Identification of cell-surface markers for detecting breast cancer cells in ovarian tissue

Immunohistochemistry was performed on 4-μm thick FFPE sections of control ovaries and 4-μm thick slices of a tissue microarray (TMA) containing invasive breast tumor cores. To generate the TMA, tissue biopsies measuring 1.0 mm in diameter were taken in triplicate from representative regions of the FFPE tumor samples and arrayed into a new recipient paraffin block using TMA Master (3DHistech, Hungary). The tissue sections were deparaffinized in xylene, rehydrated in a stepwise series of graded alcohol solutions, and rinsed in distilled water. After blocking endogenous peroxidase activity with 0.3 % hydrogen peroxide for 20 min, heat-induced antigen retrieval was performed by placing the slides in EnVision Flex Target Retrieval Solution high pH/low pH in PT Link (Dako, Denmark). EpCAM and αvβ6 integrin epitopes were unmasked by 30-min incubation with 0.125 % trypsin and 0.4 % pepsin, respectively, at 37 °C. The sections were incubated overnight in a humidified chamber at room temperature with primary antibodies against Her2/neu (ERBB2, rabbit polyclonal, Dako), E-cadherin (NCH38, mouse monoclonal, Dako), EpCAM (323/A3, mouse monoclonal, provided by the Department of Pathology, LUMC, the Netherlands), CEA (A0115, rabbit polyclonal, Dako), αvβ6 integrin (6.2A1, mouse monoclonal, Cell Essentials), uPAR (ATN615, mouse monoclonal, kindly provided by Prof. A.P. Mazar, Northwestern University, Evanston, IL), or EMA (E29, mouse monoclonal, Dako); all primary antibodies were used at their predetermined optimal dilution. Some sections were incubated with an antibody against FR-α (26B3.F2, mouse monoclonal, Biocare Medical) for 60 min in accordance with the manufacturer’s instructions. After incubation with the primary antibody, the sections were rinsed with PBS, incubated with secondary antibody (anti-mouse or anti-rabbit EnVision; Dako) for 30 min, and visualized using liquid DAB+substrate buffer (Dako). The sections were counterstained with Mayer’s hematoxylin solution, dehydrated, and permanently mounted with Pertex (Leica Microsystems, Germany). For each immunostain, a positive control expressing the antigen of interest was included. The primary antibody was omitted as a negative control.

The immunostained slides were scanned using a Pannoramic MIDI digital slide scanner (3DHistech, Hungary). Immunohistochemical staining of the ovary sections was evaluated by the primary researcher (I.P.) and an experienced pathologist specialized in gynecology (V.S.). In each breast tumor tissue core sample, the percentage of breast tumor cells and the percentage of positively stained membranes among the malignant cells were scored by two independent observers (I.P. and R.V.). In the event of a major discrepancy, the observers reached consensus regarding a final score. The tumor cell membranes were considered positive if they showed immunoreactivity of any intensity. A weighted scoring method based on the size of the tumor area in each tumor core was used to calculate the percentage of positive membrane-stained tumor cells in each sample.

Statistical analysis was performed using SPSS version 20.0 (IBM, Armonk, NY). Inter-observer agreement was calculated using the Pearson correlation coefficient. The suitability threshold for the putative NIRF probe targets was set at 80, 90, or 100 % of tumor cells expressing the antigens.

A histological analysis showed that all ovaries contained follicles. The cortex of each ovary was negative for immunohistochemical staining by all markers tested. In contrast, all markers (except CEA and uPAR) were detected at the plasma membrane of epithelial cells in inclusion cysts (Fig. 1a, b). These inclusion cysts were present in five of the ten ovaries. In addition, E-cadherin was expressed at moderate levels in the granulosa cells of primary follicles (Fig. 2).

The median age at the time of diagnosis was 32 years (range 21–35 years) for the 24 patients included in the TMA analysis. Twenty-three patients were diagnosed with ductal breast cancer, and the remaining patient was diagnosed with lobular breast carcinoma. The characteristics of these 24 patients and their tumors are summarized in Table 1.

Here, we identified several proteins that could potentially serve as a suitable target for detecting breast cancer cells within ovarian autografts. One clear application for these markers is the use of NIRF imaging, a technique that can differentiate malignant tissues from non-malignant tissues without reducing the tissue’s viability [10, 33‐35]. Designing a NIRF probe directed against E-cadherin shows particular promise, as E-cadherin was expressed by the majority (94 %) of invasive breast tumor cells and was absent on the surface of normal ovarian cells. However, a combination of tumor-selective probes will likely be needed to detect all tumor cells. Based on our results, a combination of probes against E-cadherin, EMA and Her2/neu seems suitable.

