Clinicopathologic and Molecular Features of Paired Cases of Metachronous Ovarian Serous Borderline Tumor and Subsequent Serous Carcinoma.

Of 1,025 cases with a confirmed diagnosis of SBT by centralized histopathology review, 42 patients (4%, consisting of 34 APST and 8 niLGSC) subsequently developed invasive serous carcinoma (39 LGSC and 3 HGSC). Clinicopathologic and demographic characteristics, in aggregate and subdivided by type of SBT, are presented in Table 1 . The median interval from the diagnosis of SBT to subsequent carcinoma was 9.0 years (range: 0.6 – 24.9), with a significantly shorter time for cases of LGSC compared to those associated with HGSC (8.1 versus 21.8 years, respectively, p = 0.004, Mann-Whitney U test). In one patient, there was a recurrence as non-invasive implants 2 years after the initial surgical procedure, followed by subsequent development of invasive LGSC 15 years after the first recurrence. The histologic features were characteristic of APST (n = 34, 81%) and niLGSC (n = 8, 19%) ( Fig. 1A , B ) ( 7 ). APSTs were characterized by progressive hierarchical branching papillae lined by a heterogeneous population of cells, consisting of cuboidal, columnar, and eosinophilic cells. Cilia were typically abundant. Psammoma bodies were frequently found. Four APSTs exhibited foci of micropapillary growth measuring ≥1 mm but <5 mm in greatest dimension, which, due to the limited extent, did not qualify for classification as niLGSC ( 1 ). A prominent (cyst)adenofibromatous component was identified in 14 APSTs, and in 5 cases, comprised >50% of the tumor. In all niLGSCs, the extent of micropapillary growth measured ≥5 mm, and often comprised the majority of the tumor. None of niLGSCs had a significant (cyst)adenofibromatous component. Psammoma bodies were readily identifiable. The FIGO Stage distribution is as follows: 17 (40%) Stage I, 12 (29%) Stage II, 12 (29%) Stage III, and 1 case (2%) undetermined stage ( Table 1 ). Bilateral ovarian involvement was observed in 31 (74%) cases. Microinvasion was identified in 3 tumors (conventional (eosinophilic cell) type, n = 1; microinvasive carcinoma type, n = 2). Implants were identified in 24 cases (57%), of which 18 (75%) were non-invasive and 6 (25%) were of invasive type. Invasive implants were more common in niLGSC compared to APST (p = 0.035, χ 2 test). Notably, across the entire cohort, 14 (33%) cases were characterized by low-risk pathologic features, specifically stage I APST, of which 6 were unilateral. The level of tissue sampling was 1.3 (mean) blocks per centimeter of tumor in greatest dimension (range: 0.6 to 3.5). [See section Mutational analyses and associations with clinicopathologic features for correlation between genotype and specific features]. Invasive serous carcinomas typically involved the omentum or peritoneum ( Table 2 ). Involvement of pelvic organs, lymph nodes, intestine, liver, and other extra-abdominal sites by serous carcinoma was also observed. Thirty-nine of 42 women (93%) developed LGSC. The most common pattern (28/39; 72%) was exclusively the infiltrating type, which was characterized by destructive stromal invasion by tumor cells forming compact nests, solid sheets, or glands, often within clear lacunar spaces ( Fig. 2A , B ). The neoplastic epithelium was haphazardly arranged within desmoplastic or fibrotic stroma, and the nests/sheets/glands exhibited variable degrees of crowding throughout the tumor, but foci with marked crowding were typically present. The second most common pattern (5/39; 13%) consisted of exophytic micropapillary growth resembling ovarian niLGSC and may be associated with desmoplasia at the periphery ( Fig. 2C , D ). This pattern occasionally may be within a cyst, but the majority of the lesion, nonetheless, does not exhibit the classic pattern of