Comprehensive molecular characterization of pheochromocytoma and paraganglioma

Authors

Lauren Fishbein

Ignaty Leshchiner

Vonn Walter

Ludmila Danilova

A. Gordon Robertson

Amy R. Johnson

Tara M. Lichtenberg

Bradley A. Murray

Hans K. Ghayee

Tobias Else

Shiyun Ling

Stuart R. Jefferys

Aguirre A. de Cubas

Brandon Wenz

Esther Korpershoek

Antonio L. Amelio

Liza Makowski

W. Kimryn Rathmell

Anne-Paule Gimenez-Roqueplo

Thomas J. Giordano

Sylvia L. Asa

Arthur S. Tischler

The Cancer Genome Atlas Research Network

Karel Pacak

Katherine L. Nathanson

Matthew D. Wilkerson

Doi

10.1016/j.ccell.2017.01.001

PMID: 28162975 · DOI: 10.1016/j.ccell.2017.01.001 · Journal: Cancer Cell (2017)

TL;DR

The TCGA Pheochromocytoma and Paraganglioma (PCPG) Analysis Working Group performed multi-platform molecular profiling (whole-exome sequencing, RNA-seq, miRNA-seq, DNA methylation arrays, SNP arrays, and RPPA) on 173 PCC/PGL tumors. Despite a low somatic mutation rate (~0.67/Mb), the tumors are driven by diverse alterations: pathogenic germline mutations in eight susceptibility genes (most commonly SDHB, RET, VHL, NF1), recurrent somatic mutations in five driver genes including the newly identified CSDE1, and recurrent fusion genes involving MAML3, BRAF, NGFR, and NF1. Integrated mRNA clustering yielded four molecular subtypes — kinase signaling, pseudohypoxia, Wnt-altered, and cortical admixture — and the Wnt-altered subtype (defined by MAML3 fusions and CSDE1 mutations) was associated with poor clinical outcome and metastatic disease.

Cohort & data

Key findings

  • A driver mutation, fusion gene, or copy number alteration was identified in 95% of PCC/PGLs, providing molecular etiology for the vast majority of the cohort (PMID:28162975).
  • Pathogenic germline mutations were detected in eight previously known susceptibility genes in 46/173 patients (27%). Highest germline rates: SDHB (9%), RET (6%), VHL (4%), NF1 (3%); rarer (≤2% each): SDHD, MAX, EGLN1 (PHD2), TMEM127.
  • Mean somatic sequence mutation rate was 0.67 mutations/Mb, low relative to other TCGA cancers.
  • Five MutSig2 significant somatic driver genes (q < 0.05): HRAS, NF1, EPAS1, RET, and the newly identified CSDE1. Additional cancer-relevant somatic events: BRAF p.G469A, IDH1 p.R132C, FGFR1, VHL, ATRX, TP53, SETD2, ARNT.
  • Among the 21 germline-or-somatic-driver genes, mutations were strikingly mutually exclusive (p < 1e-4); only four tumors had two events in this set (three SDHB-germline + ATRX-somatic, one with both germline RET p.V804M and somatic RET p.M918T).
  • RET mutations partitioned by location: germline mutations clustered at extracellular codon C634 vs. somatic mutations at intracellular tyrosine kinase codon M918 (p < 0.001); RET-mutant tumors (germline or somatic) overexpressed RET vs. wild-type (p < 0.003).
  • Recurrent fusion genes were discovered in PCC/PGL: a UBTF–MAML3 fusion (in 7 of 10 tumors with co-amplification of 4q31.1 and 17q21.31; p < 2e-9) and a TCF4–MAML3 fusion in one additional tumor; in total 10 tumors were MAML3 fusion-positive. Other fusions: KIAA1737–NGFR (NGFR 3.0-fold overexpressed), RUNDC1–BRAF (BRAF 5.2-fold overexpressed), and an underexpressed NF1–RAB11FIP4 (NF1 9.9-fold underexpressed). All but one of the fusion break points localized to 17q.
  • MAML3 fusion-positive tumors overexpressed MAML3 2.7-fold (p < 5e-6) and showed a unique, expansive hypomethylation signature (4,229 significant probes vs. 352 expected by chance; p < 0.002); hypomethylation was inversely correlated with mRNA expression of corresponding target genes (p < 4e-10). Wnt and Hedgehog pathway genes (e.g., WNT4, WNT11, WNT5A, NKD1, GLI2) were overexpressed; miR-375 (a negative regulator of Wnt-pathway member FZD8) was underexpressed; RPPA showed elevated β-catenin, DVL3, and GSK3 in fusion-positive tumors.
  • CSDE1 (formerly UNR; located at 1p13.2) was identified as a novel somatically-mutated driver: 4 tumors carried CSDE1 mutations (2 frameshift, 2 splice-site) clustered proximally. Splice-site mutants showed intron retention and exon skipping in mRNA-seq; CSDE1-mutant tumors showed marked genomic deletion and mRNA underexpression, supporting a loss-of-function role. Expression profiles of CSDE1-mutant PCC/PGLs significantly correlated with published Csde1-knockout mouse ESC microarrays.
  • Four mRNA expression subtypes (SigClust p < 0.001), validated in an independent Burnichon et al. cohort:
    • Kinase signaling (predominantly PCCs): highest PNMT expression, enriched for somatic/germline mutations in NF1, RET, TMEM127, HRAS, and rare BRAF/NGFR/NF1 fusions; enriched in RPPA cluster 3 (RAS-MAPK signaling); 95% of NF1-mutated tumors had 17q11.2 focal deletions, 86% in this subtype.
    • Pseudohypoxia (mixed PCC 57% / PGL 43%): germline SDHB, SDHD, VHL and somatic VHL/EPAS1 mutations were completely specific to this subtype. Most genome-doubled tumors (74%) and the hypermethylated and intermediate-methylated DNA-methylation clusters; miRNA cluster 3 (with mir-210 hypoxia marker overexpression) tightly associated.
    • Wnt-altered (sporadic adrenal PCCs): contained ALL MAML3 fusion-positive tumors (p < 4e-9) and 3 of 4 CSDE1 mutants (p < 0.01); highest CHGA overexpression (p < 0.002); no germline susceptibility-gene mutations.
    • Cortical admixture: overexpressed adrenal cortex markers (CYP11B1, CYP21A2, STAR); reduced tumor purity, elevated leukocyte infiltration; significant association with histologically interspersed cortical cells (p < 5e-5); both germline MAX mutations occurred here (p < 0.032).
  • Three SDHB-germline tumors carried concurrent ATRX somatic mutations (previously reported); ATRX mutations have been linked to alternative lengthening of telomeres.
  • Pathway analysis grouped drivers into four signaling axes: kinase signaling (NF1, HRAS, RET, BRAF, FGFR, NGFR, plus subunits of cAMP-dependent PKA — including PRKAR1A and PRKACA reported in other adrenal pathologies); hypoxia signaling (VHL, ARNT/HIF1β, EPAS1/HIF2α, EGLN1/PHD2 — mutually exclusive); Krebs cycle / electron transport (SDHB, SDHD, IDH1); and Wnt signaling (MAML3, CSDE1).

