN frequency, 11.7%), KRAS (PFN 6.4%), HRAS (PFN 4.3%)… No RAS mutations were found in PFP tumors… NF1 (PFN 5.3% mutated, 17q11.2 LOH 9%)… PIK3CA mutations (PFN 7.4%) occurred at the known oncogenic codons Q546 or H1047… FGFR4 (PFP 0%, PFN 9.6%), PDGFRA (1.4%), and ERBB2 (1.4%)” (Results, p.4–5).
  • FBXW7 and BCOR: “FBXW7… mutated in 7.4% of PFN tumors. All mutations… at conserved arginine residues (R387P, R441G, R367P) within the WD40 repeat regions… BCOR, located on chromosome Xp11.4, in 7% of all RMS cases” (Results, p.5).
  • CNA frequencies: “LOH of 11p15.5 was found in 50%… 12q13-q14… 9.7% of the tumors… The 12q13-q14 amplicon was found predominantly in PFP tumors (12 PFP vs. 1 PFN)… Recurrent focal amplification of 12q15 (9%)… FRS2 and MDM2 occurred predominantly in PFN tumors (9 PFN vs. 1 PFP)… Amplification of 2p24 involving MYCN (5%) occurred predominantly in PFP tumors” (Results, p.6).
  • Pathway analysis: “Reactome over-representation analysis indicated that FGFR signaling was the most significantly altered pathway (P= 4.6×10⁻⁵) with 29/112 candidate genes represented… mutations in this pathway were found in 88% of PFN samples (22/25 tumors)” (Results, p.6).
  • Convergent axis statement: “Evidence for alteration of this common genetic axis can be found in 93% (41/44) of the tumors surveyed by WGS” (Discussion, p.9).
  • Trametinib rationale: “the recent success of the MEK1/2 inhibitor, Trametinib, in melanomas with mutated NRAS demonstrates the utility of inhibiting the effector pathways altered by the mutation… Early preclinical evidence has found efficacy of this method in RMS” (Discussion, p.8).

This page was processed by crosslinker on 2026-05-09.