Integrative clinical genomics of advanced prostate cancer

Authors

Dan Robinson

Eliezer M. Van Allen

Yi-Mi Wu

Nikolaus Schultz

Robert J. Lonigro

Juan-Miguel Mosquera

Bruce Montgomery

Mary-Ellen Taplin

Colin C. Pritchard

Gerhardt Attard

Himisha Beltran

Wassim M. Abida

Charles L. Sawyers

Arul M. Chinnaiyan

Doi

10.1016/j.cell.2015.05.001

PMID: 26000489 · DOI: 10.1016/j.cell.2015.05.001 · Journal: Cell (2015)

TL;DR

The SU2C–PCF Dream Team performed prospective whole-exome and transcriptome sequencing on bone or soft-tissue biopsies from 150 men with metastatic castration-resistant prostate cancer (mCRPC), establishing the first large prospective integrative-genomics cohort for this disease state. Aberrations of AR, ETS genes, TP53, and PTEN were each present in 40–60% of cases, with TP53 and AR enriched in mCRPC vs. primary prostate cancer. Novel recurrent alterations were identified in PIK3CA/B, R-spondin, BRAF/RAF1, APC, CTNNB1, and ZBTB16/PLZF. DNA-repair pathway aberrations (BRCA2, BRCA1, ATM) occurred in 19.3% of cases—substantially higher than in primary disease—and 8% of patients harbored clinically actionable pathogenic germline alterations. Overall, 89% of patients harbored at least one clinically actionable aberration, demonstrating that prospective clinical sequencing is feasible in mCRPC and could materially affect treatment decisions.

Cohort & data

  • N=150 men with mCRPC with complete integrative clinical sequencing (whole exome, matched germline, transcriptome) from a starting enrollment of 189 individuals; 175 cases were sequenced after pathology review, of which 150 had >20% tumor content as defined by computational analysis.
  • Dataset: prad_su2c_2015 (SU2C–PCF International Dream Team).
  • Biopsy sites: lymph node (42%), bone (28.7%), liver (12.7%), other soft tissues (16.7%).
  • Histology: 96.4% typical high-grade prostate adenocarcinoma; 2.9% with neuroendocrine differentiation; 1 case (0.7%) small-cell neuroendocrine features.
  • Sequencing: SureSelect Exome v4 (whole-exome-seq) at mean target coverage of 160× (tumor) / 100× (normal); paired-end rna-seq on Illumina HiSeq 2500 (2×100 nt, ~50M paired reads). Reference genome: hg19/GRCh37.
  • Analytics: somatic mutation calling via mutect (Cibulskis 2013, not in corpus); recurrence analysis via mutsig (Lawrence 2013, not in corpus); fusion detection via Tophat2 with --fusion-search plus Tophat-Post-Fusion (tophat-fusion); copy-number segmentation via circular binary segmentation; FPKM via Cufflinks. Eight clinical sites contributed; two sequencing/analysis centers were used (Broad Institute, University of Michigan).
  • Clinical context: patients being considered for abiraterone acetate or enzalutamide as standard of care or as part of clinical trials, including trials of PARP inhibitors and aurora-kinase inhibitors.

