A Variably Occupied CTCF Binding Site in the Ultrabithorax Gene in the Drosophila Bithorax Complex

Abstract

Although the majority of genomic binding sites for the insulator protein CCCTC-binding factor (CTCF) are constitutively occupied, a subset show variable occupancy. Such variable sites provide an opportunity to assess context-specific CTCF functions in gene regulation. Here, we have identified a variably occupied CTCF site in the Drosophila Ultrabithorax (Ubx) gene. This site is occupied in tissues where Ubx is active (third thoracic leg imaginal disc) but is not bound in tissues where the Ubx gene is repressed (first thoracic leg imaginal disc). Using chromatin conformation capture, we show that this site preferentially interacts with the Ubx promoter region in the active state. The site lies close to Ubx enhancer elements and is also close to the locations of several gypsy transposon insertions that disrupt Ubx expression, leading to the bx mutant phenotype. gypsy insertions carry the Su(Hw)-dependent gypsy insulator and were found to affect both CTCF binding at the variable site and the chromatin topology. This suggests that insertion of the gypsy insulator in this region interferes with CTCF function and supports a model for the normal function of the variable CTCF site as a chromatin loop facilitator, promoting interaction between Ubx enhancers and the Ubx transcription start site.

INTRODUCTION

There is considerable evidence indicating a major role for the CCCTC-binding factor (CTCF) in genome organization (reviewed in references 1 and 2). CTCF binds to insulator elements and is required for their function in blocking interactions between enhancers and promoters (3). It has been shown to be involved in the formation of chromatin loops (4), and CTCF binding is enriched at the boundaries of topological chromatin domains (5–8). However, it is remains to be determined how much of CTCF function is linked to a specifically architectural role in genome organization and how much is more directly involved in the control of gene expression.

CTCF was originally identified as a transcription factor (9). Subsequent genome-wide mapping of CTCF binding revealed that 20% of binding sites are within 2.5 kb upstream from transcription start sites (10) and that CTCF sites are enriched at gene promoters (11, 12). A current unifying hypothesis is that the molecular function of CTCF is to mediate chromosomal loop formation and that this may give rise to a variety of context-dependent roles; in some contexts, loop formation may serve an architectural purpose, and in others, it may be more intimately associated with gene regulation. One way to partition CTCF binding sites into possible functional classes is to differentiate between sites that are constantly occupied and sites that show variable occupancy. The first comparisons between whole-genome maps of CTCF binding in different cell lines indicated that the majority of sites are constitutively bound (10, 13, 14). However, more recent studies have revealed higher proportions of variable sites (15, 16) and, interestingly, the variable sites are preferentially associated with enhancers (12). However, very few individual variable CTCF sites have yet been analyzed, and more examples are required to build an understanding of their association with gene regulation.

The classical example of a variable CTCF site is at the imprinted control region (ICR) of the mammalian insulin-like growth factor 2 gene (Igf2)/H19 locus, where CTCF binding is regulated by DNA methylation of the binding sites. On the maternal chromosome, CTCF binds the unmethylated ICR and the enhancer-blocking action of CTCF prevents Igf2 expression. However, on the paternal chromosome, methylation of the ICR prevents CTCF binding and the lack of insulator function enables Igf2 expression (17–20). A second example involves a CTCF site in the chicken lysozyme locus, where CTCF binding is regulated by the chromatin structure. Activation of the lysozyme gene is linked to eviction of CTCF, and this is mediated through the transcription of a noncoding RNA, chromosome remodeling, and repositioning of a nucleosome over the CTCF binding site (21). Recently, in Drosophila, Wood et al. provided evidence for two classes of regulated insulator (22). In one class, the occupancy of DNA-binding insulator proteins [e.g., BEAF, CTCF, and Su(Hw)] at insulator sites is regulated. In a second class, the DNA-binding insulator proteins are constitutively bound, but the insulators are regulated by the variable recruitment of other components (e.g., CP190) required to build a functional insulator complex.

Here, we present an analysis of a variably occupied CTCF site in the Drosophila Bithorax complex (BX-C). The BX-C contains three Hox genes, Ultrabithorax (Ubx), abdominal A (abd-A), and Abdominal B (Abd-B) and has a clear regulatory domain structure with independent regulatory elements controlling gene expression in the parasegmental (PS) units along the anteroposterior axis of the developing embryo (reviewed in reference 23). The regulatory domains are separated by boundaries that constrain the activation of PS-specific enhancers. Genetic deletion of boundaries leads to inappropriate enhancer activation and ectopic expression of Hox genes. CTCF binding is associated with BX-C boundaries, and CTCF mutations cause misexpression of Abd-B (24–26). The CTCF binding at boundary elements appears to be constitutive, and this may fit with an architectural role for these sites. Here, we report the identification of a variable CTCF site within the Ubx gene that preferentially binds CTCF when the Ubx gene is active and is associated with different chromatin topologies in active and inactive states. We present a model where CTCF has a role in facilitating the interaction between Ubx enhancers and the Ubx promoter.

MATERIALS AND METHODS

Fly lines.

The wild-type Drosophila melanogaster strain Oregon R was used in the chromatin immunoprecipitation (ChIP)-array, ChIP-quantitative PCR (qPCR), and chromosome conformation capture (3C) experiments. In addition, homozygous bx^83Ka mutants (27) from the bx^83Ka/TM6B strain were used in ChIP-PCR and 3C experiments.

Antibodies.

The following antibodies were used in the ChIP experiments: anti-CTCF-C antiserum (24), anti-CP190 antiserum (28), anti-RNA polymerase II (RNA Pol II) (0.9 mg/ml affinity-purified IgG, ab5131; Abcam), and anti-GAGA factor antibody (0.2 mg/ml IgG, SC-98263; Santa Cruz Biotechnology).