stromal invasion as seen in the more common infiltrating pattern. The remaining LGSCs were comprised of admixtures of both patterns in variable proportions (2/39; 5%) or had non-specific architectural patterns that could not be classified into these categories (4/39; 10%). The cells had varying amounts of eosinophilic cytoplasm, but the nuclear-to-cytoplasmic ratios tended to be high. The nuclei showed low-grade atypia, typical of LGSC with round shapes, uniform small to medium size throughout the tumor, small nucleoli, and infrequent mitotic figures. The remaining 3 women (7%) developed HGSC ( Table 3 ), which were histologically composed of solid sheets and non-specific papillary architecture. The cells had variable amounts of eosinophilic cytoplasm, and the nuclei were hyperchromatic and pleomorphic, containing prominent nucleoli. Mitotic figures were readily identifiable. Two of the HGSCs were associated with prior APST (cases #1 and #3); one with a prior niLGSC (case #2). In case #3, HGSC was a discrete focal lesion that arose within a peritoneal serous cystadenofibroma. Immunohistochemical staining results were generally consistent with HGSC ( Table 3 ). Of note, p53 immunohistochemical expression in the HGSCs was strong and diffuse in cases #2 and #3, and patchy (“wild-type pattern”) in case #1. In case #3, there was a sharp transition between diffuse p53 staining in the tumor, and wild-type p53-staining pattern in the adjacent benign cystadenofibromatous epithelium. Endosalpingiosis was identified at a significantly higher frequency in women with metachronous SBT and subsequent carcinoma [33%; 14/42, either at the time of SBT diagnosis (n = 9) or in the subsequent carcinoma specimen (n = 5], compared to women with SBTs not associated with subsequent malignancy from the entire Denmark SBT cohort (9%; 89/983, p < 0.0001, χ 2 test). Tumor tissue for molecular analysis was available from at least one specimen for 41 cases ( see Materials and Methods section). Of 39 SBTs with molecular analysis, 5 (13%) harbored a BRAF V600E mutation, 22 (56%) had a KRAS mutation, and the remaining 12 (31%) were wild-type at the mutation hotspots for both genes. KRAS and BRAF mutations were mutually exclusive in SBTs. An amino acid substitution from glycine to aspartate at codon 12 (G12D) was the most common KRAS mutation, occurring in 11 (50%) cases, followed by the glycine to valine substitution (G12V), which was present in 8 (36%) cases. Of the remaining three KRAS -mutated SBTs, one was a glycine to alanine substitution (G12A) and two could not be classified further. Of the 38 serous carcinomas with mutational analysis performed, 5 (13%) had a BRAF V600E mutation, 22 (58%) had a KRAS mutation [G12D (n = 12), G12V (n = 7), G12A (n = 1), G12C (n = 1), unclassifiable (n = 1)] and 11 (29%) were wild-type. In one BRAF -mutant LGSC, a KRAS G12V mutation was also detected in the same tumor at a low allelic frequency. Notably, 2 of 3 HGSCs had a KRAS mutation ( Table 3 ). Stratification of clinicopathologic features by gene mutations revealed a significant association with SBT type. Notably, all BRAF -mutant tumors were APST, and there is a high frequency of niLGSC in the wildtype group (5/12 (42%) vs 1/27 KRAS/BRAF -mutated tumors (p = 0.002), Table 4 ). It has previously been suggested that a characteristic morphologic feature of tumor cells in SBTs harboring the BRAF V600E mutation is the presence of abundant eosinophilic cytoplasm ( 27 ). These cells are typically round, present individually or in small clusters, and detached within cystic spaces or budding from the underlying tumor epithelium ( Fig. 3A , B ). In the present study, such cells were present diffusely in 4/5 (80%) BRAF -mutant tumors, in 4/22 (18%) KRAS -mutant (1 diffuse, 3 focal in extent), and in none of the wild-type