Pleural mesothelioma is a rare cancer with an often-challenging diagnosis because of its potential to be a great mimicker of many other tumors. Among them, primary lung and breast cancers are the 2 main causes of pleural metastasis. The development and application of targeted therapeutic agents have made it even more important to achieve an accurate diagnosis. In this setting, international guidelines have recommended the use of 2 positive and 2 negative immunohistochemical biomarkers.
To define the most highly specific and sensitive minimum set of antibodies for routine practice to use for the separation of epithelioid malignant mesothelioma from lung and breast metastasis and to determine the most relevant expression cutoff.
To provide information at different levels of expression of 16 mesothelial and epithelial biomarkers, we performed a systematic review of articles published between 1979 and 2017, and we compared those data to results from the Mesothelioma Telepathology Network (MESOPATH) of the standardized panel used in routine practice database since 1998.
Our results indicate that the following panel of markers—calretinin (poly)/thyroid transcription factor 1 (TTF-1; clone 8G7G3/1) and calretinin (poly)/estrogen receptor-α (ER-α; clone EP1)—should be recommended; ultimately, based on the MESOPATH database, we highlight their relevance which are the most sensitive and specific panel useful to the differential diagnosis at 10% cutoff.
Highlighted by their relevance in the large cohort reported, we recommend 2 useful panels to the differential diagnosis at 10% cutoff.
Mesothelioma is a rare cancer derived from the transformation of mesothelial cells that line the serosal surfaces. It represents 0.2% of all cancers and has a dismal prognosis with limited therapeutic options.1 Mesothelioma is related to asbestos exposure in up to 90% of cases in men2 and is located to the pleura in 90%, the peritoneum in nearly 10%, and rarely in the tunica vaginalis testis and pericardium.3 Diagnosis is based on biopsy tissue samples according to the 2015 World Health Organization classification. The diagnosis is often challenging because of mesothelioma tumor heterogeneity and to the potential of this tumor to be a great mimicker of many other tumors in the setting of small biopsies. Among them, primary lung and breast carcinomas are the 2 main causes of pleural metastasis that mimic mesothelioma, accounting for more than 150 000 cases a year in the United States, with 11% (n = 16 500) from lung and 14% (n = 21 000) from breast carcinomas.4 The development and application of drugs that target specific molecules expressed in breast and lung carcinomas have made it even more important to achieve an accurate diagnosis so that appropriate therapy can be administered.
During the past 5 years, several guidelines and recommendations for a definitive diagnosis of malignant mesothelioma have been issued from European and international mesothelioma panels of pathologists,5 such as the following scientific and academic societies: the European Respiratory Society task force, the European Society for Medical Oncology task force, the British Thoracic Society guidelines, the Australian Society of Pathology, and more recently from the International Collaboration on Cancer Reporting. A Consensus Statement from the International Mesothelioma Interest Group and the International Mesothelioma Panel have been recently published by Husain et al.6 It recommends the use of immunohistochemical tests comprising 2 positive and 2 negative biomarkers before making a definitive diagnosis of mesothelioma. The pathologist faces many different antibodies from different sources, different clones, different techniques, and different preanalytic processes that in the end can confuse him or her with the variable results obtained from previous published reports. As an example of the potential for misdiagnosis the selection of ER-β (positive in epithelioid mesothelioma)7 is preferred to that of ER-α (100% negative). To our knowledge there is no publication that has analyzed systematically the sensitivity (Se) and specificity (Sp) of biomarkers used in the differential diagnoses, according to tumor cell positivity and at various cutoff points. The main objectives of this study were to provide information on the Se and Sp at different levels of expression for the 18 mesothelial and epithelial biomarkers that have been used in the diagnosis of epithelioid malignant mesothelioma8 and to compare those data to the standardized panel used in routine practice for the diagnosis of malignant mesothelioma by the French Mesothelioma Telepathology Network (MESOPATH) national center (Léon Bérard Cancer Center, Lyon, France).