Genes & alterations

  • HRAS — recurrent somatic Q61 hotspot mutations activating MAPK signaling; enriched in kinase signaling subtype.
  • RET — germline C634 (extracellular) vs somatic M918 (intracellular tyrosine kinase) mutations (p < 0.001); RET-mutant tumors overexpress RET (p < 0.003).
  • EPAS1 (HIF2α) — somatic hotspots A530, P531, Y532 driving HIF stabilization; specific to pseudohypoxia subtype.
  • NF1 — germline (3%) and somatic mutations (no positional bias); 17q11.2 focal deletion in 95% of NF1-mutated tumors; underexpressed NF1–RAB11FIP4 fusion identified.
  • CSDE1 — newly identified driver; 4 tumors with truncating frameshift / splice-site mutations clustered proximally; integrated analysis (deletion + underexpression + correlation with Csde1-KO ESC signature) supports tumor suppressor / loss-of-function role.
  • MAML3 — recurrent in-frame fusion gene, predominantly UBTF–MAML3 (also TCF4–MAML3); 10 fusion-positive tumors; 2.7-fold overexpression (p < 5e-6); drives DNA hypomethylation and Wnt/Hedgehog signaling rather than canonical NOTCH targets; defines the Wnt-altered subtype; clinical marker of metastasis.
  • BRAF — somatic p.G469A point mutation; recurrent RUNDC1–BRAF fusion (5.2-fold BRAF overexpression).
  • NGFR — KIAA1737–NGFR fusion (3.0-fold NGFR overexpression).
  • SDHB, SDHD — germline mutations specific to pseudohypoxia subtype; SDHB germline associated with hypermethylated subtype, ATRX co-mutation, and aggressive disease.
  • VHL — germline (4%) and somatic mutations; pseudohypoxia subtype specific; co-occurring 3p deletion when germline.
  • MAX, EGLN1, TMEM127 — rare germline events; MAX germline mutations occurred in cortical admixture subtype (p < 0.032).
  • ATRX — somatic mutations co-occurring with SDHB germline (3 tumors); marker of aggressive disease and metastasis.
  • IDH1 — single tumor with R132C mutation.
  • SETD2, TP53, ARNT, FGFR1 — somatic mutations identified in cancer-relevant gene scan; SETD2 somatic mutation was a poor-outcome marker.
  • UBTF, TCF4 — 5′ partners of MAML3 fusion genes; their promoters drive MAML3 overexpression.
  • CHGA — highest expression in Wnt-altered subtype (p < 0.002); chromogranin A is a clinical neuroendocrine tumor marker.
  • PNMT — highest expression in kinase signaling subtype; encodes the enzyme converting norepinephrine to epinephrine.
  • CYP11B1, CYP21A2, STAR — adrenal cortex markers overexpressed in cortical admixture subtype.

