The Molecular Taxonomy of Primary Prostate Cancer

Author

The Cancer Genome Atlas Research Network

Doi

10.1016/j.cell.2015.10.025

PMID: 26544944 · DOI: 10.1016/j.cell.2015.10.025 · Journal: Cell (2015)

TL;DR

The TCGA Research Network performed integrated multi-platform molecular characterization of 333 primary prostate adenocarcinomas (cohort prad_tcga_pub, PRAD) using whole-exome and whole-genome sequencing, SNP arrays, DNA methylation arrays, mRNA/miRNA sequencing, and RPPA. 74% of tumors fell into one of seven mutually exclusive molecular subtypes defined by fusions of ERG (46%), ETV1 (8%), ETV4 (4%), or FLI1 (1%), or hotspot mutations in SPOP (11%), FOXA1 (3%), or IDH1 (1%). 25% of tumors carried potentially actionable PI3K/MAPK pathway lesions, and 19% had DNA repair gene defects relevant to PARP inhibitor therapy.

Cohort & data

  • N = 333 primary prostate adenocarcinomas from radical prostatectomy specimens; final cohort drawn from 425 cases after pathology and quality-control review (PMID:26544944).
  • Cancer type: PRAD (prostate adenocarcinoma); dataset: prad_tcga_pub.
  • Platforms: whole-exome-seq (333 tumor/normal pairs), whole-genome-seq (low-pass on 100 pairs; high-pass on 19), affymetrix-snp6 for copy number, hm450-methylation-array, rna-seq, mirna-seq (330 samples), and rppa (152 samples). 19 matched non-malignant adjacent prostate samples were profiled for methylation and RNA/miRNA expression.
  • Integrative clustering with icluster; significantly mutated genes via mutsig (MutSigCV); focal SCNAs via gistic (GISTIC 2.0); tumor purity estimated via absolute and other DNA/RNA methods.
  • Average follow-up under two years; outcome analysis was not possible.

