Parasegment-specific expression of the Ubx, abd-A, and Abd-B homeotic genes in the Drosophila melanogaster Bithorax complex (BX-C) is controlled by nine autonomous regulatory domains, which are separated by special elements called boundaries or insulators [1]. Boundaries ensure autonomy by blocking contacts between regulatory elements in one domain with regulatory elements in adjacent domains. Boundaries can also prevent enhancers from interacting with promoters [24]. In addition to insulator activity, some boundaries have an ability to specifically interact with their target gene in BX-C, enabling enhancers in distant regulatory domains to stimulate their target promoter [5]. These properties of the boundaries ensure correct parasegment-specific expression of the BX-C genes during Drosophila development. Consistent with this idea, Fab-6, Fab-7, and Fab-8 were shown to specifically interact with the promoter upstream region of Abd-B gene [6]. It is likely that this interaction determines the correct topological positioning of the corresponding regulatory domains (iab5 – iab7) with Abd-B promoter in parasegments 10–12.

Most of the BX-C boundaries contain binding sites for Drosophila CTCF (dCTCF), and these sites are important for the insulator activity of these boundaries (Fig. 1) [7]. In the intron of the Ubx gene 30 kb downstream from the promoter, a variable dCTCF binding site (vCTCF) was identified (Fig. 1) [8]. dCTCF does not occupy this site in tissues where Ubx is inactive (imaginal discs of the first pair of legs), but binds to it when the Ubx gene is transcriptionally active (imaginal discs of the third pair of legs). Moreover, dCTCF binding to vCTCF is associated with changes in the topology of the abx/bx regulatory domain: in tissues where Ubx is active an increase in the frequency of vCTCF contacts with the Ubx promoter is observed [8]. A model was proposed according to which binding of dCTCF to vCTCF facilitates tissue-specific interaction of the abx, BRE enhancers with the Ubx promoter [9, 10]. The aim of this study was to test this hypothesis.

Fig. 1.
figure 1

Schematic organization of the genes and regulatory domains in BX-C. abx/bx, bxd/pbx and iab-2–iab-8 domains responsible for the regulation of Ubx, abd-A, and Abd-B genes and for the development of parasegments 5-13/T3-A8 segments are shown. Ubx embryonic enhancers are shown as purple boxes. The lines with colored circles mark boundaries. The binding sites for insulator proteins dCTCF, Pita, and Su(Hw) are shown as red, blue, and yellow circles. On the lower part of the figure, regulatory regions containing dCTCF variable site and Fab-7 boundary, as well as their deletions are shown. Fab-7 boundary Deoxyribonuclease I hypersensitive sites HS*, HS1, HS2, HS3 are shown as black rectangles on the coordinate bar. The bx PRE hypersensitive site is depicted as a green box, sites for Pho and GAGA-factor proteins as orange and blue ovals. attP, lox, and frt sites used for genetic manipulations are shown as white, gray, and blue triangles.

To study vCTCF function in enhancer-promoter interactions, we used the CRISPR/Cas9 system to delete a 3408-bp DNA fragment (3R:16701239..16704646) that spans the vCTCF site and the bx PRE (polycomb response element) 1 kb downstream, and in its place we introduced an attP site (Δ3.4attP, Fig. 1). Flies homozygous for Δ3.4attP deletion show evidence of variable LOF transformations. The deletion transforms the anterior third thoracic segment toward the anterior second thoracic, a phenotype known as bithorax (bx) [11, 12]. In mutant flies the anterior third leg resembles the second leg, and in ~10% of flies anterior notal tissue is present on the dorsal surface of the third thoracic segment (Fig. 2). These transformations are caused by a disruption in the interactions of enhancers downstream of vCTCF with Ubx promoter. The Δ3.4attP deletion overlaps with a previously described 9.5 kb deletion, bx34e-prv. Like Δ3.4att, it also has a variable bx phenotype which is caused by a decrease in Ubx expression in the imaginal discs of segment T3 [11]. Next, we used attP site in Δ3.4attP as an integration platform to find minimal element that can rescue the mutant phenotype. We carried out attP-attB mediated integration of the 831-bp bx PRE fragment (PRE831, 3R:16702487..16703317) into Δ3.4attP deletion and discovered that PRE831 completely reverts bx phenotype to wild type. This finding suggests that vCTCF is redundant, while bx PRE may play a role in facilitation of enhancer-promoter interaction.

Fig. 2.
figure 2

(a) Phenotypic comparison of T3-A1 tergites of wt, Δ3.4attP and PRE831 flies. Δ3.4attP has a variable phenotype, ~10% of flies have an enlarged A1 segment, a subset of dorsal T3 cells (marked with a red arrow) are transformed toward mesonotum, while their neighbors are untransformed. PRE831 integration restores the mutant phenotype to wild type. (b) Phenotypic comparison of T3 legs of wt, Δ3.4attP and PRE831 flies. In wild type flies, T2 legs have a pair of long bristles, which are absent on T3 legs. Δ3.4attP flies develop one long bristle on T3 legs (marked with a red arrow), which indicates a partial transformation toward T2. T2 legs of PRE831 flies look wild type. (c) Bright and dark field images of abdominal cuticle of wt, Fab-7attP50, vCTCF+PRE males. In wt males, A7 segment is absent, A6 sternite is banana-shaped and has no bristles, while A5 sternite is rectangular and covered with bristles. A5 tergite is completely covered with trichomes, while A6 has bristles only along anterior and ventral margins (see dark field). In Fab-7attP50 males, A6 segment is transformed toward A7 (does not develop) due to the fusion of iab-6 and iab-7 regulatory domains. vCTCF+PRE males also do not develop A6 segment.

In order to test vCTCF insulator activity we used Fab-7attP50 replacement platform (Fig. 1). In this platform, Fab-7 boundary is removed, resulting in the fusion of iab-6 and iab-7 regulatory domains. This leads to ectopic activation of the iab-7 regulatory domain in PS11, which in turn results in the loss of the sixth abdominal segment in adult males [1315]. It was demonstrated previously that PREs are often located in close proximity to insulators and contribute to the formation of a functional boundary [16, 17]. Therefore, a fragment containing both bx PRE and vCTCF in reverse orientation, vCTCF+PRE (2161-bp, 3R:16702487..16704647) was tested in Fab-7attP50. We found that the 6th abdominal segment is still missing in males carrying vCTCF+PRE insertion. This finding indicates that the vCTCF+PRE sequence does not have insulator activity.

Altogether, our data do not support a model in which vCTCF is a necessary mediator of enhancer-promoter interactions in abx/bx domain. Moreover, the data suggest that the bx PRE may play that role. However, further research is needed to explore the functions of this element in Ubx regulation. Since the loss of the bx PRE leads only to a variable LOF phenotype, it can be assumed that, in contrast to the Abd-B enhancers, Ubx enhancers are much more autonomous and less dependent on other regulatory elements to form appropriate promoter contacts.