Chromatin preparation.

Dissected head segments of late 3rd instar larvae were inverted and fixed with 2% formaldehyde in phosphate-buffered saline (PBS) for 20 min at room temperature. These were washed with twice with PBS–125 mM glycine–0.01% Triton X-100, followed by a single wash with PBS and then with PBS containing 1% protease inhibitor cocktail (catalog number P8340; Sigma). The T1 and T3 leg imaginal discs were then dissected, snap-frozen in liquid nitrogen, and stored at −80°C prior to use. Approximately 150 leg discs were combined in PBS–0.01% Triton X-100 and centrifuged in a microcentrifuge at 1,200 rpm for 1 min. The discs were resuspended in 20 μl cell lysis buffer {5 mM PIPES [piperazine-N,N′-bis(2-ethanesulfonic acid)], pH 8, 85 mM KCl, 0.5% NP-40} containing 1% protease inhibitor cocktail and homogenized using a motorized pestle at 2-min intervals for 8 min. After a brief microcentrifuge centrifugation (13,200 rpm for 10 s), the pellet was resuspended in 300 μl nuclear lysis buffer (50 mM Tris-HCl, pH 8.1, 10 mM EDTA-Na₂, 1% SDS) with protease inhibitors and incubated for 20 min at room temperature. The extracts were sonicated in a Bioruptor standard device (Diagenode) at the high setting for 4 min 15 s (cycles of 30 s on, 30 s off), producing 0.5- to 3.0-kb fragments. One hundred-milliliter aliquots of chromatin extracts were flash-frozen in liquid nitrogen and then stored at −80°C prior to use.

ChIP.

ChIP was performed as described by Birch-Machin et al. (29). One hundred-milliliter aliquots of chromatin were precleared with 13 μl blocked Staphylococcus aureus cells (SAC) and mixed with 200 μl of IP dilution buffer (16.7 mM Tris-HCl, pH 8, 167 mM NaCl, 1% EDTA, 1.1% Triton X-100, 0.01% SDS) with protease inhibitors. Two microliters of antibody was added, and the mixture was incubated on a roller overnight at 4°C. Then, 13 μl of SAC was added to each IP reaction mixture and the samples were incubated for 35 min at 4°C on a roller. The mixture was centrifuged in a microcentrifuge at 13,200 rpm at room temperature, and the pellets were washed successively with 1 ml each of low-salt buffer (0.1% SDS, 1% Triton X-100, 2 mM EDTA-Na₂, pH 8.0, 20 mM Tris-HCl, pH 8, 150 mM NaCl), high-salt buffer (0.1% SDS, 1% Triton X-100, 2 mM EDTA-Na₂, pH 8, 20 mM Tris-HCl, pH 8, 500 mM NaCl), and LiCl buffer (0.25 M LiCl, 1% NP-40, 1% sodium deoxycholate, 1 mM EDTA-Na₂, pH 8.0, 10 mM Tris-HCl, pH 8.0) and twice with Tris-EDTA (TE) buffer, pH 8.0, for 5 min at 4°C on a roller for each solution. The immune-precipitated chromatin was then eluted twice from the SAC pellet with 300 μl of IP elution buffer (50 mM NaHCO₃, 1% SDS) by vigorously vortexing for 15 min at room temperature. One microliter of RNase A (catalog number R4642; Sigma) and 24.3 μl of 4 M NaCl (0.3 M final concentration) were then added to the eluate, and the mixture was incubated for 4 h at 65°C to reverse the cross-linking. The DNA was then precipitated by adding 812 μl of 100% ethanol and incubating overnight at −20°C. The samples were centrifuged in a microcentrifuge at 4°C for 20 min, and the pellets were air dried for 1 h at room temperature. The pellets were resuspended in 100 μl TE buffer, followed by the addition of 25 μl of 5× PK buffer (50 mM Tris-HCl, pH 7.5, 25 mM EDTA-Na₂, pH 8, 1.25% SDS) and 1.5 μl of 20 mg/ml proteinase K, incubated at 45°C for 2 h, and purified using the QIAquick PCR purification kit (catalog number 28104; Qiagen). The DNA was eluted in 30 μl of buffer EB and stored at −20°C until use.

CTCF ChIP-array.

Five microliters of CTCF ChIP and 5 μl of control ChIP DNA from T1 and T3 leg discs obtained from Oregon R larvae were amplified using the GenomePlex single-cell whole-genome amplification kit (product number WGA4; Sigma-Aldrich) according to the manufacturer's instructions. The samples were amplified for 21 cycles, and the amplified DNA purified using the QIAquick PCR purification kit. One microgram each of amplified ChIP and control DNA were labeled with Cy5 and Cy3 in the presence of Cy3- or Cy5-dCTP (GE Healthcare) using the BioPrime DNA labeling kit (Invitrogen) and hybridized onto Nimblegen ChIP-chip (ChIP with microarray technology) 2.1 M probe whole-genome tiling arrays according to the manufacturer's instructions.

Microarray data processing.

Two biological replicates were prepared for each sample with a Cy3/Cy5 dye swap for one biological replicate of each sample. ChIP DNA prepared with preimmune serum was used as the reference control to assay ChIP enrichment in the array experiments. Arrays were scanned and processed as previously described (30). The enrichment profiles were visualized using the Integrated Genome Browser (http://bioviz.org/igb/index.html). Patser position-specific weight matrix analysis was as described previously (24). Analysis of conservation used the PhastCons multiple alignment data available from http://genome.ucsc.edu.

Quantitative PCR.

