Immunohistochemical evaluation of epithelial ovarian carcinomas identifies three different expression patterns of the MX35 antigen, NaPi2b

Ovarian tumor tissues were subjected to analysis by quantitative polymerase chain reaction (QPCR) (n = 73, benign n = 5, borderline n = 11, malignant n = 57) and immunohistochemistry (IHC) (n = 150, malignant n = 108, borderline n = 42) (histology, stage and grade are described in Table 1.). The tumor samples were collected prospectively and consecutively from patients diagnosed from March 2001 to September 2010 with suspected cystic pelvic tumor as part of another study [22]. Ovarian biopsies from 14 women without ovarian cancer were used as control tissue. The local ethical committee at the University of Gothenburg approved the study, and each patient gave her informed, written consent. All case diagnoses were reviewed by a gynecological pathologist using established morphologic criteria according to World health organization (WHO) 2003 [23]. Fresh frozen biopsies from each tumor were divided into two samples one was used for RNA extraction and the other one was paraffin embedded and used in the tissue micro array (TMA). The staining intensity and pattern were evaluated according to histology and the dualistic model presented by Shih et al. [24] Type I included low-grade (G1) serous, low-grade (G1) endometrioid, all clear cell, and mucinous carcinomas. Type II included high-grade (G2–G3) serous, high-grade (G2–G3) endometrioid, undifferentiated carcinoma, and malignant mixed mesodermal tumors [25].

RNA extraction was performed using QIAGEN RNeasy plus Mini Kit (QIAGEN, Germany) according to the manufacturer’s manual, and the RNA concentration was measured with the NanoDrop instrument (ND1000 software, Thermo Fisher Scientific, Wilmington, DE) (Table 1). The RT-PCR High-Capacity cDNA Reverse Transcription kit (Applied Biosystems, Foster City, CA) was used to produce cDNA from the RNA samples. TaqMan Universal PCR Master Mix (Applied Biosystems, Focter City, CA), probe and primers for SLC34A2 (Hs 00197519_m1) as the target gene and GUSB (Hs 99999908_m1) as the reference gene (Life Technologies Corporation, San Diego, CA) were used. A 7000 sequence detection system (Applied Biosystems, Foster City, CA) was used to determine the expression levels by QPCR of the target gene for all samples. Pooled normal ovarian tissue (n = 7) was used as control since it was previously reported to contain low levels of SLC34A2 [2, 5, 21, 26, 27]. The Ct values were used to calculate ΔΔCt and fold change (FC) for each tumor sample.

MX35 is a murine IgG1 monoclonal antibody specifically directed towards a membrane phosphate transporter protein (NaPi2b). The murine MX35 antibody was produced from a hybridoma line and was kindly provided by The Ludwig Institute for Cancer Research (New York, NY, USA). The hybridoma cells were cultured at the Department of Cell and Molecular Biology at the University of Gothenburg (Gothenburg, Sweden) and the antibody was purified from hybridoma supernatant by protein-A chromatography at the Department of Radiation Physics at the University of Gothenburg (Gothenburg, Sweden) [28].

For the TMA, the whole biopsy was sectioned and stained with Hematoxylin (Histolab Products AB, Sweden). Three representative tumor areas were identified under the light microscope (Olympus BX45, Olympus Corporation, Tokyo, Japan), and three cores of 1,0 mm-diameter were punched with a manual tissue microarrayer (Beecher MTA-1, Estigen,Tartu, Estonia) and re-embedded into a predefined position on a new, empty, paraffin block. The TMA block was heated at 45 °C in 1 h, sectioned, 4 μm, and mounted onto slides.

For IHC analysis, the TMA slides were immunostained by UltraVision Quanto Detection System HRP DAB kit (Thermo Fisher Scientific, Wilmington, DE) and incubated overnight with the MX35 antibody at a concentration of 1:1000. All slides were counterstained with hematoxylin and mounted with Pertex (Histolab Products AB, Sweden). All TMAs were scanned by a Leica SCN400 (Leica Microsystems, Milton Keynes, UK). SlidePath Gateway LAN software was used for the evaluation of the NaPi2b distribution.

Staining for NaPi2b were estimated for each tumor and the amount of positive cells were evaluated and given a value; no cells stained = 0, 1/3 = 1, 1/3 > 2/3 = 2, >2/3 = 3. The intensity was estimated for each tumor no staining (negative) = 0, light yellow to yellow (weak) = 1+, light brown (moderate) = 2+, and dark brown (strong) = 3+. For the correlation analysis between QPCR and IHC we used a scoring system were we combined the intensity with the amount of cells stained in the tumor sample described by Tomic et al. [29]. The amount of cells stained (0–3) was used together with the intensity, to calculate a score that describes a combination of both the intensity and the amount of stained cell for each tumor. The product (amount of cells stained multiplied with intensity) ranging from zero to nine were grouped into four final scores as follows: score 0, score 1 (low 1–3), score 2 (intermediate 4–6) score 3 (high >7) [29].

The differences in expression of SLC34A2 between the groups previously described, were evaluated using unpaired two-sample Student’s t-test (IBM® SPSS® Statistics) and were considered significant if P < 0.05. In the analysis all samples were compared to the normal ovarian expression level, which was set to one. Correlation between the gene and protein expression was calculated using Pearson correlation. ANOVA test was used to analyze the variance of NaPi2b expression between the groups. Box plots of tumors grouped into stage, grade and histology were drawn to illustrate the ANOVA analysis results. Two researchers (MC and KL) independently evaluated the IHC staining of the TMAs. In order to evaluate their inter-rater agreement Cohen’s kappa coefficient was calculated.

