Immunohistological Insight into the Correlation between Neuropilin-1 and Epithelial-Mesenchymal Transition Markers in Epithelial Ovarian Cancer

Human ovarian tissue arrays were used for immunohistochemical staining (Cat# OVC1501; Pantomics, Richmond, CA). These arrays contained tissues fixed in 10% formalin (pH 7.0) for 24 hr and were processed with identical standard operating procedures. Each tissue array had 150 cores collected from 75 individuals, with two cores representing one tumor tissue or biopsy from the same patient. Among the 75 cases there were 2 normal, 3 benign (mucinous, serous cystadenoma and thecoma) and 73 EOC cases, each provided with pathological, grading and staging data (Supplemental Table S1). The regions of each block chosen for inclusion in the TMA were reviewed by at least three pathologists who examined the H&E-stained slides. The main inclusion criteria were to select a region of the tissue that provided sufficient and representative target cells, and this region was marked on the slides. The marked areas were then transferred to corresponding re-marked blocks (formalin-fixed paraffin-embedded blocks) by a pathologist. To ensure the representative nature of a case, duplicated cores per case for most of the TMAs were taken randomly from two marked areas of a block or from two blocks of the same case (normally multiple areas containing representative target cells were marked on a block by a pathologist). After a TMA block was made, a section was cut and stained with H&E. Each core of the stained TMA slide was examined by a pathologist.

Clinical samples, consisting of five normal, four benign (two mucinous and two serous cystadenoma) and four neoplasia (two mucinous and two serous cystadenocarcinoma), were collected from the pathology archives at Sultan Qaboos University Hospital SQUH Oman, and the paraffin blocks were cut into 4-µm-thick sections for immunohistochemistry (described below). Patients routinely provide their full consent for the donation of the tissue prior to any surgical procedure and their information is kept confidential and used for research purposes only.

TMA slides and patient individual ovary sections were deparaffinized in xylene, rehydrated in a series of ethanol (100%, 95% and 75%) followed by tap water, and then subjected to antigen retrieval using 1 mM EDTA (pH 9.0), in 95C water bath for 30–40 min. The activity of endogenous peroxidases was blocked by 2% hydrogen peroxide for 15 min. The slides were washed twice in phosphate-buffered saline (PBS), followed by PBS containing 0.05% Triton X-100, for 5 min each wash. The slides were then incubated with a blocking solution of 5% normal goat serum for 30 min at room temperature, then incubated overnight at 4C with appropriate primary antibody solutions containing one of the following monoclonal antibodies: Neuropilin-1, 1:250; vimentin, 1:250; and N-cadherin, 1:500 (Epitomics Inc.; Burlingame, CA; cat # 2621-1, 2707-1 and 2019-1, respectively). E-cadherin and Slug antibodies were used at 1:200 and were purchased from Cell Signaling Technologies (Beverly, MA) (cat #3195 and #9585, respectively). All other chemicals used were purchased from Sigma Aldrich (Munich, Germany).

After incubation with the primary antibodies, the tissues were washed twice in PBS and incubated with biotinylated secondary goat anti-rabbit antibody (1:100; Vector Laboratories; Switzerland). The signal was enhanced by one-step incubation with Vectastain Elite ABC reagent (Vector Laboratories) for 30 min at room temperature. Colorimetric detection was achieved by incubation with 3,3’-diaminobenzidine (DAB) (Dako; Warrington, PA) for 5 min. This was followed by counterstaining with Mayer’s hematoxylin solution, which was added for approximately 1–2 min. The tissues were dehydrated in a series of ethanol (75%, 95% and 100%) and then immersed into xylene. Finally, the slides were mounted using DPX (Di-n-butyl phthalate in Xylene) (Sigma- Aldrich). Tissues were visualized using an Olympus (BX 40; Olympus Optical Co. Ltd; Tokyo, Japan) light microscopy with digital camera (DP50). The images of cores were captured by 40× using OLYSIA Bio-report and Twin Viewfinder Light (Version 1.0) software (Olympus). A negative primary antibody control slide was stained simultaneously to confirm staining specificity (Supplementary Fig. S1).

