Clonality, Heterogeneity, and Evolution of Synchronous Bilateral Ovarian Cancer Free

Multiple synchronous gynecological lesions represent a relatively frequent occurrence in female patients suffering from ovary-associated malignant diseases (1–4). Clinicians are often faced with the diagnostic dilemma of whether the simultaneously presented tumors arise independently or result from metastatic dissemination (5–7). Such a distinction is especially important yet challenging in the case of epithelial ovarian cancer (EOC), which contributes to the highest mortality among gynecological malignancies (8–10). At the time of diagnosis, the majority of women with EOC display multisite tumor spread and are empirically treated as metastatic disease, although individual foci could possibly originate from independent ancestral clones (11–15). Of particular perplexity, advanced EOC often affects both sides of ovaries, and synchronous bilateral ovarian cancers (SBOC) are even observed in 20% to 25% of early-stage patients whose neoplasm is still limited to the ovaries without pelvic extension or widespread metastasis (16). However, currently there is no consensus on the clonal architecture of SBOCs, the delineation of which may hold profound implications for patient prognosis and management. Early evidence based on X chromosome inactivation, microsatellite alterations, cytogenetic analyses, loss of heterozygosity, and somatic mitochondrial DNA mutations has led to unclear conclusions (5, 16–20). In addition, these previous studies only provided incomplete molecular portraits of SBOCs and did not address the potential impact of genetic heterogeneity and evolutionary trajectories on shaping ovarian tumorigenesis.

Recent advancement of next-generation sequencing allows for accurate diagnosis and comprehensive characterization of synchronous gynecological tumors. For example, massively parallel sequencing-based analyses reveal that synchronous endometrial and ovarian carcinomas are often clonally related and constitute metastatic deposits disseminated from one site to the other, despite undergoing notable genetic divergence upon separation (21–23). Furthermore, genomic studies centered on understanding the phylogenetic relationships of primary gynecological cancer and concurrent abdominopelvic metastases depict their clonal lineages and progression patterns, yielding new insights into the mechanisms of transcoelomic spread (24, 25). In principle, in-depth bioinformatics interrogation of high-resolution sequencing data should not only enable definitive discrimination between two independent primary tumors and metastatic disease, but also help infer the molecular origins and evolutionary course of SBOCs.

In the current study, we subjected a series of 12 SBOC cases to whole-exome or whole-genome sequencing (WGS), with the aims to (i) clarify the clonal relatedness of synchronous ovarian tumors present in a given patient and (ii) determine the temporal dynamics and spatial heterogeneity underlying SBOC development. Our data unambiguously indicated that SBOCs in all individuals were derived from a common ancestry, irrespective of histological types or clinical stages. We further demonstrated that the dissemination of unifocal bilateral tumors could occur early during the pathogenesis of ovarian cancer, resulting in disparate repertoires of somatic mutations and varied degree of intratumor heterogeneity. These findings lay the basis for prospective prognostic investigations of unilateral versus bilateral EOC with matched clinicopathologic criteria and may potentially change current guidelines for the prevention, early detection, precise diagnostication, and rational intervention of ovarian cancer.

The study was conducted in accordance with ethical guidelines of U.S. Common Rule and approved by the Ren Ji Hospital Ethics Committee. Written informed consent was obtained from all patients. Between 2008 and 2011, women diagnosed with EOCs were selected from a review of patients who had been treated and followed up in Ren Ji Hospital, Shanghai Jiao Tong University. Our cohort for retrospective analysis comprised a total of 121 ovarian cancer patients. Clinical information, including age at diagnosis, FIGO stage, pathological grade, CA-125 concentration, survival status and clinical intervention, was obtained from medical records. The χ² test was used to evaluate the association between ovarian cancer subtype and the clinicopathologic parameters of the patients. P < 0.05 was considered statistically significant. Biospecimens for sequencing were collected from patients diagnosed with pathologically confirmed ovarian cancer during initial debulking surgery in Ren Ji Hospital, who had received no prior treatment for their disease. The sequenced cohort was selected based on the availability of paired ovarian tumor tissues and archived blood cell controls, as well as the proportion of tumor cell nuclei (>50%) and necrosis (<30%; reviewed by a certified pathologist). Genomic DNA quantity and quality were determined to further exclude unqualified specimens. Eventually, 33 formalin-fixed and paraffin-embedded (FFPE) samples from 11 patients with SBOC (RJOC1-11) were subjected to whole-exome sequencing (WES), concerning limited DNA materials, and 5 fresh-frozen samples from RJOC12 with metastatic disease were subjected to WGS.