Key findings

  • Mutation burden. Mean somatic mutation rate was 4.4 mutations/Mb; four cases were hypermutated at ~50/Mb, three of which were attributable to alterations in MMR genes MLH1 or MSH2.
  • Recurrent CNAs. Amplification peaks at AR and 8q; deletion peaks at CHD1, PTEN, RB1, TP53. Additional focal amplifications at CCND1, PIK3CA, and PIK3CB. Novel recurrent focal homozygous deletion at chr11q23 narrowed to ZBTB16 (8 cases, 5%).
  • Fusion landscape. 4,122 chimeras across 2,247 unique gene pairs (mean 15 fusions/tumor). Recurrent ETS fusions in 84/150 (56%), majority ERG; novel partners included FLI1, ETV4, and ETV5. Potentially actionable fusions involving BRAF, RAF1, PIK3CA/PIK3CB, or RSPO2 seen in 8 cases.
  • Top recurrently aberrant genes in mCRPC: AR 62.7%, ETS family 56.7%, TP53 53.3%, PTEN 40.7%.
  • Selective enrichment vs. primary prostate cancer (n=440 exomes; Barbieri 2012 + TCGA provisional 2015; cancer-gene-only set of 550 genes). TP53 most selectively mutated in mCRPC (q<0.001, Benjamini-Hochberg), followed by AR, KMT2D, APC, BRCA2, GNAS (q<0.1). AR and GNAS were mutated exclusively in mCRPC; no genes were selectively mutated in primary prostate cancer.
  • Driver coverage. 99% of mCRPC cases harbored a potential driver SNV/indel; 60% harbored a driver gene fusion, 50% a driver homozygous deletion, 54% a driver amplification. Mean of 7.8 biologically relevant alterations per case.
  • AR pathway (71.3% overall; 107/150). AR aberrations were dominated by amplification and mutation. Recurrent hotspot AR mutations included T878A (flutamide agonism), W742C (bicalutamide agonism), and L702H (glucocorticoid agonism). One case (Case 89) carried two AR mutations (T878A and Q903H), suggesting intra-tumor heterogeneity. AR-V7 splice variant was detectable in most pre-abiraterone/enzalutamide cases but at very low ratios relative to full-length AR (treatment-response implications unknown). FOXA1 mutations clustered near the end of the Forkhead DNA-binding domain.
  • PI3K pathway (49%; 73/150). Biallelic PTEN loss, hotspot PIK3CA mutations, and PIK3CA amplifications. Novel: activating PIK3CB mutations at positions equivalent to canonical PIK3CA hotspots, occurring in the context of PTEN deficiency. Activating AKT1 p.E17K observed. Two patients harbored PIK3CA/B fusions resulting in gene over-expression.
  • WNT pathway (18%; 27/150). Hotspot CTNNB1 activating mutations; novel recurrent APC alterations in mCRPC. Alterations in RNF43 and ZNRF3 were mutually exclusive with APC alterations. R-spondin fusions involving RSPO2 (associated with RSPO2 over-expression) overall affected 6% of patients (RNF43/ZNRF3/RSPO2 combined).
  • Cell cycle. RB1 loss in 21%; focal CCND1 amplification in 9%; less common (<5%) events in CDKN2A/CDKN2B, CDKN1B, and CDK4.
  • DNA repair pathway. BRCA2 loss in 19/150 (12.7%), with ~90% biallelic; 8 patients (5.3%) had pathogenic germline BRCA2 plus a subsequent somatic event causing biallelic loss. Broader DNA repair/recombination alterations in 34/150 (22.7%), including biallelic ATM loss (some germline), plus events in BRCA1, CDK12, FANCA, RAD51B, and RAD51C. Combined BRCA2/BRCA1/ATM aberration prevalence: 29/150 (19.3%).
  • Actionability. 89% of patients harbored a clinically actionable aberration: 62.7% AR alterations, 65% non-AR cancer-gene alterations, 8% actionable pathogenic germline. Non-AR actionable categories: PI3K pathway 49%, DNA repair 19%, RAF kinases 3%, CDK inhibitors 7%, WNT pathway 5%.