To compare the expression levels of the SLC34A2 gene, coding for NaPi2b, among the different classifications of the ovarian tumors (OT), we subdivided the samples into groups based on histology, grade and stage (Table 1). The gene expression analysis of SLC34A2 displayed considerable variation in expression levels of this gene within the material, with values ranging from no expression up to a FC > 1600 (mean = 237; median = 126) compared to the expression in normal ovaries. The mucinous OT demonstrated a significantly lower expression of SLC34A2 than both the serous and the clear cell OT (P = 0.007 and P = 0.002 respectively) (Fig. 1a). We found no significant difference between mucinous and endometrioid OT (P = 0.062). Endometrioid OT had significantly lower expression levels than the serous OT (P = 0.038) (Fig. 1a).

For evaluation of the NaPi2b expression in the tumors, six TMAs containing a total of 150 ovarian OT samples were stained with the MX35 antibody and scanned for analysis, 108 malignant and 42 borderline tumors (Table 1). Of the 150 OT 14 samples (9%) were excluded from the analysis (ten malignant and four borderline) due to lack of tumor cells in the TMA. For the remaining 136 samples, quantification of the staining intensity and localization of NaPi2 was performed. For interrater reliability, Cohen’s kappa coefficient was calculated for intensity (κ =0.77) and for pattern (κ =0.89), which established a robust IHC assessment. We found that 127 (93%) out of the 136 samples were positively stained. A Pearson correlation analysis between the gene and protein expression in the tumor tissues was performed and a positive correlation was established (r = 0.302, P < 0.05) (Fig. 1d).

Among the 136 samples there were 41 tumors (30%) stained at the highest level (3+), 48 tumors (36%) as 2+, 38 tumors (28%) as 1+ and nine tumors (7%) did not show any staining at all (Fig. 2a-b). Six of the negatively stained tumors were mucinous borderline tumors. The three malignant tumors with no staining were all type 1, two were mucinous adenocarcinoma, one highly and one moderately differentiated, and one was highly differentiated serous adenocarcinoma. The borderline tumors had a higher number of cases with 3+ staining compared to the malignant tumors, 47% and 29% respectively.

We subdivided the material according to histology, grade and stage. When comparing the histologies we were able to identify differences in the staining intensity between the groups (Fig. 2b). The serous tumors showed highest number of tumors with 3+ staining between the histologies, and among the mucinous tumors only two out of 29 were considered to be 3+ both of them were borderline tumors (Fig. 2c). The majority of the mucinous samples had negative or 1+ staining in both the malignant (75%) and in the borderline tumors (81%). The low expression of the target protein NaPi2b in the mucinous tumors correlates well with the low gene expression of SLC34A2 in this group. The malignant tumors were grouped according to type (type 1: n = 33, type 2: n = 65) [22, 24], among the type 1 tumors 48% (n = 16) had 1+ or no staining, compared to only 25% with 1+ tumors in the more aggressive type 2. Further, there were 75% of the tumors that were considered as 2+ or 3+ in the type 2 tumors (Fig. 3).

Because of the protein function and results from previous studies it was expected to find NaPi2b located to cell membranes [3, 4, 6, 7]. When we evaluated the TMAs we identified differences in the staining pattern between the tumors (Fig. 4). We identified three different patterns for the distribution of NaPi2b (Fig. 4). In pattern A) NaPi2b was primarily located in the cellmembranes of cells close to the surface of the tumor. Even when the tumor consisted of several layers of epithelial cells staining was only detected for epithelial cells that were located in the external layer of the tumor (Fig. 4a). In pattern B) NaPi2b was not limited to the cell membranes in the cells at the surface of the tumor but was found in all of the tumor cell membranes. Finally, in pattern C) there was a mixed staining pattern with NaPi2b localized to both the cellmembranes and the cytoplasm of the same cell (Fig. 4c). Staining of normal ovarian tissue (n = 4 women) showed absence of MX35 in follicles, stroma and ovarian surface epithelium in ¾, and one had typical pattern A staining of ovarian surface epithelium only. Two of 3 women with endometriosis, originating from the uterus, had pattern A staining (data not shown). Of the 136 tumors, borderline and malignant, we were able to subdivide 126 samples into three groups according to the staining pattern (A, B and C). The majority of borderline tumors (n = 29, 90%) had pattern A, and only three borderline tumors had both membrane and cytoplasmic, pattern C and none showed pattern B (Fig. 4d). Conversely, the majority of samples among the malignant tumors displayed pattern C (n = 57, 61%) (Fig. 4d). Both pattern A and B were represented among these tumors, 23% and 16% respectively. With regard to histology, pattern C was most common in the serous (64%), endometrioid (71%) and the undifferentiated OT (67%). Pattern A was the most common in mucinous OT (67%) (Fig. 4e). All three patterns were represented in the seven clear cell OC (A n = 2, B n = 3, C n = 2). None of the mucinous and endometrioid tumors displayed pattern B (Fig. 4e). In type 1 OT pattern A was present in 50% (n = 13) of the cases, contrary to type 2 OT were the most frequent was pattern C which was detected in 71% (n = 46).