The homie insulator has sub-elements with different insulating and long-range pairing properties

Abstract

Chromatin insulators are major determinants of chromosome architecture. Specific architectures induced by insulators profoundly influence nuclear processes, including how enhancers and promoters interact over long distances and between homologous chromosomes. Insulators can pair with copies of themselves in trans to facilitate homolog pairing. They can also pair with other insulators, sometimes with great specificity, inducing long-range chromosomal loops. Contrary to their canonical function of enhancer blocking, these loops can bring distant enhancers and promoters together to activate gene expression, while at the same time blocking other interactions in cis. The details of these effects depend on the choice of pairing partner, and on the orientation specificity of pairing, implicating the 3D architecture as a major functional determinant. Here, we dissect the homie insulator from the Drosophila even skipped (eve) locus, to understand its substructure. We test pairing function based on homie-carrying transgenes interacting with endogenous eve. The assay is sensitive to both pairing strength and orientation. Using this assay, we found that a Su(Hw) binding site in homie is required for efficient long-range interaction, although some activity remains without it. This binding site also contributes to the canonical insulator activities of enhancer blocking and barrier function. Based on this and other results from our functional dissection, each of the canonical insulator activities, chromosomal loop formation, enhancer blocking, and barrier activity, are partially separable. Our results show the complexity inherent in insulator functions, which can be provided by an array of different proteins with both shared and distinct properties.

Graphical Abstract

Graphical Abstract.

Introduction

Genomic elements called boundaries or insulators function to separate genes from each other, or, in some cases, to separate distinct regulatory domains within a gene (e.g. in the Hox gene complexes). However, they have properties that have proven to be difficult to reconcile mechanistically. For example, while these elements can block enhancer–promoter interactions when inserted between regulatory elements, they can also mediate long-distance regulatory interactions (Chetverina et al. 2017; Schwartz and Cavalli 2017; Özdemir and Gambetta 2019).

One of the first insulator elements discovered is located within the Drosophila gypsy retrotransposon (Bender et al. 1983; Modolell et al. 1983). The gypsy transposon was found to be responsible for many spontaneous mutations in flies. In many cases, the mutant phenotypes arose because the gypsy transposon inserted between a gene and its regulatory elements, blocking regulatory interactions (Parkhurst and Corces 1985; Geyer et al. 1986; Peifer and Bender 1986; Jack et al. 1991). The insulator properties of gypsy have been attributed to the binding of a set of chromatin proteins, the sequence-specific binding component of which is Su(Hw) (Suppressor of Hairy Wing) (Parkhurst et al. 1988; Spana et al. 1988; Holdridge and Dorsett 1991; Geyer and Corces 1992; Roseman et al. 1993; Mallin et al. 1998). Genetic and biochemical analysis identified Mod(mdg4) (Gerasimova et al. 1995; Gause et al. 2001; Ghosh et al. 2001) and Cp190 (Pai et al. 2004) as Su(Hw)-interacting proteins. While Su(Hw) is a zinc finger protein that binds DNA, both Mod(mdg4) and Cp190 contain BTB/POZ protein–protein interaction motifs. Studies have shown that the BTB/POZ domain of Cp190 interacts both with itself and with multiple other insulator-binding proteins (Vogelmann et al. 2014; Golovnin et al. 2023), and Cp190 pull-down analysis identified additional insulator proteins in embryo nuclear extracts (Kaushal et al. 2022).

Several insulators found in the bithorax complex (BX-C) of Drosophila have been extensively studied, including Mcp (Karch et al. 1994), Fab7 (Karch et al. 1994; Hagstrom et al. 1996), and Fab8 (Zhou et al. 1999; Barges et al. 2000), as well as FS1 in the Antennapedia complex (Belozerov et al. 2003), and scs and scs′, which flank the hsp70 locus (Udvardy et al. 1985; Kellum and Schedl 1991). Many insulator-binding proteins that are known to contribute to insulator activity have been identified. These include Zw5 (Gaszner et al. 1999), GAF (Ohtsuki and Levine 1998; Belozerov et al. 2003), CTCF (Moon et al. 2005), BEAF-32 (Zhao et al. 1995; Roy et al. 2007; Jiang et al. 2009), Ibf1 and Ibf2 (Cuartero et al. 2014), Elba and Insensitive (Aoki et al. 2012; Fedotova et al. 2018), Pita and ZIPIC (Maksimenko et al. 2015; Zolotarev et al. 2016), M1BP (Li and Gilmour 2013; Bag et al. 2021), and Chromator (Rath et al. 2004; Gortchakov et al. 2005; Sexton et al. 2012). Genome-wide analysis of the sequences associated with various insulator proteins suggested that there are different classes of insulators, based on binding by specific combinations of these proteins (Negre et al. 2010).

The Drosophila even skipped (eve) gene is flanked by 2 insulators, nhomie (neighbor of homie) and homie (homing insulator at eve), at its 5′- (upstream-) and 3′- (downstream-) ends, respectively (Fig. 1a and b). These 2 elements define the eve TAD (topologically associating domain) (Fujioka et al. 2009, 2016; Bing et al. 2024; Ke et al. 2024). Genome-wide analysis showed that both homie and nhomie are bound by most of the insulator-binding proteins mentioned above, as well as by Rad21, a component of the cohesin complex (http://chorogenome.ie-freiburg.mpg.de; Wood et al. 2011; Matzat et al. 2012; Schwartz et al. 2012; Soshnev et al. 2012; Van Bortle et al. 2012, 2014; Li and Gilmour 2013; Cuartero et al. 2014; Li et al. 2015; Maksimenko et al. 2015; Cubeñas-Potts et al. 2016; Zolotarev et al. 2016; Baxley et al. 2017; Ramírez et al. 2018; Bag et al. 2021). The properties of homie have been well established. Like many other insulators, it has enhancer-blocking activity (Fujioka et al. 2009, 2016) and an ability to prevent the spread of Polycomb-dependent silencing (Fujioka et al. 2013). homie abuts the promoter of an essential housekeeping gene, TER94 (Pinter et al. 1998; Leon and McKearin 1999; Ruden et al. 2000). Since the eve TAD is assembled into a Polycomb-group (PcG)-silenced domain in most cells during all but the early stages of development (Schwartz et al. 2006; Tolhuis et al. 2006; Negre et al. 2010; Van Bortle et al. 2012), this barrier activity is thought to be a critically important function. In eve-TER94 “pseudo-locus” transgenes, homie is required to keep the eve Polycomb response elements (PREs) from shutting down TER94 through the spreading of the repressive PcG-dependent histone modification H3K27me3 (trimethylation of histone H3 residue lysine-27) (Fujioka et al. 2013). Consistent with barrier activity being an essential function of many insulators genome-wide, insulator protein binding sites are located at roughly half of the borders of Polycomb domains in Drosophila (De et al. 2020).

Fig. 1.

Maps of the eve TAD and transgenes inserted at −142 kb from the eve locus. a) Map of the eve locus. Stripe enhancers (blue) (Small et al. 1992, 1996; Fujioka et al. 1999; Sackerson et al. 1999) for the “late” (stages 7–8) 7-stripe pattern (late), early (stage 5) stripes 3 and 7 (3 + 7), early stripe 2 (2), 7 ftz-like stripes (ftz-like, which drive reporter expression only when the enhancer is taken out of context), stripes 4 and 6 (4 + 6), stripe 1 (1), and stripe 5 (5). Neuronal enhancers (pink) (Fujioka et al. 1999; Sackerson et al. 1999): EL (EL), CQ (CQ), RP2 + a/pCC (RP2) neuroectodermal cells. Tissue-specific enhancers (orange) (Fujioka et al. 1999; Sackerson et al. 1999): APR, mesodermal cells (Me). 3’ PRE (yellow; upstream eve promoter region also has PRE activity) (Fujioka et al. 2008). Insulators nhomie (N) and homie (H) (red) (Fujioka et al. 2009, 2016). Arrowheads (green): DNase I hypersensitive sites (Sackerson et al. 1999). b) The homie sub-elements, ABCDEFGH. The TER94-RA transcript starts in G. Both the TER94-RD transcript and the lncRNA:CR45324 start in H (Flybase, Gramates et al. 2022). The positions of the Su(Hw) binding site and the Cp190-associated site (Negre et al. 2010; Cuartero et al. 2014; Baxley et al. 2017) are shown as ovals (red). c,d) The eZ-CDEF and eZ-CDEF-eG transgenes. The −142 kb attP site is located in the 5′ UTR of the hebe transcription unit. Because the attP site was originally inserted via P-element transposition, the 5′- and 3′-P-element ends flank it in the genome. The 3′-P-end (3′P) is located closer to the CNS midline enhancer of the hebe gene, while the 5′-P-end (5′P) is closer to the eve locus. The hebe-RD transcription unit (Flybase, Gramates et al. 2022) is shown as an arrow labeled hebe in d (blue). Insertion of a transgene via RMCE results in 2 recombined versions of the original attP and attB sites (Bateman et al. 2006). Transgene insertion can occur in either orientation: “Z5”, where lacZ is closer to 5′P, or “3Z”, where lacZ is closer to 3′P. In the case of Z5, a modified homie is located between 3′P and the eve-lacZ reporter gene (lacZ: direction of transcription is shown as an arrow). In the case of 3Z, the modified homie is located between eve-lacZ and 5′P, as illustrated. The orientation of homie in the chromosome is indicated by the order of its sub-elements: either “CDEF” for the same orientation in the chromosome as the endogenous homie, or “FEDC”, for the opposite orientation. Note that the 2 transgenes in C are identical except for their orientation in the chromosome, as are the 2 transgenes in D. The eZ-CDEF-eG vector is the same, except for the additional presence of an eve-GFP reporter (“GFP”, green arrow), transcribed in the opposite direction as eve-lacZ. Each reporter is driven by the eve basal promoter, which confers no eve-like expression on its own (Fujioka et al. 2016).

