Notes on BDNF and CREB

CREB is a transcription factor, i.e., a protein that binds to specific sites in the genome and thereby affects the transcription of a gene or genes near the binding site. (Actually, in a few cases the gene being controlled is up to 1Mb away from the binding site.) To see what the binding sites look like, i.e., the DNA pattern recognized by the transcription factor, let's try TFSEARCH. The main function there is to search a user-specified DNA sequence for binding sites, but we can look for binding patterns by clicking on "TFFACTOR". Unfortunately, the 1483 binding sites aren't sorted by name. Also, there are several sites labeled "CREB", namely numbers 478, 480, 488, 653 and 739. For us, the worse news is that these seem to give just the protein sequences, not the binding sites.

Undaunted, we try TESS. Under "Search TRANSFAC", we type "CREB" and select "Matrices". We are led to some so-called weight matrices that specify binding sites that look roughly like TGACGTAA. Actually, TESS gives 9 slightly different matrices, each derived from a different experiment to determine the sequences to which CREB binds. (Later in the course we'll see how one should search a sequence to find regions that match such a "frequency matrix" of short, experimentally determined sequences.) At the UCSC browser, you can type "CREB" to get an idea of how many CREB genes there are in human.

Searching a given sequence for predicted binding sites frequently produces to many matches. To see this, submit a sequence of length, say, a thousand, such as this one to TFSEARCH or TESS. Here is what I got back from TESS after about 10 minutes. By requiring that the matrix match human, mouse, rat, ... at the same place (which is what UCSC does), the number of matches is reduced about 100-fold.

CREB is regulates the BDNF gene, as shown in this paper. More generally, regulatory pathways involving CREB are being studied in hopes of learning why anti-depressant drugs have a low success rate, as summarized in this abstract.

Somewhere around the BDNF gene, there may be CREB binding sites. (Do the experiments reported by Conti et al. leave open the possibility that CREB regulates another gene, which codes for a transcription factor that regulates BDNF?) Let's go to UCSC's Human Genome Browser and type in "BDNF". Notice the number of alternate splice forms. There seem to be at least three transcription start sites, at around chr11:27,653,000-27,660,000, chr11:27,675,000-27,685,000 and chr11:27,695,000-27,705,000. For small regions like these, it makes sense to dial "CpG Islands", "FirstEF" and "TFBS Conserved" (all three are under "Expression and Regulation") up to "full".

Be warned that identifying the precise way that a transcription factor regulates a gene is typically difficult, and purely bioinformatic approaches rarely give the full picture.

Now turn on "GNF Atlas2" (under Expression and Regulation) and turn off other tracks. Do you see expression in the brain?

What else can we learn about BDNF? Clicking on the the KnownGenes entry, we can find the Protein Browser page for BDNF. It is interesting to follow the the KEGG link at the bottom of the Protein Browser page. Could we look at the KEGG link for the CREB gene and get an idea of which genes CREB could regulate, which in turn might be direct regulators of BDNF?

Then, follow the Gene Sorter link. You can sort by expression pattern, sequence similarity, and other properties. For BDNF, we don't find any genes whose expression pattern is significantly similar, but we can sort by blastp scores and find many similar protein sequences. Another generally worthwhile approach is to sort by GO (Gene Ontology) categories, which should in theory identify the proteins with a similar function. However, it doesn't seem to help with BDNF.

You can learn what is known about the human-disease aspects of BDNF (or any gene) by following the OMOM (Online Mendelian Inheritance of Man) link from the KnownGenes page.

Given some genes with similar expression patters or similar functions, and hence genes that might be co-regulated (turned on or off by the same combination of transcription factors), how do we look for patterns in the genomic DNA around the gene that might be the regulatory signal? There are a number of websites that that take a set of generally dissimilar DNA sequences and look for short conserved patterns. We can use Galaxy. to extract genomic sequences. I've put the three segments mentioned above that span putative transcription start sites of human BDNF into this file. I submitted the file to MEME, using the default settings, except that I specified "Any number of repetitions" and changed the "Maximum number of motifs to find" from 3 to 20. After roughly 30 minutes, the MEME server sent email messages containing both MEME results and MAST results.