Long-range insulator interactions can facilitate locally biased enhancer-promoter interactions

Previous studies showed that homie and nhomie can mediate long-range regulatory interactions when included in a reporter transgene that is inserted within several Mb of the eve locus [17,24]. Because the pairing interactions of the eve insulators are orientation-dependent, one of the two reporters in a dual reporter transgene is preferentially activated by the eve enhancers [17]. Although there are eve enhancers located both upstream (on the nhomie side) and downstream (on the homie side) of the eve transcribed region (Fig 1A), there is no apparent bias in their ability to activate reporter expression. This is reflected in the pattern of physical contacts with the transgene seen in MicroC experiments, which extend across the entire eve locus, with peaks at the insulators, nhomie and homie [26]. One interpretation of the MicroC data is that the transgene insulator interacts simultaneously with endogenous homie and nhomie, forming a tripartite complex. An alternative possibility is that there are three different pairwise interactions, transgene insulator with either endogenous nhomie or homie, and endogenous nhomie with homie, generating three different types of loops that potentially compete with each other. This led us to test how the transgene would interact with the eve enhancers if there were only one endogenous eve insulator.

Endogenous homie and nhomie are required for the long-range interaction

To answer this question, we first deleted either endogenous nhomie or homie. We used CRISPR [30,31] to remove endogenous nhomie (creating ∆N; see Fig 1C for map) [25], but were unsuccessful at obtaining a similarly “clean” removal of endogenous homie. However, we were able to obtain a small deletion that includes endogenous homie, along with the 3’ eve PRE and the TER94 basal promoter region and first exon (Fig 1B; see S1 Fig for sequence). This mutant chromosome (∆H) is homozygous lethal after embryogenesis. In earlier studies, we showed that two copies of a transgenic eve rescue construct that did not include homie were able to rescue the lethality of eve homozygous null mutants (both eve^R13/eve^R13 and eve^R13/Df(2R)eve) [32]. This rescue construct contained the eve locus from –6.4 to +8.4 kb (relative to eve + 1 transcription start site), which included all of the eve enhancers (the 3’-most of which is the RP2 neuronal element), but not nhomie or homie. The 3’ endpoint of this rescue construct (2R:9987733) is close to the 5’ endpoint of the ∆H deletion (2R:9987987), which also retains this neuronal element but is missing homie. Therefore, it seems likely that the homozygous ∆H chromosome is able to provide sufficient eve function for full rescue of viability, and so the loss of eve function is not the cause of the ∆H lethality. In order to test whether the lethality of ∆H comes instead from removal of the TER94 promoter, we complemented it with an eve-deficient chromosome, Df(2R)eve, carrying an intact TER94 locus. To do this, we first confirmed that Df(2R)eve is a complete null for eve function [33], but has intact TER94 function (Fig 1B; see S1 Fig for junction sequence). When the ∆H chromosome was placed over Df(2R)eve, some adult flies survived. Some of the surviving adults showed abnormal segmentation that is not evident in Df(2R)eve heterozygotes. Nonetheless, rescue to adulthood by Df(2R)eve indicates that the recessive lethality of ∆H is mainly caused by lack of TER94 function. Segmentation defects in ∆H/Df(2R)eve adult flies suggest that the eve function provided by ∆H is significantly weakened, and this is confirmed below by analyzing embryonic defects. For further analysis, we generated a chromosome lacking both homie and nhomie (∆N∆H), replacing nhomie on the ∆H chromosome with lambda DNA, using CRISPR (Fig 1C; see also Materials and Methods). The fertility of ∆N∆H/Df(2R)eve seems to be lower than that of ∆H/Df(2R)eve, as we could maintain ∆N∆H/Df(2R)eve for only a few generations.

In order to analyze interactions between transgenes and the modified endogenous eve locus, we used CRISPR to insert a pair of attP sites (flanking a mini-white marker gene) onto the wild-type (wt), ∆N, ∆H, and ∆N∆H chromosomes, near our previously used attP site (142kb upstream of eve). The mini-white marker was then replaced with our lacZ-gfp dual reporter transgene vector [11,17] (Fig 1D). In previous studies [17,26], we showed that transgene activation is biased according to how homie or nhomie is oriented in the transgene relative to the two reporter genes. The reporter located either upstream of homie or downstream of nhomie is preferentially activated by the eve enhancers (see Fig 1A and 1D; red or green arrow shows the orientation of homie or nhomie, respectively). Throughout this study, we used the two transgenes that preferentially express lacZ, which we call Z-homie-G and G-nhomie-Z. The negative control (lambda DNA in place of a transgenic insulator) did not show any eve-like lacZ expression (S2 Fig). However, various patterns of ectopic expression that we previously reported are seen [11]. S3 Fig shows positive control expression of each of the two reporters with homie and nhomie at the new attP site on a wild-type chromosome. In both lines, lacZ is expressed as 7 stripes at stages 5–9 (S3 Fig, stages 5 and 7 are shown), and in a tissue-specific eve pattern in the mesoderm, anal plate ring (APR), and central nervous system (CNS) at later stages (S3 Fig, stages 11 and 13 are shown). gfp is not expressed in the eve pattern, instead, ventral midline expression in the CNS driven by an enhancer near the insertion site is observed. In each case, these expression patterns match previously reported results with the same dual-reporter transgenes inserted at the original attP site [11,17].

In previous experiments, we found that long-range (LR) interactions between reporters in the transgene and the eve enhancers only take place when either homie or nhomie is included in the transgene [11,17]. We concluded that the LR interaction between the eve locus and transgenic reporter genes is mediated by directional (orientation-specific) pairing between homie or nhomie in transgenes and their endogenous counterparts. Further analysis using live-cell imaging [27] and physical interaction studies using MicroC [26] were consistent with this explanation. This model predicts that the LR interaction will disappear without endogenous homie and nhomie. So, we tested whether removing both endogenous homie and nhomie (∆N∆H) causes loss of the LR interaction. As predicted, no eve-like lacZ expression was seen from either Z-homie-G or G-nhomie-Z when they were carried on the ∆N∆H chromosome (Fig 2, compare wild type and ∆N∆H). These findings indicate that physical pairing between the endogenous and transgenic copies of the eve insulators are required for the LR interaction.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 2. Transgenic homie and nhomie require endogenous homie and nhomie for long-range interaction with eve enhancers, which can occur in trans at later embryonic stages.

RNA expression from lacZ reporter transgenes carrying homie (Z-homie-G) or nhomie (G-nhomie-Z) located at –142kb on homozygous wild-type (wt) or ∆N∆H chromosomes, and heterozygous ∆N∆H over wild-type (∆N∆H/ + , showing a trans interaction). Arrowheads indicate eve tissue-specific expression: red: APR; blue: CNS; green: mesoderm. Embryonic stages 5, 7, 11, and 13 are shown. Ventral views of stage 13 are shown in the bottom row. Scale bar: 50μm.

https://doi.org/10.1371/journal.pgen.1011940.g002

To test this idea further we took advantage of the fact that once homologs in somatic cells pair with each other, trans-regulatory interactions (transvection) are observed if enhancers on one homolog are brought into close proximity with target genes on the other homolog. Thus, if insulators at the eve locus mediate interactions with the transgene, then it should be possible to restore lacZ expression with a wild-type eve locus is trans to either Z-homie-G,∆N∆H or G-nhomie-Z,∆N∆H. Homolog pairing is limited at the blastoderm stage, and we do not detect eve-like lacZ expression. However, later in development, during stages 11–13, the APR enhancer drives lacZ expression in the posterior, and eve-like expression is also detected in the CNS and mesoderm (Fig 2, compare ∆N∆H and ∆N∆H/+).

