Comprehensive immunohistochemical study of mesothelin (MSLN) using different monoclonal antibodies 5B2 and MN-1 in 1562 tumors with evaluation of its prognostic value in malignant pleural mesothelioma

Mesothelin (MSLN) is a glycophosphatidylinositol (GPI)-linked cell surface protein highly expressed in several types of malignant tumors sometimes in association with increased tumor aggressiveness and poor clinical outcome. In the present study, 1562 tumors were immunohistochemically analyzed for mesothelin expression using two different types of mouse monoclonal antibodies (5B2 and MN-1) to determine the clinical usefulness of mesothelin immunohistochemistry as well as to pinpoint potential targets for future anti-mesothelin therapy. Also, characterization of selected mesothelin-positive tumors was performed by immunohistochemistry and oncogene sequencing. Among the tumors analyzed, the highest frequencies of mesothelin-positivity were detected in ovarian serous carcinoma (90% in 5B2 and 94% in MN-1). Both antibodies showed frequent positivity in pancreatic adenocarcinoma (71% using 5B2 and 87% using MN-1) and malignant pleural mesothelioma (75% using 5B2 and 78% using MN-1). In malignant mesothelioma, overall survival was significantly longer in the cohort of patients with diffuse membranous expression of mesothelin (P < 0.001). Both antibodies showed positive staining in thymic carcinoma (77% in 5B2 and 59% in MN-1), however, no expression was detected in thymoma. No correlation was detected between mesothelin expression and mismatch repair system deficient phenotype or gene mutation (BRAF and RAS) status in gastrointestinal adenocarcinomas. Mesothelin immunohistochemistry may assist the differential diagnosis of thymoma vs. thymic carcinoma as well as prognostication of mesothelioma patients. Our results demonstrate that patients with solid tumors expressing mesothelin could be targeted by anti-mesothelin therapies.


INTRODUCTION
Mesothelin (MSLN) is a glycophosphatidylinositol (GPI)-linked cell surface protein normally expressed in mesothelial cells that line the pleura, peritoneum, and pericardium. The MSLN gene encodes a precursor protein of 71 kDa that is processed to a 31 kDa shed protein called MPF (megakaryocyte potentiating factor) and a 40 kDa membrane bound protein, mesothelin [1]. The biologic function of mesothelin is not well known, however, no detectable abnormalities were reported in growth and reproduction in a mesothelin deficient mouse model [2].
Mesothelin is reported to be highly expressed in several types of malignant tumors, such as malignant mesothelioma, ovarian cancer, pancreatic adenocarcinoma, and lung adenocarcinoma. In some cases, mesothelin expression has been associated with increased tumor aggressiveness and poor clinical outcome, however, its impact on the clinical outcome of malignant pleural mesothelioma patients has not been extensively evaluated [3][4][5][6][7][8][9][10]. In ovarian cancer, it has been shown that mesothelin binds to ovarian cancer antigen MUC16 (CA-125) and may contribute to dissemination into the abdominal cavity [11][12][13]. It has also been shown that mesothelin plays a pivotal role in tumor cell proliferation, invasion, and chemotherapy resistance through the activation of oncogenic signaling [14][15][16]. Although the mechanism(s) and/or tumor biological significances were unclear, high mesothelin expression was associated with KRAS gene mutation in lung adenocarcinoma [8,9].
The aim of this study was to evaluate differential reactivity of different types of mouse monoclonal antibodies against mesothelin as well as MPF/precursor mesothelin for immunohistochemistry. It was also aimed to determine the clinical usefulness of mesothelin immunohistochemistry as well as to highlight tumor types for future mesothelintargeting therapy. Additional immunohistochemical and oncogene mutation analyses were performed to characterize the mesothelin-positive tumors.

Comparison of two mesothelin and three MPF antibodies in 218 selected tumor tissues
Immunohistochemical staining using two mesothelin (5B2, and MN-1) and three MPF (MPF25, MPF44, and MPF49) antibodies were performed in 218 selected tumor tissues including ovarian serous carcinoma, pancreatic ductal carcinoma, thymic tumors, and malignant mesothelioma. Among them, MPF49 antibody did not generate sufficient specific staining signals in a selection of mesothelin-positive tumors and was not studied further (data not shown). The results of immunohistochemistry using the two anti-mesothelin and two anti-MPF antibodies have been summarized in Table 1. The mesothelin antibodies (5B2 and MN-1) showed higher rates of positivity than MPF antibodies (MPF 25 and MPF 44) in all of the tumors analyzed. In malignant mesothelioma, mesothelin antibodies (5B2 and MN-1) showed membrane positivity, whereas MPF44 showed predominantly cytoplasmic staining. (Figure 1A and 1B) From this preliminary experiment, 2 mesothelin antibodies, 5B2 and MN-1, were chosen for the study of normal tissues and a larger cohort of tumors because of their higher rates of positivity on initial screening.