In addition to these activities, homie and nhomie share another characteristic property of fly insulators, namely, an ability to physically pair with themselves and with each other (Fujioka et al. 2009, 2016; Chen et al. 2018; Bing et al. 2024; Ke et al. 2024). A primary function of boundary elements/insulators in Drosophila is the subdivision of the chromosome into a series of looped domains, or TADs. TAD formation is thought to depend upon the physical pairing of neighboring boundary elements. In most cases that have been examined in detail, these pairing interactions are orientation-dependent (Kyrchanova et al. 2008; Fujioka et al. 2016). Orientation dependence determines the topology of the TAD. When boundaries pair with their neighbors head-to-head, a circle-loop is generated, while head-to-tail pairing generates a stem–loop (Fujioka et al. 2016). Since homie and nhomie pair with each other head-to-tail, the eve TAD is a stem–loop. As we have shown (Ke et al. 2024), this loop topology enhances the physical isolation of the eve TAD from its neighbors. A second function of fly boundaries is mediating the pairing of homologous chromosomes, and this depends upon their ability to self-pair (Viets et al. 2019). While pairing between homie and nhomie is head-to-tail, homie and nhomie self-pairing is head-to-head (Fujioka et al. 2016), like that of other fly boundaries that have been tested (Kyrchanova et al. 2008). This orientation preference is thought to be important for mediating the juxtapositioning and precise alignment of homologous chromosomes, and it has been shown to facilitate transvection (regulatory cross-talk) between paired homologs (Fujioka et al. 2016; Lim et al. 2018), a phenomenon that is widespread in Drosophila.

As has been observed for the gypsy insulator and for boundaries from the BX-C (Geyer et al. 1990; Sigrist and Pirrotta 1997; Muller et al. 1999; Kravchenko et al. 2005; Vazquez et al. 2006; Li et al. 2011), both homie and nhomie can also mediate long-distance (100 kb to Mb) regulatory interactions (Fujioka et al. 2009, 2016; Chen et al. 2018; Bing et al. 2024). As an example, when a homie-containing reporter transgene is inserted at an attP site located in the hebe gene, 142 kb upstream of the eve gene, enhancers in the eve TAD can drive reporter expression. Reporter activation depends on the orientation of the homie element relative to the reporter in the transgene (e.g. eZ-CDEF in Fig. 1c vs eZ-FEDC, not shown), but does not depend on the orientation of the transgene in the chromosome (Z5 vs 3Z in Fig. 1c and d). This constraint arises because homie in the transgene pairs with homie in the eve locus head-to-head and with eve-locus nhomie head-to-tail (Fujioka et al. 2016; Bing et al. 2024). This orientation-specific pairing can put a reporter gene in either a favorable or an unfavorable position for the eve enhancers to access a transgenic promoter (Fujioka et al. 2016).

Here, we have undertaken a functional dissection of homie, focusing on 3 centrally important activities of this class of elements: long-range pairing (LR pairing), enhancer blocking, and PRE blocking. Detailed slicing and dicing of homie reveals a general correlation between the 3 activities. However, the correlation is not perfect, indicating that while the activities are related, there are some mechanistic differences. An extensive comparison of sub-element combinations provides clear cases of divergence between the requirements for each of the 3 insulator functions. Our results have implications for the mechanistic connections between these seemingly disparate activities.

Materials and methods

Plasmid construction, transgenic fly production, and assay systems

Construction of the LR pairing assay vectors eZ (illustrated in Fig. 1c; used in Figs. 2 and 5a, Supplementary Figs. 1a–d, 3, and 8) and eZ-eG (illustrated in Fig. 1d; used in Figs. 3, 4, and 5b, Supplementary Figs. 1e, 2, 5, 6, 7, and 8b) were described previously (Fujioka et al. 2009, 2016). In short, the lacZ and GFP reporter genes are each driven by the eve promoter, and terminated by the eve 3′ UTR and poly-A signal and the α-tubulin poly-A signal, respectively. These vectors, carrying modified homie sequences, gypsy, or control λ DNA, were inserted into the attP site at −142 kb relative to the eve locus (Fujioka et al. 2009). The −142 kb attP site is in the 5′ UTR of the hebe gene. Since this attP site was introduced into the chromosome using P-element insertion, both the 5′P-element end (5′P) and the 3′P-element end (3′P) are present (Fig. 1c and d), and since these constructs were inserted into the attP site using recombinase-mediated cassette exchange (RMCE) (Bateman et al. 2006), they can be inserted in either orientation (Z5 or 3Z). Transgenes that carry wild-type homie interact with the eve locus and can express a reporter gene in a partial eve pattern, independent of their orientation, Z5 or 3Z (Fujioka et al. 2009, 2016; Bing et al. 2024). Modified homie sub-elements and combinations DE, DF, EF, D, DΔSH, E, EΔCp190, and F in the eZ vector carry 413 bp of λ phage DNA (λ DNA) as a spacer between the eve-lacZ promoter and the homie fragment.

Fig. 2.

Functional dissection of the LR pairing activity of homie. The previously described 800 bp “full-length” homie (Fujioka et al. 2009) was further dissected, by dividing it into roughly 100 bp segments (A–H, see Fig. 1b) and testing them in our LR interaction assay. a) Expression from transgenic reporters at embryonic stage 5. For reference, the expression of endogenous eve RNA is shown (eve). For each tested region, lacZ expression from eZ-vector transgenes carrying the indicated homie derivative or negative control (500 bp of phage λ DNA) inserted at −142 kb is shown. The color of the label above each image indicates the transgene orientation in the chromosome: black (all except eve and CEF) is Z5, red (CEF) is 3Z (see maps in Fig. 1c). Arrowheads (gray) indicate the prominent “head stripe” of background expression seen when there is weak or no eve-like expression (see text). Scale bar (in CDEFGH) = 50 μm. b, c) Quantification of LR pairing activity. Images like those in a were analyzed (as described in Materials and Methods) to give both the number of cells expressing the reporter (lacZ) in an eve pattern (b) and the intensity of that expression in each cell (c). These quantities (number of cells expressing per embryo and average intensity of that expression per expressing cell per embryo) were then standardized by scaling them (linearly) relative to those of CDEF (set at 100) and λ DNA (set at 0). Graphs of averages of these quantities are shown, ± standard deviations as error bars. The results of pairwise statistical comparisons (t-tests as implemented by Microsoft Excel) are shown as brackets connecting key pairs (see Supplementary Table 1 and Materials and Methods for more details). Significance of the difference is indicated by the number of asterisks within each bracket: *P < 0.05, **P < 0.01, ****P < 0.0001, ***** P < 0.00001, 9*P < 10^–9.

Fig. 5.

Enhancer blocking the activity of homie and its derivatives. Box-and-whiskers plots of the average and distribution of the number of CNS midline cell clusters expressing the lacZ reporter at late embryonic stages 12 and 13. Each box represents the 25th to 75th percentile range. The horizontal line and “X” mark the median and average, respectively. Whiskers represent nonoutlier data points, and open circles represent outliers. The averages for each transgene at stage 12 and at stage 13 are connected with dotted and solid lines, respectively. Representative CNS images are shown in Supplementary Fig. 5b, and the sample sizes (# of embryos analyzed) for each transgene are given in Supplementary Table 1. Those with a higher number of visible clusters (which is plotted along the y-axis) have less enhancer-blocking activity. The homie sub-elements present in the transgene are listed below each pair of plots: names are as in the text; e.g. “ΔSH” has the Su(Hw) consensus binding site mutated, and “ΔCp” has the Cp190-associated site mutated. Constructs were ordered from left to right based on the average number of midline cell clusters expressing. The results of key pairwise statistical comparisons (t-tests as implemented by Microsoft Excel; all results are given in Supplementary Table 1) are shown as brackets connecting each pair. Stage 12 comparisons are shown above the plots, while those at stage 13 are shown below the plots. Significance of the difference is indicated by the number of asterisks within each bracket: *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001, *****P < 0.00001. a) eZ vector transgenes. b) eZ-eG vector transgenes.

Fig. 3.

Quantitative comparison of the LR pairing activities of homie derivatives. a) Transgenic reporter (lacZ) expression is shown from the indicated eZ-eG transgenes in the Z5 orientation at embryonic stage 5. In D3EF, the 5′ (left) half of D was removed from DEF (the Su(Hw) site is still present), while in D3FE, the orientation of section EF was additionally reversed. In D3EFG5, D3EF is extended to just downstream of the TER94-RA start site; in D3EFGH5, it is extended to just downstream of the CR45324 start site; in D3EFGH, it is extended through the end of homie sub-region H, which includes the TER94-RD start site. Scale bar (in “CDEF”) = 50 μm. b, c) Images like those in (a) were quantified and graphed as described in Fig. 2b and c. Significance of key differences is indicated by the number of asterisks within each bracket: *P < 0.05, **P < 0.01, ***P < 0.001, *****P < 0.00001, ******P < 10^–6.

Fig. 4.

The one consensus binding site for Su(Hw) in homie contributes strongly to long-range pairing. smFISH analysis was performed on stage 15 DEF and DEFΔSH eZ-eG transgene embryos. Expression of lacZ and GFP in the anal plate region is shown in (a). Quantification of lacZ expression is shown in (b) (**** indicates significance of the difference at the P < 0.0001 level). Each image is 80 μm in width.

The same constructs were used to analyze enhancer-blocking activity. As diagrammed in Fig. 1d, the hebe ventral midline enhancer is located upstream of the attP site in the chromosome. When the construct is inserted in the Z5 orientation, the test fragment is between the enhancer and lacZ. Wild-type homie prevents activation of lacZ by the ventral midline enhancer, while λ DNA is unable to do so, resulting in midline expression (Fujioka et al. 2009, 2016). On the other hand, in the 3Z insertion, the test fragment is not between the enhancer and lacZ, so the enhancer-blocking activity of 3Z insertions cannot be determined (Fig. 1c).