Pairwise insulator interactions bias enhancer-promoter interactions

We next tested whether endogenous homie and nhomie individually can mediate the LR interaction with transgenic copies of either themselves or each other. To do this, mini-white-carrying attP sites were introduced on each of the ∆H and ∆N chromosomes, as described above (Fig 1D). At stage 5, each eve stripe is controlled by a distinct “early” stripe enhancer [32,34–38]. As diagrammed in Fig 1A, the enhancers for stripes 3 + 7 and 2 are located upstream of the eve transcription unit, while enhancers for stripes 4 + 6, 1, and 5 are located downstream. An imprecisely mapped stripe 1 enhancer or “influencer” near the 5’ end of the eve locus has also been reported (not included in Fig 1D) [38]. At stages 6–8, the 7 stripes are under control of a single “late stripe” enhancer located near nhomie [37,38]. Tissue-specific enhancers for APR, mesoderm, and CNS expression are located downstream of the eve transcription unit [32,34].

On either ∆N or ∆H, Z-homie-G and G-nhomie-Z each showed an LR interaction, expressing lacZ in a partial eve pattern. Importantly, however, these patterns differed from the interactions seen with the intact eve locus (Fig 3). In a wild-type background (Fig 3A and 3B, wt), at stage 5, the level of lacZ expression from either Z-homie-G or G-nhomie-Z is close to equal among the different stripes, except that the stripe 2 enhancer typically drives less lacZ expression than the others at this stage. This could be due to promoter competition, as the stripe 2 enhancer is located next to the endogenous eve promoter. At stage 7, the 7 stripes are also expressed at similar levels. All aspects of later-stage tissue-specific eve expression are also observed.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 3. When one endogenous eve insulator is removed, long-range interactions become biased toward enhancers near the remaining endogenous insulator.

A. RNA expression from lacZ reporter transgenes carrying homie (Z-homie-G, left panel) or nhomie (G-nhomie-Z, right panel) located at –142kb on homozygous wild-type (wt), ∆N, or ∆H chromosomes. Arrowheads indicate eve tissue-specific expression: red: APR; blue: CNS; green: mesoderm. Embryonic stages 5, 7, 11, and 13 are shown. Ventral views of stage 13 are shown in the bottom row. Scale bar: 50μm. B. Model for the pairwise LR interaction. Top panel: location of the eve enhancers. Bottom 4 diagrams (simple pair-wise interaction model): blue boxes represent the transgenes (Z-N/H-G: Z-homie-G or G-nhomie-Z at -142kb attP site), interacting with either homie (thick red arrow) or nhomie (thick green arrow) at the eve locus. Red curved arrows indicate activation of the lacZ transgene reporter by the eve enhancers. The thickness of these arrows indicates the strength/frequency of interaction. wt: nhomie or homie in the transgene interacts alternatively with either endogenous nhomie or homie. The 5’ enhancers activate lacZ when the transgene is paired with nhomie. Similarly, the 3’ enhancers activate lacZ when the transgene is paired with homie. ∆N: nhomie/homie in the transgene interacts with endogenous homie more frequently (or more stably) than it does in wild type, where there is competition with endogenous nhomie. As a result, the 3’ enhancers activate lacZ more than in wild type. ∆H: nhomie/homie in the transgene interacts with endogenous nhomie more frequently than in wild type, where there is competition with endogenous homie. As a result, the 5’ enhancers activate lacZ more than in wild type.

https://doi.org/10.1371/journal.pgen.1011940.g003

The pattern of lacZ expression in ∆N differs from that in a wild-type background in that the level of lacZ expression in the seven stripes at stage 5 is much more uneven. With either Z-homie-G or G-nhomie-Z on ∆N (Fig 3A and 3B, ∆N), lacZ is expressed more strongly than in wild-type in stripes 4, 5, and 6. In contrast, expression is weaker than in wild-type in stripes 2, 3, and 7. At stage 7, the 7-stripe pattern is observed, but stripes 4, 5, and 6 appear stronger, likely because there are transcripts persisting from stage 5. The eve tissue-specific pattern at stages 11–13 in mesodermal, CNS, and APR is also observed, and the level of lacZ expression again appears stronger than when the eve locus is intact (Fig 3A, compare wild type and ∆N). Strikingly, the enhancers for each of the aspects of the eve pattern that are stronger than in wild type are located 3’ of the eve transcription unit, closer to the remaining endogenous insulator, homie. However, there are two exceptions. First, although the precisely mapped stripe 1 enhancer is located near homie, stripe 1 expression was not as strong as that in stripes 4, 5, and 6. This is likely due to the location of the other stripe 1 element nearer nhomie, as noted above. Second, although the late-stripe enhancer is located near nhomie, this pattern appears to be less reduced than that of early stripes 2, 3, and 7. The likely explanation for this anomaly is a combination of proximity and competition with the endogenous eve promoter: in ∆N, although the late-stripe enhancer is further from the tethering point of the transgene (at endogenous homie) than are the other enhancers (for early stripes 2, 3, and 7), there is also less competition with the endogenous eve promoter, which is further away. Therefore, the reduced competition outweighs the effect of the extra distance, making the late-stripe pattern (minus the extra contribution from the early stripes) more similar between ∆N and wild type than is the early stripe 2, 3, and 7 expression pattern.

A reciprocal pattern of expression is observed for the transgene combinations with ∆H (Fig 3A and 3B, ∆H). At stage 5, lacZ expression from either Z-homie-G or G-nhomie-Z is stronger in stripes 2, 3, and 7, and weaker in stripes 4, 5, and 6, than they are in wild type. The late-stripe expression at stage 7 is similar to that seen in wild type. Furthermore, APR expression at stages 11–13 is less frequently observed, and then only faintly (Fig 3A, compare wild type and ∆H). The enhancers for stripes 2, 3, and 7, as well as the late-stripe enhancer, are all located upstream of the eve promoter, closer to nhomie. Despite the location near homie of the mapped stripe 1 enhancer, stripe 1 expression is stronger than that of the other early stripes with enhancers located in the downstream region. As mentioned above, a second enhancer that contributes to stripe 1 expression was reported upstream of the eve promoter [38]. This can explain why stripe 1 does not behave like the other stripes that have enhancers only downstream of the eve promoter.

Transgene – endogenous eve interactions: cis vs. trans

The distinct bias in reporter gene activity in either ∆H or ∆N raises an intriguing question: when homie- or nhomie-carrying transgenes engage in LR interactions in a wild-type background, do they pair with both of the endogenous insulators at the same time, or do they pair with only one of them at a time? The idea that interactions between insulators might be strictly pairwise would fit with results in which three different insulators were tested for partner preferences in an insulator competition assay [12].