MATERIALS AND METHODS
Systematic Review of the Literature
This review was conducted with references from PubMed (National Center for Biotechnology Information, Bethesda, Maryland), the Cochrane Handbook for Systematic Reviews of Interventions (http:// handbook.cochrane.org; Cochrane, London, England), the European Respiratory Society guidelines and task force (Lausanne, Switzerland), which served as a guide for the study. We used PubMed search (http://www.ncbi.nlm.nih.gov/pubmed), updated in January 2017, with the following key words as inclusion criteria for collecting articles: mesothelioma, pulmonary, breast, adenocarcinoma, and immunohistochemistry.
Then, we reviewed the literature referenced in PubMed and Medline (Ovid, New York, New York) databases between 1979 and 2017 (1874 references). The authors selected studies and extracted data comparing the results of immunohistochemical staining performed from formalin-fixed, paraffin-embedded (FFPE) tissues samples of malignant pleural mesothelioma and pulmonary and breast carcinoma metastases.
Finally, we noted the type of the study, the number of items selected, and the major contents of each article. There were observational series, case series, and diagnostic-accuracy series. Abstracts, case reports, meta-analyses, reviews, and guideline articles were excluded from the analysis. Articles analyzing fewer than 10 lung or breast tumors, cytologic material, or frozen tissue sections were also excluded from the review. The analysis included the immunohistochemical staining results obtained in the epithelioid component of epithelioid or biphasic pleural mesotheliomas (EMMs). Only the relative numbers of EMMs, lung adenocarcinomas (LAs), or breast adenocarcinomas (BAs) diagnosed on FFPE tissues reported in 88 articles were included in this study. Two panels of immunohistochemical stains were selected, the first one composed of mesothelial biomarkers: epithelial membrane antigen (EMA) clone E29, keratin AE1/AE3, thrombomodulin clone 1009, mesothelin clone 5B2, calretinin clone Z11-E3, Wilms tumor (WT1) clone 6F-H2, keratin 5/6 clone D5/16 B4, and D2-40; and the second panel comprised epithelial carcinoma biomarkers: carcinoembryonic antigen (CEA), BER-EP4, B72.3, CD15 clone Leu-M1, MOC-31, thyroid transcription factor-1 (TTF-1) clone 8G7G3/1, BG-8 clone F3, and estrogen receptor α (ER-α) clone 6F11.
The biomarker Se and Sp of these 2 panels were calculated based on the sum of EMMs, LAs, and BAs resulting from selected articles published in PubMed (Table 1). The Se of a positive biomarker was defined as the proportion of positive staining among the cases affected by the disease. The Sp of a positive biomarker was defined as the proportion of negative staining among the cases not affected by the disease. An Se or Sp greater than 80% was considered high. Because of high variation among articles about the percentage of staining considered positive, the Se and Sp were calculated at 5 cutoffs: 1%, 10%, 25%, 50%, and 75%. The Clopper and Pearson method was used to calculate 95% CI.9
MESOPATH Cohort Analysis
For the comparative study, we retrieved 6571 EMMs and 602 pleural metastases from carcinomas of 149 LAs and 41 BAs with consecutive FFPE tissue biopsy samples certified by the MESOPATH group according to the standardized procedure of certification between 1998 and 2017.10 The standardized procedure of certification from MESOPATH routinely used 2 different panels of antibodies for the diagnosis of epithelioid mesothelioma, depending on whether the tumor was observed in a man or in a woman (Table 2): keratin 5/6, EMA, calretinin, mesothelin, AE1/AE3, p53, and WT1; and the most-encountered carcinoma biomarkers: monoclonal CEA, BER-EP4, TTF-1, and ER-α. p16 and BRCA1-associated protein (BAP1) were added to the panel in 2009 and 2015, respectively.
