Proliferation of the Fallopian Tube Fimbriae and Cortical Inclusion Cysts: Effects of the Menstrual Cycle and the Levonorgestrel Intrauterine Contraceptive System.

We identified 60 patients with tissue blocks containing sufficient FTF tissue for immunohistochemistry; an ovarian tissue block with at least one CIC successfully stained was obtained from 50 of these 60 patients. The breakdown of samples by menstrual cycle and BRCA status is shown in Table 1 . The median FTF Ki67%+ was 4.9% in the follicular-phase samples and 1.5% in the luteal-phase samples (p fl = 0.003; Table 2 ). The median Ki67%+ in the postmenopausal samples was lower than in the follicular samples (1.7% vs 4.9%; p fp = 0.051) and approximately the same as that of the luteal phase samples (1.7% vs 1.5%; p lp = 0.70). The median PAX8%+ was 70%, 72% and 90% in the follicular, luteal and postmenopausal samples, resulting in increased Ki67%+ figures for secretory cells (PAX8 cells) of 7.0%, 2.1% and 2.2%. The decreased level in the postmenopausal samples compared to the follicular-phase samples was highly statistically significant (2.2% vs 7.0%; p fp = 0.006). The BRCA1+ and BRCA2+ patients showed the same pattern of FTF Ki67%+, as did the small number of other patients. None of the differences between the BRCA1+ and BRCA2+ patients were large or statistically significant. There were no PAX8-/CAL- CICs, and the Ki67%+ values were extremely low in the PAX8-/CAL+ CICs. The Ki67%+ results for the PAX8+/CAL- CICs and the PAX8+/CAL+ CICs were similar. The combined PAX8+ CIC Ki67+% results are shown in Table 2 ; there was little difference between the follicular, luteal and postmenopausal results: 1.9%, 1.7% and 1.6%. These Ki67%+ figures increased to 2.1%, 2.0% and 2.5% when adjustment was made for the proportion of cells that were PAX8+; the median PAX8%+ was 88%, 90% and 89% in the follicular, luteal and postmenopausal samples. None of the differences between the BRCA1+ and BRCA2+ patients were large or statistically significant. The only marked difference between the Ki67%+ results for the FTF and PAX8+ CICs was the increased Ki67%+ of the follicular FTF samples: 4.9% in the FTF and 1.9% in the PAX8+ CICs (ranksum p=0.067), and 7.0% in the FTF PAX8+ cells and 2.1% in the PAX8+ cells in the PAX8+ CICs (ranksum p=0.013). When analysis was restricted to the samples (n=11) where matched pairs of FTF and PAX8+ CICs were available ( Figure 1 ), the magnitude of the differences remained the same for the unadjusted comparison and increased in the PAX8%+ analysis. The statistical significance of the difference of the adjusted results decreased from p=0.013 to p=0.050 because of the decreased sample size of the FTF data (decreased from 19 to 11). The median FTF Ki67%+ for the 10 women using the Lng-IUS was 8.3% ( Table 3 ). This was higher than the median FTF Ki67%+ in the normally cycling premenopausal patients described in Table 2 (follicular 4.9%, luteal 1.5%) even after the normally cycling results had been adjusted for PAX8%+ (follicular 7.0%, luteal 2.1%). The four patients who were long-term users of the Lng-IUS had FTF Ki67%+ values of 0.4%, 1.6%, 10.0% and 10.5%. The median age of the women using the Lng-IUS was 41; no different from the median ages of the follicular and luteal samples of 42 and 41.5 and there was no apparent relation of age to the FTF Ki67%+ results. There was no evidence that the FTF proliferation of patients on the LngIUS was lower than was seen in normally cycling women.

