Mechanisms of Groucho-mediated repression revealed by genome-wide analysis of Groucho binding and activity

Abstract

Background: The transcriptional corepressor Groucho (Gro) is required for the function of many developmentally regulated DNA binding repressors, thus helping to define the gene expression profile of each cell during development. The ability of Gro to repress transcription at a distance together with its ability to oligomerize and bind to histones has led to the suggestion that Gro may spread along chromatin. However, much is unknown about the mechanism of Gro-mediated repression and about the dynamics of Gro targeting.

Results: Our chromatin immunoprecipitation sequencing analysis of temporally staged Drosophila embryos shows that Gro binds in a highly dynamic manner primarily to clusters of discrete (<1 kb) segments. Consistent with the idea that Gro may facilitate communication between silencers and promoters, Gro binding is enriched at both cis-regulatory modules, as well as within the promotors of potential target genes. While this Gro-recruitment is required for repression, our data show that it is not sufficient for repression. Integration of Gro binding data with transcriptomic analysis suggests that, contrary to what has been observed for another Gro family member, Drosophila Gro is probably a dedicated repressor. This analysis also allows us to define a set of high confidence Gro repression targets. Using publically available data regarding the physical and genetic interactions between these targets, we are able to place them in the regulatory network controlling development. Through analysis of chromatin associated pre-mRNA levels at these targets, we find that genes regulated by Gro in the embryo are enriched for characteristics of promoter proximal paused RNA polymerase II.

Conclusions: Our findings are inconsistent with a one-dimensional spreading model for long-range repression and suggest that Gro-mediated repression must be regulated at a post-recruitment step. They also show that Gro is likely a dedicated repressor that sits at a prominent highly interconnected regulatory hub in the developmental network. Furthermore, our findings suggest a role for RNA polymerase II pausing in Gro-mediated repression.

Keywords: ChIP-seq; Chromatin-associated RNA-seq, RNA polymerase II pausing; Drosophila embryogenesis; Groucho; RNA-seq; Transcriptional repression.

Fig. 1

Gro binding is highly dynamic. a Analysis of Gro binding sites: Gro ChIP-seq was carried out in duplicate on 1.5–4 h, 4–6.5 h, and 6.5–9 h embryo collections. Putative binding sites (ChIP-seq peaks) were identified as described in Materials and Methods. The Venn diagram indicates overlap between binding sites at the three time points analyzed. b Clustering of Gro binding sites by temporal pattern. The majority of Gro binding sites are unique to a single time point, while many were observed at all three time points. Although a substantial fraction of sites overlap between time points 2 and 3, very few overlap between time points 1 and 2. A small number of sites (38) are bound in only time points 1 and 3 without being bound in 2, indicating that loss of Gro from a locus tends to be a permanent regulatory decision. c Analysis of Gro-bound genes: Each Gro binding site was assigned to the closest gene. The Venn diagram indicates overlap between Gro bound genes at the three time points analyzed. d Distribution of the number of Gro binding sites per gene: About 45% of all Gro-bound genes exhibit two or more distinct Gro binding sites peaks, a fraction that is greater than that expected by chance (p < 10⁻¹⁰ by Monte Carlo simulation)

Fig. 2

The pattern of Groucho recruitment to chromatin differs from that of the corepressor dCtBP. a Distribution of Gro and dCtBP binding site size. As indicated by a box plot, Groucho sites exhibit median widths of between 500 and 750 basepairs at the three time points sampled, although significantly larger peaks were also identified. dCtBP binding sites exhibits a median width of about 1000 bp. Thus, the difference between the short-range repression mediated by dCtBP and the long-range repression mediated by Gro cannot be attributed to one-dimensional spreading along chromatin. b Distribution of Gro and dCtBP binding sites relative to the nearest transcriptional start site (TSS). Gro ChIP-seq read density is enriched around TSS’s compared to dCtBP, which exhibits a small asymmetric depletion in these areas. c Distribution of Gro and dCtBP binding sites with respect to gene feature. Compared to dCtBP, Gro more often binds in intergenic regions, and therefore less frequently within either intronic or exonic regions

