Redundant enhancers in the iab-5 domain cooperatively activate Abd-B in the A5 and A6 abdominal segments of Drosophila

In Drosophila melanogaster segment identity in the posterior two-thirds of the body is controlled by the three homeotic genes, Ultrabithorax (Ubx), abdominal-A (abd-A) and Abdominal-B (Abd-B), which form the bithorax complex (BX-C) (Lewis, 1978). The specification of parasegments (PS) and their corresponding segment (A) depends on the expression patterns of these three homeotic genes (Duncan, 1987; Karch et al., 1985, 1990; Maeda and Karch, 2015; Peifer et al., 1987). The genes are controlled by an array of nine regulatory domains, each of which is thought to direct the expression of one of the homeotic genes in a spatiotemporal pattern appropriate for the particular PS (segment) that the regulatory domain specifies. The Abd-B gene, for example, is responsible for the differentiation of PS10 (A5), PS11 (A6), PS12 (A7) and PS13 (A8). Its expression pattern in each of these parasegments is controlled by four regulatory domains: iab-5, iab-6, iab-7 and iab-8, respectively (Fig. 1A).

Analysis of BX-C regulatory domains, including those controlling Abd-B indicate that they are composed of the same set of elements (Kyrchanova et al., 2015; Maeda and Karch, 2015). Each domain has an initiator element that sets the activity state (on or off) of the domain early in embryogenesis (Maeda and Karch, 2015; Mihaly et al., 2006; Peifer et al., 1987). Initiators respond to the maternal gap and pair-rule gene products that subdivide blastoderm stage embryos along the antero-posterior axis into 14 parasegments (Busturia and Bienz, 1993; Casares and Sánchez-Herrero, 1995; Drewell et al., 2014; Ho et al., 2009; McCall et al., 1994; Qian et al., 1991; Shimell et al., 1994; Starr et al., 2011). For example, in PS10 (A5), the iab-5 initiator turns on the iab-5 domain, while the adjacent iab-6 and other more distal (relative to centromere) domains remain in the off state (Iampietro et al., 2010). In PS11 (A6), the initiator in iab-6 turns the domain on. Although iab-5 is also active in PS11, iab-7 and iab-8 are off. The gene products that set the activity state of the BX-C domains disappear during gastrulation and different mechanisms are deployed to remember the on or off state. The on state is maintained by Trithorax group proteins, while the off state is maintained by Polycomb group proteins (Busturia and Bienz, 1993; Kassis et al., 2017; Kuroda et al., 2020; Simon et al., 1992; Shimell et al., 2000; Ciabrelli et al., 2017; Müller and Bienz, 1992). These factors interact with special elements in each domain called Trithorax or Polycomb response elements (TREs or PREs). Finally, each domain has stage- and tissue-specific enhancers that are responsible for activating patterns of homeotic gene expression that drive PS (segment) differentiation (Maeda and Karch, 2015). Each domain is bracketed by chromatin boundary elements (Barges et al., 2000; Bender and Lucas, 2013; Bowman et al., 2014; Galloni et al., 1993; Gyurkovics et al., 1990; Hagstrom et al., 1996; Iampietro et al., 2010; Karch et al., 1994; Kyrchanova et al., 2020; Mihaly et al., 2006; Zhou et al., 1996). The boundaries in the Abd-B region (Fab-6, Fab-7 and Fab-8) have two important functions. The first is to block crosstalk between adjacent regulatory domains so that they can function autonomously. The loss of one of these boundaries leads to the ectopic activation or silencing of neighbouring regulatory domains. For example, deletion of the Fab-6 boundary element can result in the ectopic activation of iab-6 and silencing iab-5 in PS10 (A5) leading respectively to gain-of-function and loss-of-function transformation of PS (segment) (Iampietro et al., 2010; Postika et al., 2021). The second function is boundary bypass. This function enables enhancers in the Abd-B regulatory domains to bypass intervening boundaries and activate Abd-B (Kyrchanova et al., 2019a,b; Postika et al., 2018). However, Fab-7 replacement experiments suggest that bypass activity may be a special property of the Abd-B boundaries, as boundaries from elsewhere in the genome do not support bypass (Hogga et al., 2001; Kyrchanova et al., 2019a,b).

