Morphological and phenotypical features of ovarian metastases in breast cancer patients

Cryopreservation of ovarian tissue is the only option to preserve fertility and restore ovarian activity in prepubescent girls and women who cannot postpone the start of adjuvant chemotherapy [ 1]. Although autotransplantation of frozen-thawed cortical ovarian tissue has resulted in more than 86 live births worldwide [ 2], this method has not yet been endorsed by the American Society for Reproductive Medicine (ASRM) [ 3]. One of the reasons that the ASRM committee has put forward is that the safety of the procedure has not been substantiated in patients with cancer. Cortical ovarian tissue may contain malignant cells that could lead to reseeding of cancer upon autotransplantation. This risk of reintroducing malignant cells cannot be eliminated, since the current tumor detection methods (e.g. PCR, immunohistochemistry) jeopardize the ovarian tissue’s viability [ 4]. These methods can therefore only be used to examine cortical ovarian strips that are not transplanted. Hence, the presence of tumor cells in the actual ovarian autografts remains questionable.

Whether the current approach for tumor detection is accurate depends on the distribution of metastatic tumor cells in the ovarian tissue [ 5, 6]. If tumor cells are diffusely dispersed throughout the ovary, examination of one or two cortical ovarian strips might be sufficient. By contrast, if tumor cells are confined to a specific area in the ovarian cortex, this approach is inadequate. Then, cortical ovarian strips that are examined may turn out to be devoid of tumor cells whereas ovarian fragments that harbor metastases may be transplanted, possibly resulting in cancer relapse.

The implementation of a detection method that allows examination of the cortical ovarian strips that will be transplanted, will significantly reduce the risk of transferring malignant cells. Near-infrared fluorescence (NIRF) imaging might be an appropriate approach, as this technique discriminates malignant cells from non-malignant tissue in real time while leaving the tissues viable [ 7]. A NIRF probe consists of a fluorophore that emits light in the near-infrared spectrum (λ = 700–900 nm) and an antibody or peptide with high affinity for a protein expressed specifically at the cell surface of tumor cells [ 8, 9].

In order to use tumor-specific imaging to exclude malignant cells in cortical ovarian autografts, tumor markers should be identified that are present at the cell surface of ovarian metastases. Since a substantial proportion of patients who undergo ovarian tissue cryopreservation is diagnosed with breast cancer [ 10– 12], we tested a panel of cell-surface markers known to be expressed by breast cancer cells, including E-cadherin [ 13], epithelial membrane antigen (EMA, also known as MUC1) [ 14, 15], human epidermal growth factor receptor type 2 (Her2/neu) [ 16, 17], carcinoembryonic antigen (CEA) [ 18], αvβ6 integrin [ 19] and epithelial cell adhesion molecule (EpCAM) [ 20– 22]. The markers cytokeratin CAM 5.2, gross cystic disease fluid protein-15 (GCDFP15), Wilms’ tumor antigen-1 (WT1), mammaglobin 1, and cytokeratin 7 (CK-7), which were used by Sánchez-Serrano et al. [ 23] and Rosendahl et al. [ 6], were excluded, as they are not expressed at the cell surface and therefore not suitable as a target for tumor-specific imaging.

In this study, we assessed the distribution of breast tumor cells in ovarian tissues from patients with ovarian metastases and determined which cell-surface proteins are suitable as a target for tumor-specific imaging of ovarian metastases derived from invasive breast cancer. Because it is crucial to select a target prior to the administration of the NIRF probe, we also examined whether invasive breast cancer tissue can be used to predict the most suitable target for the detection of ovarian metastases in a particular patient.

One of the purposes of the present study was to examine the histological features of ovarian metastases in breast cancer patients to evaluate the current tumor detection approach [ 4] in ovarian tissues considered for autotransplantation. We found that 71% of ovarian metastases consisted of a solitary metastasis or multiple distinct nodules separated by uninvolved ovarian tissue. These findings suggest that tumor cells might have been missed if the current tumor detection approach would have been used. The patients included in this study however, underwent oophorectomy after a median time interval of 42 months. In patients undergoing ovarian tissue cryopreservation an oophorectomy is performed soon after cancer diagnosis. In these patients, disseminated tumor cells may not yet have outgrown into overt metastases and may appear as micrometastases in the ovarian tissues [ 23, 26, 27]. The chance that tumor cells will then be overlooked is presumably greater. We therefore recommend examination of the actual ovarian autografts on the presence of malignant cells prior to autotransplantation.

