Many men diagnosed with prostate cancer can be cured with radiation or surgery; however, those whose cancer has breached the boundaries of the prostate will relapse clinically and must undergo androgen ablation therapy, which ultimately leads to incurable, androgen-independent (AI) disease. Although AI tumor biopsies are difficult to obtain and study, the molecular characteristics and mechanisms must be identified to enable the development of a curative or even preventative therapy. In this study, we directly compared the gene expression profiles of ten AI tumor biopsies with ten primary, untreated tumors (presumed androgen-dependent, AD). Each sample was laser capture microdissected, and the RNA amplified, labeled, and hybridized to Affymetrix U133A Gene Chips. The data were examined for the technical validity and robustness of this approach and utilized to identify gene expression levels unique to the AI samples. Principal component analysis identified the three dimensions of highest variation, which incorporated 34.7 % of the data, and showed that the AD and AI tumors segregated from one another on this basis. We filtered the data to include only probe sets marked as present (P) or marginal (M) in at least one sample (13,177) and differentially expressed at least twofold (663). Hierarchical clustering showed that clear separation of the AD and AI tumor groups persisted. We then filtered the data to include only those probe sets P or M in at least 20% of the samples and differentially expressed at p<.005, which identified 239 genes. We used the Expression Analysis Systematic Explorer (EASE) to identify gene ontology over-representation, and found macromolecule biosynthesis and cell adhesion to be most clearly differentially regulated between the two groups. Of the individual genes identified as differentially expressed, many were related to IL-6 signaling, which has known significance in prostate cancer, as well as processes of carcinogenesis including angiogenesis, cell adhesion, apoptosis, oxidative stress, and dysregulation of hormone-responsive genes. Finally, we used DIfferential Gene locus MAPping (DIGMAP), which maps differential gene expression to the relevant chromosomal loci, to identify potential areas of chromosomal deletion. Nine regions were identified including 1p36, 3p21, 6p21, 8p21, 11p15, 11q12, 12q23, 16q12, and 16q21. Several of the genes in these regions matched those in the list of 239 differentially expressed genes from our initial analysis. These data identify several unique characteristics of AI prostate cancer that may hold potential for the development of targeted therapeutic intervention.

[Proc Amer Assoc Cancer Res, Volume 46, 2005]