Although identity of PS10-PS13 (A5-A8) is determined by the pattern of Abd-B expression in both sexes, the phenotype of the adult cuticle in segments A5 and A6 in Drosophila melanogaster differs in males and females (Jeong et al., 2006; Kopp et al., 2000; Massey and Wittkopp, 2016; Williams et al., 2008). In females, cuticle pigmentation and morphology in A5 and A6 are similar to those in more anterior segments, the identity of which is determined by abd-A. In these segments, the tergite has a posterior stripe of dark pigmentation, whereas the sternite has a quadrilateral shape and has multiple bristles. The pigmented stripe in tergites A2-6 is generated by the yellow (y) and tan genes, which are regulated by the optomoter blind (omb) gene (Kopp and Duncan, 1997). The bric-a-brac (bab) complex encodes DNA-binding proteins that repress the expression of the genes responsible for cuticle pigmentation (Couderc et al., 2002; Kopp et al., 2000; Roeske et al., 2018). While female pupae express bab in abdominal segments A2-A6, bab expression in males is limited to segments A2-A4. The sex-specific pigmentation pattern and cuticle morphology in A5 and A6 in males depend upon Abd-B and the male product of the double-sex gene (dsx^M), which together function to repress expression of bab genes in cells giving rise to the A5 and A6 cuticle (Kopp et al., 2000; Massey and Wittkopp, 2016; Wang et al., 2011). Abd-B is also thought to interact with the y gene to activate its expression, while it positively regulates tan indirectly (Roeske et al., 2018).

The level of Abd-B expression in PS10 (A5) and PS11 (A6) is not the same and correlates with their distinctive morphology. The Abd-B expression in A5 is relatively low and this segment has morphological features of the A4 segment, where Abd-B is not expressed: the A5 sternite has a quadrilateral shape and has multiple bristles, while the A5 tergite is covered by small trichome hairs. However, owing to the expression of Abd-B in the A5 segment of males, differences are observed: the sternite becomes wider, the tergite is completely pigmented and trichomes are less dense (Celniker et al., 1990; Maeda and Karch, 2015) (Fig. 1). The higher levels of Abd-B in A6 are accompanied by specific morphological features in both the sternite and tergite. The A6 sternite lacks bristles and has a unique ‘banana’ shape, while the trichomes on the fully pigmented tergite are restricted to the anterior and dorsal margins, instead of covering nearly the entire tergite.

Here, we have investigated the mechanisms responsible for regulating Abd-B expression during the differentiation of the male cuticle in segments A5 and A6. We have found that iab-6 is not on its own able to direct expression of Abd-B in the manner that is required for differentiation of the A6 cuticle in male flies. Instead, the iab-5 and iab-6 domains share a common set of partially redundant cuticle enhancers located in iab-5 that are crucial for male-specific differentiation of the cuticle of A5 and A6. In this respect, the functioning of iab-5 and iab-6 domains fits well with the additive model for BX-C regulation suggested by Lewis (1978).

Experiments in which Fab-7 was replaced by heterologous boundaries have shown that the three boundaries in the Abd-B region of the complex, Fab-6, Fab-7 and Fab-8, have both blocking and bypass activity (Kyrchanova et al., 2019a,b; Postika et al., 2018). In contrast, heterologous boundaries such as scs, Mcp or artificial DNA fragments consisting of multimerized binding sites for C2H2 zinc-finger proteins such as Pita or dCTCF have only blocking activity (Kyrchanova et al., 2016, 2017; Hogga et al., 2001). For example, when a multimer consisting of five binding sites for Pita (Pita^×5) is used to replace Fab-7, the identity of PS12 (A7), which is specified by iab-7, is the same as in wild type; however, Pita^×5 blocks iab-6 from activating Abd-B in PS11 (A6). Instead, iab-5 regulates Abd-B activity in both PS10 (A5) (where it normally functions) and PS11 (A6), and adult Pita^×5 males have a duplicated A5 segment. While these experiments showed that boundaries at the position of Fab-7 (between iab-6 and iab-7) must have both blocking and bypass activity for the proper regulation of Abd-B, a similar requirement has not been established for boundaries at the location of Fab-6 or Fab-8. To test whether the regulation of Abd-B by iab-5 requires that the boundary located between iab-5 and iab-6 needs bypass activity, we took advantage of the F6^1attP replacement platform, in which a 1389 bp sequence spanning the Fab-6 boundary was substituted by an attP site (Postika et al., 2021). Reintegration of the 529 bp core Fab-6 boundary, including two dCTCF sites, completely restored the wild-type male phenotype, suggesting that the 1389 bp deletion does not include important regulatory elements other than the Fab-6 boundary.

We inserted the Pita^×5 insulator into the F6^1attP platform. Based on how it functions as a replacement for Fab-7, we expected that it would block crosstalk between iab-5 and iab-6, and would also prevent iab-5 from regulating Abd-B. To assess the activity state of iab-5 and iab-6 cuticle enhancers, we included a mini-yellow (mini-y) reporter that we placed either upstream of Pita^×5 or downstream so that it would be located in the iab-5 (mini-y Pita^×5) or iab-6 (Pita^×5 mini-y) domains, respectively (Fig. S1). The reporter consists of a yellow (y) cDNA fused to the 340 bp y promoter. As it lacks the enhancers of the endogenous y gene, its expression depends upon nearby enhancers. Expression of mini-y was examined in a y¹ background. In flies carrying the null y¹ allele, the tan gene is appropriately expressed in A5 and A6, reflecting the Abd-B activity, and the resulting pigmentation in the tergite is light brown, not black (Camino et al., 2015; Rebeiz and Williams, 2017).