If insulator pairing interactions are exclusively pairwise, then there should be a competition between potential pairing partners when there are three or more of them. As a test of this idea, we manipulated the number of competing insulators by generating trans combinations between homologous chromosomes, one carrying the transgene and a wild-type eve locus, and the other carrying either a wild-type or mutant version of the eve locus. The homie- or nhomie-carrying transgene can interact with the endogenous eve locus either in cis or in trans (Fig 2), particularly at later embryonic stages, when wild-type homologs are fully paired. If there is a competition between the cis and trans interactions, then later-stage interactions might become stronger without a trans copy of the eve locus, resulting in stronger expression in the mesoderm, CNS, and APR. To remove the trans copy of endogenous eve completely, we used the large chromosomal deficiency Df(2R)BSC158 (BSC158; see Fig 1B) [39,40]. When we compared lacZ expression in Z-homie-G/ + embryos with Z-homie-G/BSC158, we saw stronger expression with Z-homie-G/BSC158 in the CNS over that seen with the wild-type chromosome (marked with Sco). This occurred with either nhomie or homie in the transgene (Fig 4: cis interaction only, 2^nd and 5^th columns, vs. in cis with a trans copy of eve on the Sco-carrying chromosome, 1^st and 4^th columns). At stages 5–7, the stripe expression may also be increased, which is consistent with some trans interactions occurring even at these early stages [23].

Download:

• PNG
  larger image
• TIFF
  original image

Fig 4. Late-stage interactions, especially in the CNS, become stronger without a trans copy of the eve locus.

RNA expression from heterozygous lacZ reporter transgenes carrying homie (Z-homie-G, left panel) or nhomie (G-nhomie-Z, right panel) located at –142kb, in trans with either a chromosome carrying a wild-type eve locus (reporter transgene/+), Df(2R)BSC158 (reporter transgene/BSC), or ∆N∆H (Z-homie-G/∆N∆H). Embryonic stages 5, 7, 11, 13 are shown. Ventral views of stage 13 are shown in the bottom row. Note that lacZ expression, especially in the CNS at stage 13 (blue arrowheads), is stronger when the eve locus is absent in trans (compare BSC to +), but not when only the eve insulators are missing on the trans chromosome (∆N∆H). Scale bar: 50μm.

https://doi.org/10.1371/journal.pgen.1011940.g004

These results are consistent with the occurrence of trans competition between multiple possible pairwise interactions, but it is not the only possible explanation. To test specifically whether insulator pairing competition is responsible for this effect, we used the ∆N∆H chromosome in trans to remove the potentially competing copies of the eve insulators. Here, when we compared Z-homie-G/+ with Z-homie-G/∆N∆H, we no longer saw the increased expression (Fig 4, 3^rd column vs. 1^st column). These data indicate that the increased expression with BSC158 is probably not due to competition between the transgene and the cis and trans copies of the endogenous insulators, but to a disruption of normal homolog pairing in BSC158 heterozygotes (see Discussion).

MicroC analysis of physical contacts between homie- and nhomie-containing transgenes and the eve locus

In the ∆N or ∆H deletions, transgene reporter expression is preferentially activated by enhancers located on the same side of the eve transcription unit as the remaining endogenous insulator (Fig 3). This bias in reporter expression is strikingly reflected in the pattern of physical interactions between the transgene reporter and sequences in the eve locus (Fig 5). We used MicroC to analyze the physical contacts that are generated between each of the transgenes, G-nhomie-Z or Z-homie-G, and sequences in the eve locus, both in a wild-type eve background and in the ∆N lines. Panels 5A-D show the MicroC contact profiles between the transgene and eve, while panels 5E and F show viewpoints from the transgene lacZ reporter. In a wild-type background, the lacZ sequences in Z-homie-G (Fig 5A, top panel and C, 3^rd panel) and G-nhomie-Z (Fig 5B, top panel and C, top panel) physically engage eve sequences located throughout the eve locus, with endpoint peaks that coincide with endogenous nhomie and homie.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 5. Physical interactions between the transgene and eve in wild type and ∆N.

A. MicroC contact profiles between Z-homie-G in wild type (top) and ∆N (bottom). A blow-up of the interaction between the transgene and eve in each background is shown on the right of each panel. B. MicroC contact profiles between G-nhomie-L in wild type (top) and ∆N (bottom). A blow-up of the interaction between the transgene and eve in each background is shown on the right of each panel. C. View points from the transgene lacZ reporter. From top to bottom, G-nhomie-Z in wild type, G-nhomie-Z in ∆N, Z-homie-G in wild type, and Z-homie-G in ∆N. D. View points from the transgene gfp reporter. From top to bottom, G-nhomie-Z in wild type, G-nhomie-Z in ∆N, Z-homie-G in wild type, and Z-homie-G in ∆N. E. Left panel: the ratio of contacts between lacZ and enhancers upstream (5’ enhancers) and downstream (3’ enhancers) of the eve gene in G-nhomie-Z, in a wild-type and a ∆N background. Right panel: the ratio of contacts between lacZ and enhancers upstream (5’ enhancers) and downstream (3’ enhancers) of the eve gene in Z-homie-G, in a wild-type and a ∆N background. F. Normalized contacts between eve and lacZ in either G-nhomie-Z or Z-homie-G, and in either wild type or ∆N, as indicated.

https://doi.org/10.1371/journal.pgen.1011940.g005

While the loop configuration that is generated by the pairing of the transgene with nhomie and homie in the wild-type eve locus is unclear, the pairing interactions between homie and either Z-homie-G or G-nhomie-Z in ∆N should generate a stem-loop structure (illustrated in Fig 6A, 6B), while they would generate circle-loop in ∆H (illustrated in Fig 6C and 6D). In these topologies, sequences within the loop on either side near the paired insulators would be brought into close proximity. Thus, without elements elsewhere in the eve locus that can anchor sustained contacts with the transgene, one would expect that, as the distance increases from the paired insulators to the interacting sequences in the transgene and eve, the weaker the contacts between them will be. Indeed, this is what is observed in ∆N (Fig 5A and 5B: 2^nd panels; Fig 5C: 2^nd and 4^th panels). The physical interactions between lacZ and eve peak at the 3’ end of the eve locus (where homie is located). The interactions then become progressively weaker, the farther the eve sequences are located from the paired insulators (i.e., moving to the left from the peak). We quantitated the contact frequency between lacZ and sequences upstream and downstream of the eve transcription unit in the wild-type and ∆N backgrounds. This analysis showed that there is a bias in favor of the enhancers downstream of the eve transcription unit (Fig 5E). We also examined the interactions between the gfp reporter (which is expressed very weakly relative to lacZ) and eve (Fig 5D). In wild type, there are gfp peaks at nhomie and homie, while interactions within the eve locus are lower than those with lacZ as the viewpoint. In ∆N, interactions are partially retained on the homie side, but much reduced over the remainder of the eve locus. Additionally, gfp contacts sequences beyond the two boundaries of the neighboring TADs more than lacZ does, especially toward the 3’ neighboring TAD. This is consistent with the fact that the stem-loop topology brings sequences beyond the paired insulators, which include gfp, into closer proximity (illustrated in Fig 6A and 6B).

Download:

• PNG
  larger image
• TIFF
  original image

Fig 6. Loop topologies generated by the pairing of transgene and endogenous eve insulators.