In order to localize the hebe midline enhancer, the 1st intron of the hebe-RD transcript was split into 3 regions (Supplementary Fig. 8a: hebe-1, 2R:9829859–9832120; hebe-2, 2R:9831531–9834080; and hebe-3, 2R:9833624–9836462, based on genome assembly dm6 coordinates). To reduce the chance of missing an activity that spans the junction of 2 fragments, overlapping regions were tested. The regions were cloned into the eZ vector and inserted into the attP site at cytological location 74A2 (Fujioka et al. 2013). Enhancer activity was assessed as lacZ expression in a hebe-like pattern, using in situ hybridization.

The eve pseudo-locus construct used for analyzing PRE-blocking activity was described previously (Fujioka et al. 2013). In short, the region located between either −6.4 kb (Supplementary Fig. 9) or −6.6 kb (Fig. 6a) and +11.4 kb relative to the eve transcription start site was modified by replacing the region from +167 bp to +1.3 kb with the lacZ coding sequence and the eve poly-A signal. The 3′ end point of +11.4 kb, which is in the 3rd exon of TER94-RA and -RD, was fused to the GFP coding region, followed by the α-tubulin poly-A signal. The attP sites used for this analysis (Fujioka et al. 2013) are at cytological locations 95E5 (Supplementary Fig. 9a) and 74A2 (Fig. 6a and Supplementary Fig. 9b). The smaller construct used to analyze PRE-blocking activity (Fig. 6b) is the same as the pseudo-locus, except that the sequence upstream of +8.4 kb has been removed. The constructs using this vector were inserted into a MiMIC attP site, Mi{MIC}Drgx [MI04684] (Venken et al. 2011) (cytological location 24B1, used in Fujioka et al. 2021). As illustrated in Fig. 1a, the eve 3′ PRE is located between +8.4 and +9.2 kb, just upstream of homie (Fujioka et al. 2008). In both constructs, the homie region from +9.2 to +9.8 kb (ABCDEF) was replaced with a modified homie and 500 bp of λ DNA. When wild-type homie is in this position, it prevents the spreading of PRE-dependent repression activity into the TER94-driven GFP gene, and TER94-GFP is expressed strongly in ovaries (Fujioka et al. 2013). When λ DNA is in this position, it fails to prevent the spreading of repression, and TER94-GFP is repressed. In order to maintain approximately normal positions relative to the eve PRE, modified homie derivatives DEF, CDE, CDF, and CEF each carry 217 bp of λ DNA, while AB, DE, DF, EF, C, D, E, F, DΔSH, and EΔCp each carry 413 bp of λ DNA, between the 3′ PRE and homie.

Fig. 6.

Multiple, nonoverlapping homie sub-elements are sufficient for PRE blocking. Top: Diagram of the eve pseudo-locus transgene (Fujioka et al. 2013). Images: Live images of GFP expression in dissected ovarioles. Scale bars (in “wt”) = 50 μm. Graphs: quantification of GFP expression by RT-PCR. GFP RNA levels were normalized to endogenous TER94 RNA in each sample. The averages and standard deviations of 3 independent data sets for each sample are shown. The results of key pairwise statistical comparisons (t-tests as implemented by Microsoft Excel; all results are given in Supplementary Table 1) are shown as brackets connecting each pair. Significance of the difference is indicated by the number of asterisks within each bracket: *P < 0.05, **P < 0.01. a) wt: intact pseudo-locus inserted at 74A2 (Fujioka et al. 2013). DEF, D, E, F, AB, C: the ABCDEF region of homie was replaced with each of these derivatives. In each case, λ DNA was used to make the spacing between the 3′ PRE and the homie element similar to that in the “wt” pseudo-locus. λ DNA: the same sequence used in Fig. 2 replaced ABCDEF. For λ DNA, AB, and C, longer exposures of the same ovaries are shown at the right. b) A shorter assay construct, diagrammed at the top, was used, consisting of the 3′ end of the eve pseudo-locus, starting just upstream of the PRE, inserted at 24B1. The ABCDEF region of homie was replaced with each of these derivatives, as in a: wt, D, E, F, λ DNA; DΔSH, EΔCp: the same point mutations of the Su(Hw) and Cp190-associated sites in homie as in Fig. 3 were introduced; ΔPRE, Δhomie: the eve PRE and ABCDEF homie together were replaced by 1.3 kb of phage λ DNA. For λ DNA and DΔSH, longer exposures of the same ovaries are shown at the right.

Analysis of transgenic lines

In situ hybridization was performed based on previously published methods (Kosman et al. 2004), except that RNA was visualized using a histochemical reaction. Stages 4–7, 4–11, and 12–15 embryos were collected separately for several days. These samples were combined to make 60 μL embryos per sample (so that stages 4–15 embryos are present in each sample). DIG-labeled antisense RNA probes against lacZ or GFP were produced using T7 RNA polymerase with DIG-RNA labeling mixture (Roche). Antisense RNA was visualized using alkaline phosphatase-conjugated anti-DIG antibody (Roche), using CBIP and NBT as substrates (Roche). Once the color was developed, which was determined by a positive control's expression level (which carried either CDEF or DEF in the same vector as the other transgenes in the experiment), the reactions of all samples were stopped simultaneously. Stained embryos were washed with ethanol to remove the pinkish color, resulting in an intensified dark blue color. Ethanol washing also prevents the bleeding out of pinkish color after embryos are mounted in Fluoromount (Southern Biotechnology). Images were obtained using a Zeiss Axioplan2 microscope with the same camera settings each time. Each experiment was performed at least twice, with independent in situ procedures. Representative images are shown in the figures.

Single-molecule fluorescence in situ hybridization (smFISH) (Trcek et al. 2017; Little and Gregor 2018) was performed as previously described (Ke et al. 2024). Dechorionated embryos (stages 14–16) were fixed in 5 mL 4% paraformaldehyde in 1× PBS and 5 mL heptane for 15 min with horizontal shaking. After devitellinization, embryos were washed 2× with 1 mL of methanol. Methanol was then removed and replaced by PTw (1× PBS with 0.1% Tween-20) through serial dilutions of 7:3, 1:1, and 3:7 methanol:PTw. The embryos were then washed 2× in 1 mL of PTw and 2× in 1 mL smFISH wash buffer (4× SSC [where a 20× SSC stock solution is 3 M sodium chloride, 300 mM trisodium citrate, adjusted to pH 7.0 with HCl], 35% formamide, and 0.1% Tween-20), and incubated with ∼5 nM coupled smFISH probes (Biosearch) in hybridization buffer (0.1 g/mL dextran sulfate, 0.1 mg/mL salmon sperm ssDNA, 2 mM ribonucleoside vanadyl complex, 20 μg/mL RNase-free BSA, 4× SSC, 1% Tween-20, and 35% formamide) for 16 h. Embryos were then washed 2× for 2 h in 1 mL smFISH wash buffer, followed by 4× 30 min washing in 1 mL PTw. For DAPI/Hoechst staining, the embryos were stained with 1 μg/mL DAPI or Hoechst in PTw for 15 min, then washed 3× for 5 min with 1 mL PTw. For imaging, the embryos were mounted on microscope slides with Aqua PolyMount and a #1.5 coverslip.

For assaying LR pairing activity, we assessed the number of cells expressing lacZ in an eve pattern, and their intensities, at stages 5–8, when the reporter is expressed in stripes, and when eve is tissue-specifically expressed (mesoderm, anal plate ring [APR], and central nervous system [CNS]) at stages 11–13. First, we made an overall assessment of the activity ranking of the reporter based on at least 2 independent experiments. When lacZ expression levels and cell numbers were similar between reporters, stage 5 embryos from sets of embryos stained in parallel were subjected to cell counting and analysis. Transgenes with expression similar to that of λ DNA (i.e. ABCD, CDF, DF, D, E, F, DEFΔSH, D3FE, and gypsy) were excluded from this analysis, as they had few, if any, cells expressing the lacZ reporter in an eve-like pattern. For those chosen for detailed analysis (see Figs. 2 and 3), we manually counted lacZ-expressing cells, and measured the staining intensity of each of those cells using the ROI tools of Image J software (Schneider et al. 2012). When a cell's shape was recognizable from the lacZ pattern, they were counted as expressing. To obtain the average intensity of expression per expressing cell for an embryo, we first chose 10 cell-sized locations each, outside the embryo and inside (where there were no lacZ-expressing cells), and measured the “background” light intensity. The outside background was taken as the maximum light intensity, representing the lowest possible expression signal. The average light intensity of each expressing cell within each embryo, as well as the average of the “inside” background, were each subtracted from this maximum light intensity to get the signal strength for each cell and the average signal background, respectively. The average signal background was then subtracted from the signal strength to get the signal intensity above the background for each cell, and these individual cell signals were averaged to get the average signal intensity above the background per expressing cell for each embryo. For this cell count and intensity analysis, we took the following precautions. First, the expression of eve, and therefore of the lacZ reporter, is rapidly changing around stage 5, so small differences in developmental timing can affect the results. So, for each sample, we used Nomarski microscopy to identify closely matched embryos based on the extent of invagination of cell membranes as cellularization of the blastoderm proceeds. Also, the viewing angle of mounted embryos can affect the number of visible cells. We compared one lateral side of each embryo to the other side (2 focal planes), and chose embryos in which the cell numbers were similar in the 2 focal planes. The clearer of these two focal planes (usually the closer one) was used for the counting. After applying these restrictions, 5–7 embryos/construct/staining could be analyzed. In order to increase the statistical power of the analysis, data from 2 to 3 independent stainings were combined. For this purpose, and to standardize the data to provide for easier comparison of activities in the different assays, values for each set of constructs were linearly scaled relative to those of CDEF (positive control = 100) and λ DNA (negative control = 0). Sample sizes are given in Supplementary Table 1 (including total number of cells counted and quantified, total number of embryos analyzed, and number of different independent experiments/stainings included in the analysis, for each construct), along with the results of pairwise t-tests of the differences in LR pairing for all constructs quantified, for both number of cells expressing per embryo and average intensity of expression per expressing cell per embryo.