Males homozygous for the starting F6^1attP deletion differ from wild type in that segment A5 has an incomplete gain-of-function and loss-of-function transformation (Fig. 1). The A5 sternite has a shape like that normally observed in A6, but with several bristles; the A5 tergite has patches of cuticle that lack trichomes, which is indicative of a gain-of-function transformation towards A6 identity. On the other hand, large regions of the A5 cuticle also lack tan pigmentation, indicating that the cells have an A4 identity. There are also unexpected (based on Fab-7 and Fab-8 boundary deletions: Mihaly et al., 1997; Barges et al., 2000) loss-of-function phenotypes in A6, including bristles on the sternite and regions of the tergite that are depigmented or have ectopic trichomes.

As expected, the Pita^×5 replacement (with or without mini-y) blocks crosstalk between iab-5 and iab-6 and the gain-of-function transformations of A5 are eliminated (Fig. 1). In addition, consistent with the idea that a boundary located between iab-5 and iab-6 must have bypass activity, A5 resembles A4: the sternite has a quadrilateral shape and is covered in bristles, while the tergite is covered in trichomes and, instead of being fully pigmented, there is only a posterior stripe. This result shows that iab-5 is unable to activate Abd-B in A5 when Pita^×5 is present.

However, there is also an unexpected result: the differentiation of A6 is altered compared with wild type. The defects are most clearly evident when mini-y is excised and y⁺ allele is introduced. Fig. 1 shows that pigmentation of the A6 tergite resembles A4: there is only a stripe of pigment along the posterior margin of the tergite. In addition, the A6 tergite is covered in trichomes just like A4. Although the A6 sternite has a nearly normal shape, there are multiple bristles. These loss-of-function transformations indicate that the Pita^×5 insulator disrupts Abd-B dependent cuticle differentiation not only in A5, but also in A6. As the insulator is located between iab-5 and the Abd-B gene, this would imply that it is blocking enhancers in iab-5 that are required for the proper activation of Abd-B in the cells that form the A5 and also the A6 cuticle in males.

Further support for this conclusion comes from analysis of mini-y expression in the y¹ background. In Pita^×5mini-y males the reporter located in the iab-6 domain it is not turned on in either A5 or A6. Instead, only the tan gene is expressed, and importantly it is expressed in an A4-like pattern. This result would indicate that enhancers in iab-5 are required to drive expression of mini-y inserted in iab-6 in the A6 tergite. Finally, when the mini-y reporter is in iab-5 we observe a mosaic pattern of pigmentation in the posterior stripes of the A4, A5 and A6 segments regulated by omb, but not by Abd-B.

To map enhancers in the iab-5 regulatory domain responsible for Abd-B expression in the cells that give rise to the male cuticle in A5 and A6, we linked 1-3 kb overlapping DNA sequences from the iab-5 domain (i5¹-i5⁷ and i5ⁱⁿⁱ; Busturia and Bienz, 1993) to a y reporter in a transgene that also carries a mini-white (w) (Fig. 2, Fig. S2). To reduce potential position effects, we placed Pita^×5 upstream of the iab-5 DNA fragments. Using phiC31-mediated recombination (Gao et al., 2008), we integrated a collection of eight i5 transgenes into a well-characterized 86 Fb platform (Bischof et al., 2007). Of these i5 fragments, only three, i5¹ (1013 bp), i5² (2145 bp) and i5⁷ (2524 bp), activated y expression in the cuticle (Fig. 2, Fig. S2). For all three, pigmentation was observed in the A5 and A6 tergites. Interestingly, we found that the i5⁷ fragment was only able to activate y in the forward (genomic) orientation. We tested two i5⁷ subfragments from the proximal (i5^S5) and distal (i5^S6) ends relative to the centromere. Of these, only i5^S6, activated y. Thus, in the larger i5⁷ fragment, the enhancer in i5^S6 must be located next to the promoter to function. As the i5¹ and i5² overlapped, it seemed possible that they share the same enhancer. To test this, we generated three smaller fragments (i5^S1, i5^S2 and i5^S3) spanning most of i5¹ and i5². Of these, only i5^S2, which includes the overlap between i5¹ and i5², activates mini-y.

We next determined whether iab-5 sequences are able to regulate Abd-B in A6 when placed in iab-6. We used the F6^1attP landing platform to insert the same collection of iab-5 sequences into the iab-6 regulatory domain (Fig. S3). The transgenes included Pita^×5 to block crosstalk between iab-5 and testing fragments, and excisable mCherry and mini-y reporters arranged so that in the replacement they are located in iab-6 (Fig. S1).