tumors (0/12) (p = 0.001). Patient age, bilateral disease, FIGO stage, type of implants, microinvasion, and grade of subsequent carcinoma (as well as growth pattern, for LGSC cases) were not significantly associated with SBT genotype ( Tables 4 ). None of the BRAF -mutant APSTs had invasive implants. Microinvasion was identified in 1 BRAF -mutant case and was of the conventional type only, whereas microinvasive foci in KRAS or wild-type cases (n = 2) were of microinvasive carcinoma type. Mutation data was obtained from both SBT and subsequent carcinoma in 36 women, of which 33 (92%) matched pairs exhibited concordant mutation profiles (23 with concordant BRAF/KRAS mutations and 10 wild-type in both tumors). Considering only cases with a BRAF or KRAS gene mutation, 23 of 26 (89%) harbored an identical gene mutation in both tumors. To assess the likelihood that this level of concordance could occur by chance while maintaining the distribution of mutation frequencies, a permutation test was applied, randomly re-ordering the mutation status of SBT/carcinoma pairs. The maximum level of concordance seen in 1,000,000 random permutations of the data was 69% (18/26; p<0.000001). The features of the 3 discordant cases are summarized in Table 5 . In first case, the initial APST carried a KRAS G12V mutation, which was not present in the subsequent LGSC diagnosed in the liver after more than 17 years, which instead harbored a KRAS G12C mutation. In the second case, the primary ovarian niLGSC was wildtype for BRAF and KRAS . After 8 years, metastatic LGSC was found in the axillary and supraclavicular lymph nodes. Given the location of the metastasis and the patient’s known history of breast carcinoma, immunohistochemical staining for GATA3 and WT-1 was performed, and the results (GATA3 negative, WT1 diffuse staining) support the morphologic impression of gynecologic origin. Immunohistochemical staining and molecular analysis revealed that the BRAF V600E mutation/expression in tumor cells was not present in an adjacent focus of endosalpingiosis in the lymph node capsule. In the third case, the HGSC (which harbored a KRAS G12D mutation), arising in a background of a cystadenofibroma, was negative for the BRAF mutation detected in the APST diagnosed 25 years prior. The benign epithelium adjacent to the HGSC, extracted via LCM, was also positive for the KRAS mutation. For a subset of cases with normal tissue available for analysis (n = 11), the mutation data were supplemented with LOH analysis of 9 microsatellite/short-tandem repeat (STR) markers to assess for clonality ( Fig. 4 ). In 3 cases (#2, 4, and 5), including the discordant wild-type niLGSC/ BRAF -mutant LGSC (same patient as case #2 from Table 5 ), identical LOH of the same marker was detected in both SBT and carcinoma, supporting a clonal relationship. In another 3 cases, an LOH event was found only in carcinoma but not the SBT, which may represent the accumulation of genetic alterations during tumor progression (cases #3, 6, and 11). In Case #1 (same patient as case #1 from Table 5 ), the APST ( KRAS G12V ) and LGSC ( KRAS G12C ) showed LOH of opposite alleles at the D7S820 marker locus, which further supports the tumors being clonally distinct. In case #9, loss of an allelic marker was only detected in the primary ovarian tumor, presenting initially as a Stage III APST with non-invasive implants. The subsequent carcinoma was diagnosed 2.6 years later. While this may be due to an independent origin for these tumors, an alternative possibility is that dissemination occurred early, and subsequently the LOH event arose exclusively in the primary tumor and not in the extra-ovarian lesion. In the remaining cases (#7, 8, and 10), all profiled loci were heterozygous in both tumors, which was interpreted as an inconclusive result.

