Gene expression profiling in lung adenocarcinomas reveals molecular signatures of potential biological significance

Abstract

Background Lung cancer remains the world-wide leading cause of cancer morbidity and mortality as a result of late presentation or diagnosis. Surgical resection could be curative for early stage lung cancer while prognosis for advanced stage lung cancer is still poor despite continuous development and trial of new therapeutic strategies and chemotherapeutic agents. Study in lung carcinogenesis is essential for better understanding of lung cancer biology that may open up new diagnostic and therapeutic targets for intervention. Gene expression represents the underlying genetic changes and reflects the activities of cancer cells. Microarray is a high throughput method that allows for parallel study of the expression of thousands of genes, which will allow for gene expression profiling and class discovery. Our hypothesis is that differential gene expression in lung adenocarcinomas could be correlated with clinical characteristics and molecular classification of tumour subtypes. Methods We have collected resected lung adenocarcinomas (n = 49) and normal lung tissue (n = 9). Total RNA from these specimens was processed according to standard procotol from Affymetrix and then hybridized onto Affymetrix GeneChip HG-U133A with one sample per GeneChip. Data was analysed with dChip 1.3 and correlated with clinical characteristics. Results Differential gene expression was obtained with two sample t-tests of log transformed signal intensities, yielding 318 genes having up- or downregulation by four fold difference, and some of these can be classified according to their known gene functions such as involvement in cell cycle or signal transduction. Unsupervised data analysis with hierarchical clustering aiming at detection of co-regulated or co-expressed genes showed significant gene clustering of molecules involved in cell cycle and proliferation, namely TOP2A, CENPF and MCM2 (P = 0.008); as well as consistent sample clustering of normal lung tissues and groups of adenocarcinomas. Supervised data analysis with principal component analysis were applied for class discovery, with a group of 35 filtered genes being used as potential molecular classifier for subtypes of adenocarcinomas. Further correlation with clinical characteristics and prognosis will be analysed. Conclusion We have identified differentially-expressed genes in lung adenocarcinomas of potential biological significance that deserve further investigation as potential diagnostic or therapeutic targets in the management of lung cancer.