Three of the iab-5 sequences, i5¹, i5³ and i5⁷ are able to stimulate mini-y expression to different extents in the A6 segment (Fig. S3). Although both i5¹ and i5⁷ also stimulated y when inserted in 86 Fb platform, i5³ did not (Fig. 2). Conversely, i5² failed to function when placed in iab-6, while it is active in the 86 Fb platform. It seems likely that ‘position effects’ are responsible for the differences in the activity of the iab-5 sequences when linked to the y reporter in 86 Fb or inserted in iab-6. As would be expected from their placement in iab-6, i5¹, i5³ and i5⁷ do not activate the reporter in more-anterior segments. Surprisingly, insertion of the i5⁷ fragment in the reverse orientation (i5^7R) stimulates mini-y expression in posterior stripes not only in the A6 tergite, but also in the A5 and A4 tergites (Fig. S3). As the distal part of i5⁷, i5^S6, induces a much stronger activation of the mini-y, it would appear that sequences elsewhere in i5⁷ contain a silencer.

As the reporters compete with Abd-B for enhancer activity, we assessed the cuticle phenotypes after removing the reporters and introducing a y⁺ allele (Fig. 3). When i5¹, i5³ or i5⁷ are included in the Pita^×5 replacements, the phenotype of A6 is close to wild type. Even though i5^S2 and i5² activate mini-y at 86 Fb, neither could rescue the loss-of-function phenotypes induced by Pita^×5. On the other hand, i5^S1 and i5³, which do not stimulate mini-y at 86 Fb, completely rescue the Pita^×5 induced loss-of-function phenotypes in A6 (Figs 2 and 3). The remaining fragments that are active when introduced into iab-6 are i5ⁱⁿⁱ and i5⁷. The former is not active at 86 Fb, while the latter is. Both partially rescue the Pita^×5-induced defects in pigmentation and trichome distribution in A6, whereas the weak loss-of-function phenotype (bristles) in the sternite is rescued by i5⁷ but not by i5ⁱⁿⁱ. As was the case in 86 Fb, the i5⁷ enhancer activity is orientation dependent and it is not observed in i5^7R. Thus, there are several enhancers in iab-5 that could help drive Abd-B expression in the cuticle and generate morphological features that are characteristic of A6.

The results in the previous section indicate that enhancers in iab-5 are important for the proper differentiation of the adult cuticle in A6. If this suggestion is correct, one would predict that the deletion of the iab-5 initiation element will disrupt the development of the adult cuticle not only in A5 but also in A6. To test this prediction, we used CRISPR/Cas9 to delete a 1975 bp genomic DNA segment that spans the iab-5 initiator (Busturia and Bienz, 1993) and replace it with an attP site and an excisable dsRed reporter under control of the 3×P3 hsp70 promoter (Fig. S1). As expected for an initiator deletion, the A5 segment in i5^attP males resembles A4 (Fig. 4, Figs S4 and S5). Crucially, this is not the only phenotypic alteration in i5^1attP males. Instead of the characteristic banana shape, the A6 sternite has an intermediate quadrilateral shape and also has bristles, whereas the A6 tergite has an irregular and variable pigmentation. In addition, trichome hairs are found in large patches that often coincide with areas of depigmentation. These results show that deletion of the iab-5 initiator affects Abd-B expression in both the A5 and A6 segments. To confirm that iab-5 is not properly activated in i5^1attP, we integrated a mini-y reporter using the attP site. As expected, the mini-y reporter introduced into iab-5 is off in A5. In A6, black pigmentation is restricted to several patches on the tergite, while tan-only dependent pigmentation occupies a somewhat larger area (Fig. 4, Fig. S5).

To confirm that the observed effects on mini-y and A6 morphology are induced by deletion of the iab-5 initiator, we introduced a 1025 bp i5ⁱⁿⁱ fragment together with mini-y into i5^1attP. The resulting flies have wild-type morphology except for one or two bristles on the A6 sternite, and the mini-y reporter is expressed throughout the tergite in A5 and also inA6 (Fig. 4). The presence of bristles on the A6 sternite is due to competition between the mini-y and Abd-B promoters.

To further evaluate the functional role of the i5 enhancers in both A5 and A6, we have created a platform by deletion of most iab-5 sequence to test the functional role of individual i5 regulatory elements and their various combinations. For this purpose, we used Cre-mediated recombination between lox sites located in the i5^attP and an Mcp boundary deletion, M^3attP, in which a 3333 bp sequence spanning the region around the Mcp boundary was substituted for attP and lox sites (Fig. 4, Figs S6 and S7). After Cre recombination, the final deletion, M-i5^attP, is 10,935 bp. It extends from the centromere proximal side of the Mcp boundary through the iab-5 initiator, leaving the 2126 bp i5⁷ sequence (Fig. 4) and single attP and lox sites. M-i5^attP males have a pigmented A4 segment and display other signs of gain-of-function transformation of not only A4 and A5, but also A3: the sternites have two lobes somewhat like the A6 sternite, whereas there is a depletion of the trichomes on the tergites (Fig. 4, Fig. S4).