Quantitative real-time PCR experiments were performed with a LightCycler 480 II (Roche Diagnostics) in 10-μl reaction mixtures using SYBR green PCR master mix (catalog number 04707516001; Roche). Each reaction mixture consisted of 5 μl SYBR green PCR master mix, 3 μl water, 1 μl 10 μM primer mix, and 1 μl DNA. Amplification was carried using the following conditions: 1 cycle at 95°C for 15 min and 45 cycles of 95°C for 10 s, 58°C for 10 s, and 72°C for 10 s. The primer pairs used for the amplification are listed in Table 1. Serial dilutions of Drosophila genomic DNA (100 to 0.01 ng/μl) were used as standards for quantification.

TABLE 1.

ChIP-qPCR primers

IDa	Primer	Chr	Directionb	Position		Sequence
				Start	End
0	Neg	3R	F	12526683	12526702	CCTAAATGGCAGAGGATTGG
			R	12526792	12526773	AAATTCAGGATGCAGGATGC
1	R1	3R	F	12528866	12528885	ATCAGCAGCCGTTGAGTAGG
			R	12528971	12528952	ATTCCTCAGCGACAAAGAGC
2	R2	3R	F	12529660	12529679	GAGTTGCCATAAAGCACTCG
			R	12529764	12529745	TTCTCTTCGCAGCCTATTCC
3	R3	3R	F	12529861	12529880	TTACAGCCGACACCTCATCA
			R	12529987	12529968	CTGGCTTGACACTGGGCTAC
4	R4	3R	F	12530745	12530769	CTCGCTGGTTCCTAATATGATATAC
			R	12530863	12530846	GTGCCTTTCGGTGACTTC
5	R5	3R	F	12531112	12531129	GCACAGATTCCGTTGAGC
			R	12531253	12531234	CCTTCTATGCTCTGCTCTCG
+ve	BXC-49	3R	F	12760726	12760707	ATCGATAAAAAGCGCCAACA
			R	12760565	12760584	GCTCTTACTGCCCGATTCTG
−ve	SuVar 3-9	3R	F	11087377	11087396	AGCCGCTACTATTGCTTGGA
			R	11087573	11087554	GCAGCGACAGCAGTATGAAA
Ubx-P	F-675	3R	F	12559800	12559819	AATACTTGGATTGCGCTTGC
			R	12560001	12559982	TTTCCACTAGATTGGCGTCC

^aID, identifiers used in Fig. 1, 2, and 6.

^bF, forward; R, reverse.

Preparation of 3C DNA from T1 and T3 leg discs.

Approximately 450 each of the T1 and T3 leg discs from 3rd instar larvae were dissected and frozen as described above. The discs were thawed on ice and transferred to a 1.5-ml microcentrifuge tube. The pooled discs were briefly centrifuged at 13,200 rpm for 10 s. The excess liquid was discarded and the discs were resuspended in 20 μl lysis buffer (31) containing 10 mM Tris-Cl, pH 8.0, 10 mM NaCl, 0.2% Igepal CA360 (catalog number I8896; Sigma), and 10 μl/ml of protease inhibitor (Sigma). The discs were homogenized with a plastic motorized pestle at 2-min intervals for a total of 8 min. After a brief centrifugation, 500 μl of lysis buffer with 50 μl of protease inhibitor was added to the homogenate, and the suspension was centrifuged at 5,000 rpm for 5 min at room temperature.

The 3C DNA was prepared based on the protocol described by Hagège et al. (32). The leg disc lysate pellet was washed twice with ice-cold 1.2× NEBuffer 3 (catalog number B7003S; New England BioLabs) at 5,000 rpm for 5 min at room temperature. The pellet was then resuspended in 500 μl 1.2× NEBuffer 3 and 7.5 μl 20% SDS. The mixture was incubated at 37°C, 900 rpm for 1 h in a Thermomixer (catalog number 5355000038; Eppendorf). Then, 50 μl 20% Triton X-100 was added and the mixture further incubated at 37°C, 900 rpm for 1 h. The lysate was then digested with 400 U of DpnII at 37°C, 900 rpm overnight. The enzyme was inactivated by heat treatment at 65°C for 20 min, and the mixture was ligated at 16°C for 16 h in a 10-ml reaction mixture with 10,000 U of T4 DNA ligase (New England BioLabs). The ligated chromatin digest was then de-cross-linked and purified as described by Hagège et al. (32). The purified 3C DNA was resuspended in 50 μl TLE buffer (10 mM Tris-Cl, pH 8.0, 0.1 mM EDTA), and the DNA concentration was measured by using the Qubit dsDNA HS assay kit (catalog number Q32854; Invitrogen). 3C DNA samples were stored at −20° until use.

PCR amplification of 3C DNA.

3C interactions were determined according to the protocol of Dekker et al. (33). To investigate the chromatin conformation and interactions in the Ubx region in T1 and T3 leg discs, 29 primers spanning chromosome (Chr) 3R:12400341.0.12695484 were designed based on the expected fragments generated by DpnII digestion (Table 2). In addition, primer pairs located in DpnII fragments containing the CTCF differential peak in Ubx, the Ubx promoter, and the Mcp region were also designed to serve as anchor fragment internal primers (Table 3).

TABLE 2.

