Germline BRCA2 mutations drive prostate cancers with distinct evolutionary trajectories

Abstract

Germline mutations in the BRCA2 tumour suppressor are associated with both an increased lifetime risk of developing prostate cancer (PCa) and increased risk of aggressive disease. To understand this aggression, here we profile the genomes and methylomes of localized PCa from 14 carriers of deleterious germline BRCA2 mutations (BRCA2-mutant PCa). We show that BRCA2-mutant PCa harbour increased genomic instability and a mutational profile that more closely resembles metastastic than localized disease. BRCA2-mutant PCa shows genomic and epigenomic dysregulation of the MED12L/MED12 axis, which is frequently dysregulated in metastatic castration-resistant prostate cancer (mCRPC). This dysregulation is enriched in BRCA2-mutant PCa harbouring intraductal carcinoma (IDC). Microdissection and sequencing of IDC and juxtaposed adjacent non-IDC invasive carcinoma in 10 patients demonstrates a common ancestor to both histopathologies. Overall we show that localized castration-sensitive BRCA2-mutant tumours are uniquely aggressive, due to de novo aberration in genes usually associated with metastatic disease, justifying aggressive initial treatment.

ARTICLE

Received 26 Jul 2016 | Accepted 20 Oct 2016 | Published 9 Jan 2017

DOI: 10.1038/ncomms13671 OPEN

Germline BRCA2 mutations drive prostate cancers with distinct evolutionary trajectories

Renea A. Taylor1,*, Michael Fraser2,*, Julie Livingstone3,*, Shadrielle Melijah G. Espiritu3,*, Heather Thorne4,5,*, Vincent Huang3, Winnie Lo2, Yu-Jia Shiah3, Takafumi N. Yamaguchi3, Ania Sliwinski5,6, Sheri Horsburgh2, Alice Meng2, Lawrence E. Heisler3, Nancy Yu3, Fouad Yousif3, Melissa Papargiris7, Mitchell G. Lawrence7,Lee Timms8, Declan G. Murphy9, Mark Frydenberg7, Julia F. Hopkins3, Damien Bolton7, David Clouston10, John D. McPherson8, Theodorus van der Kwast2, Paul C. Boutros3,11,12,**, Gail P. Risbridger7,**& Robert G. Bristow2,11,**

Germline mutations in the BRCA2 tumour suppressor are associated with both an increased lifetime risk of developing prostate cancer (PCa) and increased risk of aggressive disease. To understand this aggression, here we prole the genomes and methylomes of localized PCa from 14 carriers of deleterious germline BRCA2 mutations (BRCA2-mutant PCa). We show that BRCA2-mutant PCa harbour increased genomic instability and a mutational prole that more closely resembles metastastic than localized disease. BRCA2-mutant PCa shows genomic and epigenomic dysregulation of the MED12L/MED12 axis, which is frequently dysregulated in metastatic castration-resistant prostate cancer (mCRPC). This dysregulation is enriched in BRCA2-mutant PCa harbouring intraductal carcinoma (IDC). Microdissection and sequencing of IDC and juxtaposed adjacent non-IDC invasive carcinoma in 10 patients demonstrates a common ancestor to both histopathologies. Overall we show that localized castration-sensitive BRCA2-mutant tumours are uniquely aggressive, due to de novo aberration in genes usually associated with metastatic disease, justifying aggressive initial treatment.

1 Monash Partners Comprehensive Cancer Consortium and Cancer Program, Biomedicine Discovery Institute, Department of Physiology, Monash University, Melbourne, Victoria 3800, Australia. 2 Princess Margaret Cancer Centre, University Health Network, Toronto, Ontario, Canada M5G 1L7. 3 Informatics& Biocomputing Program, Ontario Institute for Cancer Research, Toronto, Ontario, Canada M5G 0A3. 4 kConFab, Research Department, Peter MacCallum Cancer Centre, Melbourne, Victoria 3000, Australia. 5 The Sir Peter MacCallum Department of Oncology University of Melbourne, Parkville, Victoria 3010, Australia. 6 Department of Urology, University of Melbourne, Austin Hospital, Heidelberg, Victoria 3084, Australia. 7 Monash Partners Comprehensive Cancer Consortium and Cancer Program Biomedicine Discovery Institute, Department of Anatomy and Developmental Biology, Monash University, Melbourne, Victoria 3800, Australia. 8 Genome Technologies Program, Ontario Institute for Cancer Research, Toronto, Ontario, Canada M5G 0A3. 9 Department of Cancer Surgery, Peter MacCallum Cancer Centre, Melbourne, Victoria 3000, Australia. 10 Urological Pathology, Tissupath, Mt Waverley, Victoria 3149, Australia. 11 Department of Medical Biophysics, University of Toronto, Toronto, Ontario, Canada M5G 1L7. 12 Department of Pharmacology & Toxicology, University of Toronto, Toronto, Ontario, Canada M5S 1A8. * These authors contributed equally to this work. ** These authors jointly supervised this work. Correspondence and requests for materials should be addressed to P.C.B. (email: mailto:[email protected]

Web End [email protected] ) or to G.P.R. (email: mailto:[email protected]

Web End [email protected] mailto:[email protected]

Web End =) or to R.G.B. (email: mailto:[email protected]

Web End [email protected] ).

NATURE COMMUNICATIONS | 8:13671 | DOI: 10.1038/ncomms13671 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications 1

ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms13671

Germline mutation of the BRCA2 tumour suppressor gene substantially increases the lifetime risk of developing prostate cancer (PCa)1,2. In BRCA2-mutation carriers,

localized PCa rapidly progresses to metastatic castrate-resistant prostate cancer (mCRPC) with 5-year cancer-specic survival rates of B5060% (refs 1,3,4). BRCA2-mutant tumours also exhibit an increased frequency of intraductal carcinoma (IDC), a pathology that predicts adverse outcome in both familial5 and sporadic PCa6,7. Prostate tumours arising in men with an inactivating BRCA2 germline mutation (BRCA2-mutant PCa) are uniquely aggressive, associated with younger age of onset, have higher rates of lymph node and distant metastasis, and increased mortality relative to sporadic, non-BRCA2-mutant disease3,5,6. Precision surgery or radiotherapy with curative intent inexplicably fails, with a rapid onset of incurable mCRPC in BRCA2-mutant PCa79. The molecular origins of the clinical aggressiveness of BRCA2-mutant PCa are unknown. Understanding them may allow better treatment decision-making; for example, given their DNA repair decient genotype, these cancers may be amenable to treatment with PARP inhibitors, based on genetic synthetic lethality1012.

In this study, to identify the drivers of aggression in BRCA2-mutant PCa, we characterized genomic alterations in localized PCa tumour specimens from 14 men carrying a deleterious germline mutant BRCA2 allele using either whole-genome sequencing (WGS) and/or SNP array-based copy-number analyses (CNAs). We nd that BRCA2-mutant PCa harbours a mutational prole more akin to that of mCRPC than localized sporadic PCa, and shows broad dysregulation of pathways associated with aggressive disease, including the MED12/MED12L axis. Micro-dissection and subclonal reconstruction shows distinct evolutionary trajectories and identies a common precursor to IDC and adjacent invasive carcinoma. Taken together, these data demonstrate the genomic correlates underlying aggressive BRCA2-mutant PCa.

