ABSTRACT: The dismal lethality of lung cancer is due to late stage at diagnosis and inherent therapeutic resistance. The incorporation of targeted therapies has modestly improved clinical outcomes, but the identification of new targets could further improve clinical outcomes by guiding stratification of poor-risk early-stage patients and individualizing therapeutic choices. We hypothesized that a sequential, combined microarray approach would be valuable to identify and validate new targets in lung cancer. We profiled gene expression signatures during lung epithelial cell immortalization and transformation, and showed that genes involved in mitosis were progressively enhanced in carcinogenesis. 28 genes were validated by immunoblotting and 4 genes were further evaluated in non-small cell lung cancer tissue microarrays. Although CDK1 was highly expressed in tumor tissues, its loss from the cytoplasm unexpectedly predicted poor survival and conferred resistance to chemotherapy in multiple cell lines, especially microtubule-directed agents. An analysis of expression of CDK1 and CDK1-associated genes in the NCI60 cell line database confirmed the broad association of these genes with chemotherapeutic responsiveness. These results have implications for personalizing lung cancer therapy and highlight the potential of combined approaches for biomarker discovery. In these studies, we systematically profiled gene expression in normal (NHBE), immortalized (BEAS-2B) and fully transformed (NNK-BEAS-2B) human bronchial epithelial cells, as well as a non-small cell lung cancer (NSCLC) cell line (H157) from a smoker. Expression profiles that accompany the immortalization and/or transformation of bronchial epithelial cells were generated, and expression of 28 genes was validated by immunoblotting. 4 of them were further evaluated in immunohistochemical analyses of tissue microarrays that contain NSCLC specimens, surrounding non-diseased tissues and non-pulmonary normal tissues. Although all 4 genes were predominantly expressed in tumor tissues, loss of expression of cytoplasmic CDK1 was clinically important because it was associated with a poor prognosis for NSCLC patients. This poor prognostic value may be associated with therapeutic resistance because decreasing levels of cytoplasmic CDK1 in vitro increased resistance to standard chemotherapies used in the treatment of NSCLC, especially microtubule agents where resistance was almost complete. These studies illustrate how a combined microarray approach can facilitate the identification of new, relevant targets in cancer. Fluorescently labeled cDNA were synthesized from 100 microgram RNA by oligo(dT)-primed reverse transcription in the presence of Cy3- or Cy5-dUTP (Amersham Bisciences, Piscataway, NJ). Purified Cy3/Cy5-labelled probes were combined and hybridized in the presence of 2x Denhart's solution, 3.2x saline sodium citrate (SSC), and 0.5% sodium dodecyl sulfate (SDS) in a humidified chamber at 65°C overnight. Prior to scanning (Agilent Technologies, Foster City, CA), slides were successively washed at 22°C in 0.5x SSC/0.1% SDS for 2 min, 0.5x SSC/0.01% SDS for 2 min, and 0.06x SSC for 2 min. Image analyses were performed with the IPLab software (Fairfax, VA). The reference cell (NHBE) was included in every individual hybridization to allow for normalization of each clone’s expression relative to the reference for each cell line (BEAS-2B, NNK-BEAS-2B or H157). A self-to-self hybridization with dye reversal was performed to exclude preferential differences in probe labeling (not included here). Every sample was labeled with Cy5 and Cy3 and hybridized twice. The two fluorescent images (red and green channels) obtained from the scanner constituted the intensity raw data from which differential gene expression ratios and quality control values were calculated. All data were entered into a relational database, using the FileMaker Pro 5 software (Santa Clara, CA). Genes were identified as differentially regulated only if corrected red/green hybridization signals differed by at least two-fold. The genes in each group were alphabetically listed and the ratios of BEAS-2B/NHBE, NNK-BEAS-2B/NHBE or H157/NHBE were graphically visualized by Cluster and TreeView programs (http://rana.lbl.gov/EisenSoftware.htm). These methods fulfilled the MIAME criteria (http://www.mged.org/miame).
