Boundaries support specific long-distance interactions between enhancers and promoters in Drosophila Bithorax complex

Drosophila bithorax complex (BX-C) is one of the best model systems for studying the role of boundaries (insulators) in gene regulation. Expression of three homeotic genes, Ubx, abd-A, and Abd-B, is orchestrated by nine parasegment-specific regulatory domains. These domains are flanked by boundary elements, which function to block crosstalk between adjacent domains, ensuring that they can act autonomously. Paradoxically, seven of the BX-C regulatory domains are separated from their gene target by at least one boundary, and must “jump over” the intervening boundaries. To understand the jumping mechanism, the Mcp boundary was replaced with Fab-7 and Fab-8. Mcp is located between the iab-4 and iab-5 domains, and defines the border between the set of regulatory domains controlling abd-A and Abd-B. When Mcp is replaced by Fab-7 or Fab-8, they direct the iab-4 domain (which regulates abd-A) to inappropriately activate Abd-B in abdominal segment A4. For the Fab-8 replacement, ectopic induction was only observed when it was inserted in the same orientation as the endogenous Fab-8 boundary. A similar orientation dependence for bypass activity was observed when Fab-7 was replaced by Fab-8. Thus, boundaries perform two opposite functions in the context of BX-C – they block crosstalk between neighboring regulatory domains, but at the same time actively facilitate long distance communication between the regulatory domains and their respective target genes. Author Summary Drosophila bithorax complex (BX-C) is one of a few examples demonstrating in vivo role of boundary/insulator elements in organization of independent chromatin domains. BX-C contains three HOX genes, whose parasegment-specific pattern is controlled by cis-regulatory domains flanked by boundary/insulator elements. Since the boundaries ensure autonomy of adjacent domains, the presence of these elements poses a paradox: how do the domains bypass the intervening boundaries and contact their proper regulatory targets? According to the textbook model, BX-C regulatory domains are able to bypass boundaries because they harbor special promoter targeting sequences. However, contrary to this model, we show here that the boundaries themselves play an active role in directing regulatory domains to their appropriate HOX gene promoter.


Introduction
The three homeotic (HOX) genes in the Drosophila Bithorax complex (BX-C),
Each regulatory domain contains an initiator element, a set of tissue-specific enhancers and Polycomb Responsible Elements (PREs) and is flanked by boundary/insulator elements (Fig 1A; Maeda and Karch 2006). BX-C regulation is divided into two phases, initiation and maintenance [15,16]. During the initiation phase, a combination of gap and pairrule proteins interact with initiation elements in each regulatory domain, setting the domain in the on or off state. In PS10, for example, the iab-5 domain, which regulates Abd-B, is activated by its initiator element, while the more distal Abd-B domains, iab-6 to iab-8 are set in the off state ( Fig 1B). In PS11, the iab-6 initiator activates the domain, while the adjacent iab-7 and iab-8 domains are set in the off state. Once the gap and pair-rule gene proteins disappear during gastrulation, the on and off states of the regulatory domains are maintained by Trithorax (Trx) and Polycomb (PcG) group proteins, respectively [17,18].
While these findings indicate that boundaries are needed to ensure the functional autonomy of the regulatory domains, their presence also poses a paradox [14,28]. Seven of the nine BX-C regulatory domains are separated from their target HOX gene by at least one 5 intervening boundary element. For example, the iab-6 regulatory domain must "jump over" or "bypass" Fab-7 and Fab-8 in order to interact with the Abd-B promoter (Fig 1A). That the blocking function of boundaries could pose a significant problem has been demonstrated by experiments in which Fab-7 is replaced by heterologous elements such as scs, gypsy or multimerized binding sites for the architectural proteins dCTCF, Pita or Su(Hw) [26,[29][30][31].
In these replacements, the heterologous boundary blocked crosstalk between iab-6 and iab-7 just like the endogenous boundary, Fab-7. However, the boundaries were not permissive for bypass, preventing iab-6 from regulating Abd-B.
A number of models have been proposed to account for this paradox. One is that BX-C boundaries must have unique properties that distinguish them from generic fly boundaries.
Since they function to block crosstalk between enhancers and silencers in adjacent domains, an appealing idea is that they would be permissive for enhancer/silencer interactions with promoters ( Fig 1B). However, several findings argue against this notion. For one, BX-C boundaries resemble those elsewhere in the genome in that they contain binding sites for architectural proteins such as Pita, dCTCF, and Su(Hw) [24,[31][32][33][34][35]. Consistent with their utilization of these generic architectural proteins, when placed between enhancers (or silencers) and a reporter gene, BX-C boundaries block regulatory interactions just like boundaries from elsewhere in the genome [20,[36][37][38][39][40][41][42]. Similarly, there is no indication in these transgene assays that the blocking activity of BX-C boundaries are subject to parasegmental regulation. Also arguing against the idea that BX-C boundaries have unique properties, the Mcp boundary, which is located between iab-4 and iab-5, is unable to replace Fab-7 [31]. Like the heterologous boundaries, it blocks crosstalk, but it is not permissive for bypass. A second model is that there are special sequences, called promoter targeting 6 sequence (PTS), located in each regulatory domain that actively mediate bypass Chen et al. 2005;Lin et al. 2003). While the PTS sequences that have been identified in iab-6 and iab-7 enable enhancers to "jump over" an intervening boundary in transgene assays, they do not have a similar function in the context of BX-C and are completely dispensable for Abd-B regulation [19,30].
A third model (Fig 1C) is suggested by transgene "insulator bypass" assays [46,47]. In one version of this assay, two boundaries instead of one are placed in between an enhancer and the reporter. When the two boundaries pair with each other, the enhancer is brought in close proximity to the reporter, thereby activating rather than blocking expression. Consistent with a possible role in BX-C bypass, these pairing interactions can occur over large distances and even skip over many intervening boundaries [48][49][50][51]. The transgene assays point to two important features of boundary pairing interactions that are likely to be relevant in BX-C.
First, pairing interactions are specific. For this reason boundaries must be properly matched with their neighborhood in order to function appropriately. A requirement for matching is illustrated in transgene bypass experiments in which multimerized binding sites for specific architectural proteins are paired with themselves or with each other [52]. Bypass was observed when multimerized dCTCF, Zw5 or Su(Hw) binding sites were paired with themselves; however, heterologous combinations (e.g. dCTCF sites with Su(Hw) sites) did not support bypass.
The fact that both blocking and bypass activities are intrinsic properties of fly boundaries suggests that the BX-C boundaries themselves may facilitate contacts between the regulatory domains and their target genes ( Fig 1C). Moreover, the non-autonomy of both blocking and bypass activity could potentially explain why heterologous Fab-7 replacements 7 like gyspy and Mcp behave anomalously while Fab-8 functions appropriately. Several observations fit with this idea. There is an extensive region upstream of the Abd-B promoter that has been implicated in tethering the Abd-B regulatory domains to the promoter [53][54][55][56] and this region could play an important role in mediating bypass by boundaries associated with the distal Abd-B regulatory domains (iab-5, iab-6, iab-7). Included in this region is a promoter tethering element (PTE) that facilitates interactions between iab enhancers and the Abd-B promoter in transgene assays [57,58]. Just beyond the PTE is a boundary element, AB-I. In transgene assays AB-I mediates bypass when combined with either Fab-7 or Fab-8. In contrast, a combination between AB-I and Mcp fails to support bypass [59,60]. The ability of both Fab-7 and Fab-8 to pair with AB-I is recapitulated in Fab-7 replacement experiments.
Moreover, its' bypass but not blocking activity is orientation-dependent. When inserted in the same orientation as the endogenous Fab-8 boundary, it mediates blocking and bypass, while it does not support bypass when inserted in the opposite orientation.
In the studies reported here we have tested this model by replacing the endogenous Mcp boundary with heterologous boundaries. Mcp defines the border between the set of regulatory domains that control abd-A and those that control Abd-B expression ( Fig 1A).

