Building promoter aware transcriptional regulatory networks using siRNA perturbation and deepCAGE

Abstract

Perturbation and time-course data sets, in combination with computational approaches, can be used to infer transcriptional regulatory networks which ultimately govern the developmental pathways and responses of cells. Here, we individually knocked down the four transcription factors PU.1, IRF8, MYB and SP1 in the human monocyte leukemia THP-1 cell line and profiled the genome-wide transcriptional response of individual transcription starting sites using deep sequencing based Cap Analysis of Gene Expression. From the proximal promoter regions of the responding transcription starting sites, we derived de novo binding-site motifs, characterized their biological function and constructed a network. We found a previously described composite motif for PU.1 and IRF8 that explains the overlapping set of transcriptional responses upon knockdown of either factor.

Figure 1.

DeepCAGE and microarrays detect overall similar expression changes. The transcriptome-profiling technologies deepCAGE and microarrays showed overall similar transcriptional response (log2 expression fold-change) comparing before and after siRNA-based knockdown of the transcription factors IRF8, MYB, PU.1 and SP1. The Pearson correlation values for these two platforms are: (a) 0.389 (P = 1.3e⁻¹²) for IRF8, (b) 0.453 (P = 2.2e⁻¹⁶) for MYB, (c) 0.450 (P = 1.2e⁻¹¹) for PU.1 and (d) 0.404 (P = 6.7e⁻¹⁰) for SP1.

Figure 2.

DeepCAGE identified individual transcription starting sites responding to transcription factor knockdown. DeepCAGE profiling of the transcriptome quantitatively measures individual transcription starting sites (TSS) of capped mRNA indicated by the vertical bars (a) before and (b) after the knockdown of the IRF8 transcription factor. Red bars indicate CAGE tags that do not change upon knockdown while the black bars represent tags showing significant change upon knockdown. One transcript cluster (TC) is shown in the promoter region of the XAF1 gene on chromosome 17 (positions 6 600 047–6 600 115, hg18) together with the defining TSSs.

Figure 3.

TFBS motifs derived for PU.1 and IRF8 as activators. The 50 strongest downregulated TCs after knockdown of each of the two TFs PU.1 and IRF8 and their corresponding promoter regions were used as training data set to identify binding-site motifs and their respective PWMs (a and b). The PU.1 motif was present in 47 out of 50 sequences with an E-value of 4.6e⁻²³ and is 20 nucleotides wide while the IRF8 motif was present in 20 out of 50 sequences with an E-value of 2.2e⁻⁹ and is 21 nts wide. The expression levels of deepCAGE TSSs containing the motif in their promoter sequences excluding the training data were contrasted to all other TSSs (c and d). The same comparisons were performed on promoter regions containing the TRANSFAC motif as well as for regions where the TFs bound to DNA according to ChIP-chip measurements. P-values were calculated using Student’s t-test on microarrays values.

Figure 4.

Overlapping motifs for PU.1 and IRF8 transcription factors. (a) The binding-site motifs we found for IRF8 and PU.1 were longer than the TRANSFAC motifs and both our motifs contained each of the TRANSFAC motifs as sub-motifs. Our motif for IRF8 was longer than the motifs of other IRF family members (data not shown). (b) Trimming the characteristic TTT sub-motif from either side of the IRF8 motif reduced the ability of the motif to explain the changes in expression levels. P-values were calculated using Student’s t-test.

Figure 5.

Network inferred from deepCAGE knockdown data. Our data can be transferred into network view using Cytoscape (24). The transcription factors represent the nodes and the promoters associated to their genes are the edges. Edges drawn in red indicate upregulation after TF knockdown while edges drawn in blue indicate downregulation. The dotted lines present edges that are detected by CAGE, while solid lines represent the edges that have a motif found by our method. For easier viewing, we have only shown those nodes from the training set that are influenced by more than one transcription factor.