ResultsThe Mutational Landscape of BRCA2-mutant PCa. To address intra-prostatic and intra-lesion heterogeneity, we microdissected regions of invasive carcinoma (IC) and of IDC from ve tumours, resulting in 19 PCa specimens (Supplementary Table 1; Supplementary Fig. 1).

We rst identied single-nucleotide variants (SNVs), genomic rearrangements (GRs) and CNAs from these samples using validated pipelines (Supplementary Table 1; Supplementary Fig. 2). We compared the mutational proles of BRCA2-mutant PCa to a set of 200 sporadic PCa with whole-genome sequencing13. We separated this comparison cohort into patients 50 years of age or younger at the time of treatment (n 8) and those older than 50 years (n 192), to control for

age effects. BRCA2-mutant PCa showed elevated genomic instability, as measured by percent genome alteration (PGA), an established prognostic factor for relapse in localized disease (P 7.53 10 4; two-sided t-test; Fig. 1a)14,15. Similarly,

BRCA2-mutant tumours showed elevated numbers of SNVs and GRs than sporadic tumours arising in either older or younger men (Supplementary Table 2; Supplementary Fig. 3).

Overall, BRCA2-mutant PCa shows the same C-class character16 as sporadic PCa, with a large number of CNAs and a relative paucity of SNVs. Regions of the genome recurrently gained or deleted in BRCA2-mutant PCa but not sporadic PCa include a large amplication on chromosome 3q (Fig. 1b). Genomic changes associated with aggressive disease were more common, including gain of the MYC oncogene, which occurred in 72% (13/18) of BRCA2-mutant PCa which was signicantly

higher than the 20.1% rate in sporadic PCa of equivalent Gleason score (57/283; P 1.42 10 3; two-sided proportion test17.

Gains in MYCN, GSK3B and MTOR were also more common in BRCA2-mutant PCa. By contrast, losses of the tumour suppressors TP53 and NKX3-1 occurred at similar rates in BRCA2-mutant and sporadic PCa (Fig. 1c). Supplementary Data 1 lists all genes with differential rates of copy-number aberrations. No individual gene was altered by SNVs in more than one BRCA2-mutant patient (Supplementary Fig. 4). In fact, of the 11 most frequently mutated genes in a cohort of 477 sporadic localized PCa (ref. 13), only 1 was mutated in BRCA2-mutant PCa: TP53 (Supplementary Fig. 5). Similarly of the 32 genes mutated in at least two BRCA2-mutant PCa, only 12 were mutated in 477 sporadic PCa. One BRCA2-mutant PCa harboured a kataegic event (Supplementary Fig. 6).

Although the TMPRSS2:ERG gene fusion on chromosome 21 was observed at similar frequencies in BRCA2-mutant and sporadic PCa (Fig. 2a; 3/12 versus 68/200; P 0.74, proportion

test; Supplementary Table 1), we identied several GRs in BRCA2-mutant PCa (Fig. 2b; Supplementary Fig. 7; Supplementary Data 2). The most prominent of these was an inversion on 3p26.1 which occurred in 4/12 BRCA2-mutant specimens but 0/200 sporadic PCa. This region contains only four genes: a metabotropic glutamate receptor (GRM7) and two of its antisense RNAs, along with a mitochondrial ribosomal pseudogene. GRM7 inactivation suggests loss of negative regulation of NMDA signaling, as previously associated with the neuroendocrine disease (NED) phenotype. NED commonly arises in late-stage disease and is associated with treatment resistance and poor prognosis18. There was also a recurrent translocation on chromosome 7q in a non-genic region proximal to several solute carriers and membrane proteins. BRCA2-mutant PCa tumours showed similar chromothriptic character to sporadic disease (Supplementary Fig. 8).

Surprisingly, BRCA2-mutant PCa showed global hypomethylation relative to sporadic PCa (Supplementary Fig. 9) which was retained even after controlling for the increased CNA rate of BRCA2-mutant PCa (Supplementary Figs 10 and 17). Overall B1% (4,979/435,377) of copy-number neutral probes showed differential methylation between BRCA2-mutant and sporadic PCa specimens (Supplementary Data 3), and 74% (3,677/4,979) of these were hypomethylated, including 21 of the 25 most differentially methylated genes (Supplementary Fig. 11). Differentially methylated genes in BRCA2-mutant PCa were enriched in pathways associated with VEGF signaling, neuronal development, and other neuronal processes characteristic of NED (Supplementary Fig. 12; Supplementary Table 3). Other hypomethylated NED-related genes include modulators of WNT signaling (for example, AXIN, APC, ROCK1/2), IL6 signaling (for example, IL6ST, BMP6), and mTOR signaling (for example, TSC2), and many of these genes were also altered by CNAs (Supplementary Fig. 13). The 30 UTR of the BRCA2 locus was hypomethylated in BRCA2-mutant PCa specimens (Supplementary Fig. 14).

MED12/MED12L amplication is associated with aggressive PCa. To identify other driver events that might help explain the aggressiveness of BRCA2-mutant PCa, we investigated the amplication of a region on chromosome 3q (Fig. 1b) that is common in BRCA2-mutant PCa (11/18; 61%), but rare in sporadic PCa (55/809; 6.8% P 1.64 10 9, two-sided pro

portion test). The consensus amplicon of this locus contains the WNT/b-catenin pathway modulator MED12L (Supplementary

Data 4 and 7), as well as ATR, which encodes the DNA damage response factor ataxia telangiectasia-and-Rad3 related protein.

2 NATURE COMMUNICATIONS | 8:13671 | DOI: 10.1038/ncomms13671 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications

NATURE COMMUNICATIONS | DOI: 10.1038/ncomms13671 ARTICLE

a c

40

20

MYC

5,0000

# of SNVs # of CNAs

# of GRsPGA

GSK3B

2,500

MTOR

0

TP53

500

100

NKX3-1

10

1

BRCA2

1,000

250

MYCN

50

0

BRCA2 Sporadic 50 Sporadic > 50

0.25 0.5 0.75

10

Frequency

1

BRCA2

Sporadic

50 yrs

Sporadic

> 50 yrs

b

10

Q

10

0

0

1

1

0

0

F BRAC2 50 yrs > 50 yrs

1

MED12L ATR MED12 AR

CHD1 NKX3-1MYC PTEN CDKN1B RB1 CDH1TP53

Patient

BR0450 BR0556 BR0001 BR9001 BR9002 BR9003 BR9004 BR9005 BR9013 BR9014 BR9015 BR9016 BR9017 BR9018

Sample type

Gleason score3+3 3+4 4+3 4+4 4+5 5+5

Invasive carcinoma IDCMixed

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18

Gain

Neutral

Loss

19 20 21 22 X Y

Figure 1 | The genomics of BRCA2-mutant prostate cancer. (a) Percent genome altered (PGA) and number of SNVs, GRs and CNAs per tumour sample is shown. GRs are stratied by type: orangedeletions (del), greeninversions (inv), purpleduplications (dup), yellowtranslocations (ctx). Y axis values for CNAs and GRs are in log10-scale. Box plots compare the number of events observed in BRCA2-mutant specimens versus sporadic PCa samples split into two groups by age. Y axis values are the same as in the adjacent bar plot. Whiskers indicate the maximum and minimum values, the box outline indicates the third and rst quartile and the bar indicates the mean. Multiple foci from the same patient are indicated in the same patient covariate colour, where sample type colour indicates IC in blue, IDC in yellow and a mix of IDC and IC in green. Unavailable data is indicated by grey background. (b) Per gene copy-number proles of 18 prostate tumour specimens from 14 germline BRCA2-mutation carriers, including 4 patients with multifocal disease. Red indicates gain; blue indicates loss. Rows represent specimens and columns represent genes. Top plots show frequency of each gene per group (BRCA2, n 14; sporadic PCa arising in individuals 50 years of age or younger, n 7; sporadic PCa arising in individuals older than 50 years of age, n 276) and the

