Nuclear factor I revealed as family of promoter binding transcription activators

Abstract

Background: Multiplex experimental assays coupled to computational predictions are being increasingly employed for the simultaneous analysis of many specimens at the genome scale, which quickly generates very large amounts of data. However, inferring valuable biological information from the comparisons of very large genomic datasets still represents an enormous challenge.

Results: As a study model, we chose the NFI/CTF family of mammalian transcription factors and we compared the results obtained from a genome-wide study of its binding sites with chromatin structure assays, gene expression microarray data, and in silico binding site predictions. We found that NFI/CTF family members preferentially bind their DNA target sites when they are located around transcription start sites when compared to control datasets generated from the random subsampling of the complete set of NFI binding sites. NFI proteins preferably associate with the upstream regions of genes that are highly expressed and that are enriched in active chromatin modifications such as H3K4me3 and H3K36me3. We postulate that this is a causal association and that NFI proteins mainly act as activators of transcription. This was documented for one member of the family (NFI-C), which revealed as a more potent gene activator than repressor in global gene expression analysis. Interestingly, we also discovered the association of NFI with the tri-methylation of lysine 9 of histone H3, a chromatin marker previously associated with the protection against silencing of telomeric genes by NFI.

Conclusion: Taken together, we illustrate approaches that can be taken to analyze large genomic data, and provide evidence that NFI family members may act in conjunction with specific chromatin modifications to activate gene expression.

Figure 1

NFI family of DNA binding proteins preferentially binds upstream of initiation sites of transcription in mouse genome. A set of 61,492 NFI predicted sites was defined in the mouse genome, NCBI build 37 or mm9. Sites were defined with a previously established position weight matrix (sequence score threshold > 85, out of maximum 100) [28]. Within this set, 2,852 predicted sites overlapped regions 5 kb upstream from RefSeq annotated transcriptional start sites (TSS). As a negative control for this experiment, a complementary dataset of 58,640 sites was selected (panel A) or the same number of NFI predicted sites (2,852) were randomly selected from the initial set of 61,492 predicted sites (panel B). Four independently performed random selection are shown. Average NFI ChIP-Seq tag counts were calculated in windows of 50 bp for a region of 5 kb up- and down-stream of the selected NFI predicted sites. Tag counts were normalized globally, as a fold increase over the genome average tag count in a window of 50 bp. Obtained data points were connected to form a continuous line.

Figure 2

Genes that contain NFI in vivo site in their 5 kb upstream regions exhibit higher expression levels. Using ChIP-Seq data, 39,807 in vivo NFI sites were defined genome-wide in mouse embryonic fibroblasts. Out of this number 3,120 in vivo sites were located within the 5 kb upstream regions of RefSeq annotated genes. 2881 RefSeq genes were selected that contained one or more in vivo NFI sites in their 5 kb upstream regions and the same number of genes were randomly chosen from the RefSeq gene annotation. Random selection was repeated 3 times. Histograms represent the distribution of gene expression levels for each of such defined groups. Affymetrix expression data were obtained from the same cell type (mouse embryonic fibroblasts), from the same embryo, and using the same culturing conditions. A-C. Randomly selected groups. D. Group of NFI occupied genes.

Figure 3

NFI binding correlates with histone H3 methylations: H3K4me3, H3K36me3, H3K9me3. 2881 NFI occupied genes in mouse embryonic fibroblasts and 3 groups of randomly selected genes were defined as in Figure 2. ChIP-Seq data for 4 different histone H3 methylations were obtained from the same cell type (mouse embryonic fibroblasts) [4]. Average ChIP-Seq tag counts were calculated in windows of 50 bp for a region of 5 kb up- and down-stream of the orientated transcription start sites (TSS). Tag counts were normalized globally, as a fold increase over the genome average tag count in a window of 50 bp for the following modifications: A. H3K4me3, B. H3K36me3, C. H3K27me3, and D. H3K9me3. Obtained data points were connected to form a continuous line. Arrows indicate the orientation of transcription in each panel.

Figure 4

NFI-C transcription factor acts as more potent activator of gene expression than repressor. A. Differences in the gene expression levels in wild-type (WT) and NFI-C knock-out (KO) mouse embryonic fibroblasts were depicted for 1000 most up-regulated genes and 1000 most down-regulated genes by NFI-C, as well as for 3 independently selected random groups of 1000 genes. On the y-axis: for each gene in such defined groups expression level in knock-out cells was subtracted from the WT expression level (E_wt-E_ko) and plotted as an absolute value. B. Affymetrix expression levels in WT cells (E_wt) for 1000 most up-regulated genes by NFI-C, 1000 most down-regulated genes by NFI-C, and 3 sets of 1000 randomly chosen genes.