Diverse transcription factor binding features revealed by genome-wide ChIP-seq in C. elegans

Abstract

Regulation of gene expression by sequence-specific transcription factors is central to developmental programs and depends on the binding of transcription factors with target sites in the genome. To date, most such analyses in Caenorhabditis elegans have focused on the interactions between a single transcription factor with one or a few select target genes. As part of the modENCODE Consortium, we have used chromatin immunoprecipitation coupled with high-throughput DNA sequencing (ChIP-seq) to determine the genome-wide binding sites of 22 transcription factors (ALR-1, BLMP-1, CEH-14, CEH-30, EGL-27, EGL-5, ELT-3, EOR-1, GEI-11, HLH-1, LIN-11, LIN-13, LIN-15B, LIN-39, MAB-5, MDL-1, MEP-1, PES-1, PHA-4, PQM-1, SKN-1, and UNC-130) at diverse developmental stages. For each factor we determined candidate gene targets, both coding and non-coding. The typical binding sites of almost all factors are within a few hundred nucleotides of the transcript start site. Most factors target a mixture of coding and non-coding target genes, although one factor preferentially binds to non-coding RNA genes. We built a regulatory network among the 22 factors to determine their functional relationships to each other and found that some factors appear to act preferentially as regulators and others as target genes. Examination of the binding targets of three related HOX factors--LIN-39, MAB-5, and EGL-5--indicates that these factors regulate genes involved in cellular migration, neuronal function, and vulval differentiation, consistent with their known roles in these developmental processes. Ultimately, the comprehensive mapping of transcription factor binding sites will identify features of transcriptional networks that regulate C. elegans developmental processes.

Figure 1.

Binding sites for 22 transcription factors. Chart of the total number of binding sites genome-wide, for each TF listed in alphabetical order. These binding sites are divided into those that ultimately were assigned to specific gene targets, and those that remained unassigned.

Figure 2.

Analysis of gene targets for 22 transcription factors. (A,B) Heat map showing the distribution of binding relative to the TSS (“0”) for each factor for coding (A) and non-coding (B) gene targets.

Figure 3.

Non-coding RNA genes are frequent targets of TFs. (A) Chart showing the fraction of each gene class (miRNA, snoRNA, snRNA, tRNA, and coding) bound by each TF. (B) Chart showing the relative enrichment of the fraction of each non-coding class bound by each TF relative to the fraction of coding genes bound by that TF.

Figure 4.

Correlation of GO categories between factors. The similarity between GO categories for each TF was calculated and organized into a heat map, which was clustered by TF based on this similarity.

Figure 5.

A C. elegans TF regulatory network. The relationships between TF genes as proximal targets and TF proteins as regulators. Edges represent a binding event, with the arrow pointing to the target gene. (Thick red edges) Highlight interactions discussed in the text; (green squares) TFs that bind other TF genes but are not targets themselves; (red squares) TF genes that are primarily targets and less frequently regulators; (blue squares) factors that are both targets and regulators; (squared corners) factors analyzed in embryos; (rounded corners) factors analyzed in larvae.

Figure 6.

Correlation of target gene binding by the 22 TFs. (A) Peak regions of all 22 factors (without HOT sites) were collected and binned into 100-bp windows. The signal in each bin for each factor was used in pairwise correlations as shown in the heat map. The heat map reflects the co-bound frequency (correlation coefficient) of each pair of TFs, with red indicating more co-bound regions and blue indicating less. TFs have been clustered along both axes based on similarity in their bound regions. (B) A histogram showing the number of genes targeted by PHA-4 at one, two, or more stages of development. (C) A histogram showing the number of unique PHA-4 binding sites at each stage.

Figure 7.

Common and unique targets of LIN-39, MAB-5, and EGL-5. (A) Venn diagram overlap of gene targets for LIN-39, MAB-5, and EGL-5. (B) Example of common binding sites in the targets unc-62 and kin-1. (C) Example of unique target genes for each factor. Note that the lin-39 locus exhibits increased reads across the gene in the LIN-39 and input tracks because of the extra copies of the LIN-39:GFP transgene in the genome, and therefore does not necessarily represent binding. (D) Enriched consensus DNA motifs for each set of unique target genes, with e-value.