Context.—The distinction of lung adenocarcinoma from other types of primary lung malignancies is important clinically. Accurate morphologic classification is often hindered because 70% of lung cancers are diagnosed on limited fine-needle aspiration or transbronchial biopsy specimens. Although thyroid transcription factor 1 (TTF-1) has historically been the most specific marker for lung adenocarcinoma, a relatively new marker, napsin A, has recently been shown to be more sensitive and specific than TTF-1.
Objective.—To find the most cost-effective panel to reliably distinguish lung adenocarcinoma from squamous cell carcinoma.
Design.—A total of 291 lung cancers were evaluated morphologically (197 adenocarcinomas [75%]; 66 squamous cell carcinomas [25%]; 28 cases could not be classified into either and were dropped). Immunohistochemistry for napsin A, Cytokeratin 5/6, p63, and TTF-1 was performed on a formalin-fixed tissue microarray obtained from Toyama, Japan. Cases were scored as positive or negative against a negative control.
Results.—Napsin A had 83% sensitivity and 98% specificity and TTF-1 had 60% sensitivity and 98% specificity for adenocarcinoma. Cytokeratin 5/6 had 53% sensitivity and 96% specificity and p63 had 95% sensitivity and 86% specificity for squamous cell carcinoma. A panel of napsin A and p63 has a specificity of 94% and a sensitivity of 96% for distinguishing adenocarcinoma from squamous cell carcinoma.
Conclusions.—The source of the antibody is important in avoiding false-negative results. The most cost-effective tissue-preserving panel for small biopsy specimens in the differential diagnosis of lung adenocarcinoma versus squamous cell carcinoma is a combination of p63 and napsin A.
Lung cancer is the leading cause of cancer-related mortality, both in the United States and worldwide.1–5 Lung cancer is the second most-common malignancy diagnosed in men and women in the United States. The American Cancer Society estimates that lung cancer will cause 29% of cancer deaths in men and 26% in women in 2010.4 Overall, 222 520 new cases of lung cancer and 157 300 deaths are estimated for 2010.4 There are a number of different histologic types of lung cancer, with malignant epithelial tumors being the most common. The 2 most common types are classified as either small cell carcinomas or non–small cell lung carcinomas (NSCLC). The NSCLCs are the most common type and account for approximately 85% of cases. Currently, NSCLC is further classified into adenocarcinoma (ACA) (approximately 40%–50%), squamous cell carcinoma (SCC; 30%), and large cell carcinoma (9%).5 Although it was previously acceptable to classify lung carcinomas as either small cell or NSCLC without further division, newer treatment modalities require that ACA be distinguished from SCC. Recent advances in oncology suggest that lung ACA in patients without a smoking history may have sensitizing mutations, which are the single most important predictor of response to epidermal growth factor receptors–tyrosine kinase inhibitors when compared with other lung carcinomas.6–8 Also, bevacizumab (Avastin, Genentech, Inc, South San Francisco, California), a recombinant humanized monoclonal antibody inhibitor of vascular endothelial growth factor (VEGF), has been associated with life-threatening hemorrhage in lung SCC. Therefore, SCC histology is currently considered a relative contraindication for bevacizumab therapy; in addition, ACA and SCC have differing response rates for pemetrexed (ACA) and gemcitabine (SCC) doublets.9
Although hematoxylin-eosin (H&E) evaluation is sufficient to classify many NSCLCs, that can be more difficult in poorly differentiated cases. In addition, patients often present with advanced, inoperable disease. Histologic classification is further complicated by most lung cancers being diagnosed by cytology or small biopsies. A study of more than 300 primary lung cancer resections showed that 72% were diagnosed by biopsy or cytology, either alone or in conjunction.10 Concordance rates as low as 81% between pathologists have been reported in subtyping NSCLC on H&E alone.11,12 The use of immunohistochemistry (IHC) has been well documented as an important ancillary tool in diagnosing lung carcinoma. Immunohistochemistry is also a valuable tool in the distinction of metastatic lesions from primary lung carcinomas.2
Although there has been a long-standing quest to identify a “lung-specific tumor marker,” these efforts have, until recently, largely been directed at distinguishing primary from metastatic lesions.13 Given the important therapeutic and prognostic information described above, identification of a “histologic-specific tumor marker” has recently emerged as a valuable goal, and a number of markers have been studied. Given the inherent difficulties of trying to rely on a single antibody, panels of immunohistochemical markers have been used to improve sensitivity and specificity. Panels have included combinations of thyroid transcription factor 1/carcinoembryonic antigen (TTF-1/CEA) for ACA and cytokeratin (CK) 5/6 and p63–desmoglein 3 for SCC.13,14 We elected to focus our current study on p63, CK5/6, TTF-1, and napsin A. The first 3 of these markers are widely available and in common laboratory use. Although TTF-1 has historically been the most-specific marker for lung ACA, napsin A, a relatively new marker, has recently been shown1 to be more sensitive and specific than TTF-1, especially when dealing with small cell carcinoma from various sites and was, therefore, included in our study. More specifically, TTF-1 can be positive in small cell carcinoma from various sites, whereas napsin A has been uniformly negative in this context. We and others have found that TTF-1 and napsin A are a dynamic pair that help rule out small cell carcinoma (TTF-1+, napsin−) and rule in ACA simultaneously (TTF+, napsin+).
TTF-1 is a 38-kDa, homeodomain protein that shows nuclear-specific staining. It regulates gene expression in the thyroid, lungs, and diencephalon during embryogenesis. Follicular cells of the thyroid, areas of the developing brain, type II pneumocytes, and nonciliated bronchiolar epithelial cells all normally express TTF-1.15 The use of TTF-1 has been well established for the differentiation between primary and metastatic ACA of the lung. Although TTF-1 has been considered a relatively restricted marker with high sensitivity, the reported sensitivity of TTF-1 for lung ACA has been as low as 54%.1,16
Napsin A is a functional aspartic proteinase involved in the maturation of prosurfactant protein B in type II pneumocytes, and in the maturation of the biologically active surfactant protein B. It is a single-chain protein that is normally expressed in type II pneumocytes, alveolar macrophages, renal tubules, exocrine glands, and pancreatic ducts.15–17 The role of napsin A in differentiating primary from metastatic ACA of the lung has previously been reported.12,15,18,19 Positive immunohistochemical staining shows intense granular cytoplasmic reactivity.1,18,20 The reported sensitivity of napsin A for lung ACA is 74% to 87%.1,16,21
p63 is a member of the p53 family, located on chromosome 3q27–29. It is involved in the regular growth and development of epithelial tissue and is typically expressed in SCC as well as other neoplastic lesions.22 To correctly interpret p63, only nuclear staining should be considered as a positive result. Sensitivity for lung SCC was reported to be as high as 100% in a study of 30 well-differentiated tumors by Wang et al23 but decreased to 80% in poorly differentiated SCC.
Cytokeratins are the dominant, intermediate filament proteins of the epithelial cells. Cytokeratin 5 and 6 are related proteins, often detected with the CK5/6 antibody. CK5/6 can be found in normal breast myoepithelial cells, prostate basal cells, the basal layer of skin, and basal cells of the salivary gland. Positive IHC demonstrates membranous staining. In a study by Marson et al,24 100% of primary lung SCC stained positive for CK5/6; however, CK5/6 can be positive in a variety of other neoplasms, including breast carcinomas with the basal phenotype and mesothelioma.25
Recently, a 5-antigen commercial panel has been marketed to subclassify NSCLC. This panel consists of CK5/6, MUC1, the carcinoembryonic adhesion molecule CEACAM5, TRIM29, and SLC7A5. Although a detailed description of how these immunohistochemical markers were selected is beyond the scope of this article, a thorough explanation is available in the article by Ring et al.26 Of note, 2 of the most widely used immunostains, p63 and TTF-1, are not included in the newly marketed panel. The 2009 study from Ring et al26 showed the 5-antigen commercial panel to have a greater sensitivity, when compared with a panel of TTF-1/p63 (88.6% versus 74.1%); specificity was equal between the 2 methods.
Given the incidence of lung cancer, one can easily see the overall cost savings that would be realized by a smaller, but equally accurate, immunohistochemical panel. The aim of our current study was to identify a cost-effective immunohistochemical panel that can accurately distinguish primary lung ACA from SCC.
MATERIALS AND METHODS
Tissues studied were made available by the Laboratory of Pathology, Toyama University Hospital (Toyama, Japan). These materials consisted of an unstained, formalin-fixed tissue microarray. The microarray consisted of tissue derived from resected specimens with a confirmed H&E diagnosis. One core per tumor was represented on each slide. The microarray was divided geographically on the glass slide into blocks of ACA, SCC, and nonneoplastic lung disease. Based on the H&E appearance of each specimen on the microarray, similar to what one would encounter in small lung biopsies, we thought it would be instructive to keep separate data on poorly differentiated ACA tissues. In the current study, we defined poorly differentiated ACA as those microarray specimens that had no H&E characteristics of ACA, despite the confirmation of ACA on a larger tissue resection. The diagnoses on the microarray were previously confirmed to be representative of the original block by one of us (J.F.27). A total of 291 lung cancers were evaluated morphologically (197 ACAs [75%]; 66 SCCs [25%]; 28 cases could not be classified into either ACA or SCC). Each specimen was 0.6 mm in diameter. Immunohistochemistry was performed at our IHC laboratory.
Antibodies used included napsin A (prediluted rabbit polyclonal antibody; Cell Marque, Rocklin, California), CK5/6 (prediluted; Cell Marque), p63 (prediluted; Biocare Medical, Concord, California), TTF-1 (1∶80; Dako, Carpinteria, California), and TTF-1 (SPT-24, Novocastra mouse monoclonal antibody; Leica Microsystems Inc, Buffalo Grove, Illinois). All slides were treated for 30 minutes at 100°C and then allowed to cool for 20 minutes using the Revel antigen retrieval solution (Biocare Medical). The slides were incubated in the primary antibody for 1 hour, followed by detection with Mach 2 universal polymer-horseradish peroxidase (Biocare Medical) for 30 minutes, and developed with diaminobenzidine (Ventana, Tucson, Arizona). After staining, the tissue microarrays were scanned by Spectrum/Spectrum Plus (Aperio Technologies, Inc, Vista, California) to a dedicated server. To simulate what is often encountered with small biopsies, diffuse, brown, granular cytoplasmic staining in more than 1% of cells was scored as positive for napsin A and CK5/6 as was nuclear staining of more than 1% of the nuclei for TTF-1 and p63; otherwise, the case was scored as negative. All results were evaluated against a negative control of the same tumor. See Figures 1 through 4 for examples.
The likelihood-ratio χ2 test was used for contingency tables with frequency count data. A Fisher exact test was performed to take into account small frequencies that occurred in some tables. A multiple logistic regression was used to evaluate different contributions of the markers studied for the differential diagnosis between ACA and SCC in these data. These were calculated in the usual format.
Sensitivity (SEN), specificity (SPEC), positive predictive value (PPV), negative predictive value (NPV), and accuracy (ACC) were calculated for ACA as a positive result and for SCC as a negative result and were used in a differential diagnosis sense. That is, sensitivity is listed as a percentage of ACA cases classified correctly, and specificity is given as a percentage of SCC cases classified correctly.
Statistical analysis was performed with SAS (SAS Institute, Cary, North Carolina) version 9.2 software in Microsoft (Redmond, Washington) Windows.
The SEN, SPEC, PPV, and NPV results of IHC for napsin A, TTF-1 (Dako and SPT24), CK5/6, and p63 are summarized in Tables 1 and 2. Positive immunoreactivity (SEN) was seen in 164 of 197 pulmonary ACA specimens (83%) for napsin A; SPEC was 98% for ACA, whereas TTF-1 (Dako) had 20% SEN and 100% specificity for ACA. TTF-1 (SPT 24 clone) had a SEN of 60% and 98% SPEC. For SCC, CK5/6 had 53% SEN and 96% SPEC, and p63 had 95% SEN and 86% SPEC. The SEN and SPEC for napsin A and TTF-1 dropped for poorly differentiated ACA from 88% to 64% and from 68% to 59%, respectively. A poorly differentiated ACA positive for napsin A and negative for TTF-1 is illustrated in Figure 1, A and B. A panel of napsin A and p63 had a SPEC of 94% and a SEN of 96%; results are summarized in Tables 3 and 4 and illustrated in Figures 2, A through D; 3, A through D; and 4, A and B. Logistic-regression results in a prediction equation for distinguishing ACA from SCC with napsin A and p63 (P < .001) classified 96% of the cases correctly (Figure 5, A and B). The Hosmer-Lemeshow goodness-of-fit test for distinguishing between ACA and SCC in this data set (n = 263) indicates a very good fit (P = .95) to the data using p63 and napsin A.
The Novocastra SPT24 clone for TTF-1 performed significantly better than did the clone provided by Dako. TTF-1 (SPT24) had 60% SEN and 98% SPEC, whereas the Dako clone had 20% SEN and 100% SPEC. The risk of false-negative staining results was greatly reduced by using the SPT 24 clone.
CK5/6 was more specific than p63 for SCC (96% versus 86%) but did not contribute overall to distinguishing ACA from SCC. Of the 66 SCC cases that we evaluated, there were no cases (0%) that stained positive for CK5/6 but were negative for p63; overall, p63 was more sensitive for SCC than was CK5/6 (95% versus 53%). CK5/6 did not detect any additional cases that p63 missed. We noticed a similar pattern with napsin A and TTF-1 (clone SPT24) for ACA. Of 197 cases of ACA, TTF-1 (SPT24) positively stained 6 cases (3%) that were negative for napsin A staining (see Figure 6, A and B).
Several recent studies have addressed the issue of using IHC panels to correctly distinguish ACA from SCC lung carcinomas. Two recent articles on this subject are the studies of Terry et al28 and Mukhopadhyay and Katzenstein.29 One observation regarding the use of IHC to classify ACA or SCC has been that the clinical trials validating the various histologic-specific treatment modalities were based on H&E classification. Therefore, the new International Association for the Study of Lung Cancer, American Thoracic Society, and European Respiratory Society International Multidisciplinary Classification of Lung Adenocarcinoma30 includes statements that, until more data are available, cases diagnosed on biopsy that require IHC for classification should be assigned a diagnosis of NSCLC favoring ACA or SCC. This observation applies to both of the previous articles as well as to our current study.
Our results also underscore the importance of antibody choice in IHC. The SEN and SPEC of different clones can profoundly affect the diagnosis. The Novocastra SPT 24 clone for TTF-1 performed significantly better than did the clone provided by Dako. Antibody choice is also a significant consideration for napsin A. We found the polyclonal antibody from Cell Marque to have a greater SEN and SPEC (83% SEN; 98% SPEC) than what has been reported previously (58%–59% SEN; 94%–100% SPEC) for the IP64 clone from Dako and the TMU-Ad02 clone from Immuno-Biological Laboratories (Takasaki, Japan).1,14,28,29 The previously published results were based on monoclonal antibodies. The specificity of our antibody would likely decrease if the differential was expanded to include metastatic nonpulmonary adenocarcinomas. However, in the differential of primary ACA versus SCC, the polyclonal napsin A antibody was an appropriate and useful tool.
The goal of our study was to focus on the most cost-effective tissue-preserving panel that would allow accurate separation of these 2 entities. Although the ideal classification of lung tumors is based on resection specimens that allow inspection of the entire tumor, this is often not the case for pathologists who are increasingly faced with small biopsies. During the updating of the lung cancer classification on resection specimens, neither the 1999 nor the 2004 World Health Organization classification of lung tumors addressed small biopsy diagnoses.30,31 The recent International Multidisciplinary Classification of Lung Adenocarcinoma further highlights the importance of correctly separating ACA from SCC.31
The most accurate and practical method for applying our results to IHC panels is for individual cases. We found that a napsin A+ profile, regardless of p63 staining, was most likely an ACA (see Figure 7, A and B), although in our sample the risk of a false-positive ACA was greater when p63 was also positive (1.5% false-positive results; Figure 7). A napsin A− p63+ profile is most likely an SCC. Interestingly, a napsin A− p63− profile is also most likely to be an ACA (90% ACC). The combination of TTF-1 and napsin A could be useful in the double-negative category (napsin A− p63−) for excluding small cell carcinoma and eliminating non–lung primary tumors. Overall, we found that 3% of cases (n = 8) would require additional markers.
Another important aspect of using fewer immunostains is tissue preservation. The new classification system has a strong recommendation that EGFR mutation status, in addition to other molecular tests, be assessed in patients with advanced lung ACA.30,31 Clearly, if the tissue is exhausted with an extensive IHC workup, there is a greater risk that material will not be available for molecular evaluation. This is especially valuable for cytologic specimens, in which there is frequently minimal tissue with which to work.
In conclusion, we found that the best IHC panel for distinguishing primary lung ACA from SCC to consist of p63 and napsin A. This panel offered the best discrimination and correctly separated ACA from SCC with 96% ACC. Further study, from a multidisciplinary approach, including oncologists, pulmonologists, and pathologists, will be necessary to validate the treatment efficacy in patients diagnosed by IHC on small biopsy.
From the Department of Pathology, University of Texas Health Science Center, San Antonio (Drs Whithaus, Prihoda, and Jagirdar); and the Department of Pathology, University of Toyama, Toyama, Japan (Dr Fukuoka).
The authors have no relevant financial interest in the products or companies described in this article.
Presented as a poster at the 2011 annual meeting of the United States and Canadian Academy of Pathology; March 1, 2011; San Antonio, Texas.