CTCF genomic binding sites in Drosophila and the organisation of the bithorax complex

Abstract

Insulator or enhancer-blocking elements are proposed to play an important role in the regulation of transcription by preventing inappropriate enhancer/promoter interaction. The zinc-finger protein CTCF is well studied in vertebrates as an enhancer blocking factor, but Drosophila CTCF has only been characterised recently. To date only one endogenous binding location for CTCF has been identified in the Drosophila genome, the Fab-8 insulator in the Abdominal-B locus in the Bithorax complex (BX-C). We carried out chromatin immunopurification coupled with genomic microarray analysis to identify CTCF binding sites within representative regions of the Drosophila genome, including the 3-Mb Adh region, the BX-C, and the Antennapedia complex. Location of in vivo CTCF binding within these regions enabled us to construct a robust CTCF binding-site consensus sequence. CTCF binding sites identified in the BX-C map precisely to the known insulator elements Mcp, Fab-6, and Fab-8. Other CTCF binding sites correlate with boundaries of regulatory domains allowing us to locate three additional presumptive insulator elements; "Fab-2," "Fab-3," and "Fab-4." With the exception of Fab-7, our data indicate that CTCF is directly associated with all known or predicted insulators in the BX-C, suggesting that the functioning of these insulators involves a common CTCF-dependent mechanism. Comparison of the locations of the CTCF sites with characterised Polycomb target sites and histone modification provides support for the domain model of BX-C regulation.

Figure 1. Drosophila CTCF In Vivo Binding-Site Consensus

(A) Shown is the sequence logo of the CTCF consensus. (B and C) Correspondence between ChIP enrichment and closeness of match to the CTCF consensus is presented. (B) The top 600 enriched fragments are ordered by Mn (equivalent to log₂ratio, black line), and the corresponding Patser p-values are plotted (blue diamonds). The blue horizontal dotted line indicates Patser p < 10⁻¹³, which approximately marks the boundary of the p-value distribution associated with nonenriched fragments. The black vertical dotted line indicates the position in the rank order of Mn = 0.45. (C) The lowest Patser p-value for each fragment is plotted against the Mn for that fragment (blue diamonds). Open diamonds indicate fragments with p > 10⁻¹³ that are neighbours to fragments with lower p-values. The horizontal dotted line indicates Mn = 0.45. Candidate enriched fragments were selected on Mn > 0.45 and CyberT p < 0.05. The vertical dotted lines indicate the positions of Patser p = 10^−13.5 and 10⁻¹⁵. (D) PhastCons scores are shown across all 855 predicted genomic CTCF binding sites with a Patser p < 10⁻¹⁵. The binding sites are centred over position 0, and 100 bp left and right of the site are shown. The blue line indicates the median PhastCons score for a given position. There is a prominent peak corresponding to the CTCF motif. The flanking sequences show some minor fluctuations in conservation of unclear significance.

Figure 2. CTCF Binding Profile across the BX-C

The top track shows the CTCF Mn per fragment across the region. Black asterisk marks fragment UBX200. CTCF sites are sites with Patser p < 10⁻¹³, which also show significant enrichment (mean > 0.45; p < 0.05, red bars). Patser sites are positions of Patser matches to CTCF consensus with p < 10⁻¹³ (blue bars). The numbers above the blue bars relate the Patser sites to the fragments used in the validation ChIPs and EMSA (Figure 3); the sites 67 and 168 (grey) are not associated with significant enrichment. The positions 148, 100, and 65 have closely spaced double sites. The positions of the mapped insulator elements are indicated above the sequence coordinate line and RefSeq genes below. Enriched fragments correlate well with Patser sites and with the positions of three of the mapped insulators; Mcp, Fab-6, and Fab-8.

Figure 3. Validation of CTCF Sites by ChIP and In Vitro Binding

(A) ChIP was performed with chromatin from Drosophila Schneider cell line 2 (S2) cells and from embryos. Fragments are numbered according to coordinates from the original Drosophila Genome Project sequencing of the BX-C (see Figure 2) [20]. CTCF-specific antibodies (C- or N-terminal specific) immunopurify the same set of CTCF binding sites as were enriched in the ChIP-array. Negative controls were pre-immune serum or a nonbinding sequence (Fab-8 5′-control). BXC-67 and 168 show very weak, if any, enrichment. (B) In vitro binding assays (EMSA) show direct binding of CTCF to predicted CTCF sites in the BX-C. Radioactively labelled probes were incubated with GST, GST-CTCF-ZF, or GST-CTCF full length. All predicted sites are bound. A negative control sequence from the Su(var)3–9 gene does not bind to CTCF.

Figure 4. CTCF Binding Sites Demarcate Boundaries in the BX-C

The top track shows the BX-C cis-regulatory regions above a coordinate line with numbering according to Martin et al. [20]. The coloured bar indicates the regulatory domains according to Maeda and Karch [28]. The orange/yellow regions (abx/bx and bxd/pbx) control Ubx expression. The blue shaded regions (iab-2, 3, and 4) control abd-A expression, and the green regions (iab-5 to 9) regulate Abd-B. Above the Drosophila genome Release 4 coordinate line, are shown the insertion points and associated regulation of a set of enhancer-trap insertions [27]. The locations of CTCF sites are indicated by red vertical bars. The positions of the CTCF sites correlate well with the boundaries of the cis-regulatory regions and separate sets of enhancer-traps with different regulation.

Figure 5. Genomic Context of CTCF Sites in the BX-C

(A) Shown is a comparison between the locations of CTCF binding sites, Polycomb target sites, and histone H3 lysine 27 (H3K27) methylation from the data of Schwartz et al. [42]. For the Polycomb targets the Psc track is shown but the Pc and E(Z) binding profiles identify the same targets sites in this region. CTCF sites are closely related to Polycomb targets sites as illustrated by the schematic with CTCF sites in green and Polycomb sites in red. (B) Detailed view of the Mcp region shows the relationship between the CTCF site, the mapped domains of the insulator, the PRE, and the PhastCons conservation track. The CTCF site sits within the mapped insulator and lies over a clear discrete conservation peak.

Figure 6. CTCF Sites across the 3-Mb Adh Region

(A) CTCF binding profile shows the Mn (equivalent to log₂ratio) for the array fragments. Enriched fragments row plots fragments with enrichment (mean > 0.45; p < 0.05). Patser score plots the scores for Patser hits with p < 10⁻¹². CTCF binding sites depict Patser sites with p < 10⁻¹², which are associated with significant enrichment (Mn > 0.45; p < 0.05). SuHw binding sites show the Patser p < 10⁻¹⁵ Su(Hw) binding sites from (B. Adryan, G. Woerfel, I. Birch-Machin, S. Gao, M. Quick, L. Meadows, S. Russell and R. White; unpublished data). There is no clear correlation between Su(Hw) sites and CTCF sites. Neighbourhoods row depicts the gene expression neighbourhoods of Spellman and Rubin [48]; the neighbourhood boundaries in some instances map close to CTCF sites, but the overall correspondence is not compelling. (B and C) Selected regions show the arrangement of CTCF sites (marked by red bars) around the CycE gene and the osp locus.

Figure 7. Vertebrate CTCF Consensus

Shown is the sequence logo of a CTCF consensus sequence derived from sites collated in Moon et al. [13]. Both the vertebrate and the Drosophila motifs share the AGGNGGC consensus sequence, and the strong CC at positions1,2 in the vertebrate motif corresponds to the weaker CC preference at positions 4,5 in the Drosophila motif.