Evidence for autoregulation and cell signaling pathway regulation from genome-wide binding of the Drosophila retinoblastoma protein

Abstract

The retinoblastoma (RB) tumor suppressor protein is a transcriptional cofactor with essential roles in cell cycle and development. Physical and functional targets of RB and its paralogs p107/p130 have been studied largely in cultured cells, but the full biological context of this family of proteins' activities will likely be revealed only in whole organismal studies. To identify direct targets of the major Drosophila RB counterpart in a developmental context, we carried out ChIP-Seq analysis of Rbf1 in the embryo. The association of the protein with promoters is developmentally controlled; early promoter access is globally inhibited, whereas later in development Rbf1 is found to associate with promoter-proximal regions of approximately 2000 genes. In addition to conserved cell-cycle-related genes, a wholly unexpected finding was that Rbf1 targets many components of the insulin, Hippo, JAK/STAT, Notch, and other conserved signaling pathways. Rbf1 may thus directly affect output of these essential growth-control and differentiation pathways by regulation of expression of receptors, kinases and downstream effectors. Rbf1 was also found to target multiple levels of its own regulatory hierarchy. Bioinformatic analysis indicates that different classes of genes exhibit distinct constellations of motifs associated with the Rbf1-bound regions, suggesting that the context of Rbf1 recruitment may vary within the Rbf1 regulon. Many of these targeted genes are bound by Rbf1 homologs in human cells, indicating that a conserved role of RB proteins may be to adjust the set point of interlinked signaling networks essential for growth and development.

Keywords: Drosophila; Rbf1; cell-cycle; retinoblastoma.

Figure 1

Rbf1 exhibits dynamic promoter occupancy. Rbf1 occupancy of regulated promoters measured by ChIP was low in 0- to 6-hr embryos and peaked at 12 to 18 hr. Formaldehyde cross-linked chromatin was prepared from embryos of different ages and immunoprecipitated using the indicated antibodies. No specific enrichment was found at a nontarget gene promoter (sloppy paired 1). No Ab indicates immunoprecipitation carried out without antibody; IgG, nonspecific mouse polyclonal antibodies; α-H3, antihistone H3 antibody; and α-Rbf1, rabbit anti-Rbf1 antibody.

Figure 2

Rbf1 promoter-proximal occupancy of diverse classes of genes suggests autoregulatory effects. Strong peaks were noted on cell-cycle related genes, such as DNApol-α60 and cyclin-dependent kinase 30A (A, B). Autoregulation is suggested by occupancy of the Rbf gene (C). The 5′ region of the Rbf1-related corepressor l(3)mbt is also associated with Rbf1 (D). The promoter of the dNURF 301/E(bx) gene, a chromatin remodeling component important for RB function in development, is also bound (E). Numerous components of cell signaling pathways, including the insulin receptor InR, also are targeted by Rbf1 (F). Relative peak intensities are shown on the Y-axis. Representative individual peaks are visualized on the UCSC genome browser. Bent arrows indicate the direction of transcription of the genes and absence of arrows indicates 3′ region of a gene.

Figure 3

Rbf1 occupies multiple nodes in conserved signaling pathways. Genes in the insulin (A), JAK/STAT (B), Notch (C), and Hippo/Warts/Yorkie (D) signaling pathways are targeted at multiple levels by Rbf1. Proteins of genes targeted by Rbf1 are indicated in red.

Figure 4

Rbf1 exhibits a strong promoter-proximal targeting bias. (A) The distribution of peaks relative to the nearest TSS. The majority of peaks are centered 205 bp 5′ of the TSS. Distances were grouped into 100-bp bins and points fitted with a smooth curve. (B) Distribution of peak intensities. Most peaks had an intensity within a few fold of the average, although some peaks were >10-fold greater.

Figure 5

Rbf1 target genes represent diverse GO categories. A total of 42% of 1890 Rbf1 target genes were enriched for GO terms. Of these, only approximately one-quarter were associated with Cell Cycle and DNA replication, whereas the majority of targets grouped into other gene regulatory and developmental processes. The GO terms are arranged in the pie chart in decreasing order of significance of enrichment from Cell Cycle and DNA replication to Other DNA/RNA Metabolic Process. “Others” indicates numerous smaller groups of enriched genes. The P values for the categories are as follows: Cell Cycle, 7E-16, Cytoskeleton, 5E-13; Chromatin Modification, 2E-12, Recombination/Repair, 7E-09; Phagocytosis, 1E-03; Apoptosis, 5E-03; Transcription, 2E-02; Neurogenesis, 2E-02; and other DNA/RNA Metabolic Process, 4E-02.