In our study, merely 63 out of 2648 patients (2.4%) who were diagnosed with primary invasive breast cancer at age < 41 years had ovarian metastases. For the determination of suitable targets for NIRF imaging, it might have been relevant to focus on malignancies with a higher risk of ovarian contamination in the actual patient population, for instance leukemia [ 28– 30]. Nonetheless, breast cancer can be perfectly used as a starting point to investigate whether NIRF imaging is feasible for the detection of ovarian metastases.

Considering the expression of Her2/neu in primary invasive breast cancers and ovarian metastases a high percentage of Her2/neu positive tumor cells (67–88%) was found, as we considered tumor cell membranes positive if they showed immunoreactivity of any intensity. This is in contrast to the diagnostic setting, where Her2/neu overexpression is determined because of its potential prognostic value [ 17, 31]. We applied a lower cut-off point, because for NIRF imaging the staining intensity is less important as long as a significant tumor-to-background-ratio can be achieved. In the NIR spectrum, non-specific fluorescence background signal is substantially decreased compared to wavelengths lower than NIR [ 8]. Hence, since ovarian stromal cells do not immunohistochemically express Her2/neu [ 32], Her2/neu-targeting NIRF probes will detect metastatic breast cancer cells within ovarian tissues if these cells show immunohistochemical reactivity.

In individual patients, no correlation was found between the expression of the investigated markers in breast tumors and their corresponding metastases in the ovary. This might be due to the fact that breast cancer is known as a heterogeneous disease [ 17] or be in line with the hypothesis that disseminated tumor cells autonomously evolve from the primary tumor [ 33]. For the clinical application of these markers there should not be an obstacle, since a combination of three markers enhances the ability to detect breast tumor cells in ductal and lobular ovarian metastases. Furthermore, only the histological subtype of the invasive breast tumor needs to be known to determine which combination of markers is pertinent for the detection of the corresponding ovarian metastases, making the selection of suitable NIRF probes simple and straightforward. For the non-invasive detection of metastases in the actual ovarian autografts by tumor-specific imaging, NIRF probes could be administered intravenously, after which the removed ovary is dissected into cortical ovarian strips. Subcellular detailed fluorescent images of tumor cells within ovarian autografts could then be obtained by multiphoton microscopy [ 34]. Beside breast cancer cells, inclusion cysts will likely also be illuminated by NIRF imaging, as we previously showed that in normal ovaries, all markers (except CEA) were expressed on epithelial cells in inclusion cysts [ 32]. Nevertheless, we additionally demonstrated that full-field optical coherence tomography (FF-OCT), which creates histology-like images without the need for tissue manipulation, can be perfectly used to differentiate between inclusion cysts and metastases in the ovary [ 35]. On a tomographic FF-OCT image, an inclusion cyst is characterized by a thin dark outer layer and lack of interior structure, whereas micrometastatic lesions from primary invasive ductal carcinomas present as ‘web-like’ structures in which tumor cells appear light gray. Metastatic lesions derived from primary invasive lobular carcinomas often show an Indian file pattern, defined as infiltrating single rows of cells [ 36]. Since ovarian inclusion cysts are separately identifiable within the ovarian parenchyma [ 32], a distinction between these structures can also be made. In addition, FF-OCT and NIRF imaging might be combined to enhance their sensitivity and specificity rates, as both methods are noninvasive. The a priori probability that other benign epithelial ovarian abnormalities will be detected by our panel of cell-surface markers is low, since ovaries that present as an adnexal mass on preoperative ultrasonography are generally not used for ovarian tissue cryopreservation. In case primary ovarian cancer cells are present, these cells will be detected as E-cadherin [ 37], EMA [ 38], and EpCAM [ 39] are virtually always expressed in ovarian cancer, and approximately 33% of primary ovarian carcinomas show Her2/neu amplification [ 40].

In conclusion, we showed that in young breast cancer patients with ovarian metastases, metastatic breast tumor cells may be confined to a specific area in the ovarian cortex. A non-invasive tumor detection technique by which cortical ovarian fragments that are transplanted can be examined, is recommended to minimize the risk of reintroducing metastatic tumor cells by ovarian tissue autotransplantation in breast cancer patients. NIRF imaging is a promising technique to discriminate malignant from benign tissues while leaving the examined tissues vital. Our research opens a new avenue for the development of tumor-specific NIRF probes that can be used for non-invasive detection of breast cancer metastases in ovarian tissues prior to autotransplantation.