Mesothelin Expression in Human Tumors: A Tissue Microarray Study on 12,679 Tumors

Mesothelin (MSLN) represents an attractive molecule for targeted cancer therapies. To identify tumors that might benefit from such therapies, tissue microarrays including 15,050 tumors from 122 different tumor types and 76 healthy organs were analyzed for MSLN expression by immunohistochemistry. Sixty-six (54%) tumor types showed at least occasional weak staining, including 50 (41%) tumor types with at least one strongly positive sample. Highest prevalence of MSLN positivity had ovarian carcinomas (serous 97%, clear cell 83%, endometrioid 77%, mucinous 71%, carcinosarcoma 65%), pancreatic adenocarcinoma (ductal 75%, ampullary 81%), endometrial carcinomas (clear cell 71%, serous 57%, carcinosarcoma 50%, endometrioid 45%), malignant mesothelioma (69%), and adenocarcinoma of the lung (55%). MSLN was rare in cancers of the breast (7% of 1138), kidney (7% of 807), thyroid gland (1% of 638), soft tissues (0.3% of 931), and prostate (0 of 481). High expression was linked to advanced pathological tumor (pT) stage (p < 0.0001) and metastasis (p < 0.0001) in 1619 colorectal adenocarcinomas, but unrelated to parameters of malignancy in 1072 breast-, 386 ovarian-, 174 lung-, 757 kidney-, 171 endometrial-, 373 gastric-, and 925 bladder carcinomas. In summary, numerous important cancer types with high-level MSLN expression might benefit from future anti-MSLN therapies, but MSLN’s prognostic relevance appears to be limited.

To better understand the prevalence and significance of MSLN expression in cancer, a comprehensive study analyzing a large number of neoplastic and non-neoplastic tissues under highly standardized conditions is needed. Therefore, MSLN expression was analyzed in more than 15,000 tumor tissue samples from 122 different tumor types and subtypes, as well as 76 non-neoplastic tissue categories, by immunohistochemistry (IHC) in a tissue microarray (TMA) format in this study.

Tissue Microarrays (TMAs)
The normal tissue TMA was composed of 8 samples from 8 different donors for each of 76 different normal tissue types (608 samples on one slide). The cancer TMAs contained a total of 15,050 primary tumors from 122 tumor types and subtypes. Detailed histopathological data on grade, pathological tumor stage (pT), or pathological lymph node status (pN) were available from 7625 tumors (cancers of the colon, bladder, ovarian, endometrium, lung, stomach, breast, and kidney tumors). Clinical follow-up data were available from 1178 breast cancer, 865 kidney cancer, and 254 bladder cancer patients with a median follow-up time of 49/39/14 months (range 1-88/1-250/ . No data on previous therapies were available. The composition of both normal and cancer TMAs is described in detail in the results section. All samples were from the archives of the Institutes of Pathology, University Hospital of Hamburg, Germany, the Institute of Pathology, Clinical Center Osnabrueck, Germany, and Department of Pathology, Academic Hospital Fuerth, Germany. Tissues were fixed in 4% buffered formalin and then embedded in paraffin. The TMA manufacturing process was described earlier in detail [48,49]. In brief, one tissue spot (diameter: 0.6 mm) was transmitted from a cancer containing donor block (≥70% cancer cells) in an empty recipient paraffin block. The use of archived remnants of diagnostic tissues for manufacturing of TMAs and their analysis for research purposes, as well as patient data analysis, has been approved by local laws (HmbKHG, §12) and by the local ethics committee (Ethics commission Hamburg, WF-049/09, 25 January 2010). All work has been carried out in compliance with the Helsinki Declaration.