All studies described here were approved by the Institutional Review Board (IRB) at Memorial Sloan Kettering Cancer Center (MSKCC) and followed recognized ethical guidelines, including the Declaration of Helsinki, the Belmont Report and the US Common Rule. All women who participated as subjects in these studies had signed an IRB-approved banking protocol and research authorization permitting use of their tissue for IRB-approved research. Women, not recently pregnant or using exogenous hormones, who had undergone a risk-reducing salpingo oophorectomy (RRSO) in 2000–2014 were considered for inclusion in the study. Only patients with an intact uterus, bilateral ovaries and fallopian tubes were eligible. Further exclusion criteria were pregnancy, exogenous estrogen, progestin, or anti-estrogen therapy within 6 months prior to RRSO, or any history of cancer (except non-melanoma skin cancer), or tissue samples showing early malignancy. For patients identified as premenopausal in their medical record at RRSO, we collected material only if concurrent endometrial sampling had been performed. We used this endometrial tissue to determine the phase of the menstrual cycle (follicular vs luteal) at RRSO as the dating of the endometrium based on histological features as described in the classic paper by Noyes and colleagues ( 30 ) is a well-established practice in pathology. For patients identified as postmenopausal in their medical record at RRSO, we confirmed this against their endometrial sample (available in 18 of the 21 cases used in this study) using standard criteria ( 31 ); all were confirmed. We obtained the BRCA1/2 status of the patients signing an IRB-approved release permitting use of their germline genetic information. This study was approved by the MSKCC IRB and did not require any additional patient consent. Archival tissue specimens, which had been evaluated using the SEE-FIM protocol ( 32 ), were used for this study. One of us (KJP) reviewed all hematoxylin and eosin slides for each patient and chose by visual inspection the FTF block containing the greatest amount of tissue and the ovarian block (from the same patient) containing the greatest number of CICs. Ovarian cysts were counted if they were lined by tubal type epithelium comprised of columnar secretory and/or ciliated cells, as well as those with flattened or cuboidal cells with or without cilia. Other types of cysts, such as endometriosis, corpus luteum cysts or follicle cysts were excluded. Adjacent 3 µm sections were cut, deparaffinized and hydrated. We used Ki67 as a proliferation marker, paired-box gene 8 (PAX8) as a Müllerian cell marker, and calretinin (CAL) as a mesothelial cell marker. Most, if not all, OSE cells are negative for PAX8 and positive for CAL (PAX8-/CAL+) ( 33 , 34 ). The FTF secretory cells are PAX8+/CAL-; the FTF ciliated cells are PAX8-/CAL- ( 35 ). CICs are generally regarded as being either PAX8-/CAL+ or PAX8+/CAL- ( 35 , 36 ), but we recently found that approximately 16% were PAX8+/CAL+ and that approximately 40% of these CICs demonstrated PAX8+ and CAL+ staining in the same cells ( 37 ). The following immunohistochemical stains were used: Ki67 [mouse monoclonal antibody (MIB-1) at 1:100; Dako Cytomation, Carpenteria, CA, USA]; PAX8 [rabbit polyclonal antibody (PAb) at 1:1000; Proteintech, Rosemount; IL, USA]; and calretinin [rabbit PAb at 1:100; Zymed, San Francisco, CA, USA] using our institution’s standard methods. These methods include antigen retrieval CC1 (mild) regimen for CAL and PAX8 antibodies; primary Ab-s incubation on Ventana Discovery XT platform, 60 min; secondary biotinylated anti-mouse and anti-rabbit Ab-s incubation on Ventana platform, 60 min. The Ventana DAB MAP kit was used to complete the staining process. Slides were digitized and analyzed using the Aperio® scanning and analysis system. Digitized fallopian tube slides were analyzed to obtain counts of total FTF cells and number of cells that were positive for each immunohistochemical marker in five randomly chosen subareas spread across the FTF. The surface epithelial cell layers of each subarea were outlined using the Aperio Annotations tool and analyzed using the Aperio Nuclear Algorithm tool (~400 cells per slide). Digitized ovarian slides were analyzed to obtain counts of the total cells in individual CICs, and the number of cells that were positive for each immunohistochemical marker. To be included as a CIC, the cyst had to be present in all consecutive sections of the Ki67, PAX8 and CAL slides. The cell layer lining each cyst was outlined using the Annotations tool and analyzed using the Nuclear Algorithm tool. Only CICs that were completely enclosed within the ovarian parenchyma were analyzed. The Nuclear Algorithm tool allows for adjustment of three parameters. These were optimized across all stains in a subset of cases to achieve close agreement with the counts made by one of us (KJP). While calretinin staining is both nuclear and cytoplasmic, we identified only rare instances when the staining was solely cytoplasmic; so that the Nuclear Algorithm accurately approximated the count of total positivity. The percentages of cells