q-value shown is from a two-sided proportion test comparing BRCA2-mutant and all sporadic PCa samples. Genes are ordered by genomic co-ordinates per chromosome. Multiple foci from the same patient are indicated in the same patient covariate colour, where sample type colour indicates IC in blue, IDC in yellow and a mix of IDC and IC in green. (c) Key prostate cancer driver genes that are mutated at elevated proportions in BRCA2-mutant PCa relative to sporadic PCa. Columns are patients, rows are genes, a black square indicates a CNA. Bar plots to the right show the frequency of each gene in the three groups (BRCA2, n 14; sporadic PCa arising in individuals 50 years of age or younger, n 7; sporadic PCa arising in individuals older than 50 years of age,

n 276).

NATURE COMMUNICATIONS | 8:13671 | DOI: 10.1038/ncomms13671 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications 3

ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms13671

a

10 105

10

BRCA2 50 yrs > 50 yrsQ

0

Variant

CTX DEL INV

Patient

BR0450 BR0556 BR9001 BR9002 BR9003 BR9004 BR9005

Sample type

Invasive carcinoma IDC

Gleason score 3+33+44+44+5

0.25

0.20

0

0.5

0

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 2122 X Y

b

chr5

Figure 2 | Genomic rearrangements in BRCA2-mutant prostate cancer. (a) Survey of genome-wide somatic genomic rearrangements (GRs) in BRCA2-mutant tumour PCa from seven patients. CTXinter-chromosomal translocation; DELdeletion; INVinversion. Multiple foci from the same patient are indicated in the same patient covariate colour, where sample type colour indicates microdissected invasive carcinoma (IC) in blue and intraductal carcinoma of the prostate (IDC) in yellow. Top bar plots show the frequency of specimens in BRCA2-mutant PCa (bottom) and sporadic PCa from men 50 years of age or younger (middle) or sporadic PCa from men older than 50 years of age (top). Scatterplot at top indicates FDR-corrected P values of a two-sided proportion test comparing frequency of GRs observed in BRCA2-mutant versus sporadic PCa. (b) A circos plot showing the proportions of BRCA2-mutant (blue) and sporadic PCa (red) with breakpoints in each mega-basepair bin. Bins with signicantly different proportions between BRCA2-mutant and sporadic PCa (FDR-correct proportions test) are highlighted in yellow, and have the number of BRCA2-mutant samples indicated. Only chromosomes with at least one differentially recurrent GR are shown.

4 NATURE COMMUNICATIONS | 8:13671 | DOI: 10.1038/ncomms13671 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications

NATURE COMMUNICATIONS | DOI: 10.1038/ncomms13671 ARTICLE

a

MED12L

MED12

b

BRCA2 statusBRCA2 Sporadic 50

Sporadic > 50

Gleason score3+3 3+4 4+3 4+4 4+5

Methylation status

Low High

Patient

BR0001 BR0450 BR0556 BR9001 BR9002 BR9003 BR9004 BR9005 BR9013 BR9014 BR9015 BR9016 BR9017 BR9018

Sample type IDC Invasive Ca

BCL6

MTOR

CDK2

ERBB3

TIMELESS

Mutation

CNA gain CNA loss

SNV

0 0.5 1

IDC IC

Frequency

Figure 3 | BRCA2-mutant PCa harbours multiple hallmarks of aggressive disease. (a) CNA, SNV and methylation proles for MED12L and MED12. Patients are sorted based on MED12L CNA, methylation and SNV status. b-values are median dichotomized, as reected by the background shading of each cell. The covariate bar at the top of the graph indicates BRCA2 status and Gleason score. (b) Multiple genes show differential rates of CNAs in tumours harbouring IDC and those not. Rows are genes, columns are samples. Red indicates copy-number gain; blue indicates copy-number loss; white indicates neutral copy-number status. The barplot to the right shows the differential frequency of mutation between IDC and IC components in BRCA2-mutant PCa.

Importantly, the MED12L homolog, MED12 (chromosome Xq13.1), is amplied in 44% of BRCA2-mutant PCa specimens (8/18) compared with only B0.1% of localized sporadic PCa (1/809, P 3.64 10 44, two-sided proportion test). Further,

MED12L (but not MED12) showed frequent amplication in distant metastases from metastatic castrate resistant prostate cancer (mCRPC) patients (7/37 cases, 19%). Supporting this model, four loci within the MED12L gene were signicantly hypomethylated in BRCA2-mutant versus sporadic PCa (Fig. 3a; Supplementary Fig. 15). Consistent with their alteration at the copy-number and methylation levels in aggressive BRCA2-mutant PCa, MED12/MED12L are also amplied and/or show elevated RNA expression levels in 18% (26/150) of mCRPC6 but only 7.5% (25/333) of sporadic, localized PCa18 (P 0.002,

proportion test). Thus MED12L/MED12 amplications are preferentially mutated in aggressive diseaseboth BRCA2-driven primary tumours and sporadic mCRPC.

MED12L/MED12 amplication was also enriched in BRCA2-mutant PCa harbouring IDC (6/8 IDC-positive specimens with MED12L/MED12 gain versus 1/6 IDC negative specimens). Importantly, these mutations are not enriched in sporadic tumours harbouring IDC. Therefore we evaluated global differences between IDC and IC. Presence of IDC in BRCA2-mutant PCa associates with CNAs portending poor prognosis. Several genomic regions were preferentially amplied or deleted in tumours harbouring IDC, including gain of BCL6 and loss of MTOR, along with gains of a region harbouring CDK2 and ERBB3 (Fig. 3b). This underpins the observation that the presence of IDC in BRCA2-mutant PCa associates with an aggressive clinical course and increased PCa-specic mortality rates5.

Evolutionary trajectories of BRCA2-mutant and sporadic PCa. Enrichment for negative prognostic factors in BRCA2-mutant PCa harbouring IDC suggested that these tumours may have distinct evolutionary trajectories. To assess the subclonal architectures of sporadic and BRCA2-mutant PCa harbouring IDC we selected six sporadic PCa and microdissected the IDC and IC

components (Supplementary Fig. 16). Both components were subjected to whole-genome sequencing and CNA proling and tumour phylogenies were reconstructed. This procedure was repeated for four BRCA2-mutant PCa cases that harboured both IDC and IC components. Reconstruction began with subclonal CNA proling using TITAN19, followed by clustering of SNV frequencies with PhyloWGS20. In all four BRCA2-mutant PCa (Fig. 4a; Supplementary Data 5) the IDC and IC components arose from the same founding clone, with no evidence of multiple independent tumours21. The parental population was found in both the IDC and IC regions, providing no inference as to where the tumour may have initiated. MYC amplications were observed in three of these four patients, and in each occurred before divergence of IDC and IC components. By contrast, MED12L gain was clonal in two samples and sub-clonal in two others, in one case only occurring in the IC component.

