PRC2 is recurrently inactivated through EED or SUZ12 loss in malignant peripheral nerve sheath tumors

Authors

William Lee

Sewit Teckie

Thomas Wiesner

Leili Ran

Carlos N. Prieto Granada

Mingyan Lin

Sinan Zhu

Zhen Cao

Yupu Liang

Andrea Sboner

William D. Tap

Jonathan A. Fletcher

Kety H. Huberman

Li-Xuan Qin

Agnes Viale

Samuel Singer

Deyou Zheng

Michael F. Berger

Yu Chen

Cristina R. Antonescu

Ping Chi

Doi

10.1038/ng.3095

PMID: 25240281 · DOI: 10.1038/ng.3095 · Journal: Nature Genetics (2014)

TL;DR

Lee et al. characterized the genomic landscape of malignant peripheral nerve sheath tumors (MPNSTs), an aggressive soft-tissue sarcoma, using whole-exome sequencing, SNP6.0 copy number, RNA-seq, and targeted IMPACT sequencing on a discovery cohort of 15 MPNSTs (12 patients) plus a validation cohort of 37 MPNSTs and 7 neurofibromas from 32 patients. They discovered that the Polycomb repressive complex 2 (PRC2) core components EED or SUZ12 are inactivated by loss-of-function alterations in 92% of sporadic, 70% of NF1-associated, and 90% of radiotherapy-associated MPNSTs, and that PRC2 loss co-occurs significantly with NF1 and CDKN2A alterations. PRC2-deficient tumors lose H3K27me3 and aberrantly re-express developmentally suppressed homeobox master regulators; restoring SUZ12 in a PRC2-null cell line rescued H3K27me3 and suppressed proliferation, implicating PRC2 as a tumor suppressor critical to MPNST pathogenesis PMID:25240281.

Cohort & data

  • Discovery cohort: 15 fresh-frozen tumor/normal-paired MPNSTs from 12 patients (6 NF1-associated, 4 sporadic, 4 radiotherapy-associated, 1 epithelioid) profiled by whole-exome-seq, affymetrix-snp6 SNP6.0 arrays, and rna-seq (51 bp paired-end, Illumina HiSeq-2500, hg19, STAR v2.3).
  • Validation cohort: 37 FFPE MPNSTs and 7 neurofibromas from 32 NF1 patients profiled by msk-impact-panel targeted hybrid-capture sequencing (panel includes NF1, SUZ12, EED, CDKN2A, TP53).
  • Functional validation: Two human MPNST cell lines, ST88-14 (NF1-associated, SUZ12-deficient) and MPNST724 (PRC2-wt), with Flag-HA-tagged SUZ12 or EED reintroduction.
  • Cancer type: MPNST (malignant peripheral nerve sheath tumor).
  • Dataset: mpnst_mskcc — MSKCC MPNST cohort, hg19.

Key findings

  • PRC2 inactivation is highly recurrent in MPNST. Loss-of-function alterations in EED or SUZ12 occur in 80% (12/15) of discovery-cohort MPNSTs and, combining discovery + validation, in 92% (12/13) of sporadic, 70% (19/27) of NF1-associated, and 90% (8/9) of radiotherapy-associated MPNSTs PMID:25240281.
  • EED and SUZ12 alterations are mutually exclusive. Discovery cohort: 5 EED mutations (4 frame-shift, 1 splice-site) all with LOH (3 by heterozygous deletion, 2 by copy-neutral LOH); 2 SUZ12 homozygous deletions and 5 heterozygous deletions, three of which had local genomic rearrangements producing structural transcript variants at exons 4, 6, or 10 (only detectable by RNA-seq, not by standard WES).
  • No alterations in other PRC2 core members. EZH1 and EZH2 were not altered in any MPNST in this cohort.
  • NF1 and CDKN2A co-altered with PRC2. NF1 nonsense mutations or homozygous deletion in 87.5% (7/8) of sporadic/radiotherapy-associated discovery samples and 82% (18/22) overall in sporadic/radiotherapy-associated MPNSTs. CDKN2A homozygous deletion in 73% (11/15) and heterozygous loss in 13% (2/15) of discovery samples; 81% (42/52) overall across cohorts. TP53 non-synonymous mutations in 13% (2/15) and heterozygous loss in 20% (3/15) of discovery samples; 42% (22/52) overall.
  • Three-way significant co-occurrence. NF1, CDKN2A, and PRC2 (EED or SUZ12) alterations significantly co-occur by Fleiss’ kappa (κ=0.21, p=0.001).
  • PRC2 loss produces a distinct transcriptome. PCA of 16 MPNSTs separated PRC2-loss from PRC2-wt samples on PC1. Of 479 differentially expressed genes (FDR<0.05, fold-change >8), 455/479 (95%) were upregulated in PRC2-loss tumors. GO analysis identified homeobox transcription factors and developmental/morphogenesis genes as enriched; GSEA showed strong enrichment of the “PRC2 module” and H3K27me3 target gene sets from neural precursor cells and brain tissue.
  • H3K27me3 IHC tracks PRC2 functional status. PRC2-wt MPNSTs show robust nuclear H3K27me3; PRC2-loss MPNSTs show complete tumor-cell H3K27me3 loss with preserved stromal staining. IHC concordance was high for homozygous PRC2 loss but heterozygous loss alone was not predictive — RNA-seq–detected structural variants resolved the discrepancy.
  • H3K27me3 loss marks malignant progression from neurofibroma. All 7 neurofibromas (PRC2-wt except one SUZ12 het loss) retained H3K27me3. At MPNST/neurofibroma interfaces, H3K27me3 transitions from robust (neurofibroma) to absent (MPNST). 56% (19/34) of NF1-associated MPNSTs lost H3K27me3, vs. >90% of sporadic and radiotherapy-associated MPNSTs.
  • NF1 mutation precedes EED loss in a sporadic MPNST. In sample 16T, subclonal analysis showed the NF1 D1237_splice mutation in 84% of cells and the EED E249fs in 57%, suggesting NF1 loss preceded PRC2 loss in this tumor.
  • SUZ12 re-expression rescues H3K27me3 and suppresses growth. Introducing FH-SUZ12 (but not FH-EED) into SUZ12-deficient ST88-14 cells restored H3K27me3 and significantly decreased proliferation; in PRC2-wt MPNST724, neither construct affected growth. SUZ12 re-expression localized to promoters of FOXN4, IGF2, PAX2, TLX1, with reciprocal gain of H3K27me3/EZH2 and loss of H3K4me3/H3K27ac activating marks at these promoters and reduced transcript levels.