positive for Ki67 and PAX8 were calculated for each FTF and each CIC. For CICs the percentage of cells positive for CAL were also calculated. An FTF was only considered for analysis if it was successfully stained for Ki67 and PAX8. A CIC was only considered for analysis if it was successfully stained for Ki67, PAX8 and CAL, and the associated FTF was successfully stained for Ki67 and PAX8. The CICs were classified into four groups (PAX8-/CAL-, PAX8-/CAL+, PAX8+/CAL-, PAX8+/CAL+) by designating PAX8+ and CAL+ if ≥10% of the cells showed nuclear staining for PAX8 or nuclear staining for CAL ( 34 , 35 ). The total number of cells and the total number of cells positive in the PAX8-/CAL+ CICs of a patient was calculated by combining the results from all PAX8-/CAL+ CICs from an individual subject, and similarly for PAX8+/CAL+ and PAX8+/CAL- CICs. The distribution of the Ki67 results was highly skewed, and the results are presented as medians and in the groups (0%, >0% - <1%, ≥1% - <5%, ≥5% - <15%, and ≥15%) used by George et al ( 17 ) or combining the last two groups. The statistical significance between the follicular and luteal results, p fl , were calculated using the non-parametric rank-sum test. This test was also used to compare the follicular and postmenopausal results, p fp ; and the luteal and postmenopausal results, p lp . We estimated the Ki67%+ in secretory cells (PAX8+ cells) by adjusting the total number of Ki67+ and Ki67- cells in the Ki67-stained slides by multiplying the total numbers by the percentage of cells staining PAX8+ in the PAX8-stained slides. This avoided the necessity of requiring double staining of Ki67 and PAX8 which we did not have the wherewithal to carry out. We estimated the Ki67%+ in PAX8+ cells within the PAX8+ CICs in a similar manner. The statistical significance of the differences between the FTF and CIC results were made using the non-parametric rank-sum test and checked with the non-parametric signed-rank test, which compares results within individual patients. All statistical analyses were conducted using Stata 16 (Stata Corporation, College Station, TX, USA). All statistical-significance levels (p-values) quoted are 2-sided. We used the same inclusion and exclusion criteria for women in this study as in the menstrual cycle and menopausal status studies described above except that they had to be premenopausal and using a Lng-IUS at RRSO. We identified four such patients, noted as premenopausal at RRSO in their medical record, who had a Lng-IUS in place at RRSO (three for approximately 2 years; the other for an unknown duration). In addition, we inserted a Lng-IUS in six volunteers, who met the same inclusion and exclusion criteria, approximately 3 months prior to RRSO. These six volunteers were confirmed to be premenopausal by hormone assay for estradiol. All these patients gave signed informed consent for use of their tissue for research, including use of their genetic information. The recruitment and inclusion of these six patients was approved by the MSKCC IRB. This study comparing information on Ki67%+ in LngIUS users and in normally cycling premenopausal women was approved by the MSKCC IRB and registered with ClinicalTrials.gov (Trial Registration: NCT02477202 ). The data underlying this article are available on reasonable application from the corresponding author.

The FTF follicular phase Ki67%+ was 3.3-fold greater than that of the luteal phase (4.9% vs 1.5%), confirming the results previously reported ( 16 , 17 ). A similar increase was also seen after adjusting the Ki67%+ values for the PAX8%+ values. These results provide an explanation for the protective effects of COC use when progestin exposure is increased from the two weeks of the normal cycle to the three weeks of active pill use per cycle. Although 52 of the 60 patients were BRCA1/2 mutation carriers, George et al ( 17 ) found “no significant difference in the proliferative index in histologically normal FTE between BRCA1/BRCA2 and non-BRCA” in their large study of fimbriae (BRCA1/BRCA2: n=94; non-BRCA: n=39). The FTF Ki67%+ in the postmenopausal period was 1.7%; this had not previously been studied. The average Ki67%+ over the menstrual cycle is approximately (4.9% + 1.5%)/2 = 3.2%, roughly double the proliferation rate in the postmenopausal period. Calculation of the effect of a 10-year later menopause (age at menopause of 55 years vs 45 years) using these figures with a simple model of ovarian cancer incidence ( 12 ), shows a relative risk of 1.62. This relative risk of 1.62 compares to the relative risk of 1.93 reported by Rosner et al ( 13 ) using the Nurses’ Health Study results and 1.39 reported by Li et al ( 14 ) using the EPIC Study results. The 65% decrease (4.9% vs 1.7%) in the FTF Ki67%+ rate between the follicular phase and postmenopausal period is slightly less than one would have predicted based on the approximately 85% decline in serum estradiol level between the follicular phase and postmenopausal period ( 18 , 19 ); a careful study of this would require obtaining serum estradiol and sex-hormone binding globulin concentrations at the time of RRSOs. Proliferation in PAX8-/CAL+ CICs, which resemble OSE, was extremely low as has been previously reported ( 38 ). In contrast to the results for FTF, in the PAX8+ CICs there was little difference in Ki67%+ between the follicular and luteal phase samples. The FTF Ki67%+ in the follicular phase (4.9%) was 2.6-fold higher than the proliferation of the PAX8+ CICs (1.9%): this increased to 3.3-fold higher when proliferation was estimated for the PAX8+ cells (7.0% vs 2.1%). In the only other published report comparing the Ki67%+ of the FTF to that of the PAX8+ CICs, Li et al ( 38 ) studied 50 patients and found that the Ki67%+ was 2.9-fold higher in the FTF than in the PAX8+ CICs, but no information on age or menopausal status was given. We previously suggested, following the arguments put forward by Karnezis et al ( 39 ), that the reason intraepithelial carcinomas analogous to serous tubal intraepithelial carcinomas (STICs) are never seen in CICs is that the microenvironments of the ovary and fallopian tube are quite different. The higher proliferation rate of FTF compared with CICs could also provide part of the explanation. We restricted attention to the FTF because the overwhelming majority of HGSOCs arise there. The proliferation rate in the fallopian tube ampulla was measured by Donnez et al ( 16 ) and George et al ( 17 ); it was no different from the proliferation rate in the FTF. The reason for the preponderance of HGSOCs arising in the fimbriae is, again following the arguments of Karnezis et al ( 39 ), possibly due to developmental differences ( 40 ) and toxic effects of follicular fluid ( 41 ). Our results shed no light on why CICs appear to be linked to low-grade serous neoplasia or why low-grade and high-grade serous tumors appear to arise from different tissues ( 39 , 42 , 43 ). Our study has a number of strengths: It is the first study of proliferation of the FTF in postmenopausal women and the first study of CIC proliferation by cycle phase and menopausal status. Our approach was also to make actual counts of cells as opposed to estimating counts: the cell counts were conducted using a digital imaging platform in combination with expert pathologist review. It also has several weaknesses: Most of the patients studied were BRCA1 or BRCA2 mutation carriers, the estrogen and progesterone serum levels of the patients at RRSO were not known, and the determination of the proliferation of the secretory cells was not determined directly but by estimating the proliferation using the proportion of FTF cells staining positive for PAX8. George et al ( 17 ) found that FTF cell proliferation was no different between mutation carriers and non-carriers, strongly suggesting that our results are generalizable to non-carriers. However, it would clearly be better if future studies include a significant number of non-carriers; for FTF studies, this can now be achieved much more easily because one can include FTFs from opportunistic salpingectomies, which are now frequently performed in place of tubal ligations. Future studies should determine estrogen and progesterone serum levels at RRSO so that one could obtain information on the dose-response relationship of these hormone levels to proliferation. Future studies would benefit from dual staining with Ki67 and PAX8 to directly measure proliferation of secretory cells. This is the first study of proliferation of the FTF in premenopausal women using the Lng-IUS. The FTF Ki67%+ in patients using the Lng-IUS was higher, although not statistically significantly higher, than in patients not using hormonal contraception. Direct studies of the progestin concentration in the FTF of women using an Lng-IUS are needed to better understand this observation. The two recent small epidemiological studies on the association of the Lng-IUS with ovarian cancer ( 25 , 26 ) found reduced relative risks of 0.84 (95% CI 0.53–1.35) and 0.53 (95% CI 0.32–0.88). In contrast to previous publications, these studies adjusted for known ovarian cancer risk factors. If these reduced risks are confirmed in larger studies with detailed histopathology of the ovarian cancers, our results, which also need confirmation, suggest that any reduced risk of HGSOC will not have been due to lower proliferation of the fallopian tube fimbriae. Overall, our study has shown the difference in proliferation in the FTF at different reproductive phases of a woman’s life, providing some insight into the mechanism of action for the protection against HGSOCs of COC use. The finding that proliferation of FTF in women with an Lng-IUS was similar to women not on hormonal contraception was unexpected. Studying women on different hormonal regimens is critical. This should include women taking menopausal estrogen therapy and menopausal estrogen plus progestin therapy to better understand their differing impact on ovarian cancer risk ( 44 ). More generally, little is truly understood about the underlying mechanisms of well-established risk and protective factors for ovarian cancer. This should be a focus of future research to both shed light on ovarian cancer etiology and to also identify opportunities for primary prevention of this deadly disease.

