Identification of conserved regulatory elements in mammalian promoter regions: a case study using the PCK1 promoter

Abstract

A systematic phylogenetic footprinting approach was performed to identify conserved transcription factor binding sites (TFBSs) in mammalian promoter regions using human, mouse and rat sequence alignments. We found that the score distributions of most binding site models did not follow the Gaussian distribution required by many statistical methods. Therefore, we performed an empirical test to establish the optimal threshold for each model. We gauged our computational predictions by comparing with previously known TFBSs in the PCK1 gene promoter of the cytosolic isoform of phosphoenolpyruvate carboxykinase, and achieved a sensitivity of 75% and a specificity of approximately 32%. Almost all known sites overlapped with predicted sites, and several new putative TFBSs were also identified. We validated a predicted SP1 binding site in the control of PCK1 transcription using gel shift and reporter assays. Finally, we applied our computational approach to the prediction of putative TFBSs within the promoter regions of all available RefSeq genes. Our full set of TFBS predictions is freely available at http://bfgl.anri.barc.usda.gov/tfbsConsSites.

Fig. 1

Raw score distribution family types and their relationship with the length and information content of matrices. The occurrences of raw scores within each bin of size 0.001 were recorded and plotted (red). The corresponding normal distribution was fitted to the observed values and superimposed on the histogram plots (black). A representative histogram is presented for each family type. Histograms of Types 1–4 (A, B, C and D) have means centered close to 0.5. Type 1 histograms have symmetric Gaussian bell-shaped curves; Type 2 histograms have asymmetric Gaussian bell-shaped curves; Type 3 histograms are symmetric but do not follow a Gaussian bell curve; and Type 4 histograms are asymmetric and do not follow a Gaussian bell curve. Types 5–8 (E, F, G and H) correspond to Types 1–4, respectively, except that Types 5–8 histograms are shifted to the left. I. Distributions of the motif length and information content for each histogram type.

Fig. 2

Known and predicted TFBSs as custom tracks on hg18 in the UCSC Genome Browser. Known and predicted TFBSs are represented as separate tracks in human genome assembly chr20: 55,569,120-55,569,542. Known TFBSs are in black and newly identified TFBSs are in blue. Predictions are organized into three tracks according to the JASPAR collection: MA (green), MF (gray), and PF (red). Displayed UCSC browser tracks include 5-way (5X) and 7-way (7X) regulatory potential, repeating elements by RepeatMasker, and Human/Mouse/Rat conserved TFBSs.

Fig. 3

Experimental verification of newly identified proximal SP1 binding site in the PCK1 promoter. A. Schematic illustration of the location of the proximal SP1 binding site (chr20: 55,569,514–55,569,528) in the rat PCK1 promoter. The rat PCK1 promoter sequence (−490/+72) was linked to a luciferase gene to create a reporter construct, p490-Luc. B. Endogenous SP1 interacts with the proximal SP1 binding site. EMSA was performed using DNA fragments containing wild-type (SP1 WT) (Lane 1) or mutated (SP1 Mut) (Lane 2) binding site. The binding of SP1 to DNA fragment was confirmed via a super-shift with antibody against SP1 (SP1 IgG) (Lane 3). NE, nuclear extracts. C. Overexpression of SP proteins alters the PCK1 promoter activity. Control plasmid (Con) and plasmids over-expressing SP1 or SP3 were co-transfected with a luciferase reporter plasmid p409-Luc into HepG2 cells. The results are expressed as the means of relative luciferase activity ± S.E.M. for three experiments.