Details of the population-based study cohort have been previously reported ( 2 , 7 ). Briefly, all women with a pathologic diagnosis of SBT between 1978 and 2002 in Denmark were identified in the Danish Pathology Data Bank and/or Danish Cancer Registry. Diagnostic slides were retrieved from 1,487 cases and reviewed by 2 gynecologic pathologists (R.V. and R.J.K.) blinded to all clinical information. Diagnostic terminology and criteria have been described in our previous publication ( 7 ). Of note, SBTs were subdivided into APST and niLGSC. Centralized histologic review confirmed the diagnosis of SBT in 1,042 cases. Women with concurrent invasive carcinoma in the ovary (n = 17) were excluded. Of the remaining 1,025 cases, 42 (4%) were associated with a subsequent diagnosis of serous carcinoma, with a follow-up period of at least 11 years from the time of diagnosis of SBT, up to 36 years. No patients were lost to follow-up. Slides of the carcinomas were re-reviewed (R.V. and R.J.K.). For all SBTs, clinicopathologic and histologic features were assessed from review of H&E slides, relevant information from the surgical pathology report (i.e. related to laterality, staging, etc.), and clinical records. The following features were specifically assessed: SBT type (APST/niLGSC), involvement of one or both ovaries, FIGO stage, non-invasive or invasive implants (the latter designated as metastatic LGSC in the 2014 WHO Classification) ( 1 ), microinvasion (either conventional type microinvasion or microinvasive carcinoma) ( 1 ), and semi-quantitative assessment of tumor cells with prominent eosinophilic cytoplasm (focal/diffuse). The subsequent serous carcinomas were classified as low-grade or high-grade, based on the degree of nuclear atypia and mitotic index, as previously described ( 18 ). For LGSCs, architectural features were assessed (infiltrative versus exophytic micropapillary growth pattern). Information on anatomic sites of involvement was retrieved from imaging and/or clinical reports. All resection specimens were evaluated for the presence of endosalpingiosis. Immunohistochemical staining was performed to confirm the diagnosis and to further characterize the HGSCs in this cohort (n = 3), using antibodies against p53 (mouse monoclonal, Ventana), p16 (mouse monoclonal, Ventana), PAX8 (rabbit polyclonal, Proteintech Group, Chicago, IL), WT-1 (mouse monoclonal, Ventana, Tucson, AZ), ER (clone 6F11, mouse monoclonal, Ventana), PR (clone 16, mouse monoclonal, Ventana), and Ki-67 (mouse monoclonal, Ventana). In addition, WT-1 and GATA3 (mouse monoclonal, Biocare Medical, Concord, CA) staining was performed on one metastatic LGSC in an axillary lymph node, to exclude the possibility of a breast carcinoma. All immunohistochemical stains were performed with the automated XT iVIEW DAB V.1 procedure on the BenchMark XT IHC/ISH Staining Module (Ventana, Tucson, AZ). Antigen retrieval was carried out with CC1 (Ventana) and staining was detected with the I-View DAB detection system. For molecular studies, representative tissue blocks were accessible for 41 cases: 36 had both SBT and carcinoma tissues available, 3 had tissue for SBT only, and 2 had tissue for the carcinoma only. To ensure high confidence in genotyping results, two orthogonal methods for mutation detection were employed: traditional direct sequencing and ddPCR, a relatively new technology with a limit-of-detection of ~0.01%. ddPCR allows for identification of mutations missed by Sanger sequencing. However, each ddPCR probe is designed to test for one specific mutation. As a number of nucleotide substitutions are possible at codon 12 of KRAS, a validated multiplex assay screening for 8 possible KRAS mutations was first used, and positive results were followed up with separate assays for each of 4 different point mutations commonly found in SBT/LGSC ( 19 – 22 ). In a few cases, due to DNA quality and/or quantity, it was only possible to confirm the presence of