In sporadic PCa with evidence of IDC (Fig. 4b; Supplementary Data 6) there was again no evidence of multiple independent tumoursthe IDC and IC components arose from a common ancestor and there was no clear evidence as to which compartment this ancestor arose in. MYC amplication was observed in 4/6 tumours, once clonally, once in a clone only present in IDC and twice only in a clone present in IC. NKX3-1 deletion was present in the trunk of 3/6 tumours. Intriguingly, PIK3CA SNVs were observed in 2/6 patients with IDC (once clonally and once only in the IC component), as compared with 7/333 patients in the TCGA localized PCa dataset (P 5.92 10 4; proportion test)9.

DiscussionWhile the clinical aggressiveness of germline BRCA2-mutant PCa remains unexplained, it is distinguished from sporadic PCa by a unique mutational prole. BRCA2-mutant PCa is characterized by elevated global genomic instability, which is a well-established biomarker of biochemical failure and metastasis14,15. BRCA2-mutant PCa harboured specic mutations that are rarely, if ever, observed in sporadic localized PCa. These lead to dysregulation of

NATURE COMMUNICATIONS | 8:13671 | DOI: 10.1038/ncomms13671 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications 5

ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms13671

a b

BR9001

BR9002

CPCG0100

CPCG0196

CPCG0260

TP53 PTEN CDH1

A

MED12L MYC

A MYCN

C

MKI67 SMAD2 MYC MYCN MED12L PTEN NKX3-1 CDH1

A

APOBEC3B

A

BRCA1 TP53 PTEN RB1 NKX3-1

A

B

C

D

MED12L

B PIK3R1

D

BRCA2 PTEN RB1

TP53 PIK3CA MED12L MYC

Sample type

Variant

B

C

Normal

Point mutations CNA gainCNA loss

D C

Invasive carcinoma IDC

C

BRCA2 TP53 PTEN RB1 NKX3-1

PTEN

CPCG0364

BR9003

BR9005

CPCG0361

MYC

NKX3-1

BRCA2 TP53 PTEN

RB1 NKX3-1

A

A

G

NOTCH2 RB1 NKX3-1

MYC

PIK3CA CTNNB1

A

MED12L MYC TP53 MTOR

MED12L MYCN PTEN

NKX3-2

B

VHL MYC BRCA1

D

D

E

B

C

C

B

C

D

E

Kataegis

C

E

D

F

PTEN

c

BRCA2-mutant PCa

Sporadic PCa

Figure 4 | Evolutionary trajectory of BRCA2-mutant and sporadic PCa harbouring IDC. (a) Subclonal reconstruction of four BRCA2-mutant PCa from microdissected IDC and IC components using CNA and SNV data. (b) Reconstructions of six sporadic PCa with microdissected IDC and IC components using CNA and SNV data. Each tree gives the inferred phylogeny for a single patient. The gray node is the germline. Blue nodes are present in the IC specimen, gold nodes in the IDC specimen, and the relative proportions of each reect their relative cellular prevalence. Node size reects the sum of the cellular prevalence of the IDC and IC components. A set of key mutations that occur on each branch are annotated, and the length of the branch is proportional to the number of SNVs that accumulated on it. (c) Early in the development of BRCA2-Mutant PCa, the disease is characterized by amplications in 3q, MED12, MED12L and MYC, global genomic instability, a global hypomethylation prole and presents as a more aggressive disease. Tumour suppressors appear to be lost early in sporadic PCa, leading to subsequent development of global genome instability, and increased amplications as the disease advances.

pathways associated with NED and mCRPC, despite no exposure to androgen deprivation therapy in these patients. These observations help rationalize our recent clinical association between disease outcome and the presence of IDC5, in part, with increased activation of the MED12L/MED12 axis. Surprisingly, and counter to previous expectations, the IDC and IC components share a common precursor, and share a large portion of their mutational prole.

MED12L/MED12 are components of the mediator complex and important regulators of the WNT pathway, which has been implicated in cell proliferation and neuroendocrine differentiation in PCa2225. The MED12L/MED12 axis is implicated in metastatic sporadic PCa, with MED12 nuclear overexpression common in mCRPC and localized recurrent PCa but not in

localized primary PCa or benign prostate tissue26. MED12 SNVs, which occur in both localized and advanced PCa, are also implicated in prostate tumourigenesis27 and genome instability28,29.

BRCA2-mutant PCa are also enriched for aberrant DNA methylation in genes involved in neurogenesis and neuron development, consistent with recent reports that autonomic nerve development contributes to PCa progression and metastasis30. Overall, our data strongly support a model in which germline BRCA2-mutant PCa are genotypically-similar to mCRPC and frequently harbour genomic aberrations in key pathways (for example, WNT/APC, IL6-BMP6, mTOR, ATR, MYCN, and so on; Supplementary Data 6; Fig. 4c) leading to genomic instability, NED, castrate-resistance and metastasis, together explaining

6 NATURE COMMUNICATIONS | 8:13671 | DOI: 10.1038/ncomms13671 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications

NATURE COMMUNICATIONS | DOI: 10.1038/ncomms13671 ARTICLE

adverse patient prognosis. It remains to be shown if the temporal change in genomic stability of BRCA2-mutation carrier PCa tumours is relevant to other solid tumours with specic predispositions for aggressive disease such as those arising in breast, colon and pancreas.

Collectively, our data provide genome-wide, nucleotide-resolution characterization of the molecular architecture of localized PCa and IDC that develops in men who harbour a mutant germline BRCA2 allele. While the population frequency is relatively low (o1%)31, the aggressive natural history of localized tumours may justify a change in clinical management to a more intensied approach including upfront use of PARP inhibitors10 and chemotherapy (for example, success of CHAARTED, STAMPEDE trials)32 in addition to ADT to block the rapid progression to mCRPC observed in these patients. Based on our provocative data, trials could also explore molecular inhibition of ATR, mTOR and GSK3A pathways in these patients, especially when IDC is present. Our results may ultimately help to elucidate multiple key pathways to therapeutically target men who are carriers of a BRCA2 mutation with PCa to increase rates of cure by preventing rapid progression to mCRPC.

Methods

Patient cohorts. Men were treated for sporadic or BRCA2-mutant PCa by radical prostatectomy at Princess Margaret Cancer Centre (University Health Network, Toronto, Canada) or were recruited by kConfab (Kathleen Cuningham Foundation Consortium for research into Familial Breast cancer) at the Peter MacCallum Cancer Centre (Melbourne, Australia). Informed consent for analysis was obtained through institutional Research Ethics Board protocols (UHN #11-0024 and #11-0091-T; #97/27).

In Canada, patients with a known BRCA2 germline mutation were approached for retrospective informed consent based on the following inclusion criteria:(1) intermediate or high risk localized Prostate cancer, as dened by NCCN guidelines, treated in the Princess Margaret Cancer Centre BRCA clinic;(2) conrmed presence of primary tumour tissue in the UHN genitourinary biobank; (3) whole blood (5 ml) was obtained at the time of consenting. Consent was for use of tumour and normal tissues and associated DNA, RNA and protein, including whole-genome and whole-transcriptome sequencing of both somatic and germline specimens.