3C PCR primers

ID	Primer	Chr	Position		Sequence
			Start	End
1	223	3R	12400341	12400360	GCGAGACGATAAACGACGAC
2	237	3R	12412997	12413016	AAGAAGTGGTAAAGTGGCGG
3	372	3R	12444906	12444925	CTGTGCATCTCCACCACATC
4	396	3R	12449306	12449325	CAGAAGCTGCCTCTCGTAGG
5	444	3R	12465581	12465600	CAAAGCCACCTTCCTGAAAC
6	478	3R	12474725	12474744	ATCTCGCCCAGCACTATTTG
7	504	3R	12480871	12480890	TTTGAGTGGGTTAAGCTGCC
8	559	3R	12508313	12508332	TAAATACGAAGTGCATGCGG
9	589	3R	12529861	12529880	TTACAGCCGACACCTCATCA
10	590	3R	12530474	12530494	GGAACACGCATATAGCATTGG
11	636	3R	12549178	12549196	TTTGAAATGCAAACACGGC
12	674	3R	12559159	12559178	GGAGGCCTGTTCAAAGTACG
13	675	3R	12559351	12559332	CAAAGGAGGCAAAGGAACAG
14	677	3R	12561570	12561589	CGAGAAGACCCAGAGCAAAG
15	698	3R	12574489	12574509	AAGAAATATGCGTTTCCCACC
16	699	3R	12575770	12575788	CGCCAGACAATGGAAACTG
17	745	3R	12592412	12592433	GTGCTATCAACTCGCTTTCTTG
18	751	3R	12593896	12593915	CTCTTTGTTAGCGGAGGCAG
19	789	3R	12608923	12608942	TAAGCGAGTGCGTGTCATTC
20	842	3R	12625282	12625303	TCATCTGGAACTGGTTCTATCG
21	858	3R	12633588	12633607	AATCCGGTTGTGAAACAAGG
22	875	3R	12640691	12640710	TCAGTCTCACAGCCATTTCG
23	899	3R	12649777	12649797	GCATGTGCATTTAAGGAGTGG
24	918	3R	12657009	12657031	CCAGTTAATGTGCTTCCTACCTG
25	918	3R	12657020	12657043	GCTTCCTACCTGTCTATTTGTTGG
26	919	3R	12658026	12658046	GTGTCGAGTTTCGGTTGAGTC
27	923	3R	12660715	12660734	AAATGTTTGGACGGGAAATG
30	961	3R	12683796	12683817	GCTTTAACTTTAACCTCTGGCG
31	983	3R	12695484	12695507	CTGCTCTGCTTATCAGTTTATTGG

TABLE 3.

Anchor fragment internal primers for 3C PCR

Anchor position	Fragment ID	Chr	Directiona	Position		Sequence
				Start	End
Ubx promoter	675	3R	F	12559800	12559819	AATACTTGGATTGCGCTTGC
			R	12560001	12559982	TTTCCACTAGATTGGCGTCC
Variable CTCF site_1	589	3R	F	12529861	12529880	TTACAGCCGACACCTCATCA
			R	12529987	12529968	CTGGCTTGACACTGGGCTAC
Variable CTCF site_2	590	3R	F	12530221	12530240	AGGGTTAATTCGTTCATCGC
			R	12530362	12530343	CTGATGATGACGCTGTTGTG
Mcp	983	3R	F	12694755	12694774	ATTGTATGTATCCGCTCCGC
			R	12694917	12694898	AAGCCCTTATTTGCAGACCC

^aF, forward; R, reverse.

For each anchor fragment investigated, individual 10 μM primer mixes composed of the anchor fragment internal primers and individual anchor primer/target primer pairs were prepared. The 3C PCRs were carried out in a 25-μl mixture using the Thermo-Start Taq DNA polymerase kit (product number AB-1057; Thermo Scientific). Each reaction mixture contained 18.3 μl water, 2.5 μl 10× PCR buffer, 1.5 μl 25 mM MgCl₂, 0.5 μl 10 mM deoxynucleoside triphosphate (dNTP) mix, 0.2 μl Taq DNA polymerase, 1 μl 10 μM primer mix, and 1 μl (1 ng/μl) of 3C DNA sample. Amplification was carried out in an iCycler 582BR thermal cycler (Bio-Rad) using a touchdown protocol with 1 cycle at 95°C for 15 min and then 10 cycles at 95°C for 30 s, annealing from 69 to 59°C for 30 s, and 72°C for 30 s. This was followed by 30 cycles at 95°C for 30 s, 59°C for 30 s, and 72°C for 30 s, followed by a final extension at 72°C for 10 min. PCR products were then subjected to electrophoresis on a 2% agarose gel in 0.5× Tris-borate-EDTA (TBE).

Quantification of 3C PCR products.

Gel images were digitized, and the bands were quantified using ImageJ software (http://imagej.nih.gov/ij). The relative interaction between the different primer pairs was then expressed as the ratio of the signal strength between the anchor/target 3C PCR product and the anchor fragment PCR product. The relative interactions between the 3C primer pairs and each specific anchor fragment were plotted to visualize interactions.

Microarray data accession number.

The ChIP-array data have been submitted to GEO under accession number GSE62234.

RESULTS

Identification of a variably occupied CTCF site in the Ubx gene.

The individual Hox genes of the BX-C are expressed in different segments along the anteroposterior axis (23), presenting a useful experimental system for the isolation of in vivo tissues with different states of gene expression in sufficient quantities for genomic analysis. Here, we have used the imaginal discs from Drosophila larvae to compare the genome-wide CTCF binding profile in leg imaginal discs from the 1st thoracic segment (T1) to that of leg discs from the 3rd thoracic segment (T3). The Hox gene Ubx is not expressed in T1 but is active in T3. The other two genes of the BX-C, abd-A and Abd-B, are inactive in both T1 and T3. The activity state of these BX-C genes is regulated by Polycomb (Pc) silencing, which imposes a repressive chromatin state on inactive genes. Comparing the CTCF ChIP-array profiles of the T1 leg disc and the T3 leg disc, we find that the profiles are generally extremely similar, with very few clear differential peaks. However, we identify a clear differential CTCF binding peak in the Ubx gene (Fig. 1A). There is strong CTCF binding at this position in the T3 leg disc, where Ubx is expressed, but we find little binding at this site in the T1 leg disc, where the Ubx gene is repressed. In contrast, the binding of CTCF in the repressed abd-A and Abd-B regions is very similar in both discs.

