Molecular signatures generated with RNA isolated from formalin-fixed paraffin-embedded tumor specimens differentiate metastatic and non-metastatic primary colorectal cancer

Abstract

3615 Background: Colorectal cancer (CRC) can be either non-metastatic (locally-advanced) or metastatic, with the latter having a considerably worse prognosis. Because tissue from patients is generally available only as FFPE specimens, we investigated whether distinct molecular signatures could be obtained for these CRC types using RNA isolated from archival FFPE specimen blocks in gene expression arrays. Methods: FFPE specimens were available from 12 non-metastatic tumors and 9 primary tumors along with their matching metastases. Up to 5 10-micron sections of each were microdissected to isolate areas of tumor tissue. RNA was extracted using a proprietary procedure of Response Genetics, Inc. and was then amplified and labeled. The resulting cRNA was hybridized to the U133 plus 2.0 GeneChip. Unsupervised PCA analysis of the samples resulted in the first principal component separating 2 distinct groups, which consisted of the non-metastatic and metastatic tumors. A differentially expressed gene list between metastatic and non-metastatic CRC was determined. These data were also analyzed for differential canonical pathways using Ingenuity Pathway Analysis. Results: Hierarchical clustering analysis segregated locally advanced primary tumors and metastatic primary tumors into two clusters with distinct gene signatures. A T-test with unequal variance assumption identified 609 differentially expressed probe sets with FDR (false discovery rate) = 0.05. Comparison of primary tumors with their liver metastases using a paired T-test showed only 2 differentially expressed genes at FDR = 0.05 but 526 genes with significance p value < 0.005. Pathway analysis showed significant deregulation of VEGF, hypoxia, B Cell receptor, PI3K/AKT, ERK/MAPK and G-protein coupled receptor signaling between locally advanced and metastatic tumors. Pathway analysis of primary tumors and metastases showed deregulated TGF-b, integrin and chemokine signaling pathways. Conclusion: We have demonstrated the feasibility of identifying metastatic and non-metastatic tumors by microarray analysis using FFPE CRC tissue. This result is currently being validated in a separate larger cohort of patients. [Table: see text]