Tumor DNA was extracted from tissue shavings of fresh-frozen specimens (RJOC12) or 10 sections with a thickness of 10 μm from FFPE tissues (RJOC1-11) using the QIAamp DNA FFPE Tissue Kit. Paired blood cell DNA was extracted following instructions using the QIAamp DNA Blood Mini Kit. DNA was quantified by Qubit (Life Technologies), and DNA integrity was examined by agarose gel electrophoresis.

A paired-end DNA library was constructed according to the manufacturer's instructions (Agilent). Genomic DNA from patients RJOC1 to 11 was sheared into 180 to 280 bp fragments by Covaris S220 sonicator and purified using AMPure SPRI beads from Agencourt. The DNA fragments were enriched by 6 cycles of PCR using SureSelect Primer and SureSelect ILM Indexing Pre Capture PCR Reverse Primer. The size distributions of the libraries were examined on an Agilent Bioanalyzer DNA 1000 chip. DNA (500 ng) was subjected to whole-exome capture, using the Agilent's SureSelect Human All Exon V5 Kit. The captured DNA–RNA hybrids were recovered using Dynabeads MyOne Streptavidin T1 from Dynal. DNA was eluted from the beads and desalted using Qiagen MinElute PCR purification columns. The purified capture products were then amplified using the SureSelect ILM Indexing Post Capture Forward PCR Primer and PCR Primer Index 1 through Index 16 (Agilent). DNA sequences (50 Mb) of 334,378 exons from 20,965 genes were captured. DNA libraries were sequenced on Illumina Hiseq 4000 sequencing platform (Illumina) according to the manufacturer's instructions for paired-end 150 bp reads (Novogene). The targeted sequencing depth was 200×.

One microgram genomic DNA from each sample of patient RJOC12 was used as input for the DNA library preparations. Sequencing library was prepared using Truseq Nano DNA HT Sample Prep Kit (Illumina) following the manufacturer's recommendations. Briefly, genomic DNA sample was fragmented, end-polished, A-tailed, and ligated with the full-length adaptor for Illumina sequencing, followed by further PCR amplification. The clustering of samples was performed on a cBot Cluster Generation System using Hiseq PE Cluster Kit (Illumina) according to the manufacturer's instructions. After cluster generation, the DNA libraries were sequenced on Illumina Hiseq 4000 sequencing platform (Illumina) and 150 bp paired-end reads were generated (Novogene). The targeted sequencing depth was 60×.

Clean reads in FastQ format generated by the Illumina platform were aligned to UCSC hg19 human genome by Burrows-Wheeler Aligner to obtain mapping results in BAM format (26). SAMtools, Picard (http://broadinstitute.github.io/picard/), and GATK were used to filter BAM files, local realignment, and base quality recalibration to generate final BAM files for computation of the sequence coverage and depth (27, 28). The somatic single-nucleotide variations (SNV) were detected by MuTect with an additional filter described below (29). The somatic indels detected by GATK Somatic Indel Detector. ANNOVAR was performed to do annotation for VCF (Variant Call Format; ref. 30). Variants obtained from previous steps were compared against SNPs present in the dbSNP and 1000 Genomes databases (1000 Genomes Project Consortium) to discard known SNPs. The retained nonsynonymous SNVs were submitted to PolyPhen and Sorting Intolerant From Tolerant (SIFT) for functional prediction (31, 32). Control-FREEC was utilized to detect somatic copy-number variations (33). For samples from patient RJOC12, which were subjected to WGS, Breakdancer was implemented to identify potential structural variants (34).

Additional filters were applied to exclude artifactual mutations introduced by FFPE specimens. In brief, duplicates and soft clipped reads removed data were analyzed in MuTect with these parameters (align quality: 30; strand bias: 0.05; keep the mutation site with highest align quality if more than one mutation sites were examined within 11 bp; keep the mutation sites supported by at least three different reads). Furthermore, we filtered out single strand bias based on a read pair orientation of larger than 20:1.

