Gene expression profile of human colorectal cancer using oligonucleotide microarray

Abstract

e22182 Background: Current microarray technologies allow the analysis of thousands of genes simultaneously. We analyze the gene expression profile of different samples from human colorectal carcinoma in relation with the clinical stage at diagnosis. Methods: Thirty-one patients were included: 13 with stage II (transmural invasion of the colonic wall without lymph node invasion), 10 with stage III (regional lymph node invasion) and 8 with stage IV at diagnosis (metastatic). RNA was extracted from fresh tumour samples, and the quantity and quality were assessed by spectrophotometric method and capilar electrophoresis. Tumoral samples were hibrizided with a pool containing non-tumoral colonic mucosa from 33 patients with colorectal cancer. Normalized crude data were used for the initial selection of the genes, using several univariate metrics (Matusita, Kullback-Leibler, Bhattachryya). A high confidence Bayesian network was used to compare the best ranked genes with the stage of the tumour at diagnosis. Results: We identified a group of 20 genes with potential prognostic value. Of them, nine genes (HAS1, LO9861, NHLRC2, ATP6V0E2, GTFH2H2, FLJ34077 , CASP12 and RP5–1077B9.4) were able to discriminate cases with stage II-III and those with stage IV, and 11 genes (MEA1, SPRY4, TLR8, SAMSN1, DGCR14, MRCL3, SMAD2, SFRS12, LAPTM4A, CD18 and FAM111A) could discriminate between patients with stage II or III, in 95% of the cases. Data were validated with 18 new patients with colorectal cancer using a quantitative method (RT- PCR). Conclusions: Only few genes are able to discriminate the clinical stage of the tumour at diagnosis with a fresh sample of the primary. Patients with an unfavourable signature might benefit of a more extensive initial work-up. No significant financial relationships to disclose.