For analyzing enhancer-blocking activity, late stages 12 and 13 embryos were separately subjected to counting of CNS ventral midline cell clusters expressing the reporter gene. When cell shapes were visible, it was counted as an expressing cluster. Embryos from at least 2 independently stained samples were analyzed, and the results were combined. The distribution of expressing cell clusters per embryo was graphed as a box-and-whiskers plot, and the pairwise significance of differences (P-values) between constructs was calculated using the t-test function in Excel (Microsoft). Expression in the CNS for each line is shown in Supplementary Fig. 8b. The number of embryos analyzed for each construct and stage is given in Supplementary Table 1.

For analyzing PRE-blocking activity, ovaries were dissected from adult female flies aged 18–24 h at room temperature. Live GFP images of ovaries were obtained using the same camera settings throughout, except as noted in the figures. In order to show the consistency of GFP expression among ovarioles, several ovarioles are shown in each picture. In order to quantify GFP reporter RNA expression, ovaries were subjected to reverse transcription polymerase chain reaction (RT-PCR) (Fig. 6 and Supplementary Fig. 9). Total RNA was extracted from 5 to 8 pairs of ovaries dissected from 18 to 22 h adult females using an RNA extraction kit (Roche). Fifty nanograms of total RNA was used to make cDNA using the Transcriptor First Strand cDNA Synthesis kit with a random primer (Roche). One microliter out of 50 μL cDNA solution was used for each quantitative PCR (qPCR) reaction, each time performed in triplicate. Data are shown as GFP expression normalized to endogenous TER94 RNA expression, similarly quantified in each sample. Each set of lines was analyzed in 3 independent assays, and their averages and standard deviations were graphed. The pairwise significance of the difference (P-value) between these construct averages was calculated using the t-test function in Excel (Microsoft). These t-test results are given in Supplementary Table 1. Primers used were, for GFP: GGGCACAAGCTGGAGTACAACTACAA and TGGCGGATCTTGAAGTTCACCTTG, and for TER94: TGAAGCCACCGCGTGGTATTCTTA and TTTGGACATGATCTCCGGTCCGTT.

Binding site analysis

Binding site searches were done using MacVector software (MacVector Inc). We used DNA-binding motif logos presented in genome-wide studies. When the frequency of nucleotide occurrence was similar at a position, it was represented by the appropriate single-letter code. In order to identify motifs in homie, several mismatches were allowed in the search. Su(Hw) DNA-binding motifs used were YWGCATACTTTT (Negre et al. 2010) and NWWWWNYRTWGCATACTTTTNKGSDB (Baxley et al. 2017). The Cp190-associated consensus sequence used was GGTTBDWRWMYYNGCTD (Cuartero et al. 2014). Mutations were introduced at base pairs that occur at high frequency, according to the motif. Additionally, a sequence motif for Pita binding, TAGCVDRKDHNHVMWCC (Maksimenko et al. 2015) was searched for. The results are shown in Supplementary Fig. 10. For cloning purposes, BamHI and HindIII restriction enzyme recognition sequences were added for the Su(Hw) and Cp190 site mutations, respectively. The ChIP-seq data shown in Supplementary Fig. 4 for Su(Hw) (Wood et al. 2011), Cp190 (Wood et al. 2011), and Pita (Zolotarev et al. 2016), NIH GEO accession numbers GSM762839, GSM762836, and GSM2042225, respectively, were visualized using integrative genomics viewer (IGV) (Thorvaldsdóttir et al. 2013).

Results

Multiple insulator sub-elements contribute to long-range pairing

In previous studies on LR pairing interactions, we used a “minimal” homie, CDEF, which consisted of 4 contiguous ∼100 bp sub-elements, as it appeared to have nearly full activity (Fujioka et al. 2016; Chen et al. 2018; Bing et al. 2024). To better understand how these homie sequences contribute to LR pairing, we examined the long-distance pairing activity of different combinations of sub-elements from a larger 800 bp fragment, ABCDEFGH (Fig. 1b), that was previously identified as the “homie” insulator (Fujioka et al. 2009). According to Flybase (Gramates et al. 2022), this larger fragment extends from just downstream of the eve PRE to just beyond the first exon of the TER94-RD transcript, and it spans a DNase I hypersensitive site (Sackerson et al. 1999), as seen in boundaries from the BX-C. To test for LR pairing interactions, we inserted homie-containing reporter gene constructs into an attP site in the 1st exon of the hebe gene, 142 kb upstream of eve (Fujioka et al. 2009, 2016). For this analysis, we used either the single reporter transgene shown in Fig. 1c or the dual reporter transgene shown in Fig. 1d. When homie or its derivatives have LR pairing activity, and when the insulator is in the “correct” orientation relative to the lacZ reporter in the transgene (e.g. eZ-CDEF in Fig. 1c, but not eZ-FEDC, not shown), lacZ is subject to regulation by enhancers in the eve TAD and is expressed in an eve-like pattern (e.g. CDEF in Fig. 2a and Supplementary Fig. 1a and b). For example, homie CDEF in the reporter genes diagrammed in Fig. 1c and d pairs with CDEF in the eve locus (diagrammed in Fig. 1b) head-to-head (CDEF–CDEF). When this happens, the lacZ reporter is placed in proximity to the eve enhancers, while the GFP reporter is placed away from them, on the opposite side (in 3D space) of the paired insulators, transgene CDEF and endogenous CDEF. This results in lacZ being expressed in the eve pattern, while GFP is expressed very little (Fujioka et al. 2016).

By contrast, no eve-like expression is seen when the same reporter transgene has an equal-length DNA segment from phage Lambda (“λ DNA” in Fig. 2a and Supplementary Fig. 1a and b). However, there is some non-eve-related expression. Near-ubiquitous expression occurs early, but fades away by early stage 5 (“λ DNA” in Fig. 2a). At stage 4, embryos also show spotty expression in yolk cells, which lasts to early stage 5 (Supplementary Fig. 2, olive arrowhead with black outline). At stage 5, a head stripe located anterior to eve stripe 1 is seen, which is more prominent with the single reporter construct (Fig. 2a and Supplementary Fig. 1a and b: “λ DNA”, gray arrowhead with red outline). At stage 9/10, there is some expression laterally, ventral to eve mesodermal expression (Supplementary Fig. 2, sky blue arrowhead with black outline), which is more prominent in the 3Z orientation. Some of these non-eve-related patterns are also seen, with varying intensities, in some of the homie derivatives we analyzed (e.g. the head stripe is seen with varying intensity in a number of constructs; Supplementary Fig. 1: gray arrowhead with red outline). We note that MicroC analysis showed a weak interaction between the λ DNA transgene and the eve locus, although this did not result in any detectable eve-like lacZ expression (Bing et al. 2024). Also, as described further below in the section on enhancer-blocking activity, a hebe midline enhancer is located upstream of the attP site at −142 kb (see map in Fig. 1d for the location). When homie is interposed between this enhancer and the transgene reporter, it blocks the enhancer from activating the reporter (for example, CDEFGH and CDEF in Supplementary Fig. 1a, which at stage 13 show only eve-like CNS expression on the ventral side). However, with λ DNA in place of homie, the hebe midline enhancer activates both the lacZ and GFP reporters (black arrowheads in λ DNA, Supplementary Figs. 1 and 2).

We compared the LR pairing activity of different constructs using 2 criteria: the number of cells expressing the lacZ reporter gene in an eve pattern (Fig. 2b) and the level of expression in those cells (Fig. 2c), as determined by quantifying the average intensity of staining per expressing cell in individual embryos (see Materials and methods for more detail). These 2 measures generally correlated positively with each other (with some exceptions, see below). We first extended the minimal CDEF homie endpoint by ∼200 bp downstream, into the beginning of the TER94-RD transcription unit, to give CDEFGH. The addition of the GH sub-region increased the LR pairing frequency and/or stability. This is reflected in a greater number of cells expressing the reporter at early stages (compare CDEF with CDEFGH in Fig. 2a and b; see also Supplementary Fig. 1a), and in stronger expression at later stages, most noticeably in the mesoderm (Supplementary Fig. 3, left column) and CNS (Supplementary Fig. 3, right column) of stage 13 embryos.

Next, we assayed the LR pairing activity of 4 combinations of 3 different CDEF sub-elements (DEF, CEF, CDF, and CDE). Of these sub-element combinations, only DEF has LR pairing activity comparable to that of CDEF. There is a significant difference between CDEF and DEF with regard to the number of cells expressing, but not the intensity of expression (Fig. 2b and c). There may be a weaker expression with DEF in the mesoderm and the CNS, but this difference is small and somewhat variable (Supplementary Fig. 1a and b, stages 11/12 and 13, and Supplementary Fig. 3). Sub-element F contributes to LR pairing more than does C, since expression with DEF is stronger, especially at later stages, and seen in more cells, than is expression with CDE (Fig. 2a and b and Supplementary Fig. 1b). CEF shows very little stripe expression and has only weak APR expression (Fig. 2a and b and Supplementary Fig. 1b, red arrowhead), while CDF expression is similar to that with λ DNA, showing no eve-like expression (Supplementary Fig. 1b). Therefore, both D and E are required for LR pairing activity to generate eve-like expression in stripes, while enough activity remains in the absence of D, but not E, for some APR expression. We note that APR expression has been found in previous studies to be the most persistent aspect of eve expression produced by LR pairing, when either pairing activity is reduced or the intervening distance is increased (Fujioka et al. 2009, 2016).