All identified nonsynonymous mutations were compared with potential driver genes in ovarian cancer or present in the COSMIC cancer gene census. Putative driver mutations were determined if they satisfy one of the following criteria: (i) either the exact mutation, the same mutation site or at least three mutations located within 15 bp of the variant were found in COSMIC and (ii) if the candidate gene was marked as recessive in COSMIC and the variant was predicted to be deleterious and had a SIFT score <0.05 or a PolyPhen score >0.995. The putative driver mutations (variant allele frequency, VAF, >0.1) were subjected to Sanger sequencing for validation.

All somatic SNVs were included to calculate relative weights of mutational signatures in a given sample. The R package “deconstructSigs,” based on the Wellcome Trust Sanger Institute Mutational Signature Framework, was used to statistically quantify the contribution of each signature for each tumor (35).

SciClone R package (https://github.com/genome/sciclone) and PyClone (http://compbio.bccrc.ca/software/pyclone/) were used to detect subclonality of a given sample (36, 37). Clonally relatedness was determined by clonality indices calculation, which was performed as described previously (21). Phylogenetic tree was constructed through PHYLIP version 3.695. For each variant, the cancer cell fraction (CCF) was defined as |VAF = {\frac{{\alpha \times CCF}}{{CT \times \alpha + CN( {1 - \alpha } )}}}$|⁠, where CT is the copy number of the tumor, CN is the copy number of the matched normal sample, and α is the tumor purity. Tumor purity and copy number were determined by ABSOLUTE (38). VAF is defined with respect to the number of reads supporting the variant allele (⁠|{x^{var}}$|⁠) and the number of reads supporting the reference allele (⁠|{x^{ref}}$|⁠): |VAF = {\frac{{{x^{var}}}}{{{x^{var}} + {x^{ref}}}}}$|⁠. HLA typing was performed using Polysolver (39). Nonsilent mutations were used to generate a list of peptides ranging 9 to 11 amino acids in length with the mutated residues represented in each position. Prediction for binding affinity of every mutant peptide and its corresponding wild-type peptide to the patient's germline HLA alleles was performed using the NetMHCpan (v3.0; ref. 40). Candidate neoantigens were identified as those with a predicted mutant peptide binding affinity of <500 nmol/L.

Statistical analysis was performed with GraphPad Prism software. In all experiments, comparisons between two groups were based on two-sided Student t test, and one-way analysis of variance (ANOVA) was used to test for differences among more groups. P values of <0.05 were considered statistically significant.

We set out to assess the prevalence and characteristics of SBOC by performing a retrospective analysis of 121 consecutive ovarian cancer patients who underwent debulking surgery in our hospital between 2008 and 2011 (Supplementary Table S1). A total of 46 women (39%) with EOC had pathologically confirmed concurrent tumors in both ovaries. At diagnosis, the median age of patients presenting SBOCs was similar to that of individuals with unilateral ovarian cancer (54 vs. 57). Most SBOCs (63%) were diagnosed as serous carcinomas, in comparison with 32% of unilateral EOC cases (P = 0.001, χ² test). Bilaterality was associated with EOC of higher histological grade (P < 0.001, χ² test) and more advanced FIGO stage (P < 0.001, χ² test). The estimated 5-year survival rate of SBOCs was significantly lower than that of unilateral EOCs (30.7% vs. 65.4%, P = 0.007, χ² test). Although these data implied that SBOCs might represent metastatic disease with a dismal prognosis, the exact molecular origin and relationship of the synchronous bilateral tumors remained largely elusive.