A. Z-homie-G ∆N: Pairing of the dual reporter containing the homie insulator with homie on the ∆N chromosome. Homologous homie-homie pairing (head-to-head) generates a stem-loop. Since transgene homie is oriented in the dual reporter so that it is “downstream” of the lacZ reporter, lacZ interacts preferentially with the eve enhancers. B. G-nhomie-Z ∆N: Pairing of the dual reporter containing the nhomie insulator with homie on the ∆N chromosome. Heterologous nhomie-homie pairing (head-to-tail) generates a stem-loop. Since transgene nhomie is oriented in the dual reporter so that it is “upstream” of the lacZ reporter, lacZ interacts preferentially with the eve enhancers. C. Z-homie-G ∆H: Pairing of the dual reporter containing homie to nhomie on the ∆H chromosome. Heterologous homie-nhomie pairing (head-to-tail) generates a circle-loop. Since transgene homie is oriented in the dual reporter so that it is “downstream” of the lacZ reporter, lacZ interacts preferentially with the eve enhancers. D. G-nhomie-Z ∆H: Pairing of the dual reporter containing the nhomie insulator to nhomie on the ∆H chromosome. Homologous nhomie-nhomie pairing (head-to-head) generates a circle-loop. Since transgene nhomie is oriented in the dual reporter so that it is “upstream” of the lacZ reporter, lacZ interacts preferentially with the eve enhancers. E. Z-homie-G wt: Pairing of the dual reporter containing homie with the wild-type eve locus. In the endogenous eve locus, heterologous homie-nhomie pairing (head-to-tail) generates a stem-loop. Transgene homie pairs head-to-head with endogenous homie, while it pairs head-to-tail with endogenous nhomie, generating a complicated multi-loop topology. Since transgene homie is oriented in the dual reporter so that it is “downstream” of the lacZ reporter, lacZ interacts preferentially with the eve enhancers. F. G-nhomie-Z wt: Pairing of the dual reporter containing the nhomie insulator to the eve locus. In the endogenous eve locus, heterologous homie-nhomie pairing (head-to-tail) generates a stem-loop. Transgene nhomie pairs head-to-head with endogenous nhomie, while it pairs head-to-tail with endogenous homie, generating a complicated multi-loop topology. Since transgene nhomie is oriented in the dual reporter so that it is “upstream” of the lacZ reporter, lacZ interacts preferentially with the eve enhancers.

https://doi.org/10.1371/journal.pgen.1011940.g006

As noted above, one possible explanation for the increase in lacZ expression in the single-insulator deletions is that insulator pairing is strictly pairwise. In this model, the transgene insulator in wild type would have to compete with the pairwise interaction between the endogenous insulators. Since the endogenous insulators are less than 20 kb apart, pairing between them would be favored. In contrast, this competition is removed in the insulator deletions. This could greatly boost pairing between the transgene insulator and the remaining endogenous insulator, if the endogenous insulators are paired most of the time in wild type. So, if insulator pairing were strictly pairwise, there would likely be a significant increase in the contact frequency between the lacZ reporter and eve in ∆N compared to wild type. However, the frequency of physical contacts between the lacZ reporter and eve in ∆N is not significantly different from wild type (Fig 5F). This would suggest that we need an alternative to the simple pairwise interaction model. In one such model, the topology in wild type, when the transgene is “engaged” with the endogenous locus, involves the formation of a 3-way complex (illustrated in Fig 6E and 6F). In the simplest form of this model, the expression of the lacZ reporter in an eve-like pattern is facilitated by this single 3-way complex, rather than being a composite of two alternative pairwise complexes (see Discussion for more detail and additional considerations).

Effects of endogenous insulator removal on eve function

Does removal of an insulator from the eve locus affect its function? In a previous study using the nhomie deletion ∆N, we found that homozygous mutants were viable and fertile. However, ∆N is weakly haploinsufficient, so that when in trans to the BSC158 deletion, nearly 10% of embryos were missing two or more ventral abdominal denticle bands [25]. To extend this analysis, we have now analyzed embryos derived from self-crossing ∆N∆H/SM6a,Cy, ∆H/SM6a,Cy, and homozygous ∆N, for cuticular segmentation defects. As controls, we used Sco/SM6a,Cy (“Sco/Cy”), controlling for the SM6a chromosome in two of the lines, and yw, as the control for the ∆N self-cross. In all these cases, more than 98% of embryos showed a wild-type larval cuticle phenotype at the end of embryogenesis (S4 Fig). In contrast, embryos from a cross of Sco/Cy x BSC158/Cy showed a missing ventral abdominal denticle band (most of them in the abdominal segment A6) in ~10–25% of embryos (Fig 7, “WT” in “% affected”). The Sco/BSC158 and Cy/BSC158 progeny (which each have a single copy of eve) together represent about 50% of these embryos, so this phenotype is not fully penetrant. However, the fact that progeny from the above self-crosses did not show such segmentation defects suggests that homozygotes for the insulator deletions have more than one wild-type copy’s worth of normal eve activity (see below and Discussion).

Download:

• PNG
  larger image
• TIFF
  original image

Fig 7. Deleting both nhomie and homie from endogenous eve compromises function more than does deleting only homie or nhomie, when the genetic background is sensitized to a reduction in eve function.

Sco/Cy, ∆N/Cy, ∆H/Cy, and ∆N∆H/Cy lines were crossed with Df(2R)BSC158/Cy, and embryonic cuticle defects were counted. The parental source of the chromosomes analyzed is indicated as –f (female) or –m (male). The percentage of embryos showing segmentation defects (left graph, % affected) and the average number of ventral abdominal denticle bands deleted per non-wild-type embryo (right graph, severity score) are shown as box-and-whiskers plots. The pair-wise significance of differences (p-values) is indicated as: *: p < .05, **: p < .01, ***: p < .001, ****: p < .0001, 6*: p < 10^–6. The data are from two separate 3-way comparison experiments, involving [Sco/Cy, ∆N/Cy, and ∆H/Cy] and [Sco/Cy, ∆H/Cy, and ∆N∆H/Cy], and these two data sets were combined as described in Materials and Methods. The number of cuticle preparations used (n) and the total number of embryos counted were as follows: WT-f (n = 7, 1414 embryos), ∆N-f (n = 6, 1676), ∆H-f (n = 7, 1885), ∆N∆H-f (n = 8, 1391), WT-m (n = 6, 1311), ∆N-m (n = 6, 1676), ∆H-m (n = 6, 1928), ∆N∆H-m (n = 6, 2541).

https://doi.org/10.1371/journal.pgen.1011940.g007

Since we found that ∆N is weakly haploinsufficient, we crossed each of the three insulator deletion lines with a series of different eve mutant chromosomes, as diagrammed in Fig 1B. For ∆nhomie, we used ∆N/SM6a,Cy in order to make the crosses equivalent, even though the ∆N lines are homozygous viable and fertile. To compare the number and severity of segmentation defects, we calculated two measures: the percentage of affected embryos in each population (Fig 7A, % affected) and the average number of ventral abdominal denticle bands deleted in the affected population (Fig 7B, severity score). When placed over the BSC158 chromosome, both ∆H and ∆N∆H significantly increased the percentage of embryos showing segmentation defects over that of the control Sco/Cy (“WT”, Fig 7A and S1 File) in at least one direction of the cross. In order to present all of the data with this set of crosses together, the data from the experimental set [WT, ∆N, and ∆H] (S5A Fig) were merged with those from the experimental set [WT, ∆H, and ∆N∆H] to produce Fig 7 (see Materials and Methods). The statistical analysis used the WT data from the latter experiment, which had a larger variance than the WT data from the former experiment. In the former experiment, ∆N did show a significant increase in the percentage of defects over WT (S5A Fig and S1 File), even though this same comparison in the merged data set did not show a significant difference (Fig 7 and S1 File). Importantly, the severity of defects in each of the 3 cases was significantly increased over that of WT. Furthermore, the percentage of defective embryos and the severity of segmentation defects were both greater with ∆N∆H than with either ∆N or ∆H alone (Fig 7B and S1 File).