Since DEF has significantly more LR pairing activity than the other tripartite combinations, we tested the relative contributions of its 3 sub-elements DE, EF, and DF. As shown in Fig. 2 and Supplementary Fig. 1c, DF showed no clear evidence of LR pairing activity. While EF rarely showed stripe expression at stages 5–8, it did give APR expression at stage 11/12 (Supplementary Fig. 1c, red arrowhead). DE, on the other hand, clearly has more LR pairing activity than either DF or EF. In every DE embryo, a small number of cells expressed lacZ at stages 5–8 (Fig. 2 and Supplementary Fig. 1c), as well as in the APR (stages 11–13, Supplementary Fig. 1c: red arrowheads) and CNS (stage 11/12, Supplementary Fig. 1c: yellow arrowhead). However, as is evident from the large number of cells expressing lacZ in stages 5–8 DEF embryos (Fig. 2b), the F sub-element clearly bolsters the LR pairing activity of the DE combination (Fig. 2a and b and Supplementary Fig. 1b and c, compare DEF with DE). None of the single sub-regions (D, E, or F alone) gave any eve-like expression (Supplementary Fig. 1d). We note that DE, DF, D, E, and F gave broad, “background” expression at stage 5 similar to that of the λ DNA control (Fig. 2a and Supplementary Fig. 1). Interestingly, this phenomenon may be analogous to the background expression that is often more prominent when small enhancer fragments that retain little or no “specific” activity are tested in reporter transgenes. Perhaps strong insulators, like strong enhancers, tend to harbor some repressive activity. Alternatively, the absence of background expression with our stronger insulator fragments may be a consequence of their stronger enhancer-blocking activity.

To further test the contribution of the GH region to LR pairing activity, we combined GH with EF (EFGH). As shown above, EF on its own has almost no LR pairing activity, except weak APR expression at stage 11/12 (Supplementary Fig. 1c, red arrowhead). In contrast, a spotty eve-like pattern of lacZ expression is observed with EFGH throughout embryogenesis: lacZ is expressed in more cells than with EF at stages 5–8 (Fig. 2a and b and Supplementary Fig. 1a), in the APR at stages 11–13 (Supplementary Fig. 1a, red arrowheads), and in the mesoderm at stage 13 (Supplementary Fig. 1a, green arrowhead). However, the intensity of expression is weaker than with either CDEF or DEF, and far fewer cells express the reporter gene (Fig. 2 and Supplementary Fig. 1a and b), indicating that GH can only partially substitute for the D sub-element.

On the other hand, the AB sub-element combination cannot replace EF, as ABCD has no LR pairing activity (Supplementary Fig. 1a). In other experiments, we further tested the AB region by creating ABDE in the context of the double reporter transgene (Fig. 3 and Supplementary Fig. 1e). The AB region does not increase the LR pairing activity of DE, since neither the average number of expressing cells nor the average intensity of reporter expression with ABDE is significantly increased over that with DE alone (Fig. 3b and c).

Insulator protein binding sites are required for long-range pairing

The experiments in the previous section show that DE harbors more of homie's LR pairing activity than any other 2-sub-element combination. Genome-wide analysis of the distributions of the Su(Hw) and Cp190 proteins showed that both proteins localize to homie and nhomie (Negre et al. 2010; Wood et al. 2011; Schwartz et al. 2012; Soshnev et al. 2012; Cuartero et al. 2014; Baxley et al. 2017). Sequence-searching of homie identified a Su(Hw) binding site motif (Negre et al. 2010; Baxley et al. 2017) in D, and a Cp190-associated sequence motif (Cuartero et al. 2014) in E. We tested whether these sites are involved in LR pairing. We mutated the Su(Hw) site in the context of DEF (DEF-ΔSu(Hw), or DEFΔSH for short). As shown in Fig. 4 and Supplementary Fig. 5, mutating the Su(Hw) site results in a loss of most of the LR pairing activity; however, weakened APR expression remains (Supplementary Fig. 5, red arrowheads). Interestingly, the gypsy insulator, which has multiple Su(Hw) binding sites, did not show any LR pairing activity (Supplementary Fig. 1e), indicating that Su(Hw) binding sites are not sufficient for this activity. Mutating the Cp190-associated sequence in the context of DEF reduced LR pairing (DEF-ΔCp190, or DEFΔCP for short, Fig. 3 and Supplementary Fig. 5), although the effect is not as drastic as that of the Su(Hw) site mutation (Supplementary Fig. 5). We note that this Cp190-associated sequence motif from Cuartero et al. (2014) is similar to a Pita binding motif reported by Maksimenko et al. (2015). That study reported that the Pita motif is similar to the Cp190 motif identified by Schwartz et al. (2012), and that Pita is capable of recruiting Cp190 (see Supplementary Fig. 4 for excerpts of genome-wide ChIP data).

To further quantify the effects of the Su(Hw) site mutation, we used smFISH on stage 15 embryos. In the experiments of Fig. 4, DEF and DEFΔSH dual reporter transgenes were inserted into the attP site at −142 kb in each orientation, Z5 and 3Z (see diagram in Fig. 1d). We then probed for lacZ and GFP expression driven by the APR enhancer. We focused on this LR activity because the expression is less stochastic in this tissue than in other expressing tissues, and most eve-expressing APR cells also show reporter expression from the starting DEF construct. In both the Z5 and 3Z transgene orientations, eve-like lacZ expression is observed in the APR, while GFP expression is only rarely detected (Fig. 4). This is consistent with the digoxigenin staining (Supplementary Fig. 5 for lacZ and Supplementary Fig. 7 for GFP). Quantitation of lacZ mRNA in the APR showed that mutation of the Su(Hw) recognition sequence in the D sub-element very significantly impairs LR pairing activity, resulting in a substantial drop in lacZ expression (Fig. 4b). This is true when the transgene is inserted in either orientation, 3Z or Z5.

Next, we tested whether the Su(Hw) site is fully responsible for the activity of D, using our dual reporter transgene. Here, we added back the 3′ half of D (D3), which contains the Su(Hw) site, to EF, which has only very weak LR pairing activity (see above), to make D3EF. The addition of the D3 fragment clearly restores some LR pairing activity, since the lacZ reporter is expressed in an eve pattern throughout embryogenesis (Fig. 3 and Supplementary Fig. 5). However, the activity of D3EF is significantly weaker than that of DEF (Fig. 3 and Supplementary Fig. 5), suggesting that sequences in the 5′ (left) half of D interact with a factor(s) that contributes to LR pairing. On the other hand, the activity of D3EF is significantly stronger than that of DE (Fig. 3), indicating that the F region can more than substitute for the 5′ portion of D.

Finally, we tested whether the order of sequences within homie is important for its activity. To do this, we inverted EF in D3EF, to make D3FE. As shown in Supplementary Fig. 4, this change abolishes LR pairing activity, as the eve enhancers fail to activate lacZ expression. It seemed possible that the change in the order of the sub-elements altered the orientation dependence of D3EF. That is, if the orientation dependence were due to EF alone, then D3FE would align with the FE of the endogenous homie (see Fig. 1b). In this case, the lacZ reporter would be placed away from the eve enhancers (and not be expressed). Since D3FE is inserted in the dual reporter, we were able to test this possibility by assaying GFP expression. Supplementary Figure 7 shows that GFP is also not expressed in the D3FE dual reporter, indicating that LR pairing activity has indeed been lost with the change in sub-element order, not switched in its orientation preference. Consistent with the maintenance of orientation specificity with these constructs whenever there is significant LR pairing, we found that neither of the other dual reporter constructs tested (DEF-ΔSu(Hw) and DEF) detectably expressed GFP in an eve pattern (Supplementary Fig. 7; however, they do express GFP in the pattern of the hebe midline enhancer at stage 13, since the enhancer is not able to be shielded from the GFP reporter in this orientation of the transgene).

Contribution of TER94-associated elements to LR pairing activity

In the experiments described above, GH contributed to LR pairing activity in the context of both CDEF and EF homie. G contains the 5′-most TER94 transcription start site (which produces transcript TER94-RA; TER94 is transcribed away from the eve locus), while H contains a noncoding lncRNA start site (CR45324, which is transcribed toward the eve locus) and the initiation site for transcript TER94-RD (Flybase, Gramates et al. 2022). Because D3EF showed weak but clear-cut LR pairing activity (Fig. 3, Supplementary Figs. 5 and 6), we tested whether the adjacent GH region enhances the LR pairing activity of D3EF. Indeed, D3EFGH showed considerably stronger activity than does D3EF (Fig. 3 and Supplementary Fig. 6). We then trimmed back the added region to exclude the start sites for both the noncoding RNA and TER94-RD. This resulted in no apparent loss of LR pairing activity (D3EFGH5, Fig. 3 and Supplementary Fig. 6). We further trimmed it to just downstream of the TER94-RA start site, again with no apparent loss of activity (D3EFG5, Fig. 3 and Supplementary Fig. 6). This suggests that it is the ∼50 bp region just downstream of the F region that harbors most, if not all, of the LR pairing activity of GH. Since this facilitating region includes the TER94-RA start site (to +2 nt), it may contain a basal promoter element from this housekeeping gene. This increased activity is still a bit less than that of CDEF (Fig. 3 and Supplementary Fig. 6). It is also worth noting our previous finding that TER94 promoter activity is enhanced by sequences between the 1st and 3rd exons of the TER94-RA transcript (Fujioka et al. 2013), which are mostly not included in this insulator-facilitating fragment. So, the sequences facilitating homie's insulating activity and those responsible for TER94's transcriptional activity seem to be mostly, if not entirely, separable, with the caveat that we can’t rule out a minor contribution of homie's 3′ end (the 5′ portion of region G) to the level of TER94 expression.