FIG 1.

A variably occupied CTCF site in the Ubx gene. (A) CTCF binding profiles from T1 (Ubx inactive) and T3 (Ubx active) leg imaginal discs. The arrow indicates the variably occupied CTCF site. Ubx, abd-A, and Abd-B are transcribed from right to left. (B) The CTCF ChIP peak aligns with a match to the CTCF position-specific weight matrix. The positions of the PCR primers used in the PCR whose results are reported in panel D are shown. (C) PhastCons conservation plot across 15 insect species (http://genome.ucsc.edu). The sequence at the variable CTCF site is compared with the Drosophila consensus (red) (36). The conserved CC motif (34) and conserved T in module number 1 of Rhee and Pugh (35) are indicated in blue. (D) ChIP-PCR confirming the differential binding of CTCF at the variable site. Ubx-P is at the Ubx promoter; for the −ve and +ve primers, see Table 1. Error bars show standard errors of the means.

The variably occupied CTCF site lies in an intron within the Ubx transcription unit. Motif analysis with the CTCF position-weight matrix revealed a strong sequence match at this position (Fig. 1B). It has been proposed that CTCF sites serving different functions may be identifiable at the sequence level, and subfamilies of CTCF binding sites have been identified. We examined the variable site for sequence features that might place it in a defined subfamily. In general, the variable site has features associated with high occupancy, having, in addition to a strong match to the core motif (Patser score = 12.3), the conserved T of module number 1 described by Rhee and Pugh and the CC motif (Fig. 1C) that are both associated with higher levels of CTCF binding (34, 35). The variable site is on the edge of a sequence block that is highly conserved across 15 insect genomes (Fig. 1C), and CTCF binding at this site is clearly identified in pupal-stage chromatin from four Drosophila species (D. melanogaster, D. simulans, D. yakuba, and D. pseudoobscura) covering a range of evolutionary divergence of up to 25 million years (36).

We validated the differential CTCF binding at this site using quantitative PCR with a set of primer pairs spanning the CTCF peak (Fig. 1B and D). We see clearly enriched CTCF binding in T3 versus T1 leg disc chromatin specifically at this CTCF site.

Protein complex formation at the variable CTCF site.

To investigate whether the DNA binding protein CTCF is involved in building a protein complex together with other insulator proteins or transcription factors at this site, we analyzed the binding of other protein components (Fig. 2). Centrosomal protein 190 (CP190) does not bind DNA directly but associates with CTCF [and other DNA-binding insulator components, such as Su(Hw)] through a BTB domain interaction and is required for the enhancer-blocking function of insulator complexes (25, 37, 38) and for looping interactions of CTCF insulators (22). We find no evidence for CP190 association with the variable CTCF site in T1 leg disc chromatin, but CP190 is significantly associated with this site in T3 leg disc chromatin. This suggests that differential binding of CTCF in T3 enables the formation of a protein complex involving proteins associated with insulator function.

FIG 2.

ChIP-PCR analysis of binding of CP190, GAF, and RNA Pol II (Ser5) in the region of the variably occupied CTCF site. RNA Pol II (Ser5) refers to the Ser5-phosphorylated form of RNA Pol II. The results for T1 and T3 chromatin are color coded as shown in the key. Primers are as described in the legend to Fig. 1 and shown in that figure. *, P = 0.02 (t test). Error bars show standard errors of the means.

GAGA factor (GAF) appears to participate in a diverse range of transcriptional processes and is required for the activity of some insulators (39–41). GAF does not bind at the variable CTCF site, but there is substantial binding in the region of primer pair 1 that lies about 1 kb away from the CTCF site (Fig. 2). This strong GAF binding is similar in both T1 and T3 leg imaginal disc chromatin. We also examined the binding of the insulator components Su(Hw), mod(mdg4 isoform N), and BEAF32 but found no evidence for binding in the region of the variable CTCF site in leg discs (data not shown).

Intronic CTCF sites have been implicated in splicing regulation and Pol II pausing (42). We examined the binding profile of Pol II across the region spanning the variable CTCF site and at the Ubx promoter using an antibody that recognizes the Ser5-phosphorylated (Ser5P) Pol II (Fig. 2). Pol II-Ser5P is found preferentially bound across the region in T3 versus T1 discs, which fits with the specific Ubx expression in T3; however, there is no pronounced peak at the CTCF site, and thus, we see no evidence of Pol II pausing at this site. At the promoter, Pol II-Ser5P shows strong binding in T3 and no binding in T1, indicating the engagement of Pol II with the active promoter and a lack of paused Pol II when the Ubx promoter is inactive.

Chromatin topology in the active and inactive states.

We next investigated whether the variable CTCF-dependent protein complex that assembles on the active Ubx gene is associated with alteration in chromosomal topology between the inactive and active states of Ubx transcription. We used chromosome conformation capture (3C) (33) to analyze interactions from the viewpoint of the variable CTCF site as an anchor fragment and 28 nearby target sites, including the Ubx promoter, the abd-A promoter, and CTCF sites across the Ubx and abd-A regions. The overall interaction profiles are shown in Fig. 3A, and the interaction scores for selected primers closest to particular features, e.g., the Ubx promoter and the abd-A promoter, are detailed in Fig. 3B. We find that the variable CTCF site shows a marked preferential interaction with the Ubx promoter in the Ubx active (T3) state (Fig. 3B, anchor 1, Ubx 5′ primers). In contrast, the interaction of the variable CTCF site with the repressed abd-A promoter shows the reverse preference; in T3, there is no interaction, but in the Ubx inactive state (T1), the variable CTCF site is associated with the repressed abd-A promoter (Fig. 3B, anchor 1, abd-A 5′ primers).