Aiming to prevent the iab-4 domain from activating Abd-B, we reintroduced a M⁴¹³ insulator, characterized previously (Kyrchanova et al., 2007), with the mini-y reporter using the phiC31 integration system (Fig. S1). The resulting M⁴¹³mini-y replacement contains only the i5⁷ sequence. As would be expected, as there is no initiator in iab-5, the domain is inactive in A5 and mini-y is not expressed in this segment. However, in spite of the fact that the iab-5 domain is inactive, the phenotype of A6 resembles wild type, and the mini-y reporter, which is located in the inactive iab-5 domain, is expressed throughout the A6 tergite (Fig. 4). When the reporters are excised, the minimal Mcp⁴¹³ boundary is not able to prevent iab-4 from activating the enhancer in i5⁷ or Abd-B directly. In addition to having a wild-type A6 segment, the tergites in A4 and A5 are nearly covered in pigmentation, indicating that the Abd-B gene is active in both of these segments (Fig. S8).

A plausible interpretation of these findings is that the iab-4 domain somehow activates the remaining iab-5 enhancers (i5⁷) in this deletion. To test the possible role of the iab-4 regulatory region in the gain-of-function transformation of A4 in Mcp⁴¹³, we deleted a 4401 bp sequence (iab-4^Δ), including the iab-4 initiator, as described previously (Postika et al., 2018). The deletion of these iab-4 sequences not only reverts the gain-of-function transformations in M⁴¹³, but also results in a dramatic loss-of-function transformation of both A5 and A6 (Fig. 4). Although A5 resembles A4 in M⁴¹³iab-4^Δ males, the pigmentation patterns in the A6 tergite range from a few dark spots to almost ubiquitous pigmentation (Fig. S9). The A6 sternite is also mis-shapen and covered in bristles. As A6 appears wild type when the iab-4 domain is intact, it would appear that in the M⁴¹³ platform, sequences in iab-4 are able, either on their own or in collaboration with elements in iab-6, to activate enhancers in i5⁷ and to help direct the proper expression of Abd-B in A6.

We next used the M-i5^attP platform to reconstruct a minimal iab-5 regulatory domain. As Mcp⁴¹³ in combination with the two reporters is more effective in insulating against elements in iab-4, we will first consider the functioning of different iab-5 sequences in the presence of the reporters. In the first set of experiments, we tested i5^S2 and i5³. As the M-i5^attP deletion retains the i5⁷, it is included in all of the replacements we tested. Thus, the three combinations are i5^S2+i5⁷; i5³+i5⁷ and i5^S2+i5³+i5⁷ (Fig. 5). In both i5⁷ and i5^S2+i5⁷, the mini-y reporter is expressed in a mosaic pattern along the posterior margin of A4 and A5. In contrast, we observed only rare spots of dark pigmentation in the A5 segment in combinations containing i5³. Thus, the i5³ region has a negative effect on mini-y expression in cis. In all three combinations (i5^S2+i5³+i5⁷), the anterior two-thirds of the A5 tergite is largely devoid of pigmentation, indicating that the tan gene is also not expressed in much of the tergite. At the same time, the A6 segment has a nearly wild-type phenotype.

We next tested the same combinations of i5 enhancers with the initiator, i5ⁱⁿⁱ. The i5ⁱⁿⁱ+i5⁷ combination expands the expression domain of mini-y in A5, while having minimal effect on expression in A4. However, there are regions in the anterior of the A5 tergite where mini-y is not expressed (Fig. 5). Although adding i5^S2 has little effect on the pattern of mini-y expression (i5^S2+i5ⁱⁿⁱ+i5⁷), there is a noticeable expansion in the expression area in A5 when i5³ is combined with the initiator (i5³+i5ⁱⁿⁱ+i5⁷) (Fig. 5). This is the opposite of what was observed for i5^S2+i5^S7 and i5³+i5^S7combinations without the initiator sequence. However, even in this case, mini-y expression is not observed throughout the A5 tergite. On the other hand, when the initiator is combined with all three sequences (i5^S2+i5³+i5ⁱⁿⁱ+i5⁷), mini-y is expressed throughout the entire A5 tergite, as is tan, while the ectopic activation in A4 is absent (Fig. 5). Thus, this combination appears to be sufficient for full domain function.

