Purpose: Lung cancer is the leading course of cancer deaths in Japan and a number of other countries. The treatments of lung cancer are principally based on the stage of TNM classification. However, remarkably different overall outcome and treatment response suggests the heterogeneity of patients with lung cancer at each TNM stage. Lymph node metastasis and tumor recurrence are major factors associated with poor prognosis in the cancer, but little is known of its molecular mechanism. Recent studies suggested that information from gene expression profiles could be used to develop molecular classifications of cancers. The aim of this study was to identify genes differentially expressed between normal tissue and lung cancer. We investigated gene expression profiles of 64 resected primary lung cancers by using an oligonucleotide microarray and attempted to select a set of genes predicting to lymph node metastasis and tumor recurrence in lung cancer. Methods: Fifty two pairs of primary lung cancer tissues (32 cases were diagnosed as adenocarcinoma, 12 as squamous cell carcinoma and 8 as others) and corresponding adjacent non-cancerous tissues including the clinicopathological information were obtained from patients who underwent surgical resection at Chiba University Hospital, Japan. We have analyzed gene expression profiles using an oligonucleotide microarray containing 10,800 probes (AceGene, Hitachi Software Engineering Co., Ltd. Japan). We performed hierarchical clustering analysis with expression levels of the selected genes and calculated statistical significance. To validate the predicting systems, we analyzed an additional independent set of 12 primary lung cancer cases in a completely blinded fashion. Results: In our search for candidate genes, 23 genes showed differentially expressed in lung cancer tissues with lymph node metastasis and 27 genes with tumor recurrence within a year. Using the most suitable set of genes, it was possible to classify the patients with lung cancer for lymph node metastasis and tumor recurrence or not. The predicting system yielded 75% accuracy for forecasting lymph node metastasis and 70% accuracy for forecasting tumor recurrence within a year after surgery in independent cases. Conclusions: Based on the expression analysis of 10,800 genes and clustering algorithm, we successfully distinguished clinical status of lung cancer. We developed a predictive score system that could clearly separate the poor prognosis group from the patients with lung cancers. These systems may influence the management strategy of primary lung cancer.
[Proc Amer Assoc Cancer Res, Volume 46, 2005]