Divergent clonal evolution of castration resistant neuroendocrine prostate cancer

Authors

Himisha Beltran

Davide Prandi

Juan Miguel Mosquera

Matteo Benelli

Loredana Puca

Joanna Cyrta

Clarisse Marotz

Eugenia Giannopoulou

Balabhadrapatruni V.S.K. Chakravarthi

Sooryanarayana Varambally

Scott A. Tomlins

David M. Nanus

Scott T. Tagawa

Eliezer M. Van Allen

Olivier Elemento

Andrea Sboner

Levi A. Garraway

Mark A. Rubin

Francesca Demichelis

Doi

10.1038/nm.4045

PMID: 26855148 · DOI: 10.1038/nm.4045 · Journal: Nature Medicine (2016)

TL;DR

Beltran et al. profiled 114 metastatic biopsies from 81 men with castration resistant prostate cancer — 51 with adenocarcinoma histology (CRPC-Adeno) and 30 with neuroendocrine histology (CRPC-NE) — using whole-exome sequencing, RNA-seq, and genome-wide DNA methylation (eRRBS). CRPC-NE and CRPC-Adeno share a largely overlapping somatic copy-number landscape, but CRPC-NE is enriched for concurrent RB1 loss + TP53 alteration and is depleted of AR genomic alterations. Serial samples support a divergent clonal evolution model: CRPC-NE arises from a CRPC-Adeno precursor under AR-targeted therapy pressure rather than from pre-existing neuroendocrine clones. Strong epigenetic segregation (including SPDEF hypermethylation and EZH2 over-expression) distinguishes the two states and yielded a 70-gene NEPC classifier with precision/recall >0.99 in the discovery cohort. (PMID:26855148)

Cohort & data

  • 81 patients / 114 metastatic biopsies of castration-resistant prostate cancer: 51 PRAD-Adeno (70 samples) and 30 CRPC-NE (44 samples); matched normals from all patients; multi-region/serial biopsies in 17 patients (PMID:26855148).
  • Cohort identifier: nepc_wcm_2016 (Weill Cornell Medicine NEPC cohort); BAM files deposited at dbGaP phs000909.v.p1 and made available on the cBioPortal for Cancer Genomics (PMID:26855148).
  • Assays: whole-exome sequencing using SureSelect v2/v4 or HaloPlex Exome (Agilent), Illumina HiSeq, mean target coverage ~100×; RNA-seq (TruSeq, HiSeq 2500, 2×75 bp paired-end) on 49 specimens; eRRBS genome-wide single-cytosine methylation; custom NanoString assay for samples with limited material; immunohistochemistry for AR, synaptophysin, chromogranin A, CD56, EZH2, MMR proteins; FISH for CYLD verification (PMID:26855148).
  • Computational tools: MuTect, Oncotator, MutSig, and CLONET for tumor purity/ploidy and allele-specific copy-number clonality (PMID:26855148).
  • External validation cohorts queried for the NEPC classifier: TCGA prostate (prad_tcga, n=460 treatment-naïve adenocarcinomas), Grasso et al. Michigan 2012 (prad_mich), and Robinson et al. SU2C/PCF 2015 (prad_su2c_2015) — totaling 683 prostate samples (PMID:26855148).

