Abstract
3555 Background: A more accurate method of identifying stage 2 and 3 colorectal cancer (CRC) patients at highest risk for recurrence after surgical resection is needed. Gene expression signatures utilizing microarray-derived gene expression data from fresh frozen primary CRCs to predict risk of recurrence have been developed by us and others. Advances in technology platforms for gene expression measurements and their applicability to formalin-fixed, paraffin-embedded (FFPE) specimens offer new opportunity to develop clinically useful diagnostics based on molecular profiles. Methods: 58 patient FFPE samples of all stages stored from 1-12 years were collected from the Vanderbilt GI SPORE Translational Pathology and Imaging Core and annotated with appropriate clinicopathologic data. 414 genes were selected from our 34-gene prognostic classifier and other published CRC gene signatures, as well as gene elements associated with intestinal stem cell biology and epithelial-to-mesenchymal transition (EMT). RNA was extracted from the tumors, and gene expression analysis was completed using the nCounterplatform. Results: Quality of extracted RNA from tumor blocks was similar among the tumors and adequate for analysis. No significant differences were seen in signal strength (p=0.94, Kruskal-Wallis test) or intra-class variation (correlation coefficient = 0.99) across material extracted from new and old blocks. Fold change values for the 70 most highly differentially expressed genes on the nCounter platform correlated well with Affymetrix U133 plus 2 microarray (R2=0.819). Genes associated with EMT clustered according to prognosis, with poorer prognoses seen in patients with high TWIST expression or low E-cadherin and SMAD4 expression. There was a trend toward better survival outcomes with high expression of E-cadherin and SMAD4 (p=0.072, log-rank test). Conclusions: This preliminary study demonstrates the feasibility of this approach to determine gene expression patterns in FFPE tumor tissue samples. Our data suggest that this approach may be applied to identify clinically applicable prognostic gene expression profiles that may be validated in archived patient samples that are well annotated with patient outcome data.