Background: Gene signatures of sporadic colorectal carcinoma tissues and microdissected colorectal tumor cells were analyzed to identify stromal and tumor cell-specific markers, respectively.
Methods: Serial sections of frozen colorectal tumors (n=29) were subjected to RNA isolation of (1) entire tissue sections with a various tumor cell content and of (2) microdissected invasive tumor cells. Three matching samples of microdissected normal colorectal epithelial and invasive tumor cells were similarly obtained. RNA samples were analyzed using the HG95A and HG95Av2 GeneChip microarrays (Affymetrix). The microarray data was evaluated by established methods and validated by Q-RT-PCR.
Results: Unsupervised hierarchical cluster analysis of 18 sample pairs (training set) clearly distinguished tumors from microdissected tumor cells. A 149-gene signature was identified using statistical methods, which was then validated by a hierarchical clustering analysis of 11 independent sample pairs (test set). Genes specifically associated with microdissected invasive tumor cells were for example CKS2 and NME1. In contrast, genes associated with stromal cells were for example MMP2, SDF1 and FBLN2. Finally, a 65-gene signature distinguished normal colorectal epithelial cells and invasive tumor cells, including down-regulation of BMP2 and ANPEP mRNA expression as well as up-regulation of TKT, SPARC, MCM5 mRNA expression.
Conclusions: Our approach allowed precise evaluation of molecular signatures in morphologically defined cell populations and identified novel target genes related to stroma-tumor interactions in colorectal cancer. The approach enables further analysis of gene signatures in different tumor areas and cell types, such as within invasive margins to decipher molecular mechanisms of colorectal cancer invasion and metastasis.