Key findings

  • Seven mutually exclusive molecular subtypes capture 74% of primary prostate cancers (PMID:26544944):
  • TMPRSS2 is the dominant 5’ fusion partner; additional androgen-regulated partners include SLC45A3 and NDRG1. Tumors with elevated full-length ETS transcripts but no detectable fusion (12 ETV1-high, 6 ETV4, 2 FLI1) were identified, potentially driven by cryptic translocations or epigenetic mechanisms.
  • Overall mutational burden is 0.94 mutations/Mb (median; range 0.04–28), or ~19 non-synonymous mutations per tumor (median; IQR 13–25) — lower than most epithelial tumors without an exogenous mutagen.
  • 13 significantly mutated genes (q < 0.1): SPOP, TP53, FOXA1, PTEN, MED12, CDKN1B, plus newly identified BRAF, HRAS, AKT1, CTNNB1, ATM, ZMYM3, and NKX3-1. Subthreshold but biologically meaningful: KMT2C, KMT2D, APC, IDH1, PIK3CA.
  • Focal SCNAs by GISTIC 2.0: 20 amplifications and 35 deletions (q < 0.25). Homozygous PTEN deletion occurred in 15% — one of the highest rates of any TCGA tumor type. Other recurrent amplifications: MYC 8q24.21 (8%), FGFR1/WHSC1L1 8p11.23 (8%), CCND1 11q13.2 (2%). Heterozygous BRCA2 loss (13q13.1) almost always co-occurred with RB1 loss (13q14.2).
  • SCNA burden stratification identified three groups (quiet, intermediate, high-burden); the high-burden group had significantly higher Gleason scores, PSA, and tumor cellularity.
  • Four epigenetic clusters from methylation profiling. ERG fusion-positive tumors split between a low/moderate-methylation cluster (3) and a distinct hypermethylated cluster (1) with ~twice the number of hypermethylated loci.
  • IDH1 R132 mutations define a CIMP-like subgroup with even greater genome-wide hypermethylation than IDH1-mutant GBM or AML, fewer SCNAs, ETS-fusion-negative, SPOP wild-type, and associated with younger age at diagnosis (Figure 3B).
  • Epigenetic silencing of 164 genes identified by integrating methylation and mRNA expression. SHF, FAXDC2, GSTP1, ZNF154, KLF8 silenced in >85% of tumors; STAT6 silenced predominantly in ETS-fusion-positive tumors; HEXA silenced in 86.5% of SPOP-mutant vs 14.5% of ERG-fusion tumors (p<5.4·10⁻¹⁵).
  • Androgen receptor (AR) activity is variable across all subtypes. Tumors with SPOP or FOXA1 mutations had the highest AR transcriptional activity (p = 1.1·10⁻⁶ and 0.04 respectively, t-test).
  • AR splice variants (notably AR-V7) detected at low levels in hormone-naive primary tumors and occasionally in adjacent benign tissue — previously assumed restricted to metastatic castration-resistant disease.
  • SPOP/FOXA1 co-mutation observed in 4 tumors, both alterations clonal in all cases (i.e., same tumor cells).
  • DNA repair gene inactivation in 19% of tumors (Figure 5A) — including BRCA2 3% (germline + somatic), BRCA1 1 case, CDK12 1%, ATM 3 cases, FANCD2 lesions (6% focal heterozygous loss + truncating/homozygous deletions), RAD51C 3%.
  • PI3K/MAPK pathway lesions in ~25%: PTEN deletion/mutation 17%, PIK3CA hotspot mutations (6 tumors), PIK3CB mutations in 2 PTEN-deleted tumors, AKT1 E17K and D323Y mutations (3 tumors total), HRAS Q61R hotspot (3/4 mutants), and 8 BRAF mutations (none V600E; includes K601E, G469A, L597R, and a K601 in-frame 3-aa deletion).
  • Comparison with metastatic castration-resistant prostate cancer (Robinson et al. 2015 cohort, n=150): subtype distribution similar except no IDH1 mutants in mets; metastatic tumors had higher SCNA burden, higher mutational burden, more frequent AR amplification/mutation (essentially absent in primary), more frequent DNA repair, TP53, RB1, KMT2C, KMT2D alterations; SPOP mutations slightly less frequent in mets (8% vs 11%).