FIG 3.

Chromatin interactions in the BX-C in T1 and T3. (A) 3C interactions at 29 sites in the BX-C. Top, overview of the BX-C, showing the T3 CTCF ChIP profile with 3C anchor positions highlighted in gray. The graphs below show the 3C profiles. The genomic sites of anchors 1 (primer 589), 2 (primer 675), and 3 (primer 983) are indicated. Anchor positions are indicated by red bars, and orange bars indicate positions whose results are detailed in panel B. Arrows in anchor 2 data indicate interactions of the Ubx promoter with sites in the abx (left) and pbx (right) regulatory regions. The dashed line indicates the boundary between the Ubx and abd-A regulatory domains (61). Primers are listed in Tables 2 and 3. (B) T1 versus T3 comparisons focusing on selected primers that are closest to key genomic features; for the interactions between anchors and the variable CTCF site, we show data for primers 9 and 10; for the Ubx promoter, primers 12 and 13, and for the abd-A promoter primers 24 and 25. Error bars show standard errors of the means.

Since using the variable CTCF site as the 3C anchor indicated a specific preferential interaction with the Ubx promoter in the active state, we next examined interaction from the viewpoint of a 3C anchor at the Ubx promoter. This confirmed the preferential interaction between the variable CTCF site and the Ubx promoter in the active (T3) state (Fig. 3B, anchor 2, CTCF site primers). In contrast, in T1, the repressed Ubx promoter shows evidence of a preferential interaction with the repressed abd-A promoter.

We also examined a third viewpoint using a 3C anchor at the Mcp boundary element, which contains a CTCF binding site and is in the repressed abd-A domain in both T1 and T3. The Mcp anchor shows a peak of interaction with the abd-A promoter in both T1 and T3 but shows a preferential interaction with the Ubx promoter and the variable CTCF site in the inactive (T1) state (Fig. 3B, anchor 3). Since there is little CTCF associated with the variable site in the inactive state, these interactions may involve the nearby Polycomb response element (bx-PRE) (Fig. 4).

FIG 4.

Ubx regulation and bx mutations. (A) Map of the Ubx regulatory region. Regulatory regions defined by mutation are in blue. The rectangle on the gypsy transposable element indicates Su(Hw) binding sites. Coordinates are indicated for abx enhancer (abx20 [45]), bx and bxd PREs (62), pbx and bxd mutations (61), abx¹ and bx alleles (27), and BRE (44). The gypsy insertion in bx^83Ka was mapped by sequencing: the insertion is at Chr 3R:12528835, with a 6-bp duplication of the target site at position 12528830 to 12528835. In addition to the indicated cluster of bx alleles (gypsy-associated alleles in black and non-gypsy alleles in gray), there is also an outlier, bx^F31, associated with an I element insertion at approximately position 12516500 (27). (B and C) Immunofluorescence labeling of Ubx expression in wild-type (B) and bx^83Ka (C) T3 leg imaginal discs. The discs are oriented with the anterior to the left; the in bx^83Ka T3 leg imaginal disc, Ubx expression is strongly reduced in the anterior compartment.

Overall, the 3C analysis indicates that the Ubx region adopts a different chromatin topology in the active versus the inactive state. The active (T3) state is characterized by increased interaction between the variable CTCF site and the Ubx promoter and decreased association of both the variable CTCF site and the Ubx promoter with repressed regions, specifically, the abd-A promoter and the Mcp boundary element.

Chromatin topology in the bx^83Ka mutation.

The variable CTCF site lies close to the bx-PRE (43), the BRE embryonic enhancers (44), and the abx enhancers (45), which are active both in the embryo and in imaginal discs (Fig. 4). This arrangement, together with the interaction between the variable CTCF site and the Ubx promoter, suggests a model where the variable CTCF site may play a role in facilitating interaction between the abx/bx enhancers and the Ubx promoter. Deletion of a 9.5-kb region that includes the variable CTCF site gives a bx phenotype (bx^34e-prv) (27) caused by decreased Ubx expression in T3 discs, and it is intriguing that the variable CTCF site lies in the heart of the region defined by the cluster of bx mutations. There is a strong connection between bx mutations and insulator function since, of the 10 bx mutations, seven are caused by the insertion of gypsy transposable elements (27, 46), which carry a cluster of binding sites for the Su(Hw) insulator protein, the most studied insulator in Drosophila (47). These gypsy-induced bx alleles are all suppressed in a su(Hw) mutant background (27, 46), indicating that it is not simply the presence of the 7.5-kb gypsy element but, rather, the binding of the Su(Hw) insulator protein that causes the bx mutant phenotype. This suggests that this region is topologically sensitive and that the gypsy insertions may interfere with interactions between the abx/bx enhancers and the Ubx promoter. Specifically, in terms of the above-described model for the function of the variable CTCF site, the insertion of a second topological regulator, Su(Hw), in this region may interfere with the interaction between the variable CTCF site and the Ubx promoter.

To test this hypothesis, we examined the effect of a bx mutation on chromatin topology by carrying out 3C analysis on homozygous bx^83Ka T1 and T3 leg discs. The phenotype of bx mutations is a loss of Ubx expression in the anterior compartment of the T3 imaginal discs, haltere and T3 leg (Fig. 4B and C) (48). In the anterior compartment, Ubx expression may depend on interactions between the promoter and the downstream enhancers, abx and bx, whereas in the posterior compartment, the Ubx promoter may contact the upstream pbx region. This fits with the presence of both upstream and downstream preferential interactions with the Ubx promoter in the active state that we observed in the 3C analysis (Fig. 3A, arrows). The bx mutations might be expected to interfere specifically with the downstream interaction.