a KRAS mutation (by the screening assay) without specification of the specific nucleotide alteration. Tumor tissue was manually microdissected from 10-micron-thick unstained sections in areas with >70% tumor cellularity identified on corresponding H&E slides. Laser-capture microdissection (LCM) was performed to enrich for lesional tissue on select cases with low cellularity. Microdissected tissues were subjected to genomic DNA extraction using the QIAamp DNA FFPE Tissue Kit (Qiagen, Valencia) as per the manufacturer’s instructions. Mutational analysis was performed by two independent methods: traditional Sanger sequencing of PCR-amplified products and ddPCR. Sanger sequencing of KRAS at exon 2 (including codons 12–13) and BRAF at exon 15 (including codon 600) was performed as previously described ( 23 ), with the following amplification primers: exon 15 of BRAF : forward 5′-TGCTTGCTCTGATAGGAAAATGA-3′ and reverse 5′-CCACAAAATGGATCCAGACAAC-3′; for exon 2 of KRAS : forward 5′-TAAGGCCTGCTGAAAATGACTG-3′ and KRAS reverse 5′-TGGTCCTGCACCAGTAATATGC-3′. Digital droplet PCR (ddPCR) was performed using the BioRad QX200 system as an orthogonal genotyping method. The following validated ddPCR mutation assays were obtained from Bio-Rad (Hercules, CA): BRAF p. V600E c. 1799T>A (dHsaMDV2010027); KRAS G12/13 Mutation Screening Kit (cat#1863506); KRAS p. G12C c.34G>T (dHsaMDV2510584); KRAS p. G12V c. 35G>T (dHsaMDV2510592); KRAS p. G12D c.35G>A (dHsaMDV2510596); and KRAS p. G12A c. 35G>C (dHsaMDV2510586). All samples were subjected to mutation analysis using the BRAF -V600E assay and the KRAS G12/G13 Mutation Screening Kit, a multiplex assay which screens for 7 common mutations in codons 12 and 13. Samples were subjected to KRAS G12C, G12V, G12D, and G12A mutation-specific ddPCR assays for definitive genotyping if found to carry a KRAS mutation by multiplex ddPCR. The ddPCR reaction was comprised of 2X ddPCR Supermix (no dUTP), 0.5 μL of Uracil-DNA Glycosylase (UDG) (New England BioLabs, Ipswich, MA), 1μL primer/probe assay reagent, and sample, up to a total volume of 20 μL. Droplets were generated using the Droplet Generator with an eight-channel DG8 cartridge and cartridge holder. Droplets contained 70 μL of DG oil per well and 20 μL of fluorescent PCR reaction mixture and were transferred to a 96-well PCR plate, which was subsequently heat-sealed with foil. PCR amplification was performed with the following cycling conditions: initial incubation at 37°C for 30 min, then 10 min at 95°C, followed by denaturation for 30 s at 94°C, annealing for 60 s at 55°C for 40 cycles; and final incubation for 10 min at 98 °C, ending at 4°C. After amplification, the 96-well plate was placed into the Droplet Reader (Bio-Rad, Hercules, CA). Data were analyzed using the QuantaSoft analysis software (Bio-Rad). The threshold for a positive mutation call was set at an allelic frequency of ≥1.0%. LOH analysis was performed for cases that had sufficient high-quality DNA from SBT, carcinoma, and adjacent normal tissue. Fluorescent-labeled primer sets for 9 microsatellite markers, representing chromosome arms frequently lost in low-grade serous tumors ( 24 – 26 ), were used for PCR amplification. Fragment lengths were analyzed by capillary electrophoresis (Applied Biosystems 3730xl DNA Analyzer), and chromatograms were visualized with PeakScanner v. 1.0 software. Frequency distributions were compared by the Fisher exact test or χ 2 test, when the number of groups exceed 2×2. The Mann-Whitney U test was used to compare duration of time from SBT to carcinoma between different groups. The significance of co-occurrence of the concordant mutational status between SBT and subsequent carcinoma was assessed by a permutation test. The test statistic was the number of SBT/carcinoma pairs with concordant mutational status. To generate a null distribution, the mutational status of SBT/carcinoma pairs was randomly reordered. All statistical tests were two-sided.