Immunohistochemistry (IHC)
Freshly prepared TMA sections were immunostained on one day in one experiment. Slides were deparaffinized with xylol, rehydrated through a graded alcohol series, and exposed to heat-induced antigen retrieval for 5 min in an autoclave at 121 • C in pH 9 DakoTarget Retrieval Solution TM (Agilent, Santa Clara, CA, USA; #S2367). Endogenous peroxidase activity was blocked with Dako Peroxidase Blocking Solution TM (Agilent, CA, USA; #52023) for 10 min. Primary antibody specific against MSLN protein (mouse monoclonal, MSVA-235, MS Validated Antibodies, Hamburg, Germany) was applied at 37 • C for 60 min at a dilution of 1:150. Bound antibody was then visualized using the EnVision Kit TM (Agilent, CA, USA; #K5007) according to the manufacturer's directions. The sections were counterstained with hemalaun. A trained pathologist scored all tissue spots and marked tissue spots with questionable findings for revision by a second pathologist. For normal tissues, the staining intensity of positive cells was semi-quantitively recorded (+, ++, +++). For tumor tissues, the percentage of MSLN positive tumor cells was estimated and the staining intensity was semi-quantitatively recorded (0, 1+, 2+, 3+). For statistical analyses, the staining results were categorized into four groups as follows: Negative: no staining at all, weak staining: staining intensity of 1+ in ≤70% or staining intensity of 2+ in ≤30% of tumor cells, moderate staining: staining intensity of 1+ in >70%, staining intensity of 2+ in > 30% but in ≤70% or staining intensity of 3+ in ≤30% of tumor cells, strong staining: staining intensity of 2+ in >70% or staining intensity of 3+ in >30% of tumor cells.

Statistics
Statistical calculations were performed with JMP 14 software (SAS Institute Inc., Cary, NC, USA). Contingency tables and the chi 2 -test were performed to search for associations between MSLN and tumor phenotype. Survival curves were calculated according to Kaplan-Meier. The Log-Rank test was applied to detect significant differences between groups. A p-value of ≤0.05 was defined as significant.

Technical Issues
A total of 12,679 (84.2%) of 15,050 tumor samples were interpretable in the TMA analysis. The remaining 2371 (15.8%) samples were not analyzable due to the lack of unequivocal tumor cells or loss of the tissue spot during the technical procedures. On the normal tissue TMA, a sufficient number of samples was always interpretable per tissue to determine MSLN expression.

MSLN Immunostaining in Normal Tissues
In normal tissues, the strongest MSLN expression was observed in the squamous epithelium of tonsil crypts ( Figure 1A), where a fraction of cells (intermediate to superficial cell layers) showed strong (+++) MSLN staining. Strong MSLN immunostaining was also seen in some cells and cell groups of the rectal mucosa (+++), the anal transitional epithelium (+++) ( Figure 1B) where staining was often particularly prominent in superficial mucinous cells, amnion cells (+++) ( Figure 1C) and some chorion cells (++) of the mature placenta, and some elements of corpuscles of Hassall's of the thymus (++). A somewhat weaker MSLN staining was seen in scattered cells and groups of cells of endocervical mucosa (++) and endometrium (++), epithelial cells of fallopian tube (apical cell border and cilia; ++) ( Figure 1D), some intermediate (neck) cells of the stomach antrum (+), some scattered glands in sublingual ( Figure 1E) and Brunner glands, few cells of respiratory epithelium (goblet cells;++), some cells or groups of cells in bronchial glands (++), seminal vesicle (+; not in all samples), and in the cytoplasm of few cells of the adenohypophysis (+). MSLN immunostaining was absent in endothelium and media of the aorta, heart muscle, striated muscle, tongue muscle, myometrium of the uterus, muscular wall of the appendix, esophagus, ileum, kidney pelvis, and urinary bladder, corpus spongiosum of the penis, ovarian stroma, fat, skin, hair follicles and sebaceous glands of the skin, non-keratinizing squamous epithelium from the lip, oral cavity, ectocervix, and the esophagus, urothelium of the kidney pelvis and urinary bladder, spleen, antrum and corpus of the stomach, gallbladder epithelium, liver, kidney ( Figure 1F), epididymis, testis, lung, decidua cells, cerebellum, cerebrum, salivary glands, prostate, breast, adrenal gland, and lymphatic tissue.
bronchial glands (++), seminal vesicle (+; not in all samples), and in the cytoplasm of few cells of the adenohypophysis (+). MSLN immunostaining was absent in endothelium and media of the aorta, heart muscle, striated muscle, tongue muscle, myometrium of the uterus, muscular wall of the appendix, esophagus, ileum, kidney pelvis, and urinary bladder, corpus spongiosum of the penis, ovarian stroma, fat, skin, hair follicles and sebaceous glands of the skin, non-keratinizing squamous epithelium from the lip, oral cavity, ectocervix, and the esophagus, urothelium of the kidney pelvis and urinary bladder, spleen, antrum and corpus of the stomach, gallbladder epithelium, liver, kidney ( Figure 1F), epididymis, testis, lung, decidua cells, cerebellum, cerebrum, salivary glands, prostate, breast, adrenal gland, and lymphatic tissue.