We also examined the activity of the iab-5 enhancers after reporter excision (Fig. S8). The gain-of-function transformations (mis-shapen sternites and loss of trichome hairs) in the morphology of segments A3-A6 in the starting M-i5^attP platform are largely rescued by the introduction of the Mcp insulator M⁴¹³. However, as mentioned above, the pigmentation patterns in A4 and A5 are still abnormal. The former has patches of ectopic pigmentation, whereas the latter is not fully pigmented. The pigmentation patches in anterior of A4 and A5 mostly disappear in the i5^S2+i5⁷ combination. When i5^S2+i5⁷ are combined with i5³, there is a further suppression in A5 pigmentation and a loss of pigmentation in A4. Thus, the i5³ and i5⁷ sequences, in cooperation with i5^S2, can block the activation of Abd-B expression mediated by sequences in the iab-4 domain. However, as was observed when mini-y is present, the iab-5 enhancers in i5^S2, i5³ and i5⁷ are unable to direct the proper development of A5 unless the iab-5 initiator is also present. Addition of i5ⁱⁿⁱ to i5⁷, or i5^S2+i5⁷, or i5³+i5⁷ substantially expands the area of pigmentation not only in the A5 tergite, but also in A4 (Fig. S8). As was observed for the mini-y reporter, combining i5ⁱⁿⁱ with i5^S2+i5^S3+i5^S7 yields what appears to be a fully wild-type pattern of pigmentation in both A4 and A5. Thus, the i5^S2+i5^S3+i5^S7 combination blocks incorrect activity of the iab-4 and iab-5 initiators in the A4 segment, and is sufficient, in cooperation with i5ⁱⁿⁱ, for the proper stimulation of Abd-B in the A5 segment.

The temporal and spatial patterns of expression of the three BX-C homeotic genes are generated by nine regulatory domains that are arranged in the same order in the chromosome as the segments that they are thought to specify (Duncan, 1987; Karch et al., 1985; Peifer et al., 1987). Two different models have been proposed to explain the functional properties of these regulatory domains. In the first, the additive model, which was envisioned nearly 50 years ago by Lewis (1978), the regulatory domains for each segment are sequentially activated from anterior to posterior along the main body axis. Once activated, the domain is also active in all of the more posterior segments, and thus could contribute to the differentiation of these segments. This would mean that although each domain is necessary for directing the differentiation of a specific segment, it may not be sufficient. In the case of Abd-B, for example, it first turns on in PS10 (A5). In this PS (segment), the iab-5 domain is responsible for controlling Abd-B expression and it would be both necessary and sufficient for differentiation of this PS (segment). This would not be true for iab-6, which is activated in the next posterior PS11 (A7). Although iab-6 would be necessary for the differentiation of PS11 (A6), it would not be sufficient. Instead, differentiation would require the combined action of enhancers in iab-6 and iab-5 (Fig. 6). For the next posterior PS12 (A7), the iab-7 domain would be essential for differentiation; however, the requisite pattern of Abd-B expression would only be achieved by the contributions of iab-5 and iab-6. Several lines of evidence are consistent with this model. Antibody staining experiments in the embryo showed that the levels of Abd-B and the number of cells expressing this protein increase in a stepwise fashion between PS10 and PS12 (Celniker et al., 1990; Delorenzi and Bienz, 1990; Sanchez-Herrero, 1991). It also fits with the finding that mutations that inactivate an Abd-B regulatory domain result in a loss-of-function transformation in which the PS (segment) assumes the identity of the immediately anterior PS (segment). For example, an iab-6⁴ deletion (Iampietro et al., 2010) transforms PS11 into a copy of the more anterior PS10. Similarly, an iab-7 deletion, iab-7^Sz, transforms PS12 to the PS11 (Mihaly et al., 1998). In the additive model, this transformation arises because iab-5 and iab-6 already have a role in activating Abd-B in PS12 (A7) together with iab-7 in wild type. In the iab-7^Sz mutant, they still have this function in PS12 (A7), but are only able to drive a pattern of Abd-B expression appropriate for PS11 (A6) identity. This model leaves unanswered the question of whether the enhancers in different domains are strictly additive or have complementary activities. In the former case, the enhancers in, for example, iab-5, would work in concert with enhancers in iab-6 to generate a higher level of Abd-B expression in each cell in PS11 (A6). Consistent with the notion that there are redundant tissue-specific enhancers spread throughout the complex, Crosby et al. showed that a deletion that removes sequences extending from iab-4 to iab-7 induces the segments A4 through A7 to assume an A6-like identity (Crosby et al., 1993). In the latter case, the enhancers in iab-5 would drive expression in the same set of cells in both PS10 (A5) and PS11 (A6). The enhancers in iab-6 would then be responsible for activating Abd-B in a set of cells that do not express Abd-B in PS10 (A5). Consistent with this idea, a comparison of the pattern of Abd-B expression in the embryonic CNS in PS10 and PS11 suggests that there are greater number of Abd-B-positive cells in PS11.