In the 3C analysis, we find that the mutation has several effects on chromatin topology in the Ubx region (Fig. 5). First, contrary to the expectations of the model, the gypsy insertion enhances interaction between the variable CTCF site and the Ubx promoter. This enhancement is seen in both T1 and T3, although the interaction remains stronger in T3 (Fig. 5B, anchor 1, Ubx5′ primers, and anchor 2, CTCF site primers). Second, fitting the predictions of the model, the preferential interaction seen in the active state (T3) between the downstream abx enhancer region and the variable CTCF site is lost in the mutant (Fig. 5B, anchor 1, abx primer). Similarly, for the interaction between the abx enhancer region and the Ubx promoter (Fig. 5B, anchor 2, abx primer), there is evidence for stronger interaction in T3 than in T1 in the wild type, and this differential is lost in the mutant. Also fitting the model, in contrast to the abx region, the pbx region preferentially interacts with the Ubx promoter in the active state (T3) in the bx^83Ka mutant (Fig. 5B, anchor 2, pbx primer).

FIG 5.

Chromatin interactions in the BX-C in T1 and T3, comparing data for the wild type and the bx^83Ka mutant. (A) 3C interactions at 29 sites in the BX-C. Top, overview of the BX-C showing the T3 CTCF ChIP profile with 3C anchor positions highlighted in gray. The positions of the abx and pbx regulatory regions are indicated, corresponding to abx¹ deletion (27) and pbx deletions (61). The graphs below show the 3C profiles. The genomic sites of anchors 1 (primer 590) and 2 (primer 675) are indicated. Anchor positions are indicated by red bars, and orange bars indicate positions whose results are detailed in panel B. The dotted line indicates the boundary between the Ubx and abd-A regulatory domains (61). Primers are listed in Tables 2 and 3. (B) Comparisons of interactions at specific sites focusing on selected primers that are closest to key genomic features; for the interactions between anchors and the abx region, we show data for primer 8; for the variable CTCF site, primers 9 (left bars) and 10 (right bars); for the Ubx promoter, primers 12 (left bars) and 13 (right bars); for the pbx region, primer 17; and for the abd-A promoter, primers 24 (left bars) and 25 (right bars). Error bars show standard errors of the means.

Overall, although some predictions of the model are borne out, it appears that the effects of the gypsy insertion are more complex than simply blocking interactions between the variable CTCF site and the Ubx promoter.

The bx^83Ka insertion affects protein binding in flanking regions.

To investigate this further, we examined protein binding in the region of the variable CTCF site in homozygous bx^83Ka T1 and T3 leg discs (Fig. 6). Strikingly, we find that, in the mutant, CTCF is strongly associated with the site not only in T3 but also in T1. In addition, we find that the gypsy insertion in the bx^83Ka mutation also strongly affects GAF binding; compared to the wild type, it is markedly reduced in both T1 and T3. Pol II binding shows, as expected, clear occupancy in the T3 discs, where Ubx is expressed in posterior compartment cells.

FIG 6.

Binding of CTCF, GAF, and RNA Pol II (Ser 5) in the region of the variably occupied CTCF site in the bx^83Ka mutant. (A) ChIP-PCR analysis of T1 and T3 in the bx^83Ka mutant. RNA Pol (Ser5) refers to the Ser5-phosphorylated form of RNA Pol II. Data are in blue for T1 chromatin and in green for T3 chromatin. Primers are indicated on the x axes and are as described in the legend to Fig. 1 and shown in that figure. (B) Comparison of wild type versus bx^83Ka at T1 and T3 for the CTCF peak (primer 3) and for the GAF peak (primer 1). Data are color coded as shown in the key. Error bars indicate standard errors of the means. The GAF binding interval is from reference 63.

Overall, perhaps the most striking effect of the bx^83Ka insertion is the increase in CTCF binding at the variable CTCF site, particularly in T1. This indicates that the gypsy insulator can affect the loading of insulator proteins onto a nearby site, and this fits with an increased association between the variable insulator site and the Ubx promoter. It is possible that this interaction excludes the abx regulatory region, since the preferential contact between the abx regulatory region and the variable CTCF site seen in the active state in the wild type is lost in the mutant.

DISCUSSION

We have identified a variably occupied CTCF binding site in the Ubx gene in the Drosophila BX-C. This site lies close to characterized Ubx regulatory elements, and we find that CTCF occupancy is associated with a specific interaction between the variable site and the Ubx promoter in the transcriptionally active state. These observations suggest a model that CTCF binding at this site facilitates interaction between the regulatory elements and the Ubx promoter.