MSLN Immunostaining in Neoplastic Tissues
A significant MSLN immunostaining was observed in 2413 (19.0%) of 12,679 analyzable tumors, including 8.0% with weak, 3.4% with moderate, and 7.6% with strong staining intensity. The staining pattern was variable. Most positive tumors showed a predominantly apical membranous MSLN staining, often accompanied by a less intense cytoplasmic coloration. Other tumors showed a pure membranous staining or a predominantly cytoplasmic positivity. Representative images are shown in Figure 2. At least an occasional weak MSLN positivity was detected in 66 of 122 (54.1%) different tumor types and tumor subtypes and 50 (41.0%) tumor types and subtypes had at least one tumor sample exhibiting strong positivity. The highest frequencies of MSLN positivity were seen in different subtypes of ovarian (65% to 97%) and endometrium (45% to 71%) carcinomas, pancreatic adenocarcinoma (75% and 81%), malignant mesothelioma

MSLN Immunostaining in Neoplastic Tissues
A significant MSLN immunostaining was observed in 2413 (19.0%) of 12,679 analyzable tumors, including 8.0% with weak, 3.4% with moderate, and 7.6% with strong staining intensity. The staining pattern was variable. Most positive tumors showed a predominantly apical membranous MSLN staining, often accompanied by a less intense cytoplasmic coloration. Other tumors showed a pure membranous staining or a predominantly cytoplasmic positivity. Representative images are shown in Figure 2. At least an occasional weak MSLN positivity was detected in 66 of 122 (54.1%) different tumor types and tumor subtypes and 50 (41.0%) tumor types and subtypes had at least one tumor sample exhibiting strong positivity. The highest frequencies of MSLN positivity were seen in different subtypes of ovarian (65% to 97%) and endometrium (45% to 71%) carcinomas, pancreatic adenocarcinoma (75% and 81%), malignant mesothelioma (69%), and adenocarcinoma of the lung (55%). Rare or absent MSLN positivity was observed in different subtypes of breast tumors (0% in 50 phyllodes, 0.9% in 294 lobular, 5.3% in 27 tubular, 6.6% in 1391 invasive no special type, and 10.9% in 58 mucinous subtypes), renal cell (0% in 177 oncocytomas, 1.6% in 131 chromophobe, 6.6% in 858 clear cell, and 8.7% in 255 papillary subtypes), and thyroid carcinomas (0% to 1.4%), as well as soft tissue tumors (0.3%), and adenocarcinomas of the prostate (0%). A detailed description of the immunostaining results is given in Table 1 and Figure 3.
(69%), and adenocarcinoma of the lung (55%). Rare or absent MSLN positivity was observed in different subtypes of breast tumors (0% in 50 phyllodes, 0.9% in 294 lobular, 5.3% in 27 tubular, 6.6% in 1391 invasive no special type, and 10.9% in 58 mucinous subtypes), renal cell (0% in 177 oncocytomas, 1.6% in 131 chromophobe, 6.6% in 858 clear cell, and 8.7% in 255 papillary subtypes), and thyroid carcinomas (0% to 1.4%), as well as soft tissue tumors (0.3%), and adenocarcinomas of the prostate (0%). A detailed description of the immunostaining results is given in Table 1 and Figure 3. Pattern of MSLN immunostaining in cancer. MSLN immunostaining was found to be strong and predominantly membranous (A) or predominantly situated at the apical/luminal cell border (B) in two serous high-grade ovarian cancers, variable but predominantly apical in a cholangiocellular carcinoma (C), strong and predominantly cytoplasmic in a diffuse type adenocarcinoma of the stomach (D), strong and predominantly membranous in a malignant (epitheloid) mesothelioma (E), and weak to moderate, cytoplasmic and membranous in a colorectal adenocarcinoma (F).