Biomarkers Used Routinely on the Standardized Procedure of Certification at the MESOPATH Center
Immunohistochemical studies were performed on 4-μm-thick, FFPE tissue sections. Sections were stained with the BenchMark ULTRA immunohistochemistry slide staining system (Ventana Medical Systems, Oro Valley, Arizona). The staining protocol used antigen retrieval in Ventana Cell Conditioning 1 solution (pH 8.4). A Ventana UltraView Universal diaminobenzidine detection kit was used, followed by Ventana hematoxylin as a nuclear counterstain.
To evaluate the Sp of the immunoreaction, known positive and negative tissues were used as controls. The immunostaining was graded according to the percentage of reactive cells (0, 1%–9%, 10%–25%, 25%–49%, 50%–74%, and 75%–100%).
RESULTS
This systematic review was conducted through a literature review (Tables 3 and 4) combined with a pathologic review of cases enrolled in the MESOPATH cohort.
Sensitivity (Se) and Specificity (Sp) of Mesothelial Biomarkers of Epithelioid Malignant Mesothelioma (EMM) From the Literature for Each Cutoff
Sensitivity (Se) and Specificity (Sp) of Epithelial Biomarkers of Lung Adenocarcinoma (LA) and Breast Adenocarcinoma (BA) From the Literature for Each Cutoff
This study incorporated the selected data from 88 articles published between 1979 and 2017 and comprised 2933 malignant mesotheliomas, 3123 lung adenocarcinomas, and 383 breast adenocarcinomas. Among the lung adenocarcinomas studied, 205 (7%) were acinar, 182 (6%) were nonmucinous lepidic, 100 (3%) were papillary, 57 (2%) were biphasic, 62 (2%) were solid with or without mucin production, 1 (<1%) was a clear cell subtype, and the remaining 2516 (80%) were not specified. The 2933 malignant mesotheliomas with an epithelioid component comprised 2540 epithelioid (87%) and 276 biphasic (9%) subtypes. Because neither of these 2 histologic subtypes were specified in some articles, 115 mesotheliomas (4%) were either epithelioid or biphasic.
Additionally, 73 articles were retrieved from PubMed using the key words mesothelioma, breast adenocarcinoma, and immunohistochemistry, of which, only 5 articles (7%) were selected during the period of time from 1979 to 2017.11–15 These articles included 280 breast adenocarcinomas.
Mesothelioma Biomarkers Studied
Epithelial membrane antigen is a mucinlike, transmembrane glycoprotein initially described in 1984.16 Seventeen articles were analyzed with clone E29 membranous staining.17–33
Pancytokeratin AE1/AE3 is a cocktail of 2 different clones of anti-cytokeratin monoclonal antibodies studied since 1987.34 AE1 detects the high–molecular-weight cytokeratins 10, 14, 15, and 16 and the low–molecular-weight cytokeratin 19. Clone AE3 detects the high–molecular-weight cytokeratins 1, 2, 3, 4, 5, and 6 and the low–molecular-weight cytokeratins 7 and 8. Six articles evaluating its expression were included in this study.27,29,33–36
Thrombomodulin (CD141) is a 75-kDa endothelial, transmembrane glycoprotein expressed on the surface of endothelial cells reported in 1992.37 This study included 18 referenced articles in PubMed; the authors used clone 1009.11,26–28,31,33,38–49
Mesothelin is a 40-kDa cell glycoprotein that presents on the surface of normal mesothelial cells, described on frozen-tissue specimens in 199250 and on FFPE tissue sections in 2003.12,44 Nine articles* using clone 5B2 were studied.
Calretinin is a 29-kDa calcium-binding protein, expressed in the mesothelial cytoplasm and nuclei with characteristic fried-egg staining, described in 1996.40 Twelve articles† were evaluated.
Wilms tumor 1, a zinc-finger protein encoded by the Wilms tumor gene located on chromosome band 11p13, was first reported by Amin et al57 and Kumar-Singh et al.58 Seven articles15,30,33,49,55,58,59 using clone 6F-H2 were evaluated.