Clinical implications

  • The authors propose this comprehensive map as a foundation for PCC/PGL precision medicine. Aggressive-disease-free survival (ADFS) was significantly associated with nine molecular markers; metastatic-free survival (MFS) with seven of those nine.
  • Markers of poor ADFS / MFS: MAML3 fusion gene, SDHB germline mutation, somatic SETD2 or ATRX mutation, high somatic mutation total, Wnt-altered and pseudohypoxia mRNA subtypes, hypermethylated DNA methylation subtype.
  • Markers of better ADFS: plasma/urine metanephrine and epinephrine positivity, kinase signaling subtype, low-methylation subtype.
  • Ki-67 protein expression by immunohistochemistry (n = 62) positively correlated with metastatic disease; the highest-Ki-67 tumor was MAML3 fusion-positive.
  • Therapeutic suggestions raised by the authors (not tested in this paper):
    • SDH-mutant tumors accumulate glutamine; glutaminase inhibitors (e.g., NCT02071862) merit investigation.
    • Because MAML3 fusions activate Wnt signaling, downstream antagonists of β-catenin (e.g., PRI-274) and STAT3 (e.g., BB1608) merit investigation.
    • ATRX-loss tumors with alternative lengthening of telomeres may be sensitive to ATR inhibitors.
    • FDA-approved targeted therapies (unspecified by name in the paper) are noted as available for tumors carrying VHL, RET, BRAF, EPAS1, and FGFR1 mutations.

Limitations & open questions

  • Head and neck PGLs were excluded due to insufficient post-embolization tumor tissue, so the cohort is biased away from a clinically distinct PGL subgroup.
  • The metastatic event rate was modest (11/173 with distant metastases; 16 aggressive cases overall), limiting statistical power for survival analyses despite the multi-marker signal.
  • The authors cannot exclude that the upstream fusion partners (UBTF, TCF4) themselves contribute tumorigenic properties beyond simply driving MAML3 overexpression.
  • The DNA hypomethylation observed in MAML3 fusion-positive tumors is hypothesized to result from aberrant MAML3 binding to the genome, but direct binding evidence (e.g., ChIP-seq) is not provided.
  • CSDE1 had not previously been described as a driver in any cancer type; while integrated analysis is consistent with a tumor-suppressor role, functional validation in PCC/PGL models is left to future work.
  • Cortical admixture subtype’s relationship to multifocal MAX-mutant disease (vs. tissue-sampling artifact) remains a hypothesis.
  • Drug-target hypotheses (glutaminase inhibitors, β-catenin/STAT3 antagonists, ATR inhibitors) are not evaluated in this paper.

Citations from this paper used in the wiki

  • “We identified CSDE1 as a somatically-mutated driver gene, complementing four known drivers (HRAS, RET, EPAS1, NF1). We also discovered fusion genes in PCC/PGL, involving MAML3, BRAF, NGFR and NF1.” (Summary)
  • “Integrated analysis classified PCC/PGLs into four molecularly-defined groups: a kinase signaling subtype, a pseudohypoxia subtype, a Wnt-altered subtype, driven by MAML3 and CSDE1, and a cortical admixture subtype.” (Summary)
  • “Pathogenic germline mutations were detected within eight previously reported PCC/PGL susceptibility genes in 46 patients (27% of the cohort).” (Results — Germline and Somatic Mutations)
  • “PCC/PGLs exhibited a low somatic sequence mutation rate (mean 0.67 mutations per megabase) relative to other cancer types.” (Results — Germline and Somatic Mutations)
  • “Ten of 16 primary tumors with focal 4q31.1 amplification also had 17q21.31 focally amplified (p < 3e-8)… seven of these 10 co-amplified tumors (p < 2e-9) possessed in-frame RNA fusion transcripts spanning the 5′ portion of UBTF… and the 3′ portion of MAML3.” (Results — Fusion Gene Discovery)
  • “We identified a driver mutation, fusion gene, or copy number alteration in a majority of PCC/PGLs (95%).” (Discussion)
  • “Markers associated with poor ADFS included MAML3 fusion gene, SDHB germline mutation, somatic mutation in SETD2 or ATRX, high somatic mutation total, the Wnt-altered and pseudohypoxia expression subtypes, and the hypermethylated subtype.” (Results — Clinical outcome associations)
  • “FDA-approved targeted therapies are available for patients whose tumors carry VHL, RET, BRAF, EPAS1 and FGFR1 mutations.” (Discussion)

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