High-grade serous ovarian carcinoma (HGSOC), the most lethal of ovarian cancers, accounts for approximately 70% of epithelial ovarian cancers ( 1 ). Epithelial ovarian cancers are of Müllerian cell origin and the long-favored “incessant ovulation hypothesis” for the etiology of ovarian cancer postulated that the cancers arose from the mesothelial ovarian surface epithelium (OSE) entrapped in ovarian cortical inclusion cysts (CICs) that underwent metaplastic transformation to Müllerian-type epithelium ( 2 , 3 ). Although some HGSOCs may arise from Müllerian CICs with the cell of origin being OSE (as described above) or Müllerian fallopian tubal cells entrapped within the ovary post-ovulation ( 3 , 4 ), it is now clear that most arise directly from the Müllerian fallopian tube fimbriae (FTF) ( 5 – 7 ). A major way in which hormones affect cancer risk is by altering the proliferation rate of the cells of origin ( 8 , 9 ) – this is clearly seen with endometrial cancer where estrogen ‘unopposed’ by a progestin is the major risk factor ( 10 ). That the risk of HGSOC is also affected by ovarian hormones is shown by the increased risk from use of menopausal estrogen therapy ( 11 ) and is suggested by the fact that the rate of increase with age of the incidence of HGSOC after menopause is much slower than the rate of increase in younger women ( 12 – 15 ). The premenopausal period is marked by relatively high levels of estrogen throughout the menstrual cycle, and relatively high levels of progesterone in the luteal phase. Donnez et al ( 16 ) found that the FTF mitotic index was 4.0-fold higher in the follicular phase than in the luteal phase of premenopausal women. Similar results were reported by George et al ( 17 ), who found that the FTF proliferation rate, as measured by Ki67% positivity (Ki67%+), was 3.7-fold higher in the follicular phase than in the luteal phase. The reduced FTF proliferation in the luteal phase suggests that proliferation may be reduced by progesterone exposure. There are no studies of FTF proliferation in the postmenopausal period, when estrogen levels are 10%−15% of premenopausal levels and progesterone is effectively not present ( 18 , 19 ). There are no studies of CIC proliferation either by menstrual cycle phase or menopausal status. The FTF consist of secretory, ciliated, and rare non-proliferating intercalated cells. In comparison with secretory cells, ciliated cells rarely proliferate ( 20 , 21 ), and a more complete proliferation rate analysis should also include proliferation of secretory cells (PAX8+ cells, see below). There are no studies reporting proliferation of secretory cells. Combination (estrogen + progestin) oral contraceptives (COCs) yield a more potent progestin over a 28-day cycle than is seen in a physiological ovulatory cycle ( 22 ); this can account for their significant protective effect on ovarian cancer including in BRCA1/2 mutation carriers ( 23 , 24 ). The standard levonorgestrel (Lng) intra-uterine contraceptive system (Lng-IUS; Mirena®) produces a circulating level of Lng that is only approximately 10% of that seen in women using a standard Lng-COC ( 25 , 26 ). Although the one study of the concentration of Lng in fallopian tube tissue was similar in women using a Lng-IUS or a Lng-COC ( 27 ), it was of homogenized tissue from the whole fallopian tube and was too small (3 and 2 samples respectively) to allow any conclusions to be drawn. The two small epidemiologic studies ( 28 , 29 ) of ovarian cancer and Lng-IUS use, which adjusted for other ovarian cancer risk factors, showed a decreased risk of ovarian cancer, one statistically significant. There are no studies of the proliferation of FTF (or CICs) in women using a COC or an Lng-IUS. In the present study: (i) we studied FTF and CIC cell proliferation in premenopausal and postmenopausal women to explore if differences in cell proliferation of FTF and CICs may help explain the preponderance of HGSOCs arising in the FTF rather than in CICs; (ii) we studied FTF cell proliferation in premenopausal and postmenopausal women to investigate whether there was a reduced rate in the postmenopausal period which would provide an explanation for the slower rate of increase in ovarian cancer incidence after menopause; and (iii) we studied FTF cell proliferation in premenopausal women using a standard Lng-IUS to evaluate whether this would provide a mechanism for the epidemiologic findings that suggest that Lng-IUS use may be associated with a reduced risk of ovarian cancer.