Microscopic examination was performed independently in a blinded fashion by two scientists and one pathologist, and the average score was considered for the final analysis. The total 150 cores in the TMA obtained from 75 patient specimens (2 cores per specimen) and another 13 representative patient ovarian tissue sections (5 normal, 4 benign, and 4 malignant) were studied and analyzed for each biomarker. Each specimen was visualized individually under the light microscope (400×) and assigned to an appropriate category. Finally, the scores of each case (2 cores) were combined and given a specific code for statistical analysis. The staining intensity (SI) for the expression of epithelial proteins—E- Cadherin, NRP-1 and Slug—was graded based on DAB staining intensity. The grading system used was: absence of staining (negative), 0; weak staining, 1; moderate staining, 2; strong staining, 3. The score for the distribution of the positively stained cells (percentage of positive cells, PP) was based on the average score observed in ten random fields at 400× (five fields from each TMA duplicated cores). Based on the cell staining proportion, all cases were classified as 0, no positive cells; 1, 1–20% positive cells; 2, 21–50% positive cells; 3, 51–100% positive cells. The IRS value was calculated according to a previously published method, as follows: IRS = SI × PP (Chui et al. 1996); the mean IRS for each marker (NRP-1, Slug, E-cadherin) for each case was calculated as the final value used for statistical analysis.

Because N-cadherin and vimentin expression (intensity and distribution) was homogenous in all TMA studied, these two proteins were categorized based on tissue localization: group 1, mainly stromal; group 2, mainly epithelial; group 3, both epithelial and stromal. In addition to the IRS score, Slug was also grouped for its tissue location (0, negative staining; 1, epithelial staining; 2, stromal staining) and for its cellular localization (0, negative staining; 1, nuclear staining; 2, cytoplasmic staining). Similarly, NRP-1 was expressed in both the cytoplasm and nucleus; this localization was categorized as: 1, nuclear staining; 2, cytoplasmic staining; 3, nuclear and cytoplasmic staining.

Statistical differences between tested variables were determined by Pearson Chi-Square and Fisher’s Exact Test (two-sided). The Spearman’s rank correlation was used to determine whether there was a positive or negative correlation between variables. Post hoc tests were used to analyze the multiple comparisons of the measured IRS values with grade, stage and pathology. To account for multiple testing, we used Bonferroni correction tests, considering α of 0.05; the Bonferroni’s adjustment p value was 0.0071. This p value was used as a cut-off point for significance. SPSS for Windows (v19.0; Chicago, IL) was used to analyze the data.

A total of seven normal ovarian tissues were used, in which two (four cores) were included within the TMA and another five were normal ovaries retrieved from pathology archives. Similarly, of another seven benign cases that were examined, three were included in the TMA slides (one each of mucinous cystadenoma, serous cystadenoma and thecoma) and another four cases were retrieved from the pathology archives (two serous and two mucinous cystadenoma). The normal and benign cases were used as controls to track changes in protein expression and compare them to the neoplasia cases (Fig. 1A). E-cadherin and Slug were weakly expressed in all normal cases (TMA and individual specimens), as represented by the mean IRS values of 0.7143 ± 0.19 (SEM) and 0.1714 ± 0.1749, respectively. NRP-1 was not detected in those specimens (IRS value, 0.0; Fig. 1A, Table 1). NRP-1, Slug and E-cadherin were detected in all benign cases with different intensities. However, the univariate analysis of the variance did not show any statistical differences between the normal and benign cases, and the Bonferroni-adjusted p value for IRS-NRP-1, IRS-E-cadherin and IRS-Slug was p>0.05 (Fig. 1A, 1B). Slug expression was detected only in one of the two normal ovarian tissue cases included in the TMA slide (Fig. 1C) and was not detected in any of the other individual patient cases studied. As shown in Fig. 1A and Table 1, the mean IRS values of the three markers NRP-1, Slug and E-cadherin were significantly higher in the three subtypes of EOC when compared with the normal and benign cases; however, no statistical differences were detected among the three subtypes. The Bonferroni-adjusted p value between the three pathologies was p>0.05 (Fig. 1A, Table 1).

Three pathological subtypes of EOC were studied and statistically analyzed for their expression of the five proteins (E-cadherin, N-cadherin, Vimentin, Slug and NRP-1). The studied pathologies were serous cystadenocarcinoma, endometrioid adenocarcinoma and mucinous cystadenocarcinoma, as these pathologies had higher frequencies of the proteins among the test cases. These three pathological subtypes represented 35.8%, 40.3% and 17.9% of the 67 carcinoma cases, respectively, with the other subtypes accounting for 3% or less of cases evaluated (Supplemental Table S1). These percentages represent the random distribution of the different pathologies on the TMA slides and do not reflect actual statistical frequencies of the different subtypes of ovarian carcinomas.