Substitution of Mcp by an attP integration site in the BX-C
The Mcp boundary is defined by 340 bp core sequence that has enhancer blocking activity in transgene assays [36] and blocks crosstalk between iab-6 and iab-7 when substituted for Fab-7 [31]. Located just distal to the boundary is a PRE that negatively regulates the activity of the iab-5 enhancers [61]. We used CRISPR to delete a 789 bp DNA  [62].

Multimerized dCTCF sites substitute for Mcp
The Mcp boundary marks the division between the set of regulatory domains that control the abd-A and Abd-B genes ( Fig 1A). The Abd-B protein is master regulator of pigmentation in the male abdominal A5 and A6 segments due to the regulation of genes involved in melanin synthesis [63][64][65]. Flies carrying the null y 1 allele lack black melanin but still have brown melanin that is also regulated by the Abd-B protein [65,66]. In order to be able to recover recombinants and also to monitor the blocking activity of the replacement sequence and the on/off state of the iab-5 domain, we included a minimal yellow (mini-y) reporter in our Mcp replacement construct.
The mini-y reporter consists of the cDNA and the 340 bp yellow promoter and lacks the wing, body and bristle enhancers of the endogenous yellow gene. As a result, activity of the mini-y reporter depends upon proximity to nearby enhancers. Expression of the mini-y reporter was examined in the y 1 background.
Based on previous studies [5,22,67], we assume that the activity of this reporter will be determined by the activity state of the iab-5 domain. When iab-5 is off in PS9 and more anterior parasegments, the mini-y reporter will also be off. When iab-5 is on in PS10 and more posterior parasegments, the mini-y reporter will be expressed. This parasegment-specific regulation of the reporter activity will be reflected in the segmental pattern of black melanin pigmentation in the adult cuticle. In replacements in which blocking activity is compromised, mini-y will be expressed in PS9 in adults the A4 tergite will be black, just like the A5 and A6 tergites.
When we replaced the Mcp deletion by the iab-5 PRE alone (Mcp PRE ) the mini-y reporter was active not only in A5 (PS10) and more posterior segments, but also in A4 (PS9).
As shown in Fig 2, the pigmentation in A4 is black like that in A5 indicating that the reporter is expressed in both segments (Fig 2). This finding shows that, similar to classical Mcp deletions, the Mcp PRE replacement does not have blocking activity. In these Mcp deletions iab-5 is ectopically activated in PS9 by the iab-4 initiator and as a consequence there is a gain-of-function transformation in parasegment identity from PS9 to PS10. We used two approaches to test whether this was true for the Mcp PRE replacement. In the first, we excised the mini-y reporter and introduced an y + X chromosome. Since Abd-B directly regulates endogenous yellow expression in abdomen [64,66], a transformation of PS9 into PS10 should be accompanied by a PS10-like pattern of pigmentation. Fig 2 shows that this is indeed the case. We also examined the pattern of Abd-B protein expression in the embryonic CNS. In wild type embryos Abd-B is not expressed is PS9, while it is expressed at low levels in PS10.  2) and the mini-y reporter or for that matter in wild type y 1 males (see Fig 4). The presence of yellow-brown pigmentation throughout most of the A4 tergite suggests that the cells in this segment (PS9) are not properly specified. This is the case. When the mini-y 13 reporter was excised and replaced by the endogenous X-linked y + gene, the A4 tergite has a black pigmentation like A5 and A6 (Fig 4). Since expression of the yellow gene is controlled by Abd-B, this observation indicates that Abd-B must be ectopically activated throughout A4.
Antibody staining experiments of the CNS in Mcp F7 embryos indicate that this inference is correct (Fig 3B).
A simple interpretation of these findings is that Mcp F7 is unable to block crosstalk

Ectopic expression of Abd-B in A4 (PS9) requires a functional iab-4 domain
In the Fab-7 replacement experiments, the relative orientation of the Fab-8 boundary was thought to be important because it determined whether the chromatin loops formed between the replacement boundary and the AB-I element and/or the PTE sequence upstream of the Abd-B transcription start site were circle loops or stem loops [30,74]. In the forward orientation circle loops are expected to be formed and in this configuration, the downstream iab-5 regulatory domain is brought into close proximity with the Abd-B promoter. In the reverse orientation, iab-6 and iab-7 are predicted to form stem loops, and this configuration would tend to isolate the iab-5 regulatory domain from the Abd-B promoter.  (Fig 3B). Interestingly, the loss of trichomes along the anterior margin of the A6 tergite in Mcp F8 also seemed to depend on a functional iab-4 domain. As can be seen in Fig 5, the trichome pattern in the A6 tergite of iab-4 Δ Mcp F8 flies resembled that of wild type.

Conclusion
Mcp defines the boundary between the regulatory domains that control expression of abd-A and Abd-B. In this location, it is required to block crosstalk between the flanking domains iab-4 and iab-5, but it does not need to mediate bypass. In this respect, it differs from the boundaries that are located within the set of regulatory domains that control either abd-A are transformed towards an A6 identity, while A6 is also misspecified. Similar though somewhat less severe effects are observed when Fab-7 is inserted in the reverse orientation.
Although the mechanisms responsible for these novel phenotypic effects are uncertain, a plausible idea is that pairing interactions between the Fab-7 insert and the endogenous Fab-7 boundary disrupt the normal topological organization of the regulatory domains in a manner similar to that seen in boundary competition transgene assays [76]. Further studies will be required to test this idea.

Generation of Mcp attP by CRISPR/Cas9-induced homologous recombination
The backbone of the recombination plasmid was designed in silico and contains Genewiz. The two multiple cloning sites MCS5 and MCS3 were used to clone homology 20 arms into this plasmid. The orientations of two the attP sites are inverted relative to each other and serve as targets for фC31-mediated recombination mediated cassette exchange [77]. The 3x3P-EGFP reporter [78] was introduced as a means to isolate positive recombination events.
The Flp-recombinase target FRT [79] were includedl for the deletion of the selectable miniyellow marker after recombination mediated cassette exchange.
Homology arms were PCR-amplified from y w genomic DNA using the following

Generation of iab-4 Δ by CRISPR/Cas9-induced homologous recombination
For generating dsDNA donors for homology-directed repair we used pHD-DsRed Targets for Cas9 were selected using "CRISPR optimal target finder"the program from O'Connor-Giles Lab. The recombination plasmid was injected into Mcp F8 vasa-Cas9 embryos together with two gRNAs containing the following guides: ATAGCAAGTAGGAGTGGAGT and GAACTTCTTCCCTTTCCGAGCGG.

Cuticle preparations
Adult abdominal cuticles of homozygous enclosed 3-4 day old flies were prepared essentially as described in (Kyrchanova et al. 2017) and mounted in 100% glycerol. Photographs in the bright or dark field were taken on the Nikon SMZ18 stereomicroscope using Nikon DS-Ri2 digital camera, processed with ImageJ 1.50c4 and Fiji bundle 2.0.0-rc-

Embryo immunostaining
Primary antibodies were mouse monoclonal anti-Abd-B at 1:100 dilution (1A2E9,         Mcp region that was deleted (coordinates according to complete sequence of BX-C in SEQ89E numbering) and replaced by two attP sites for the integration of the tested constructs. 3xP3-eGFP was used as a marker gene. frt site was used for excision of yellow