In the second model, each regulatory domain is both necessary and sufficient to drive the expression of one of the homeotic genes in a pattern appropriate for the differentiation of a given segment (Peifer et al., 1987; Simon et al., 1990). Several observations support the notion that BX-C regulatory domains are sufficient on their own to specify segment differentiation. The first comes from deletions that span iab-5 and iab-6 (Mihaly et al., 2006). While these deletions transform both PS10 (A5) and PS11 (A6) towards a PS9 (A4) identity, they have no apparent effect on the differentiation of PS12 (A7). The second comes from boundary replacement experiments. Deletions of Fab-7 transform PS11 (A6) into a copy of PS12 (A7), but have no discernible effect on the development of PS12 (A7). Hogga et al. (2001) showed that this gain-of-function transformation can be rescued by the scs boundary (Hogga et al., 2001); however, unlike Fab-7, scs prevents iab-6 from regulating Abd-B, and PS11 (A6) is transformed into a copy of PS10 (A5). Importantly, although iab-6 is prevented from regulating Abd-B, this has no effect on the development of PS12 (A7). This result indicates that iab-7 is sufficient on its own to direct a pattern of Abd-B expression appropriate for the differentiation of PS12 (A7). Subsequent experiments in which Fab-7 was replaced by a variety of different boundaries that lack bypass activity, but block crosstalk have supported this conclusion (Hogga et al., 2001; Kyrchanova et al., 2016, 2017, 2019a).

One of the boundaries tested in these Fab-7 replacement experiments was Pita^×5 (Kyrchanova et al., 2017, 2019a). Like other heterologous boundaries, it blocks crosstalk between iab-6 and iab-7, and rescues the gain-of-function transformation of the Fab-7 boundary deletion. However, it lacks bypass activity and A6 assumes an A5 identity. A7, on the other hand, is specified correctly. As shown here, it functions in a similar fashion when used to replace the Fab-6 boundary. It blocks crosstalk between iab-5 and iab-6, and rescues the transformations A5 induced by the deletion. At the same time, A5 assumes an A4 identity. This is the result expected for a boundary element at the position of Fab-6 that fails to support bypass of the iab-5 regulatory domain. However, unlike the Pita^×5 replacement of Fab-7 which has no effect on the specification of A7, the Pita^×5 replacement of Fab-6 also disrupts the development of A6. Just like A5, the pigmentation of the A6 tergite, and the pattern of trichome hairs, resemble that seen in A4 in wild-type males. As insulators must be interposed between enhancers and/or silencers and promoters, in order to block regulatory interactions, the defects in the differentiation of the cuticle in A6 in adult males cannot be due to insulator-dependent interference with the cuticle enhancers in iab-6. Instead, Pita^×5 has to block enhancers in iab-5 from correct regulation of Abd-B not only in A5, but also in A6. This conclusion is supported by two other results. First, the defects in A6 induced by replacing Fab-6 with Pita^×5 can be partially or completely rescued by introducing several different iab-5 sequences into iab-6. This finding indicates that there are cuticle enhancers in iab-5 that can support the proper differentiation of the A6 tergite if they are not blocked by the Pita^×5 insulator. Second, deletion of the iab-5 initiator expectedly inactivates the iab-5 domain, and A5 is transformed into a copy of A4; however, the partial loss-of-function transformation of A6 is also observed. These loss-of-function transformations indicate that the formation of the characteristic sex-specific cuticular features in the A6 segment requires regulatory elements located in iab-5. Taken together, these findings indicate that iab-6 is not, on its own, able to direct expression of Abd-B in the manner that is required for differentiation of the A6 cuticle in male flies. Instead, the adjacent iab-5 domain is needed to supplement the intrinsic enhancer activities of the iab-6 domain. In this respect, the functioning of iab-5 and iab-6 domains fit well with the model for BX-C regulation suggested by Lewis (1978). On the other hand, this does not seem to be true for iab-7, which appears to be both necessary and sufficient for the development of PS12 and A7 (Fig. 6).

Although our results demonstrate that the proper differentiation of A6 requires several distinct enhancers in iab-5 to drive Abd-B expression in the appropriate manner, the iab-5 domain alone cannot substitute for iab-6. Experiments by Iampietro et al. (2010) show that deletion of the iab-6 initiator results in a loss-of-function transformation of A6 into A5. Thus, it is possible that the differentiation of the A6 cuticle in adult males requires that enhancers in iab-5 and iab-6 work in concert (Fig. 6). In the case of iab-5, we have found that at least four distinct DNA sequences (i5^S2, i5³, i5⁷ and the iab-5 initiator i5ⁱⁿⁱ) are required for proper differentiation of A5. Several lines of evidence (orientation or position dependent activation of the yellow reporter) suggest that the i5^S2, i5³ and i5⁷ sequences contain not only tissue-specific enhancers but also silencers that are not coincident with regions bound by Polycomb proteins (Fig. 2). With respect to the cells in the cuticle in which these enhancers and silencers are active, it would appear that they have overlapping rather than completely distinct activities. The i5 sequences also seem to help the Mcp⁴¹³ boundary block interactions between the iab-4 and iab-5 initiators and regulatory elements. At this point, it is not yet clear whether there are also multiple physically distinct cuticle enhancers in iab-6. Even if this is not true, it would appear that Abd-B expression in the cells in PS10 and PS11 that give rise to adult male cuticle in A5 and A6 depends on interactions with several physically distinct enhancer elements. Whether these interactions occur simultaneously or only individually remains to be determined. Likewise, it is not clear whether the effects are strictly additive or whether the enhancers in iab-5 are active in sets of cells that are distinct from the cells in which the iab-6 enhancers are active.