Key findings

  • Concurrent RB1/TP53 loss is the genetic hallmark of CRPC-NE. RB1 deletion: 70% of CRPC-NE vs. 32% of CRPC-Adeno (P = 0.003, proportion test). TP53 mutation or deletion: 66.7% vs. 31.4% (P = 0.0043). Concurrent RB1 + TP53 loss: 53.3% vs. 13.7% (P < 0.0004) (PMID:26855148).
  • CRPC-NE lacks the AR alterations that dominate CRPC-Adeno (P < 0.0001, Wilcoxon test). 29 cases showed AR focal amplification or point mutation and 21 had alterations in AR co-activators (FOXA1, NCOR1/NCOR2, ZBTB16); AR point mutations were notably absent in CRPC-NE and AR gains were low-level and explained by polyploidy. The ARv7-to-wild-type AR ratio was significantly decreased in CRPC-NE (P = 0.0025) (PMID:26855148).
  • No global differences in non-silent SNV rate, polyploidy, or copy-number burden between subtypes; both have >30% of the genome aberrant on average; median 41 non-silent SNVs (range 2–729). 5 of 6 hyper-mutated samples (115–663 SNVs) carried mismatch-repair gene alterations (PMID:26855148).
  • CYLD focal deletion enriched in CRPC-NE. CYLD deleted in 51% of CRPC-NE samples (FDR<10% for both DNA and mRNA, FISH-verified), with concordant down-regulation of mRNA and modest decrease of AR-signaling genes — also confirmed in the SU2C/PCF cohort and cell lines (PMID:26855148).
  • Focal allelic imbalance of DEK in CRPC-NE vs. CRPC-Adeno (P = 0.04, binomial test) (PMID:26855148).
  • Divergent clonal evolution from a CRPC-Adeno precursor. Serial sampling in patient WCMC7520 (homozygous BRCA2 deletion and TP53 mutation in primary, lymph-node, and CRPC-NE metastases; divergent allelic states of MYCN) and patient WCMC161 (CRPC-Adeno LN, CRPC-Adeno bone, and CRPC-NE liver-on-abiraterone with divergent SBDS allelic states) — together with multi-tumor phylogenetic trees — favor model V (divergent clonal evolution) over linear or independent parallel models (PMID:26855148).
  • Strong epigenetic segregation of CRPC-NE from CRPC-Adeno on unsupervised eRRBS analysis, despite genomic similarity. 22% of top dysregulated CRPC-NE transcripts had concordant DNA-methylation changes (P < 0.0002). Methylation-based clustering re-classified 3 histologic adenocarcinomas with clinical AR-independence (radiographic progression with stable/low PSA) as CRPC-NE (PMID:26855148).
  • Hypermethylation and silencing of SPDEF in CRPC-NE (P < 10⁻⁹, Wilcoxon test); confirmed in NCI-H660 (NE) vs. LNCaP (adeno) cell lines (PMID:26855148).
  • EZH2 over-expression and PRC2 target down-regulation in CRPC-NE. EZH2 mRNA 2× higher than in CRPC-Adeno (P < 10⁻⁶); EZH2-repressed targets including WNT genes DKK1 (P = 0.0002), NKD1 (P = 0.0046), and HOX genes (P = 0.001) significantly down-regulated. EZH2 inhibitor GSK343 preferentially reduced viability of NCI-H660 vs. non-NE prostate cell lines and down-regulated CRPC-NE genes NCAM1 (CD56), MYCN, and PEG10 (PMID:26855148).
  • 70-gene integrated NEPC classifier (built from genes prioritized by genomic, transcriptomic, and epigenomic signal — including AURKA at P < 10⁻⁵ and MYCN at P < 10⁻⁴) achieved precision and recall >0.99 in the discovery cohort, identified an elevated NEPC score in up to 8% of metastatic prostate tumors across prad_tcga, prad_mich, and prad_su2c_2015 (n=191 metastatic) and 0% of treatment-naïve adenocarcinoma (n=460) or benign prostate (n=32). >80% of high-score cases on pathology re-review had CRPC-NE features (PMID:26855148).
  • In vitro plasticity: LNCaP cells treated long-term with enzalutamide acquired CRPC-NE molecular features including SPDEF promoter methylation (PMID:26855148).

Genes & alterations

  • AR — focal amplification, activating point mutations, and ARv7 splice variant common in CRPC-Adeno but largely absent / low-level in CRPC-NE; AR signaling attenuated in CRPC-NE (PMID:26855148).
  • RB1 — deletion in 70% CRPC-NE vs. 32% CRPC-Adeno (P = 0.003) (PMID:26855148).
  • TP53 — mutation/deletion in 66.7% CRPC-NE vs. 31.4% CRPC-Adeno (P = 0.0043); concurrent with RB1 loss in 53.3% of CRPC-NE (P < 0.0004) (PMID:26855148).
  • BRCA2 — homozygous deletion shared across primary, lymph-node, and CRPC-NE metastases in patient WCMC7520, marking a common ancestor (PMID:26855148).
  • MYCN — divergent allelic states across metastatic sites in patient WCMC7520; over-expressed in CRPC-NE (P < 10⁻⁴); included in NEPC classifier; down-regulated by GSK343 in NCI-H660 (PMID:26855148).
  • CYLD — focal deletion in 51% of CRPC-NE with concordant mRNA loss (FISH-verified); deletion modestly suppresses AR signaling (PMID:26855148).
  • DEK — focal allelic imbalance enriched in CRPC-NE (P = 0.04, binomial test) (PMID:26855148).
  • SBDS — divergent allelic state across CRPC-Adeno LN, CRPC-Adeno bone, and CRPC-NE liver metastases in patient WCMC161, illustrating clonal divergence (PMID:26855148).
  • FOXA1, NCOR1, NCOR2, ZBTB16 — AR co-activator alterations enriched in CRPC-Adeno (21 cases) (PMID:26855148).
  • TMPRSS2–ERG — fusion concordance between adenocarcinoma and neuroendocrine foci of mixed tumors cited as supporting common cell of origin (PMID:26855148).
  • EZH2 — mRNA and protein up in CRPC-NE (P < 10⁻⁶); EZH2 inhibitor GSK343 preferentially kills NCI-H660 (PMID:26855148).
  • SPDEF — promoter hypermethylation and mRNA loss in CRPC-NE (P < 10⁻⁹); also induced by long-term enzalutamide treatment of LNCaP (PMID:26855148).
  • AURKA — over-expressed in CRPC-NE (P < 10⁻⁵); part of NEPC classifier (PMID:26855148).
  • NCAM1 (CD56), PEG10 — CRPC-NE-associated transcripts down-regulated by EZH2 inhibition (PMID:26855148).
  • DKK1, NKD1 — WNT-pathway EZH2 targets significantly down-regulated in CRPC-NE (P = 0.0002 and P = 0.0046 respectively) (PMID:26855148).