The EMT inducer Slug was detected in both epithelial and stromal parts of the tissue (Fig. 2C). Furthermore, 52 of the 73 tumors (72.2%) expressed Slug in the epithelium as compared to only 19.4% with expression in the stroma (Supplemental Table S1). Comparing the three studied pathologies, we found a significant difference among them in terms of their tissue localization (epithelial/stromal) (p=0.001) (Fig. 2C; Table 2). Twenty-one of 24 serous cystadenocarcinomas (87.5%) and 23 of 27 endometrioid adenocarcinomas (85.2%) displayed higher percentages of epithelial slug expression when compared with mucinous cystadenocarcinoma tumors, for which only 8 of 12 cases (61.5%) showed slug localization in the epithelium (Table 2).

NRP-1 was expressed exclusively in the epithelial parts of EOC tumors. Therefore, we investigated its expression through IRS and cellular localization. We found that NRP-1 was not differentially expressed among the different pathologies (p>0.05) (Fig. 1A, Table 1, and Fig. 3A). However, its cellular localization (nuclear, cytoplasmic or both) showed a significant difference among the pathological subtypes (p=0.004) (Table 2). The highest percentage of NRP-1 localization (expression) in both sites was observed in 12 of 27 endometrioid adenocarcinomas (44.4%), 5 of 24 serous cystadenocarcinomas (20.8%) and 4 of 13 mucinous cystadenocarcinomas (30.80%) (p=0.004). Cytoplasmic-restricted NRP-1 localization was not detected in any of the 12 mucinous cystadenocarcinomas (0.0%). However, the number of patients with NRP-1 localized in the nucleus was highest in patients with mucinous tumors (6 of 13, 46.20%) as compared with the other two pathologies. Patients with serous tumors had higher nuclear NRP-1 expression as compared with endometrioid (8 of 24 serous, 33.3%; and 5 of 27 endometrioid; 18.5%) (Fig. 3, Table 2).

Descriptive statistics was used to ascertain an association for Vimentin tissue distribution (stromal, epithelial and both stromal and epithelial) in normal, benign and three subtypes of ovarian carcinomas (cross tabulation 3×5). Vimentin was expressed either in the stroma or both stroma and epithelia of both normal and benign tissues (Table 4). Vimentin was exclusively expressed in the epithelial cancer cells of two endometrioid cases only (Fig. 2). The remaining tumors expressed vimentin either only in the stroma (38 cases) or in both the stroma and epithelia (27 cases) (Fig. 2, Supplemental Table S1). Vimentin was exclusively expressed in the stroma of all mucinous tumors present in TMAs and in the additional two retrieved pathologies, with no expression in the epithelial compartment (Fig. 2B). In contrast, Vimentin expression in both stroma and epithelia was detected in 14 of 24 serous carcinomas (61.5%) and 10 of 27 endometrioid carcinomas (37.0%) (Table 3).

Unlike E-cadherin, N-cadherin was expressed in both stroma and epithelium. Its tendency to be expressed in both epithelium and stroma was higher than its expression in only the stroma and/or epithelia (Fig. 4; Supplemental Table S1). (In 50 of the 70 EOC cases, N-cadherin was expressed in both locations.) Multiple comparisons showed that N-cadherin tissue distribution was associated with the different ovarian tissue types, as shown in Table 4.

Although none of the proteins studied showed a significant relationship with tumor grade (Supplemental Table S2), the N-cadherin tissue distribution (stromal, epithelial and both) was significantly related to tumor stage (p>0.001 and Table 4). N-cadherin was observed in 27 of 35 stage I tumors (56.3%), 5 of 11 stage II tumors (50.0%), 5 of 5 stage III tumors (100.0%) and 10 of 16 stage IV tumors (66.7 %) (Table 4). Figure 4 shows the staining patterns observed in four representative tissues from four different stages.

Spearman’s rank correlation test showed that the expression of both E-cadherin and NRP-1 had a significant positive correlation with Slug (ρ=0.861, p<0.001 and ρ=0.602, p<0.001, respectively). (Supplemental Table S3).