Immunohistochemistry using 5B2 and MN-1 antibodies in normal tissues and 1562 tumors
In normal tissues, both 5B2 and MN-1 antibodies showed limited mesothelin expression in the epithelium of fallopian tubes and seminal vesicles (Figure 2A and 2B). Also, Hassall's corpuscles of the thymus and a subset of squamous epithelial cells of the tonsils showed mesothelin expression ( Figure 2C and 2D). Table 2 summarizes the results of immuno histochemistry for mesothelin using mouse monoclonal antibodies 5B2 and MN-1 in different types of tumors. Among them, high frequencies of mesothelin-positivity were detected in ovarian serous carcinoma (90%, median value of positive cells 90% in 5B2 and 94%, median value of positive cells 100% in MN-1) ( Figure 3A and 3B).
In other tumors with adenocarcinoma-like differentiation, both clones showed variable positivity on the cell membrane and cytoplasm of the neoplastic cells. MN-1 showed higher positive staining rate than 5B2 on the cell membrane in invasive ductal carcinoma of the pancreas (87% vs. 71%, Figure 4A and 4B), adenocarcinomas of the colorectum (61% vs. 48%) and stomach (49% vs. 38%).
Adenocarcinoma of the prostate and lobular carcinoma of the breast were usually negative but occasionally showed cytoplasmic mesothelin-positivity (1-5% of cases).
Forty-one percent of thymic carcinomas were positive for mesothelin on the cell membrane with MN-1 antibody. 5B2 showed dominant cytoplasmic staining in these tumors ( Figure 5A). On the other hand, benign thymomas showed no positivity with either antibody ( Figure 5B).
Urothelial carcinomas and clear cell renal cell carcinomas showed rare positivity for mesothelin with both antibodies (4-13%, Figure 6A and 6B). In urothelial carcinoma, positive signals were especially detected in components with squamous cell differentiation ( Figure 6A).

Survival analysis of mesothelioma patients
Characteristics of the 66 mesothelioma patients analyzed for survival have been summarized in Table 3. Patients were followed up for up to 120 months. Survival was significantly longer for patients with mesothelioma expressing mesothelin diffusely (100% of positive cells) on the cell membrane using 5B2 (36.0 months median vs. 10.0 months with heterogeneous or no mesothelin expression; P = 0.002, Figure 7A) or MN-1 (33.0 months median vs. 9.0 months with heterogeneous or no mesothelin expression; P < 0.001, Figure 7B) antibodies. Multivariable Cox hazards regression analysis revealed diffuse membranous mesothelin expression in mesothelioma tumor cells to be a favorable prognostic factor (HR, 0.36; 95% CI, 0.21-0.64; P < 0.001) ( Table 4).