To resolve the clonal composition of SBOCs, we selected archived biospecimens from the 121-sample cohort for molecular analysis, on the basis of availability of paired ovarian tumor tissues and companion blood cell controls, as well as the proportion of tumor cell nuclei (>50%) and necrosis (<30%). Twenty-two tumors from paired ovaries of 11 EOC patients (Fig. 1A; Table 1), spanning various FIGO stages (from IB to IV) and different histological subtypes (Fig. 1B), passed sequencing quality control and were included in our study. Both representative hematoxylin and eosin (H&E) staining and computed tomography (CT) images verified the presence of SBOC in each patient (Supplementary Fig. S1). Genomic DNA extracted from tumors and matched peripheral blood cells were subjected to WES with a median depth of 212× (ranging from 99× to 393×). At least 97.9% of targeted bases were covered to a depth of more than 10× (Supplementary Table S2). We applied stringent filters to exclude enriched C>T sequence artifacts characteristic of DNA from FFPE tissues (41–43), and identified a median of 132 somatic exonic alterations per tumor (27–2,324), including 107 nonsynonymous mutations (17–1,537; Fig. 1C; Supplementary Table S3). A subset of somatic mutations with VAF of >0.1 were analyzed with Sanger sequencing, achieving a 90% validation rate (Supplementary Table S4). Consistent with previous reports (44), TP53 mutations were prevalent and detected in all EOC patients (Fig. 1C). Analysis of genetic predisposition on the basis of 24 hereditary gynecologic cancer genes uncovered 13 single-nucleotide polymorphisms (SNP) possibly associated with ovarian cancer susceptibility in 9 patients (Supplementary Fig. S2; Supplementary Table S5). Notably, patient RJOC8 harbored germline mutations in mismatch repair genes related to lynch syndrome, including MSH2/6, and displayed a hypermutator phenotype (Fig. 1C; Supplementary Fig. S2). Overall, the number of genomic abnormalities, mutation spectra, and mutational signatures of SNVs were fairly consistent between two ovarian tumors from each of the individuals (Fig. 1C–E). Additionally, we identified putative driver genetic aberrations involving genes regulating multiple pivotal cancer-promoting pathways (Supplementary Table S6) and synchronous ovarian lesions from a given patient often shared mutations enriched in the same functional categories (Fig. 1D), further supporting monoclonality of bilateral tumors within our SBOC cohort.

Rooted in the clonal evolution theory of tumorigenesis, SBOCs are assumed to inherit an identical set of somatic variants if they arise from a single ancestral progenitor cell; on the other hand, distinct genomic profiles of bilateral ovarian tumors would most likely indicate multifocal origins. Following this conceptual framework, we probed the genetic aberrations of SBOCs identified by WES in order to yield important information about the molecular relationship of paired bilateral tumors in each patient. Our results revealed that the two synchronous tumors of a given case contained considerably overlapping repertoires of somatic exonic mutations including SNVs and indels (Fig. 2A), though to diverse extents in different patients (11%–71%). Notably, clonal architecture of SBOCs indicated that in many cases, the number of identical shared variants was larger than or comparable with that of private somatic events. Nonetheless, several patients such as RJOC3, RJOC8, and RJOC10 displayed relatively less similarity between bilateral ovarian carcinomas, raising the possibility that shared mutations in these samples could had occurred by chance (Fig. 2B). To formally evaluate the phylogeny of SBOCs, we used a conservative analytic approach to estimate their clonality indices based on all SNVs and indels (21) and found that paired bilateral tumors from all 11 patients were clonally related with no exception (Fig. 2C). Therefore, we concluded that SBOCs formed through metastatic spread rather than independent multifocal oncogenesis.