In Australia, prostate cancer cases were identied from 1,600 families recruited into the Kathleen Cunningham Consortium for Research into Familial Breast Cancer (kConFab), the Australian and New Zealand consortium for families at high-risk of breast cancer (http://www.kconfab.org

Web End =http://www.kconfab.org). Inclusion in kConFab requires a strong multi-case family history of breast and/or ovarian cancer, or a germline mutation in a breast cancer predisposition gene such as BRCA1and BRCA2(see http://www.kconfab.org

Web End =www.kconfab.org for full recruitment criteria). Male kConFab members were eligible for inclusion in this study if: (i) they had a veried diagnosis of prostate cancer, (ii) complete diagnostic and treatment notes were available and (iii) family BRCA mutation status was known. Human Research Ethics Committee approval (protocol: 97/27) was obtained from Peter MacCallum Cancer Centre to allow recruitment of prostate cancer patients at the time of diagnosis for fresh tissue collection or access to retrospectively archived formalin xed parafn embedded tissues. A comparative cohort of patients with sporadic prostate cancer was recruited for pathology review; these men were spouses of the multi-case breast cancer women recruited into kConFab.

Tissue collection and microdissection. Hematoxylin and Eosin-stained sections from fresh-frozen or formalin-xed parafn-embedded (FFPE) prostate cancer tissues were assessed by a clinical genitourinary pathologist (TvdK or DC). Regions of at least 70% tumour cellularity were manually microdissected from fresh-frozen or FFPE tissues from BRCA2 or sporadic tissues.

Copy-number proling. Genome-wide gene CNA proles were generated with Affymetrix OncoScan v3 SNP microarrays. Genomic DNA was extractedfrom microdissected tumour specimens using phenol:chloroform, as previously described23), and quantied using a Qubit Fluorometer. Microarray analysis used 30150 ng of dsDNA for each sample. Analysis of Affymetrix OncoScan FFPE Express 3.0 SNP probe assays was performed using. OSCHP les generated by OncoScan Console 1.1 using a custom reference when starting material was Fresh Frozen (n 12). A custom reference which included 119 normal blood samples

from male patients with prostate cancer, 2 normal blood samples from females with anaplastic thyroid cancer and 10 female HapMap cell line samples was created to combat artefacts created due to the difference in sample preparation (FFPE versus Fresh Frozen).

For FFPE samples,. OSCHP les generated by Affymetrix using their in-house reference (n 8).

BioDiscoverys Nexus Express for OncoScan 3 Software were used to call CNAs with the SNP-FASST2 algorithm. Default parameters were used, except that the minimum number of probes per segment was changed from 3 to 10. When necessary, samples were re-centered using the Nexus Express software, choosing regions that showed diploid log2ratio and B allele frequency proles (n 2). Gene

level copy-number aberrations for each patient were identied by overlapping CN segments with RefGene annotation using BEDTools (v2.17.0)33. Tumour purity was assessed computationally (Supplementary Fig. 15). PGA was calculated for each sample by dividing the number of base-pairs that are involved in a copy-number change by the total length of the genome.

To discover differences in copy-number states between germline BRCA2 mutated samples (n 14) and sporadic samples (n 284) two-sided proportion

tests were performed as implemented in R (v3.1.3). A by bvote prole was created for patients with both IDC and IC samples. When the copy-number state per gene differed between samples, gain and deletions were kept over neutral states. Copy-number segment data was mapped to the RefGene annotation, classifying each genes state as amplication, deletion or neutral and then amplications and deletions were grouped together for this analysis. P values were FDR adjusted to account for multiple testing. Two-sided, unpaired t-tests were performed to test differences in PGA between BRCA2 and sporadic samples implemented inR (v3.1.3).

The copy-number states of specic prognostic genes were investigated to determine if the frequency differed between samples that had an IDC component and those that did not. Samples that had both IDC and IC specimens were combined to get a complete prole, in which the union of all aberrations was taken. Only CNAs that are in the same state (amplication versus deletion) as known prognostic genes are shown. For example if a specimen has amplication of NKX3-1, it will not be shown, as only NKX3-1 deletions have prognostic value.

A Wilcoxon test was performed comparing the rank sum of CNAs in three prognostic genes (that is, NXK3-1, MYC, PTEN) between patients with IDC and those without.

Whole-genome sequencing. Whole-genome sequencing was performed on Illumina HiSeq 2,500 instruments at the Ontario Institute for Cancer Research. Sequencing libraries were prepared on 150 ng of input genomic DNA, as previously reported21. Tumour and normal DNA were sequenced to target depths of50 and 30 , respectively (actual sequencing depths shown in Supplementary

Table 1).

Whole-genome sequencing data analysis. Each lane of raw sequencing reads was aligned against human reference build GRCh37 with decoy (hs37d5) using bwa (v0.7.11)34. Lane level BAMs from the same library were merged, marking duplicates using picard (v1.92). Library level BAMs from each sample were merged without marking duplicates. The Genome Analysis Toolkit (GATK v2.4.9) was used for local realignment and base quality recalibration, processing tumour/ normal pairs together35. Separate tumour and normal sample level BAMs were extracted, headers were corrected using samtools (v0.1.9)36 and les were indexed with picard (v1.107). To evaluate cross-individual contamination, ContEst (v1.0.24530) was used for all the normal and tumour sequences37. Both the sample and lane-level analysis were performed and conrmed no signicant contamination. Regarding the required input VCFs, genotype information was gained from the germline SNVs generated by GATK (v2.4.9) and the VCF for population allele frequencies for each SNP in HapMap (hg19) was downloaded from https://www.broadinstitute.org/cancer/cga/contest_download

Web End =https://www.broadinstitute.org/cancer/cga/contest_download .

Somatic SNVs were predicted using SomaticSniper (v1.0.2)38. First, somatic SNV candidates were detected using bam-somaticsniper with the default parameters except -q option (mapping quality threshold). The -q was set to 1 instead of 0 as recommended by the developer. To lter the candidate SNVs,a pileup indel le was generated for both normal BAM and tumour BAM le using SAMtools (v0.1.6). The SomaticSniper package provides a series of Perl scripts to lter out possible false positives (http://gmt.genome.wustl.edu/somatic-sniper/1.0.3/documentation.html

Web End =http://gmt.genome.wustl.edu/somatic-sniper/ http://gmt.genome.wustl.edu/somatic-sniper/1.0.3/documentation.html

Web End =1.0.3/documentation.html ). First, standard and LOH ltering were performed using the pileup indel les and then, bam-readcount lter was also performed (bam-readcount downloaded on 10 January 2014) with a mapping quality lter -q 1 (otherwise default settings). In addition, we ran the false positive lter. Subsequently, a high condence lter was used with the default parameters.

The nal VCF le that contains high condence somatic SNVs was used in the downstream analysis.