The deletions were obtained by CRISPR/Cas9 method (see Fig. S1). As a reporter, we used pHD-DsRed vector (Addgene plasmid #51434). The plasmid was constructed in the following order: proximal arm-attP-lox-3×Р3:DsRed-SV40polyA-lox-distal arm. Arms were amplified by PCR from DNA isolated from Oregon line. For generation of the i5^1attP deletion, homology arms were obtained by DNA amplification between the following primers: TGTCGAGGTCCCGAAATG and ACGTCACTTGGCTGAAATGC; and CAGACAGGTCCATCGGGG and TTGTTGAGGGTTGGTTGTG. For M^3attP, the primers were: ATAACTAGTCCTAAATTACGACCACGAC and ATACTCGAGCCCATAAACAGCACGGC; and ATAGCGGCCGCATTTTAATCGAGCCATC and CGAGAATTCCTAGAATGAGTAG. The guide RNAs were selected using the program ‘CRISPR optimal target finder’ (O'Connor-Giles Lab). For i5^1attP deletion, the primers were: TTTCGGGACCTCGACACGTT_TGG and TTGGCCCCGATGGACCTGTC_TGG. For M^3attP deletion, the primers were: CACTGACAGAGTCAGGCTCG_TGG and CATACTTGCCCCGTACTTGC_CGG. The breakpoints of the designed deletion were: for i5^1attP, 3R:16877730..16879686 (1957 bp); and for M^3attP, 3R:16872084..16868751 (3333 bp), according Genome Release r6.36.

To generate the deletions, the plasmid construct was injected into embryos: y¹ M{Act5C-Cas9.P.RFP-}ZH-2A w¹¹¹⁸ DNAlig4[169] (BL 58492 stock, Bloomington Drosophila Stock Center) together with two gRNAs. The F0 progeny were crossed with y w; TM6/MKRS flies. Flies with potential deletions were selected on the basis of dsRed-signal in the posterior part of their abdomens and these flies were crossed with y w; TM6/MKRS flies. All independently obtained flies with dsRed reporter were tested by PCR. The successful deletions events were confirmed by sequencing of PCR products. Next, dsRed reporter was deleted by Cre/lox recombination.

To create M-i5^attP (Fig. S4) the i5^1attP and M^3attP were crossed with line expressing Cre recombinase (#1092, Bloomington Drosophila Stock Center). Then, i5^1attP/+; CyO, P{w[+mC]=Crew}DH1/+ was crossed with M^3attP/+; CyO, P{w[+mC]=Crew}DH1/+. Next, the i5^1attP/M^3attP; CyO, P{w[+mC]=Crew}DH1/+ males and females were crossed with each other and male offspring with the expected phenotypes were crossed with y w; TM6/MKRS flies. The deletion was confirmed by PCR and sequencing.

The replacement vector was a plasmid with the mini-yellow and mCherry reporters as shown in Fig. S1. The iab-5 fragments were obtained by PCR amplification. Their coordinates are: i5¹, 112812-113529; i5², 111101-113245; i5³, 109349-111346; i5⁴, 107265-109694; i5⁵, 105709-107851; i5⁵, 105030-105750; i5⁷, 101629-104152; i5ⁱⁿⁱ, 104011-105035; i5^S1, 113227-113824; i5^S2, 112455-113245; i5^S3, 111607-112829; i5^S4, 104016-104537; i5^S5, 103516-104152; and i5^S6, 101629-102685 (according to the published sequences of the Bithorax complex; Martin et al., 1995).

Integration of the plasmids in the landing platforms was achieved by injecting the plasmid and a vector expressing the фC31 recombinase into embryos of yw; i5^1attP/i5^1attP, or yw; M^3attP/M^3attP, or yw; M-i5^attP/M-i5^attP lines. The successful integrations were selected on the basis of expression of mini-y in abdominal segments. The integration of the replacement DNA fragments was confirmed by PCR. The yellow and mCherry reporters were excised by Cre-mediated recombination between the lox sites.

Cuticle preparations were carried out as described previously (Postika et al., 2018).

We are grateful to François Karch for discussing the results and editing the discussion. We thank Farhod Hasanov for fly injections, Kate O'Connor-Giles for the pHD-DsRed plasmid (Addgene plasmid #51434), Bloomington Drosophila Stock Center for Drosophila lines. We thank Natalia Klimenko for bioinformatics analysis.

The peer review history is available online at https://journals.biologists.com/dev/article-lookup/doi/10.1242/dev.199827