Genes & alterations

  • ERG — Most common ETS fusion (46%); TMPRSS2-ERG dominant. ERG-fusion-positive tumors split between low-methylation and hypermethylated epigenetic subclusters. Co-occurs preferentially with PTEN deletion.
  • ETV1, ETV4, FLI1 — Define less common ETS-fusion subtypes (8%, 4%, 1%). Cryptic ETV1 rearrangement to chr14 near MIPOL1/FOXA1 identified in one high-expressing case lacking a canonical fusion. Three tumors had fusions involving more than one ETS gene, likely reflecting convergent clonal evolution.
  • TMPRSS2 — Dominant 5’ androgen-regulated fusion partner.
  • SLC45A3, NDRG1 — Alternative androgen-regulated 5’ fusion partners.
  • SPOP — Hotspot mutations in 11%; mutually exclusive with ETS fusions; defines a subtype with CHD1 deletion, elevated methylation, homogeneous expression, SPINK1 overexpression, and highest AR transcriptional output.
  • FOXA1 — Mutated in 3–4%; missense mutations cluster in the winged-helix DNA-binding domain but not at DNA-binding residues; mutually exclusive with all other subtype-defining alterations except SPOP (4 co-mutated tumors, both clonal). FOXA1 mutants also have elevated AR output.
  • IDH1 — R132 hotspot mutations in ~1% define a CIMP-like, ETS-negative, SPOP-wt, SCNA-quiet subtype with younger age at diagnosis. Genome-wide hypermethylation exceeds that of IDH1-mut GBM and AML.
  • TP53 — Recurrently mutated; enriched in the 26% “unclassified” subset with high SCNA burden.
  • PTEN — Homozygous deletion in 15% (highest rates among TCGA tumor types); total alteration 17%; preferentially co-occurs with ERG fusion.
  • CHD1 — Co-deletion with SPOP-mutant subtype (5q15-q21). MAP3K7 (6q12-22) and CHD1 co-deletion associated with aggressive ETS-negative disease.
  • MED12, CDKN1B, NKX3-1, ZMYM3 — Significantly mutated; ZMYM3 (2%) is novel in prostate cancer; NKX3-1 somatically mutated in 1% (also commonly deleted).
  • BRAF — 2.4% mutation rate (higher than previously reported); 8 mutations, none V600E. Includes K601E, G469A, L597R (MEK-inhibitor sensitive), a K601 in-frame deletion, F468C, and others.
  • HRAS, RAC1, RRAS2 — HRAS Q61R in 3/4 mutants; RAC1 Q61R and RRAS2 Q72L each in one tumor (paralogous to RAS Q61).
  • AKT1 — E17K hotspot in 2 tumors; D323Y in 1 tumor (likely activating, abuts E17K in 3D structure).
  • PIK3CA — Hotspot mutations (E545K, Q546K, N345I, C420R, E542A) in 6 tumors; focal amplification with overexpression in ~1%.
  • PIK3CB — Mutated in 2 PTEN-homozygous-deleted tumors (E552K paralogous to PIK3CA helical-domain hotspot) — suggests combined PI3K/AR inhibition strategy.
  • ATM — Nonsense mutation in 1 case; kinase-dead N2875 hotspot in 2 cases.
  • BRCA1 — 1 germline frameshift (V923, ClinVar RCV000083190.3).
  • BRCA2 — Inactivation in 3% (germline + somatic). All six BRCA2 germline mutations were K3326* (C-terminal truncation, debated impact). Two tumors had focal homozygous deletions with low transcript.
  • CDK12 — Loss-of-function or homozygous deletion in 1% (4 tumors).
  • FANCD2 — Truncating mutation (1 tumor), homozygous deletion (2), focal heterozygous loss in 6%.
  • RAD51C — Focal DNA losses in 3% (mostly heterozygous).
  • CTNNB1, APC — Beta-catenin pathway mutations (APC truncating).
  • KMT2C, KMT2D, KDM6A — Truncating chromatin-modifier mutations; enriched in the 26% unclassified subset.
  • MYC, CCND1 — Recurrent focal amplifications (8% and 2%).
  • FGFR1, WHSC1L1 — Co-amplified at 8p11.23 (8%).
  • MAP3K1, MAP3K7 — Recurrent focal deletions; MAP3K7 co-deletion with CHD1.
  • RB1 — Heterozygous loss often co-incident with BRCA2 loss (13q).
  • SPINK1 — Frequently overexpressed in SPOP-mutant/CHD1-deleted tumors.
  • AR — Mutation/amplification essentially absent in primary disease but frequent in mCRPC. AR-V7 and other splice variants detectable at low levels in primary tumors.
  • FOXP1/RYBP/SHQ1 — Complex 3p13 deletion locus.
  • SPOPL — Recurrent focal deletion (2q22.1).
  • MTOR — Rare activating mutations contributing to PI3K-pathway aberration.

Clinical implications

  • PARP inhibitors (e.g., olaparib) — 19% of primary tumors have DNA-repair gene defects (BRCA2, BRCA1, ATM, CDK12, FANCD2, RAD51C). The authors explicitly link this to TOPARP-A trial results in mCRPC and propose that PARP-inhibitor strategies could be considered at earlier disease stages (PMID:26544944).
  • PI3K/MAPK pathway inhibitors — ~25% of tumors carry actionable PI3K or MAPK pathway lesions. PTEN-deleted/PIK3CB-mutant cases may benefit from combined PI3K + androgen-signaling inhibition. Non-V600E BRAF mutants (K601E, L597R) may be MEK-inhibitor sensitive.
  • IDH1-targeted therapy — Rare IDH1 R132-mutant prostate cancers (~1%) may be candidates for IDH1 inhibitors, by analogy with GBM and AML, though clinical follow-up was inadequate to assess prognosis directly.
  • AR-pathway considerations — Variable AR transcriptional output across subtypes (highest in SPOP/FOXA1-mutant tumors) implies subtype-stratified responses to AR-directed therapies such as enzalutamide and abiraterone. AR-V7 expression — previously a marker of resistance to AR-targeted therapy in mCRPC — is detectable in some hormone-naive primary tumors.
  • Molecular subtyping for risk stratification — The seven-class taxonomy and SCNA burden subgrouping (quiet/intermediate/high) provide a framework for prognostic and treatment-selection studies, although outcome analysis was precluded in this cohort.