Since the BSC158 deficiency is large and deletes many other genes besides eve, we further tested these effects for ∆H and ∆N∆H using both Df(2R)eve and eve^R13 [41]. Df(2R)eve deletes most of the eve locus plus several kilobases upstream (sequences of junction fragments are shown in S1 Fig), while eve^R13 is a point mutant causing premature termination of the Eve protein [32]. The effects of combining these two eve mutants with the insulator deletions are similar to those observed for the BSC158 deficiency (S5 and S6 Figs). This indicates that the BSC158 results can be attributed to the removal of one functional copy of eve in combination with the loss (in trans) of one or both eve insulators in the remaining copy, which is only partially functional without the insulators. This interaction is shown in a box plot for Df(2R)eve in S5B Fig, and a stacked graph in S6B Fig, while S5C and S6C Figs show the data for eve^R13. We also note that in the Df(2R)eve crosses, the effect of ∆N∆H (S5B and S6B Figs, and S1 File) is significantly more severe when it comes from the female parent, for unknown reasons.

In all of the cases described above, virtually all of the missing abdominal denticle bands occurred in even-numbered segments (A2, A4, and A6). This is significant because it is mostly the even-numbered segments that are deleted in eve deficiency mutants [33,41]. This indicates that the defects are most likely due to a loss of eve function in early embryos.

Endogenous insulator removal allows the eve Polycomb domain to spread

In previous studies, we found that homie blocked the spread of the repressive Polycomb histone modification H3K27me3 written by Polycomb Repressive Complex 2 [11]. We also showed that removing homie from a modified transgenic eve “pseudo-locus” (in which the eve coding region is replaced with a lacZ reporter gene, and the TER94 coding region is replaced with a gfp reporter) caused the spreading of H3K27me3 into TER94, and reduction of TER94-gfp transcript levels [51]. Those data suggested that homie prevents the spreading of repressive H3K27me3 in order to prevent the repression of TER94. Here, we tested whether we could see similar spreading of H3K27me3 outside of the endogenous eve locus in the absence of nhomie and homie, using the ∆N∆H mutants. It is important to note that this experiment differs from the transgenic experiment in that ∆N∆H is homozygous in only 1/4 of the embryos from the cross and is heterozygous in another 1/2 of the population, so H3K27me3 is expected to increase outside the eve locus in at most 1/2 of the chromosomes being analyzed in the experiment. Despite this difference, which is expected to reduce measured Polycomb spreading by at least 2-fold, we measured a significant increase in H3K27me3 outside the endogenous eve locus in both directions (Fig 8: 2.0-fold increase upstream of nhomie, 2.8-fold increase downstream of homie) with the removal of both nhomie and homie. In the previous experiments using the eve pseudo-locus [51], we found that removing homie (along with the 3’ PRE) resulted in a 5.7-fold increase in H3K27me3 levels just downstream. Since in the present study, less than half of the chromosomes in the analyzed population have homie (along with the PRE) deleted, whereas in that study all the chromosomes had the deletion, this compares very well to the 2.8-fold change seen here. So, while these increases are not dramatic, they may be significant for the proper expression of neighboring genes.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 8. H3K27me3 levels upstream and downstream of eve increase when nhomie and homie are removed.

Both Sco/Cy (WT) and ∆N∆H/Cy (∆∆) lines were subjected to ChIP analysis using anti-H3K27me3 (me3) antibody or negative control IgG (IgG). Immunoprecipitated DNA was analyzed by qPCR at 4 positions, as follows. N: upstream of the nhomie deletion (nhomie is replaced by λ DNA), eve: eve coding region, H: downstream of the homie deletion, en: the engrailed coding region (positive control). The average and standard deviation from 3 independent experiments are shown as % input. The pair-wise significance of differences (p-values) between ∆∆ and WT is indicated as: *: p < .05, **: p < .01. The increase upstream of nhomie (∆∆/ WT, after each IgG background is subtracted) is 2.00-fold, while that downstream of homie is 2.81-fold, as indicated.

https://doi.org/10.1371/journal.pgen.1011940.g008

Insulators function to isolate regulatory domains

Insulators function to prevent enhancers or silencers in one TAD from influencing genes in neighboring TADs. Our analysis of the phenotypic consequences of the homie and nhomie deletions provides strong support for this hypothesis. In Ke, et al. [25], we showed that the eif3j gene, which is located upstream of the nhomie insulator, is inappropriately activated by eve enhancers when nhomie is deleted. In addition to restricting the activity of enhancers, another key function of the eve insulators in their endogenous setting is to “contain” repressive chromatin within the eve domain, presumably thereby facilitating the full levels of expression of neighboring transcription units. We previously showed that homie has this function in the context of an eve “pseudo-locus” transgene, where a modest increase in H3K27me3 levels was sufficient to significantly reduce TER94 promoter activity [51]. Consistent with that study, there is a similar spreading of the eve Polycomb domain in both directions in the absence of the endogenous eve insulators, measured here as an increase in the characteristic histone modification written by Polycomb Repressive Complex 2, H3K27me3 (Fig 8).

One of the key activities of insulators is the establishment of units of independent gene activity [6,52–54]. The studies here also support a related function, that insulators help to facilitate full levels of expression of “insulated” transcription units. For eve, this manifests itself in both the single and the double insulator deletions as an increase in embryonic defects. These defects are characteristic of reduced eve function at a critical early stage (Figs 7, S5, and S6), when pattern formation is highly dependent on properly formed “pair-rule” eve stripes [33,42–45,55]. The establishment of fully functional parasegment spacing is known to be highly sensitive to the levels of eve function [46,47] relative to that of its “complementary” pair-rule gene fushi tarazu (ftz) [45,56]. Specifically, the positioning of the 14 stripes of the segment polarity gene engrailed (en), which is immediately downstream of eve and ftz in the segmentation hierarchy [46,47], become abnormally spaced when eve function is reduced [42–46] or when ftz function is increased [57]. In both cases, each odd-numbered en stripe becomes closer to the next (posteriorly adjacent) even-numbered en stripe. This effect occurs in heterozygotes for eve null mutations, yet this abnormal spacing is almost always corrected as pattern formation continues, and few abnormal embryos are seen at the end of embryogenesis (Figs 7, S5, and S6). However, when eve function is reduced further, a critical point is reached where the narrow spacing is no longer efficiently corrected, and defects increase significantly. This is seen as “skips” in the pattern of denticles within even-numbered abdominal segments (this is the phenotype for which eve is named, which was based on the initially isolated, strong hypomorphic allele eve¹) [33,55]. The results here show that homie deletion causes segmentation defects in a sensitized background with one copy of endogenous eve function. Deletion of nhomie in addition to homie increases the severity of the phenotype, indicating that proper activation of eve, mediated by efficient interaction between eve enhancers and the promoter, requires both homie and nhomie. We note that because our homie deletion is imprecise, the extra deleted sequence could contribute to its more severe phenotype relative to deleting nhomie alone. For example, the deletion of the eve 3’ PRE might reduce the stability of eve enhancer-promoter interactions. Nonetheless, our results indicate that the eve insulators, both together and individually, contribute to establishing the proper level of expression and function of the eve locus. While interfering with each of these three functions of the eve insulators, namely facilitating full eve activity, preventing the spread of repressive chromatin, and restricting the activity of eve enhancers, produce incompletely penetrant defects on their own, they are probably a sufficient driving force for the evolution of a well-insulated eve TAD. Similarly subtle yet important functions of insulators are likely to apply genome-wide.

