Regulation of E2F dependent transcription is disturbed in nearly 100% of human tumors and cell lines as a result of the loss of a functional pRB pathway. We have used inducible transcription factors E2F1, -2, and -3 to identify the target genes affected. Highly reproducible changes in gene expression were observed in three independent replicates for each transcription factor. We developed a method based on the McNemar test that allows us to evaluate three parameters of data quality: 1. Equivalence of chip performance and chip handling. 2. Presence of differences in transcript levels between test and control target. 3. Reproducibility of measurement of such differences. The test involves triangulation of replicates and randomization of gene order. Identified gene expression patterns display a low level of false positives (2%) as verified in numerous Northern blots. Furthermore, we have developed a method that allows us to identify statistically relevant biases of gene expression patterns to arbitrary groups of genes, such as functional gene groups, genes in certain signaling pathways, or genes that are part of expression patterns identified by others. Application of this type of analysis to E2F induced expression patterns identifies significant biases towards genes involved in cell cycle regulation and apoptosis. This was expected since E2F is known to regulate these processes and proves the validity of the approach. Significant biases were also observed to genes involved in TGF beta signaling and transcription factors, in particular to homeobox genes that are key regulators of differentiation and development, respectively. Together with previous reports that linked E2F transcription to TGF beta signaling and homeobox genes, our analysis suggests a key role of the E2Fs in integrating cell cycle and apoptosis with development and differentiation. Our approach can also be used in signaling pathway analysis since we observed significant overlaps in the gene expression patterns induced by serum, receptor tyrosine kinases, activated ras, and E2F (known to be activated in response to serum). Full exploitation of this approach requires gene classifications of high quality and databases containing statistically robust expression patterns.
Copyright. All Rights Reserved.