Overview | Start at bench | Install TOUCAN | Get Sequences | Annotate | MotifScanner | Statistics | ModuleSearcher | MotifSampler | Return to bench | References
Biological Background
Endothelial cells are lining the inner wall of blood vessels and build the barrier between the blood stream and the surrounding tissue. Therefore, endothelial cells also represent a major site of immunological processes and inflammation. Interleukin-1 (IL-1) is one of the major inducers which provoke an inflammatory response of endothelial cells (Mantovani et al., 1997). We were interested to obtain a comprehensive view of the gene regulation program of endothelial cells treated with IL-1, in particular concentrating on the immediate-early to early phase, and to delineate expression-specific regulatory patterns.
The Experiment
The figure below depicts an overview how the initial experiment was performed. Human Umbilical Vein Endothelial Cells (HUVEC) were stimulated with IL-1 for various periods of time, up to 6 hours (Mayer et al., 2004). Following RNA preparation, cDNA synthesis, and in vitro transcription, each sample was hybridized to Affymetrix human U133 microarrays, which contain about 45.000 ProbeSets representing approximately 33.000 genes. The chips were scanned and image analysis was performed, which finally produced a table of values describing the expression levels of all genes over the chosen time-course. (Please proceed below the image.)
These data have been subjected to thorough procedures of sorting and filtering, which was based on two main criteria. The quality of the hybridization signal was essential, and we concentrated on those genes which showed at least 4-fold induction or repression at one of the time-points as compared to the control. This is a quite stringent value, but we wanted to focus on pronounced effects. This process finally yielded a list of 137 human genes.
Cluster Analysis
Next, we wanted to group those genes which show a similar expression profile over the time-course. Several programs are available for this purpose, and we chose EPCLUST which is part of the program package Expression Profiler (Kapushesky et al., 2004) provided by the European Bioinformatics Institute (EBI). It would be beyond the scope of this tutorial to fully describe the functionality of this program, but there is a very good documentation which can be found at the EPCLUST web-site.
The K-means clustering algorithm groups genes into a user-defined number of separate clusters. The image below shows the result of such a K-means clustering performed on our dataset of 137 gene profiles. You can see that each gene is represented by one line. The horizontal axis represents the time-course which is also indicated in cluster 1. On the vertical axis, the log ratio of the signals is displayed, red meaning up- and green meaning down-regulation.
Questions
We ask the question, if it is possible to delineate common regulatory elements in clusters of genes which show a common behavior upon stimulation. As example, we will take the genes from cluster 1, extract their proximal promoter regions, and perform a comprehensive TOUCAN analysis, in order to predict three kinds of data: known transcription factor binding sites, combinations of TFBS, and finally over-represented sequence motifs, which might represent novel regulatory elements.