Fig. 3

Gro binding sites are enriched for binding motifs of multiple sequence-specific transcription factors. The four most significantly enriched sites as identified by de novo motif discovery (DREME) are shown for each timepoint. For each motif, the factor with the most similar binding site is listed. In cases where the discovered motif corresponds with similar likelihoods to binding sites for multiple factors, all potential factors are listed. a Motifs enriched in 1.5–4 h Gro binding sites. b Motifs enriched in 4–6.5 h Gro binding sites. c Motifs enriched in 6.5–9 h Gro binding sites

Fig. 4

Groucho is recruited to both VRRs and VARs. a Genome browser views showing Gro ChIP-seq signal in genomic regions containing the two Dorsal repression targets zen and dpp. The positions of Dorsal binding site-containing VRRs in each gene are indicated. b Genome browser views showing Gro ChIP-seq signal in genomic regions containing the two Dorsal activation targets sna and rho. Dorsal binds and activates sna through a primary enhancer and a secondary (shadow) enhancer [65, 66]. Dorsal binds and activates rho through its neuroectodermal enhancer (nee). c Dorsal binds with high frequency to all three classes of Dorsal binding sites. Sites are categorized as described previously [69]. See text for a discussion of the roles of the three types of sites

Fig. 5

Integrating binding data (ChIP-seq) with expression data (RNA-seq) to identify high confidence Gro targets. A score corresponding to the extent of Gro occupancy within genes and adjacent areas was calculated for each gene using a previously published algorithm [35]. The algorithm was adjusted to allow for increased score contribution from regions binding more distantly from the target gene (see Methods). Plotted for each time point are the number of genes either down-regulated (top) or up-regulated (bottom) upon overexpression of Gro (vertical axis) out of the total number of genes meeting a score cutoff of decreasing stringency (horizontal axis). As the threshold Gro occupancy score decreases from left to right the number of genes that exceed this threshold (indicated by the horizontal axis labels) increases from left to right. Where a change in slope is clearly evident, the score cutoff selected for the high-confidence set of Groucho targets is indicated (circles). These are the data obtained using Gro overexpression line A. The “Experimental” curves show the data obtained using the experimentally determined up and down-regulated genes. The “Randomized” curves were obtained by generating random gene sets of the same size as the gene set used for the corresponding “Experimental” curve. Each “Randomized” curve is the average of the results obtained with 100 such random gene sets

Fig. 6

Gro target genes form a highly-interconnected network with multiple hubs. Genes that are differentially significantly down-regulated upon Gro overexpression and that exceed the ChIP-seq score cutoffs indicated by the circles in Fig. 5 define a set of 187 high-confidence targets (the full list is provided in Additional file 6: Table S3A). a The most significantly enriched (FDR < 0.01) gene ontology groups of high-confidence Gro target genes are uniformly related to transcriptional (red bars) and developmental (black bars) regulation, confirming the role of Gro as a high-level regulatory node in the establishment of tissue fate during development. The plot indicates fold-enrichment of the indicated groups relative to random. The numbers next to the bars indicate the numbers of genes that fall into the indicated groups, and the bars are ordered according to this number with the group containing the most genes at the top. b Potential Groucho-target genes were integrated into a network analysis to visualize genetic and physical interactions of these target genes. Genetic (blue edges) and physical (orange edges) interactions were obtained from a curated set maintained by FlyMine [72]. The target gene set results in highly-connected networks with multiple hubs (8 or more edges, yellow nodes) interconnected by multiple genetic interactions

Fig. 7

Gro regulated genes are enriched for stalled Pol II. Published data classifying all Drosophila genes into four categories of Pol II enrichment or depletion in 2–4 h embryos were used to classify all Groucho-regulated genes at each timepoint [77]. a Predicted Gro-regulated genes are enriched for genes classified as possessing stalled Pol II and depleted for genes possessing actively elongating Pol II. b Chromatin-associated transcript density across all expressed genes was calculated independently for different sets of genes at three time points. At each time window, genes predicted to be Gro regulatory targets are enriched for 5’ proximal transcript density, suggesting that these genes are enriched for stalled Pol II