Epithelial Biomarkers Studied
Carcinoembryonic antigen is an oncofetal glycoprotein. It is expressed by fetal epithelial cells and in small amounts by normal adult epithelial cells and benign tumors. This biomarker described in 197969 is available commercially as either a monoclonal or a polyclonal antibody. It is represented by more than 44 epitopes. Twenty articles§ were studied for the polyclonal antibody A115 and 14 articles for the monoclonal antibodies clone II-7,22,32,52 clone CEJ065,19,22,38,76,83,84 and clone 12.140.10.27–29,31,34,73
Clone B72.3 represents a monoclonal antibody directed against tumor-associated glycoprotein-72 evaluated in 1986.81 This antigen is expressed in a limited range of benign tissues but in a wide range of adenocarcinomas, including lung adenocarcinomas. Nineteen articles** were analyzed.
CD15 is a monocyte/granulocyte-related biomarker, expressed in the cytoplasm and on the surface membranes of carcinoma cells, as reported in 1986.72 The analysis of this antibody, clone Leu–M1, included 22 articles.11,19–24,30,73,81–89
BG-8 is a monoclonal antibody that recognizes the blood group LewisY. The relevance of this biomarker in the differential diagnosis of mesothelioma was first described in 1985.90 Five articles11,30,38,90,91 were included in the analysis of clone F3.
Reported in 1990,84,92 the clone BER-EP4 is present on the surface membrane and in the cytoplasm of all epithelial cells except the superficial layers of squamous epithelium. Seventeen articles†† were analyzed.
MOC-31 is the clone of a monoclonal antibody which recognizes a 40-kDa transmembrane protein reported in 1995.95 Ten articles‡‡ were analyzed.
TTF-1 is a tissue-specific transcription factor, a 38-kDa member of the NK-X2 family of DNA-binding proteins. It was initially described in 1989 as a highly specific biomarker for adenocarcinoma arising from the distal lung parenchyma (>80% Sp). The analysis of clone 8G7G3/1 included 16 articles.§§
Mouse anti-human ER-α antibody, clone 6F11, recognizes the human ER-α chain, also known as the estradiol receptor or nuclear receptor subfamily 3 group A member 1. Estrogen receptor α is an approximately 65-kDa steroid hormone receptor containing an N-terminal (AF-1) ligand-independent transactivation domain, a DNA binding domain and a C-terminal ligand-binding domain which overlaps with an AF-2 domain. The ER-α binds to DNA as a homodimer105,106 and can also form heterodimers with estrogen receptor β. The detection of ER and progesterone receptors using immunohistochemical staining on FFPE tissue became the most common method for the determination of the ER/progesterone status of breast tumors.107 Approximately 75% to 80% of breast tumors have ER and/or progesterone, and the presence of these receptors helps to determine the origin of the patient's cancer, the prognosis, and the effectiveness of hormonal therapy.108 The analysis of this biomarker included only 2 articles.13,15
BAP1 and p16
The BAP1 gene is located on the chromosome band 3p21, which encodes BAP1 and belongs to the C-terminal hydrolase family. It is a deubiquitinase involved in the removal of ubiquitin from H3a (mono-ubiquitin), coordinating cell proliferation. The mutation of the nuclear deubiquitinase is a loss of nuclear expression.109,110 Clone C-4 from Santa Cruz Biotechnology (Dallas, Texas) was selected by the MESOPATH Center. Nuclear staining loss with an internal positive control was considered positive, whereas cytoplasmic staining was not.
The CDKN2a (p16) gene is located on chromosome band 9p21 encoding the p16 protein tumor suppressor gene. p16 is located on the short arm of chromosome 9 (9p21) in the 21.3 region and it has an important role in cell cycle regulation.111 The detection of CDKN2a (p16) homozygous deletion by fluorescence in situ hybridization analysis is routinely performed on FFPE and is observed in more than 40% of epithelioid mesotheliomas, compared with sarcomatoid mesotheliomas (MESOPATH, unpublished data). Few studies have analyzed the usefulness of p16 immunohistochemistry to detect the presence of the CDKN2a (p16) homozygous deletion, and they have showed some discrepancies. The exact role of p16 expression detected by immunohistochemistry has not yet been elucidated. MESOPATH used the clone E6-H4.