Consistent with our results with endogenous homie and nhomie, a recent study showed that LR enhancer-promoter interactions were increased by flanking a transgene with homie and nhomie [58]. How do the eve insulators upregulate transcription? Loss of TAD boundaries could cause more frequent accessing of neighboring promoters by eve enhancers. Where eve enhancer activity is limiting, these interactions could reduce eve transcriptional output by promoter competition. Consistent with this idea, Chen et al. [27] found that the presence of a homie-containing reporter transgene at -142 kb, similar to those used in this study, induced a weak loss-of-function phenotype, much like the insulator deletions described here.

Competition between cis and trans insulator interactions

We observed that transgenic reporter expression is stronger in late-stage embryos when a trans copy of the eve locus is absent (due to the chromosomal deficiency BSC158). However, when the ∆N∆H chromosome was in trans, the expression was not increased (Fig 4), consistent with a lack of competition. Therefore, this effect is probably not due to competition between the trans copy of the endogenous insulators and the pairing of the transgene with the cis copy of the endogenous insulators. However, it should be noted that trans pairing has been suggested to be weaker than cis pairing [59], potentially explaining this result without dismissing the importance of pairwise competition among insulators.

A more likely explanation for the effect with Df(2R)BSC158 is that with a large chromosomal deficiency, homologous chromosomes are no longer paired in the “missing” region, which here contains the eve locus plus about 100 kb upstream of eve (and ends about 40 kb downstream of the transgene insertion site; see Fig 1 legend for precise locations). This may lead to a greater flexibility of the chromatin fiber for most of the region between the eve locus and the transgene, increasing the frequency with which they find each other by random flexing of the chromosome. This flexibility could thus lead to a more efficient LR cis interaction [60], and increase the number of cells that express the transgene reporter in an eve pattern. Furthermore, the wild-type chromosome is expected to pair with the BSC158 deficiency chromosome both upstream and downstream of the missing region, creating a “loop out” of the hemizygous region (which contains the eve locus). The transgene insertion site, which is upstream of the missing region (and so is presumably in a mostly paired region) is thus brought closer to the region downstream of eve, beyond the missing region. This would then bring the transgene closer to the eve locus in 3-D space, likely increasing their frequency of interaction. Such a mechanism would be consistent with our earlier observation of distance-dependent transgene – endogenous eve interactions [24].

Pairwise interactions between transgenic and endogenous eve insulators change both transgene reporter expression and its physical interactions

As would be expected from the known properties of Drosophila insulators, the LR regulatory interactions seen between homie- and nhomie-containing transgenes and the eve enhancers require endogenous homie and nhomie (Fig 2). On the other hand, LR trans pairing by transgenic copies of either homie or nhomie can occur when homie and nhomie are both absent in cis, but present in trans (Fig 2). Our studies also examine how pairing interactions change depending on the number and position of potential pairing partners. When a transgenic insulator has a choice of either of the two endogenous eve insulators as a partner (in the wild-type background), the results are distinctly different than when only one endogenous eve insulator is present in the chromosome (in the ∆N or ∆H background) (Fig 3). In the latter case, some enhancer-promoter interactions are weakened, consistent with the loss of an interaction with one end of the eve locus, while others are strengthened: specifically, those that involve enhancers that are closer to the remaining insulator. This strengthening suggests that the remaining insulator may form a more stable or a more frequently forming complex when the other endogenous insulator is missing. This, in turn, is suggestive of a competition occurring in wild-type between the two endogenous insulators for the transgenic copy.

There are other likely contributing factors for the observed changes in eve enhancer interactions with a transgenic promoter when one endogenous insulator is missing. Assuming that transgenic and endogenous promoters are competing for enhancer activity, then anything that reduces endogenous promoter access is expected to increase transgene reporter expression (in this case, lacZ). We showed here that removing either homie or nhomie from the eve locus reduces eve function at early stages, and this likely occurs through a small decrease in expression in early stripes. This may allow enhancers to access the lacZ promoter more efficiently and increase reporter gene expression. Additionally, Micro-C data showed that the endogenous homie region interacts with the endogenous promoter, as well as with the nhomie region [25], and disruption of these interactions by deleting either homie or nhomie may well reduce the efficiency of endogenous enhancer-promoter interactions, allowing greater expression of the transgenic promoter. Furthermore, the endogenous homie-nhomie interaction loop can no longer form on either the ∆N or the ∆H chromosome. This may reduce endogenous eve promoter access to the more distant enhancers, as they are further away, on average, than they are when the two endogenous insulators are paired. This, in turn, can reduce competition between the eve and transgene promoters, allowing greater reporter gene expression.

Interestingly, beyond the distinct bias in reporter gene activity driven by either the upstream or downstream eve enhancers when the transgene insulator pairs with the remaining endogenous insulator in ∆H and ∆N, we did not see evidence that activation of lacZ depends on distance of the individual enhancers from the tethering point (which is the remaining endogenous insulator). For example in ∆N, stripes 4 and 6 are expressed as strongly as is stripe 5, although the position of the stripe 4 + 6 enhancer is considerably further away from endogenous homie than is the stripe 5 enhancer. Rather, the upstream and downstream enhancers of eve seem to behave as two separate units in activating the transgene reporter. This could be due to the upstream eve promoter region acting as a weak insulator [61]. In fact, an interaction between homie and this region is seen in MicroC data [25], as mentioned above.

Pairwise versus tripartite interactions

As described above, when one of the eve insulators is deleted, there is not only a change in the pattern of regulatory interactions between the lacZ reporter and the eve enhancers, there also appears to be an increase in the level of lacZ expression within the stripes that are preferentially activated (Fig 3). One possible explanation for such increased expression is that insulator pairing interactions are strictly pairwise. In this case, in a wild-type background, there would be a competition between the transgene insulator and the two eve insulators. Since proximity would favor pairing of the eve insulators with each other at the eve locus, the frequency of pairing interactions between the transgene insulator and either one of the eve insulators would be suppressed by competition. That is, assuming endogenous homie and nhomie are paired with each other most of the time, they would need to disassociate before either could pair with the transgene insulator, and this would suppress pairing interactions with the transgene insulator. So, because in the ∆H and ∆N deletions the competing endogenous insulator is removed, this would be expected to increase the frequency of insulator pairing between the transgene and the remaining eve insulator. Consistent with this possibility, the level of lacZ expression from the Z-homie-G and G-nhomie-Z transgenes is noticeably higher in the ∆N deletion, and to a lesser extent in ∆H, than in wild type (Fig 3). However, the MicroC data challenge this explanation. In this case, when one of them is deleted, we would expect the transgene insulator interactions with the eve locus to increase significantly. Contrasting with this prediction, there is no significant change in total transgene – eve locus interactions when one endogenous insulator (nhomie) is deleted (Fig 5F). Therefore, the MicroC results are incompatible with this simple model based on strictly pairwise competition.

One possible way to reconcile the MicroC data with a pairwise-only interaction model is to include homolog interactions in the model. When homologs are paired, there are two copies of endogenous homie present near each other, even when nhomie is deleted. If, instead of local cis interactions (since nhomie and homie no longer pair in cis), homie-homie interactions in trans are present most of the time in most cells when nhomie is deleted (in ∆N), then the competition barrier to transgene interactions with endogenous eve might not be reduced very much. Thus, we might expect the total amount of transgene – endogenous eve interactions to remain about the same when nhomie is deleted, consistent with what is seen (Fig 5F). Arguing against this model is that homolog pairing between enhancers and promoters has been shown to be considerably weaker than local cis pairing, at least in some parts of the genome [59].

Another way to explain the MicroC results is to modify the simple pairwise-only interaction model by including a multi-component complex in the mix. In the simplest form of this model, a tripartite complex can form, involving simultaneous endogenous nhomie-homie interactions with binding by the transgene insulator. In this model, the pre-existence of the local nhomie-homie complex does not present a barrier to transgene interactions with the eve locus, and so no increase in physical interactions is expected when one endogenous insulator is deleted (this situation is illustrated in Fig 6E and 6F).

A hybrid model can also potentially reconcile transgene expression with the MicroC data. In the simplest version of this model, the interaction that initially “captures” the distant transgene is a tripartite complex like that described above. This complex, involving the two endogenous eve insulators and the transgene insulator, holds the transgene in proximity to the eve locus and mediates most of the physical contacts seen by MicroC. This is plausible because a pairwise complex between two “matched” insulators like homie and nhomie likely involves a number of individual pairwise contacts between insulator proteins arrayed along their length. At any given time, some of those contacts may not be engaged, and could therefore present an interaction interface. These unengaged contact points would be free to interact with a 3^rd compatible partner to form a tripartite complex. This configuration could itself be quite stable, particularly if individual insulator-bound proteins with multiple interactions domains can interact simultaneously with proteins bound to more than one other insulator. Furthermore, it would be expected to “isomerize” at some frequency to either of 3 pairwise complexes, only one of which would disengage the attachment of the transgene to the endogenous locus. The remaining two, namely the transgene insulator paired with either nhomie or homie, might be more conducive to lacZ expression driven by the nearby eve enhancers than is the tripartite complex. In this more dynamic, hybrid model, expression of lacZ comes more from the pairwise complexes, while the MicroC signal is a composite of all the alternative complexes, most of which may be tripartite. This would naturally explain both the “extra” enhancement of lacZ expression when one endogenous insulator is removed (particularly in stripes 4, 5, and 6 in the ∆N combinations), and the lack of a corresponding increase in MicroC signal in ∆N over wild type. This type of model is also compatible with the results of insulator competition experiments [62–64], which were analyzed based on reporter gene expression only, before chromosome conformation capture methods were developed. These results were consistent with pairwise complexes being primarily responsible for the observed expression patterns. Why might pairwise complexes give more reporter expression? The tripartite complex could be more constrained, while a pairwise complex might be more flexible, allowing enhancer-promoter interactions to form more efficiently or to be more stable.

While the MicroC experiments suggest that insulator pairing interactions may not be exclusively pairwise, further studies are required to disentangle the properties of tripartite and pairwise interactions. Since insulator pairing depends on protein-protein interactions, the combinations that are possible may be different when insulator interactions are pairwise versus when they are tripartite. Given the increase in stripe expression in stage 5 embryos that is observed when one endogenous insulator is removed, it seems possible that pairwise complexes between insulators are formed more rapidly than tripartite complexes. As the cell cycle in most of the cells in late-stage embryos is lengthy compared to those in pre-cellular blastoderm embryos, there would be much more time to establish tripartite complexes. The situation is also different in stage 5 embryos in that the contacts between the transgene insulator and eve insulators have to be established de novo [27]. It would certainly be interesting to follow the dynamics of pairwise versus tripartite pairing interactions in early embryos using live imaging.

Creation of nhomie and homie deletions, and breakpoint analysis of chromosomal deficiencies

Creation of the ∆N line has been described previously [25]. In short, we inserted two attP sites flanking endogenous nhomie. To do this, we used a donor plasmid, P-attPx2-mw, carrying two 102 bp attP sequences flanking a modified mini-white (mw) gene. The following modification was made to mw: the Wari insulator [65] was deleted from the standard mw, and Glass binding sites were added to boost the eye color. The endogenous nhomie region was replaced with this attP-mw-attP using CRISPR. This chromosomal modification resulted in one attP site being inserted in the intron of CG12134, and the other being inserted between the eve late-stripe enhancer and the 3 + 7 stripe enhancer (Fig 1C). This process also deleted 2.2kb of endogenous sequence, including nhomie and the eve late-stripe enhancer. After identifying a successful insertion (NattPmw), mw was replaced by the same 2.2kb sequence, but with 600 bp of phage λ DNA in place of 600 bp nhomie, using recombinase-mediated cassette exchange (RMCE) [66].

Several attempts to insert attP sites flanking homie by CRISPR [30,31] using different gRNA sets failed. While we do not fully understand why our multiple trials of this strategy did not work, it is possible that removing homie is either dominantly lethal, or that it reduces viability such that recovery of the resulting chromosome is difficult, perhaps due to the function of homie in preventing repression of the adjacent, essential gene, TER94. However, we have a homed transgenic line that inserted at +9111 bp relative to the eve transcription start site [24]. Using this line, we successfully mobilized the transgene, creating a small deletion (2R:9987986–2R:9989354 fusion, the two ends share the sequence AAA) (Fig 1B, sequence is shown in S1 Fig). The deletion includes the eve 3’ PRE, homie, and the first exon of TER94 (Fig 1B). We described this line as ∆H. ∆H is homozygous lethal. The same strategy that was used to make the nhomie deletion on a wild-type chromosome was used to modify nhomie on the ∆H chromosome, creating the ∆N∆H deficiency chromosome (Fig 1C). As expected, ∆N∆H is homozygous lethal. Lines carrying either the ∆H or ∆N∆H chromosome over a balancer were self-crossed to make ∆H or ∆N∆H, respectively, (“∆∆” in some figure labels) homozygous embryos.

For Df(2R)eve [33,41], information in Flybase [39] suggested that its breakpoints are in Mef-2 and TER94, so we tested potential PCR primers to find ones that amplified junction fragments. These PCR products were then sequenced to identify the specific breakpoints (S1 Fig). The PCR primers used to amplify the breakpoint junction fragments are: for the ∆H deletion, CAGTCGAGCCTCCGTAAGGG and CCTCCAGCAAAGGATGACTTG, and for Df(2R)eve, TTTCAACCGCACACAATCC and CATTCATTCCAAATCACGCAC.