Genes & alterations

  • AR — Amplification and hotspot mutation (T878A, W742C, L702H, plus novel mutations of unclear function); ~62.7% of cases; AR pathway aggregate 71.3%. AR-V7 splice variant detectable in most pre-abiraterone/enzalutamide cases at low ratios.
  • TP53 — Most selectively mutated gene in mCRPC vs. primary (q<0.001); ~53.3% of cases.
  • PTEN — Biallelic loss; 40.7% of cases.
  • PIK3CA, PIK3CB — Hotspot mutations, focal amplifications, and (novel) activating fusions; PIK3CB mutations at positions equivalent to canonical PIK3CA hotspots co-occur with PTEN loss.
  • AKT1 — Activating p.E17K mutation.
  • BRAF, RAF1 — Potentially actionable fusions in ~3% of patients.
  • APC, CTNNB1, RNF43, ZNRF3, RSPO2 — WNT pathway: APC alterations recurrently observed in mCRPC for the first time; CTNNB1 hotspot activating mutations; RNF43/ZNRF3 mutually exclusive with APC; RSPO2 fusions with overexpression.
  • ZBTB16 — Novel recurrent focal homozygous deletion at chr11q23 in 8 (5%) cases. ZBTB16 (PLZF) is androgen-regulated; preclinical models link loss to androgen resistance and MAPK upregulation.
  • BRCA2, BRCA1, ATM — Biallelic loss (somatic + germline) totaling 19.3% of cases; multiple patients with germline BRCA2 + somatic second hit. Some patients with biallelic BRCA2 or ATM exhibited clinical responses to PARP inhibition in a parallel trial.
  • CDK12, FANCA, RAD51B, RAD51C, BRIP1 — Broader DNA-repair pathway alterations contributing to the 22.7% aggregate.
  • MLH1, MSH2 — MMR alterations in 3 of 4 hypermutated cases (~50/Mb).
  • ERG, FLI1, ETV4, ETV5 — Recurrent ETS fusions in 56% of cases; majority involve ERG.
  • SPOP, FOXA1, NCOR1, NCOR2 — AR pathway regulators; FOXA1 mutations cluster near the end of the Forkhead DNA-binding domain.
  • RB1, CCND1, CDKN2A, CDKN2B, CDKN1B, CDK4 — Cell-cycle pathway aberrations potentially actionable via CDK4 inhibition.
  • MYC, CHD1 — Recurrent CNAs (8q gain; CHD1 deletion).
  • KMT2D, GNAS — Selectively mutated in mCRPC vs. primary (q<0.1); GNAS mutated exclusively in mCRPC.

Clinical implications

  • PARP-inhibitor candidates. Biallelic inactivation of BRCA2, BRCA1, or ATM in 19.3% of patients identifies a substantial mCRPC subset that may benefit from olaparib or other PARP inhibitors, or from platinum-based chemotherapy. Multiple patients on a parallel PARP-inhibitor trial who achieved clinical benefit harbored biallelic BRCA2 loss (Mateo et al., 2014).
  • Germline testing. 5.3% pathogenic germline BRCA2 with somatic second hit, plus additional germline ATM events, support clinical consideration of germline genetic testing in mCRPC patients. 8% of patients harbored actionable pathogenic germline alterations overall.
  • PI3K-pathway therapeutics. PIK3CB-specific inhibitors may have utility in patients with PIK3CB mutation, amplification, or fusion; the authors cite durable (>1 year) responses to PIK3CB-specific inhibition in carriers of activating PIK3CB mutation/amplification (de Bono 2015).
  • RAF/MEK inhibition. RAF-kinase fusions in 3% of mCRPC support pan-RAF or MEK inhibitor strategies.
  • WNT-pathway therapeutics. RNF43/ZNRF3 mutation or RSPO2 fusion (6% combined) are predicted to confer sensitivity to porcupine inhibitors; R-spondin antibodies may also warrant investigation.
  • CDK4/6 inhibition. CCND1 amplification or CDKN2A/B loss could sensitize to CDK4 inhibition, by analogy to other tumor types (Finn 2015) and preclinical mCRPC models.
  • AR-directed therapy resistance. Recurrent AR hotspots (T878A, W742C, L702H) that confer agonism to bicalutamide, flutamide, or glucocorticoids predict differential responses to second-generation ADT. AR-V7 was detectable at low ratios relative to full-length AR; clinical-response implications were not assessable from this cross-sectional dataset.
  • Actionability headline. 89% of mCRPC patients harbored at least one clinically actionable somatic or germline alteration; even excluding AR, 65% had a non-AR actionable aberration.

Limitations & open questions

  • Cross-sectional snapshot. Only baseline biopsies are reported; longitudinal treatment-response data are deferred to future reports, so the genomic-to-clinical link (e.g., whether AR T878A predicts flutamide failure in this cohort) is not yet established.
  • Tumor content threshold. Cases with ≤20% tumor content (39 of 189 enrolled) were excluded from integrative analysis, potentially biasing toward higher-purity, more accessible metastases.
  • Biopsy-site sampling. Bone metastases—the most common mCRPC site—account for only 28.7% of biopsies because of technical recovery challenges; lymph node (42%) is over-represented.
  • AR-V7 functional interpretation. AR-V7 was present in most pre-abiraterone/enzalutamide cases but at low ratios versus full-length AR; the clinical-response implications are explicitly stated as unknown.
  • Heterogeneity not directly resolved. Some cases carried multiple AR mutations in the same biopsy, hinting at clonal heterogeneity, but multi-region or longitudinal sampling was not part of this initial report.
  • Functional validation pending. Several rare or novel AR mutations were identified but their functional significance is of unclear; the impact of PIK3CB activating mutations and ZBTB16 deletion on mCRPC biology warrants exploration in organoid and preclinical systems.
  • Epigenetic profiling not performed. The authors explicitly flag epigenetic profiling as an immediate next step for mCRPC.

Citations from this paper used in the wiki

  • “we established a multi-institutional clinical sequencing infrastructure to conduct prospective whole exome and transcriptome sequencing of bone or soft tissue tumor biopsies from a cohort of 150 mCRPC affected individuals” (Summary).
  • “Aberrations of BRCA2, BRCA1 and ATM were observed at substantially higher frequencies (19.3% overall) than seen in primary prostate cancers” (Summary).
  • “89% of affected individuals harbored a clinically actionable aberration including 62.7% with aberrations in AR, 65% in other cancer-related genes, and 8% with actionable pathogenic germline alterations” (Summary).
  • “A novel recurrent focal homozygous deletion event was observed in chr11q23, encompassing the transcriptional repressor ZBTB16” (Results — Landscape of mCRPC alterations).
  • “recurrent ETS fusions … were observed in 84 cases (56%), with a majority of these fusions to ERG and also novel fusions to FLI1, ETV4, and ETV5” (Results).
  • “somatic TP53 mutations were the most selectively mutated (q < 0.001; Benjamini-Hochberg), followed by AR, KMT2D, APC, BRCA2, and GNAS (q < 0.1; Benjamini-Hochberg). Both AR and GNAS were mutated exclusively in mCRPC” (Results).
  • “Recurrent hotspot mutations in AR were observed at residues previously reported to confer agonism to AR antagonists such as flutamide (T878A) and bicalutamide (W742C), as well as to glucocorticoids (L702H)” (AR signaling pathway).
  • “mutations in another member of the PI3K catalytic subunit, PIK3CB, were observed in this cohort for the first time, at equivalent positions to canonical activating mutations in PIK3CA” (New PI3K pathway discoveries).
  • “Affected individuals with aberrations in RNF43, ZNRF3, or RSPO2 (overall 6% of affected individuals) are predicted to respond to porcupine inhibitors” (New Wnt pathway discoveries).
  • “If aberrations of BRCA2, BRCA1 and ATM all confer enhanced sensitivity to PARP inhibitors, 29/150 (19.3%) of mCRPC affected individuals would be predicted to benefit from this therapy” (DNA Repair pathway).
  • “3 out of 4 mCRPC tumors exhibited hypermutation and harbored alterations in the mismatch repair pathway genes MLH1 or MSH2” (DNA Repair pathway).

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