Introduction

Studies on lampbrush chromosomes were the first to suggest that the key organizing principle for chromosomes in animals is the subdivision of the chromatin into a series of topologically independent loops (now called TADs) that extend from the main chromosomal axis (Gall and Callan, 1962; Callan, 1986; Callan, 1987). Each lampbrush chromosome has a stereotypic pattern of loops, and in most loops there is only a single actively transcribed gene. The loops are anchored to the main axis of the chromosome, and the anchors pull apart when the chromosomes are stretched (Callan, 1986, 1987). Since then, a variety of approaches have shown that TADs are pervasive features of chromosomes in multicellular organisms (Rao et al., 2014; Dekker and Heard, 2015; Eagen et al., 2015; Stadler et al., 2017; Hsieh et al., 2021; Krietenstein et al., 2020; Goel et al., 2023; Dolstein et al., 2025). In flies, TADs range in size from only a few kb to over 100 kb, while recent MicroC and Region Capture MicroC (RCMC) studies indicate that the scale is similar in mammals. The arrangement of TADs along a given chromosome tends to be invariant and is largely, but not completely, independent of cell type or developmental stage (Bing et al. 2024; Ke et al., 2024; Dolstein et al., 2025; Wang et al., 2026).

This regular and heritable organization is generated by a special class of cis-acting elements called boundaries or insulators. Boundary elements were first discovered and have been most fully characterized in Drosophila; however, similar elements have been identified in other species (Ghirlando et al., 2012; Ghirlando and Felsenfeld, 2016; Chetverina et al., 2017; Matthews and White 2019; Cavaleiro et al., 2021). Boundary elements in flies consist of one or more large (>150 bp) nuclease hypersensitive regions and can span DNA sequences of up to 1.5 kb in length. These nuclease-hypersensitive regions are targets for a large and diverse collection of DNA binding proteins that have been implicated in boundary function (Parkhurst et al., 1988; Gasner et al., 1999; Golovin et al., 2007; Maksimento et al., 2015; Zolotarev et al, 2016; Fedotova et al., 2017; Chetverina et al. 2021; Bonchuk et al., 2021; Bonchuk et al., 2023). Of these, the most prevalent class are polydactyl C2H2 zinc finger proteins {e.g., CTCF, Zw5, Su(Hw), Pita, Zipic, Clamp, Zad1}(Bonchuk et al., 2021; Fedotova et al., 2017). Other boundary factors include several BEN domain proteins (Insv, Elba1-2), GAF, Ibf1/2, BEAF, CP190 and the 31 Mod(mdg4) isoforms (Zhao et al., 1995; Bonchuk et al., 2011; Aoki et al., 2012; Dai et al. 2013; Cuartero et al., 2013; Avva and Hart, 2016). As there are over 200 mostly uncharacterized polydactyl C2H2 zinc finger proteins, including some 70 proteins that have the ZAD homodimerization domain, encoded in the fly genome (Bonchuk et al., 2021), it is likely that a large number of chromosomal architectural proteins have yet to be identified.

In flies, TADs are formed by boundary:boundary pairing. The first evidence of physical pairing came from experiments showing that when either the Bithorax (BX-C) boundary Mcp or the gypsy transposon su(Hw) boundary are included in transgenes, they can induce long-distance regulatory interactions (PcG-dependent silencing and enhancer activation) between transgene inserts located many megabases (Mb) apart (Vazquez et al., 1993; Sigrest and Pirrotta, 1997; Muller et al., 1998). Regulatory interactions were even observed for transgenes inserted on different chromosomes. Subsequent in vivo imaging studies showed that 2, 3, and even 4 copies of these distant transgenes co-localize in imaginal disc cells (Vazquez et al., 2006; Li et al., 2011). The parameters governing boundary pairing interactions have been defined using boundary bypass, transvection and boundary competition assays (Cai and Shen, 2001; Muravyova et al., 2001; Gohl et al., 2011; Li et al., 2018). Though boundary elements can engage in promiscuous pairing interactions, there are clear partner preferences. For example, in boundary bypass assays using boundaries from the Abd-B region of BX-C, Kyrchanova et al. (2011) found that Mcp can pair with Fab-8 but not with Fab-7. The other key feature is orientation dependence (Kyrchanova et al., 2008a). With a few exceptions, pairing interactions are orientation-dependent. Orientation dependence can differ depending upon whether boundaries are pairing with themselves or with other boundaries. The self-pairing interactions that have been examined in detail are head-to-head (Kyrchanova et al., 2008a; Kyrchanova et al., 2008b; Fujioka et al., 2016). Self-pairing interactions take place in trans, and these interactions are likely responsible for the alignment and pairing of sister chromosomes and homologs (Fujioka et al., 2016; Alhai Abed et al., 2019; Viets et al., 2019; Child et al., 2021). That these interactions are head-to-head is not surprising, as head-to-tail self-pairing would generate unpaired loops and potentially disrupt transvection (Fujioka et al., 2016). Heterologous pairing interactions typically take place in cis, and unlike self-pairing, can be either head-to-head or head-to-tail. The former generates a TAD with a circle-loop topology while the latter generates a TAD with a stem-loop topology (Kyrchanova et al., 2008a; Fujioka et al., 2016; Li et al., 2018; Bing et al., 2024; Ke et al., 2024).

Thus far, three different mechanisms are thought to be involved in boundary:boundary pairing interactions. The simplest mechanism in flies is based on shared binding sites for architectural proteins that can form dimers or multimers (Bonchuk et al. 2021; Fedotova et al. 2017). Most of the factors implicated in fly boundary function form such dimers or higher order multimers. A second mechanism for physically linking two boundaries is protein:protein interaction between heterologous DNA binding factors (Blanton et al. 2003). Yet a third mechanism would be proteins that function to bridge DNA binding proteins that are associated with each boundary. Several proteins that can potentially function as “linkers” have been identified. Included in this group are two proteins, CP190 and the 67.1 isoform of Mod(mdg4), which interact with the polydactyl zinc finger protein Su(Hw) (Harrison et al.,1993; Kim et al., 1996; Pai et al., 2004; Golovnin et al., 2007; Vogelmann et al., 2014; Melnikova et al., 2017; Melnikova et al., 2018; Melnikova et al., 2019; Kaushal et al., 2022; Golovnin et al., 2023). CP190 and Mod(mdg4) family proteins, including the 67.1 isoform, have BTB domains that assemble into dimers (CP190) or multimers (Mod(mdg4)), and thus could form a bridge linking Su(Hw) proteins bound to sites in different boundaries.

