Get 20M+ Full-Text Papers For Less Than $1.50/day. Subscribe now for You or Your Team.

Learn More →

High grade serous ovarian carcinomas originate in the fallopian tube

High grade serous ovarian carcinomas originate in the fallopian tube ARTICLE DOI: 10.1038/s41467-017-00962-1 OPEN High grade serous ovarian carcinomas originate in the fallopian tube 1,13 2,14 2 2,3 2 2 S. Intidhar Labidi-Galy , Eniko Papp , Dorothy Hallberg , Noushin Niknafs , Vilmos Adleff , Michael Noe , 2,4 1,14 5 2 2 6 Rohit Bhattacharya , Marian Novak , Siân Jones , Jillian Phallen , Carolyn A. Hruban , Michelle S. Hirsch , 6,15 7 1 8 6 Douglas I. Lin , Lauren Schwartz , Cecile L. Maire , Jean-Christophe Tille , Michaela Bowden , 9,10,11,12 2 2 2,12 2,12 Ayse Ayhan , Laura D. Wood , Robert B. Scharpf , Robert Kurman , Tian-Li Wang , 2,12 2,3 1,6,16 2 Ie-Ming Shih , Rachel Karchin , Ronny Drapkin & Victor E. Velculescu High-grade serous ovarian carcinoma (HGSOC) is the most frequent type of ovarian cancer and has a poor outcome. It has been proposed that fallopian tube cancers may be precursors of HGSOC but evolutionary evidence for this hypothesis has been limited. Here, we perform whole-exome sequence and copy number analyses of laser capture microdissected fallopian tube lesions (p53 signatures, serous tubal intraepithelial carcinomas (STICs), and fallopian tube carcinomas), ovarian cancers, and metastases from nine patients. The majority of tumor-specific alterations in ovarian cancers were present in STICs, including those affecting TP53, BRCA1, BRCA2 or PTEN. Evolutionary analyses reveal that p53 signatures and STICs are precursors of ovarian carcinoma and identify a window of 7 years between development of a STIC and initiation of ovarian carcinoma, with metastases following rapidly thereafter. Our results provide insights into the etiology of ovarian cancer and have implications for pre- vention, early detection and therapeutic intervention of this disease. 1 2 Department of Medical Oncology, Dana-Farber Cancer Institute and Harvard Medical School, Boston, MA 02215, USA. Sidney Kimmel Comprehensive Cancer Center, Johns Hopkins University School of Medicine, Baltimore, MD 21287, USA. Department of Biomedical Engineering, Institute for Computational Medicine, Johns Hopkins University, Baltimore, MD 21218, USA. Department of Computer Science, Institute for Computational Medicine, 5 6 Johns Hopkins University, Baltimore, MD 21218, USA. Personal Genome Diagnostics, Baltimore, MD 21224, USA. Department of Pathology, Brigham and Women’s hospital and Harvard Medical School, Boston, MA 02115, USA. Department of Pathology, University of Pennsylvania Perelman School of Medicine, Philadelphia, PA 19104, USA. Division of Clinical Pathology, Faculty of Medicine, Geneva University Hospital, 1205 Geneva, Switzerland. 9 10 Department of Pathology, Seirei Mikatahara Hospital, Hamamatsu 433-8558, Japan. Department of Tumor Pathology, Hamamatsu University School of Medicine, Hamamatsu 431-3192, Japan. Department of Molecular Pathology, Hiroshima University School of Medicine, Hiroshima 739-0046, Japan. Departments of Gynecology and Obstetrics and Pathology, Johns Hopkins University School of Medicine, Baltimore, MD 21287, USA. Present address: Department of Oncology, Geneva University Hospitals, Geneva 1205, Switzerland. Present address: Personal Genome Diagnostics, Baltimore, MD 21224, 15 16 USA. Present address: Department of Pathology, Beth Israel Deaconess Medical Center, Boston, MA 02215, USA. Present address: Department of Obstetrics and Gynecology, Penn Ovarian Cancer Research Center, University of Pennsylvania Perelman School of Medicine, Philadelphia, PA 19104, USA. S. Intidhar Labidi-Galy and Eniko Papp contributed equally to this work. Ronny Drapkin and Victor E. Velculescu jointly supervised this work. Correspondence and requests for materials should be addressed to R.D. (email: rdrapkin@pennmedicine.upenn.edu) or to V.E.V. (email: velculescu@jhmi.edu) NATURE COMMUNICATIONS 8: 1093 DOI: 10.1038/s41467-017-00962-1 www.nature.com/naturecommunications 1 | | | ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/s41467-017-00962-1 varian cancer is the leading cause of death from gyne- In addition, we analyzed isolated STIC lesions from four 1, 2 cologic cancers . The 10-year survival is < 30% and has patients (CGOV64, CGOV65, CGOV303, and CGOV304), three Onot improved significantly over the last 30 years . Despite of whom had germline pathogenic BRCA alterations and significant efforts, various screening and therapeutic strategies underwent prophylactic bilateral salpingo-oophorectomy, and a 4, 5 have generally not led to improved overall survival . One of the fourth who had bilateral salpingo-oophorectomy and hysterect- major challenges to improved diagnostic and therapeutic inter- omy in the context of a pelvic mass (Supplementary Data 1). For vention in ovarian cancer has been a limited understanding of the all patients, laser capture microdissection (LCM) was used to natural history of the disease. Ovarian carcinoma is a highly isolate lesions after immunohistochemistry (IHC) staining of p53 heterogeneous group of diseases including different histological in STICs and p53 signatures if these contained a TP53 missense subtypes with distinct clinicopathological and molecular genetic mutation or after hematoxylin staining if the samples contained a features that can be generally classified as Type I and Type II TP53 nonsense mutation (Fig. 1). All other samples were tumors . Among them, high-grade serous ovarian carcinoma microdissected after hematoxylin staining. Whole blood, normal (HGSOC, the major Type II tumor) is the most common histo- ovarian stroma, normal FT stroma, or normal cervix were used as logic subtype of ovarian cancer, accounting for three quarters of control samples. 7–10 ovarian carcinoma . Genomic analyses of HGSOC have To identify genetic alterations in the coding regions of these identified genetic alterations in TP53, BRCA1, BRCA2, PTEN, and cancers, we used next-generation sequencing platforms to other genes although few of these discoveries have affected clin- examine entire exomes in matched tumor and normal specimens 11, 12 ical care . HGSOC is diagnosed at advanced stages in ~70% of of all patients (Fig. 1). This approach allowed us to identify non- cases, and these women have a significantly worse outcome than synonymous and synonymous sequence changes, including those with early stage disease. Until recently, the prevailing view single base and small insertion or deletion mutations, as well of HGSOC was that it developed from the ovarian surface epi- as copy number alterations in coding genes. Given the thelium. However, early in situ lesions that arise from the ovarian challenges of exome-wide analyses of small tumor samples surface epithelium and progress to invasive HGSOC have never observed in STICs and p53 signature lesions, we developed been reproducibly identified. experimental and bioinformatic approaches for detection of Insights into the pathogenesis of HGSOC have emerged from somatic alterations from laser capture microdissected tissue. investigating the prevalence of occult ovarian and fallopian tube These included optimized approaches for microdissection of (FT) carcinomas in women with germline mutations of BRCA1/ STICs and p53 signatures after immunohistochemical staining, 13–17 BRCA2 genes . Potential precursor lesions of HGSOC were improved DNA recovery from laser captured material, library identified in the fimbriae of the FTs removed as part of pro- construction from limited and stained tissue samples, and error phylactic surgery . Such lesions, including a TP53 mutant single- correction methods in next-generation sequence analyses cell epithelial layer (p53 signature) and serous tubal intraepithelial (Methods section). The analyses of p53 signatures were 17, 18 carcinoma (STIC) , have been identified in patients with particularly challenging because these are extremely small advanced stage sporadic HGSOC of the ovary, FT and perito- lesions, representing 10–30 cells per section and less than neum . Immunohistochemical as well as targeted sequencing several hundred cells total that result in minute amounts (less analyses have shown that FT lesions harbor the same TP53 than a few ng) of isolated DNA. We optimized these approaches 17–21 mutation as surrounding invasive carcinomas . These ana- using a targeted next-generation sequencing approach analyzing lyses suggest a clonal relationship among such tumors but given 120 genes in a subset of samples from patient CGOV62, and the limited number of genes analyzed do not conclusively identify then used whole-exome analyses to evaluate coding sequence the initiating lesions nor exclude the possibility of FT metastases alterations in all samples (Supplementary Data 2–4). We 21, 22 from primary ovarian carcinomas . Yet additional studies obtained a total of 719 Gb of sequence data, resulting in an have evaluated clonal intraperitoneal spread of ovarian cancer average per-base sequence ~178-fold total coverage (~112-fold using whole genome analyses, but these efforts did not analyze distinct coverage) for each tumor analyzed by whole-exome precursor lesions such as STICs that may give rise to this sequencing (Supplementary Data 2). disease . In this study, we use exome-wide sequence and structural analyses of multiple tumor samples from the same individual to Analysis of sequence and structural changes. Whole-exome examine the origins of HGSOC. We have previously shown that sequence analyses of the tumor samples from each patient the acquisition of somatic alterations can be used as a molecular identified somatic mutations that were present in all neoplastic marker in the development of human cancer . Here, we examine samples analyzed as well as specific changes that were present in whether the compendium of somatic alterations identified in individual or subsets of tumors (Fig. 2). As expected, we identified different lesions may provide insights into the evolutionary sequence changes in the TP53 tumor suppressor gene, a well- relationship between primary FT lesions, including p53 signatures known driver gene in HGSOC, in all cases. The TP53 alterations and STIC lesions, ovarian carcinomas, and intraperitoneal were identical in all samples analyzed for each patient including metastases. in the p53 signatures, the STIC lesions, and other carcinomas. These data suggest that mutation of TP53 was among the earliest initiating events for HGSOC development as all lesions harbored Results this alteration. Overall approach. To elucidate the relationship among tumors in IHC staining for p53 did not identify any nuclear positive patients with HGSOC, we performed whole-exome sequencing of staining of p53 on the ovarian surface epithelium in any of the 37 samples from five patients diagnosed with sporadic HGSOC cases that had TP53 missense mutation, whereas all carcino- who underwent upfront debulking (Supplementary Data 1). This mas, STICs, and p53 signatures in the FT were positive. included STIC lesions, FT carcinomas, and ovarian cancers in all Whole-exome sequence analyses of normal ovarian stroma five patients; appendiceal, omental, or rectal metastases in three of (no p53 staining) microdissected from three patients patients (CGOV62, CGOV280, CGOV278); p53 signatures in two (CGOV64, CGOV65, CGOV280) did not find any genomic patients (CGOV62, CGOV63); and a STIC lesion in the con- abnormalities. Analysis of the resected tissues revealed tralateral FT from the affected ovarian cancer (CGOV280). that none of the nine cases had ovarian inclusion cysts. 2 NATURE COMMUNICATIONS 8: 1093 DOI: 10.1038/s41467-017-00962-1 www.nature.com/naturecommunications | | | NATURE COMMUNICATIONS | DOI: 10.1038/s41467-017-00962-1 ARTICLE p53 signature STIC Left FT tumor Omental metastasis Left ovarian tumor Appendiceal metastasis Left ovarian tumor Rectal metastasis Right ovarian tumor Identification of tumor cells Laser capture microdissection Infrared laser Attached Transfer captured pulse tumor cells tumor cells H&E p53 IHC p53 signature Whole-exome analyses from laser capture microdissected cells Evolutionary analysis Evolutionary SCHISM timeline Point Chromosomal mutations alterations Fig. 1 Schematic of sample isolation and next-generation sequencing analyses. (Top panel) Tumor sites analyzed from CGOV62 with stage III HGSOC. For each sample, slides were stained with hematoxylin and eosin as well as analyzed by immunohistochemical staining of p53. (Middle panel) Tumor samples were microdissected for genomic analyses. For microdissection for STIC and p53 signature lesions, tumor cells were identified using immunohistochemical staining of p53 and isolated through laser capture microdissection. (Bottom panel, left) Next-generation sequencing analyses were performed for tumor specimens using either whole-exome or targeted analyses. (Bottom panel, right) Somatic mutations and chromosomal alterations were used to evaluate tumor evolution using the tumor subclonality phylogenetic reconstruction algorithm SCHISM and to determine a timeline for tumor progression These observations suggest that there is no early lesion with analyzed sequence alterations in all samples with estimated tumor TP53 mutation in the surface epithelium or other normal purities > 50%, while four samples with tumor cellularities below regions within the ovary. this threshold (omental metastasis from CGOV279 and right Because TP53 mutations are expected to be clonal and were all ovarian tumor from CGOV278) or that were miliary carcinomas homozygous due to loss of heterozygosity (LOH) of the (rectal and sigmoidal metastases from CGOV63) were only remaining wild-type allele (as determined in our subsequent analyzed for structural changes. allelic imbalance analyses), we used the mutant allele fraction of Using a high-sensitivity mutation detection pipeline, we TP53 in each sample to estimate tumor purity. We further identified an average of 33 non-synonymous and synonymous NATURE COMMUNICATIONS 8: 1093 DOI: 10.1038/s41467-017-00962-1 www.nature.com/naturecommunications 3 | | | ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/s41467-017-00962-1 CGOV62 CGOV63 CGOV280 CGOV279 CGOV278 TP53:Y220C TP53:S90Lfs*59 TP53:Y126N TP53:chr17:7.6 CC>AG TP53:R273H AHDC1:N1218K ATP2B3:L606S KCNJ10:K335T ANKFN1:S172S ABCA13:H878Y BAHCC1:Q484E ATN1:Q485Q C12orf43:S223A ANKRD35: T908T ASB15:D503Y ATP13A3:A1175S COL6A5:K88E CAPZA3:N217I C5orf25:T635M C1orf173:Q897Q CREB3L1:E328D CD14:Q250Pfs*17 CLSPN:A1261S C5orf46:S37S CACNA1C:A1292D IGSF21:T251N CTNND2:G692R CSMD1:chr8:3 A>T KIAA1614:L83V ELL:L12L EPHA5:L1037F CCR2:E15X DGKBP12P NOL4:V28V EPPIN–WFDC6:Q43Q SI:chr3:166.2 T>C DNAH3:A3011T FAM135BT1242T OR8S1:R258C GPATCH1:R922W DUSP27:D880N SLC25A3:chr12:97.5 G>A HCK:G186S SLC13A3:W52X IGHMBP2:S508W FLG2:D64D SLC41A3:V71A TRIO:L111L KIF13A:T1034M KCNAB1:I46V TRPC7:E65X GPR158:D1002E KRT3:G133E WDR11:A371T LINGO4:V207A chr16:56.4–70.8 GPR25:L21L LMAN2:E235X ZNF71:E358D LRRC15:D125N chr16:77.8–90.1 MACF1:D242Y MACF1:A1296A INPP5E:S582S chr1:155.2–211.1 chr17:29.9–73 METAP2:chr12:95.9 A>T NID1:D405E ISM1:D350D chr10:1.2–129.8 chr17:9.8–21.1 MPZ:G213G RPS6KA5:A802P chr8:120.7–144.9 chr11:106.7–129.5 LPIN3:G623G NLRP5:G363V chr8:41.5–66.5 SEPT6:chrX:118.6 T>G chr11:82.3–102.7 LRP1:D2764D PTPRND793D GRIN2B:E47E SFMBT2:T599M RAD9A:A190A chr13:20.8–114 MS4A14:I56I OR10W1:A116A SILV:G100G SERINC2:A379A chr14:19.3–104.8 MTA3:V42L SV2C:C145C chr14:60.5–72.5 SMARCA4:V830V chr18:50.7–77.9 chr17:0.3–70.6 PCDHGB4:L408I chr14:91–101.9 chr2:98.3–109.1 TMPPE:S440C chr18:29.6–66.1 PDE2A:chr11:72.3 C>T chr16:17.1–65.5 chr1:1.2–16.4 chr3:0.4–13.4 PREX1:A283V chr11:67.1–134.2 chr20:31.1–54.6 chr16:70.6–87.4 chr8:75.9–95.5 SETBP1:L227L chr9:0.3–21 chr13:22.3–53.4 chr22:16.6–39.1 chr17:37.5–59.4 chr9:71.1–101 chr13:67.8–115.1 TMEM64:N378S chrX:44.9–154.2 chr4:154.7–177.3 chr2:135.2–153.5 chr15:39–101.9 chr4:178.5–189.3 TRHR:R328H APOB:K3630KN ARAP3:L1034H chr16:53.3–70.8 chr5:57.8–180.6 TRIML1:P433L BTNL9:P43PS C20orf194:L502L chr17:26.7–81.1 chr10:19–31.2 TTN:R25052H C1S:V225V DLG5:Q837Q chr17:0.6–19.7 chr10:42.9–93.7 XIRP2:L2419L EVC2:K450E chr18:21–34.9 C4orf43:K114N ACVR1C:G343E KCNQ2:N289N ZMIZ2:V393F chr18:43.3–77.9 C6:N343D LAMA5:R3419G C1orf174:A14E chr19:0.3–12.2 ZMYM6:P855P CDKAL1:T212T MB:G74G C2orf21:N97Y chr2:0.2–11.9 DNAH6:R1043H chr15:22.9–102.3 MCL1:E168Q chr21:11.1–48 C9orf152:Q107Q SP9:G170G DUSP27:V100I chr17:0.1–20.9 chr22:17.3–48.9 CA6:G102R VPS16:L162L chr17:26.7–72.9 chr3:29.5–52 FAM5C:M673I CAPRIN2:K307N chr11:16.1–46.4 chr4:4.2–190.9 chr18:0.2–77.9 KIF17:R769W CARD8:I13V chr18:7–19.3 chr5:137.4–178.4 chr19:30.3–41.4 chr2:69.7–96.8 KIF3C:V184I CDH5:V189L chr5:74.7–135.7 CEP135:R292X chr4:8.6–189.1 PRSS23:G347V chr8:0.2–33.4 CWC22:M550I chr6:122.8–168.5 RANBP2:M933I chr9:0.2–20.8 DLEC1:L305L chr6:0.3–29.8 chr9:27.5–140.8 RHBDD1:S5X DNAH3:R2256K chr6:39.9–51.7 chrX:2.8–154 ROBO1:S879S ERBB3:R103R chr6:156.3–170.9 chr6:62.4–121.6 SEMA3F:D65G FAM135B:A566A ACTL10:L219M chr7:1.1–18.8 SLC1A2:G257E HKDC1:R642G GJB1:D169D chr9:0.1–140.9 USP34:D1540H INTS9:T395T LCK:chr1:32.7 T>G chrX.23.9–153.7 ZBBX:A669S MAGEC1:S822X NFS1:R434P chrX:3–23.9 NPFFR1:G330G chr16:45.2–84.9 OR4F15:H6H RNF148:I163V ADAMTSL4:L1066L chr6:86.3–170.7 PHOX2B:M4T SETD1A:R18Q PTPRJ:A670S NALCN:R694R chr9:0.3–38.4 SH3PXD2B:R819Q RDH10:V218L SHPRH:R744Q chrX:9.8–40.4 SYT12:A280A SHH:N115I SLX4:L782L chr9:70.8–131.7 chr10:97.1–134.1 SLC17A6:S582X ST18:D162N chr20:4.8–14.8 SLC44A3:K44N ABCG8:P523P FOXG1:V264V SORCS3:S431L PCK2:Q216Q NOS1AP:S464S SZT2:V1975V STX11:G169A SETD2:S717L chr10:15.1–26.5 TARBP1:F918F NPAP1:V362V chr2:201.3–230.8 TTN:I18465N IQSEC1:S70S UNC13A:V1277M POU3F2:P326P UROC1:R173Q EFCAB4B:Y103Y ZNF462:T1671T GRM7:V153V chr4:120.6–154.7 PARN:F366F C16orf68:V124M ZFHX4:D2817D IGF2R:Y1094Y AADACL2:S205C ZNF804B:S710S ABI2:Q319K chr7:112.7–158.9 C7orf62:G195X MRPS12:F55F CLCC1:P182A REV3L:T903T p53 gene somatic mutation CSMD1:P1962L RHBDL3:G128G HTR1E:L90V CDC5L:K718K p53 sig somatic mutation KIAA1217:M1200I PPP1CC:R191R MUC7:chr4:71.4del p53 sig LOH STK36:P892P NPHS2:D190E CACNB4:A266A NRK:P611S STIC somatic mutation DNAH6:A2821A TMEM62:R554W EYA3:T223T STIC LOH TRIM35:R186Q GALNT16:G433G TRIP12:R798R STIC somatic mutation TWISTNB:K329R GP6:A71A ZNF345:H56Y GPRC5B:F308F STIC localized somatic mutation chr11:65.4–105.1 PER1:G310G FT somatic mutation SAMD4B:P465P ZKSCAN8:R472R FT LOH chr1:1.7–28.2 chr13:103.3–115 FT localized LOH chr13:21.1–102.4 chr22:17.3–32.6 Ovarian tumor somatic mutation Ovarian tumor LOH Ovarian tumor localized mutation Metastasis somatic mutation Metastasis LOH Mutation lost due to chr loss Fig. 2 Somatic mutation and allelic imbalance profiles among different tumor lesions. Somatic mutations and segments of allelic imbalance detected by whole-exome analyses are indicated as colored cells in rows for all patients. Darker shades of each color indicate somatic mutations while lighter shades indicate allelic imbalances. The tumor samples analyzed for each patient are indicated in columns (p53 sig, p53 signature; STIC, serous tubal intraepithelial carcinoma). For ovarian tumors in CGOV62 and STIC lesions in CGOV63 multiple blocks are indicated, including one ovarian tumor where multiple sections were analyzed after hematoxylin and eosin staining or after immunohistochemistry (IHC) staining of p53. These analyses indicated that staining methods did not affect detection of somatic alterations. The color of mutations indicates the degree of relatedness among tumor samples: red, shared among all tumor samples with TP53 highlighted at the top row; green, shared among all tumor samples except p53 signature lesion; purple, shared among fallopian tube tumor and omental metastasis; blue indicates mutations that were first detected in the ovarian tumors; and gray indicates mutations that were only detected in metastatic lesions. Additional color shades or patterns indicate mutations that are localized to specific lesions or lost due to chromosome loss as shown in the legend sequence alterations per tumor sample. Candidate alterations were distinguish the STIC lesions and FT carcinomas from ovarian evaluated across samples in an individual to determine if they cancers or intraperitoneal metastases. were present in multiple neoplastic lesions or were unique to a Given the importance of chromosomal instability in HGSOC , particular sample. To allow for the possibility that a subclone may we extended our analyses to examine structural variation in the have developed in a tumor lesion prior to becoming a dominant multiple tumors of each patient. We focused on regions of allelic clone at another location, we determined if genetic alterations that imbalance that can result from the complete loss of an allele were present in one tumor were also present in a low fraction of (LOH) or from an increase in copy number of one allele relative neoplastic cells of other lesions. This method required high to the other. We divided the genome into chromosome segments coverage of analyzed alterations in all samples and excluded and for each segment compared the minor allele (B-allele) potential artifacts related to mapping, sequencing or PCR errors, frequency values in tumor and normal samples using the ~17,000 allowing specific detection of alterations present in ≥ 1% of whole-exome germline heterozygous single-nucleotide poly- sequence reads (see Methods section for additional information). morphisms (SNPs) observed (Fig. 3, Supplementary Figs. 1–9 The composition of sequence alterations was relatively similar and Supplementary Data 7–11). Overall, we observed that an among the affected lesions of each patient. For example, for average of ~26% (range 12–39%) of the genome had chromoso- CGOV62, the STIC lesion, FT carcinomas, left and right ovarian mal imbalances in the samples analyzed (Fig. 3). cancers, and all three metastatic lesions harbored a common set Integration of sequence and structural alterations identified an of somatic mutations (Fig. 2). In CGOV63, CGOV279, and average of 47 alterations per sample (range 21–74) (Fig. 2). The CGOV278, while most of the sequence alterations were the same combination of both types of alterations allowed robust genomic among the tumors of each patient, a subset of mutations could differentiation between STICs and ovarian cancers or metastatic 4 NATURE COMMUNICATIONS 8: 1093 DOI: 10.1038/s41467-017-00962-1 www.nature.com/naturecommunications | | | p53 sig STIC Left FT tumor Left ovarian tumor A4 Left ovarian tumor A4 IHC Left ovarian tumor A7 Right ovarian tumor Rectal metastasis Appendiceal metastasis Omental metastasis p53 sig STIC D1 STIC D2 STIC D3 Right FT tumor D3 Right ovarian tumor Right FT STIC Right FT tumor Right ovarian tumor Omental metastasis Left FT STIC Right FT STIC Right FT STIC, tumor Right FT tumor Right ovarian tumor Left FT STIC Left ovarian tumor Omental metastasis NATURE COMMUNICATIONS | DOI: 10.1038/s41467-017-00962-1 ARTICLE CGOV62 CGOV63 CGOV280 CGOV279 CGOV278 0.00 0.00 0.00 0.00 0.00 1 1 1 1 0.05 0.05 0.05 0.05 0.05 2 2 2 2 0.10 3 3 3 0.10 3 0.10 3 0.10 0.10 4 4 4 4 4 0.15 0.15 0.15 0.15 0.15 5 5 5 5 5 6 6 6 6 6 0.20 0.20 0.20 0.20 0.20 7 7 7 7 7 8 8 8 8 8 0.25 0.25 0.25 0.25 0.25 9 9 9 9 9 10 10 10 10 0.30 0.30 0.30 0.30 0.30 11 11 11 11 11 12 12 12 12 12 0.35 0.35 0.35 0.35 0.35 13 13 13 13 13 14 14 14 14 14 15 15 15 15 15 0.40 0.40 0.40 0.40 0.40 16 16 16 16 16 17 17 17 17 17 18 18 18 18 18 19 19 19 19 0.45 0.45 0.45 20 0.45 20 20 0.45 20 21 21 21 21 21 22 22 22 22 X X X X X 0.50 0.50 0.50 0.50 0.50 Fig. 3 Genome-wide allelic imbalance profile. Minor allele frequency of heterozygous SNPs identified from normal tissue in each patient are derived in each tumor sample, enabling assessment of allelic imbalance in ~17,000 loci across the exome. Circular binary segmentation (CBS) is applied to minor allele frequencies of SNPs with minimum coverage of 10× in each tumor sample, and the resulting segment means are shown as a heatmap. Asterisks indicate samples where corresponding mutation analyses were not performed due to low tumor purity (omental metastasis of CGOV279, right ovarian tumor of CGOV278) or miliary pattern of tumor samples (peritoneal metastases of CGOV63). Given the relatively lower number of distinct DNA molecules available from the p53 signature samples from CGOV62 and CGOV63, these samples were subjected to a more sensitive LOH analysis (Methods, Genome-wide imbalance analysis) and are not shown here lesions in all patients analyzed. In patient CGOV62, a LOH of 9q A SCHISM tree node represents cells harboring a unique (70.8–131.7 Mb) provided a clear difference between the STIC compartment of mutations defining a subclone whereas an edge and all other carcinomas analyzed (Figs. 2 and 3). Likewise, represents a set of mutations acquired by the cells in the progeny chromosomal changes in 7q represented a distinguishing feature nodes that distinguish them from the cells in the parental node. between the right STIC or right FT tumors and the remaining By definition, for an individual cancer there could only be one lesions (ovarian cancers, omental metastasis, and left STIC) in parental clone, although there could be many different progeny CGOV280 (Figs. 2 and 3). In patient CGOV279, multiple regions subclones representing invasive or metastatic lesions or further of allelic imbalance were present in a STIC near the FT evolution of the primary tumor. The optimal hierarchy among carcinoma, while these were absent in a STIC that was not subclones is determined by examining all possible pairwise adjacent to this lesion. relationships between somatic alterations, and performing a heuristic search over the space of phylogenetic trees to identify a model that best explains the observed alterations. Evolutionary relationship of neoplastic lesions. As somatic In all samples, the SCHISM analysis of sequence and structural genetic alterations can be used to recreate the evolutionary history alterations suggested that the p53 signature or STIC lesions of tumor clones, we used the somatic sequence mutations and contained the ancestral clone for the observed cancers (Fig. 4). chromosomal alterations observed in each patient to determine This evolutionary relationship was strengthened by the observa- the history of tumor clonal evolution. We employed a subclone tion that nearly all of the alterations within the p53 signature and hierarchy inference tool called SCHISM (SubClonal Hierarchy STIC lesions were shared by all other lesions. For example, the Inference from Somatic Mutations) which enables improved ovarian tumors of all cases displayed alterations that were shared phylogenetic reconstruction by incorporating estimates of the in FT lesions but also contained additional changes, suggesting fraction of neoplastic cells in which a mutation occurs (mutation that these represented daughter clones of the latter tumors cellularity) . We estimated the cellularity of each mutation by (Fig. 2). Likewise, the ovarian cancers or their immediate correcting the observed allele frequencies for tumor purity and precursors were likely the direct parental clones for the metastases copy number levels (Methods section). In addition to the in CGOV62, CGOV278, and CGOV280 as demonstrated by the observed structural alterations, this approach allowed us to use shared alterations that were not contained in earlier FT lesions. 213 synonymous and non-synonymous somatic sequence Overall, the phylogenetic model generated by these data suggests alterations to construct the phylogenetic trees illustrated in Fig. 4 a progression from FT epithelium to p53 signatures and to STIC and Supplementary Data 5. lesions which are then precursors of FT carcinoma, ovarian NATURE COMMUNICATIONS 8: 1093 DOI: 10.1038/s41467-017-00962-1 www.nature.com/naturecommunications 5 | | | STIC Left FT tumor Left ovarian tumor (A4) Left ovarian tumor (A4) - IHC Left ovarian tumor (A7) Right ovarian tumor Rectal Metastasis Appendiceal metastasis Omental metastasis STIC D1 STIC D2 STIC D3 Right FT tumor D3 Right ovarian tumor Rectal metastasis* Sigmoidal metastasis* Right FT STIC Right FT Tumor Right ovarian tumor Omental metastasis Left FT STIC Right FT STIC Right FT STIC, near tumor Right FT Tumor Right ovarian tumor Omental Metastasis* Left FT STIC Left ovarian tumor Omental metastasis Right ovarian tumor* ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/s41467-017-00962-1 CGOV62 CGOV63 CGOV280 CGOV279 CGOV278 Normal FT Normal FT Normal FT Normal FT Normal FT epithelium epithelium epithelium epithelium epithelium TP53 TP53 TP53 TP53 TP53 SI SFMBT2 TTN IGSF21 MACF1 Chr17 LOH Chr17 LOH Chr17 LOH Chr17 LOH Chr17 LOH Precursor to right p53 signature p53 signature Right FT STIC FT STIC and tumor Left FT STIC APOB TTN IQSEC1 GRM7 Chr18q LOH SETD2 SETD1A Chr16 LOH Chr4 LOH Chr10q LOH Right Right FT tumor STIC D1,3, right FT FT STIC STIC tumor LAMA5 Chr2 LOH Chr7 LOH Chr18p LOH SZT2 KIAA1217 IGF2R Precursor to ovarian ABCG8 Chr10p LOH Chr11 LOH Chr9 LOH tumor, omental met and Left FT STIC PER1 Chr1 LOH REV3L Right ovarian Right FT STIC tumor (near tumor), Left ovarian tumor Omental metastasis Left FT tumor, STIC D2 Right ovarian tumor right FT tumor Omental metastasis Left FT STIC ovarian tumors, metastases CDC5L Right ovarian tumor Fig. 4 Schematic of tumor evolution. The history of tumor evolution in each patient is modeled as a subclonal hierarchy inferred from the somatic mutations and large scale genomic regions harboring loss of heterozygosity (LOH features) using the SCHISM framework, and is depicted as a tree. Each tree starts from a root node corresponding to the normal fallopian tube epithelium (germline). In all patients, mutations in TP53 (red boxes) are among the earliest somatic alterations and are ubiquitously present in all tumor samples. Somatic alterations (boxes) are acquired along edges (arrows) of the tree, and example alterations are indicated in each case. Nodes of the tree represent cells whose genotype is described by the presence of somatic mutations and LOH features on the path connecting the node to the root of the tree. Each node is labeled with tumor samples harboring all upstream and lacking any downstream alterations. The trees inferred for all patients support a pattern of evolution with p53 signatures and STIC lesions as early events in tumorigenesis. Mutation clusters and LOH feature groups follow the same color code as Fig. 2 carcinoma, and metastatic lesions. In addition to the sequential of these patients (BRCA1 Q1200X, BRCA2 L2653P, and a BRCA2 accumulation of alterations in this linear evolution, we also 55 kb hemizygous deletion in CGOV65, CGOV64, and observed branching phylogenetic trees due to continued evolution CGOV304, respectively), as well as somatic mutations in TP53, within STIC lesions as well as FT carcinomas and ovarian and LOH of both chromosome 13 and 17, encompassing the carcinomas (Fig. 4). We compared evolutionary trees resulting BRCA1, BRCA2, and TP53 loci in all of these cases (Supple- from SCHISM analysis with those derived by maximum mentary Figs. 6, 7, 8, and 9). Whole-exome analyses showed that parsimony phylogeny using PHYLIP and the results were similar the STIC lesions contained a total of 91, 23, 34, and 46 non- in all cases (Fig. 4 and Supplementary Fig. 11). synonymous and synonymous somatic mutations, in CGOV65, Interestingly, patient CGOV280 had a right STIC, a right CGOV64, CGOV303, and CGOV304, respectively. Overall, these fallopian carcinoma, and a right ovarian cancer but also had a analyses revealed that STICs in isolation in patients with or STIC in the left FT (Supplementary Fig. 5). In this case the without germline BRCA changes have a roughly similar number SCHISM analysis suggested that the lesion in the left FT which of sequence changes to STICs in patients with sporadic tumors. was pathologically determined to be a STIC actually represented a These observations provide evidence that isolated STICs may act metastatic lesion of the right ovarian cancer (Fig. 4). This lesion as precursors in the same manner as those identified in patients shared nearly all the alterations of the ovarian cancer but with sporadic advanced stages HGSOC analyzed in this study. contained 10 single base substitutions and four additional regions of allelic imbalance on chromosomes 1, 13, and 22, and both the left STIC and right ovarian cancer had an additional region of Recurrent molecular alterations. We examined tumors from the allelic imbalance on chromosome 7 that was absent in the right nine patients to identify recurrent non-silent sequence or chro- STIC (Figs. 2 and 3). These observations are consistent with the mosomal changes. Although no genes other than TP53 were above model of STIC to ovarian cancer progression, but suggest mutated in all patients analyzed, we identified mutations in ten that in advanced disease ovarian cancers may also seed metastatic genes that were altered in two or more patients (Supplementary deposits throughout the peritoneum, including to the FT on the Data 6). These included mutations in the tumors of two patients of the PIK3R5 gene that encodes a regulatory subunit of the PI3- contralateral side. kinase complex. CGOV64 also had a somatic alteration in PTEN that together with changes in PIK3R5 highlight the importance of Genomic alterations in isolated STICs. Neoplastic cells observed the PI3K pathway in ovarian cancer . Additional genes that were in the FTs rather than the ovaries removed from carriers of observed to be altered in other ovarian cancers through other germline mutation of BRCA1 and BRCA2 provided the first large scale sequencing efforts such as TCGA are indicated in 15, 26 indication of the FT as a potential cell of origin of HGSOC . Supplementary Data 6. Since < 1.25% of HGSOC are diagnosed with stage I disease , In addition to recurrent sequence changes, we found altera- BRCA carriers provide a unique opportunity to analyze genomic tions in regions of allelic imbalances encompassing several tumor alterations in isolated STICs without associated HGSOC. We suppressor genes involved in ovarian cancer. Remarkably, these examined neoplastic samples from three individuals with germ- included losses of BRCA1, BRCA2, and TP53 in all nine patients, line BRCA alterations where STIC lesions were incidentally and loss of PTEN for CGOV62, CGOV63, CGOV280, and identified after prophylactic bilateral salpingo-oophorectomy, and CGOV64 (in addition to the somatic sequence alterations of these one patient where two STICs were identified after resection of a genes) (Supplementary Figs. 1–9). In all cases, the LOH observed pelvic mass (Supplementary Data 1). We identified BRCA1 or in the metastatic lesions and ovarian tumor lesions for regions BRCA2 sequence alterations or deletions in the germline of three encompassing these genes were already present in the FT tumor 6 NATURE COMMUNICATIONS 8: 1093 DOI: 10.1038/s41467-017-00962-1 www.nature.com/naturecommunications | | | NATURE COMMUNICATIONS | DOI: 10.1038/s41467-017-00962-1 ARTICLE and STIC lesions. Considering the evolutionary model above, (including STICs or p53 signatures) are infrequent and unlikely these data suggest that a combination of sequence changes in a to be the source of most FT lesions. few genes including TP53 together with loss of the TP53 wild-type Despite these potential limitations, the data we have obtained allele as well as BRCA1, BRCA2, and PTEN may be crucial early provide new insights into the etiology of ovarian Type II carci- 27, 28 events that are needed for the initiation of STICs . noma and have significant implications for the prevention, early detection and therapeutic intervention of this disease. The results suggest that ovarian cancer is a disease of the FTs, with the Evolutionary timeline of ovarian cancer development. To esti- development of p53 signatures and STICs as early events. The mate the time between the development of the earliest neoplastic subsequent formation of a cancer in the ovaries represents a clones in the FT and the development of ovarian and other seeding event from a primary tumor in the FT that already metastatic lesions we used a mathematical model for comparative contains sequence and structural alterations in key driver genes, 24, 29 lesion analysis . This model estimates the time interval including those in TP53, PI3K pathway, and BRCA1/BRCA2 between a founder cell of a tumor of interest and the ancestral genes. The recurrent allelic imbalances observed in chromosomes precursor cell assuming that mutation rates and cell division 1, 6, 16, 18, 20, and 22 may suggest additional genes that are times are constant throughout a patient’s life (Methods section). involved in this process. The timing of the progression from In patient CGOV62, this model would suggest ~1.9 years between STICs to ovarian cancer in the cases we analyzed was on average the development of the STIC lesion and the ovarian cancer (90% 6.5 years, but seeding of metastatic lesions in these patients CI, 0.5–4.2 years). For other patients this transition appears to occurred rapidly thereafter. This timing is consistent with recent have been slower as the average time between STICs and ovarian reports showing a difference of 7.7 years in the age of BRCA cancer among all patients was 6.5 years (1.4–10.7 years). carriers with localized vs. advanced adnexal lesions .This evo- Importantly, in patients with metastatic lesions, the time between lutionary timeline can help explain why most HGSOC patients are the initiation of the ovarian carcinoma and development of diagnosed at advanced stage (III/IV) with pelvic and peritoneal metastases appears to have been rapid (average 2 years). There spread of disease, and why among asymptomatic BRCA germline were either no additional mutations in metastatic lesions (e.g., mutation carriers half of the cases diagnosed with asymptomatic CGOV62 omental, rectal or appendiceal metastasis or CGOV280 adnexal neoplasia have already seeded to pelvis or peritoneum (> omental metastasis) or the number of additional changes was IA) . These observations are largely similar to other genomic 19, 20, 23, 32 small (e.g., three changes in CGOV278 omental metastasis), analyses of the evolution of ovarian cancer as well as the reflecting the ease with which cancer cells located on the ovaries recent analyses of STIC lesions that were reported while this can subsequently seed additional peritoneal sites. Although the study was under review . Our study highlights the role of precise timing of this progression depends on assumptions related p53 signatures as early lesions in this evolutionary paradigm. to mutation rates, which may change during tumor progression, Our genomic analyses are consistent with population-based models employing different rates all showed longer timeline from studies of the effects of salpingectomy on the risk of ovarian STIC lesions to ovarian tumors followed by rapid development of cancer. Prophylactic bilateral salpingo-oophorectomy has been metastatic lesions (Methods section). shown to reduce the risk of developing ovarian cancer in BRCA 34, 35 mutation carriers to below 5% . Likewise, bilateral sal- pingectomy, performed as a contraceptive method instead of Discussion tubal sterilization, reduced the risk of ovarian cancer by 61% at 10 years . Our study provides a mechanistic basis for these obser- These results provide a comprehensive evolutionary analysis of sporadic HGSOC in five patients. Given the unique nature of the vations and has implications for clinical management in pre- multiple samples we examined from each patient, our study may vention of ovarian cancer. In high risk BRCA carriers, bilateral have certain limitations not typical of genome-wide efforts. First, salpingectomy with delayed oophorectomy should be con- the small size of the tumor samples compared to surrounding sidered through participation in ongoing clinical trials non-neoplastic tissue could potentially lead to low tumor purity. (NCT02321228; NCT01907789). In non-carriers, our work The high mutant allele fraction of TP53 among cancer samples implies that for women who undergo surgery for benign uterine (average of 56–85%) indicates that this issue was largely over- causes, total abdominal hysterectomy and bilateral salpingectomy come through LCM. Second, the small number of cells in with sparing of the ovaries should be considered , and that p53 signature samples may have limited our genomic analyses for bilateral salpingectomy may be a preferred contraceptive alter- these lesions. The observation that all sequence changes in native to tubal ligation. The dual concepts in these recommen- p53 signatures were also present in STIC and other carcinomas of dations for BRCA carriers and non-carriers are that removal of multiple sites is consistent with our evolutionary model and the FTs (rather than the ovaries) may be curative as it eliminates suggests that these cells are likely to represent a parental clone of the underlying cellular precursors of ovarian cancer, and that other neoplastic lesions. Third, our analysis was limited to preservation of the ovaries provides long term benefits due to ovarian cancers where STICs and other concomitant lesions were decreased risk and fatalities from coronary heart disease and identified, and may therefore not be representative of all HGOCs. other illnesses . A limitation of this approach is that as the The absence of STIC lesions in ~40% of sporadic HGSOCs is precise timing of when potentially malignant cells shed from the likely due to an incomplete sampling of the FT or the overgrowth FT and microscopically seed the ovary is unknown, removal of of the STIC by the carcinoma in the context of bulky disease, but the tubes may not provide optimal risk-reduction. may also reflect another site of origin that has yet to be deter- Our observations also have implications for improved detec- mined for these cancers . Fourth, this study did not intend to tion of ovarian cancer. Unfortunately, < 1.25% of HGSOC are address the intra-tumoral heterogeneity within the carcinomas confined to the ovary at diagnosis . Earlier detection of this but rather focused on clonal changes within each tumor. Fifth, as disease is likely to benefit from the identification of a precursor in any evolutionary analyses, the genomic alterations we observed lesion, as has been the case for many other tumor types. Our data provide the most likely model of tumor development but do not suggest that FT neoplasia is the origin of ovarian serous carci- exclude the possibility of other relationships. Nevertheless, our nogenesis, and can directly lead to cancer of the ovaries and of analyses of somatic alterations suggest that models where the other sites. Currently, the typical histopathologic evaluation of 20, 21 ovarian cancer or metastatic lesions seed the FT tumors FTs typically involves a cursory evaluation of one or two NATURE COMMUNICATIONS 8: 1093 DOI: 10.1038/s41467-017-00962-1 www.nature.com/naturecommunications 7 | | | ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/s41467-017-00962-1 representative sections. Our study suggests that systematic sec- Sample preparation and next-generation sequencing. DNA was extracted from patient whole blood using a QIAamp DNA Blood Mini QIAcube Kit (Qiagen tioning and extensive examination of total FTs should become Valencia, CA). Genomic DNA from FFPE blocks was extracted from the micro- common practice in pathology, and not confined to academic dissected tissues using the QIAamp DNA FFPE Tissue kit (Qiagen, Valencia, CA). tertiary care centers. Depending on whether the FTs are removed In brief, the samples were incubated in proteinase K for 16 h before DNA for benign conditions, risk-reducing bilateral salpingectomy, or extraction. The digested mixture was transferred to a microtube for DNA frag- mentation using the truXTRAC™ FFPE DNA Kit with 10 min shearing time as per gynecological cancers, specific examination protocols should be 16, 40 the manufacturer’s instructions (Covaris, Woburn, MA). Following fragmentation, applied . Given the window of time that appears to exist the sample was further digested for 24 h followed by 1 h incubation at 80 °C. DNA between the formation of FT lesions and development of ovarian purification was performed using the QIAamp DNA FFPE Tissue kit following the cancer, these insights open the prospect of novel approaches for manufacturer’s instructions (Qiagen, Valencia, CA). Fragmented genomic DNA from tumor and normal samples were used for Illumina TruSeq library con- screening. Such approaches may be especially important given the struction (Illumina, San Diego, CA) according to the manufacturer’s instructions limited therapeutic options currently available for ovarian can- or as previously described . Exonic or targeted regions were captured in solution 4, 5 cer . Recent advances for ultrasensitive detection of genetic using the Agilent SureSelect v.4 kit or a custom targeted panel according to the alterations in blood-based liquid biopsies, pap smears, and other manufacturer’s instructions (Agilent, Santa Clara, CA). Paired-end sequencing, 41, 42 bodily fluids , or imaging approaches may provide opportu- resulting in 100 bases from each end of the fragments for exome libraries and 150 bases from each end of the fragment for targeted libraries, was performed using nities in early diagnosis and intervention. Illumina HiSeq 2000/2500 and Illumina MiSeq instrumentation (Illumina, San Diego, CA). Methods Next-generation sequencing data and identification of somatic mutations. Specimens obtained for sequencing analysis. The study was approved by the Somatic mutations were identified using VariantDx custom software for identi- Institutional Review Board at Brigham and Women’s Hospital and the Johns fying mutations in matched tumor and normal samples. Prior to mutation calling, Hopkins Hospital and all patients gave informed consent before inclusion. Five primary processing of sequence data for both tumor and normal samples were sequential patients with stage III sporadic HGSOC, in whom a STIC was identified performed using Illumina CASAVA software (v1.8), including masking of adapter in their FTs (FT), were included. In addition, we included isolated STICs from sequences. Sequence reads were aligned against the human reference genome three patients with germline BRCA deleterious alterations who underwent pro- (version hg18 or hg19) using ELAND. Candidate somatic mutations, consisting of phylactic bilateral salpingo-oophorectomy as well as a fourth patient who had point mutations, insertions, and deletions were then identified using VariantDx bilateral salpingo-oophorectomy and hysterectomy in the context of a pelvic mass. across either the whole exome or regions of interest . For samples analyzed using All cases underwent complete tubal examination using the SEE-FIM protocol . targeted sequencing, we identified candidate mutations that were altered in > 10% Formalin-fixed paraffin embedded (FFPE) blocks were retrieved from the pathol- of distinct reads. For samples analyzed using whole-exome sequencing, we iden- ogy files at Brigham and Women’s Hospital and Johns Hopkins Hospital within the tified candidate mutations that were altered in > 10% of distinct reads with ≥ 5 3 months following surgical diagnosis and stored at 4 °C to slow down nucleic acids altered reads in at least one sample, where coverage at the altered base was at least degradation. All the cases were reviewed by a gynecologic pathologist (M.S.H., D.I. as high as the TP53 alteration in that sample, and where the ratio of the coverage of L., L.S.) that confirmed the diagnosis of STIC and/or p53 signature in the FT. Slides the mutated base to the overall sequence coverage of that sample was > 20%. from each FFPE block, including early lesions, invasisve carcinomas and metas- Identified mutations were reported as present in other samples of the same patient tases, were stained with hematoxylin and eosin, and analyzed by p53 IHC staining. if the mutation was present in at least two distinct altered reads. Mutations present In each FT, at least one STIC and/or p53 signature was identified and micro- in polyN tract ≥ 5 bases, or those with an average distinct coverage below 50× were dissected separately. Importantly, STICs were not pooled together if they were in removed from the analysis. the same section and were considered separate STICs. An analysis of each candidate mutated region was performed using BLAT. For each mutation, 101 bases including 50 bases 5ʹ and 3ʹ flanking the mutated base was used as query sequence (http://genome.ucsc.edu/cgi-bin/hgBlat). Candidate Immunohistochemistry and laser capture microdissection. For accurate mutations were removed from further analysis, if the analyzed region resulted in microdissection of early lesions including STIC and p53 signature, IHC staining of > 1 BLAT hits with 90% identity over 70 SCORE sequence length. All candidate p53 was specifically adapted for LCM as previously described . PEN membrane alterations were verified by visual inspection. frame slides Arcturus (Life technologies, Carlsbad, CA) were used. Each slide was coated with 350 ul of undiluted poly-L-lysine 0.1% w/v (Sigma, St. Louis, MO). For drying, the slides were placed in a slide holder for 60 min at room temperature. Genome-wide allelic imbalance analysis. We performed comparative analysis of Tissue sections were cut and mounted on the pretreated membrane slides. LOH across the tumor samples from each patient to identify copy number Deparaffinization was performed in fresh xylene for 5 min twice, followed by 100% alterations occurring in the course of tumor evolution. Minor allele frequency ethanol for 2 min, 95% for ethanol 2 min, and 70% ethanol for 2 min. Subsequently, (MAF) of germline heterozygous SNPs with minimum coverage of 10× in each the slides were transferred into distilled water for 5 min. Heat-epitope antigen tumor sample were segmented using circular binary segmentation algorithm retrieval (AR) was performed in Citrate Buffer (Dako, Carpinteria, CA) at low (CBS) . Genomic segments where the difference between tumor and normal MAF temperature (60 °C) for 44 h instead of 120 °C for 10 min to reduce tissue and DNA exceeded a threshold of 0.10 were labeled as harboring LOH. In each tumor sample, damage by high temperature. Retrieval solution was pre-warmed to 60 °C before the minimum MAF across segments with minimum size of 10 Mb was calculated to usage. After incubation in the oven, the AR solution was left to cool down to room provide a measure of sample purity. Each segment marked as LOH was assigned to temperature and the slides were rinsed for 30 seconds in fresh 1×PBS then incu- one of the three confidence categories: (1) high confidence, segment MAF within bated for 40 min with primary antibody anti-p53 (Epitomics, Burlingame) at 1:100 0.1 of the minimum sample MAF. (2) Intermediate confidence, segment MAF in a humidifying chamber. Before adding the secondary antibody, slides were within 0.1–0.2 of the minimum sample MAF. (3) Low confidence, segment MAF washed twice for 1 min in fresh 1×PBS. The secondary antibody, labeled polymer- exceeding the minimum sample MAF by > 0.2. HRP anti-mouse (Dako EnVision System-HRP (DAB), Carpinteria, CA) was Next, sample level segments were intersected across the entire set of samples applied for 30 min. Then, slides were washed twice for 1 min in fresh 1×PBS. from each patient to derive patient level segments while accounting for the Chromogenic labeling was performed with 3,3-DAB substrate buffer and DAB possibility of variable segment break points in different samples (Supplementary chromogen (Dako EnVision System-HRP (DAB), Carpinteria, CA) for 5 min. Data 7–11). Patient level segments were filtered to keep those covering a minimum Slides were washed again for 30 s in fresh distilled water. Dehydration was per- of 20 SNPs and with minimum length of 10 Mb. The resulting segments were formed as follows: 70% ethanol for 30 s, 95% ethanol for 30 s, 100% ethanol for 30 further narrowed down to only include those with high confidence LOH in at least s, and xylene for 30 s. The stained slides were microdissected within 2 h with the one of the samples. Genomic segments with LOH in a subset of samples can serve Arcturus XT LCM system (Life technologies, Carlsbad, CA). as informative markers to track tumor evolution similar to somatic mutations. To increase the specificity in identifying this class of genomic segments, we required a minimum distance of 0.1 between the MAF of samples with and without LOH. To Hematoxylin staining for laser capture microdissection. Invasive carcinomas minimize the possibility of over-segmentation which could result in inflated from the ovaries, the FTs and intraperitoneal metastases or STICs from patients estimates of the number independent structural alterations, we evaluated patient with negative p53 IHC staining were microdissected after Hematoxylin staining. level segments with boundaries within a 5 Mb window. In cases where the LOH Briefly, deparaffinization was performed in fresh xylene for 1 min twice followed by calls were identical and the difference of segment MAFs were ≤0.05 in all tumor 100% ethanol for 1 min, 95% for ethanol 1 min, and 70% ethanol for 1 min. The samples, the segments were merged. slides were transferred into distilled water for 2 min before staining with Hema- For CGOV62 and CGOV63, the number of germline heterozygous SNPs toxylin for 2 min. Subsequently, slides were rinsed in distilled water until they meeting the coverage criteria in p53 signature samples was significantly lower than became clear before undergoing dehydration in 70% ethanol for 1 min, 95% the other samples from the same patient. Thus, we modified the approach above in ethanol for 1 min, 100% ethanol for 1 min, and xylene for 1 min. The stained slides these two patients to enable sensitive analysis of LOH in p53 signature samples. were microdissected within 2 h. Initially, the patient level genomic segments of interest were defined excluding 8 NATURE COMMUNICATIONS 8: 1093 DOI: 10.1038/s41467-017-00962-1 www.nature.com/naturecommunications | | | NATURE COMMUNICATIONS | DOI: 10.1038/s41467-017-00962-1 ARTICLE p53 signature samples. Next, in each genomic segment, the minor allele of each we can estimate mutation cellularity (or cancer cell fraction) from the observed overlapping germline SNP was determined by taking a majority vote over their reference and alternate read counts, and estimates of copy number, and tumor minor alleles in the other samples. The coverage and minor allele read count for purity as follows. each SNP was derived using samtools (v0.1.19) mpileup module . The segment MAF in p53 signature samples were calculated by dividing the sum of minor allele mαC V ¼ exp read counts across all SNPs by the total coverage of SNPs, circumventing the α CN þðÞ 1  α CN T N variance resulting from low coverage at individual SNPs. In each p53 signature sample, segments with MAF lower than that of the normal by at least 0.1 were In the equation above, V is the expected variant allele frequency of the exp marked as LOH. mutation, m is the multiplicity of the mutation which is set to 1, α is the purity of the tumor sample, C is the cellularity of the mutation, and CN and CN are the T N integer copy number of tumor and normal sample at the locus of the mutation. The Copy number analysis. The genome-wide copy number profiles were determined observed alternate read count of the mutation can be modeled as a binomial by analysis of the ratio of read counts in the tumor and matched normal whole- random variable drawn from a distribution with probability parameter equal to exome sequenced samples. In each sample, the number of reads mapping to V and number of trials equal to the total sequence coverage of the mutation. We exp genomic bins located in target and off-target regions were corrected for biases calculated the likelihood for observation of the alternate read counts for cellularity arising from GC-content, repetitive sequences, and target capture process using values spanning the range of 0–1 in increments of 0.01, and derived the maximum CNVKit (v.0.7.6) (https://doi.org/10.1371/journal.pcbi.1004873). The log ratio of likelihood estimate and confidence interval for the mutation cellularity. the processed tumor to normal read counts provides a measure of copy number in To obtain reliable estimates of mutation cellularity, we clustered mutations by each bin, and was segmented to yield genomic intervals at constant copy number joint presence or absence across the available tumor samples. This approach makes levels. The difference in sequencing library size between the tumor and normal phylogenetic reconstruction more tractable and the cellularity of the resulting samples is another factor that needs to be accounted for when analyzing reads clusters can be estimated with higher accuracy than that of individual mutations. ratios in NGS-based copy number pipelines. In CNVKit, the log ratio values in For each patient, a mutation was called as present or absent in each of the available each sample are adjusted by setting the median of autosomal bins to 0 in log space, tumor samples (10 samples from CGOV62, 6 samples from CGOV63, 5 samples assuming a median ploidy of 2 for the genome. Given the high prevalence of copy from CGOV280, 4 samples from CGOV279, and 3 samples from CGOV278). To number aberrations in ovarian cancer and the high frequency of allelic imbalance call the mutation present, we used a minimum allele frequency of 2% and 2 distinct in the present cohort, this assumption may not be accurate, and will manifest itself mutant reads. Mutation clustering was performed by a greedy algorithm. Tumor as a genome-wide bias or shift of log ratio values. purity in each tumor sample was estimated as the read count fraction of TP53 Therefore, an alternative approach for normalization of log ratio values was mutation in each patient. Each patient harbored a single distinct TP53 mutation adopted, which takes into account the level of allelic imbalance in each genomic that was present in all tumor samples, and we assumed the wild-type allele was lost, region. Briefly, genomic regions with the least degree of allelic imbalance were as supported by the ubiquitous LOH of chromosome 17. To derive a more identified in each tumor sample, and used in a normalization process based on the comprehensive view of the evolution of these samples, we extended the original notion that these regions can only be present in an even number of copies. The SCHISM framework to model acquisition of large scale somatic copy number distribution of log ratio values among these regions was inspected to ensure that alterations, which can be detected by analysis of allelic imbalance (including LOH). they belong to the same copy number level. Otherwise, a subset of regions at a First, we extracted a set of high confidence genomic regions with ubiquitous, common log ratio (and thus copy number) level were selected. By fixing the copy partially shared, or private LOH in tumor samples of each patient (Methods number of these segments at a specified level, one can solve for the genome-wide section). These regions of LOH served as binary features that could be used for bias of log ratio values as follows, and thus identify the genome-wide integer copy evolutionary analysis, and were clustered into LOH feature groups with identical number profile. patterns of presence or absence across samples (Fig. 2). Each LOH feature group was compared to the somatic mutation clusters in each patient, with respect to its α CN þðÞ 1  α CN T N R ¼ log  δ pattern of presence or absence across samples. In cases where a mutation cluster with the identical pattern could be found, the cluster and the LOH feature group In the equation above, R represents the observed log ratio of read counts, α is were assumed to have occurred together in the course of tumor evolution. the purity of the tumor sample, CN and CN are the integer copy number of Otherwise, the LOH feature groups were modeled as distinct features, and added in T N tumor and normal samples at a locus, and δ is the genome-wide bias term. Given post-hoc analysis by application of the lineage precedence rule from SCHISM; the value of tumor purity and copy number, δ is the only unknown in the equation. which requires cellularity of ancestor alterations to be greater than or equal to To favor solutions with less complex genomes, the copy number of regions with cellularity of descendant alterations in all tumor samples. complete allelic balance was initially set to 2. If the resulting solution was deemed SCHISM was run with the above inputs and default parameter settings to infer implausible (e.g., by implying chromosome or chromosome arm scale homozygous the order of somatic alterations and thus define subclonal hierarchy in each patient. deletions), the copy number of regions with complete allelic balance was assigned SCHISM software is freely available for non-profit use at http://karchinlab.org/ to 4 and an alternative solution was found (Supplementary Fig. 10). appSchism. Details of the genomic segments selected to solve for the genome-wide bias term Evolutionary trees resulting from SCHISM analysis were compared with those δ are as follows. In CGOV62, chromosomes 4 and 12 did not have allelic imbalance derived by maximum parsimony phylogeny using PHYLIP (Phylip-3.695, PARS in any tumor samples. The solution assigning copy number two to these regions method). For CGOV280, an adjustment to the tree was applied to account for implied homozygous deletion of the p-arm of chrX in multiple samples; therefore, multiple subclones in Right FT STIC. the simplest plausible solution assigned them to four copies. In CGOV63, chromosomes 6 and 15 did not have allelic imbalance in any of the tumor samples, Estimating an evolutionary timeline. Following the approach of Jones et al. , the and were assigned to two copies. No complete chromosome with absence of allelic observed data are the number of somatic mutations in the STIC (n ), the number of imbalance across all tumor samples could be found in CGOV278. Therefore, four mutations in the metastasis (n ), and the age at which the patient was diagnosed genomic regions with no allelic imbalance were selected for the normalization (t ), where somatic mutations include both sequence and structural alterations. process above. These regions were chr8:38–69 Mb, chr12:62–85 Mb, chr18:7–19 Unknown is the birthdate (t ) of the cell that was the last common ancestor of the Mb, chr20:23–35 Mb. The solution assigning these regions to two copies resulted in STIC and the metastasis. Assuming the mutation rate of somatic passenger an implausible assignment of homozygous deletion to chr5:50–136 Mb. Therefore, mutations and the length of the cell cycle is constant, the number of somatic assignment of four copies to the selected regions results in the simplest solution. In mutations in the metastasis cell that were present in the STIC follows a binomial CGOV279, two genomic regions were selected for the normalization procedure: distribution with parameters n and probability t /t .As t is unknown, we posit a k j k j chr5: 64–131 Mb, chr20:17–36 Mb. Evaluation of log ratio values suggested that the conjugate beta probability distribution on the rate t /t with shape parameters a and j k two regions are present at different copy levels, as evidenced by a difference of b estimated from previous studies as described below. The posterior distribution of ~0.60 in the log ratio values. The region on chr5, which had the lower log ratio t /t is β (a + n , b + n −n ) from which 90% highest posterior density intervals can j k j k j level, was assigned to copy number 2. In CGOV280, chr16q had no allelic be constructed with point estimates for the birthdate reported as the posterior imbalance in any samples excluding the left FT STIC. Examination of log ratio mean. For simplicity, we refer to the highest posterior density as a confidence values of chr16q in the left FT STIC supports a copy loss in that sample. The interval. To construct a prior for t /t , we draw on a previous study of four col- j k genome-wide bias term δ was determined by assignment of two copies to chr16q in orectal cancer patients where a small number of additional passenger mutations the four samples with no allelic imbalance, and one copy in the left FT STIC. were acquired by the cell that gave birth to the metastasis. On average, 95% of the mutations in the original adenocarcinoma were present in the metastases. We Subclonal hierarchy analysis. The tumor subclonality phylogenetic reconstruc- center the mean for the beta prior at 0.95 using shape parameters a = 34 and b = tion algorithm SCHISM was used to infer tumor subclonal hierarchies from the 1.6. Our prior is equivalent to one patient having 34 passenger somatic mutations set of confidently called somatic mutations in each patient. Given the estimates of in the original lesion and 1.6 additional mutations to be acquired by cells that gave genome-wide copy number profile, most copy number aberrations seem to occur birth to the metastases. For patients with three samples in a linear tree as deter- early in the evolution of disease and are common across the lesions analyzed from mined by evolutionary analyses (say, samples j, k, and l where sample j is the STIC, each patient. Thus, the majority of somatic mutations can be assumed to occur l is the metastasis, and k is an intermediate sample), we first derived the posterior following the acquisition of copy number aberrations, and can be present in cancer distribution for t comparing mutations in samples k and l. Next, we derived the cells with multiplicity of one (one mutated copy per cell). Using this assumption, posterior distribution of t integrating over all possible values of t , thereby fully j k NATURE COMMUNICATIONS 8: 1093 DOI: 10.1038/s41467-017-00962-1 www.nature.com/naturecommunications 9 | | | ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/s41467-017-00962-1 incorporating the uncertainty of the intermediate timepoint in the estimate of t . 26. Leeper, K. et al. Pathologic findings in prophylactic oophorectomy specimens in We evaluated three additional prior models, and found that that posterior inference high-risk women. Gynecol. Oncol. 87,52–56 (2002). under these alternative models given by 90% credible intervals for t −t , results in 27. Roh, M. H. et al. High-grade fimbrial-ovarian carcinomas are unified by altered k j qualitatively similar timelines among different lesions in tumor progression. p53, PTEN and PAX2 expression. Mod. Pathol. 23, 1316–1324 (2010). 28. Perets, R. et al. Transformation of the fallopian tube secretory epithelium leads to high-grade serous ovarian cancer in Brca;Tp53;Pten models. Cancer Cell 24, Data availability. Sequence data have been deposited at the European Genome- 751–765 (2013). phenome Archive, which is hosted at the European Bioinformatics Institute, under 29. Jones, S. et al. Comparative lesion sequencing provides insights into tumor study accession EGAS00001002589. evolution. Proc. Natl Acad. Sci. USA 105, 4283–4288 (2008). 30. Perets, R. & Drapkin, R. It’s totally tubular…riding the new wave of ovarian Received: 26 January 2017 Accepted: 9 August 2017 cancer research. Cancer Res. 76,10–17 (2016). 31. Conner, J. R. et al. Outcome of unexpected adnexal neoplasia discovered during risk reduction salpingo-oophorectomy in women with germ-line BRCA1 or BRCA2 mutations. Gynecol. Oncol. 132, 280–286 (2014). 32. Karnezis, A. N. & Cho, K. R. Of mice and women - non-ovarian origins of “ovarian” cancer. Gynecol. Oncol. 144,5–7 (2016). References 33. Eckert, M. A. et al. Genomics of ovarian cancer progression reveals diverse 1. Ferlay, J. et al. Cancer incidence and mortality patterns in Europe: estimates for metastatic trajectories including intraepithelial metastasis to the fallopian tube. 40 countries in 2012. Eur. J. Cancer 49, 1374–1403 (2013). Cancer Discov. 6, 1342–1351 (2016). 2. Siegel, R. L., Miller, K. D. & Jemal, A. Cancer statistics, 2015. CA Cancer J. Clin. 34. Rebbeck, T. R. et al. Prophylactic oophorectomy in carriers of BRCA1 or 65,5–29 (2015). BRCA2 mutations. N. Engl. J. Med. 346, 1616–1622 (2002). 3. Cress, R. D., Chen, Y. S., Morris, C. R., Petersen, M. & Leiserowitz, G. S. 35. Kauff, N. D. et al. Risk-reducing salpingo-oophorectomy in women with a Characteristics of long-term survivors of epithelial ovarian cancer. Obstet. BRCA1 or BRCA2 mutation. N. Engl. J. Med. 346, 1609–1615 (2002). Gynecol. 126, 491–497 (2015). 36. Falconer, H., Yin, L., Gronberg, H. & Altman, D. Ovarian cancer risk after 4. Menon, U., Griffin, M. & Gentry-Maharaj, A. Ovarian cancer screening-- salpingectomy: a nationwide population-based study. J. Natl. Cancer. Inst. 107, current status, future directions. Gynecol. Oncol. 132, 490–495 (2014). dju410 (2015). 5. Jacobs, I. J. et al. Ovarian cancer screening and mortality in the UK 37. Kwon, J. S. et al. Prophylactic salpingectomy and delayed oophorectomy as an Collaborative Trial of Ovarian Cancer Screening (UKCTOCS): a randomised alternative for BRCA mutation carriers. Obstet. Gynecol. 121,14–24 (2013). controlled trial. Lancet 387, 945–956 (2016). 38. McAlpine, J. N. et al. Opportunistic salpingectomy: uptake, risks, and 6. Kurman, R. J. & Shih Ie, M. The dualistic model of ovarian carcinogenesis: complications of a regional initiative for ovarian cancer prevention. Am. J. revisited, revised, and expanded. Am. J. Pathol. 186, 733–747 (2016). Obstet. Gynecol. 210, 471.e1–471.e11 (2014). 7. Kurman, R. J. & Shih Ie, M. The origin and pathogenesis of epithelial 39. Parker, W. H. et al. Ovarian conservation at the time of hysterectomy and long- ovarian cancer: a proposed unifying theory. Am. J. Surg. Pathol. 34, 433–443 term health outcomes in the nurses’ health study. Obstet. Gynecol. 113, (2010). 1027–1037 (2009). 8. Karst, A. M. & Drapkin, R. Ovarian cancer pathogenesis: a model in evolution. 40. Longacre, T. A., Oliva, E., Soslow, R. A. Association of directors of, A. & J. Oncol. 2010, 932371 (2010). Surgical, P. Recommendations for the reporting of fallopian tube neoplasms. 9. Levanon, K., Crum, C. & Drapkin, R. New insights into the pathogenesis of Hum. Pathol. 38, 1160–1163 (2007). serous ovarian cancer and its clinical impact. J. Clin. Oncol. 26, 5284–5293 41. Haber, D. A. & Velculescu, V. E. Blood-based analyses of cancer: circulating (2008). tumor cells and circulating tumor DNA. Cancer Discov. 4, 650–661 (2014). 10. Bowtell, D. D. et al. Rethinking ovarian cancer II: reducing mortality from high- 42. Kinde, I. et al. Evaluation of DNA from the Papanicolaou test to detect ovarian grade serous ovarian cancer. Nat. Rev. Cancer 15, 668–679 (2015). and endometrial cancers. Sci. Transl. Med. 5, 167ra4 (2013). 11. Cancer Genome Atlas Research Network. Integrated genomic analyses of 43. Eberle, F. C. et al. Immunoguided laser assisted microdissection techniques for ovarian carcinoma. Nature 474, 609–615 (2011). DNA methylation analysis of archival tissue specimens. J. Mol. Diagn. 12, 12. Patch, A. M. et al. Whole-genome characterization of chemoresistant ovarian 394–401 (2010). cancer. Nature 521, 489–494 (2015). 44. Bertotti, A. et al. The genomic landscape of response to EGFR blockade in 13. Cass, I. et al. BRCA-mutation-associated fallopian tube carcinoma: a distinct colorectal cancer. Nature 526, 263–267 (2015). clinical phenotype? Obstet. Gynecol. 106, 1327–1334 (2005). 45. Jones, S. et al. Personalized genomic analyses for cancer mutation discovery and 14. Piek, J. M. et al. BRCA1/2-related ovarian cancers are of tubal origin: a interpretation. Sci. Transl. Med. 7, 283ra53 (2015). hypothesis. Gynecol. Oncol. 90, 491 (2003). 46. Olshen, A. B., Venkatraman, E. S., Lucito, R. & Wigler, M. Circular binary 15. Piek, J. M. et al. Dysplastic changes in prophylactically removed Fallopian tubes segmentation for the analysis of array-based DNA copy number data. of women predisposed to developing ovarian cancer. J. Pathol. 195, 451–456 Biostatistics 5, 557–572 (2004). (2001). 47. Li, H. et al. The sequence alignment/map format and SAMtools. Bioinformatics 16. Medeiros, F. et al. The tubal fimbria is a preferred site for early adenocarcinoma 25, 2078–2079 (2009). in women with familial ovarian cancer syndrome. Am. J. Surg. Pathol. 30, 230–236 (2006). 17. Lee, Y. et al. A candidate precursor to serous carcinoma that originates in the Acknowledgements distal fallopian tube. J. Pathol. 211,26–35 (2007). We thank members of our laboratories for critical review of the manuscript. This work 18. Kindelberger, D. W. et al. Intraepithelial carcinoma of the fimbria and pelvic was supported by the Dr Miriam and Sheldon G. Adelson Medical Research Foundation serous carcinoma: evidence for a causal relationship. Am. J. Surg. Pathol. 31, (R.D. and V.E.V), Commonwealth Foundation (V.E.V.), US National Institutes of Health 161–169 (2007). grants CA121113 (V.E.V.), CA006973 (V.E.V.), CA083636 (R.D.), CA152990 (R.D.), 19. Kuhn, E. et al. TP53 mutations in serous tubal intraepithelial carcinoma and CA200469 (I.S.), US Department of Defense grant OCRP-OC-100517 (R.J.K and I.S.), concurrent pelvic high-grade serous carcinoma--evidence supporting the clonal the Honorable Tina Brozman Foundation for Ovarian Cancer Research (R.D.), the relationship of the two lesions. J. Pathol. 226, 421–426 (2012). SU2C-DCS International Translational Cancer Research Dream Team Grant (SU2C- 20. McDaniel, A. S. et al. Next-generation sequencing of tubal intraepithelial AACR-DT1415; V.E.V.), the Foundation for Women’s Wellness (R.D.), and the Richard carcinomas. JAMA Oncol. 1, 1128–1132 (2015). W. TeLinde Gynecologic Pathology Laboratory Endowment (I.S.). Stand Up To Cancer is 21. Bashashati, A. et al. Distinct evolutionary trajectories of primary high-grade a program of the Entertainment Industry Foundation administered by the American serous ovarian cancers revealed through spatial mutational profiling. J. Pathol. Association for Cancer Research. S.I.L-G. is a recipient of grants from Arthur Sachs/ 231,21–34 (2013). Fulbright/Harvard, La Fondation Philippe and La Fondation de France—“Recherche 22. Nik, N. N., Vang, R., Shih Ie, M. & Kurman, R. J. Origin and pathogenesis of clinique en cancérologie—Aide à la mobilité des chercheurs”. pelvic (ovarian, tubal, and primary peritoneal) serous carcinoma. Annu. Rev. Pathol. 9,27–45 (2014). 23. McPherson, A. et al. Divergent modes of clonal spread and intraperitoneal Author contributions mixing in high-grade serous ovarian cancer. Nat. Genet. 48, 758–767 (2016). S.I.L.-G, E.P., R.D., and V.E.V. were involved in the conception and design of this project. 24. Yachida, S. et al. Distant metastasis occurs late during the genetic evolution of S.I.L-G, E.P., V.A., N.N., R.K., R.D., and V.E.V developed the methodology. S.I.L.-G, E.P., pancreatic cancer. Nature 467, 1114–1117 (2010). V.A., M.N., M.N., M.S.H. D.I.L, L.S., C.L.M., J.-C.T, M.B., A.A., L.D.W, R.K., T.-L.W, 25. Niknafs, N., Beleva-Guthrie, V., Naiman, D. Q. & Karchin, R. SubClonal I.-M.S., R.D., and V.E.V. were involved in the acquisition of data, acquiring and hierarchy inference from somatic mutations: automatic reconstruction of managing patients, providing facilities and ect. S.I.L.-G, E.P., D.H., R.B., N.N., S.J., J.P., C. cancer evolutionary trees from multi-region next generation sequencing. PLoS A.H., R.B.S., R.K., R.D., and V.E.V. were involved in the analysis and interpretation of Comput. Biol. 11, e1004416 (2015). data through statistical analysis, biostatistics and computational analysis. S.I.L.-G, E.P., 10 NATURE COMMUNICATIONS 8: 1093 DOI: 10.1038/s41467-017-00962-1 www.nature.com/naturecommunications | | | NATURE COMMUNICATIONS | DOI: 10.1038/s41467-017-00962-1 ARTICLE D.H., R.D., and V.E.V. were involved in the writing, review and revision of the manu- Publisher's note: Springer Nature remains neutral with regard to jurisdictional claims in script. S.I.L.-G, E.P., D.H., R.B., N.N., S.J., J.P., R.B.S., R.K., R.D., and V.E.V. were published maps and institutional affiliations. involved with administrative, technical or material support by reporting or organizing data or construction databases. R.K., R.D., and V.E.V supervised the study. Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, Additional information adaptation, distribution and reproduction in any medium or format, as long as you give Supplementary Information accompanies this paper at doi:10.1038/s41467-017-00962-1. appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The images or other third party Competing interests: V.E.V. is a founder of Personal Genome Diagnostics and is a material in this article are included in the article’s Creative Commons license, unless member of its Scientific Advisory Board and Board of Directors. V.E.V. owns Personal indicated otherwise in a credit line to the material. If material is not included in the Genome Diagnostics stock, which is subject to certain restrictions under university policy. article’s Creative Commons license and your intended use is not permitted by statutory The terms of this arrangement is managed by the Johns Hopkins University in accordance regulation or exceeds the permitted use, you will need to obtain permission directly from with its conflict of interest policies. The remaining authors declare no competing financial the copyright holder. To view a copy of this license, visit http://creativecommons.org/ interests. licenses/by/4.0/. Reprints and permission information is available online at http://npg.nature.com/ reprintsandpermissions/ © The Author(s) 2017 NATURE COMMUNICATIONS 8: 1093 DOI: 10.1038/s41467-017-00962-1 www.nature.com/naturecommunications 11 | | | http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Nature Communications Springer Journals

Loading next page...
 
/lp/springer-journals/high-grade-serous-ovarian-carcinomas-originate-in-the-fallopian-tube-I56Gt30YQQ

References (51)

Publisher
Springer Journals
Copyright
Copyright © 2017 by The Author(s)
Subject
Science, Humanities and Social Sciences, multidisciplinary; Science, Humanities and Social Sciences, multidisciplinary; Science, multidisciplinary
eISSN
2041-1723
DOI
10.1038/s41467-017-00962-1
pmid
29061967
Publisher site
See Article on Publisher Site

Abstract

ARTICLE DOI: 10.1038/s41467-017-00962-1 OPEN High grade serous ovarian carcinomas originate in the fallopian tube 1,13 2,14 2 2,3 2 2 S. Intidhar Labidi-Galy , Eniko Papp , Dorothy Hallberg , Noushin Niknafs , Vilmos Adleff , Michael Noe , 2,4 1,14 5 2 2 6 Rohit Bhattacharya , Marian Novak , Siân Jones , Jillian Phallen , Carolyn A. Hruban , Michelle S. Hirsch , 6,15 7 1 8 6 Douglas I. Lin , Lauren Schwartz , Cecile L. Maire , Jean-Christophe Tille , Michaela Bowden , 9,10,11,12 2 2 2,12 2,12 Ayse Ayhan , Laura D. Wood , Robert B. Scharpf , Robert Kurman , Tian-Li Wang , 2,12 2,3 1,6,16 2 Ie-Ming Shih , Rachel Karchin , Ronny Drapkin & Victor E. Velculescu High-grade serous ovarian carcinoma (HGSOC) is the most frequent type of ovarian cancer and has a poor outcome. It has been proposed that fallopian tube cancers may be precursors of HGSOC but evolutionary evidence for this hypothesis has been limited. Here, we perform whole-exome sequence and copy number analyses of laser capture microdissected fallopian tube lesions (p53 signatures, serous tubal intraepithelial carcinomas (STICs), and fallopian tube carcinomas), ovarian cancers, and metastases from nine patients. The majority of tumor-specific alterations in ovarian cancers were present in STICs, including those affecting TP53, BRCA1, BRCA2 or PTEN. Evolutionary analyses reveal that p53 signatures and STICs are precursors of ovarian carcinoma and identify a window of 7 years between development of a STIC and initiation of ovarian carcinoma, with metastases following rapidly thereafter. Our results provide insights into the etiology of ovarian cancer and have implications for pre- vention, early detection and therapeutic intervention of this disease. 1 2 Department of Medical Oncology, Dana-Farber Cancer Institute and Harvard Medical School, Boston, MA 02215, USA. Sidney Kimmel Comprehensive Cancer Center, Johns Hopkins University School of Medicine, Baltimore, MD 21287, USA. Department of Biomedical Engineering, Institute for Computational Medicine, Johns Hopkins University, Baltimore, MD 21218, USA. Department of Computer Science, Institute for Computational Medicine, 5 6 Johns Hopkins University, Baltimore, MD 21218, USA. Personal Genome Diagnostics, Baltimore, MD 21224, USA. Department of Pathology, Brigham and Women’s hospital and Harvard Medical School, Boston, MA 02115, USA. Department of Pathology, University of Pennsylvania Perelman School of Medicine, Philadelphia, PA 19104, USA. Division of Clinical Pathology, Faculty of Medicine, Geneva University Hospital, 1205 Geneva, Switzerland. 9 10 Department of Pathology, Seirei Mikatahara Hospital, Hamamatsu 433-8558, Japan. Department of Tumor Pathology, Hamamatsu University School of Medicine, Hamamatsu 431-3192, Japan. Department of Molecular Pathology, Hiroshima University School of Medicine, Hiroshima 739-0046, Japan. Departments of Gynecology and Obstetrics and Pathology, Johns Hopkins University School of Medicine, Baltimore, MD 21287, USA. Present address: Department of Oncology, Geneva University Hospitals, Geneva 1205, Switzerland. Present address: Personal Genome Diagnostics, Baltimore, MD 21224, 15 16 USA. Present address: Department of Pathology, Beth Israel Deaconess Medical Center, Boston, MA 02215, USA. Present address: Department of Obstetrics and Gynecology, Penn Ovarian Cancer Research Center, University of Pennsylvania Perelman School of Medicine, Philadelphia, PA 19104, USA. S. Intidhar Labidi-Galy and Eniko Papp contributed equally to this work. Ronny Drapkin and Victor E. Velculescu jointly supervised this work. Correspondence and requests for materials should be addressed to R.D. (email: rdrapkin@pennmedicine.upenn.edu) or to V.E.V. (email: velculescu@jhmi.edu) NATURE COMMUNICATIONS 8: 1093 DOI: 10.1038/s41467-017-00962-1 www.nature.com/naturecommunications 1 | | | ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/s41467-017-00962-1 varian cancer is the leading cause of death from gyne- In addition, we analyzed isolated STIC lesions from four 1, 2 cologic cancers . The 10-year survival is < 30% and has patients (CGOV64, CGOV65, CGOV303, and CGOV304), three Onot improved significantly over the last 30 years . Despite of whom had germline pathogenic BRCA alterations and significant efforts, various screening and therapeutic strategies underwent prophylactic bilateral salpingo-oophorectomy, and a 4, 5 have generally not led to improved overall survival . One of the fourth who had bilateral salpingo-oophorectomy and hysterect- major challenges to improved diagnostic and therapeutic inter- omy in the context of a pelvic mass (Supplementary Data 1). For vention in ovarian cancer has been a limited understanding of the all patients, laser capture microdissection (LCM) was used to natural history of the disease. Ovarian carcinoma is a highly isolate lesions after immunohistochemistry (IHC) staining of p53 heterogeneous group of diseases including different histological in STICs and p53 signatures if these contained a TP53 missense subtypes with distinct clinicopathological and molecular genetic mutation or after hematoxylin staining if the samples contained a features that can be generally classified as Type I and Type II TP53 nonsense mutation (Fig. 1). All other samples were tumors . Among them, high-grade serous ovarian carcinoma microdissected after hematoxylin staining. Whole blood, normal (HGSOC, the major Type II tumor) is the most common histo- ovarian stroma, normal FT stroma, or normal cervix were used as logic subtype of ovarian cancer, accounting for three quarters of control samples. 7–10 ovarian carcinoma . Genomic analyses of HGSOC have To identify genetic alterations in the coding regions of these identified genetic alterations in TP53, BRCA1, BRCA2, PTEN, and cancers, we used next-generation sequencing platforms to other genes although few of these discoveries have affected clin- examine entire exomes in matched tumor and normal specimens 11, 12 ical care . HGSOC is diagnosed at advanced stages in ~70% of of all patients (Fig. 1). This approach allowed us to identify non- cases, and these women have a significantly worse outcome than synonymous and synonymous sequence changes, including those with early stage disease. Until recently, the prevailing view single base and small insertion or deletion mutations, as well of HGSOC was that it developed from the ovarian surface epi- as copy number alterations in coding genes. Given the thelium. However, early in situ lesions that arise from the ovarian challenges of exome-wide analyses of small tumor samples surface epithelium and progress to invasive HGSOC have never observed in STICs and p53 signature lesions, we developed been reproducibly identified. experimental and bioinformatic approaches for detection of Insights into the pathogenesis of HGSOC have emerged from somatic alterations from laser capture microdissected tissue. investigating the prevalence of occult ovarian and fallopian tube These included optimized approaches for microdissection of (FT) carcinomas in women with germline mutations of BRCA1/ STICs and p53 signatures after immunohistochemical staining, 13–17 BRCA2 genes . Potential precursor lesions of HGSOC were improved DNA recovery from laser captured material, library identified in the fimbriae of the FTs removed as part of pro- construction from limited and stained tissue samples, and error phylactic surgery . Such lesions, including a TP53 mutant single- correction methods in next-generation sequence analyses cell epithelial layer (p53 signature) and serous tubal intraepithelial (Methods section). The analyses of p53 signatures were 17, 18 carcinoma (STIC) , have been identified in patients with particularly challenging because these are extremely small advanced stage sporadic HGSOC of the ovary, FT and perito- lesions, representing 10–30 cells per section and less than neum . Immunohistochemical as well as targeted sequencing several hundred cells total that result in minute amounts (less analyses have shown that FT lesions harbor the same TP53 than a few ng) of isolated DNA. We optimized these approaches 17–21 mutation as surrounding invasive carcinomas . These ana- using a targeted next-generation sequencing approach analyzing lyses suggest a clonal relationship among such tumors but given 120 genes in a subset of samples from patient CGOV62, and the limited number of genes analyzed do not conclusively identify then used whole-exome analyses to evaluate coding sequence the initiating lesions nor exclude the possibility of FT metastases alterations in all samples (Supplementary Data 2–4). We 21, 22 from primary ovarian carcinomas . Yet additional studies obtained a total of 719 Gb of sequence data, resulting in an have evaluated clonal intraperitoneal spread of ovarian cancer average per-base sequence ~178-fold total coverage (~112-fold using whole genome analyses, but these efforts did not analyze distinct coverage) for each tumor analyzed by whole-exome precursor lesions such as STICs that may give rise to this sequencing (Supplementary Data 2). disease . In this study, we use exome-wide sequence and structural analyses of multiple tumor samples from the same individual to Analysis of sequence and structural changes. Whole-exome examine the origins of HGSOC. We have previously shown that sequence analyses of the tumor samples from each patient the acquisition of somatic alterations can be used as a molecular identified somatic mutations that were present in all neoplastic marker in the development of human cancer . Here, we examine samples analyzed as well as specific changes that were present in whether the compendium of somatic alterations identified in individual or subsets of tumors (Fig. 2). As expected, we identified different lesions may provide insights into the evolutionary sequence changes in the TP53 tumor suppressor gene, a well- relationship between primary FT lesions, including p53 signatures known driver gene in HGSOC, in all cases. The TP53 alterations and STIC lesions, ovarian carcinomas, and intraperitoneal were identical in all samples analyzed for each patient including metastases. in the p53 signatures, the STIC lesions, and other carcinomas. These data suggest that mutation of TP53 was among the earliest initiating events for HGSOC development as all lesions harbored Results this alteration. Overall approach. To elucidate the relationship among tumors in IHC staining for p53 did not identify any nuclear positive patients with HGSOC, we performed whole-exome sequencing of staining of p53 on the ovarian surface epithelium in any of the 37 samples from five patients diagnosed with sporadic HGSOC cases that had TP53 missense mutation, whereas all carcino- who underwent upfront debulking (Supplementary Data 1). This mas, STICs, and p53 signatures in the FT were positive. included STIC lesions, FT carcinomas, and ovarian cancers in all Whole-exome sequence analyses of normal ovarian stroma five patients; appendiceal, omental, or rectal metastases in three of (no p53 staining) microdissected from three patients patients (CGOV62, CGOV280, CGOV278); p53 signatures in two (CGOV64, CGOV65, CGOV280) did not find any genomic patients (CGOV62, CGOV63); and a STIC lesion in the con- abnormalities. Analysis of the resected tissues revealed tralateral FT from the affected ovarian cancer (CGOV280). that none of the nine cases had ovarian inclusion cysts. 2 NATURE COMMUNICATIONS 8: 1093 DOI: 10.1038/s41467-017-00962-1 www.nature.com/naturecommunications | | | NATURE COMMUNICATIONS | DOI: 10.1038/s41467-017-00962-1 ARTICLE p53 signature STIC Left FT tumor Omental metastasis Left ovarian tumor Appendiceal metastasis Left ovarian tumor Rectal metastasis Right ovarian tumor Identification of tumor cells Laser capture microdissection Infrared laser Attached Transfer captured pulse tumor cells tumor cells H&E p53 IHC p53 signature Whole-exome analyses from laser capture microdissected cells Evolutionary analysis Evolutionary SCHISM timeline Point Chromosomal mutations alterations Fig. 1 Schematic of sample isolation and next-generation sequencing analyses. (Top panel) Tumor sites analyzed from CGOV62 with stage III HGSOC. For each sample, slides were stained with hematoxylin and eosin as well as analyzed by immunohistochemical staining of p53. (Middle panel) Tumor samples were microdissected for genomic analyses. For microdissection for STIC and p53 signature lesions, tumor cells were identified using immunohistochemical staining of p53 and isolated through laser capture microdissection. (Bottom panel, left) Next-generation sequencing analyses were performed for tumor specimens using either whole-exome or targeted analyses. (Bottom panel, right) Somatic mutations and chromosomal alterations were used to evaluate tumor evolution using the tumor subclonality phylogenetic reconstruction algorithm SCHISM and to determine a timeline for tumor progression These observations suggest that there is no early lesion with analyzed sequence alterations in all samples with estimated tumor TP53 mutation in the surface epithelium or other normal purities > 50%, while four samples with tumor cellularities below regions within the ovary. this threshold (omental metastasis from CGOV279 and right Because TP53 mutations are expected to be clonal and were all ovarian tumor from CGOV278) or that were miliary carcinomas homozygous due to loss of heterozygosity (LOH) of the (rectal and sigmoidal metastases from CGOV63) were only remaining wild-type allele (as determined in our subsequent analyzed for structural changes. allelic imbalance analyses), we used the mutant allele fraction of Using a high-sensitivity mutation detection pipeline, we TP53 in each sample to estimate tumor purity. We further identified an average of 33 non-synonymous and synonymous NATURE COMMUNICATIONS 8: 1093 DOI: 10.1038/s41467-017-00962-1 www.nature.com/naturecommunications 3 | | | ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/s41467-017-00962-1 CGOV62 CGOV63 CGOV280 CGOV279 CGOV278 TP53:Y220C TP53:S90Lfs*59 TP53:Y126N TP53:chr17:7.6 CC>AG TP53:R273H AHDC1:N1218K ATP2B3:L606S KCNJ10:K335T ANKFN1:S172S ABCA13:H878Y BAHCC1:Q484E ATN1:Q485Q C12orf43:S223A ANKRD35: T908T ASB15:D503Y ATP13A3:A1175S COL6A5:K88E CAPZA3:N217I C5orf25:T635M C1orf173:Q897Q CREB3L1:E328D CD14:Q250Pfs*17 CLSPN:A1261S C5orf46:S37S CACNA1C:A1292D IGSF21:T251N CTNND2:G692R CSMD1:chr8:3 A>T KIAA1614:L83V ELL:L12L EPHA5:L1037F CCR2:E15X DGKBP12P NOL4:V28V EPPIN–WFDC6:Q43Q SI:chr3:166.2 T>C DNAH3:A3011T FAM135BT1242T OR8S1:R258C GPATCH1:R922W DUSP27:D880N SLC25A3:chr12:97.5 G>A HCK:G186S SLC13A3:W52X IGHMBP2:S508W FLG2:D64D SLC41A3:V71A TRIO:L111L KIF13A:T1034M KCNAB1:I46V TRPC7:E65X GPR158:D1002E KRT3:G133E WDR11:A371T LINGO4:V207A chr16:56.4–70.8 GPR25:L21L LMAN2:E235X ZNF71:E358D LRRC15:D125N chr16:77.8–90.1 MACF1:D242Y MACF1:A1296A INPP5E:S582S chr1:155.2–211.1 chr17:29.9–73 METAP2:chr12:95.9 A>T NID1:D405E ISM1:D350D chr10:1.2–129.8 chr17:9.8–21.1 MPZ:G213G RPS6KA5:A802P chr8:120.7–144.9 chr11:106.7–129.5 LPIN3:G623G NLRP5:G363V chr8:41.5–66.5 SEPT6:chrX:118.6 T>G chr11:82.3–102.7 LRP1:D2764D PTPRND793D GRIN2B:E47E SFMBT2:T599M RAD9A:A190A chr13:20.8–114 MS4A14:I56I OR10W1:A116A SILV:G100G SERINC2:A379A chr14:19.3–104.8 MTA3:V42L SV2C:C145C chr14:60.5–72.5 SMARCA4:V830V chr18:50.7–77.9 chr17:0.3–70.6 PCDHGB4:L408I chr14:91–101.9 chr2:98.3–109.1 TMPPE:S440C chr18:29.6–66.1 PDE2A:chr11:72.3 C>T chr16:17.1–65.5 chr1:1.2–16.4 chr3:0.4–13.4 PREX1:A283V chr11:67.1–134.2 chr20:31.1–54.6 chr16:70.6–87.4 chr8:75.9–95.5 SETBP1:L227L chr9:0.3–21 chr13:22.3–53.4 chr22:16.6–39.1 chr17:37.5–59.4 chr9:71.1–101 chr13:67.8–115.1 TMEM64:N378S chrX:44.9–154.2 chr4:154.7–177.3 chr2:135.2–153.5 chr15:39–101.9 chr4:178.5–189.3 TRHR:R328H APOB:K3630KN ARAP3:L1034H chr16:53.3–70.8 chr5:57.8–180.6 TRIML1:P433L BTNL9:P43PS C20orf194:L502L chr17:26.7–81.1 chr10:19–31.2 TTN:R25052H C1S:V225V DLG5:Q837Q chr17:0.6–19.7 chr10:42.9–93.7 XIRP2:L2419L EVC2:K450E chr18:21–34.9 C4orf43:K114N ACVR1C:G343E KCNQ2:N289N ZMIZ2:V393F chr18:43.3–77.9 C6:N343D LAMA5:R3419G C1orf174:A14E chr19:0.3–12.2 ZMYM6:P855P CDKAL1:T212T MB:G74G C2orf21:N97Y chr2:0.2–11.9 DNAH6:R1043H chr15:22.9–102.3 MCL1:E168Q chr21:11.1–48 C9orf152:Q107Q SP9:G170G DUSP27:V100I chr17:0.1–20.9 chr22:17.3–48.9 CA6:G102R VPS16:L162L chr17:26.7–72.9 chr3:29.5–52 FAM5C:M673I CAPRIN2:K307N chr11:16.1–46.4 chr4:4.2–190.9 chr18:0.2–77.9 KIF17:R769W CARD8:I13V chr18:7–19.3 chr5:137.4–178.4 chr19:30.3–41.4 chr2:69.7–96.8 KIF3C:V184I CDH5:V189L chr5:74.7–135.7 CEP135:R292X chr4:8.6–189.1 PRSS23:G347V chr8:0.2–33.4 CWC22:M550I chr6:122.8–168.5 RANBP2:M933I chr9:0.2–20.8 DLEC1:L305L chr6:0.3–29.8 chr9:27.5–140.8 RHBDD1:S5X DNAH3:R2256K chr6:39.9–51.7 chrX:2.8–154 ROBO1:S879S ERBB3:R103R chr6:156.3–170.9 chr6:62.4–121.6 SEMA3F:D65G FAM135B:A566A ACTL10:L219M chr7:1.1–18.8 SLC1A2:G257E HKDC1:R642G GJB1:D169D chr9:0.1–140.9 USP34:D1540H INTS9:T395T LCK:chr1:32.7 T>G chrX.23.9–153.7 ZBBX:A669S MAGEC1:S822X NFS1:R434P chrX:3–23.9 NPFFR1:G330G chr16:45.2–84.9 OR4F15:H6H RNF148:I163V ADAMTSL4:L1066L chr6:86.3–170.7 PHOX2B:M4T SETD1A:R18Q PTPRJ:A670S NALCN:R694R chr9:0.3–38.4 SH3PXD2B:R819Q RDH10:V218L SHPRH:R744Q chrX:9.8–40.4 SYT12:A280A SHH:N115I SLX4:L782L chr9:70.8–131.7 chr10:97.1–134.1 SLC17A6:S582X ST18:D162N chr20:4.8–14.8 SLC44A3:K44N ABCG8:P523P FOXG1:V264V SORCS3:S431L PCK2:Q216Q NOS1AP:S464S SZT2:V1975V STX11:G169A SETD2:S717L chr10:15.1–26.5 TARBP1:F918F NPAP1:V362V chr2:201.3–230.8 TTN:I18465N IQSEC1:S70S UNC13A:V1277M POU3F2:P326P UROC1:R173Q EFCAB4B:Y103Y ZNF462:T1671T GRM7:V153V chr4:120.6–154.7 PARN:F366F C16orf68:V124M ZFHX4:D2817D IGF2R:Y1094Y AADACL2:S205C ZNF804B:S710S ABI2:Q319K chr7:112.7–158.9 C7orf62:G195X MRPS12:F55F CLCC1:P182A REV3L:T903T p53 gene somatic mutation CSMD1:P1962L RHBDL3:G128G HTR1E:L90V CDC5L:K718K p53 sig somatic mutation KIAA1217:M1200I PPP1CC:R191R MUC7:chr4:71.4del p53 sig LOH STK36:P892P NPHS2:D190E CACNB4:A266A NRK:P611S STIC somatic mutation DNAH6:A2821A TMEM62:R554W EYA3:T223T STIC LOH TRIM35:R186Q GALNT16:G433G TRIP12:R798R STIC somatic mutation TWISTNB:K329R GP6:A71A ZNF345:H56Y GPRC5B:F308F STIC localized somatic mutation chr11:65.4–105.1 PER1:G310G FT somatic mutation SAMD4B:P465P ZKSCAN8:R472R FT LOH chr1:1.7–28.2 chr13:103.3–115 FT localized LOH chr13:21.1–102.4 chr22:17.3–32.6 Ovarian tumor somatic mutation Ovarian tumor LOH Ovarian tumor localized mutation Metastasis somatic mutation Metastasis LOH Mutation lost due to chr loss Fig. 2 Somatic mutation and allelic imbalance profiles among different tumor lesions. Somatic mutations and segments of allelic imbalance detected by whole-exome analyses are indicated as colored cells in rows for all patients. Darker shades of each color indicate somatic mutations while lighter shades indicate allelic imbalances. The tumor samples analyzed for each patient are indicated in columns (p53 sig, p53 signature; STIC, serous tubal intraepithelial carcinoma). For ovarian tumors in CGOV62 and STIC lesions in CGOV63 multiple blocks are indicated, including one ovarian tumor where multiple sections were analyzed after hematoxylin and eosin staining or after immunohistochemistry (IHC) staining of p53. These analyses indicated that staining methods did not affect detection of somatic alterations. The color of mutations indicates the degree of relatedness among tumor samples: red, shared among all tumor samples with TP53 highlighted at the top row; green, shared among all tumor samples except p53 signature lesion; purple, shared among fallopian tube tumor and omental metastasis; blue indicates mutations that were first detected in the ovarian tumors; and gray indicates mutations that were only detected in metastatic lesions. Additional color shades or patterns indicate mutations that are localized to specific lesions or lost due to chromosome loss as shown in the legend sequence alterations per tumor sample. Candidate alterations were distinguish the STIC lesions and FT carcinomas from ovarian evaluated across samples in an individual to determine if they cancers or intraperitoneal metastases. were present in multiple neoplastic lesions or were unique to a Given the importance of chromosomal instability in HGSOC , particular sample. To allow for the possibility that a subclone may we extended our analyses to examine structural variation in the have developed in a tumor lesion prior to becoming a dominant multiple tumors of each patient. We focused on regions of allelic clone at another location, we determined if genetic alterations that imbalance that can result from the complete loss of an allele were present in one tumor were also present in a low fraction of (LOH) or from an increase in copy number of one allele relative neoplastic cells of other lesions. This method required high to the other. We divided the genome into chromosome segments coverage of analyzed alterations in all samples and excluded and for each segment compared the minor allele (B-allele) potential artifacts related to mapping, sequencing or PCR errors, frequency values in tumor and normal samples using the ~17,000 allowing specific detection of alterations present in ≥ 1% of whole-exome germline heterozygous single-nucleotide poly- sequence reads (see Methods section for additional information). morphisms (SNPs) observed (Fig. 3, Supplementary Figs. 1–9 The composition of sequence alterations was relatively similar and Supplementary Data 7–11). Overall, we observed that an among the affected lesions of each patient. For example, for average of ~26% (range 12–39%) of the genome had chromoso- CGOV62, the STIC lesion, FT carcinomas, left and right ovarian mal imbalances in the samples analyzed (Fig. 3). cancers, and all three metastatic lesions harbored a common set Integration of sequence and structural alterations identified an of somatic mutations (Fig. 2). In CGOV63, CGOV279, and average of 47 alterations per sample (range 21–74) (Fig. 2). The CGOV278, while most of the sequence alterations were the same combination of both types of alterations allowed robust genomic among the tumors of each patient, a subset of mutations could differentiation between STICs and ovarian cancers or metastatic 4 NATURE COMMUNICATIONS 8: 1093 DOI: 10.1038/s41467-017-00962-1 www.nature.com/naturecommunications | | | p53 sig STIC Left FT tumor Left ovarian tumor A4 Left ovarian tumor A4 IHC Left ovarian tumor A7 Right ovarian tumor Rectal metastasis Appendiceal metastasis Omental metastasis p53 sig STIC D1 STIC D2 STIC D3 Right FT tumor D3 Right ovarian tumor Right FT STIC Right FT tumor Right ovarian tumor Omental metastasis Left FT STIC Right FT STIC Right FT STIC, tumor Right FT tumor Right ovarian tumor Left FT STIC Left ovarian tumor Omental metastasis NATURE COMMUNICATIONS | DOI: 10.1038/s41467-017-00962-1 ARTICLE CGOV62 CGOV63 CGOV280 CGOV279 CGOV278 0.00 0.00 0.00 0.00 0.00 1 1 1 1 0.05 0.05 0.05 0.05 0.05 2 2 2 2 0.10 3 3 3 0.10 3 0.10 3 0.10 0.10 4 4 4 4 4 0.15 0.15 0.15 0.15 0.15 5 5 5 5 5 6 6 6 6 6 0.20 0.20 0.20 0.20 0.20 7 7 7 7 7 8 8 8 8 8 0.25 0.25 0.25 0.25 0.25 9 9 9 9 9 10 10 10 10 0.30 0.30 0.30 0.30 0.30 11 11 11 11 11 12 12 12 12 12 0.35 0.35 0.35 0.35 0.35 13 13 13 13 13 14 14 14 14 14 15 15 15 15 15 0.40 0.40 0.40 0.40 0.40 16 16 16 16 16 17 17 17 17 17 18 18 18 18 18 19 19 19 19 0.45 0.45 0.45 20 0.45 20 20 0.45 20 21 21 21 21 21 22 22 22 22 X X X X X 0.50 0.50 0.50 0.50 0.50 Fig. 3 Genome-wide allelic imbalance profile. Minor allele frequency of heterozygous SNPs identified from normal tissue in each patient are derived in each tumor sample, enabling assessment of allelic imbalance in ~17,000 loci across the exome. Circular binary segmentation (CBS) is applied to minor allele frequencies of SNPs with minimum coverage of 10× in each tumor sample, and the resulting segment means are shown as a heatmap. Asterisks indicate samples where corresponding mutation analyses were not performed due to low tumor purity (omental metastasis of CGOV279, right ovarian tumor of CGOV278) or miliary pattern of tumor samples (peritoneal metastases of CGOV63). Given the relatively lower number of distinct DNA molecules available from the p53 signature samples from CGOV62 and CGOV63, these samples were subjected to a more sensitive LOH analysis (Methods, Genome-wide imbalance analysis) and are not shown here lesions in all patients analyzed. In patient CGOV62, a LOH of 9q A SCHISM tree node represents cells harboring a unique (70.8–131.7 Mb) provided a clear difference between the STIC compartment of mutations defining a subclone whereas an edge and all other carcinomas analyzed (Figs. 2 and 3). Likewise, represents a set of mutations acquired by the cells in the progeny chromosomal changes in 7q represented a distinguishing feature nodes that distinguish them from the cells in the parental node. between the right STIC or right FT tumors and the remaining By definition, for an individual cancer there could only be one lesions (ovarian cancers, omental metastasis, and left STIC) in parental clone, although there could be many different progeny CGOV280 (Figs. 2 and 3). In patient CGOV279, multiple regions subclones representing invasive or metastatic lesions or further of allelic imbalance were present in a STIC near the FT evolution of the primary tumor. The optimal hierarchy among carcinoma, while these were absent in a STIC that was not subclones is determined by examining all possible pairwise adjacent to this lesion. relationships between somatic alterations, and performing a heuristic search over the space of phylogenetic trees to identify a model that best explains the observed alterations. Evolutionary relationship of neoplastic lesions. As somatic In all samples, the SCHISM analysis of sequence and structural genetic alterations can be used to recreate the evolutionary history alterations suggested that the p53 signature or STIC lesions of tumor clones, we used the somatic sequence mutations and contained the ancestral clone for the observed cancers (Fig. 4). chromosomal alterations observed in each patient to determine This evolutionary relationship was strengthened by the observa- the history of tumor clonal evolution. We employed a subclone tion that nearly all of the alterations within the p53 signature and hierarchy inference tool called SCHISM (SubClonal Hierarchy STIC lesions were shared by all other lesions. For example, the Inference from Somatic Mutations) which enables improved ovarian tumors of all cases displayed alterations that were shared phylogenetic reconstruction by incorporating estimates of the in FT lesions but also contained additional changes, suggesting fraction of neoplastic cells in which a mutation occurs (mutation that these represented daughter clones of the latter tumors cellularity) . We estimated the cellularity of each mutation by (Fig. 2). Likewise, the ovarian cancers or their immediate correcting the observed allele frequencies for tumor purity and precursors were likely the direct parental clones for the metastases copy number levels (Methods section). In addition to the in CGOV62, CGOV278, and CGOV280 as demonstrated by the observed structural alterations, this approach allowed us to use shared alterations that were not contained in earlier FT lesions. 213 synonymous and non-synonymous somatic sequence Overall, the phylogenetic model generated by these data suggests alterations to construct the phylogenetic trees illustrated in Fig. 4 a progression from FT epithelium to p53 signatures and to STIC and Supplementary Data 5. lesions which are then precursors of FT carcinoma, ovarian NATURE COMMUNICATIONS 8: 1093 DOI: 10.1038/s41467-017-00962-1 www.nature.com/naturecommunications 5 | | | STIC Left FT tumor Left ovarian tumor (A4) Left ovarian tumor (A4) - IHC Left ovarian tumor (A7) Right ovarian tumor Rectal Metastasis Appendiceal metastasis Omental metastasis STIC D1 STIC D2 STIC D3 Right FT tumor D3 Right ovarian tumor Rectal metastasis* Sigmoidal metastasis* Right FT STIC Right FT Tumor Right ovarian tumor Omental metastasis Left FT STIC Right FT STIC Right FT STIC, near tumor Right FT Tumor Right ovarian tumor Omental Metastasis* Left FT STIC Left ovarian tumor Omental metastasis Right ovarian tumor* ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/s41467-017-00962-1 CGOV62 CGOV63 CGOV280 CGOV279 CGOV278 Normal FT Normal FT Normal FT Normal FT Normal FT epithelium epithelium epithelium epithelium epithelium TP53 TP53 TP53 TP53 TP53 SI SFMBT2 TTN IGSF21 MACF1 Chr17 LOH Chr17 LOH Chr17 LOH Chr17 LOH Chr17 LOH Precursor to right p53 signature p53 signature Right FT STIC FT STIC and tumor Left FT STIC APOB TTN IQSEC1 GRM7 Chr18q LOH SETD2 SETD1A Chr16 LOH Chr4 LOH Chr10q LOH Right Right FT tumor STIC D1,3, right FT FT STIC STIC tumor LAMA5 Chr2 LOH Chr7 LOH Chr18p LOH SZT2 KIAA1217 IGF2R Precursor to ovarian ABCG8 Chr10p LOH Chr11 LOH Chr9 LOH tumor, omental met and Left FT STIC PER1 Chr1 LOH REV3L Right ovarian Right FT STIC tumor (near tumor), Left ovarian tumor Omental metastasis Left FT tumor, STIC D2 Right ovarian tumor right FT tumor Omental metastasis Left FT STIC ovarian tumors, metastases CDC5L Right ovarian tumor Fig. 4 Schematic of tumor evolution. The history of tumor evolution in each patient is modeled as a subclonal hierarchy inferred from the somatic mutations and large scale genomic regions harboring loss of heterozygosity (LOH features) using the SCHISM framework, and is depicted as a tree. Each tree starts from a root node corresponding to the normal fallopian tube epithelium (germline). In all patients, mutations in TP53 (red boxes) are among the earliest somatic alterations and are ubiquitously present in all tumor samples. Somatic alterations (boxes) are acquired along edges (arrows) of the tree, and example alterations are indicated in each case. Nodes of the tree represent cells whose genotype is described by the presence of somatic mutations and LOH features on the path connecting the node to the root of the tree. Each node is labeled with tumor samples harboring all upstream and lacking any downstream alterations. The trees inferred for all patients support a pattern of evolution with p53 signatures and STIC lesions as early events in tumorigenesis. Mutation clusters and LOH feature groups follow the same color code as Fig. 2 carcinoma, and metastatic lesions. In addition to the sequential of these patients (BRCA1 Q1200X, BRCA2 L2653P, and a BRCA2 accumulation of alterations in this linear evolution, we also 55 kb hemizygous deletion in CGOV65, CGOV64, and observed branching phylogenetic trees due to continued evolution CGOV304, respectively), as well as somatic mutations in TP53, within STIC lesions as well as FT carcinomas and ovarian and LOH of both chromosome 13 and 17, encompassing the carcinomas (Fig. 4). We compared evolutionary trees resulting BRCA1, BRCA2, and TP53 loci in all of these cases (Supple- from SCHISM analysis with those derived by maximum mentary Figs. 6, 7, 8, and 9). Whole-exome analyses showed that parsimony phylogeny using PHYLIP and the results were similar the STIC lesions contained a total of 91, 23, 34, and 46 non- in all cases (Fig. 4 and Supplementary Fig. 11). synonymous and synonymous somatic mutations, in CGOV65, Interestingly, patient CGOV280 had a right STIC, a right CGOV64, CGOV303, and CGOV304, respectively. Overall, these fallopian carcinoma, and a right ovarian cancer but also had a analyses revealed that STICs in isolation in patients with or STIC in the left FT (Supplementary Fig. 5). In this case the without germline BRCA changes have a roughly similar number SCHISM analysis suggested that the lesion in the left FT which of sequence changes to STICs in patients with sporadic tumors. was pathologically determined to be a STIC actually represented a These observations provide evidence that isolated STICs may act metastatic lesion of the right ovarian cancer (Fig. 4). This lesion as precursors in the same manner as those identified in patients shared nearly all the alterations of the ovarian cancer but with sporadic advanced stages HGSOC analyzed in this study. contained 10 single base substitutions and four additional regions of allelic imbalance on chromosomes 1, 13, and 22, and both the left STIC and right ovarian cancer had an additional region of Recurrent molecular alterations. We examined tumors from the allelic imbalance on chromosome 7 that was absent in the right nine patients to identify recurrent non-silent sequence or chro- STIC (Figs. 2 and 3). These observations are consistent with the mosomal changes. Although no genes other than TP53 were above model of STIC to ovarian cancer progression, but suggest mutated in all patients analyzed, we identified mutations in ten that in advanced disease ovarian cancers may also seed metastatic genes that were altered in two or more patients (Supplementary deposits throughout the peritoneum, including to the FT on the Data 6). These included mutations in the tumors of two patients of the PIK3R5 gene that encodes a regulatory subunit of the PI3- contralateral side. kinase complex. CGOV64 also had a somatic alteration in PTEN that together with changes in PIK3R5 highlight the importance of Genomic alterations in isolated STICs. Neoplastic cells observed the PI3K pathway in ovarian cancer . Additional genes that were in the FTs rather than the ovaries removed from carriers of observed to be altered in other ovarian cancers through other germline mutation of BRCA1 and BRCA2 provided the first large scale sequencing efforts such as TCGA are indicated in 15, 26 indication of the FT as a potential cell of origin of HGSOC . Supplementary Data 6. Since < 1.25% of HGSOC are diagnosed with stage I disease , In addition to recurrent sequence changes, we found altera- BRCA carriers provide a unique opportunity to analyze genomic tions in regions of allelic imbalances encompassing several tumor alterations in isolated STICs without associated HGSOC. We suppressor genes involved in ovarian cancer. Remarkably, these examined neoplastic samples from three individuals with germ- included losses of BRCA1, BRCA2, and TP53 in all nine patients, line BRCA alterations where STIC lesions were incidentally and loss of PTEN for CGOV62, CGOV63, CGOV280, and identified after prophylactic bilateral salpingo-oophorectomy, and CGOV64 (in addition to the somatic sequence alterations of these one patient where two STICs were identified after resection of a genes) (Supplementary Figs. 1–9). In all cases, the LOH observed pelvic mass (Supplementary Data 1). We identified BRCA1 or in the metastatic lesions and ovarian tumor lesions for regions BRCA2 sequence alterations or deletions in the germline of three encompassing these genes were already present in the FT tumor 6 NATURE COMMUNICATIONS 8: 1093 DOI: 10.1038/s41467-017-00962-1 www.nature.com/naturecommunications | | | NATURE COMMUNICATIONS | DOI: 10.1038/s41467-017-00962-1 ARTICLE and STIC lesions. Considering the evolutionary model above, (including STICs or p53 signatures) are infrequent and unlikely these data suggest that a combination of sequence changes in a to be the source of most FT lesions. few genes including TP53 together with loss of the TP53 wild-type Despite these potential limitations, the data we have obtained allele as well as BRCA1, BRCA2, and PTEN may be crucial early provide new insights into the etiology of ovarian Type II carci- 27, 28 events that are needed for the initiation of STICs . noma and have significant implications for the prevention, early detection and therapeutic intervention of this disease. The results suggest that ovarian cancer is a disease of the FTs, with the Evolutionary timeline of ovarian cancer development. To esti- development of p53 signatures and STICs as early events. The mate the time between the development of the earliest neoplastic subsequent formation of a cancer in the ovaries represents a clones in the FT and the development of ovarian and other seeding event from a primary tumor in the FT that already metastatic lesions we used a mathematical model for comparative contains sequence and structural alterations in key driver genes, 24, 29 lesion analysis . This model estimates the time interval including those in TP53, PI3K pathway, and BRCA1/BRCA2 between a founder cell of a tumor of interest and the ancestral genes. The recurrent allelic imbalances observed in chromosomes precursor cell assuming that mutation rates and cell division 1, 6, 16, 18, 20, and 22 may suggest additional genes that are times are constant throughout a patient’s life (Methods section). involved in this process. The timing of the progression from In patient CGOV62, this model would suggest ~1.9 years between STICs to ovarian cancer in the cases we analyzed was on average the development of the STIC lesion and the ovarian cancer (90% 6.5 years, but seeding of metastatic lesions in these patients CI, 0.5–4.2 years). For other patients this transition appears to occurred rapidly thereafter. This timing is consistent with recent have been slower as the average time between STICs and ovarian reports showing a difference of 7.7 years in the age of BRCA cancer among all patients was 6.5 years (1.4–10.7 years). carriers with localized vs. advanced adnexal lesions .This evo- Importantly, in patients with metastatic lesions, the time between lutionary timeline can help explain why most HGSOC patients are the initiation of the ovarian carcinoma and development of diagnosed at advanced stage (III/IV) with pelvic and peritoneal metastases appears to have been rapid (average 2 years). There spread of disease, and why among asymptomatic BRCA germline were either no additional mutations in metastatic lesions (e.g., mutation carriers half of the cases diagnosed with asymptomatic CGOV62 omental, rectal or appendiceal metastasis or CGOV280 adnexal neoplasia have already seeded to pelvis or peritoneum (> omental metastasis) or the number of additional changes was IA) . These observations are largely similar to other genomic 19, 20, 23, 32 small (e.g., three changes in CGOV278 omental metastasis), analyses of the evolution of ovarian cancer as well as the reflecting the ease with which cancer cells located on the ovaries recent analyses of STIC lesions that were reported while this can subsequently seed additional peritoneal sites. Although the study was under review . Our study highlights the role of precise timing of this progression depends on assumptions related p53 signatures as early lesions in this evolutionary paradigm. to mutation rates, which may change during tumor progression, Our genomic analyses are consistent with population-based models employing different rates all showed longer timeline from studies of the effects of salpingectomy on the risk of ovarian STIC lesions to ovarian tumors followed by rapid development of cancer. Prophylactic bilateral salpingo-oophorectomy has been metastatic lesions (Methods section). shown to reduce the risk of developing ovarian cancer in BRCA 34, 35 mutation carriers to below 5% . Likewise, bilateral sal- pingectomy, performed as a contraceptive method instead of Discussion tubal sterilization, reduced the risk of ovarian cancer by 61% at 10 years . Our study provides a mechanistic basis for these obser- These results provide a comprehensive evolutionary analysis of sporadic HGSOC in five patients. Given the unique nature of the vations and has implications for clinical management in pre- multiple samples we examined from each patient, our study may vention of ovarian cancer. In high risk BRCA carriers, bilateral have certain limitations not typical of genome-wide efforts. First, salpingectomy with delayed oophorectomy should be con- the small size of the tumor samples compared to surrounding sidered through participation in ongoing clinical trials non-neoplastic tissue could potentially lead to low tumor purity. (NCT02321228; NCT01907789). In non-carriers, our work The high mutant allele fraction of TP53 among cancer samples implies that for women who undergo surgery for benign uterine (average of 56–85%) indicates that this issue was largely over- causes, total abdominal hysterectomy and bilateral salpingectomy come through LCM. Second, the small number of cells in with sparing of the ovaries should be considered , and that p53 signature samples may have limited our genomic analyses for bilateral salpingectomy may be a preferred contraceptive alter- these lesions. The observation that all sequence changes in native to tubal ligation. The dual concepts in these recommen- p53 signatures were also present in STIC and other carcinomas of dations for BRCA carriers and non-carriers are that removal of multiple sites is consistent with our evolutionary model and the FTs (rather than the ovaries) may be curative as it eliminates suggests that these cells are likely to represent a parental clone of the underlying cellular precursors of ovarian cancer, and that other neoplastic lesions. Third, our analysis was limited to preservation of the ovaries provides long term benefits due to ovarian cancers where STICs and other concomitant lesions were decreased risk and fatalities from coronary heart disease and identified, and may therefore not be representative of all HGOCs. other illnesses . A limitation of this approach is that as the The absence of STIC lesions in ~40% of sporadic HGSOCs is precise timing of when potentially malignant cells shed from the likely due to an incomplete sampling of the FT or the overgrowth FT and microscopically seed the ovary is unknown, removal of of the STIC by the carcinoma in the context of bulky disease, but the tubes may not provide optimal risk-reduction. may also reflect another site of origin that has yet to be deter- Our observations also have implications for improved detec- mined for these cancers . Fourth, this study did not intend to tion of ovarian cancer. Unfortunately, < 1.25% of HGSOC are address the intra-tumoral heterogeneity within the carcinomas confined to the ovary at diagnosis . Earlier detection of this but rather focused on clonal changes within each tumor. Fifth, as disease is likely to benefit from the identification of a precursor in any evolutionary analyses, the genomic alterations we observed lesion, as has been the case for many other tumor types. Our data provide the most likely model of tumor development but do not suggest that FT neoplasia is the origin of ovarian serous carci- exclude the possibility of other relationships. Nevertheless, our nogenesis, and can directly lead to cancer of the ovaries and of analyses of somatic alterations suggest that models where the other sites. Currently, the typical histopathologic evaluation of 20, 21 ovarian cancer or metastatic lesions seed the FT tumors FTs typically involves a cursory evaluation of one or two NATURE COMMUNICATIONS 8: 1093 DOI: 10.1038/s41467-017-00962-1 www.nature.com/naturecommunications 7 | | | ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/s41467-017-00962-1 representative sections. Our study suggests that systematic sec- Sample preparation and next-generation sequencing. DNA was extracted from patient whole blood using a QIAamp DNA Blood Mini QIAcube Kit (Qiagen tioning and extensive examination of total FTs should become Valencia, CA). Genomic DNA from FFPE blocks was extracted from the micro- common practice in pathology, and not confined to academic dissected tissues using the QIAamp DNA FFPE Tissue kit (Qiagen, Valencia, CA). tertiary care centers. Depending on whether the FTs are removed In brief, the samples were incubated in proteinase K for 16 h before DNA for benign conditions, risk-reducing bilateral salpingectomy, or extraction. The digested mixture was transferred to a microtube for DNA frag- mentation using the truXTRAC™ FFPE DNA Kit with 10 min shearing time as per gynecological cancers, specific examination protocols should be 16, 40 the manufacturer’s instructions (Covaris, Woburn, MA). Following fragmentation, applied . Given the window of time that appears to exist the sample was further digested for 24 h followed by 1 h incubation at 80 °C. DNA between the formation of FT lesions and development of ovarian purification was performed using the QIAamp DNA FFPE Tissue kit following the cancer, these insights open the prospect of novel approaches for manufacturer’s instructions (Qiagen, Valencia, CA). Fragmented genomic DNA from tumor and normal samples were used for Illumina TruSeq library con- screening. Such approaches may be especially important given the struction (Illumina, San Diego, CA) according to the manufacturer’s instructions limited therapeutic options currently available for ovarian can- or as previously described . Exonic or targeted regions were captured in solution 4, 5 cer . Recent advances for ultrasensitive detection of genetic using the Agilent SureSelect v.4 kit or a custom targeted panel according to the alterations in blood-based liquid biopsies, pap smears, and other manufacturer’s instructions (Agilent, Santa Clara, CA). Paired-end sequencing, 41, 42 bodily fluids , or imaging approaches may provide opportu- resulting in 100 bases from each end of the fragments for exome libraries and 150 bases from each end of the fragment for targeted libraries, was performed using nities in early diagnosis and intervention. Illumina HiSeq 2000/2500 and Illumina MiSeq instrumentation (Illumina, San Diego, CA). Methods Next-generation sequencing data and identification of somatic mutations. Specimens obtained for sequencing analysis. The study was approved by the Somatic mutations were identified using VariantDx custom software for identi- Institutional Review Board at Brigham and Women’s Hospital and the Johns fying mutations in matched tumor and normal samples. Prior to mutation calling, Hopkins Hospital and all patients gave informed consent before inclusion. Five primary processing of sequence data for both tumor and normal samples were sequential patients with stage III sporadic HGSOC, in whom a STIC was identified performed using Illumina CASAVA software (v1.8), including masking of adapter in their FTs (FT), were included. In addition, we included isolated STICs from sequences. Sequence reads were aligned against the human reference genome three patients with germline BRCA deleterious alterations who underwent pro- (version hg18 or hg19) using ELAND. Candidate somatic mutations, consisting of phylactic bilateral salpingo-oophorectomy as well as a fourth patient who had point mutations, insertions, and deletions were then identified using VariantDx bilateral salpingo-oophorectomy and hysterectomy in the context of a pelvic mass. across either the whole exome or regions of interest . For samples analyzed using All cases underwent complete tubal examination using the SEE-FIM protocol . targeted sequencing, we identified candidate mutations that were altered in > 10% Formalin-fixed paraffin embedded (FFPE) blocks were retrieved from the pathol- of distinct reads. For samples analyzed using whole-exome sequencing, we iden- ogy files at Brigham and Women’s Hospital and Johns Hopkins Hospital within the tified candidate mutations that were altered in > 10% of distinct reads with ≥ 5 3 months following surgical diagnosis and stored at 4 °C to slow down nucleic acids altered reads in at least one sample, where coverage at the altered base was at least degradation. All the cases were reviewed by a gynecologic pathologist (M.S.H., D.I. as high as the TP53 alteration in that sample, and where the ratio of the coverage of L., L.S.) that confirmed the diagnosis of STIC and/or p53 signature in the FT. Slides the mutated base to the overall sequence coverage of that sample was > 20%. from each FFPE block, including early lesions, invasisve carcinomas and metas- Identified mutations were reported as present in other samples of the same patient tases, were stained with hematoxylin and eosin, and analyzed by p53 IHC staining. if the mutation was present in at least two distinct altered reads. Mutations present In each FT, at least one STIC and/or p53 signature was identified and micro- in polyN tract ≥ 5 bases, or those with an average distinct coverage below 50× were dissected separately. Importantly, STICs were not pooled together if they were in removed from the analysis. the same section and were considered separate STICs. An analysis of each candidate mutated region was performed using BLAT. For each mutation, 101 bases including 50 bases 5ʹ and 3ʹ flanking the mutated base was used as query sequence (http://genome.ucsc.edu/cgi-bin/hgBlat). Candidate Immunohistochemistry and laser capture microdissection. For accurate mutations were removed from further analysis, if the analyzed region resulted in microdissection of early lesions including STIC and p53 signature, IHC staining of > 1 BLAT hits with 90% identity over 70 SCORE sequence length. All candidate p53 was specifically adapted for LCM as previously described . PEN membrane alterations were verified by visual inspection. frame slides Arcturus (Life technologies, Carlsbad, CA) were used. Each slide was coated with 350 ul of undiluted poly-L-lysine 0.1% w/v (Sigma, St. Louis, MO). For drying, the slides were placed in a slide holder for 60 min at room temperature. Genome-wide allelic imbalance analysis. We performed comparative analysis of Tissue sections were cut and mounted on the pretreated membrane slides. LOH across the tumor samples from each patient to identify copy number Deparaffinization was performed in fresh xylene for 5 min twice, followed by 100% alterations occurring in the course of tumor evolution. Minor allele frequency ethanol for 2 min, 95% for ethanol 2 min, and 70% ethanol for 2 min. Subsequently, (MAF) of germline heterozygous SNPs with minimum coverage of 10× in each the slides were transferred into distilled water for 5 min. Heat-epitope antigen tumor sample were segmented using circular binary segmentation algorithm retrieval (AR) was performed in Citrate Buffer (Dako, Carpinteria, CA) at low (CBS) . Genomic segments where the difference between tumor and normal MAF temperature (60 °C) for 44 h instead of 120 °C for 10 min to reduce tissue and DNA exceeded a threshold of 0.10 were labeled as harboring LOH. In each tumor sample, damage by high temperature. Retrieval solution was pre-warmed to 60 °C before the minimum MAF across segments with minimum size of 10 Mb was calculated to usage. After incubation in the oven, the AR solution was left to cool down to room provide a measure of sample purity. Each segment marked as LOH was assigned to temperature and the slides were rinsed for 30 seconds in fresh 1×PBS then incu- one of the three confidence categories: (1) high confidence, segment MAF within bated for 40 min with primary antibody anti-p53 (Epitomics, Burlingame) at 1:100 0.1 of the minimum sample MAF. (2) Intermediate confidence, segment MAF in a humidifying chamber. Before adding the secondary antibody, slides were within 0.1–0.2 of the minimum sample MAF. (3) Low confidence, segment MAF washed twice for 1 min in fresh 1×PBS. The secondary antibody, labeled polymer- exceeding the minimum sample MAF by > 0.2. HRP anti-mouse (Dako EnVision System-HRP (DAB), Carpinteria, CA) was Next, sample level segments were intersected across the entire set of samples applied for 30 min. Then, slides were washed twice for 1 min in fresh 1×PBS. from each patient to derive patient level segments while accounting for the Chromogenic labeling was performed with 3,3-DAB substrate buffer and DAB possibility of variable segment break points in different samples (Supplementary chromogen (Dako EnVision System-HRP (DAB), Carpinteria, CA) for 5 min. Data 7–11). Patient level segments were filtered to keep those covering a minimum Slides were washed again for 30 s in fresh distilled water. Dehydration was per- of 20 SNPs and with minimum length of 10 Mb. The resulting segments were formed as follows: 70% ethanol for 30 s, 95% ethanol for 30 s, 100% ethanol for 30 further narrowed down to only include those with high confidence LOH in at least s, and xylene for 30 s. The stained slides were microdissected within 2 h with the one of the samples. Genomic segments with LOH in a subset of samples can serve Arcturus XT LCM system (Life technologies, Carlsbad, CA). as informative markers to track tumor evolution similar to somatic mutations. To increase the specificity in identifying this class of genomic segments, we required a minimum distance of 0.1 between the MAF of samples with and without LOH. To Hematoxylin staining for laser capture microdissection. Invasive carcinomas minimize the possibility of over-segmentation which could result in inflated from the ovaries, the FTs and intraperitoneal metastases or STICs from patients estimates of the number independent structural alterations, we evaluated patient with negative p53 IHC staining were microdissected after Hematoxylin staining. level segments with boundaries within a 5 Mb window. In cases where the LOH Briefly, deparaffinization was performed in fresh xylene for 1 min twice followed by calls were identical and the difference of segment MAFs were ≤0.05 in all tumor 100% ethanol for 1 min, 95% for ethanol 1 min, and 70% ethanol for 1 min. The samples, the segments were merged. slides were transferred into distilled water for 2 min before staining with Hema- For CGOV62 and CGOV63, the number of germline heterozygous SNPs toxylin for 2 min. Subsequently, slides were rinsed in distilled water until they meeting the coverage criteria in p53 signature samples was significantly lower than became clear before undergoing dehydration in 70% ethanol for 1 min, 95% the other samples from the same patient. Thus, we modified the approach above in ethanol for 1 min, 100% ethanol for 1 min, and xylene for 1 min. The stained slides these two patients to enable sensitive analysis of LOH in p53 signature samples. were microdissected within 2 h. Initially, the patient level genomic segments of interest were defined excluding 8 NATURE COMMUNICATIONS 8: 1093 DOI: 10.1038/s41467-017-00962-1 www.nature.com/naturecommunications | | | NATURE COMMUNICATIONS | DOI: 10.1038/s41467-017-00962-1 ARTICLE p53 signature samples. Next, in each genomic segment, the minor allele of each we can estimate mutation cellularity (or cancer cell fraction) from the observed overlapping germline SNP was determined by taking a majority vote over their reference and alternate read counts, and estimates of copy number, and tumor minor alleles in the other samples. The coverage and minor allele read count for purity as follows. each SNP was derived using samtools (v0.1.19) mpileup module . The segment MAF in p53 signature samples were calculated by dividing the sum of minor allele mαC V ¼ exp read counts across all SNPs by the total coverage of SNPs, circumventing the α CN þðÞ 1  α CN T N variance resulting from low coverage at individual SNPs. In each p53 signature sample, segments with MAF lower than that of the normal by at least 0.1 were In the equation above, V is the expected variant allele frequency of the exp marked as LOH. mutation, m is the multiplicity of the mutation which is set to 1, α is the purity of the tumor sample, C is the cellularity of the mutation, and CN and CN are the T N integer copy number of tumor and normal sample at the locus of the mutation. The Copy number analysis. The genome-wide copy number profiles were determined observed alternate read count of the mutation can be modeled as a binomial by analysis of the ratio of read counts in the tumor and matched normal whole- random variable drawn from a distribution with probability parameter equal to exome sequenced samples. In each sample, the number of reads mapping to V and number of trials equal to the total sequence coverage of the mutation. We exp genomic bins located in target and off-target regions were corrected for biases calculated the likelihood for observation of the alternate read counts for cellularity arising from GC-content, repetitive sequences, and target capture process using values spanning the range of 0–1 in increments of 0.01, and derived the maximum CNVKit (v.0.7.6) (https://doi.org/10.1371/journal.pcbi.1004873). The log ratio of likelihood estimate and confidence interval for the mutation cellularity. the processed tumor to normal read counts provides a measure of copy number in To obtain reliable estimates of mutation cellularity, we clustered mutations by each bin, and was segmented to yield genomic intervals at constant copy number joint presence or absence across the available tumor samples. This approach makes levels. The difference in sequencing library size between the tumor and normal phylogenetic reconstruction more tractable and the cellularity of the resulting samples is another factor that needs to be accounted for when analyzing reads clusters can be estimated with higher accuracy than that of individual mutations. ratios in NGS-based copy number pipelines. In CNVKit, the log ratio values in For each patient, a mutation was called as present or absent in each of the available each sample are adjusted by setting the median of autosomal bins to 0 in log space, tumor samples (10 samples from CGOV62, 6 samples from CGOV63, 5 samples assuming a median ploidy of 2 for the genome. Given the high prevalence of copy from CGOV280, 4 samples from CGOV279, and 3 samples from CGOV278). To number aberrations in ovarian cancer and the high frequency of allelic imbalance call the mutation present, we used a minimum allele frequency of 2% and 2 distinct in the present cohort, this assumption may not be accurate, and will manifest itself mutant reads. Mutation clustering was performed by a greedy algorithm. Tumor as a genome-wide bias or shift of log ratio values. purity in each tumor sample was estimated as the read count fraction of TP53 Therefore, an alternative approach for normalization of log ratio values was mutation in each patient. Each patient harbored a single distinct TP53 mutation adopted, which takes into account the level of allelic imbalance in each genomic that was present in all tumor samples, and we assumed the wild-type allele was lost, region. Briefly, genomic regions with the least degree of allelic imbalance were as supported by the ubiquitous LOH of chromosome 17. To derive a more identified in each tumor sample, and used in a normalization process based on the comprehensive view of the evolution of these samples, we extended the original notion that these regions can only be present in an even number of copies. The SCHISM framework to model acquisition of large scale somatic copy number distribution of log ratio values among these regions was inspected to ensure that alterations, which can be detected by analysis of allelic imbalance (including LOH). they belong to the same copy number level. Otherwise, a subset of regions at a First, we extracted a set of high confidence genomic regions with ubiquitous, common log ratio (and thus copy number) level were selected. By fixing the copy partially shared, or private LOH in tumor samples of each patient (Methods number of these segments at a specified level, one can solve for the genome-wide section). These regions of LOH served as binary features that could be used for bias of log ratio values as follows, and thus identify the genome-wide integer copy evolutionary analysis, and were clustered into LOH feature groups with identical number profile. patterns of presence or absence across samples (Fig. 2). Each LOH feature group was compared to the somatic mutation clusters in each patient, with respect to its α CN þðÞ 1  α CN T N R ¼ log  δ pattern of presence or absence across samples. In cases where a mutation cluster with the identical pattern could be found, the cluster and the LOH feature group In the equation above, R represents the observed log ratio of read counts, α is were assumed to have occurred together in the course of tumor evolution. the purity of the tumor sample, CN and CN are the integer copy number of Otherwise, the LOH feature groups were modeled as distinct features, and added in T N tumor and normal samples at a locus, and δ is the genome-wide bias term. Given post-hoc analysis by application of the lineage precedence rule from SCHISM; the value of tumor purity and copy number, δ is the only unknown in the equation. which requires cellularity of ancestor alterations to be greater than or equal to To favor solutions with less complex genomes, the copy number of regions with cellularity of descendant alterations in all tumor samples. complete allelic balance was initially set to 2. If the resulting solution was deemed SCHISM was run with the above inputs and default parameter settings to infer implausible (e.g., by implying chromosome or chromosome arm scale homozygous the order of somatic alterations and thus define subclonal hierarchy in each patient. deletions), the copy number of regions with complete allelic balance was assigned SCHISM software is freely available for non-profit use at http://karchinlab.org/ to 4 and an alternative solution was found (Supplementary Fig. 10). appSchism. Details of the genomic segments selected to solve for the genome-wide bias term Evolutionary trees resulting from SCHISM analysis were compared with those δ are as follows. In CGOV62, chromosomes 4 and 12 did not have allelic imbalance derived by maximum parsimony phylogeny using PHYLIP (Phylip-3.695, PARS in any tumor samples. The solution assigning copy number two to these regions method). For CGOV280, an adjustment to the tree was applied to account for implied homozygous deletion of the p-arm of chrX in multiple samples; therefore, multiple subclones in Right FT STIC. the simplest plausible solution assigned them to four copies. In CGOV63, chromosomes 6 and 15 did not have allelic imbalance in any of the tumor samples, Estimating an evolutionary timeline. Following the approach of Jones et al. , the and were assigned to two copies. No complete chromosome with absence of allelic observed data are the number of somatic mutations in the STIC (n ), the number of imbalance across all tumor samples could be found in CGOV278. Therefore, four mutations in the metastasis (n ), and the age at which the patient was diagnosed genomic regions with no allelic imbalance were selected for the normalization (t ), where somatic mutations include both sequence and structural alterations. process above. These regions were chr8:38–69 Mb, chr12:62–85 Mb, chr18:7–19 Unknown is the birthdate (t ) of the cell that was the last common ancestor of the Mb, chr20:23–35 Mb. The solution assigning these regions to two copies resulted in STIC and the metastasis. Assuming the mutation rate of somatic passenger an implausible assignment of homozygous deletion to chr5:50–136 Mb. Therefore, mutations and the length of the cell cycle is constant, the number of somatic assignment of four copies to the selected regions results in the simplest solution. In mutations in the metastasis cell that were present in the STIC follows a binomial CGOV279, two genomic regions were selected for the normalization procedure: distribution with parameters n and probability t /t .As t is unknown, we posit a k j k j chr5: 64–131 Mb, chr20:17–36 Mb. Evaluation of log ratio values suggested that the conjugate beta probability distribution on the rate t /t with shape parameters a and j k two regions are present at different copy levels, as evidenced by a difference of b estimated from previous studies as described below. The posterior distribution of ~0.60 in the log ratio values. The region on chr5, which had the lower log ratio t /t is β (a + n , b + n −n ) from which 90% highest posterior density intervals can j k j k j level, was assigned to copy number 2. In CGOV280, chr16q had no allelic be constructed with point estimates for the birthdate reported as the posterior imbalance in any samples excluding the left FT STIC. Examination of log ratio mean. For simplicity, we refer to the highest posterior density as a confidence values of chr16q in the left FT STIC supports a copy loss in that sample. The interval. To construct a prior for t /t , we draw on a previous study of four col- j k genome-wide bias term δ was determined by assignment of two copies to chr16q in orectal cancer patients where a small number of additional passenger mutations the four samples with no allelic imbalance, and one copy in the left FT STIC. were acquired by the cell that gave birth to the metastasis. On average, 95% of the mutations in the original adenocarcinoma were present in the metastases. We Subclonal hierarchy analysis. The tumor subclonality phylogenetic reconstruc- center the mean for the beta prior at 0.95 using shape parameters a = 34 and b = tion algorithm SCHISM was used to infer tumor subclonal hierarchies from the 1.6. Our prior is equivalent to one patient having 34 passenger somatic mutations set of confidently called somatic mutations in each patient. Given the estimates of in the original lesion and 1.6 additional mutations to be acquired by cells that gave genome-wide copy number profile, most copy number aberrations seem to occur birth to the metastases. For patients with three samples in a linear tree as deter- early in the evolution of disease and are common across the lesions analyzed from mined by evolutionary analyses (say, samples j, k, and l where sample j is the STIC, each patient. Thus, the majority of somatic mutations can be assumed to occur l is the metastasis, and k is an intermediate sample), we first derived the posterior following the acquisition of copy number aberrations, and can be present in cancer distribution for t comparing mutations in samples k and l. Next, we derived the cells with multiplicity of one (one mutated copy per cell). Using this assumption, posterior distribution of t integrating over all possible values of t , thereby fully j k NATURE COMMUNICATIONS 8: 1093 DOI: 10.1038/s41467-017-00962-1 www.nature.com/naturecommunications 9 | | | ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/s41467-017-00962-1 incorporating the uncertainty of the intermediate timepoint in the estimate of t . 26. Leeper, K. et al. Pathologic findings in prophylactic oophorectomy specimens in We evaluated three additional prior models, and found that that posterior inference high-risk women. Gynecol. Oncol. 87,52–56 (2002). under these alternative models given by 90% credible intervals for t −t , results in 27. Roh, M. H. et al. High-grade fimbrial-ovarian carcinomas are unified by altered k j qualitatively similar timelines among different lesions in tumor progression. p53, PTEN and PAX2 expression. Mod. Pathol. 23, 1316–1324 (2010). 28. Perets, R. et al. Transformation of the fallopian tube secretory epithelium leads to high-grade serous ovarian cancer in Brca;Tp53;Pten models. Cancer Cell 24, Data availability. Sequence data have been deposited at the European Genome- 751–765 (2013). phenome Archive, which is hosted at the European Bioinformatics Institute, under 29. Jones, S. et al. Comparative lesion sequencing provides insights into tumor study accession EGAS00001002589. evolution. Proc. Natl Acad. Sci. USA 105, 4283–4288 (2008). 30. Perets, R. & Drapkin, R. It’s totally tubular…riding the new wave of ovarian Received: 26 January 2017 Accepted: 9 August 2017 cancer research. Cancer Res. 76,10–17 (2016). 31. Conner, J. R. et al. Outcome of unexpected adnexal neoplasia discovered during risk reduction salpingo-oophorectomy in women with germ-line BRCA1 or BRCA2 mutations. Gynecol. Oncol. 132, 280–286 (2014). 32. Karnezis, A. N. & Cho, K. R. Of mice and women - non-ovarian origins of “ovarian” cancer. Gynecol. Oncol. 144,5–7 (2016). References 33. Eckert, M. A. et al. Genomics of ovarian cancer progression reveals diverse 1. Ferlay, J. et al. Cancer incidence and mortality patterns in Europe: estimates for metastatic trajectories including intraepithelial metastasis to the fallopian tube. 40 countries in 2012. Eur. J. Cancer 49, 1374–1403 (2013). Cancer Discov. 6, 1342–1351 (2016). 2. Siegel, R. L., Miller, K. D. & Jemal, A. Cancer statistics, 2015. CA Cancer J. Clin. 34. Rebbeck, T. R. et al. Prophylactic oophorectomy in carriers of BRCA1 or 65,5–29 (2015). BRCA2 mutations. N. Engl. J. Med. 346, 1616–1622 (2002). 3. Cress, R. D., Chen, Y. S., Morris, C. R., Petersen, M. & Leiserowitz, G. S. 35. Kauff, N. D. et al. Risk-reducing salpingo-oophorectomy in women with a Characteristics of long-term survivors of epithelial ovarian cancer. Obstet. BRCA1 or BRCA2 mutation. N. Engl. J. Med. 346, 1609–1615 (2002). Gynecol. 126, 491–497 (2015). 36. Falconer, H., Yin, L., Gronberg, H. & Altman, D. Ovarian cancer risk after 4. Menon, U., Griffin, M. & Gentry-Maharaj, A. Ovarian cancer screening-- salpingectomy: a nationwide population-based study. J. Natl. Cancer. Inst. 107, current status, future directions. Gynecol. Oncol. 132, 490–495 (2014). dju410 (2015). 5. Jacobs, I. J. et al. Ovarian cancer screening and mortality in the UK 37. Kwon, J. S. et al. Prophylactic salpingectomy and delayed oophorectomy as an Collaborative Trial of Ovarian Cancer Screening (UKCTOCS): a randomised alternative for BRCA mutation carriers. Obstet. Gynecol. 121,14–24 (2013). controlled trial. Lancet 387, 945–956 (2016). 38. McAlpine, J. N. et al. Opportunistic salpingectomy: uptake, risks, and 6. Kurman, R. J. & Shih Ie, M. The dualistic model of ovarian carcinogenesis: complications of a regional initiative for ovarian cancer prevention. Am. J. revisited, revised, and expanded. Am. J. Pathol. 186, 733–747 (2016). Obstet. Gynecol. 210, 471.e1–471.e11 (2014). 7. Kurman, R. J. & Shih Ie, M. The origin and pathogenesis of epithelial 39. Parker, W. H. et al. Ovarian conservation at the time of hysterectomy and long- ovarian cancer: a proposed unifying theory. Am. J. Surg. Pathol. 34, 433–443 term health outcomes in the nurses’ health study. Obstet. Gynecol. 113, (2010). 1027–1037 (2009). 8. Karst, A. M. & Drapkin, R. Ovarian cancer pathogenesis: a model in evolution. 40. Longacre, T. A., Oliva, E., Soslow, R. A. Association of directors of, A. & J. Oncol. 2010, 932371 (2010). Surgical, P. Recommendations for the reporting of fallopian tube neoplasms. 9. Levanon, K., Crum, C. & Drapkin, R. New insights into the pathogenesis of Hum. Pathol. 38, 1160–1163 (2007). serous ovarian cancer and its clinical impact. J. Clin. Oncol. 26, 5284–5293 41. Haber, D. A. & Velculescu, V. E. Blood-based analyses of cancer: circulating (2008). tumor cells and circulating tumor DNA. Cancer Discov. 4, 650–661 (2014). 10. Bowtell, D. D. et al. Rethinking ovarian cancer II: reducing mortality from high- 42. Kinde, I. et al. Evaluation of DNA from the Papanicolaou test to detect ovarian grade serous ovarian cancer. Nat. Rev. Cancer 15, 668–679 (2015). and endometrial cancers. Sci. Transl. Med. 5, 167ra4 (2013). 11. Cancer Genome Atlas Research Network. Integrated genomic analyses of 43. Eberle, F. C. et al. Immunoguided laser assisted microdissection techniques for ovarian carcinoma. Nature 474, 609–615 (2011). DNA methylation analysis of archival tissue specimens. J. Mol. Diagn. 12, 12. Patch, A. M. et al. Whole-genome characterization of chemoresistant ovarian 394–401 (2010). cancer. Nature 521, 489–494 (2015). 44. Bertotti, A. et al. The genomic landscape of response to EGFR blockade in 13. Cass, I. et al. BRCA-mutation-associated fallopian tube carcinoma: a distinct colorectal cancer. Nature 526, 263–267 (2015). clinical phenotype? Obstet. Gynecol. 106, 1327–1334 (2005). 45. Jones, S. et al. Personalized genomic analyses for cancer mutation discovery and 14. Piek, J. M. et al. BRCA1/2-related ovarian cancers are of tubal origin: a interpretation. Sci. Transl. Med. 7, 283ra53 (2015). hypothesis. Gynecol. Oncol. 90, 491 (2003). 46. Olshen, A. B., Venkatraman, E. S., Lucito, R. & Wigler, M. Circular binary 15. Piek, J. M. et al. Dysplastic changes in prophylactically removed Fallopian tubes segmentation for the analysis of array-based DNA copy number data. of women predisposed to developing ovarian cancer. J. Pathol. 195, 451–456 Biostatistics 5, 557–572 (2004). (2001). 47. Li, H. et al. The sequence alignment/map format and SAMtools. Bioinformatics 16. Medeiros, F. et al. The tubal fimbria is a preferred site for early adenocarcinoma 25, 2078–2079 (2009). in women with familial ovarian cancer syndrome. Am. J. Surg. Pathol. 30, 230–236 (2006). 17. Lee, Y. et al. A candidate precursor to serous carcinoma that originates in the Acknowledgements distal fallopian tube. J. Pathol. 211,26–35 (2007). We thank members of our laboratories for critical review of the manuscript. This work 18. Kindelberger, D. W. et al. Intraepithelial carcinoma of the fimbria and pelvic was supported by the Dr Miriam and Sheldon G. Adelson Medical Research Foundation serous carcinoma: evidence for a causal relationship. Am. J. Surg. Pathol. 31, (R.D. and V.E.V), Commonwealth Foundation (V.E.V.), US National Institutes of Health 161–169 (2007). grants CA121113 (V.E.V.), CA006973 (V.E.V.), CA083636 (R.D.), CA152990 (R.D.), 19. Kuhn, E. et al. TP53 mutations in serous tubal intraepithelial carcinoma and CA200469 (I.S.), US Department of Defense grant OCRP-OC-100517 (R.J.K and I.S.), concurrent pelvic high-grade serous carcinoma--evidence supporting the clonal the Honorable Tina Brozman Foundation for Ovarian Cancer Research (R.D.), the relationship of the two lesions. J. Pathol. 226, 421–426 (2012). SU2C-DCS International Translational Cancer Research Dream Team Grant (SU2C- 20. McDaniel, A. S. et al. Next-generation sequencing of tubal intraepithelial AACR-DT1415; V.E.V.), the Foundation for Women’s Wellness (R.D.), and the Richard carcinomas. JAMA Oncol. 1, 1128–1132 (2015). W. TeLinde Gynecologic Pathology Laboratory Endowment (I.S.). Stand Up To Cancer is 21. Bashashati, A. et al. Distinct evolutionary trajectories of primary high-grade a program of the Entertainment Industry Foundation administered by the American serous ovarian cancers revealed through spatial mutational profiling. J. Pathol. Association for Cancer Research. S.I.L-G. is a recipient of grants from Arthur Sachs/ 231,21–34 (2013). Fulbright/Harvard, La Fondation Philippe and La Fondation de France—“Recherche 22. Nik, N. N., Vang, R., Shih Ie, M. & Kurman, R. J. Origin and pathogenesis of clinique en cancérologie—Aide à la mobilité des chercheurs”. pelvic (ovarian, tubal, and primary peritoneal) serous carcinoma. Annu. Rev. Pathol. 9,27–45 (2014). 23. McPherson, A. et al. Divergent modes of clonal spread and intraperitoneal Author contributions mixing in high-grade serous ovarian cancer. Nat. Genet. 48, 758–767 (2016). S.I.L.-G, E.P., R.D., and V.E.V. were involved in the conception and design of this project. 24. Yachida, S. et al. Distant metastasis occurs late during the genetic evolution of S.I.L-G, E.P., V.A., N.N., R.K., R.D., and V.E.V developed the methodology. S.I.L.-G, E.P., pancreatic cancer. Nature 467, 1114–1117 (2010). V.A., M.N., M.N., M.S.H. D.I.L, L.S., C.L.M., J.-C.T, M.B., A.A., L.D.W, R.K., T.-L.W, 25. Niknafs, N., Beleva-Guthrie, V., Naiman, D. Q. & Karchin, R. SubClonal I.-M.S., R.D., and V.E.V. were involved in the acquisition of data, acquiring and hierarchy inference from somatic mutations: automatic reconstruction of managing patients, providing facilities and ect. S.I.L.-G, E.P., D.H., R.B., N.N., S.J., J.P., C. cancer evolutionary trees from multi-region next generation sequencing. PLoS A.H., R.B.S., R.K., R.D., and V.E.V. were involved in the analysis and interpretation of Comput. Biol. 11, e1004416 (2015). data through statistical analysis, biostatistics and computational analysis. S.I.L.-G, E.P., 10 NATURE COMMUNICATIONS 8: 1093 DOI: 10.1038/s41467-017-00962-1 www.nature.com/naturecommunications | | | NATURE COMMUNICATIONS | DOI: 10.1038/s41467-017-00962-1 ARTICLE D.H., R.D., and V.E.V. were involved in the writing, review and revision of the manu- Publisher's note: Springer Nature remains neutral with regard to jurisdictional claims in script. S.I.L.-G, E.P., D.H., R.B., N.N., S.J., J.P., R.B.S., R.K., R.D., and V.E.V. were published maps and institutional affiliations. involved with administrative, technical or material support by reporting or organizing data or construction databases. R.K., R.D., and V.E.V supervised the study. Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, Additional information adaptation, distribution and reproduction in any medium or format, as long as you give Supplementary Information accompanies this paper at doi:10.1038/s41467-017-00962-1. appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The images or other third party Competing interests: V.E.V. is a founder of Personal Genome Diagnostics and is a material in this article are included in the article’s Creative Commons license, unless member of its Scientific Advisory Board and Board of Directors. V.E.V. owns Personal indicated otherwise in a credit line to the material. If material is not included in the Genome Diagnostics stock, which is subject to certain restrictions under university policy. article’s Creative Commons license and your intended use is not permitted by statutory The terms of this arrangement is managed by the Johns Hopkins University in accordance regulation or exceeds the permitted use, you will need to obtain permission directly from with its conflict of interest policies. The remaining authors declare no competing financial the copyright holder. To view a copy of this license, visit http://creativecommons.org/ interests. licenses/by/4.0/. Reprints and permission information is available online at http://npg.nature.com/ reprintsandpermissions/ © The Author(s) 2017 NATURE COMMUNICATIONS 8: 1093 DOI: 10.1038/s41467-017-00962-1 www.nature.com/naturecommunications 11 | | |

Journal

Nature CommunicationsSpringer Journals

Published: Oct 23, 2017

There are no references for this article.