This model is supported by our studies on the bx^83Ka mutation, where the insertion of a gypsy insulator close to the variable CTCF site disrupts the chromatin topology. One explanation for the effect of the gypsy insertion on Ubx expression is that the gypsy insulator acts as an enhancer blocker, preventing interactions between the Ubx promoter and regulatory elements (e.g., abx) lying beyond the insulator insertion site (49). However, a simple enhancer blocking model does not fit with the enhanced interaction we see between the variable CTCF site and the Ubx promoter in the bx^83Ka mutant, nor does it explain the tight clustering of gypsy insertions with a bx phenotype within a specific 11-kb region centered on the variable CTCF site. Our analysis shows that the bx^83Ka insertion does not simply introduce an insulator but also has effects on flanking regions. In particular, the bx^83Ka insertion affects the binding of CTCF at the variable CTCF site, leading to clearly enhanced CTCF occupancy in both T1 and T3 discs. In the case of bx^83Ka, the gypsy insertion also lies close to a GAF ChIP binding peak and results in loss of GAF binding in both T1 and T3 discs. This effect on GAF binding is difficult to interpret functionally; GAF has a role in Ubx expression, as the GAF gene Trl interacts with Ubx alleles (50). However, Trl mutant clones in imaginal discs do not appear to affect Ubx expression (51, 52). The topological changes associated with the bx^83Ka insertion include enhanced interactions between the variable CTCF site and the Ubx promoter in both T1 and T3 and loss of the preferential interaction between the variable CTCF site and the distant abx regulatory region in T3. This suggests that the insertion of a gypsy insulator may stabilize CTCF binding and promote interactions with the Ubx promoter but in a manner that excludes interactions with distant regulatory elements. Hence, the gypsy Su(Hw) insulator element may indeed act as an enhancer blocker, but it may do so in collaboration with a CTCF complex. We speculate that the involvement of CTCF in the mechanism that generates the mutant phenotype explains the observed clustering of gypsy insertions with bx phenotypes around the variable CTCF site.

Although our observations indicate a likely role for CTCF in facilitating enhancer-promoter interaction in Ubx regulation, functional studies will be required to confirm the role of CTCF and its importance for Ubx expression. In this regard, we have looked for genetic interaction between CTCF and Ubx. As null CTCF mutants are lethal, we investigated whether the Ubx haplo-insufficient phenotype is enhanced by heterozygosity for CTCF. We have not seen clear enhancement in this situation, and further work will be required to test the proposed CTCF role.

Why are some CTCF binding sites constitutive and others variably occupied? The occupancy of CTCF sites across the BX-C sheds light on this issue but initially presents a puzzle. CTCF sites within the abd-A and Abd-B domains are occupied even when these domains are silenced by Pc-mediated repression, whereas the variable CTCF site in the Ubx gene is only occupied when the Ubx domain is derepressed. This raises questions about the ability of CTCF to access its binding site in different chromatin states. There is evidence that CTCF binding is sensitive to the chromatin configuration. In particular, CTCF binding is affected by nucleosome positioning, and CTCF is unable to bind if its target site is covered by a nucleosome (21, 53). Examination of chromatin accessibility within the repressed abd-A and Abd-B domains by DNase I sensitivity reveals that CTCF sites generally correspond to small regions of DNase I accessibility within the repressed domains (Fig. 7A), indicating that CTCF is bound at sites of open, potentially nucleosome-free chromatin. Interestingly, these sites are bound by other factors, for example, Yki and GAF, so it is unclear which factor or factors are responsible for initiating and establishing open chromatin at these positions. Importantly, the presence of other factors indicates that CTCF is not necessarily responsible for pioneering binding at these sites in repressed chromatin. The variable CTCF site in Ubx supports the idea that CTCF on its own may not be able to bind to repressed chromatin, and it is intriguing that in this particular case, the adjacent DNase I site, occupied by Yki, GAF, and Pho, does not extend over the CTCF site (Fig. 7B). Occupancy of the variable site may be dependent on Pc derepression of the Ubx domain, enabling nucleosome remodeling to expose the CTCF site for binding. A different perspective is given by the finding that, although CTCF does not bind to the variable site in the repressed Ubx domain in T1 in the wild type, it does bind in the context of the bx^83Ka mutant. The insertion of the gypsy transposon carrying the Su(Hw)-dependent gypsy insulator may stabilize CTCF binding at the variable binding site, perhaps through a general function of insulator complexes to facilitate loading of insulator components at nearby sites. Overall, our studies point to a view of CTCF binding where CTCF is in competition with nucleosomes for site occupancy. In the repressed state in T1, the nucleosome is dominant and there is very little CTCF binding to the variable site. CTCF binding may be enhanced either by decreasing nucleosome occupancy, associated with the opening of the Ubx domain in T3, or by local interactions between insulator complexes stabilizing CTCF binding.

FIG 7.

Chromatin accessibility and protein binding at CTCF sites in the BX-C. (A) In the repressed BX-C in the Drosophila Kc167 cell line (Kc), DNase I profiling reveals specific accessible sites in the repressed chromatin. Thirteen CTCF sites, bound in T3 chromatin, are numbered; 11 of the 13 are associated with DNase I sensitivity peaks. (B) Close-up of selected sites from the experiment whose results are shown in panel A; the binding peaks of several regulators align with the DNase I sites. The variable CTCF site (site 1) is offset from this alignment, whereas other, constitutive CTCF sites are more closely aligned with the DNase I sites. Data from the CTCF T1 and T3 leg are from this paper; data for Pho are from reference 64; data for Yki and GAF are from reference 63; data for DNase I Kc167 are from reference 65; and data for CTCF Kc167 are from ModENCODE (www.modencode.org) data set 908.

Our data also provide a view of the in vivo 3-dimensional organization of the BX-C, comparing the situation in T1, where all three BX-C genes are inactive, with the situation in T3, where Ubx is active and abd-A and Abd-B are inactive. In the active Ubx state, both the variable CTCF site and the Ubx promoter engage in long-range interactions over a range of about 100 kb, but the interactions we see are nevertheless confined to the Ubx domain. In the repressed state, the variable CTCF site and the Ubx promoter show more association with distant repressed regions outside the Ubx domain (Fig. 3). This fits with previous studies, both in Drosophila (54, 55) and in the mammalian Hox complexes (56–60), which support the idea of regulatory domains as dynamic topological structures where repressed domains cluster together and expressed domains are segregated into a separate compartment.