In the studies reported here, we have tested the role of the chromosomal architectural protein Su(Hw) in generating direct physical interactions between boundary elements. As a model system for this analysis we used the two boundaries, nhomie and homie, that are responsible for the formation of the ∼16 kb TAD that encompasses the Drosophila even-skipped (eve) gene (Fujioka et al., 2009; Fujioka et al., 2016; Ke et al., 2024). Previous work showed that nhomie and homie pair with themselves head-to-head, while they pair with each other head-to-tail (Fujioka et al., 2016). Like the Mcp and su(Hw) boundaries, nhomie and homie are also able to mediate long-distance regulatory interactions. We have taken advantage of this long-distance activity to probe the pairing interactions of the two eve boundaries. In one experimental paradigm, we used an attP site located about a dozen TADs upstream from eve (at -142 kb, in the 1^st exon of the hebe gene) to introduce a transgene containing two reporters driven by the eve promoters, eve-LacZ and eve-GFP (see Figure 1). When nhomie or homie is inserted between the two reporters, one of the two reporters is activated more strongly by eve enhancers than the other. Activation depends on the 5’◊3’ orientation of the boundary relative to the two reporters in the transgene, but not on the orientation of the transgene in the chromosome. Thus, eve-LacZ is preferentially activated when it is positioned “downstream” of nhomie or “upstream” of homie. The same relationship holds for eve-GFP. We also found that the ability of a minimal 271 bp homie fragment (DEF) to mediate eve enhancer-dependent reporter activation in the -142 kb assay depends on a predicted Su(Hw) binding site located in the D element (Fujioka et al., 2025). Since nhomie also has a predicted Su(Hw) binding site, this shared DNA recognition sequence could serve two functions: namely nhomie and homie self-pairing and heterologous pairing. In the studies reported here we have tested this idea. We have also analyzed interactions between the gypsy transposon Su(hw) insulator and the two eve boundaries.

Figure 1.

Results

nhomie-dependent reporter activation

To assay the long-distance pairing activity of the nhomie boundary, we inserted a “minimal” (602 bp) version of nhomie into the dual reporter in each orientation. The resulting transgenes were then integrated into the -142 kb attP site to give four different transgene configurations: G-nhomie-L and G-eimohn-L (Figure 1 and Figure 1—figure supplemental 1), and L-nhomie-G and L-eimohn-G (Figure 1—figure supplemental 2). In the transgenes G-nhomie-L and G-eimohn-L (diagrammed in Figure 1A), the LacZ reporter is located between the transgene nhomie and the eve TAD. Conversely, in the transgenes L-nhomie-G and L-eimohn-G (shown in Figure 1—figure supplemental 2), the GFP reporter is located between the transgene nhomie and the eve TAD. Upstream of the attP insertion site are enhancers for the hebe gene, which are active in stage 12-16 embryos (Fujioka et a., 2025). Shown in Figure 1 is a negative control transgene containing phage lambda DNA sequences (G-lambda-L) instead of nhomie. It is oriented so that LacZ is on the eve side of the lambda DNA, while GFP is on the same side as the hebe enhancers.

In the transgene G-nhomie-L (shown in Figure 1—figure supplemental 1A), while in stage 13 embryos, LacZ expression is driven by the eve neurogenic, mesodermal, and anal plate ring (APR) enhancers (Figure 1—figure supplemental 1A). Quantification of LacZ and GFP expression in the APR confirmed the preferential expression of LacZ (Figure 1—figure supplemental 1A), the eve enhancers drive very little GFP expression from the G-nhomie-L transgene. However, since the GFP reporter is on the same side of nhomie as the hebe enhancers, it is activated by the hebe enhancers in stage 13 embryos. The GFP reporter is also activated by the hebe enhancers in the lambda DNA control (Figure 1A); however, the control differs from G-nhomie-L in that the LacZ reporter is not insulated from the hebe enhancer, and they also activate its expression.

Reporter activation depends upon the orientation of the nhomie boundary with respect to the reporters in the transgene, and not on the orientation of the transgene in the chromosome. In G-eimohn-L, LacZ is still on the eve side of nhomie while GFP is located on the other side. However, since nhomie is inverted, the GFP reporter is now “downstream” of the transgene boundary. In this configuration the GFP reporter is activated by the eve stripe enhancers at the blastoderm stage (Figure 1—figure supplemental 1B), and by the tissue-specific enhancers in stage 13 embryos (Figure 1—figure supplemental 1B). It will also be activated by the hebe enhancers, while the LacZ reporter is not. Quantification of LacZ and GFP expression in the APR confirmed this preferential expression of GFP rather than LacZ (Figure 1D, G-eimohn-L).

An analogous set of results is observed for the two transgenes L-nhomie-G and L-eimohn-G, in which GFP is on the eve side of the nhomie, and LacZ is on the hebe enhancer side. For L-nhomie-G, the eve enhancers activate GFP while LacZ is regulated by the hebe enhancer, since GFP is in the “downstream” position (Figure 1—figure supplemental 2A). When the orientation of nhomie is flipped to give L-eimohn-G (LacZ is now in the “downstream” position), LacZ is preferentially activated by both the eve and hebe enhancers, while GFP is not (Figure 1—figure supplemental 2B).

While orientation-dependent pairing of the transgene nhomie with the endogenous nhomie and homie boundaries results in the preferential activation of the “downstream” reporter, the close physical proximity of the reporter on the “upstream” side of the transgene nhomie to the enhancers in the eve TAD can also trigger its activation. Activation of the upstream reporter is quite weak (see Figure 1) and is best visualized when LacZ is “upstream” of the transgene nhomie, and the digoxigenin procedure is used for in situ hybridization. This is shown for the G-nhomie-L and G-eimohn-L pair in Figure 1—figure supplemental 1. As was observed with HCR-FISH probes, only LacZ transcripts are detected in G-nhomie-L embryos using digoxigenin in situ hybridization. However, when GFP is downstream of the transgene nhomie in G-eimohn-L and should be activated by the eve enhancers, we detect not only GFP but also LacZ transcripts. The LacZ probe in the digoxigenin experiment is ∼2.5x longer than the GFP probe, and this difference likely explains why LacZ transcripts are observed in G-eimohn-L embryos. Similar results are obtained for the second pair of nhomie inserts (L-nhomie-G and L-eimohn-G): when GFP is downstream of the transgene nhomie, a low level of LacZ transcripts is detected (Figure 1—figure supplemental 2).

Physical interactions between the reporters at -142 kb and the eve TAD

We used MicroC to probe the physical interactions between the nhomie transgenes at -142 kb and sequences in the eve TAD. Figure 2A shows the MicroC contact profile in the chromosomal segment between the eve TAD and the attP site in the 1^st exon of the hebe gene, as well as a blowup of both the eve TAD and the insertion site. Note that there are multiple TADs in the 142 kb chromosomal segment between eve and the insertion site at the beginning of the hebe gene. As reported previously for the lambda DNA control (Bing et al. 2024), weak physical interactions can be detected in 12-16 hr embryos collected at 25°C (Figure 2B and 2C) even though we do not observe activation of either reporter by the eve enhancers (Figure 1A). We suspect that the weak contacts seen for the lambda control may be mediated by the two eve promoters in the transgene.