Metastatic spread requires the local invasion of the surrounding host tissue by cells that originated from the primary tumor, followed by intravasation in blood and lymphatic vessels, ultimately leading to the dissemination of tumor cells [36]. E-cadherin and EpCAM mediate cell–cell adhesion, and the downregulation or loss-of-function of these proteins enables cells to escape from solid tumors [19]. E-cadherin and/or EpCAM are not necessarily expressed in all tumor cells; therefore, metastatic tumor cells might not be detected in some tissues. Furthermore, the majority of metastatic lobular breast cancer cells, which lack E-cadherin expression, will not be detected using a specific anti-E-cadherin probe, even though lobular breast cancer cells are more likely to invade ovarian tissue compared to cells derived from ductal carcinomas [37, 38].

As mentioned above, we considered tumor cell membranes positive if they showed immunoreactivity of any intensity. As a result, some tumor cells might be more positive than others for the investigated markers. Yet, for NIRF imaging, the staining intensity is less important as long as a significant tumor-to-background-ratio can be achieved.

None of the premenopausal ovaries in our cohort had positive staining in either the stromal cells or the ovarian surface epithelium. However, all markers (except CEA and uPAR) were expressed on epithelial cells in inclusion cysts. Consequently, conjugating antibodies against these markers to a NIR fluorophore will illuminate invasive breast cancer cells in the ovary, as well as inclusion cysts. Because inclusion cysts might differ from metastatic breast cancer cells with respect to their fluorescent configuration, it might be possible to distinguish between these structures. The same strategy might be used to distinguish granulosa cells in primary follicles from metastatic breast cancer cells that express E-cadherin. In addition, full-field optical coherence tomography (FF-OCT), a non-invasive imaging technique that mimics conventional histopathology, might be very useful. In the field of dermatological oncology, FF-OCT has already been proven capable of visualizing sebaceous glands and adipose tissue surrounding hair follicles as well as small malignant skin tumors [39]. On high magnification, fine architectural details on the subcellular level can be recognized. Therefore, it is expected that FF-OCT will also be able to distinguish inclusion cysts from metastatic tumor cells in ovarian tissue.

One strength of our study is that we examined the expression of tumor markers in tumor tissues obtained from young breast cancer patients who met the criteria for ovarian tissue cryopreservation. Moreover, only biologically active ovaries were analyzed, giving the results clinical relevance, as ovarian tissue is generally cryopreserved before ovarian failure has occurred.

On the other hand, this study has some limitations that merit discussion. First, a relatively small sample size was examined. However, normal ovarian tissue from premenopausal patients is not readily available. We excluded ovaries that were removed due to the presence of a BRCA gene mutation, as such samples could contain primary ovarian tumor cells [40], that express the markers investigated in this study. To be certain, we also excluded ovaries from breast cancer patients with unknown mutation status. At the LUMC, normal ovarian tissue from premenopausal breast cancer patients was exclusively available from BRCA mutation carriers or women with unknown mutation status. As a consequence, ovaries from breast cancer patients were not included in this study. Malignant cells may also be present in ovaries that were removed due to endometrial carcinoma, uterine sarcoma, cervical squamous cell carcinoma, or contralateral ovarian carcinoma; however, the risk of false-positive results was relatively low in our cohort, as all primary tumors were diagnosed at an early stage and the lymph nodes in these patients were clear. Second, the expression of the markers was evaluated on primary invasive breast tumors, since a substantial cohort consisting of ovarian tissues containing breast cancer metastases is scarce. Finally, we examined the expression of tumor markers in relatively small tumor cores using a TMA approach. Yet, the TMA technique is considered an accurate method for examining protein expression in breast cancer tissues [41].

Recently, the intraoperative use of tumor-targeted fluorescent imaging yielded a high tumor identification rate and enabled the surgeon to detect metastases that could not be detected by visual observation [42]. For our purpose, tumor-specific NIRF probes could be administered intravenously prior to oophorectomy, after which the removed ovary is dissected into cortical ovarian strips. Detailed fluorescent images could then be obtained using multiphoton microscopy, which provides an inherent submicron spatial resolution that allows revelation of subcellular details with reduced phototoxicity and photobleaching [43, 44]. Because the NIRF signal lies beyond the red end of the visible spectrum, the signal has enhanced tissue penetration, enabling the identification of fluorescently labeled tumor cells that are located deep within the tissue. Moreover, the low autofluorescence of the tissue at the emission wavelength of the probe provides a high tumor-to-background ratio [45]. Because of these features and the fact that cortical ovarian fragments can be imaged from both the upper and lower side, thereby increasing the imaging depth even further, NIRF imaging is a promising technique for detecting tumor cells in cortical ovarian strips up to 2 mm in thickness.

In conclusion, we report the identification of tumor markers that may serve as a target for detecting breast cancer cells in ovarian tissue using robust imaging techniques such as NIRF imaging. Based on our analysis, E-cadherin is likely the most suitable target for designing a tumor-specific probe. Further research will focus on examining the expression of these markers on breast cancer metastases in ovaries, refining methods to distinguish breast cancer cells from ovarian inclusion cysts, and examining the clinical feasibility of applying NIRF imaging to the field of fertility preservation.