We have estimated the absolute risk and identified risk factors for subsequent development of serous carcinoma in previous epidemiologic studies analyzing this entire population-based cohort of women with SBT ( 2 , 6 , 7 ). Even though SBT is generally associated with a good prognosis, some women eventually develop serous carcinoma, including some without these risk factors. It has been unclear in the prior literature whether these subsequent neoplasms represent tumor progression versus independent primary tumors. In the present study, we characterized the clinicopathologic and molecular features of SBTs and matched subsequent serous carcinomas from the same cohort, the largest analysis of its kind in the literature. Established risk factors for subsequent development of serous carcinoma in women with SBTs include bilateral ovarian involvement, advanced stage, invasive implants, and niLGSC histologic type (for stage I cases) ( 6 , 7 ). Accordingly, the frequencies of these features are higher in SBTs associated with subsequent carcinoma compared to SBTs overall (with or without subsequent carcinoma), reported in our prior studies ( 2 , 6 , 7 ). It is important to note that a third of the cases in the present study were stage I APST (i.e., women that are not considered as being at risk) though the possibility of an unsampled focus of invasive LGSC in the original SBT or omentum can never be entirely excluded, even for tumors that appear to be adequately sampled. However, to place things in perspective, the 14 patients with stage I APST that subsequently developed carcinoma represented only 1% of all women (all stages combined) with either APST or niLGSC (n=1,025). Also, in one of our prior epidemiology studies of the entire cohort, the overall survival of women with stage I APST was not significantly different from that of the age-matched background population in Denmark ( 2 ). Nonetheless, in the analysis by Longacre et al , 19 of 276 (7%) women with SBT developed LGSC ( 3 ), of which 9 were stage I “typical” SBTs (2 with microinvasion). LGSC was diagnosed from 22 to 310 months after the initial primary ovarian tumor ( 3 ). Silva et al reported a series of 10 stage I SBTs associated with subsequent LGSC ( 12 ). It was unclear how many of these were of typical histologic type, but development of carcinoma occurred up to 39 years after diagnosis of SBT. Collectively, these findings highlight the fact that absence of risk factors for development of carcinoma does not guarantee a non-adverse outcome. They also raise the question of whether subsequent carcinomas in women with low-risk features actually represent primary peritoneal tumors that are clonally unrelated to the prior SBT. In the pathogenesis of ovarian low-grade serous tumors (starting with serous cystadenoma/adenofibroma, and thought to progress sequentially to APST, niLGSC, and then invasive LGSC), the 2 most important genes involved in this process are KRAS and BRAF . Mutations in these genes are mutually exclusive and result in activation of the MAPK signaling pathway. In the present study, significant associations between genotype and SBT pathologic features were identified. niLGSC was more common among cases wild-type for KRAS/BRAF , and BRAF -mutated tumors more frequently contained prominent tumor cells with abundant dense eosinophilic cytoplasm. This morphologic feature and its association with the BRAF mutation has been previously reported, and the eosinophilic cells have been shown to overexpress p16, a marker of cellular senescence ( 27 , 28 ). In other studies of unselected SBTs (with or without subsequent carcinoma), BRAF mutations were found in APST, niLGSC, and invasive LGSC, in order of decreasing frequency ( 29 , 30 ). While BRAF mutations were more common in low-stage SBTs ( 29 ) and associated with lower frequencies of tumor recurrence and improved prognosis, KRAS mutations were associated with invasive recurrence ( 17 , 29 ). For these reasons, and the putative senescent phenotype of tumor cells with eosinophilic cytoplasm, it has been proposed that BRAF -mutations may have a suppressive effect on malignant transformation, which is not evident in KRAS -mutated or wild-type tumors ( 27 ). As part of a future study, we are presently analyzing KRAS/BRAF mutational status in SBTs (including cases without subsequent carcinoma) from our population-based cohort to more clearly determine if genotype truly stratifies women with SBT into different risk groups with respect to development of subsequent carcinoma. While the vast majority of