MSLN Immunostaining, Tumor Phenotype, and Prognosis
A comparison of MSLN immunostaining with pT, pN, and histological grade in 1619 colorectal adenocarcinomas, 1072 invasive breast carcinomas of no special type, 386 serous carcinomas of the ovary, 174 lung adenocarcinomas, 757 clear cell renal cell carcinomas, and 171 endometrioid endometrial, in 373 gastric, and in 925 bladder carcinomas revealed only a statistically significant association between MSLN immunostaining and pT stage, as well as pN status in colorectal cancer (p < 0.0001 each, Table 2). MSLN immunostaining was unrelated to overall survival in 227 bladder carcinomas (p = 0.3302; only pT ≥ 2), 593 invasive breast carcinomas of no special types (p = 0.0976), and 502 clear cell renal cell carcinomas (p = 0.3144, Figure 4). A significant association was found between positive MSLN immunostaining and RAS mutations in colorectal carcinomas (p = 0.0010), and triple negative invasive breast carcinomas of no special type (p < 0.0001, Table 2). There was also a strong tendency towards higher MSLN expression in HPV positive than in HPV negative squamous cell carcinomas (p = 0.0098, Table 3).

MSLN Immunostaining, Tumor Phenotype, and Prognosis
A comparison of MSLN immunostaining with pT, pN, and histological grade in 1619 colorectal adenocarcinomas, 1072 invasive breast carcinomas of no special type, 386 serous carcinomas of the ovary, 174 lung adenocarcinomas, 757 clear cell renal cell carcinomas, and 171 endometrioid endometrial, in 373 gastric, and in 925 bladder carcinomas revealed only a statistically significant association between MSLN immunostaining and pT stage, as well as pN status in colorectal cancer (p < 0.0001 each, Table 2). MSLN immunostaining was unrelated to overall survival in 227 bladder carcinomas (p = 0.3302; only pT ≥ 2), 593 invasive breast carcinomas of no special types (p = 0.0976), and 502 clear cell renal cell carcinomas (p = 0.3144, Figure 4). A significant association was found between positive MSLN immunostaining and RAS mutations in colorectal carcinomas (p = 0.0010), and triple negative invasive breast carcinomas of no special type (p < 0.0001, Table 2). There was also a strong tendency towards higher MSLN expression in HPV positive than in HPV negative squamous cell carcinomas (p = 0.0098, Table 3).