Creation of attP sites at –142kb and insertion of reporter transgenes

Two attP sites were inserted near the original -142kb attP site [17,24] on the wild-type, ∆N, ∆H, and ∆N∆H chromosomes, using the same CRISPR strategy described above. Then, mini-white was replaced with each of the reporter transgenes using RMCE. Dual reporter transgenes were used (Fig 1D) [17], carrying either 400 bp homie (Z-homie-G), 600 bp nhomie (G-nhomie-Z), or 500 bp of lambda DNA (Z-lambda-G). For each of these, the orientation in which the lacZ reporter is more strongly activated due to orientation-specific LR pairing was used in this study.

Analysis of transgenic lines

In situ hybridization was performed based on previously published methods [67], except that RNA was visualized using a histochemical reaction, as described previously [11]. Briefly, approximately 60μl embryos per sample (stage 4–15 embryos were present in each sample) were subjected to the process. DIG-labeled antisense RNA probes against lacZ, gfp, eve, or en were visualized using alkaline phosphatase-conjugated anti-DIG antibody (Roche), using CBIP and NBT as substrates (Roche). Each set of experiments was carried out with the positive control and experimental samples in parallel to minimize experimental variation. Once color became visible in the positive control, the reactions of all the samples were stopped simultaneously. Each experiment was performed at least twice, with independent in situ hybridization procedures. Representative expression is shown in the figures.

MicroC library construction

Details of the MicroC procedure were described previously [25]. Micro-C libraries were prepared as previously described [68] 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 [69] with BWA-MEM [70] using parameters -S -P -5 -M. Briefly, the custom genomes are created by inserting the transgenic sequence into the –142kb integration site, as predicted from perfect integration. The resultant BAM files were parsed, sorted, de-duplicated, filtered, and split with Pairtools (https://github.com/mirnylab/pairtools). Pairs were removed 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 100 bp contact matrices using Cooler [71]. Finally, balancing and Mcool file generation was performed with Cooler’s Zoomify tool.

Virtual 4C profiles were extracted from individual replicates using FAN-C [72] at 400 bp 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 800 bp window in the gene body of either lacZ or gfp. The raw and processed sequencing data are available under NCBI GEO accession number GSE328676.

Quantification of total contacts between the lacZ reporter and the eve locus in wild type and ∆N

For G-nhomie-Z and Z-homie-G, the locations of the lacZ reporter and the eve locus in wild type and ∆N were identified. Then the interaction region between lacZ and the eve locus was visualized, in order to manually segment bounding boxes of interaction between them. After matrix balancing and masking of null values, the mean of the contact frequency within this bounding box was taken for each dataset. This became the average contact frequency between the transgene lacZ and eve.

Analysis of embryonic cuticle defects

To identify segmentation defects in developing embryos, embryos were collected and analyzed as described previously [25], with minor modifications. Briefly, embryos were collected for 2.5-3 hr, allowed to develop for 20–21 hr at 25°C, then dechorionated and mounted in a 1:1 mixture of Hoyer’s medium and lactic acid. Mounted embryos were left at room temperature until they cleared, and the patterns of ventral abdominal denticles were examined and tallied as follows. Loss of at least one-fifth of a denticle band in A1-A8 was counted as ‘missing’. Fused denticle bands, which rarely occurred, were also counted as a ‘missing’ band. Minor defects such as those within individual denticle rows were not counted.

Wild-type (WT), ∆N, Sco/SM6a,Cy, ∆H/SM6a,Cy, and ∆N∆H/SM6a,Cy lines were self-crossed for S4 Fig. For each cross, 3–4 independent cuticle preparations were analyzed. To sensitize for functional differences of these chromosomes, each of the lines was crossed with each of the eve heterozygous mutant lines Df(2R)BSC158/CyO (Figs 7, S5A, and S6A), Df(2R)eve/CyO (S5B and S6B Figs), and eve^R13/CyO (S5C and S6C Figs). The two directions of each cross were analyzed separately. For each cross, 5–8 independent cuticle preparations were analyzed.

In the sensitized condition, segmentation defects were quantified in two ways: the percentage of embryos with segmentation defects (% affected) and the average number of ventral abdominal denticle bands deleted per non-wild-type embryo (severity score). The data from each set of cuticle preparations are presented as box-and-whiskers plots (Figs 7 and S5), and the pair-wise significance of differences (p-values) between lines was calculated using the t-test function (2-tailed, unequal variances) in Excel (Microsoft). For Fig 7, two sets of data were combined, one directly comparing WT, ∆N, and ∆H (shown in S5A Fig) and the other directly comparing WT, ∆H, and ∆N∆H. The raw data from this 2nd comparison were used in Fig 7, along with scaled data for ∆N from the 1st comparison. For this purpose, individual data points from the 1st comparison were scaled based on their values relative to WT and ∆H (they were all between the average values for WT and ∆H in the 1st comparison, and were scaled to have the same fractional distance from the average WT and ∆H values in the 2nd comparison as they did in the first comparison). The statistical analysis for Fig 7 used these scaled values for ∆N in combination with the raw data for the other lines. For a different visual comparison, stacked graphs are shown in S6 Fig. For this, the percentages of embryos (average and standard deviation) with 4 different severities of phenotype are shown: wild-type, mild (1 band missing), moderate (2 bands missing), and severe (3–4 bands missing). The complete set of statistical calculations is shown in S1 File.

Analysis of H3K27me3 levels

ChIP-qPCR was performed based on previously published methods [73]. Briefly, stage 4–17 embryos (approximately 200ul) from Sco/Cy and ∆N∆H/Cy were cross-linked in 2% formaldehyde for 10 min. After sonication, 50μg of isolated chromatin was used to immunoprecipitate with anti-H3K27me3 (Millipore), with rabbit IgG (Jackson ImmunoResearch) used as the negative control. Precipitated chromatin samples were collected using ProteinG magnetic beads. After reversing cross-links, purified samples were dissolved in 50 μl TE, and 1 μl was used for each qPCR reaction. Triplicate samples were analyzed by real-time PCR (Life Technologies, StepOne Plus), using SYBR Green Master Mix with ROX dye (Roche Applied Science). Data were analyzed with StepOne software (Life Technologies), using the standard curve method. The data are presented as average % inputs from 3 independent ChIP analyses. Standard deviations were calculated using Excel software (Microsoft). The pair-wise significance of differences (p-values) between lines was calculated using the t-test function (1-tailed, unequal variances) in Excel (Microsoft).

Primers used for qPCR were as follows. N (upstream of nhomie): CGGAGAATCCGGCATTGTTA and GCTTGCGTGATTTCTTCTCC, eve (eve coding region): TCCAGTCCGGATAACTCCTTG AAC and TGTAGAACTCCTTCTCCAAGCGAC, H (downstream of homie): AAGGGCCACATCGCAGACATACTA and GTCGCGGTAAATGTCTTTGTCTCG, en (engrailed coding region): GAGAACCAGGCCAGCATATT and CTAAACTCCAGCAGATCCACTC.