To validate our findings in FFPE tissues (RJOC1-11) that are prone to sequencing errors, we recruited a patient (RJOC12) with stage III serous ovarian cancer and sampled fresh-frozen bilateral ovarian masses, in addition to two other anatomic sites (colon and stomach) with distant metastases. WGS achieved an average of 50× coverage with 99% of targets sequenced at a depth of ≥10×, thus providing more robust genetic data to better decipher the metastatic routes and evolutionary dynamics of ovarian malignancies. As with the WES-profiled subjects (RJOC1-11), RJOC12 exhibited generally similar mutation spectra and mutational signatures across all four specimens, although mutation burdens of the right ovarian tumor appeared to be relatively lower (Fig. 3A). We sought to infer the spatiotemporal origin of neoplastic development by applying two complementary approaches on the basis of somatic mutations. First, SciClone (36) was performed to identify the number and genetic composition of subclones in each sample (Supplementary Fig. S3), followed by two-dimensional cluster correlation between tumor pairs of the four samples. These analyses captured tumor-specific mutations in cluster 2, which were observed in all samples but right ovarian lesion (Fig. 3B). Alternatively, a phylogenetic clustering method (45, 46) was used and revealed that the genome of right ovarian tumor was hierarchically most similar to germline DNA of RJOC12, whereas sequenced samples from the other three anatomic sites located in parallel on the same branch (Fig. 3C). Taken together, our results suggested that the right ovarian tumor represented the primary neoplasm, from which a subpopulation of cancer cells disseminated and implanted in the left ovary, colon, and stomach. Indeed, comparison of copy-number alterations or chromosomal structural variations harbored by four samples confirmed this tumor topology (Fig. 3D; Supplementary Fig. S4). In addition, the strikingly recurrent patterns of somatic events in distal sites and appreciably early dichotomy between primary and metastatic tumors supported a model of monoclonal seeding and expansion during ovarian cancer progression of RJOC12 (Fig. 3D). Interestingly, in line with the spatiotemporal trajectories of malignant dissemination, CCF (the fraction of tumor cells harboring the SNVs) of somatic mutations in the right ovarian tumor was significantly lower than that in three presumed metastatic lesions (Fig. 3E), which elicited our attempt to define the chronology of SBOC according to the CCF information. By comparing the relative CCF abundance of shared somatic alterations between bilateral ovarian tumors, we were able to deduce the likely primary side of tumorigenesis with the smaller median CCF value for each patient (Supplementary Fig. S5). Future prospective clinical studies integrating both molecular and histopathological features to determine the metastatic directionality of SBOC are warranted.

To further investigate the evolutionary process of SBOCs, we performed phylogenetic reconstruction of the WES data for each patient (Fig. 4A). The first notable observation was that the phylogenetic structure varied considerably between different SBOC cases, i.e., disparate timing of branched evolution in each individual resulting from tumor metastasis. By taking into account the ovarian tumor volumes at surgery, we performed a quantitative analysis on the time point of metastatic spread, indicating that cancer dissemination to the contralateral ovary mostly occurred early when the primary carcinoma was still relatively small (<100 million cells; Fig. 4B). Accordingly, the spectra of point mutations in each patient displayed pronounced temporal difference between early (truncal) and late (branched) SNVs (Fig. 4A), with a statistically significant increase of C>T transitions and decrease of C>A or C>G transversions in late compared with early mutations (Fig. 4C). We used deconstructSigs framework (35) to extract known mutational signatures (47, 48) that might contribute to the specific mutation profiles of early versus late SNVs (Fig. 4D; Supplementary Fig. S6). Early (truncal) SNVs exhibited an increased enrichment of signature 3 (associated with failure of DNA double-strand break repair by homologous recombination), signature 4 (attributable to tobacco mutagens), signature 13 (implicated with APOBEC activity), and signature 24 (related to aflatoxin exposures). On the other hand, later (branched) SNVs were more consistent with mutational processes characterized by signature 1 (correlated with age of cancer diagnosis) and signature 6 (associated with defective DNA-mismatch repair). Additional investigations are required to fully uncover the mechanistic underpinnings and biological significance of these observations, although the results pinpointed differential contributions of deficiency in different DNA repair mechanisms during ovarian cancer progression. Moreover, early mutations included a large proportion of driver events, such as alterations in TP53, NOTCH1, and ARID1B; however, multiple driver mutations affecting cancer genes including EZH2 and AXIN1 were acquired late (Fig. 4E). In addition, spatial heterogeneity, as indicated by divergent mutation spectra and mutational signatures of late SNVs, was evident between geographically separated bilateral tumors (Fig. 4A; Supplementary Fig. S7), suggesting that primary and metastatic cancers continued to genetically evolve upon disassociation. Together, our findings implied that dynamic mutational processes constantly shaped tumor genomes over time, conceivably leading to the substantial spatiotemporal divergent evolution of SBOC.