Figure 6

Transcription factor motifs enriched in Rbf1-bound peaks. (A) The four most overrepresented motifs identified by the MEME motif discovery tool, including one previously unknown motif (RAM). (B) Rbf1-associated motifs are highly enriched compared with average occurrence in DNA of the same A/T composition. The sequences under Rbf1 peaks were scrambled five times, and specific motifs with P < 0.0001 were identified. E2F sites showed the greatest level of enrichment in specifically bound regions compared with randomized DNA sequences. (C) E2F, DREF, and RAM motifs preferentially associate with Rbf1-bound promoters. The presence of motifs in Rbf1-bound sequences was compared with the presence in Rbf1-unbound promoters. FOXJ2 sites are not restricted to Rbf1-associated promoters and may represent a motif for a broadly acting factor. Note that the canonical DREF sites are 8-mers (Yamaguchi et al. 1995). In our data, the eighth nucleotide was not conserved. (D) Diversity of motif composition of peaks. A total of 42% of total peaks contained only one of the four different motifs (E2F, DREF, FOXJ2, or RAM). A quarter of the peaks had a combination of two different motifs; 6%, a combination of three, and 1% contained all four motifs. Only 36% or peaks had an identifiable E2F motif (small insert). Strikingly, a quarter of the peaks did not have any of the four motifs. The heterogeneity of sequences in Rbf1-bound peaks suggests that E2F may not be the only transcription factor that recruits Rbf1 to target gene promoters. Peaks used in this analysis were drawn from the 1236 bound regions found in both Rbf1 ChIP biological experiments. A peak with multiple E2F motifs, but no other motif types, was counted as one type of motif; a similar treatment applies for the other three motifs. “None” means the peaks did not contain any motifs for E2F, DREF, FOXJ2, or RAM.

Figure 7

Enrichment of Rbf1-associated motifs indicates distinct promoter subclasses in the Rbf1 regulon. (A) Heat map for association of motifs with different GO categories. E2F sites were present in a significant fraction of bound regions as a whole, especially in GO categories Phagocytosis, Chromatin Modification, and Cell Cycle. Genes involved in Neurogenesis and Oogenesis tend to be depleted of E2F motifs. RAM sites tend to occur on a subset of E2F-containing sequences, but Chromatin Modification and Nucleotide Metabolic Process genes are depleted of RAM motifs. Apoptosis and Transcription/Translation genes are associated with DREF motifs. Chromatin Modification, Phagocytosis, and Oogenesis genes are depleted of FOXJ2 motifs, whereas a larger fraction of Neurogenesis and Oogenesis genes lack all of the four motifs. (B) Heat map for association of motifs with selected functional and physical targets of Rbf1, including functional targets of Rbf in S2 cells (Rbf); functional and physical targets of the Rbf- and Myb- containing dREAM complex; functional targets of l(3)mbt, a corepressor and a binding partner of Rbf; and physical targets of Rbf1 identified in this study that are involved in signaling pathways. The E2F motif alone is preferentially associated with Rbf1 functional targets in S2 cells (Rbf) and dREAM functional targets, however, a combination of E2F and RAM sites is found preferentially on l(3)mbt functional and dREAM physical targets. Signaling pathway Rbf1 target genes are depleted of all four motifs, suggesting a distinct promoter signature. “Percent” indicates fraction of genes in a selected GO category containing at least one occurrence of the indicated motif within the Rbf1 peak. The category “all” represents all 1236 peaks present in both ChIP replicates. The category “other” represents small clusters of genes found to be overrepresented in numerous GO categories (see Figure 5). “Signaling pathway” represents 136 genes found in diverse conserved Drosophila signaling networks (Table S3). Numerical values for heat map shown here are found in Table S5.

Figure 8

Divergence and conservation of RB regulon. A total of 2310 identifiable human orthologs of Rbf1 targets were compared with RB targets in growing (A), quiescent (B), and senescent (C) cells and p130 targets in quiescent (D) and senescent (E) cells. The overlaps in (A), (B), and (C) were further compared with each other (F) and the overlaps in (D) and (E) were compared with each other (G). Comparison of the total genes in (F) and (G) with each other (H), shows that most of the targets of Rbf1 and RB, as well as Rbf1 and p130, are the same. In all overlaps in A-E, GO terms DNA Replication, Cell Cycle, DNA Repair, and Chromatin Modification were enriched, indicating that these genes may form a conserved ancient regulon of RB proteins. The overlap of human homologs of Rbf1 targets in Drosophila embryos and RB and p130 targets in human cell culture suggests that many genes have retained regulation by RB proteins since divergence of these organisms. Other categories of genes may represent divergence of RB family function, or context-dependent differences in binding. Human homologs for Rbf1 targets were compared with published RB and p130 targets in growing, quiescent and senescent human lung fibroblasts (Chicas et al. 2010).