Limitations & open questions

  • Short clinical follow-up (average <2 years post-prostatectomy) precluded formal outcome and prognostic analyses.
  • Single tumor focus per patient, despite known multifocality and inter-focus molecular heterogeneity in primary prostate cancer (Cooper et al. 2015; Boutros et al. 2015; Lindberg et al. 2013) — biomarker panels must account for multifocal disease.
  • 26% of tumors lack a taxonomy-defining driver, even after WGS of 19 high-cellularity unclassified cases; no recurrent regulatory (e.g., TERT-promoter) mutations identified. This subset is clinically and genomically heterogeneous.
  • Lower tumor cellularity in low-SCNA tumors may partially mask alteration detection (p = 0.0002 for SCNA-low vs others); however, ETS-fusion-positive vs negative cellularity not significantly different (p = 0.32).
  • AR splice variant expression in primary tumors — functional and prognostic significance unclear since AR-V7 levels did not correlate with AR-target-gene expression or with the seven genomic subtypes.
  • Mechanism of FOXA1 missense mutations — these affect the winged-helix DNA-binding domain but not direct DNA-binding residues; impact may relate to chromatin cofactor interactions rather than DNA binding per se.
  • Drivers of ERG-fusion epigenetic heterogeneity — why ERG-positive tumors split into two methylation subclusters remains unknown.
  • IDH1-mutant prostate cancer — too rare here for outcome analysis; prognostic implications and IDH-inhibitor sensitivity remain to be tested in larger cohorts.

Citations from this paper used in the wiki

  • “we present a comprehensive molecular analysis of 333 primary prostate carcinomas. Our results revealed a molecular taxonomy in which 74% of these tumors fell into one of seven subtypes defined by specific gene fusions (ERG, ETV1/4, FLI1) or mutations (SPOP, FOXA1, IDH1).” — Abstract.
  • “25% of the prostate cancers had a presumed actionable lesion in the PI3K or MAPK signaling pathways, and DNA repair genes were inactivated in 19%.” — Abstract.
  • “In total, 53% of tumors were found to have ETS-family gene fusions (ERG, ETV1, ETV4, and FLI1)” — Results (Molecular taxonomy).
  • “Homozygous deletions spanning the PTEN locus occurred at one of the highest rates of any tumor type studied thus far (15%).” — Results (Recurrently altered genes).
  • “we found that tumors with SPOP or FOXA1 mutations had the highest AR transcriptional activity of all genotypically distinct subsets of prostate cancer (p = 1.1·10−6 and 0.04, respectively, t-test).” — Results (AR activity).
  • “we found inactivation of several DNA repair genes that collectively affected 19% of patients” — Results (Clinically actionable DNA repair defects).
  • “These IDH1 R132-mutant tumors defined a distinct subgroup of what appears to be early onset prostate cancer … that possesses fewer DNA copy-number alterations” — Results (Epigenetic changes).
  • “Curiously, IDH1-mutant prostate cancers possessed even greater levels of genome-wide hypermethylation than do either glioma or AML IDH1-mutant tumors” — Results (Epigenetic changes).
  • “BRAF mutations (2.4%) seen in this study is higher than previously reported; these include several known activating mutations, but curiously not the canonical V600E hotspot” — Results (Recurrently altered genes).
  • “The observation that nearly 20% of primary prostate cancers bear genomic defects involving DNA repair pathways is remarkably consistent with the recently announced TOPARP-A Phase II trial results in patients with metastatic castration-resistant prostate cancer indicating that clinical responses to the PARP inhibitor olaparib likely occurred in the subgroup of tumors bearing defects in DNA repair genes” — Discussion.

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