Figure 2.

A different result is observed for the two nhomie-containing transgenes, G-nhomie-L and G-eimohn-L. In both cases, the physical interactions in 14-16 hr embryos recapitulate the pattern of activity of the transgene reporters. For G-nhomie-L, the LacZ reporter preferentially interacts with sequences in the eve TAD. This interaction between the transgene and eve is shown in Figure 2D and E. As can be seen in the blow-up, the frequency of crosslinking events between LacZ and sequences in the eve TAD is much higher than those between GFP and the eve TAD.

While the LacZ reporter preferentially interacts with sequences in the eve TAD, contacts between the GFP reporter and the eve TAD are much more frequent than those observed for either reporter in the lambda control (compare Figure 2B and C with Figure 2D and E). This difference is likely due to the fact that nhomie interactions with eve nhomie and homie brings the GFP reporter into sufficiently close proximity for it to interact with sequences in the eve TAD, even though the pairing orientation of nhomie with the eve boundaries favors interactions with LacZ.

When nhomie is inverted in the transgene to give G-eimohn-L, the pattern of physical interactions between the reporters and sequences in the eve TAD is the opposite of that observed for G-nhomie-L. The primary interactions are between the GFP reporter and the eve TAD while the secondary interactions are between LacZ and the eve TAD (Figure 2F and G). This can be seen by comparing the blowups for Figure 2E and 2G.

The distinct topology of the loops generated when the nhomie elements in G-nhomie-L and G-eimohn-L interact with nhomie and homie in the eve TAD results in other differences in the contact patterns. For example, in the G-nhomie-L insert, GFP comes into contact with the TADs immediately upstream (blue arrowhead in Figure 2E) and downstream (green arrowhead) of the eve TAD. When the orientation of nhomie is flipped so that the GFP reporter is activated instead of LacZ, the interaction pattern with neighboring sequences changes. In this case, it is LacZ that interacts with sequences in the TADs upstream (blue arrowhead in Figure 2G) and downstream (green arrowhead) of the eve TAD.

Viewpoints from the LacZ and GFP reporters

To further document that boundary orientation in the transgene determines which of the two reporters preferentially interacts with the eve TAD, we used the MicroC data to generate “virtual 4C” viewpoints from the GFP reporter (Figure 3A) or LacZ reporter (Figure 3B) in the G-nhomie-L insert, and the GFP (Figure 3C) and LacZ (Figure 3D) reporters for the G-eimohn-L insert. These “virtual 4C” data can also be visualized as 1-dimensional read coverage tracks and the average contact counts between the reporters and the eve locus are shown in Figure 3E. For the G-nhomie-L insert, the LacZ reporter interacts strongly with sequences within the eve TAD, while there is little interaction with sequences in TADs located beyond either the nhomie or homie boundaries. While the GFP reporter also interacts with sequences in the eve TAD, the interactions are highest near the boundaries, and lower within the eve TAD. Additionally, the GFP reporter contacts sequences in the TADs flanking the eve gene. A similar pattern is observed for G-eimohn-L, except in this case, GFP preferentially interacts with the eve TAD, and the LacZ reporter contacts areas flanking the eve TAD. The different topologies of the loops generated by interaction between nhomie in G-nhomie-L and G-eimohn-L and nhomie / homie in the eve TAD are shown in Figure 3F.

Figure 3.

The Su(Hw) site in nhomie is required for long-distance regulatory interactions

Fujioka et al. (2025) showed that the predicted Su(Hw) binding site in a minimal 271 bp homie boundary (called DEF) is required for mediating long-distance regulatory interactions in the -142 kb assay. Since nhomie also has a predicted Su(Hw) binding site, we tested whether mutations in this site impact regulatory interactions in the -142 kb assay. Figure 4A and B show the control G-nhomie-L and the corresponding Su(Hw) binding site mutant transgene, G-nhomieΔSH-L. As evident from a comparison of wild-type and Su(Hw) mutant embryos, expression of the LacZ reporter both at the blastoderm stage and later in development is largely lost when the nhomie Su(Hw) binding site is mutated. Quantification of LacZ and GFP confirmed the staining results (Figure 4C).

Figure 4.

As the digoxigenin in situ hybridization procedure can be used to amplify weaker LacZ signals, we also used it to assay the effects of mutating the nhomie Su(Hw) binding sites, and this is shown in Figure 5A and B. While little if any LacZ expression was detected using HCR-FISH, LacZ transcripts driven by the eve enhancers can be detected in blastoderm and early gastrula G-nhomieΔSH-L embryos using digoxigenin in situ. The number of LacZ-positive cells is, however, substantially reduced compared to WT nhomie. Interestingly, in older germband extended and germband retracted embryos, eve enhancer-dependent expression in the APR and the CNS is largely absent. The loss of the Su(Hw) binding site also impacts the blocking activity of nhomie, since LacZ expression driven by the hebe enhancers in midline cell clusters is now observed (Figure 5B, stage 13, “13” and “13v”).

Figure 5.

The Su(Hw) site in homieCDEF is required for long-distance regulatory interactions

We also reexamined the effects of Su(Hw) binding site mutation in homie on long-distance regulatory interactions. Instead of the 271 bp DEF homie element used in our previous study (Fujioka et al., 2025), we used a larger 367 bp homie element, CDEF. Like nhomie, the relative orientation of the homie in the transgene determines which of the two reporters is preferentially activated by the eve enhancers. For these experiments we oriented the wild-type homie and the Su(Hw) mutant homieΔSH in the transgene so that LacZ was located “upstream” of the transgene boundary (Figure 5C and D, and Figure 6; G-eimoh-L and G-eimoh-ΔSu(Hw)-L). In this configuration, the LacZ reporter is activated by the eve enhancers when the transgene homie pairs with the eve boundaries. The transgenes were then inserted into the -142 kb attP site so that the LacZ reporter is located on the eve side of the transgene homie.

Figure 6.

Figure 6 shows G-eimohΔSH-L HRC-FISH in situs, while Figure 5C and D shows digoxigenin in situs. Little or no eve-enhancer-dependent LacZ expression can be detected in G-eimohΔSH-L in stage 5 (or 7) embryos using either in situ hybridization procedure. Thus, the loss of the Su(Hw) binding site appears to have even greater impact on homie-dependent regulatory interactions at this stage in development than observed for nhomie (compare Figure 5B and D). On the other hand, in stage 11 and 13 embryos in G-eimohΔSu(Hw)-L, weak LacZ expression can be detected in the APR, but not in G-nhomieΔSu(Hw)-L. Quantification of LacZ and GFP expression in the APR confirmed the staining results (Figure 6C). For both cases, several midline cell clusters in the CNS were observed, indicating that in addition to being largely unable to mediate long-distance regulatory interactions, the ΔSH mutant is unable to fully block the hebe enhancers from activating LacZ expression (Figure 4B, 5B, 5D and 6B).