Clinical implications

  • Diagnostic supplement to morphology. The integrated 70-gene NEPC classifier outperformed conventional neuroendocrine markers (CHGA, SYP, NSE, CD56 transcript ± PSA), AR mRNA, AR-signaling status, and individual differentially-expressed genes such as SPDEF in distinguishing CRPC-NE — and re-classified three histologic adenocarcinomas with clinical AR-independence (radiographic progression with stable/low PSA) as CRPC-NE (PMID:26855148).
  • Predictive biomarker hypothesis for AR-targeted therapy. Castration-resistant tumors with moderate or rising NEPC scores may represent AR-independent transitional states unlikely to respond to next-line abiraterone or enzalutamide, and could be candidates for platinum chemotherapy or co-targeting strategies (PMID:26855148).
  • Therapeutic rationale for EZH2 inhibition in CRPC-NE: EZH2 over-expression, PRC2-target silencing, and selective in vitro vulnerability of NCI-H660 to GSK343 nominate EZH2 inhibitors as candidate epigenetic modifiers to test for reversing or preventing the CRPC-NE state (PMID:26855148).
  • Liquid-biopsy applicability. Because the classifier accepts partial input (DNA, RNA, or methylation), it is amenable to small metastatic biopsies and potentially circulating tumor DNA — relevant where multiple immunohistochemical assays are impractical (PMID:26855148).

Limitations & open questions

  • The authors cannot fully exclude metastasis-to-metastasis seeding as an alternative to divergent evolution within individual patients (PMID:26855148).
  • An alternative model — de-differentiation of adenocarcinoma to a progenitor-like state with subsequent local acquisition of neuroendocrine features — cannot be fully excluded (PMID:26855148).
  • Cases with low AR signaling and low NEPC classifier score exist; a less-common, distinct AR-low non-NE subset cannot be ruled out (PMID:26855148).
  • Validation cohorts contained relatively few CRPC-NE cases; larger clinical studies are needed to confirm classifier superiority and prognostic/predictive utility (PMID:26855148).
  • Effects of prior therapy differences between CRPC-Adeno and CRPC-NE patients on gene expression and the apparent absence of AR alterations in CRPC-NE cannot be fully disentangled (PMID:26855148).
  • The reversibility of the CRPC-NE state with early intervention or epigenetic modifiers (e.g. EZH2 inhibitors) is hypothesized but not yet tested in patients (PMID:26855148).

Citations from this paper used in the wiki

  • “concurrent RB1 and TP53 loss was present in 53.3% of CRPC-NE vs.13.7% of CRPC-Adeno (P < 0.0004, proportion test).” (p. 3)
  • “AR point mutations were notably absent in CRPC-NE and gains when present were of low level and explained by tumor polyploidy.” (p. 3)
  • “CYLD gene, deleted in 51% of CRPC-NE samples, encodes cylindromatosis, a deubiquinating enzyme reported as a tumor suppressor…” (p. 4)
  • “the most parsimonious model that explains the data is divergent clonal evolution of metastatic CRPC towards either an AR-driven or AR-indifferent state” (p. 5)
  • “Treatment with the EZH2 inhibitor GSK343 resulted in a preferential decrease in cellular viability of NCI-H660 compared to other non-NE prostate cancer cell lines” (p. 6)
  • “This integrated 70 gene neuroendocrine prostate cancer (NEPC) classifier… demonstrated both a precision and recall of >0.99 in identifying CRPC-NE in our discovery cohort” (p. 6)
  • “a subpopulation of prostate adenocarcinoma cells (LNCaP) treated long term with enzalutamide acquired molecular features of CRPC-NE (ie., methylation of SPDEF)” (p. 7)

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