After somatic SNV calling using SomaticSniper, identied SNVs were passed through an annotation pipeline. SNVs were functionally annotated by ANNOVAR (v2015-06-17)39, using the RefGene database (downloaded on 2014-07-15). Nonsynonymous, stop-loss, stop-gain and splice-site SNVs (based on RefGene annotations) were considered functional. If more than one mutation is found in a sample for a gene, then the mutation of the higher priority functional class was used for visualization. SNVs were ltered using tabixpp (3b299cc0911debadc435fdae60bbb72bd10f6d84), removing SNVs found in any of the following databases: dbSNP141 (modied to remove somatic and clinical variants, with variants with the following ags excluded: SAO 2/3, PM, CDA,

NATURE COMMUNICATIONS | 8:13671 | DOI: 10.1038/ncomms13671 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications 7

ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms13671

TPA, MUT and OM)40, 1,000 Genomes Project (v3), Complete Genomics 69 whole genomes, duplicate gene database (v68)41, ENCODE DAC and Duke Mapability Consensus Excludable databases (comprising poorly mapping reads, repeat regions, and mitochondrial and ribosomal DNA)42, invalidated somatic SNVs from 68 human colorectal cancer exomes (unpublished data) using the AccuSNP platform (Roche NimbleGen), germline SNVs from 477 sporadic PCa patients with the intermediate GS (3 X and X 3) and additional 10 prostate cancer patients with

higher GS13, and the Fuentes database of likely false positive variants43. SNVs were whitelisted (and retained, independently of the presence in other lters) if they were contained within the Catalogue of Somatic Mutations in Cancer (COSMIC) database (v70)44.

SNV mutation load. To nd the total number of bases somatic SNVs can be called on, each position in the reference was classied as callable or not. Positions were considered callable in a sample if it has a minimum coverage of 10 in normal

and 17 in the tumour as calculated using BEDTools (v2.18.2)33. Mutation load

for each sample was then calculated as the number of somatic SNVs divided by the total number of callable bases and converted to per mega-basepair of sequence. Mutation load was not calculated for samples missing a normal specimen to avoid upwards bias on estimated mutation rates.

Genomic rearrangements. For each patient, aligned tumour and normal BAM les were used to call structural variants with Delly (v0.5.5)45 at a minimum median mapping quality of 20 and a paired-end cut-off of ve. A list of somatic variants were produced by removing germline mutations from the resulting VCF les, which were further ltered using a consolidated list of structural variants from 124 normal samples. To identify genes affected by the genomic rearrangements, bed les were generated for each sample from deleted regions, and breakpoints from inversions, inter-chromosomal translocations, and tandem duplications. The resultant bed les were examined with SnpEff (v3.5)46 and gene names were subsequently extracted for downstream analyses. Recurrent translocation events were visualized using Circos (v0.67-4)47. Input les were bed les containing paired translocation breakpoints and the number of samples the event was observed. Differentially aberrant regions between sporadic and BRCA2 patients were identied by rst dividing the genome into mega-basepair bins, and tabulating the number of samples with a deletion, translocation or inversion breakpoint in each bin. Subsequently, a difference in proportions test was performed for each bin and the resultant P values were corrected for multiple comparisons with FDR. Entire chromosomes containing mega-basepair bins with signicantly different proportions of breakpoints were visualized using Circos (v0.67-4)47.

Chromothripsis detection. ShatterProof48 was used with default settings to examine the presence of chromothripsis through the analysis of copy-number aberrations, loss of heterozygosity, and genomic translocations, and duplications. Samples with a maximum ShatterProof score of at least 0.517 were dened as having chromothriptic characteristics.

Subclonal reconstruction. For each of the four BRCA2-mutant PCa and six sporadic PCa subjects with microdissected IDC and IC components, phylogenetic trees were constructed using the predicted SNVs and CNAs inferred by TITAN19. TITAN was run, from a range of one to ve clonal clusters, on each sample and the CNA segments associated with the run having lowest S_Dbw validity index was used for further analysis. Using the predicted SNVs, a multi-sample PhyloWGS20 run was conducted on each subject to cluster the SNV frequencies. PhyloWGS Witness was used to view and determine the consensus phylogenetic tree from the many sampled trees, summarizing the co-evolution of IDC and IC (Fig. 4; Supplementary Data 5 and 6) from the normal sample (gray). For each tree, blue denotes IC while gold denotes IDC, and the relative proportions of each reect their relative cellular prevalence. Node size reects the sum of the cellular prevalence of the IDC and IC components. Length of the branch is proportional to the number of SNVs that accumulated on it, while its width is scaled to the cellular prevalence of the new arising node. Validated pipelines identied recurrent mutations across the samples and using the gene list associated with the top PhyloWGS tree, key SNV mutations (green) were annotated along the branches. With OncoScan runs as a reference, TITANs subclonal CNAs and their inferred cellular prevalence identied a set of key driver genes, both gains (red) and losses (blue) along these branches, whereas parsimony was assumed.

Methylation microarray data analysis. Illumina Innium HumanMethylation 450 k BeadChip kits were used to assess global methylation, using 500 ng of input genomic DNA at the McGill University and Genome Quebec Innovation Centre (Montreal, QC). This study comprises 10 BRCA2-mutant and 104 sporadic samples from 8 processing batches. All methylation analysis was performed in R statistical environment (v3.2.1). The IDAT les were loaded and converted to raw intensity with the use of the wateRmelon package (v1.6.0)49 from the BioConductor (v3.0) open-source project. Quality control was conducted using the min package (v1.12.0)50 (no outlier samples were detected). Batch effects were also examined across 5 batches (BRCA2 versus sporadic samples) using the mclust package

(v5.0.2) and no batch effects were found (Adjusted Rand Index 0.070). Raw

methylation intensity levels were then pre-processed using Dasen. Probe ltering was conducted after the normalization. For each probe, a detection p-value is computed to indicate whether the signal for the corresponding genomic position is distinguishable from the background noise. Probes having 1% of samples with a detection p-value greater than 0.05 were removed (2,816 probes). We also ltered probes based on SNPs (65 probes) and non-CpG methylation probes (2,944 probes). Next, we used the DMRcate package (v1.2.0)51 to further lter out 27,296 probes that are known to cross-hybridize to multiple locations in the genome and 17,079 probes that contain a SNP with an annotated minor allele frequency of greater than 5% with a maximum distance of two nucleotides to the nearest CpG site. Average intensity levels were taken for technical replicates. Probes showing differential methylations between BRCA and sporadic samples (age450 years)

were identied using the DMRcate package. Probes were deemed to be signicantly differentially methylated based on qo0.05, |log2 FC| of b-values40.1, and CNA status. The fraction of patients having a CNA event (whether it is a gain or loss) is calculated and only probes with fraction CNAo0.1 were kept. Next, probes were annotated to their respective gene symbols via IlluminaHumanMethylation450kanno. ilmn12.hg19 (v0.2.1) annotation package, resulting in 2,624 unique genes; this gene list showing differential methylations were analysed using g:GOSt in the g:proler web database (build: 2014-04-11)52. Signicantly enriched pathways (Po0.05;

hypergeometric test), as an output from g:proler52, were then visualized using Cytoscape (v3.3.0)53. Tumour purities for all the samples were assessed with LUMP54.

Public data analysis. To assess the occurrence of MED12/MED12L amplications in sporadic and mCRPC relative to BRCA2-mutant cancers, we downloaded publically available gene copy-number data from TCGA9 (n 203), Baca55

(n 57), Barbieri56 (n 109) and Taylor8 (sporadic, n 157; mCRPC, n 37).

To assess the relationship between RNA expression and MED12L/MED12 amplication, we queried cBioPortal for the presence of RNA over-expression (z-score42.0) and amplication of one or both of these genes in two studies for which both RNA-seq and CNA data were available: Robinson et al.7 (mCRPC) and TCGA9 (localized sporadic PCa). A case was counted as positive if either RNA was overexpressed or the gene was amplied.