Mutation of Su(Hw) binding sites in nhomie and homieCDEF disrupts long-distance physical interactions

To better understand how the loss of the Su(Hw) binding sites in nhomie and homie impacts the ability of these TAD boundaries to mediate long-distance regulatory interactions, we used MicroC to probe the physical interactions between the transgenes and the eve TAD. The MicroC contact maps for G-nhomie-L and G-nhomieΔSH-L are shown in Figure 7A and B, while Figure 7C and D show “virtual 4C” viewpoints from the transgene LacZ and GFP reporters (similar to the analysis in Figure 3). As expected from the substantial reduction in reporter activity, contacts between the LacZ reporter in G-nhomieΔSH-L and sequences in the eve TAD are diminished. This change can be seen in the blow-ups in panels 7A and B and in the LacZ and GFP viewpoints in panels C and D. Quantitation indicates that there is a nearly 10-fold reduction in contact frequency (Figure 7E). Interestingly, however, the pattern of physical interactions between G-nhomieΔSH-L and the surrounding TADs, with eve and the TADs surrounding eve appear to be quite similar to that of the wild-type G-nhomie-L insert (Figure 7—figure supplemental 1). This would suggest that the transgene boundary still pairs in the same orientation-dependent manner with the endogenous eve locus, just less frequently or less stably than it does when the Su(Hw) site is intact.

Figure 7.

The MicroC contact profile for the G-eimoh-L and G-eimohΔSH-L inserts are shown in Figure 8A and B. As was the case for the G-nhomieΔSH-L insertion, there is a substantial reduction in the physical interactions between the G-eimohΔSH-L and sequences in the eve TAD. This can be seen by comparing the LacZ and GFP reporter “virtual 4C” viewpoints for G-eimoh-L in Figure 8C with those for G-eimohΔSH-L in Figure 8D. Even though the frequency of physical interactions is greatly reduced (Figure 8E), the residual pattern of interactions between G-eimohΔSH-L and the surrounding TADs with eve and the TADs surrounding eve still resemble that seen for G-eimoh-L. Thus, like nhomieΔSH, the eimohΔSH retains some orientation-dependent pairing activity.

Figure 8.

Since mutating the Su(Hw) binding sites in both nhomie and homie results in the activation of the LacZ reporter by the hebe enhancers located just beyond the GFP reporter, we calculated the “insulation” score. We used Fan-C and a window of 4 kb to analyze cross-TAD contacts for lambda DNA, the wild-type and Su(Hw) mutant nhomie and homie boundaries (see Methods). For this purpose, we measured the contact frequency between sequences immediately upstream and downstream of the transgene inserts. Figure 7F shows the insulation score for wild -type and Su(Hw)-mutant nhomie relative to that of lambda, while Figure 8F shows the insulation score for wild-type and Su(Hw)-mutant homie relative to that of lambda. In both cases, there is a significant drop due to the Su(Hw) binding site mutation.

Long-distance physical interactions between the transgene boundary and the endogenous nhomie and homie

The regulatory interactions between the reporters in the transgene insert and the eve enhancers depend upon the physical pairing of the transgene boundary with the endogenous nhomie and homie. In previous experiments (Bing et al., 2024), we found that when the viewpoint is centered on homie in the transgene rather than one of the reporters, the interactions with eve peak at the nhomie and homie boundaries (Figure 9C). This is also true for transgenes containing the nhomie boundary. Figure 9A shows that the interactions between nhomie in the G-nhomie-L transgene and eve are centered on the nhomie and homie boundaries. These findings support the conclusion that pairing of the transgene boundary with the two eve boundaries is directly responsible for physically linking the transgene to the eve TAD.

Figure 9.

As would be predicted by the boundary pairing model, mutations in the Su(Hw) binding site result in a disruption of the physical interactions between the transgene boundary and nhomie and homie in the eve locus. For the Su(Hw) mutation in G-nhomieΔSH-L, the physical interactions between the transgene boundary and the two eve boundaries are substantially reduced (Figure 9B). Moreover, consistent with the idea that the shared Su(Hw) binding sites are important for nhomie self-pairing and for heterologous pairing with homie, the prominent peaks for nhomie-transgene with both nhomie and homie are substantially reduced in the nhomieΔSH viewpoint (Figure 9B). The physical interactions are not completely lost, as there are still small peaks that map to both nhomie and homie.

The physical contacts between the transgene homieΔSH and the two eve boundaries are also compromised by the Su(Hw) binding site mutation (Figure 9C and D). Moreover, as was the case for nhomie, homie self-pairing (homie-transgene with homie) and heterologous pairing (homie-transgene with nhomie) largely depend upon the shared Su(Hw) binding sites.

homie and nhomie Su(Hw) mutant boundaries retains] pairing activity in transvection assays

While the Su(Hw) mutations disrupt long-distance interactions between the transgene boundary and the boundaries in the eve TAD, weak physical interactions can still be detected in viewpoints from the reporters and the transgene boundary. This would suggest that factors other than Su(Hw) must be involved in pairing interactions, and that they might be sufficient on their own to mediate physical interactions in less demanding assays.

To test this possibility, we used a transvection assay in which a LacZ reporter transgene interacts with a transgene containing enhancers, each inserted into an attP site distant from eve, where long-range interactions with the endogenous eve enhancers are not observed (Fujioka et al., 2016). As pairing interactions in trans in this attP environment are relatively weak, the reporter is only weakly activated by the eve enhancers when the LacZ reporter carries lambda DNA and the enhancer transgene carries the homieCDEF that was used in -142kb assay (Figure 10A). In contrast, when both transgenes have the homieCDEF inserted in the same orientation relative to the reporter and the enhancers, the LacZ reporter is activated strongly by the eve mesodermal and APR enhancers (Figure 10B). Trans-activation is also observed when the 367 bp homie fragment CDEF is paired with the smaller 271 bp homie fragment DEF (Figure 10C). In contrast to the -142 kb assay, the regulatory interactions in the transvection assay are less sensitive to mutations in the Su(Hw) binding site. While mutation of the Su(Hw) site in DEF (DΔSuEF, located in the reporter transgene) weakens regulatory interactions, the eve enhancers are still able to active LacZ expression more strongly than in the lambda DNA control (Figure 10D relative to 10A).

Figure 10.

We performed a similar set of transvection experiments with nhomie. However, instead of using the full length 600 bp nhomie fragment we used a truncated 200 bp fragment. Importantly, the long-distance pairing activity of the truncated nhomie fragment is substantially reduced compared to the 600 bp fragment (Park, 2026). Despite its reduced pairing activity in the long-distance assay, it still functions in the transvection assay (Figure 10 figure supplemental 1). Moreover, mutations in the Su(Hw) site in the 200 bp fragment weakened but do not eliminate stimulation by the eve enhancers.

Shared Su(Hw) binding sites promote pairing interactions in the transvection assay

In previous studies (Fujioka et al. 2025), we found that the gypsy insulator is unable to mediate regulatory interactions between transgenes inserted at -142 kb and the eve enhancers (Figure 11—figure supplemental 1B). One interpretation of this finding is that the Su(Hw) proteins associated with homie or nhomie are not in themselves sufficient for physical interactions with the gypsy insulator. An alternative possibility is that while the shared Su(Hw) proteins in nhomie/homie can engage in physical interactions with the gypsy insulator these interactions are not strong enough to mediate the sustain physical contacts needed to activate reporter transcription over a distance spanning 142 kb and multiple intervening TADs.

Figure 11.

To distinguish between these possibilities, we used the transvection assay to test whether the single Su(Hw) binding site in homie is sufficient to mediate regulatory interactions with the gypsy insulator. Figure 11—figure supplemental 1A).

These results show that the two eve boundaries can productively partner with the gypsy insulator in a less demanding assay. This is not unexpected as fly boundaries are known to be somewhat promiscuous in their pairing interactions (c,f, Kychanova et al., 2011; Gohl et al., 2011; Li et al., 2018). In this case, the likely reason would be the shared Su(Hw) binding sites. Consistent with this idea, we found that the dCTCF dependent Fab-8 boundary (Moon et al., 2005; Kyrchanova et al., 2016) does not partner with homieCDEF in the transvection assay (Figure 11).

Discussion

TAD formation in the boundary:boundary pairing model depends upon physical interactions between proteins that are associated with interacting boundary elements. In one mechanism, boundaries that share binding sites for the same DNA binding protein can be linked together if that protein can assemble into homodimers and/or homomultimers. This mechanism is likely used by polydactyl zinc finger proteins like CTCF, Zw5, Zipic, and Pita, by the BEN domain protein Insensitive, and DNA binding proteins like GAF, Mod(mdg4), and CP190 that have BTB domains (Fedolova et al., 2017; Bonchuck et al., 2021). Unlike these chromosomal architectural proteins, Su(Hw) is not known to form homomeric complexes; however, previous studies have suggested that its chromosome architectural functions depend upon shared binding sites. The first evidence for this possibility came from studies by Sigrest and Pirrotta (1997), who showed that transgenes carrying the Su(Hw)-dependent gypsy insulator and a Polycomb response element from the Bithorax complex could mediate long-distance Polycomb-dependent silencing of a white reporter. Subsequent work by Cai and Shen (2001) and Muravyova et al. (2001) presented additional genetic evidence for the pairing of gyspy insulators, and this was extended by the boundary bypass experiments of Kyrchanova et al. (2008a). In these experiments, Kyrchanova et al. showed that bypass depended on shared binding sites for chromosomal architectural proteins. When multimerized binding sites for Su(Hw), CTCF, or Zw5 were paired with themselves, bypass was observed; however, when multimerized Su(Hw) binding sites were paired with multimerized CTCF or Zw5 binding sites, there was no bypass. The Kyrchanova et al. paper also showed that self-pairing is a common property of endogenous boundaries in flies.

While the genetic experiments showing that fly boundaries utilize shared binding sites in their pairing interactions are compelling, a direct demonstration of the physical interactions that are mediated by these shared sites is lacking. In fact, there is limited evidence linking binding sites for fly chromosomal architectural proteins with the physical interactions involved in self or heterologous boundary:boundary pairing. To address this question, we have used the eve TAD boundaries, nhomie and homie, to test the role of shared binding sites. Genetic studies have shown that, like other fly boundaries, nhomie and homie pair with themselves head-to-head. They also pair with each other, in this case in a head-to-tail configuration. This pairing interaction generates a TAD with a stem-loop topology, which has a distinctive contact profile in MicroC experiments (Bing et al. 2024; Ke et al. 2024). When nhomie or homie is included in a dual reporter transgene inserted in an attP site 142 kb from eve, the eve enhancers drive reporter expression. In previous studies on homie, we showed that these regulatory interactions accompany the physical pairing of the transgene boundary with both of the endogenous eve boundaries (Bing et al., 2024). As reported here, this is also true for nhomie (Figs. 2 and 3). This can be seen when the transgene nhomie is used as a viewpoint: it contacts both nhomie and homie in the eve locus. Also like homie, the relative orientation of nhomie in the transgene determines which of the two reporters is activated. Moreover, the pattern of activation is a reflection of the physical contacts that are visualized with MicroC between the two reporters and sequences in the eve TAD.

Mutations in the shared homie and nhomie Su(Hw) binding sites disrupt self and heterologous pairing interactions

As would be predicted if shared sites for DNA binding proteins are important for boundary function, mutations in the homie and nhomie Su(Hw) sites largely abrogate long-distance regulation by the eve enhancers (Figure 4, 5, and 6). Although the effects of the Su(Hw) mutation on transcriptional activation are in both cases substantial, we are able to detect regulatory interactions using the more sensitive digoxigenin in situ procedure (Figure 5). For nhomieΔSH, we detect residual LacZ expression in an eve-like stripe pattern at the blastoderm and early gastrulate stages, while there is little if any expression evident in stage 11 and 13 embryos, when eve is expressed in the APR, CNS, and mesoderm. In contrast to nhomieΔSH, only very few LacZ-positive cells are visible in blastoderm and early gastrula embryos in eimohΔSH. Instead, we detect weak LacZ expression in stage 11 and 13 embryos in the APR and CNS. It is not clear why enhancer-driven LacZ expression for nhomieΔSH seems to be restricted to blastoderm and early gastrulation embryos, while for eimohΔSH, enhancer-dependent expression is stronger in stage 11 and 13. One possibility is that residual self-pairing of the Su(Hw) mutant boundary is stronger than pairing with the heterologous boundary. This is suggested by the proximity of the relevant enhancers to the eve boundary whose transgene partner carries the Su(Hw) mutation. The 7-stripe enhancer that drives eve expression during early gastrulation is located next to the nhomie boundary, and this is the time when nhomieΔSH-dependent LacZ expression is highest. Likewise, the APR and mesodermal enhancers are located between eve and homie, and thus would be closer to the homie boundary. Another possibility is that nhomie and homie have different stage/tissue-specific boundary factors: the nhomie stage specific factors would be present in early embryos, while the homie stage/tissue specific factors would be present later in embryogenesis.

The homie Su(Hw) binding site mutant retains some boundary function

Although the long-distance regulatory and physical interactions of GeimohΔSHL with the eve locus is significantly reduced, it is not completely disrupted. Consistent with the idea that the mutant retains some boundary activity, we can readily detect LacZ expression driven by eve enhancers when a minimal 271 bp homie element, DΔSuEF, carrying the Su(Hw) binding site mutation is combined with the standard 367 bp (CDEF) homie boundary in our transvection assay (Figure 10). The likely reason is that this is a less demanding assay, as boundaries to either side of the reporter and enhancer transgenes in the chromosome are expected to be mediating homologous pairing (via head-to-head self-pairing interactions), and this would help align boundaries in the transgene and stabilize their physical interactions. By contrast, the -142 kb assay boundary pairing takes place across multiple TADs, and, as is evident from the novel contacts that are induced, requires a substantial distortion in the native organization of the chromatin fiber in between the transgene and the eve TAD, as well as to either side. This would likely tend to disfavor physical interactions between the distant boundary elements.

Shared binding sites and TAD formation

Taken together our findings support the idea that one of the mechanisms orchestrating the physical pairing of TAD boundaries in cis is the presence of shared binding sites for the same chromosomal architectural proteins. Consistent with this idea, recent computational analysis of the TAD boundary lexicon in Drosophila identified 575 TAD boundaries in which Su(Hw) is predicted to play a key role in TAD formation (Wang et al., 2026). Of these, 138 correspond to adjacent boundaries that like nhomie and homie share Su(Hw) binding sites and define the endpoints of a TAD. On the other hand, most of the predicted Su(Hw) dependent TAD boundaries are solo—namely, neither of its neighbors has a Su(Hw) site that is predicted to be functionally important. If these predictions are correct, the Su(Hw) protein in the solo TAD boundary would have to partner with some other protein(s) that is bound to one (or both) of the neighboring boundaries. Clearly it will be important to identify factors that can engage in heterologous interactions of this type.

Specificity versus promiscuity in pairing interactions

Although there are multiple TAD boundaries between the -142 kb attP site and eve, none have Su(Hw) binding sites, and there is no indication of pairing interactions between these boundaries and either homie or nhomie in the MicroC contact profiles and the viewpoints from the transgene towards eve (Figure 11—figure supplemental 1B). On the other hand, shared Su(Hw) sites would appear to be sufficient for pairing interactions in less demanding assays, as the gypsy insulator is able to support transvection when paired with homie (Figure 11—figure supplemental 1A). In both cases, the level of reporter expression is much closer to that observed when homie or nhomie are paired with themselves than it is to that seen when one of the transgenes carries lambda DNA in place of a boundary. Thus, the observed levels of specificity versus promiscuity depend upon the design of the assays that are used to detect pairing interactions. Interestingly, even in the less demanding transvection assay some degree of specificity is retained as the dCTCF dependent Fab-8 boundary is unable to partner with homie.

Mechanisms underpinning TAD assembly

Two different mechanism have been proposed to explain how TADs are formed. One is the physical pairing of neighboring boundary elements in cis that has been described here, while the other is the cohesin loop extrusion/CTCF road block model (Dixon et al., 2012; Rao et al., 2017; Fudenberg et al., 2017; Davidson and Peters, 2021; Dekker and Mirny, 2024). The results reported here together with those presented in a previous publication on the pairing properties of the other eve boundary, homie, cannot be explained by the cohesin loop extrusion model; however, they are completely consistent with a mechanism for TAD formation that depends on the physical pairing of compatible boundary elements.

In order to activate reporters inserted in the -142 kb attP site the eve enhancers need to be brought into close proximity. In the loop extrusion model, the cohesin complex would have to break through multiple intervening TAD boundaries and come to a halt when it encounters the transgene nhomie or homie boundary. This would generate a stem-loop in which cohesin sits atop the eve homie boundary and the transgene boundary and holds them together. In this configuration the reporter on the eve side of the transgene boundary will be in the same stem-loop as the eve enhancers and it will be activated by the eve enhancers. At the same time, the transgene boundary will block the eve enhancers from activating the reporter on the opposite, upstream side of the transgene boundary. If the homie and nhomie boundaries function as cohesin roadblocks, independent of their relative orientation with respecting to the extruding cohesin complex, the reporter on the eve side of the transgene boundary will always be activated, while the reporter on the opposite side of the transgene boundary will not.

In this paper, we tested the four possible nhomie transgene configurations: a) nhomie “pointing towards” LacZ; b) nhomie “pointing towards” GFP; c) LacZ on the eve side of the transgene boundary and d) GFP on the eve side of the transgene boundary. The results for two of the configurations—LacZ on the eve side of the transgene boundary with nhomie “pointing towards” LacZ (G-nhomie-L) and GFP on the eve side of the transgene boundary with nhomie “pointing towards” GFP (L-nhomie-G) are consistent with the predictions of the loop extrusion model. However, if one flips, for example, the orientation of the G-nhomie-L transgene in the attP insertion site so that the GFP reporter is on the eve side of the boundary (L-eimohn-G), the GFP reporter is not activated as would be predicted by the cohesin loop extrusion model. Instead, the LacZ reporter is activated by the eve enhancers. Since the orientation of the CTCF sites relative to the incoming cohesin complex is thought to determine whether the CTCF protein functions as a cohesin roadblock, it is possible that the inverted nhomie boundary is inactive as it failed to halt the extruding cohesin complex. However, this possibility is not correct as the inverted boundary blocks the hebe enhancers from turning on the GFP reporter. Essentially the same results are observed when the nhomie boundary in the G-nhomie-L insert is inverted to give G-eimohn-L. Even though the LacZ reporter is still on the eve side of the transgene boundary, it is not activated by the eve enhancers as would be predicted by the loop extrusion model. Instead, the GFP reporter is turned on by both the eve and hebe enhancers.

The shortcomings of the loop extrusion model are also clearly evident in the patterns of physical contacts seen in MicroC experiments. As predicted by the boundary pairing model, physical interactions between the two reporters and sequences in the eve TAD track completely with the relative reporter activity (Figs. 2 and 3 for nhomie; Figs. 8 and 9, see also Bing et al. 2024 for homie). Moreover, independent of the orientation of the boundary in the transgene, it physically interacts with both nhomie and homie in the eve TAD (see also Bing et al. 2024). The physical interactions generated when the nhomie boundary is reversed in G-eimohn-LacZ indicate that the ability to pair with boundaries in the eve TAD is independent of the orientation of the transgene boundary. This would rule out a model in which cohesin bypasses the “inappropriately” oriented nhomie or homie boundaries and stops at some imaginary TAD boundary beyond the hebe gene so that the reporter on the “wrong side” of the boundary is activated by the eve enhancers. In fact, there is no evidence of novel contacts between the boundaries in the eve TAD and boundaries beyond the transgene insert as would be expected if cohesin fail to stop when the transgene boundary is “inappropriately” oriented.

These are not the only shortcomings of the loop extrusion model. There is no mechanism in this model that could explain how TAD boundaries are able to promote trans regulatory interactions. However, we have shown here (Figure 10 and Figure 11—figure supplemental 1) and elsewhere (Fujioka et al., 2016) that nhomie and homie facilitate transvection either by pairing with themselves in trans or by pairing with each other in trans. For self-pairing, the nhomie or homie elements in the enhancer and reporter transgenes must be in the same orientation, while for heterologous pairing (nhomie:homie) they must be in the opposite orientation (Fujioka et al. 2016). Moreover, although they contain only a single Su(Hw) binding, both eve boundaries can promote transvection with the gyspy insulator as a pairing partner. In the case of gypsy:homie, we found that this interaction is orientation independent (Figure 11). These results could not be explained by a cohesin based loop extrusion mechanism. Nor would this mechanism provide an explanation for why nhomie and homie can partner in trans with the gypsy element but are unable to do so with a different CTCF dependent boundary Fab-8.

Materials and methods

Key resources table

Plasmid construction and transgenic lines

The dual reporters construct was described previously (Fujioka et al. 2016). In short, each reporter contains the eve basal promoter (-275 to +106 bp relative to eve start site), either the LacZ (eve-LacZ) or EGFP (eve-GFP) coding region, and the eve 3’ UTR (+1300 to +1525 bp). These two reporters are divergently transcribed. Test fragments were then inserted between the two reporters. The fragments used here are: a 500 bp fragment from lambda phage DNA, a 367 bp wild-type or Su(Hw)-mutant homie fragment (CDEF) (Fujioka et al., 2025), and a 602 bp wild-type or Su(Hw)-mutant nhomie fragment.

The -142 kb attP landing site was described previously (Fujioka et al. 2009). It contains two attP target sites for phiC31 recombinase-mediated cassette exchange (RMCE) (Bateman et al. 2006) and mini-white as a marker. RMCE can result in the insertion of the transgene in either orientation, and all four possible insertions of the transgenes can be recovered. For the experiments in this paper, we analyzed all four possible insertions for transgenes containing the nhomie fragment. For two of these, G-eimohn-L and G-nhomie-L, the eve-LacZ reporter is on the eve side of transgene nhomie, while eve-GFP is on the hebe enhancer side of the transgene homie. For the other two, L-eimohn-G and L-nhomie-G, the eve-GFP reporter is on the eve side of transgene homie, while eve-LacZ is on the hebe enhancer side of the transgene nhomie. For the nhomie Su(Hw) mutant nhomieΔSH, we analyzed the G-nhomieΔSH-L insert, where the eve-LacZ reporter is on the eve side of the transgene. For the two homie transgenes that were analyzed, G-eimoh-L and G-eimohΔSH-L the eve-LacZ reporter is also on the eve side of the transgene. RCMC events were identified by loss of mini-white, and the orientation of each insert was determined by PCR.

Two transgenes were used for the transvection assay (Fujioka et al., 2016). The reporter transgenes are the same ones used in -142 kb assay described above. For enhancer transgenes, modified boundaries (described in figures) were flanked by the anal plate ring (APR) - mesodermal enhancers and two neuronal enhancers. The transvection assay was carried out so that LacZ in the dual reporter transgene and the APR-mesodermal enhancers was placed on the same side of the boundary. With this arrangement, LacZ is activated by APR and mesodermal enhancers. As described in the text, modified homie, nhomie, lambda DNA, and gypsy were inserted into these transgenes. The homie DNAs were the 367 bp CDEF fragment, the 271 bp DEF fragment and the Su(Hw) mutant fragment DΔSuEF (Fujioka et al., 2025). The lambda DNA and gypsy fragments were also described previously (2025).

HCR-FISH and digoxigenin in situ procedure

The sequences of target genes were obtained from Flybase (https://flybase.org)(Gramates et al. 2022). To design probes, the target gene sequences were submitted to the Molecular Instruments probe design platform (https://www.molecularinstruments.com/hcr-rnafish) (Choi et al. 2016), with parameters set to a 35 probe set size for Drosophila melanogaster. A similar method was designed based on published smFISH methods (Little and Gregor 2018; Trcek et al. 2017). 100-200 flies were placed in a cage with an apple juice plate at the bottom of the cage. For early stages, the embryos were collected for 7 hours, while for later-stage embryos, collections were overnight. Embryos from each plate were washed into collection mesh and dechorionated in bleach for 2min, then fixed in 5mL of 4% paraformaldehyde in 1X PBS and 5mL of heptane for 15min with horizontal shaking. The paraformaldehyde was then removed and replaced with 5mL methanol. The embryos were then devitellinized by vortexing for 30s, and washed in 1mL of methanol twice. Methanol was then removed and replaced by PTw (1X PBS with 0.1% Tween-20) through serial dilutions of 7:3, 1:1, and 3:7 methanol:PTw. The embryos were washed twice in 1mL of PTw and pre-hybridized in 200μL of probe hybridization buffer for 30min at 37°C. 0.4pmol of each probe set were added to the embryos in probe hybridization buffer, and the embryos were incubated at 37°C for 12-14h. The embryos were then washed 3X with probe wash buffer at 37°C for 30min and 2X with 5X SSCT(5X SSC+0.1% tween) at room temperature for 5min. Then the embryos were pre-amplified with 300μL amplification buffer for 10min at 25°C. Meanwhile, 6pmol of hairpin h1 and h2 were snap-cooled separately (95°C for 90s, cool to RT with a 0.1°C drop per second), and then mixed in 100μL of amplification buffer at room temperature. After that, the pre-amplification solution was removed from the embryos, and 100μL of hairpin h1/h2 mix were added to the embryos. Next, the embryos were incubated for 12-14h at room temperature in the dark. To remove excess hairpins, the embryos were washed in SSCT as follows: 2X for 5min, 2X for 30min, and 5X for 5min. Then, the embryos were washed with 1mL PTw for 2min and stained with DAPI/Hoechst at 1μg/mL for 15min at room temperature in the dark. The embryos were then washed with PTw 3X for 5min. Finally, the embryos were mounted on microscope slides with Vectashield and a #1.5 coverslip for imaging.

The procedure for digoxigenin in situ was described previously (Fujioka et al. 2025).

Imaging, image analysis, and statistics

Embryos from HCR-FISH were imaged using a Nikon A1 confocal microscope system, with a Plan Apo 20X/0.75 DIC objective. Z-stack images were taken at interval of 2μm, 4X average, 1024x1024 resolution, and the appropriate laser power and gain were set for the 405, 488, 561, and 640 channels to avoid overexposure. Images were processed by ImageJ, and the maximum projection was applied to each of the stack images. To determine the presence of stripes in early embryos, multi-channel images were first split into single channels, and the stripe signal was then highlighted and detected by the MaxEntropy thresholding method. GraphPad Prism was used for data visualization and statistical analysis. Two-way ANOVA with Tukey’s multiple comparisons test for each pair of groups was used to determine the statistical significance for the percentage of embryos carrying stripes in eIF3j and TER94 channels in each group.

Imaging, image analysis and statistics

Embryos from smFISH were imaged using a Nikon A1 confocal microscope system with a Plan Apo 20X/0.75 DIC objective. Z-stack images were taken at intervals of 2μm, 4X average, 1024x1024 resolution, and the appropriate laser power and gain were set for the 405, 561, and 640 channels to avoid overexposure. Images were processed using ImageJ, and the maximum projection was applied to each of the stack images. To measure the stripe intensity of early embryos, multi-channel images were first split into single channels, and the stripe signal was then highlighted and selected by the MaxEntropy thresholding method. For APR intensity measurements, the ROI tool was used to crop out the APR region of late-stage embryos. The cells with APR signal were also highlighted and selected by MaxEntropy thresholding. The particle measurement tool was used to measure the average intensity of all cells that had a signal. At the same time, the background signal (average intensity) was taken from cells without a signal in the same embryo. The relative intensity (signal to background) for each embryo was calculated using the stripe signal and background from the same embryo. To make comparisons between independent biological replicates, the average background signal of all embryos from each replicate was calculated. The relative intensity of each embryo from each replicate was normalized based on the average background signal of all embryos from that replicate. GraphPad Prism was used for data visualization and statistical analysis. To compare intensity from embryos in different groups, different signals from the same embryo (e.g., LacZ and GFP) were paired, and paired two-tailed t-tests were used to calculate p-values. All raw measurements and normalized data are included in Supplemental Data S2.

MicroC library construction

Embryos were collected on yeasted apple juice plates in population cages for 4 hours, incubated for 12hr at 25℃, then subjected to fixation as follows. Embryos were dechorionated for 2min in 3% sodium hypochlorite, rinsed with deionized water, and transferred to glass vials containing 5 mL PBST (0.1% Triton-X100 in PBS), 7.5 mL n-heptane, and 1.5mL fresh 16% formaldehyde. Crosslinking was carried out at room temperature for exactly 15min on an orbital shaker at 250rpm, followed by addition of 3.7 mL 2M Tris-HCl pH7.5 and shaking for 5min to quench the reaction. Embryos were washed twice with 15 mL PBST and subjected to secondary crosslinking. Secondary crosslinking was done in 10mL of freshly prepared 3mM final DSG and ESG in PBST for 45 min. at room temperature with passive mixing. The reaction was quenched by addition of 3.7mL 2M Tris-HCl pH7.5 for 5min, washed twice with PBST, snap-frozen, and stored at -80℃ until library construction.

Micro-C libraries were prepared as previously described (Batut et al. 2022) with the following modifications: 50uL of 12-16hr embryos were used for each biological replicate. 60U of MNase were used for each reaction to digest chromatin to a mononucleosome:dinucleosome ratio of 4. Libraries were barcoded, pooled, and subjected to paired-end sequencing on an Illumina Novaseq S1 100 nt Flowcell (read length 50 bases per mate, 6-base index read).

Micro-C data processing

MicroC data for D. melanogaster were aligned to custom genomes edited from the Berkeley Drosophila Genome Project (BDGP) Release 6 reference assembly (dos Santos et al. 2015) with BWA-MEM (Li and Durbin 2009) using parameters -S -P -5 -M. Briefly, the custom genomes are simply insertions of the transgenic sequence into the –142kb integration site, as predicted from perfect integration. These events were confirmed using PCR post-integration. The resultant BAM files were parsed, sorted, de-duplicated, filtered, and split with Pairtools (https://github.com/mirnylab/pairtools). We removed pairs where only half of the pair could be mapped, or where the MAPQ score was less than three. The resultant files were indexed with Pairix (https://github.com/4dn-dcic/pairix). The files from replicates were merged with Pairtools before generating 100bp contact matrices using Cooler (Abdennur and Mirny 2020). Finally, balancing and Mcool file generation was performed with Cooler’s Zoomify tool.

Virtual 4C profiles were extracted from individual replicates using FAN-C (Kruse et al. 2020) at 400bp resolution. The values were summed across replicates and smoothed across three bins (1.2kb). Viewpoints were determined based on the most informative region for interpretation. We used an 800bp window located downstream of the eve promoter, in the gene body of either GFP or LacZ. For the nhomie and homie viewpoints, the 800 bp window centered on the boundary; however, due to the masking of the duplicated boundary and eve promoters in the data analysis, a reduced signal is expected (Fig. 9).

Calculation and comparison of the insulation score at the hebe insertion site

We used the FAN-C (https://github.com/vaquerizaslab/fanc/tree/main) package to compute insulation scores on the merged Mcool maps corresponding to all the conditions, using a window size of 4 kb. The insulation score is defined as the log (base 2)-fold change between the signal within a 4 kb window centered on a site over the mean of all signal values along chromosome 2R. First, we computed boundaries using the boundary calling tool from FAN-C and identified the hebe insertion site for all mutants. Using these centered coordinates, we computed the insulation score at the hebe insertion site from the balanced contact maps. We then normalized this score relative to the local region around the hebe insertion site by subtracting the original insulation score from the mean of the insulation scores in a 40.2 kb window centered around the hebe insertion site. This mean-scaled normalization can be interpreted as the level of insulation at the hebe insertion relative to its local neighborhood. We computed the change in the normalized insulation between G-lambda-L and all mutants.

Quantification of total contacts between the hebe transgene reporter and the eve locus

For all the mutants, we identified the location of the hebe locus and the eve locus (specifically, the region between nhomie and homie). We then visualized the interaction region between the hebe insertion and the eve locus to manually segment bounding boxes of interaction between the LacZ reporter and the eve locus. After matrix balancing and masking of null values, we took the mean of the contact frequencies within this bounding box for each dataset. This became the average contact frequency between LacZ and hebe transgene. We repeated this procedure for gfp.

Data availability

Sequence data are available at GEO (accession number GSE330020). Analysis and notes are available at https://github.com/pritykinlab/suhw-tad-boundary-mechanisms.

Acknowledgements

Part of this work was supported by NIH grants to Paul Schedl (5R35GM126975), to James B. Jaynes (1R01GM137062) and to Yuri Pritykin (DP2AI171161). Princeton Bioengineering Biocondensates Program to Yuri, and a New Jersey Commission on Cancer Research (COCR23PDF011) to Wenfan Ke. Authors would like to thank Gordon Grey for preparing fly food, members of the Lewis Sigler Genomics Core facility for the assistance with DNA sequencing and Qing Liu for excellent technical assistance.

Additional files

All supplemental figures

Additional information

Funding

National Institute of Health Sciences (NIHS) (5R35GM126975)

• Paul Schedl

National Institute of Health Sciences (NIHS) (1R01GM137062)

• James B Jaynes

National Institute of Health Sciences (NIHS) (DP2AI171161)

• Yuri Pritykin

Princeton Bioengineering Biocondensates Program

• Yuri Pritykin

NJ | NJDOH | New Jersey Commission on Cancer Research (NJCCR) (COCR23PDF011)

• Wenfan Ke