Close relation of large cell carcinoma to adenocarcinoma by hierarchical cluster analysis: implications for histologic typing of lung cancer on biopsies.

Abstract

Determining histologic types of lung cancer on biopsies can be difficult. This study addresses the role of immunohistochemistry in histologic typing, using a tissue microarray (TMA) as "model biopsies," and presents a classification generated by an unsupervised hierarchical cluster analysis. A TMA was made from resection specimens of a consecutive series of 165 lung tumors. In a "tissue-spot review" with hematoxylin and eosin sections all the large cell carcinomas (N=22) were assigned to the noncommittal class of non-small cell lung cancer (NSCLC), as were an additional 37 tumors of defined histologic types. Adenocarcinomas and squamous cell carcinomas included with these NSCLC could be diagnosed by immunohistochemistry with antibodies against TTF-1, Napsin A, cytokeratin (CK)7, p40, p63, and CK5/6 with moderate to good sensitivities and specificities. Unsupervised hierarchical clustering was done with these data and additional high-molecular-weight cytokeratins, CD56, synaptophysin, and chromogranin immunohistochemistry. This delineated separate clusters for adenocarcinomas, large cell carcinomas, neuroendocrine tumors, and squamous cell carcinomas. Notably, adenocarcinoma and large cell carcinoma clusters were closely related and clearly set off from the squamous cell carcinoma cluster. As would be expected for a clinically well-staged series CDX2, GATA3, estrogen, and progesterone receptor immunohistochemistry remained negative in the vast majority of the tumors and, if positive, were restricted to very few cells. These results, the clustering data in particular, underpin the pragmatic recommendation canvassed with the IASLC/ATS/ERS classification of lung cancers that adenocarcinoma-type molecular studies should include NSCLC with a nonsquamous cell carcinoma immunophenotype.