DICUSSION
Mesothelin (MSLN) is a cell surface protein highly expressed in several types of malignant tumors, such as malignant mesothelioma, ovarian cancer, pancreatic adenocarcinoma, and lung adenocarcinoma, sometimes in association with increased tumor aggressiveness and poor clinical outcome. Several antibody-based therapeutic agents, vaccines, and T cell therapies against mesothelin, in some case, have shown favorable results [17]. Serum mesothelin levels have been found to correlate with mesothelioma responsiveness to anti-mesothelin therapies [12,13,18,19]. However, the clinical usefulness of mesothelin immunohistochemistry, including for the prognostication of malignant pleural mesothelioma patients, has not been fully examined [3][4][5][6][7][8][9][10].
In the present study, mesothelin expression was evaluated using five different mouse monoclonal antibodies against mesothelin (clone 5B2 and MN-1) and MPF (megakaryocyte potentiating factor)/precursor mesothelin (MPF25, MPF44 and MPF49), in order to determine the optimal antibodies for mesothelin detection in common carcinomas. Since the anti-mesothelin (5B2 and MN-1) antibodies offered a superior signal quality in a preliminary study, they were used to analyze 1562 tumors to compare immunoreactivities of selected tumors as well as to test their clinical utility for pathological diagnosis and prognostication of mesothelioma patients. Furthermore, immunohistochemistry (MLH1, MSH2, MSH6, and PMS2) and gene mutation analyses (BRAF, KRAS, and NRAS) were performed to characterize the mesothelin-positive tumors.
Almost all ovarian serous carcinomas showed high-levels of mesothelin expression with both 5B2 and MN-1 anti-mesothelin antibodies. In the present study, these antibodies also showed similar high positivity (up to 70-80% of cases) in malignant mesothelioma and uterine endometrioid adenocarcinoma. Patients with these mesothelin-positive tumors are potential candidates for the  Oncotarget 26750 www.impactjournals.com/oncotarget mesothelin-targeting therapeutics such as a recombinant immunotoxin (SS1P), a high-affinity chimeric monoclonal antibody (MORAb-009), an anti-mesothelin antibody drug conjugate (BAY 94-9343), or adoptive T-cell immunotherapy using mesothelin-specific chimeric antigen receptors (CAR) in autologous T-lymphocytes [15].
Up to 80% of thymic carcinomas showed homogeneous and strong cell membrane and/or cytoplasmic staining for mesothelin. On the other hand, thymomas showed no positivity for mesothelin immunostaining with either antibody. Therefore, mesothelin immunohistochemistry with either antibody could have a role in differentiating thymic carcinomas from thymomas.
In adenocarcinomas arising in the pancreas, stomach, and colorectum, MN-1 antibodies showed a more prominent mesothelin positivity rate than 5B2. On the other hand, squamous cell carcinoma of the lung and urothelial carcinoma of the urinary tract, 5B2 showed slightly higher positivity.
In ELISA and immunoblot assay, MN-1 antibody showed much higher affinity than 5B2 antibody. Furthermore, 5B2 antibody showed lower or no affinity to native mesothelin expressed in human cells [22]. However, appropriate antigen retrieval and dilution of the antibodies led to favorable immunohistochemical staining results with 5B2 antibody [22]. It is likely that the different staining patterns and rates of positivity obtained with 5B2 and MN-1 antibodies are due to the differential expression of the various mesothelin epitopes in these tumors. Furthermore, these results suggest that appropriate type of anti-mesothelin antibody should be used when evaluating mesothelin expression in different tumor types. A thorough  Oncotarget 26751 www.impactjournals.com/oncotarget analysis of mesothelin expression in tumors being targeted by anti-mesothelin agents will be important to understand the correlation between mesothelin expression and treatment response.
Among the MPF antibodies, MPF44 has been shown to react with the full-length mesothelin precursor protein, a 71 kDa protein without cleavage, and the shed 31 kDa MPF, in immunoblot analyses [23]. These results indicate that mesothelin precursor protein might also be detected in immunohistochemistry by MPF25 and MPF 44 antibodies, however, soluble factor MPF may be harder to detect in formalin-fixed paraffin-embedded (FFPE) tissue sections. Our results suggest that malignant mesotheliomas express processed mesothelin on the cell membrane because 5B2 and MN-1 typically showed membrane positivity, whereas MPF44 showed predominantly cytoplasmic staining. We also show that MPF49 does not react with MPF protein in immunoblot analysis under SDS-denatured conditions probably indicating reactivity with a conformationsensitive epitope [23]. This might also explain why MPF49 showed no specific staining in FFPE sections in this study.
In the present study, different from other types of tumors such as adenocarcinoma of the lung and stomach, mesothelin expression itself did not show any impact on the survival of the mesothelioma patients (data not shown) [8][9][10]. However, patients with malignant mesothelioma expressing mesothelin diffusely (100% of positive cells on the cell membrane) had increased overall survival compared to patients whose tumors had heterogeneous or no mesothelin expression. Since we did not have detailed clinical information about other patient characteristics such as clinical stage of their disease or performance status it is difficult to say whether diffuse mesothelin expression by itself is an independent favorable prognostic factor. This will need to be validated in larger studies with defined patient characteristics.
Several types of lung carcinomas have been reported to express mesothelin [6]. In case of lung adenocarcinoma high-level of mesothelin expression is associated with aggressiveness, poor prognosis and KRAS gene mutation status [8,9]. Based on these observations, in the present study, gene mutation analyses were performed in colorectal adenocarcinomas since they commonly harbor KRAS mutations. Among 75 arbitrarily selected colorectal adenocarcinomas, 20 and 31 tumors carried BRAF and RAS mutations, respectively. However, no significant correlation between mesothelin-expression and gene mutation status was detected. These results indicate different relationship between mesothelin expression and gene mutations in colorectal and lung adenocarcinomas.
In normal tissues, mesothelin expression was detected in the mesothelial cells, epithelium of the fallopian tubes and seminal vesicles. Also, focal expression was seen in Hassall's corpuscles of the thymus and a subset of squamous epithelial cells of the tonsils. The biologic function of mesothelin is not well known including in these tissues.
In conclusion, many tumors such as mesothelioma and ovarian serous carcinoma showed equal immunoreactivity with both 5B2 and MN-1 antibodies, however, differences in mesothelin expression were seen in other tumor types using 5B2 or MN-1 antibodies. This has to be taken into account when selecting patients for anti-mesothelin immunotherapies. Mesothelin immunohistochemistry may also assist in the differential diagnosis of thymoma versus thymic carcinoma. Furthermore, mesothelin immunohistochemistry could be useful for the prognostication of malignant pleural mesothelioma patients. For neoplastic conditions, 1562 of extensively characterized tumors were subjected to immunohistochemical analyses. Tumor samples derived from surgical specimens were assembled into multitumor blocks containing up to 60 rectangular tissue samples as previously described [24]. The size of tumor sample was estimated to exceed the size of a single 0.6mm 2 core by a factor of 10-15.

Antibodies and cells
The information of the antibodies has been summarized in Table 5. It was noted that 5B2 was generated by immunizing mice with a recombinant prokaryotic fusion protein corresponding to 100 amino acids which is also present in the N terminal Region I of mesothelin by the company (Novocastra/Leica, Bannockburn, IL). On the other hand, MN-1 was generated in mesothelin-deficient mice by immunization with plasmid cDNA corresponding to the mesothelin extracellular domain followed by a single boost with a recombinant human mesothelin-Fc fusion protein [22]. The validity of 5B2 and MN-1 antibodies was confirmed by immunoblot analyses using malignant pleural mesothelioma and pancreatic cancer cells.

Immunohistochemistry
Immunohistochemistry was performed using the Leica Bond-Max automation and Leica Refine detection kit (Leica Biosystems, Bannockburn, IL). The conditions for immunostaining have been summarized in Table 5. Signals were visualized by DAB. Mesothelin immunoreactivity (luminal/membranous or cytoplasmic) was evaluated with a detection cut-off of 5%. To compare the mesothelin positivity in small rectangular tissue samples on the tissue array and the corresponding larger whole-slide donor samples, 31 cases were selected for comparative immunohistochemical analysis. The results showing a high concordance are illustrated in Supplementary Figure 2. Immunohistochemistry of MLH1, MSH2, MSH6, and PMS2 were performed as previously reported [26,27].

Sequencing analysis
In colorectal tumors, arbitrary selected 75 samples were subjected to sequencing analysis in BRAF, KRAS, and NRAS genes. Mutation analyses were done by PCR and Sanger sequencing as previously reported [26]. Primer sequences, PCR conditions and size of amplicons are provided in Supplementary Table 4.

Statistical analysis
All statistical analyses were performed with EZR version 1.32. software [28]. Chi-square or Fisher's exact test were performed to investigate the statistical correlation. To analyze the impact of the mesothelin expression on overall survival of mesothelioma patients, Kaplan-Meier survival estimates with log-rank test was performed. Cox proportional hazards regression analysis was used to analyze the association of survival and other factors. The initial model included age (< 65 vs. ≧ 65 years old), sex (male vs. female), tumor histology (epithelial vs. biphasic vs. sarcomatoid) and data from immunohistochemistry (diffuse mesothelin expression: 100% of positive cells on the cell membrane vs. negative or partial mesothelin expression in any location) for mesothelin with MN-1 and 5B2 antibody. A backward elimination with a threshold of P = 0.05 was used to select variables in the final model. Cases with missing information were eliminated from the statistical analysis of that parameter. A backward elimination with a threshold of P = 0.05 was used to select variables in the final model. Provided from collaborators, MO and IP Antigen retrieval was performed with heat activation in Bond low (H1) or high (H2) pH buffer.