Enhancer-blocking activity only roughly correlates with other insulator activities

Next, we investigated how LR pairing activity correlates with enhancer-blocking activity. As an assay, we tested the ability of different homie sub-elements to block the hebe enhancer from activating the lacZ transgene reporter (Fujioka et al. 2016). This enhancer is located upstream of the −142 kb attP site in a part of the first intron of the hebe gene (Supplementary Fig. 8a, hebe-2). Interestingly, the hebe intron also contains a weak APR enhancer. However, it is active only after stage 16 (Supplementary Fig. 8a, hebe-1), unlike the eve APR enhancer. Because of this, we assessed eve-like APR expression only at stages 11–15. Consistent with these enhancer activities, both expression patterns were reported by the Berkeley Drosophila Genome Project (Tomancak et al. 2007). The 3rd activity we identified, driving expression in the gut (Supplementary Fig. 8a, hebe-3), was not reported by the Genome Project, and we did not see this expression from our λ DNA transgene (Supplementary Figs. 1a, b and 2). It is possible that the gut expression driven by hebe-3 depends on something in the chromosomal environment of the attP site we used for this enhancer mapping. This activity is not relevant to our assay.

The relative position of the midline enhancer to the transgene is shown in Fig. 1d. In the Z5 orientation, when CDEF is located between the lacZ reporter and the hebe enhancer, it blocks the enhancer from activating reporter expression, and only eve-like expression is seen in the CNS (Supplementary Fig. 8b, CDEF and all the other insulators in the left column except CDE; see Supplementary Fig. 3 for details of eve-like expression). On the other hand, λ DNA has little or no blocking activity, and so the hebe enhancer is able to activate lacZ expression in CNS midline cell clusters, mimicking hebe expression at this stage (“λ DNA”, Supplementary Figs. 1a, b and 8b: in Supplementary Fig. 8b, this is also true, to varying degrees, for those in the middle and right columns, and for CDE in the left column). In the 3Z orientation, the midline enhancer can activate lacZ expression independent of the presence of homie, since homie is not located between the enhancer and the lacZ reporter (see Fig. 1d for a diagram, and CDEF, CEF, and λ DNA in Supplementary Fig. 1b).

Enhancer-blocking activity was tested at the 2 stages of embryogenesis when the hebe enhancer is active in cells in the CNS midline (Supplementary Fig. 8a and b, hebe-2, late stage 12 and stage 13, persisting into stage 16). One set of fragments was tested in the context of the single reporter transgene (the eZ vector, Fig. 5a and Supplementary Fig. 8b), and another, overlapping set was tested in the double reporter transgene (the eZ-eG vector, Fig. 5b and Supplementary Fig. 8b). The 2 assays gave a consistent order of activity with all of the constructs common to the 2 assays (CDEF, DEF, DE, and λ DNA). However, the apparent strength of the enhancer-blocking activity of DE in the eZ-eG vector is less than that seen with the eZ vector. This difference in apparent blocking strength allowed us to make clear distinctions using the eZ-eG vector between the activities of the 5 constructs whose activities lie between those of DEF and DE (Fig. 5b).

For the tripartite combinations DEF, CDE, and CDF, blocking activity follows a pattern roughly similar to that of LR pairing activity (Figs. 2 and 5a, Supplementary Figs. 1b and 8b). DEF is the most effective enhancer blocker, followed by CDE, and then CDF. Because we obtained a transgenic line carrying only one of the 2 possible orientations of CEF (the 3Z orientation, in which the insulator fragment is not between the reporter and the hebe midline enhancer), we were not able to determine its enhancer-blocking activity. For the three 2-element combinations from DEF (DE, DF, and EF) in the eZ vector, DE is almost as effective in stage 13 embryos as DEF, while DF and EF each have considerably weaker activity (Fig. 5a and Supplementary Fig. 8b), roughly paralleling their LR pairing activities. The addition of GH to EF (EFGH) increased enhancer-blocking activity (Fig. 5a), similar to the effect seen on LR pairing (Fig. 2a and b and Supplementary Fig. 1a and c). Although we did not see an increase in the LR pairing activity of ABDE over DE (Fig. 3 and Supplementary Fig. 1e), the addition of AB significantly increased the enhancer-blocking activity of DE (Fig. 5b).

Since the Su(Hw) and Cp190 sites contribute to the LR pairing activity of DEF (Figs. 3 and 4, and Supplementary Fig. 5), we tested the effects of mutations in these recognition sequences (Fig. 5b and Supplementary Fig. 8b). As was the case for LR pairing activity, mutation of the Cp190 site had a modest effect on DEF's enhancer-blocking activity (at stage 12, but not at stage 13). In contrast, mutating the Su(Hw) site led to a substantial loss in this blocking activity. Finally, we assayed the enhancer-blocking activity of individual sub-elements (D, E, and F), as well as the effects of mutating the Su(Hw) and Cp190 sites, located in fragments D and E, respectively (Fig. 5a). D alone showed substantial activity, while E alone gave a low but detectable activity, and F alone showed no significant difference from λ DNA. When the Su(Hw) site was mutated, it abolished activity in the context of D alone. Mutating the Cp190-associated site has no clear effect on the activity of E, as a low activity is retained in EΔCP (Fig. 5a). The addition of region ABC to D did not increase enhancer-blocking activity (Fig. 5a, ABCD), indicating that there is no significant enhancer-blocking activity within ABC. Again, these effects on enhancer-blocking activity parallel the effects of the same alterations on LR pairing activity.

There are exceptions to this correlation, however. Most dramatic is the case of the gypsy transposon, which showed strong enhancer-blocking activity (Fig. 5b), but no LR pairing activity (Supplementary Fig. 1e). Seven of the other constructs tested in the eZ-eG vector showed less enhancer-blocking activity than gypsy, but more LR pairing activity (see Fig. 7 for a summary). There are also examples among the homie derivatives. EF has detectable LR pairing activity (Fig. 2 and Supplementary Fig. 1c), whereas DF does not (Supplementary Fig. 1c), while DF has significantly more enhancer-blocking activity at stage 13 than does EF (P < 4.7 × 10^–8, Fig. 5a). Another example is provided by a comparison of the 2 homie derivatives D3EF and D3FE (in which the order of the EF region is reversed; see Supplementary Fig. 5a for a diagram). While D3FE is unable to mediate LR pairing with the eve TAD, D3EF did show LR pairing activity (Fig. 3 and Supplementary Fig. 5). In contrast, both showed modest enhancer-blocking capability (Fig. 5b; in fact, at stage 12, D3FE showed significantly stronger blocking activity than D3EF, P < 0.03). This suggests that, at least from the location of the hebe locus, LR pairing activity is much more sensitive to the order of binding sites for insulator proteins than enhancer-blocking activity. In summary, the correlation is not perfect for all constructs, suggesting that while the 2 activities are closely related, there are some mechanistic differences (see Discussion).

Fig. 7.

Relative strengths of LR pairing, enhancer blocking, and PRE-blocking activities do not always correlate. Constructs within each column (pale green background, first 5 columns, for LR pairing, pink, columns 6–10, for enhancer blocking, and blue, last 3 columns, for PRE blocking) were ordered based on their relative levels of activity in the indicated assay. Constructs that are common to all 3 assays and both LR pairing vectors are listed in boldface in the first column under each activity. Those within boxed groups in each column have activities that are not distinguishable in that assay. Averages are given for each activity (± standard deviations) relative to those of CDEF (set at 100) and λ DNA (set at 0), as in the graphs of Figs. 2 and 3. Asterisks indicate slightly better LR pairing than those in the group just below, based on detectable APR reporter expression. Green dotted lines connect those (in blue or red text) that clearly differ in their order of activity in 2 of the assays (see text).

PRE-blocking activity

We showed previously that homie has the ability to block the spread of repressive chromatin (barrier activity), from an eve pseudo-locus transgene into the neighboring gene, TER94. This spreading depends on the homie-adjacent eve PRE (Fujioka et al. 2013). TER94 promoter activity is repressed when homie is removed, and this correlates with the spreading of the histone modification H3K27me3, characteristic of PcG-dependent repressive chromatin. This loss of expression is not due to reduced TER94 promoter activity caused by removing homie, since removing the eve PRE in addition to homie fully restores expression (Fujioka et al. 2013). In this transgenic context, when homie is present, TER94 promoter-driven GFP expression is high in ovaries dissected from adult females (Supplementary Fig. 9a, wt), while TER94-driven GFP is repressed when homie is removed (Supplementary Fig. 9a, Δhomie).

Whether full-length homie (ABCDEF) is replaced by CDEF or by any of the sub-element combinations DEF, CDE, CEF, CDF, DE, DF, or EF, TER94 promoter-driven GFP expression is undiminished, indicating that PRE blocking is still complete (Supplementary Fig. 9a and b). As shown in Fig. 6a, we also tested other homie sub-elements. Of these, AB and C have no barrier activity, while E alone is sufficient to completely block the repressive effect of the eve PRE. D and F each have partial barrier activity: GFP expression is diminished, but not as much as when homie is replaced by λ DNA (compare “λ DNA” with D and F in Fig. 6a). It is also worth noting that this barrier activity does not depend on the orientation of homie (Supplementary Fig. 9a, CDEF vs FEDC).

