**3. Discussion**

Numerous hypotheses regarding the mechanisms involved in OC etiopathogenesis have been proposed over time as attempts to explain the multiple tumor phenotypes, poor prognosis, and chemoresistance. Endometriosis represents a precursor lesion for certain types of epithelial OC, since the identification of the same genetic alterations in both diseases are demonstrated [3,4,31]. Accordingly, the corroboration of specific clinicopathological findings with specific mutations led to the EOC and NEOC categories distinction [44].

The BMI-1 protein, involved in homeotic genes regulation by transcription inhibition [10], represents a survival factor of malignant stem cells [10], and is correlated to hormonal receptor expression, and is considered as a prognosis factor surrogate [44,45].

BMI-1 has been identified in experimental studies of OC (cell lines, clone derivation, and animal experiments) [32–38], both in protein and the protein-coding gene [39], and in human ovarian tumors or ascites fluid samples [10,32,34,36,39,46,47]. Despite these reported results, BMI-1 expression is not fully established in OC. The review of the literature shows that less than 10 studies have addressed BMI-1's involvement in OC, most of them highlighting the molecular action and potential therapeutic value of this protein. A positive correlation between BMI-1 positive expression in human epithelial OC and elevated telomerase activity was demonstrated [46,47]. Another study, based on human specimens and ovarian cancer cells, showed that BMI-1 expression is downregulated by MiR-15a or MiR-16 underexpression, with subsequent significant decreases in cell proliferation and clonal growth [40]. Therefore, BMI-1 seems to be a potential target in OC therapy. Eloquent evidences in this direction are provided in recent papers that have demonstrated the therapeutic activity of PTC-028 as a novel inhibitor of BMI-1 function in OC [37] and the role of MiR-132 in cisplatin resistance and OC metastasis by the targeted regulation of BMI-1 [41]. In terms of the number of human OC samples, the studies on BMI-1 have been generally performed on small groups, with a median number of research sample of 41 (range 5–179) [10,32,34,36,40,46,47]. These samples were collected from tumor tissue [10,32,34,36,39,40,46,47], fresh ascites [34], and frozen ovarian tissues [47].

The reported data target only BMI-1 in epithelial tumor cells, showing a high expression in 80.9% of OC and its relationship with tumor aggressiveness [46]. Moreover, a positive correlation between BMI-1 expression and advanced International Federation of Gynecology and Obstetrics (FIGO) stages, bilaterality, higher tumor grades, and serous morphology [42,47], and a progressive incremental number of BMI-1-positive cases in accordance with the increase of tumor grade and stage were demonstrated, while increased BMI-1 expression was associated with reduced patient survival [39].

This short review of data concerning the correlation between BMI-1 and OC shows that the current knowledge is predominantly based on experimental data as the first level of evidence regarding its role in carcinogenesis, while the results obtained by the investigation of BMI-1 in human tissues is very scarce. Within this general context, our study complements the knowledge on BMI-1 in OC by doing research that translates the evidences level from the experimental area to the clinical domain by reference to the clinicopathological characteristics of OC with different parameters for EOC and NEOC.

Our work has demonstrated high BMI-1 expression levels in the epithelial tumor cells in 66% of OC (26% in EOC and 93% in NEOC). Moreover, our study provides valuable data on BMI-1 profile in OC, bringing to the foreground the relationship of OC with endometriosis, and the differences between the epithelial and stromal expression. This endeavor was possible by consistent differences in the design of the patient's cohort, comprising 47 cases of OC separated in two different tumor groups: EOC and NEOC. Thus, we have demonstrated, for the first time, the possible correlations between epithelial and stromal BMI-1 profiles in EOC and NEOC and several classical clinicopathological parameters.

The segregation into EOC and NEOC has been justified by the findings that certain histological types of EOC, mainly endometrioid and clear cell carcinomas, have different clinical features, such as younger age at diagnosis, unilaterality, identification at an earlier stage, and a better survival rate, compared to the counterpart entities of NEOC [5]. Our

study supports the hypothesis of EOC development within endometriosis, showing mostly an endometrioid (42% in EOC versus 28.57% in NEOC) or clear cell phenotype (21% in EOC versus 18% in NEOC), and, implicitly, the quality of precursor lesion of ovarian endometriosis. Endometriosis and EOC represent two entities with the same target organ (ovary), the same tissue of origin (endometrial-like), and the same pathogenic mechanism which progresses from benign to atypical and malignant phenotypes. Having these in mind, tubal ligature or salpingectomy may be used as preventative maneuvers which may be applied within a screening and early therapy algorithm.

An original finding in our research is the dual staining pattern, nuclear and cytoplasmic/membrane in both study groups, although only a nuclear staining is reported in literature [27,41,42,48]. This immunostaining pattern may indicate a possible relocation of protein during the transition to tumor phenotype. Moreover, it may suggest the involvement of additional factors as a possible reflection of adhesion molecules interrelationship in the context of epithelial mesenchymal transition (EMT) [27,48] or of the involvement of variable ovarian microenvironmental factors in both EOC and NEOC.

Our study confirms the relationship between BMI-1 in epithelial tumor cells and stroma in three instances: (i) in the general OC group (*p* = 0.01), (ii) in the NEOC group (*p* = 0.001), and (iii) in the EOC group (*p* = 0.04). In parallel, the comparative analysis of BMI-1 expression in EOC and NEOC showed a statistically significant higher expression of BMI-1 in the epithelial tumor component than in the stroma (*p* = 0.0002). Our results clearly show EOC's association with BMI-1 low expression in epithelial tumor cells without a dominant expression profile in stromal cells, while NEOC is characterized by high BMI-1 expression in both the epithelial and stromal types of cells. However, stromal BMI-1 expression is reflecting EMT involvement in tumor progression and the interrelationship between the two cellular components, which result in BMI-1 synthesis as a stromal-dependent mechanism. Therefore, if present, stromal BMI-1 could be considered as a valuable marker for poor survival.

To the best of our knowledge, our study provides for the first time evidence for BMI-1 expression in human EOC. Differently from NEOC group findings, a progressive gain of BMI-1 expression in epithelial tumor cells has been noticed in the EOC group along with tumor grade, with statistically significant differences when we compared well and moderately differentiated with poorly differentiated tumors. This finding indicates a relationship between BMI-1 epithelial overexpression and a poorer prognosis in the selected EOC cases. Currently, CA125, expressed in the embryonic development of ovaries and re-expressed in endometriosis and ovarian neoplasms, can be used as a prognostic and predictive biomarker related to patient survival, independent of OC treatment [48].

CA 125 shows significant different values in the two major types of OC, suggesting that they occur as a result of different factors, following specific pathway initiations and progressions [49]. Many studies have shown that the CA125 profiles of HGSC and HGEC are different from other subtypes [50]. We also found a statistically significant correlation between stromal BMI-1 and CA 125 level, suggesting that EOC may be influenced by a microenvironment modulation specific for endometriosis-based ovarian carcinomas, supporting the rapid growth pattern and the unfavorable prognosis in a subcategory of cases. Thus, we may conclude that the interrelationship and reciprocal stimulation between a tumor's epithelial and stromal components occurs latter during the endometriosis-related carcinogenic process, with a subsequent uptake of BMI-1 expression by stromal component, which may be reflected in an increased CA-125 level. The aggressive behavior of these EOC cases has a different significance from that of aggressive type I OC, probably originating from fallopian tube epithelium. It is worth mentioning that BMI-1 was absent in the normal ovaries or ovarian surface in the study groups, while BMI-1 expression has been identified in the normal tubal surface epithelium; this finding comports with the hypothesis of some OCs development from the fallopian tube, providing another support for this pathogenic mechanism.

On the other hand, in the NEOC group, we have shown statistically significant differences between BMI-1 immunopositivity in the tumor's epithelial cells, stromal cells, and histological subtypes. In our opinion, these results may be considered as solid evidence for the association of BMI-1 with high grade OC phenotypes and, consequently, with tumor aggressiveness.

Overall, our study reveals a different BMI-1 profile in the EOC an NEOC groups, thus underlying the differences in their etiopathogeny. We are aware of the limitations of our study due to the small size of the study groups and their heterogeneity in histological types, as the selection criteria have been strictly applied. Despite these limitations, our results open promising perspectives for differentiation of EOC from NEOC that need to be further validated in a larger and homogenous cohort of study. An interesting research item can be directed to the high-grade serous phenotype of OC that may be further subdivided into subcategories according to their affiliation to the EOC or to NEOC groups.