Previous studies have revealed profound intratumor genetic heterogeneity of high-grade serous ovarian cancer, which may have predictive value for survival after chemotherapy treatment (49). To estimate the extent of intratumor heterogeneity in our SBOC cohort, we performed single-sample subclonality analysis using SciClone (Supplementary Fig. S8) and PyClone (Supplementary Fig. S9). Both methods consistently showed that the majority of sequenced samples were oligoclonal (Supplementary Table S7) and the clonal structure was more complicated in metastatic deposits relative to primary tumors (Fig. 5A). Consequently, most key putative driver mutations were subclonal events in bilateral ovarian carcinomas, as evidenced by the CCF of below 1 (Fig. 5B). In addition, we evaluated the druggability of somatic aberrations by stratifying them based on the Cancer Drivers Actionability Database (50). Although all 11 SBOC cases could potentially benefit from current or prospective anticancer targeted agents, the in silico prescription approach highlighted a patient-specific therapeutic landscape. More importantly, distinct tailored drugs should be assigned to primary or metastatic cancers (Fig. 5C). Of note, somatic alterations of DNA repair genes such as BRCA1/2 often occurred only in one side of SBOC, in contrast to the uniformly mutated germline cancer–predisposition genes (Supplementary Fig. S10), implying that currently approved PARP inhibitors might display different efficacy against paired tumors in some patients. Finally, bilateral tumors were predicted to harbor distinct repertoires of neoantigens (Fig. 5D), suggestive of heterogeneous tumor-infiltrating lymphocytes and potentially differential clinical responses to immune checkpoint inhibitors. We concluded that the prominent intratumor and intertumor heterogeneity contributed to the genetic complexity and phenotypic diversity of clonal related bilateral ovarian tumors, and might have significant impact on the clinical course and therapeutic interference of individuals who suffer from SBOC.

In this study, by performing next-generation genomic sequencing and detailed bioinformatics analysis, we have presented an explicit view of the clonality, heterogeneity, and evolution of SBOC. Our data support three major conclusions: first, bilateral tumors of SBOC most likely originate from a common ancestry, regardless of histotypes or pathological stages; second, ovarian cancer cells may disseminate extremely early to give rise to the SBOC disease; third, each ovarian lesion comprises genetically heterogeneous composition and paired ovarian cancers display notable intertumor heterogeneity. These findings provide new insights into the phylogenetic principles of ovarian tumorigenesis and malignant progression and may have considerable implications for the prognosis and management of patients bearing SBOC.

Previous studies, using a variety of experimental strategies, have attempted to evaluate the clonal relatedness of synchronous bilateral ovarian tumors and have reached inconsistent conclusions. For example, genetic markers based on the pattern of X chromosome inactivation and microsatellite instability (MSI) revealed that SBOCs shared a unifocal origin (16). Similarly, cytogenetic analyses indicated that bilateral ovarian cancer developed through metastatic spreading (19). In striking contrast, distinct somatic mitochondrial DNA variants were detected between paired tumors in 24% (4 out of 17) of SBOC cases, indicative of different clonal populations of cancer cells (20). Additionally, evidence for monoclonality was found in 14 of 16 SBOCs according to loss of heterozygosity (LOH) analysis, whereas bilateral tumors in the remaining two cases showed discordant LOH patterns, which was thought to be most likely due to independent origins (18). Moreover, genetic characterization by combining LOH and MSI information into a statistical algorithm classified 33% (5 out of 15) of SBOCs as having independently derived primary tumors (5). We speculate that one plausible reason for these contradictory results lies in the insufficient resolution of limited numbers of molecular markers. Our study, on the other hand, exercised comprehensive genomic profiling of SBOCs for the first time (to our knowledge) and conclusively demonstrated the uniform monoclonality of bilateral ovarian lesions in most, if not all, cases. However, it is noteworthy that all the paired bilateral tumors have the same histologic appearance in our SBOC cohort. Therefore, it remains to be determined whether SBOCs with dissimilar histotypes are clonally related as well or represent independent tumors.

It has been suggested that malignant dissemination can occur early in the course of SBOC, based on the fact that some cases are diagnosed as stage IB disease with tumors affecting both ovaries in the presence of seemingly intact capsules (16). Indeed, our SBOC cohort also included a stage IB patient and her bilateral ovarian carcinomas evolved as unifocal neoplasms through a yet-to-be identified mechanism of metastatic spread. We further discovered that the extraorgan metastasis likewise represented a generally early event even in the advanced-stage SBOC, by quantitatively inferring the time point of cancer cell dissemination to the contralateral ovary. Taken together, these findings imply that early distal dissemination may be an intrinsic feature of ovarian cancer, thereby causing delayed diagnosis and dismal prognosis in patients with EOC. This notion poses formidable challenges for active surveillance, early detection, and optimal intervention of ovarian pathogenesis. It should be noted, though, that a “pseudo-metastatic” scenario has been proposed in the context of gynecological cancers with a predilection for abdominopelvic implantation (22), which advocates distinguishing between clinically indolent dissemination and widespread metastasis. Therefore, it is important to recognize the complexity underneath the simple bilaterality regarding SBOC in order not to undertreat or overtreat patients.

Considering the early disassociation of bilateral ovarian tumors, perhaps it was not surprising to learn that pronounced spatiotemporal divergent evolution took place to shape SBOC tumorigenesis. As a consequence, notable differences in spatial and temporal acquisition of putative driver mutations, including those predicted targetable, were observed between paired tumors. Importantly, we showed that the repertoire of neoantigens created by genetic mutations were heterogeneous in SBOCs, which might influence inflammatory microenvironment and tumor immunoreactivity. In addition, each tumor was revealed to contain multiple subclones, and metastatic lesions seemed to bear a more complex subclonal structure relative to primary tumors. Our findings of the prominent intratumor and intertumor heterogeneity were in accordance with previous reports unveiling the extensive genomic diversity in high-grade serous ovarian cancer (49). Furthermore, dynamic mutational processes were pinpointed during SBOC progression, and our data highlighted DNA repair deficiency as a prevalent mechanism involving the whole evolutionary history of ovarian cancer. Future investigations are warranted to uncover the mechanistic underpinnings and therapeutic relevance of these observations.

In summary, we have presented a detailed delineation of the molecular origins and evolutionary trajectories of SBOCs, which holds the promise to improve the diagnosis and treatment of ovarian cancer. Our results reveal the monoclonality, generally early metastatic dissemination and profound genetic heterogeneity associated with SBOCs, and provide an impetus for further dissecting the underlying biological mechanisms. We envision that the sequencing-based approach described in this study, which is applicable to FFPE tissues, may be incorporated into routine histopathology to reliably identify the clonal relationships of synchronous gynecological lesions and other multiple malignancies.

No potential conflicts of interest were disclosed.

Conception and design: W. Di, G. Zhuang

Development of methodology: W. Di

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): X. Yin, Y. Jing, C. Xu, M. Zhang, W. Di

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): X. Yin, Y. Jing, M.-C. Cai, P. Ma, M. Zhang, W. Di

Writing, review, and/or revision of the manuscript: X. Yin, Y. Jing, M.-C. Cai, W. Di, G. Zhuang

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): X. Yin, Y. Zhang, W. Di

Study supervision: Y. Jing, W. Di, G. Zhuang

The authors thank all lab members for helpful discussions and advice regarding this work.

This work was supported by the National Natural Science Foundation of China (81472537 and 81672714 to G. Zhuang; 81502597 to Y. Jing; 81472426 to W. Di), the Grants from the State Key Laboratory of Oncogenes and Related Genes (No. 91-15-12 to G. Zhuang; SB17-06 to M.-C. Cai), the grants from Shanghai Jiao Tong University School of Medicine (DLY201505 to W. Di; YG2016MS51 to X. Yin), Shanghai Municipal Education Commission-Gaofeng Clinical Medicine Grant Support (20161313 to G. Zhuang), the Shanghai Institutions of Higher Learning (Eastern Scholar to G. Zhuang), Shanghai Rising-Star Program (16QA1403600 to G. Zhuang), Shanghai Municipal Commission of Health and Family Planning (2013ZYJB0202 and 15GWZK0701 to W. Di), the grant from Shanghai Key Laboratory of Gynecologic Oncology (FKZL-2017-01 to Y. Jing), and the grant from Science and Technology Commission of Shanghai Municipality (16140904401 to X Yin).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.