subsequent serous carcinomas were low grade, rare HGSCs were observed ( 3 , 8 , 12 ). This is not surprising as other investigators have found SBTs rarely associated with HGSC either as a synchronous or metachronous tumor ( 9 , 31 – 34 ). The pathogenesis and molecular profile of ovarian low-grade serous tumors (SBT/invasive LGSC) and HGSC are different, and pathways leading to the development of LGSC and HGSC are usually unrelated. Specifically, invasive LGSC evolves from ovarian SBT, driven by BRAF and KRAS mutations, while most HGSCs are thought to originate from the fallopian tube, arising from the precursor lesion, serous tubal intraepithelial carcinoma, with mutation of TP53 being an important early step. In this series, 2 of 3 HGSCs associated with prior SBT harbored KRAS mutations, which is a rare event in HGSCs overall (1% of all cases) ( 35 ). While TP53 mutational analysis was not performed in the current series, 2 of the HGSCs exhibited diffuse immunohistochemical expression of p53 while 1 showed a patchy pattern (in general, these patterns are consistent with mutation and wild-type TP53 status, respectively). In a study by Dehari et al , 2 of 3 cases of HGSC associated with APST also had KRAS mutations in both components, and all 3 cases did not contain a TP53 mutation in either component ( 9 ). Thus, HGSCs associated with SBT may have a molecular profile/pathogenesis distinct from the majority of HGSCs. Although women with SBTs may subsequently develop serous carcinoma, it has been unclear in the prior literature whether the carcinoma in this setting represents tumor progression versus an independent primary neoplasm. Evidence supporting tumor progression include: 1) specific clinicopathologic factors in women with SBTs have been identified which confer a statistically significant increased risk for subsequent carcinoma; and 2) a substantial proportion of cases have concordant KRAS/BRAF mutations in the SBT and metachronous carcinoma. However, some cases have features supporting interpretation as independent primary tumors. These include women initially diagnosed with stage I APST, cases with a prolonged latency between SBT and subsequent carcinoma (e.g. >20 years), discordant mutations between SBT and carcinoma, and the rare HGSCs associated with prior SBT, which do not conform to the conventional model of pathogenesis. The significantly longer time interval between the metachronous tumors for HGSC cases, compared to the LGSCs in our cohort, suggests that more time is needed for the accumulation of additional genetic alterations for high-grade transformation or that they are clonally unrelated to the primary ovarian tumor. The higher frequency of endosalpingiosis in SBTs associated with subsequent carcinoma relative to those without subsequent carcinoma, was originally described by Silva et al ( 12 ) who suggested that endosalpingiosis can be a direct precursor of primary peritoneal low-grade serous tumors. This possibility may explain some cases of serous carcinoma arising “ de novo ” independent of the primary ovarian SBT. Identical mutation profiles were observed in 92% of SBT/carcinoma pairs (i.e., BRAF -mutant/ BRAF -mutant, KRAS -mutant/ KRAS -mutant, or wild-type/wild-type). Applying a more conservative estimate by excluding cases with concordant wild-type status, which may be considered molecularly uninformative, this still leaves 89% of cases with identical BRAF/KRAS mutations in both tumors. While these findings can provide evidence for a clonal relationship between SBT and subsequent carcinoma, they may also be due to random chance. Given the limited spectrum of hotspot mutations seen in low-grade serous tumors, a plausible scenario could involve a “field effect,” exerting a common mutagenic process acting on multiple anatomic sites in the same patient, which results in an identical KRAS or BRAF mutation. However, using a permutation test for the paired SBTs/subsequent carcinomas in this study, the observed frequency of genotype concordance (89%) was significantly higher than that expected by chance alone (69%). The finding of shared allelic losses between SBT and subsequent serous carcinoma in a subset of cases (3 of 11 cases subjected to LOH analysis) further supports a clonal relationship between paired metachronous tumors. The collective evidence supports progression of SBT into invasive serous carcinoma as the explanation for most cases of serous carcinomas that arise in women with SBT. Nevertheless, without a global analysis of somatic genetic variants, it is not possible to confidently