Sensitivity and Specificity of Mesothelial and Epithelial Biomarkers
Calretinin was the most sensitive and specific mesothelial biomarker. At a cutoff of 1% of tumor cells, thrombomodulin showed 74% Se and 89% Sp. Wilms tumor 1 had an Se of 88% and an SP of 94% in lung and breast adenocarcinomas at the 1% cutoff. Keratin 5/6 presented the best Se and Sp, between the 1% and 25% cutoffs. D2-40 showed 84% Se and 87% Sp at a 10% cutoff. BAP1 loss was highly specific in LAs and BAs, but presented 65% Se at a 1% cutoff (Tables 3 and 5).
Carcinoembryonic antigen and B72.3 were highly sensitive and specific at the 1% and 10% cell-staining cutoff. Monoclonal CEA had a better Se than polyclonal CEA did for the same Sp. The Se issue from MESOPATH data was poor (30%) probably because lung adenocarcinoma metastases were mainly sent when monoclonal CEA was negative. CD15 and TTF-1 presented the best Se and Sp at the 2 lower cutoff values. BG-8, BER-EP4, and MOC31 were highly sensitive and specific at the 4 lower cutoffs. To distinguish EMM from breast adenocarcinoma, ER-α had the highest Se (81%) and Sp (100%), based on the largest series issue from the literature and from the MESOPATH data. BG-8 and B72.3 presented the highest Se and Sp but should be investigated on larger series to confirm the initial results (Table 1). BER-EP4 was also interesting for the differential diagnosis of breast adenocarcinoma based on MESOPATH data (Tables 4 and 6).
A panel of 2 antibodies comprising TTF1 and calretinin showed an Se of 98% and an Sp of 82% in EMM versus LA. A panel of ER-α and calretinin showed 98% Se and 71% Sp in EMM versus BA. The potential of the WT1 and ER-α association should be highlighted, despite the few breast metastases (Table 7).
DISCUSSION
Biomarkers that are potentially relevant for the separation of mesothelioma from lung adenocarcinoma metastasis have been evaluated since the 1970s.69 In 2006, King et al8 published an analysis based on a review of 88 published reports that included 15 antibodies, 8 negative and 7 positive mesothelial biomarkers. The Se and Sp of the analysis was performed at a 30% tumor cell cutoff.
Recently, the International Mesothelioma Interest Group6 published recommendations on relevant biomarkers for the separation of EMM from adenocarcinoma. The recommendations were to use in addition to a pancytokeratin, at least 2 positive mesothelial markers such as calretinin, keratin 5/6, WT1 protein, or D2-40; and 2 positive carcinoma markers such as MOC-31, BG-8, CEA, B72.3, BER-EP4, TTF-1, ER-α, and CD15, with a Se and Sp both higher than 80%. The percentage of tumor cell positivity, in the absence of a gold standard, was based on expert opinions and considered a 10% cutoff for membranous, cytoplasmic, or nuclear staining.
In this study, we provide information on the Se and Sp at different levels of expression for the 16 mesothelial and epithelial biomarkers that have been used in the diagnosis of EMM and were able to compare those data to a systematic review of the literature. We also used the results of the standardized panel for the diagnosis of EMM validated by the French national center MESOPATH. To our knowledge, this systematic review is the first showing the importance of considering the cutoffs for cell expression for each antibody to strongly discriminate EMM from lung and breast carcinomas spreading to the pleural cavity.
For the separation of EMM from LA, calretinin was the stronger biomarker for the diagnosis of EMM, with a high Se and Sp no matter what cutoff was used, which is in agreement with the literature. Wilms tumor 1 showed high Sp at all cutoff values but was most sensitive at a 1% cutoff. Cytokeratin 5/6 showed the best Sp and Se at a 1% cutoff. Thrombomodulin and D2-40 showed good Sp at all cutoffs. Even if AE1/AE3 is not discriminant for the separation of EMM from adenocarcinoma, the patterns of expression of AE1/AE3 may help to identify mesothelial cells by highlighting a strong perinuclear distribution of the staining in EMM compared with the haphazard distribution of expression in the cytoplasm of carcinoma metastasis. Moreover, AE1/AE3 is a crucial marker to determine invasion of adipose tissue by the mesothelial cells confirming malignancy in superficial mesothelial proliferations.
Among the epithelial markers, TTF-1 showed the highest Sp (100%) at all studied cutoffs and was more sensitive at the 1% and 10% cutoff. Moreover, monoclonal CEA was more robust than polyclonal CEA for the separation of EMM from LA.
Recently, Yoshimura et al110 reported the potential use of BAP1 in the differential diagnosis of EMM from LA and BA. The BAP1 proved to be retained in 100% of adenocarcinoma but was moderately sensitive in EMM (66%). Calretinin and WT1 were the most sensitive and specific mesothelial markers to distinguish EMM from BA while ER-α had the highest Se and Sp based on the most-robust series. One exception of major importance reported by Ordóñez et al15 was that triple-negative breast carcinoma metastases may strongly express calretinin. The author evaluated 60 EMMs and 80 BAs (40 triple negative and 40 ER-α positive) for expression of calretinin, keratin 5/6, mesothelin, podoplanin, thrombomodulin, and WT1. The carcinoma biomarker claudin-4, the breast-associated biomarkers gross cystic disease fluid protein 15 (GCDFP-15), mammaglobin, and GATA3 were also evaluated. They showed that ER-α positive breast-carcinoma metastases were keratin 5/6 positive in 5%, calretinin in 13%, mesothelin in 3%, and WT1 in 8% of the cases, whereas in cases of triple-negative carcinoma they strongly expressed calretinin with a nuclear and fried-egg appearance in 38% of the cases and positive with keratin 5/6 in 5%, WT1 in 8% and mesothelin in 56% of cases. Claudin 4 was proposed as an alternative for the diagnosis of BA because it was considered 100% negative in EMM.
The publication bias of our analysis is obviously linked to the constant research of new antibodies used in the differential diagnosis of mesothelioma. Additional limitations of our study are related to the quality of data analyzed: assessment of the exact Se and Sp of any particular biomarker related to the number of subjects studied; analytic method adopted and their quality assurance; preanalytic methods such as selection of clones, type of laboratory procedures, and the methods of evaluation of the various staining patterns. The Se and Sp extracted from the literature compared with those of the MESOPATH Reference Center are smoothed by strict criteria of inclusion to avoid bias in this statistical analysis. Our systematic review validates the recommendations proposed by the International Mesothelioma Panel guidelines6 published in 2018 and the results provide clear data on the relevance of selected panel of antibodies in the largest cohort worldwide of patients from whom specimens have been reviewed in a systematic manner for histologic diagnosis of mesothelioma during the past 20 years. Moreover, we show that the implementation of standardized protocols leads to a better accuracy even using a minimal panel of 1 positive and 1 negative marker for the separation of EMM from lung and breast carcinoma metastasis.
We suggest from our experience the use of the same systematically standardized panel on FFPE blocks coming from different laboratories to run calretinin (poly) and TTF-1 (clone 8G7G3/1) for lung adenocarcinoma metastasis and calretinin (poly) plus ER-α (clone EP1) for breast adenocarcinoma metastasis at a cutoff of 10% to optimize cost respecting the level of accuracy of the antibodies.
We wish to thank all the MESOPATH team, especially L. Barjhoux, A. Boj, G. Blaizot, F. Damiola, C. Farge, E. Ferrey, A. de Quillacq, E. Malandain, MC. Petit, C. Py, and R. Sequeiros.