Genes & alterations

  • EED — frame-shift and splice-site mutations with LOH (heterozygous deletion or copy-neutral LOH); confirmed loss-of-function tumor-suppressor role in MPNST.
  • SUZ12 — homozygous and heterozygous deletions plus local genomic rearrangements producing aberrant transcripts; mutually exclusive with EED loss; SUZ12 re-expression rescues H3K27me3 and inhibits growth.
  • NF1 — nonsense mutations and homozygous deletion in 82% of sporadic/radiotherapy-associated MPNSTs; significantly co-occurs with PRC2 and CDKN2A loss.
  • CDKN2A — homozygous deletion (most common) and heterozygous loss in 81% of MPNSTs overall.
  • TP53 — non-synonymous mutation and heterozygous loss in 42% of MPNSTs overall.
  • EZH1 / EZH2 — no alterations detected; explicit negative finding for these PRC2 catalytic subunits.
  • IGF2, FOXN4, PAX2, TLX1 — PRC2-repressed developmental master regulators and imprinted genes aberrantly upregulated in PRC2-loss MPNSTs; promoter H3K27me3 restored after SUZ12 reintroduction.

Clinical implications

  • Diagnostic biomarker. Given that >90% of sporadic and radiotherapy-associated MPNSTs and ~56% of NF1-associated MPNSTs lose H3K27me3, the authors propose H3K27me3 IHC (and PRC2 mutational status) as a biomarker to aid the diagnostically challenging distinction of MPNST.
  • Distinguishes MPNST from neurofibroma. H3K27me3 IHC is retained in benign neurofibromas and lost upon malignant progression, providing a candidate marker for malignant transformation.
  • Epithelioid MPNST is distinct. None of the three epithelioid MPNSTs harbored PRC2, NF1, or CDKN2A alterations, suggesting epithelioid MPNST is a molecularly separate entity that should not be grouped with conventional MPNST for therapeutic decisions.
  • Therapeutic implication (mechanistic). Because PRC2 acts as a tumor suppressor in MPNST (in contrast to EZH2’s oncogenic gain-of-function role in lymphoma), broad PRC2 inhibitors are unlikely to benefit MPNST; conversely, downstream PRC2-repressed pathways may be targetable.

Limitations & open questions

  • Standard WES alone misses local genomic rearrangements at the SUZ12 locus that produce aberrant transcripts — RNA-seq was required to detect three of five Het-loss SUZ12 cases. DNA sequencing alone (exome or IMPACT) cannot predict PRC2 functional status in heterozygous-loss cases; H3K27me3 IHC is more accurate.
  • The temporal order of NF1, CDKN2A, and PRC2 inactivation is inferred from one sporadic case (NF1 before EED) and remains largely correlative; the authors call for sequential inactivation studies in cell-line and mouse models.
  • Why >90% of sporadic and radiotherapy-associated MPNSTs lose H3K27me3 but only ~56% of NF1-associated MPNSTs do remains unexplained.
  • Discovery cohort is small (n=15 tumors from 12 patients); the validation IMPACT cohort (37 MPNSTs) extends but does not replace WES.
  • Functional rescue was demonstrated in a single SUZ12-deficient cell line (ST88-14); whether EED-deficient MPNSTs behave equivalently and whether re-expressing PRC2 in vivo would suppress tumor growth remains untested.

Citations from this paper used in the wiki

  • “we identified loss-of-function (LOF) somatic alterations of the Polycomb repressive complex 2 (PRC2) core components, EED or SUZ12, in 92% of sporadic, 70% of NF1-associated and 90% of radiotherapy-associated MPNSTs” (Abstract).
  • “EED and SUZ12 genetic alterations are mutually exclusive and are collectively found in 80% (12/15) of all MPNSTs” (Results).
  • “There is a significant co-occurrence of NF1, CDKN2A and PRC2 genetic alterations (Fleiss’ Kappa statistics, Kappa=0.21, p=0.001)” (Results).
  • “We observed Hom deletion and Het loss of the CDKN2A locus in 73% (11/15) and 13% (2/15) of MPNSTs, respectively. We also observed non-synonymous mutations and Het loss in TP53 in 13% (2/15) and 20% (3/15) of MPNSTs” (Results).
  • “Genetic alterations in CDKN2A and TP53 were found in 81% (42/52) and 42% (22/52) of all MPNSTs respectively” (Results).
  • “Only the FH-SUZ12, but not FH-EED, restored the H3K27me3 level in ST88-14 cells and significantly decreased cell growth” (Results).
  • “The data was further analyzed and visualized using the cBio Portal” (Methods, DNA sequencing and analysis).

This page was processed by crosslinker on 2026-05-12.