Discussion
The result of our normal tissue analysis for MSLN expression is consistent with a high potential of this protein as a therapeutic target. It is obvious that the risk for potential side effects of targeted therapies is connected to the target protein's site and level of expression in normal tissues. MSLN expression was mostly seen in organs that are not vital for adult or old-aged people, such as tonsil, thymus, gallbladder, seminal vesicle, fallopian tube, uterus, and placenta. In the stomach, duodenum, rectum, and the anal canal, the fraction of MSLN expressing cells was so low that critical side effects might not occur, even if these rare cells were disturbed by drug effects. Our normal tissue results are largely consistent with RNA expression data derived from the FANTOM5 project [50,51] and the Genotype-Tissue Expression (GTEx) project [52] which are all summarized in the protein atlas (https://www.proteinatlas.org/ENSG00000102854-MSLN/tissue, accessed on 6 January 2021). These RNA data would suggest the lung as the organ, which might be most endangered by side effects of anti-MSLN therapies. A high MSLN RNA expression in the lung could be explained by strong MSLN immunostaining in goblet cells of respiratory epithelium and occasional positivity of bronchial glands. As the alveolar system did not show any MSLN expression, we would expect that possible lung side effects derived from therapeutic anti-MSLN antibodies would rather affect the bronchial system than the alveolar space.
The successful analysis of more than 12,000 tumors identified various cancer types that might be particularly well suited for anti-MSLN drugs. Although the usage of tissue microarrays has disadvantages connected to the small size of the analyzed tissue spots (0.6 mm in diameter), including the risk of missing "relevant" tumor components in heterogenous tumors, or the impossibility to capture different tumor compartments (such as invasion front and tumor center) within one tissue spot, it allows for an unprecedented degree of experimental standardization across the analyzed tumors. Although different staining conditions may alter the absolute numbers of positive cancers, the relative ranking order of positive tumor types will remain unchanged. The cancer entities with highest prevalence and also highest levels of MSLN expression included all types of ovarian and endometrium carcinomas, pancreatic adenocarcinoma, malignant mesothelioma, and adenocarcinomas of the lung, stomach, esophagus, and the colorectum. High rates of MSLN expression have already been described for these tumor entities, although the results varied between studies. Previously described high MSLN positivity rates range from 55% to 100% in ovarian cancer [35,43,44,53], 59% to 76% in endometrium cancer [5,8], 57% to 100% in pancreatic adenocarcinoma [43,47,54], 45% to 100% in malignant mesothelioma [33,43,55,56], 38% to 69% of lung adenocarcinomas [37][38][39], 45% to 78% in stomach cancer [5,45,57,58], 29% to 46% in esophageal adenocarcinoma [6,59], and 30% to 68% of colorectal adenocarcinoma [5,8]. These cancers with a positivity rate of 40% or higher in our study appear to be the best candidates for targeted anti-MSLN therapy. Squamous cell carcinomas of various different sites of origin represent the next group of cancers that show MSLN expression at a relatively high frequency (10-40%). Clinical studies using adaptive immunotherapy (CAR-T cells), monoclonal antibodies (Amatuximab/MORAb-009), recombinant immunotoxins (SS1P, LMB-100/RG7787), antibody-drug conjugates (Anetumab, DMOT4039A), or listeria monocytogene induced anti-tumor immune response (CRS-207) therapies have so far focused on malignant mesothelioma, pancreatic adenocarcinoma, and carcinomas of the ovarian, lung, and breast [24][25][26][27][28]35,[60][61][62][63].
Our analysis of 1619 colorectal carcinomas identified significant associations with advanced pT stage and lymph node metastases. This is in line with data from 4 earlier studies all describing associations between high MSLN expression and parameters for cancer aggressiveness or poor patient prognosis [42,99,100,104]. However, the absence of statistical associations between MSLN expression and tumor phenotype and/or prognosis in carcinomas of the bladder, breast, ovary, endometrium, kidney, lung, and stomach argues against a major prognostic role of MSLN expression levels. In principle, this notion is consistent with the existing literature. Only 9 previous studies have earlier described associations between high MSLN expression and poor prognosis and/or unfavorable tumor phenotype in these tumor types [43,45,46,57,74,96,[105][106][107], while there were 7 other studies which could not find associations with clinico-pathological parameters [5,44,47,58,64,65,92]. The concept of MSLN expression not representing a universal parameter of malignancy is also supported by the frequent MSLN expression in various benign tumors, including Brenner tumors of the ovary, as well as Warthin tumors, pleomorphic adenomas, and basal cell adenomas of the salivary glands. It is of note that high MSLN expression was linked to several key molecular features in the cancer types analyzed, such as triple negative breast cancer and RAS (KRAS or NRAS) mutations. This observation fits with known interactions of MSLN with relevant molecular pathways interacting with these molecular features, such as the MAPK/ERK pathway (interacting with MMP-7, MUC16, and ERK) [15,16,108], deregulation of HER2 expression [109], and the PI3K/AkT pathway (interacting with PI3K and MUC6) [16,108,110]. Importantly, all prevalences described in this study are specific to the reagents and the protocol used in our laboratory. It is almost certain that the use of different antibodies, protocols, and interpretation criteria have jointly caused highly diverse literature data on MSLN expression in cancer (summarized in Figure 5). It is well known that different antibodies designed for the same target protein can vary to a large extent in their binding properties and that protocol modifications greatly impact the rate of immunostained cases [111]. Biomedicines 2021, 9, x FOR PEER REVIEW 15 of 21

Conclusions
Our analysis of 12,679 cancers generated a ranking order of cancers according to their frequency of MSLN expression. Top ranked tumor entities, such as ovarian carcinomas, endometrium carcinomas, pancreatic adenocarcinomas, and malignant mesothelioma, thus, may be the best candidates for therapy with drugs targeting MSLN. Despite a link

Conclusions
Our analysis of 12,679 cancers generated a ranking order of cancers according to their frequency of MSLN expression. Top ranked tumor entities, such as ovarian carcinomas, endometrium carcinomas, pancreatic adenocarcinomas, and malignant mesothelioma, thus, may be the best candidates for therapy with drugs targeting MSLN. Despite a link between MSLN positivity and aggressive colon cancer phenotype, the prognostic impact of MSLN expression appears to be low in many other tumor types. This section is not mandatory but can be added to the manuscript if the discussion is unusually long or complex.  Institutional Review Board Statement: The use of archived remnants of diagnostic tissues for manufacturing of TMAs and their analysis for research purposes, as well as patient data analysis, has been approved by local laws (HmbKHG, §12) and by the local ethics committee (Ethics commission Hamburg, WF-049/09). All work has been carried out in compliance with the Helsinki Declaration.
Informed Consent Statement: Informed consent was obtained from all subjects involved in the study (HmbKHG, §12).
Data Availability Statement: All data generated or analyzed during this study are included in this published article.