Prostate cancer, the second leading cause of cancer death in North American men, is very heterogeneous in its clinical presentation and course. This variability in clinical behavior is likely reflected by underlying molecular heterogeneity among patients’ tumors, which might be captured by gene expression profiling. Using cDNA microarrays containing ∼26, 000 human genes, we profiled gene expression in 62 primary prostate tumors, 41 normal prostate samples and 9 lymph node metastases. Unsupervised hierarchical clustering identified 3 subclasses of prostate tumors based on distinct gene expression patterns. High grade and advanced stage tumors, as well as tumors associated with recurrence, were disproportionately represented among two of the three tumor subtypes, one of which also included most lymph node metastases. To further characterize the clinical relevance of tumor subtypes, we evaluated as surrogate markers genes differentially expressed among tumor subgroups, using immunohistochemistry on tissue microarrays representing an independent set of 225 primary prostate tumors. In multivariate analysis, these immunohistochemical markers were strong predictors of tumor recurrence independent of tumor grade, stage, and preoperative prostate specific antigen levels. Our results suggest that prostate tumors can be classified according to their gene expression patterns, and this classification might provide a basis for improved prognostication and treatment stratification.

[Proc Amer Assoc Cancer Res, Volume 45, 2004]