determine which individual cases were clonally related versus independent primaries, particularly for cases that have some features suggesting independent primaries while simultaneously also having features suggesting tumor progression (e.g., stage I APST with 22-year time interval between initial diagnosis and subsequent HGSC, in which there was no identified endosalpingiosis and that both tumors shared an identical KRAS mutation); however, for cases that have tumor progression with a >10-year latency interval, the long time-interval is likely due to the slow natural history of either stage I APST or low-volume extra-ovarian disease. Additionally, since LOH events were infrequent (as assessed by a targeted panel of markers), full assessment of clonality by this type of analysis is also limited In conclusion, recognition of established risk factors for development of subsequent serous carcinoma in women with SBTs is helpful for clinical management. Nevertheless, as some women without risk factors can still develop carcinoma, sometimes decades later, lifelong surveillance for all women with SBT is important (regardless of whether the subsequent carcinoma represents true tumor progression or origin from endosalpingiosis as an independent primary tumor). Among SBTs associated with subsequent serous carcinoma, there is a high frequency of KRAS mutations overall, a higher frequency of wild-type status (for BRAF and KRAS ) in niLGSCs relative to APSTs, and a significant association between the BRAF V600E mutation and tumor cells with abundant eosinophilic cytoplasm. Despite the inherent limitations of clonality assessment based only on BRAF/KRAS mutational analysis, the combined clinicopathologic and molecular findings in this study suggest that most serous carcinomas that occur after a diagnosis of SBT represent tumor progression and that a minority are likely independent primary tumors.

Ovarian serous borderline tumor (SBT) is a known precursor of low-grade serous carcinoma (LGSC) and is divided into 2 types: atypical proliferative serous tumor (APST, also referred to as typical serous borderline tumor) and non-invasive low-grade serous carcinoma (niLGSC, also referred to as micropapillary serous borderline tumor) ( 1 ). These tumors are typically diagnosed at early stage and associated with an excellent prognosis. However, compared to the general population, women with SBT are at increased risk for subsequent development of serous carcinoma, which occurs in up to 7% of patients ( 2 – 5 ). Established risk factors for subsequent serous carcinoma in women with SBT include bilateral ovarian involvement, ovarian surface involvement, advanced stage, invasive implants, post-surgical residual disease, and niLGSC histologic type (for stage I cases) ( 6 , 7 ). However, carcinomas can still occur in some women after a diagnosis of SBT without risk factors, such as those with stage I APST. The time to development of subsequent carcinoma is typically several years, yet in some instances, carcinomas can develop decades later ( 3 , 7 , 8 ). The majority of carcinomas subsequently developing after serous borderline tumors are invasive LGSC, but high-grade serous carcinoma (HGSC) can also be encountered ( 9 , 10 ). Important molecular events in SBT/LGSC are mutations in either KRAS or BRAF . In addition, endosalpingiosis is commonly associated with SBT ( 7 , 11 – 13 ), and limited molecular data in the literature have shown identical KRAS mutations in SBTs and endosalpingiosis from the same patient ( 14 , 15 ). Moreover, some evidence has suggested that, in a subset of cases, lymph node involvement by SBT may actually represent a primary lymph node lesion arising from endosalpingiosis ( 13 ). Considering the above observations, it is unclear whether all carcinomas subsequently developing after SBT represent true tumor progression, independent primary tumors arising from endosalpingiosis, or a combination of these 2 scenarios. Furthermore, the data in the literature comparing the molecular features of matched metachronous SBTs and serous carcinomas are limited ( 16 , 17 ), and the pathologic features of SBTs that are associated with subsequent serous carcinoma have not been fully characterized with respect to genotype. The aims of this study were to characterize the relationship between matched metachronous SBTs and serous carcinomas and analyze the integrated clinicopathologic-molecular/genetic features of these types of cases from a population-based cohort, including identifying any specific genotype-phenotype associations.