Data availability. Raw sequencing and SNP microarray data from this study is available at EGA under the study accession EGAS00001001615. Methylation data from this study is available at GEO under accession GSE80685. Methylation data accessed from GEO with the accession code GSE84043 was used to support the ndings of this study. Sequencing data accessed from EGA with the accession codes EGAS000010000258 and EGAS000010000900 and from dbGaP with the accession codes phs000447.v1.p1 and phs000330.v1.p1 was used to support the ndings of this study. A subset of the TCGA-PRAD dataset was also downloaded from CGHub (data now stored at the Genomic Data Commons). The authors declare that all remaining data are contained within the Article and its Supplementary Information les or available from the author upon request.

References

1. Castro, E. et al. Effect of BRCA mutations on metastatic relapse and cause-specic survival after radical treatment for localised prostate cancer. Eur. Urol. 68, 186193 (2015).

2. Willems, A. J. et al. Loss of heterozygosity at the BRCA2 locus detected by multiplex ligation-dependent probe amplication is common in prostate cancers from men with a germline BRCA2 mutation. Clin. Cancer Res. 14, 29532961 (2008).

3. Akbari, M. R. et al. The impact of a BRCA2 mutation on mortality from screen-detected prostate cancer. Br. J. Cancer 111, 12381240 (2014).

4. Castro, E. et al. Germline BRCA mutations are associated with higher risk of nodal involvement, distant metastasis, and poor survival outcomes in prostate cancer. J. Clin. Oncol. 31, 17481757 (2013).

5. Risbridger, G. P. et al. Patient-derived xenografts reveal that intraductal carcinoma of the prostate is a prominent pathology in BRCA2 mutation carriers with prostate cancer and correlates with poor prognosis. Eur. Urol. 67, 496503 (2015).

6. Liede, A., Karlan, B. Y. & Narod, S. A. Cancer risks for male carriers of germline mutations in BRCA1 or BRCA2: a review of the literature. J. Clin. Oncol. 22, 735742 (2004).

7. Robinson, D. et al. Integrative clinical genomics of advanced prostate cancer. Cell 161, 12151228 (2015).

8. Taylor, B. S. et al. Integrative genomic proling of human prostate cancer. Cancer Cell 18, 1122 (2010).

9. The Cancer Genome Atlas Network. The molecular taxonomy of primary prostate cancer. Cell 163, 10111025 (2015).

10. Mateo, J. et al. DNA-repair defects and olaparib in metastatic prostate cancer.N. Engl. J. Med. 373, 16971708 (2015).

8 NATURE COMMUNICATIONS | 8:13671 | DOI: 10.1038/ncomms13671 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications

NATURE COMMUNICATIONS | DOI: 10.1038/ncomms13671 ARTICLE

11. Sandhu, S. K. et al. Poly (ADP-ribose) polymerase (PARP) inhibitors for the treatment of advanced germline BRCA2 mutant prostate cancer. Ann. Oncol. 24, 14161418 (2013).

12. Sandhu, S. K. et al. The poly(ADP-ribose) polymerase inhibitor niraparib (MK4827) in BRCA mutation carriers and patients with sporadic cancer: a phase 1 dose-escalation trial. Lancet Oncol. 14, 882892 (2013).

13. Fraser, M. et al. Genomic analyses of localized non-indolent prostate cancer. Nature. in the press (2016).

14. Lalonde, E. et al. Tumour genomic and microenvironmental heterogeneity for integrated prediction of 5-year biochemical recurrence of prostate cancer: a retrospective cohort study. Lancet Oncol. 15, 15211532 (2014).

15. Hieronymus, H. et al. Copy number alteration burden predicts prostate cancer relapse. Proc. Natl Acad. Sci. USA 111, 1113911144 (2014).

16. Ciriello, G. et al. Emerging landscape of oncogenic signatures across human cancers. Nat. Genet. 45, 11271133 (2013).

17. Castro, E. et al. High burden of copy number alterations and c-MYC amplication in prostate cancer from BRCA2 germline mutation carriers. Ann. Oncol. 26, 22932300 (2015).

18. Ippolito, J. E. et al. Linkage between cellular communications, energy utilization, and proliferation in metastatic neuroendocrine cancers. Proc. Natl Acad. Sci. USA 103, 1250512510 (2006).

19. Ha, G. et al. TITAN: inference of copy number architectures in clonal cell populations from tumor whole-genome sequence data. Genome Res. 24, 18811893 (2014).

20. Deshwar, A. G. et al. PhyloWGS: Reconstructing subclonal composition and evolution from whole-genome sequencing of tumors. Genome Biol. 16, 35 (2015).

21. Boutros, P. C. et al. Spatial genomic heterogeneity within localized, multifocal prostate cancer. Nat. Genet. 47, 736745 (2015).

22. Uysal-Onganer, P. et al. Wnt-11 promotes neuroendocrine-like differentiation, survival and migration of prostate cancer cells. Mol. Cancer 9, 55 (2010).

23. Yu, X., Wang, Y., DeGraff, D. J., Wills, M. L. & Matusik, R. J. Wnt/beta-catenin activation promotes prostate tumor progression in a mouse model. Oncogene 30, 18681879 (2011).

24. Yang, X. et al. A human- and male-specic protocadherin that acts through the wnt signaling pathway to induce neuroendocrine transdifferentiation of prostate cancer cells. Cancer. Res. 65, 52635271 (2005).

25. Grigore, A. D., Ben-Jacob, E. & Farach-Carson, M. C. Prostate cancer and neuroendocrine differentiation: more neuronal, less endocrine? Front. Oncol. 5, 37 (2015).

26. Shaikhibrahim, Z. et al. MED12 overexpression is a frequent event in castration-resistant prostate cancer. Endocr. Relat. Cancer 21, 663675 (2014).

27. Kampjarvi, K. et al. Somatic MED12 mutations in prostate cancer and uterine leiomyomas promote tumorigenesis through distinct mechanisms. Prostate 76, 2231 (2016).

28. Yoshida, M. et al. Frequent MED12 mutations in phyllodes tumours of the breast. Br. J. Cancer 112, 17031708 (2015).

29. Mittal, P. et al. Med12 gain-of-function mutation causes leiomyomas and genomic instability. J. Clin. Invest. 125, 32803284 (2015).

30. Magnon, C. et al. Autonomic nerve development contributes to prostate cancer progression. Science 341, 1236361 (2013).

31. Risch, H. A. et al. Population BRCA1 and BRCA2 mutation frequencies and cancer penetrances: a kin-cohort study in Ontario, Canada. J. Natl Cancer. Inst. 98, 16941706 (2006).

32. Gillessen, S. et al. Management of patients with advanced prostate cancer: recommendations of the St Gallen Advanced Prostate Cancer Consensus Conference (APCCC) 2015. Ann. Oncol. 26, 15891604 (2015).

33. Quinlan, A. R. & Hall, I. M. BEDTools: a exible suite of utilities for comparing genomic features. Bioinformatics 26, 841842 (2010).

34. Li, H. & Durbin, R. Fast and accurate long-read alignment with Burrows-Wheeler transform. Bioinformatics 26, 589595 (2010).

35. McKenna, A. et al. The Genome Analysis Toolkit: a MapReduce framework for analyzing next-generation DNA sequencing data. Genome Res. 20, 12971303 (2010).