Above, we showed that the Su(Hw) binding site in the D region is required for most, but not all, of the activity of D in the LR pairing assay. We asked whether the Su(Hw) binding site is also required for PRE-blocking activity. This set of experiments was performed in a different context than the one used for Fig. 6a and Supplementary Fig. 9. Instead of the full-length eve locus in the transgene, we used only the 3′ end of the locus, which contains all of the essential elements for this assay; namely, the eve 3′ PRE, homie, and the TER94 promoter driving GFP (diagrammed at the top of Fig. 6b). In addition to the transgene construct being different, the insertion site of the transgene is different. Despite these differences, the results are very similar to those shown in Fig. 6a (compare wt, D, E, F, and λ DNA between Fig. 6a and b). As expected, when the PRE is not present in the transgene, there is no repression of TER94-driven GFP expression, even in the absence of any insulator sequences (“ΔPRE, Δhomie”, Fig. 6b; here, both the PRE and homie are replaced by equal-length stretches of λ DNA). We tested the same mutation in this assay, and found that the Su(Hw) site is required for the D region to show any barrier activity (Fig. 6b, DΔSH). We also tested the requirement for the Cp190-associated site in region E: it does not have any apparent effect on the barrier activity of E (Fig. 6b, E vs EΔCp). In summary, while E alone showed more PRE-blocking activity than D alone, D showed stronger enhancer blocking activity, suggesting that enhancer blocking and PRE-blocking activities may have partially distinct mechanisms, and/or that their mechanisms may be differentially context-dependent (see Discussion).

Discussion

In their endogenous locations, homie and nhomie are separated by only 16 kb. However, both elements can pair with themselves and with each other when separated by multiple TADs and TAD boundaries (Fujioka et al. 2009, 2016; Chen et al. 2018; Bing et al. 2024; Ke et al. 2024). In addition to this LR pairing activity, homie can block enhancer–promoter interactions (Fujioka et al. 2009, 2016, 2021) and act as a barrier to the spread of PcG silencing (Fujioka et al. 2013). In the studies reported here, we have undertaken a functional dissection of the homie boundary and identified sequences/sub-elements that are important for these 3 activities. Figure 7 summarizes the activity of each of these elements for each of the 3 activities.

Different stringencies for different insulator activities

For LR pairing activity, DEF is our minimal element, since it gives an eve pattern of lacZ throughout embryogenesis, and the expression pattern and intensity are similar to those of CDEF. Shortening this to DE significantly weakens activity, since many fewer cells express lacZ (Figs. 2 and 3, Supplementary Fig. 1a–e). For enhancer-blocking activity, in the context of the eZ vector, DE is our minimal element. Although we observe transient hebe enhancer-driven midline expression at stage 12, by stage 13, this expression is not seen. Removal of E to give D alone causes substantial loss of enhancer-blocking activity, but considerable activity remains, while E alone shows a lower but detectable activity, and F shows no significant activity (Fig. 5a and Supplementary Fig. 8b). For PRE blocking, E alone has full activity, while D and F alone each have partial activity, F having more than D (Fig. 6). Overall, the requirements for LR pairing activity are the most stringent, followed by enhancer blocking, with PRE blocking showing the least stringent requirements in our assays.

As noted in the Results, we tested enhancer-blocking activity using both the eZ and eZ-eG vectors. While the 2 vectors gave consistent results, in that there were no discrepancies in the relative strengths of enhancer blocking between the constructs tested in both vectors, the absolute level of enhancer blocking seen with the eZ-eG vector was lower. The 2 differences between them are (1) the eZ-eG vector has 2 reporter genes (divergently transcribed), which places lacZ about 1.4 kb further away from the hebe enhancer than it is in the eZ vector, and (2) there is a stretch of λ DNA (∼400 bp) inserted as a spacer between the smaller test fragments (such as DE) and lacZ in eZ that is not present in eZ-eG. While we cannot be sure of the cause of the observed quantitative difference in enhancer blocking, it is possible that the extra distance (∼1 kb) between the hebe enhancer and lacZ in the eZ-eG vector makes it easier for the enhancer to loop around the insulator sequence and activate lacZ in the hebe pattern, thus decreasing the measured enhancer-blocking activity.

LR pairing vs enhancer blocking

We compared LR pairing activity and enhancer-blocking activity, and found that the 2 roughly correlate (summarized in Fig. 7). However, this correlation is not perfect. One clear example is provided by D3EF and D3FE, which each show substantial enhancer-blocking activity (Fig. 5b). However, while D3EF has clear-cut LR pairing activity (although weaker than that of DEF), D3FE has none (Fig. 3, Supplementary Figs. 5 and 6). Reversing the orientation of EF relative to the D3 region likely causes a change in the order of insulator protein binding across the element. Thus, the “correct” ordering of insulator-binding proteins along the chromosome may be required for LR pairing. This is readily explained by individual insulator proteins requiring specific partners for effective pairing. This also explains a preference for self-pairing, which is a property of several of the known insulator-binding proteins (Ghosh et al. 2001; Vogelmann et al. 2014; Zolotarev et al. 2016). We also showed that the single direct Su(Hw) binding site is very important for LR pairing activity by homie (Fig. 4, Supplementary Figs. 5 and 7, DEF-ΔSH). In contrast, the gypsy insulator, which contains 12 Su(Hw) binding sites (Parkhurst et al. 1988; Spana et al. 1988) and can self-pair (Cai and Shen 2001; Muravyova et al. 2001), has strong enhancer-blocking activity in our assay (Fig. 5b), yet shows no detectable LR pairing activity with the eve TAD (Supplementary Fig. 1e). These observations suggest that additional insulator proteins besides Su(Hw) are required to form specific and stable LR pairing interactions between copies of the eve insulators. Consistent with this idea, genome-wide analyses have shown that many of the known insulator-binding proteins are bound at homie and nhomie (Wood et al. 2011; Matzat et al. 2012; Schwartz et al. 2012; Soshnev et al. 2012; Van Bortle et al. 2012, 2014; Li and Gilmour 2013; Cuartero et al. 2014; Li et al. 2015; Maksimenko et al. 2015; Cubeñas-Potts et al. 2016; Zolotarev et al. 2016; Baxley et al. 2017; Ramírez et al. 2018; Bag et al. 2021).

As noted above, enhancer blocking seems less demanding than LR pairing. For example, sub-element D, by itself, has enhancer-blocking activity but no LR pairing activity. However, like its LR pairing activity in the context of larger elements, the enhancer-blocking activity of D strongly depends on the Su(Hw) site (Fig. 5a), indicating that Su(Hw) contributes substantially to both of these activities. Interestingly, region F alone shows no enhancer-blocking activity, while E alone has a very low amount (albeit significantly above that of λ DNA), but the 2 together (EF) have much more activity (in fact, significantly more than that of D alone, Fig. 5a). Likewise, the blocking activity of D is clearly augmented by adding the E region (DE, Fig. 5a). These observations suggest that, like LR pairing, a combination of proteins contributes synergistically to enhancer-blocking activity. The EF region is also capable of detectable LR pairing with the eve locus, whereas DF and CDF are not (Fig. 2 and Supplementary Fig. 1b and c). This is consistent with the importance for LR pairing of the order and spacing of protein binding sites along the insulator, whereas the relative positions of insulator proteins arrayed along an insulator may be less important for enhancer-blocking activity. This difference can account for why DF and CDF have more enhancer-blocking activity than does EF, whereas EF has more LR pairing activity (summarized in Fig. 7). The GH region increases both LR pairing and enhancer blocking when added to EF, although these activities are still not as strong as those of DEF, with its Su(Hw) site intact (compare EF, EFGH, and DEF in Figs. 2 and 5a, Supplementary Fig. 1). Overall, while Su(Hw) provides a substantial part of homie's pairing and enhancer-blocking activities, other combinations of homie-bound insulator proteins apparently provide substantial activity as well.

Together, our data speak to the specificity of our LR pairing assay, which reflects the ability of homie to specifically pair with copies of itself when the 2 are located at considerable chromosomal distances from each other. The ability to pair with other insulators in the vicinity of the insertion site may be sufficient for enhancer blocking, but not for LR pairing. Alternatively, enhancer blocking may not require insulator pairing at all, even though it is clear that pairing can affect which interactions are blocked and which are facilitated by insulators.

Enhancer blocking vs PRE blocking

We compared the enhancer blocking and PRE-blocking activities of an extensive set of homie sub-elements. These comparisons provide a clear indication of a mechanistic distinction between enhancer blocking and PRE blocking by homie. As Fig. 5a shows, the D region has clearly more enhancer-blocking activity than E, which has a small but significant amount, while the opposite is true in the PRE-blocking assay, where E has strong activity, while D is considerably weaker (Fig. 6). One possible explanation is that PRE blocking only requires the introduction of an extended nucleosome-depleted region in 1D space, along the chromosome, while enhancer blocking may require an additional ability to suppress looping in 3D between enhancers and promoters (reviewed in Gaszner and Felsenfeld 2006; Bushey et al. 2008). A specific, albeit speculative model that could explain these observations is that Su(Hw), located in region D, might recruit a complex of protein partners, some of which have weak transcriptional repressive activity in addition to forming a physical roadblock, while region E does not have repressive activity, but forms a more effective roadblock along the chromosome (see below for more discussion of this issue).

Su(Hw) and Cp190 motifs

Early studies found a genetic interaction between Su(Hw) and Cp190 (Pai et al. 2004), and suggested that the DNA-binding insulator proteins BEAF, CTCF, and Su(Hw) dictate DNA sequence specificity, while Cp190 engages in protein–protein interactions through its BTB/POZ domain (Vogelmann et al. 2014). Although Cp190 has a zinc finger domain (Pai et al. 2004), it is unclear whether it can directly bind DNA (Vogelmann et al. 2014). Genome-wide analysis showed that both Su(Hw) and Cp190 localize to homie and nhomie (Negre et al. 2010; Wood et al. 2011; Schwartz et al. 2012; Soshnev et al. 2012; Cuartero et al. 2014; Baxley et al. 2017). Indeed, we found DNA sequence motifs both for binding by Su(Hw) and “association with” Cp190 in homie's D and E regions, respectively. Our data show that the Su(Hw) site has a major role in all 3 insulator activities, while the Cp190 site modestly affects LR pairing and enhancer blocking, but not PRE-blocking activity (Figs. 3, 4, 5b, and 6b, Supplementary Fig. 5). A previous study using a mutation in the Cp190 gene concluded that it affects enhancer blocking but not LR pairing by homie (Kaushal et al. 2022). It is possible that our LR pairing analysis is more sensitive than the one done in that study because of our extensive quantitation and statistical analysis (Fig. 3), but it is also possible that mutating the Cp190-associated site (which may be a Pita binding site, as described in Results) has effects beyond a reduced recruitment of Cp190. Conversely, other sites in homie could also contribute to Cp190 recruitment. Cp190 pull-down assays identified a number of physically associated insulator-binding proteins (Kaushal et al. 2022), and the BTB/POZ domain of Cp190 is known to interact with several DNA-binding insulator proteins, including Su(Hw), Pita, and CTCF (Golovnin et al. 2023). Other studies showed interactions between Cp190 and the DNA-binding proteins Ibf1/2 (Cuartero et al. 2014) and M1BP (Bag et al. 2021). These studies suggest that Cp190 functions through interactions with other DNA-binding proteins. All of the insulator proteins mentioned above localize to homie, based on genome-wide studies. The modest effects of mutating the Cp190-associated sequence in this study are consistent with other binding sites contributing to (indirect) Cp190 association with homie.

Previous studies showed that gypsy, containing 12 Su(Hw) binding sites (Parkhurst et al. 1988; Spana et al. 1988), is capable of self-pairing (Cai and Shen 2001; Muravyova et al. 2001), as well as having enhancer blocking and barrier activities (Holdridge and Dorsett 1991; Geyer and Corces 1992; Roseman et al. 1993, 1995). As our data show, even though the Su(Hw) site in homie has a major role in each of its activities, gypsy does not show LR pairing with the eve locus. This distinction may arise from the strong requirement for pairing specificity in the LR pairing assay, as discussed above. How might Su(Hw) simultaneously harbor these different activities? A previous study showed that distinct zinc-fingers of Su(Hw) bind to different sequences in the core Su(Hw) site, and this could conceivably “activate” different functions, perhaps as a result of the recruitment of different cofactors (Baxley et al. 2017), or by facilitating interactions with a variety of other insulator-bound proteins. Consistent with this general notion, our data show that combining different sub-elements, such as adding F to DE (DEF vs DE), adding 50 bp of the G region to D3EF (D3EFG5 vs D3EF), or adding E to D (DE vs D alone) increases both LR pairing and enhancer-blocking activities (Figs. 2, 3, and 5a, and Supplementary Fig. 1). Further analysis will be required to determine the specific interactions of these sub-elements with various insulator-binding proteins, and how these interactions facilitate the 3 partially distinct insulator functions that we have studied here.

It has been suggested that the Su(Hw) consensus site should be extended beyond the core, because Su(Hw) contacts sequences flanking the core consensus binding site using parts of the protein that contribute differentially to different activities, including insulator activity (Baxley et al. 2017). Since the Su(Hw) site in homie is near the D–E junction, inverting EF separates one of the extended regions, an A/T-rich sequence, from the core site, which might cause a reduction in Su(Hw) binding or insulator activity, and thereby contribute to the observed loss of LR pairing. In the context used here, this A/T sequence in homie (TTTTT) is replaced by a sequence with G at one of the 5 positions (GATTA), both in the enhancer blocking and LR pairing assay vectors. The D region alone (which lacks the A/T sequence extension of the core Su(Hw) site) showed measurable enhancer blocking (Fig. 5a) and PRE blocking (Fig. 6b), and these activities are both totally dependent on the core Su(Hw) binding site (D vs DΔSH), suggesting that the site has strong activity without the extended A/T sequence. Therefore, it seems unlikely that the difference in LR pairing between D3EF and D3FE is due solely to a change in the activity of the Su(Hw) site.

Several studies have shown that depletion of insulator proteins only moderately affects either boundary activities or TAD structure in cell lines, or during Drosophila embryogenesis (Schwartz et al. 2012; Gambetta and Furlong 2018; Ramírez et al. 2018; Kaushal et al. 2021, 2022; Chathoth et al. 2022; Kahn et al. 2023), including effects on the eve locus and surrounding genes (Van Bortle et al. 2012). For example, Cavalheiro et al. (2023) showed that removing or reducing individually the insulator proteins CTCF, BEAF, or Cp190 during Drosophila embryogenesis had no major impact on the initial establishment of TAD structure, as assayed by Hi-C. This likely speaks to 2 issues. First, the plethora of insulator proteins in flies likely reflects a great deal of redundancy in their functions. Second, the effects of partially reducing insulator function are likely to be subtle, when viewed from a genome-scale perspective. For example, even when nhomie is completely removed from the eve locus, the changes seen at high resolution using MicroC are noticeable, but not dramatic, as are the changes in gene function that result (Ke et al. 2024). Nonetheless, these changes are functionally important, and do seem to have been the object of a considerable amount of selection pressure, given that the genome is subdivided into thousands of TADs by insulator elements.

Importance of the order of sub-elements

Our analysis of the D3EF combination uncovered another important feature of homie boundary activity. We found that inverting the order of the EF sub-elements in D3EF abolishes LR pairing interactions (Fig. 3 and Supplementary Fig. 5), but not enhancer blocking (Fig. 5b; D3FE actually shows significantly more activity at stage 12 than does D3EF). This is consistent with a model in which the order of binding sites for chromosomal architectural proteins like Su(Hw), CTCF, and Pita plays a central role in determining the specificity and stability of boundary:boundary pairing interactions. Many members of this particular class of polydactyl zinc finger proteins have self-interaction domains that can generate multimeric complexes, and potentially link boundary elements that have binding sites in common (Bonchuk et al. 2021; Fedotova et al. 2017). In this case, the ordering of those binding sites along the chromatin fiber would be expected to be important for both the specificity and stability of insulator pairing. The order of protein binding within the insulator is apparently not critical for its enhancer-blocking function.

Inverting EF in D3EF to give D3FE could interfere with LR pairing at several levels. In order for the eve enhancers to activate reporter expression at a distance of almost 150 kb, across about a dozen intervening TADs, the transgene homie must recognize potential partners (homie and/or nhomie) in the eve TAD and “initiate” physical pairing (Bing et al. 2024). Changing the order of E and F may make it impossible for proteins associated with each sub-element to simultaneously interact with their partners in endogenous homie and nhomie, and the initiation step could fail. It is also possible that the initial interaction might involve only a subset of proteins associated with each boundary. In this case, factors associated with a single sub-element might be sufficient to initiate pairing, but steric hindrance arising from the mis-ordered sub-elements could prevent the formation of a sufficient number of physical links (a zippering up) between the boundary elements to generate a stable pairing interaction. Furthermore, live imaging experiments indicate that after pairing interactions are established and stabilized, activation of the transgene reporter by the eve enhancers requires a further “compaction” that appears to involve forming contacts between the eve enhancers and the promoter of the reporter (Chen et al. 2018). Though less likely, this step might also be perturbed by steric hindrance arising from misaligned sub-element-mediated contacts.

Several mechanisms have been proposed to explain how boundaries block enhancer–promoter interactions. Boundaries could act as roadblocks or sinks (Geyer 1997; Blackwood and Kadonaga 1998; Bi and Broach 1999; Blanton et al. 2003; Gohl et al. 2011), in which case blocking activity would be autonomous, and depend only on the functional properties of the proteins bound to D3EF and D3FE. Alternatively, insulation could be achieved by organizing the chromatin fiber into looped domains (Bing et al. 2024). In this case, the blocking activity of D3EF and D3FE would depend upon whether they can pair with boundaries in the neighborhood of the transgene. Clearly, a 1D roadblock model for insulator function is insufficient, by itself, to explain the pattern of MicroC interactions within and between TADs (Bing et al. 2024), nor can it explain how E–P interactions between sequences that are 142 kb apart result in one reporter being activated, while the other, nearby reporter is not (Fujioka et al. 2016). More generally, a 1D roadblock could explain enhancer blocking only to the extent that enhancers and promoters are communicating along the chromatin fiber, while we know that enhancers can loop to promoters without such “linear” communication. The sink model is not easily compatible with the directional activity of insulator-dependent enhancer blocking, since a sink should compete for enhancer–promoter interactions even when it is not located between them, acting more like a silencer than an insulator. So, it seems likely that enhancer-blocking activity must be envisioned as a process in 3D that builds on the TAD organization of the chromosome, limiting the ability of enhancers and promoters to “find” each other by looping out the intervening DNA. Such a restriction on looping could be facilitated by the tendency of chromatin in different, insulator-defined TADs to coalesce together through copolymer cosegregation of similarly modified nucleosomes (Rowley et al. 2017; Rowley and Corces 2018; Harris et al. 2023; Ke et al. 2024).

A 1D roadblock model, while it is not sufficient to fully account for enhancer blocking, might be sufficient for PRE blocking. Silent chromatin is thought to spread along a chromosome through mechanisms involving the recruitment of modifiers to existing silent domains, which then modify histones in nearby, mostly adjacent nucleosomes (Blackledge and Klose 2021). This difference could explain our results showing that PRE blocking is less demanding than enhancer blocking, in terms of its sequence requirements, in that smaller fragments of homie are sufficient for PRE blocking than for enhancer blocking (Fig. 7). On the other hand, PRE pairing can result in the spreading of repressive chromatin in trans (Kraft et al. 2022). So, even for PRE blocking, 3D structures must be taken into account to fully understand the process. The requirements for LR pairing are more stringent still, which may reflect its requirement that distant sequences find each other and stably pair, in competition with closer pairwise arrangements of boundary elements, which would be more likely to find each other by the random motion of the chromatin fiber.

Data availability

All data underlying this publication are included in the text and figures. All plasmids and fly strains used in this study are available on request.

Supplemental material available at GENETICS online.

Funding

This work was supported by National Institutes of Health grants to P.S. (5R35GM126975) and J.B.J. (1R01GM137062).