36. Li, H. et al. The sequence alignment/map format and SAMtools. Bioinformatics 25, 20782079 (2009).

37. Cibulskis, K. et al. ContEst: estimating cross-contamination of human samples in next-generation sequencing data. Bioinformatics 27, 26012602 (2011).38. Larson, D. E. et al. SomaticSniper: identication of somatic point mutations in whole genome sequencing data. Bioinformatics 28, 311317 (2012).

39. Wang, K., Li, M. & Hakonarson, H. ANNOVAR: functional annotation of genetic variants from high-throughput sequencing data. Nucleic Acids Res. 38, e164 (2010).

40. Sherry, S. T. et al. dbSNP: the NCBI database of genetic variation. Nucleic Acids Res. 29, 308311 (2001).

41. Ouedraogo, M. et al. The duplicated genes database: identication and functional annotation of co-localised duplicated genes across genomes. PLoS ONE 7, e50653 (2012).

42. Gerstein, M. B. et al. Architecture of the human regulatory network derived from ENCODE data. Nature 489, 91100 (2012).

43. Fuentes Fajardo, K. V. et al. Detecting false-positive signals in exome sequencing. Hum. Mutat. 33, 609613 (2012).

44. Forbes, S. A. et al. COSMIC: exploring the worlds knowledge of somatic mutations in human cancer. Nucleic Acids Res. 43, D805D811 (2015).

45. Rausch, T. et al. DELLY: structural variant discovery by integrated paired-end and split-read analysis. Bioinformatics 28, i333i339 (2012).

46. Cingolani, P. et al. A program for annotating and predicting the effects of single nucleotide polymorphisms, SnpEff: SNPs in the genome of Drosophila melanogaster strain w1118; iso-2; iso-3. Fly 6, 8092 (2012).

47. Krzywinski, M. et al. Circos: an information aesthetic for comparative genomics. Genome Res. 19, 16391645 (2009).

48. Govind, S. K. et al. ShatterProof: operational detection and quantication of chromothripsis. BMC Bioinformatics 15, 78 (2014).

49. Pidsley, R. et al. A data-driven approach to preprocessing Illumina 450 K methylation array data. BMC Genomics 14, 293 (2013).

50. Aryee, M. J. et al. Min: a exible and comprehensive Bioconductor package for the analysis of Innium DNA methylation microarrays. Bioinformatics 30, 13631369 (2014).

51. Peters, T. J. et al. De novo identication of differentially methylated regions in the human genome. Epigenetics Chromatin 8, 6 (2015).

52. Reimand, J. et al. g:Proler-a web server for functional interpretation of gene lists (2016 update). Nucleic Acids Res. 44, 8389 (2016).

53. Shannon, P. et al. Cytoscape: a software environment for integrated models of biomolecular interaction networks. Genome Res. 13, 24982504 (2003).

54. Aran, D., Sirota, M. & Butte, A. J. Systematic pan-cancer analysis of tumour purity. Nature Commun. 6, 8971 (2015).

55. Baca, S. C. et al. Punctuated evolution of prostate cancer genomes. Cell 153,

666677 (2013).

56. Barbieri, C. E. et al. Exome sequencing identies recurrent SPOP, FOXA1 and MED12 mutations in prostate cancer. Nat. Genet. 44, 685689 (2012).

Acknowledgements

The authors thank all members of the Boutros, Risbridger and Bristow labs for helpful suggestions, and Eveline Niedermayr, all the kConFab research nurses and staff, the heads and staff of the Family Cancer Clinics, and the many families who contribute to kConFab. The results described here are in part based upon data generated by the TCGA Research Network: http://cancergenome.nih.gov/

Web End =http://cancergenome.nih.gov/ . This study was conducted with the support of Movember funds through Prostate Cancer Canada and with the additional support of the Ontario Institute for Cancer Research, funded by the Government of Ontario. This study was conducted with the support of the Ontario Institute for Cancer Research to PCB through funding provided by the Government of Ontario. This work was supported by Prostate Cancer Canada and is proudly funded by the Movember FoundationGrant #RS2014-01 to PCB. PCB was supported by a Terry Fox Research Institute New Investigator Award and a CIHR New Investigator Award. This project was supported by Genome Canada through a Large-Scale Applied Project contract to PCB and Drs. Ryan Morin and Sohrab P. Shah. The authors gratefully thank the Princess Margaret Cancer Centre Foundation and Radiation Medicine Program Academic Enrichment Fund for support (to R.G.B.). R.G.B. is a recipient of a Canadian Cancer Society Research Scientist Award. This study was supported with funds from National Health and Medical Research Council of Australia (NHMRC ID APP1077799), Victorian State Government Cancer Agency Program Grant (Prostate Translational Research in Victoria Collaboration; CAPTIV) and TissuPath. R.A.T. was supported by Victorian Cancer Agency Fellowship. G.P.R. was supported by NHMRC Fellowships (IDs: 1002648 & 1102752). MGL was supported by a NHMRC Fellowship (1035721) and Movember Young Investigator Grant from the Prostate Cancer Foundation of Australia. kConFab is supported by a grant from the National Breast Cancer Foundation, and previously by the National Health and Medical Research Council, the Queensland Cancer Fund, the Cancer Councils of New South Wales, Victoria, Tasmania and South Australia and the Cancer Foundation of Western Australia.

Author contributions

Project initiation: R.A.T., M.F., H.T., P.C.B., G.P.R. and R.G.B.; identied and consented patients: M.Fra, A.S., M.P., D.G.M., M.Fry and D.B.; pathology analysis: D.C. and T.v.d.K.; genomic analyses: M.Fra, W.L., L.T., J.D.M. and P.C.B.; computational and statistical analyses: J.L., S.M.G.E., Y.-J.S., V.H., T.N.Y., L.E.H., N.Y., F.Y., J.F.H. and P.C.B.; supervised research: R.A.T., H.T., M.G.L., D.G.M., J.D.M., T.v.d.K., M.F., P.C.B., G.P.R. and R.G.B.; wrote rst draft of manuscript: R.A.T., M.F., H.T., P.C.B., G.P.R. and R.G.B.; approved the manuscript: all authors.

Additional information

Supplementary Information accompanies this paper at http://www.nature.com/naturecommunications

Web End =http://www.nature.com/ http://www.nature.com/naturecommunications

Web End =naturecommunications

Competing nancial interests: D.C. is an employee of TissuPath. The remaining authors declare no competing nancial interests.

NATURE COMMUNICATIONS | 8:13671 | DOI: 10.1038/ncomms13671 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications 9

ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms13671

Reprints and permission information is available online at http://npg.nature.com/reprintsandpermissions/

Web End =http://npg.nature.com/ http://npg.nature.com/reprintsandpermissions/

Web End =reprintsandpermissions/

How to cite this article: Taylor, R. A. et al. Germline BRCA2 mutations drive prostate cancers with distinct evolutionary trajectories. Nat. Commun. 8, 13671 doi: 10.1038/ncomms13671 (2017).

Publishers note: Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional afliations.

This work is licensed under a Creative Commons Attribution 4.0 International License. The images or other third party material in this article are included in the articles Creative Commons license, unless indicated otherwise in the credit line; if the material is not included under the Creative Commons license, users will need to obtain permission from the license holder to reproduce the material. To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/

Web End =http://creativecommons.org/licenses/by/4.0/

r The Author(s) 2017

10 NATURE COMMUNICATIONS | 8:13671 | DOI: 10.1038/ncomms13671 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications