Pericentromeric heterochromatin is hierarchically organized and spatially contacts H3K9me2 islands in euchromatin

Membraneless pericentromeric heterochromatin (PCH) domains play vital roles in chromosome dynamics and genome stability. However, our current understanding of 3D genome organization does not include PCH domains because of technical challenges associated with repetitive sequences enriched in PCH genomic regions. We investigated the 3D architecture of Drosophila melanogaster PCH domains and their spatial associations with the euchromatic genome by developing a novel analysis method that incorporates genome-wide Hi-C reads originating from PCH DNA. Combined with cytogenetic analysis, we reveal a hierarchical organization of the PCH domains into distinct “territories.” Strikingly, H3K9me2-enriched regions embedded in the euchromatic genome show prevalent 3D interactions with the PCH domain. These spatial contacts require H3K9me2 enrichment, are likely mediated by liquid-liquid phase separation, and may influence organismal fitness. Our findings have important implications for how PCH architecture influences the function and evolution of both repetitive heterochromatin and the gene-rich euchromatin.


Introduction
Nuclear architecture and dynamics regulate many important genome functions (reviewed 2 in [1][2][3][4]). The development of Hi-C, which combines chromosome conformation capture 3 (3C) [5] with genome-wide sequencing [6], has led to major breakthroughs in our 4 understanding of global nuclear architecture (reviewed in [7]). However, analyses of Hi-C 5 results have focused on single copy sequences in euchromatic regions (e.g. [6,8-10]), and 6 virtually all have excluded the large Peri-Centromeric Heterochromatin (PCH) portion of 7 genomes due to its enrichment for large blocks of repetitive DNAs [11,12]. Despite being 8 gene-poor, the PCH plays vital roles in chromosome dynamics [13,14] and genome integrity 9 [15-17]. 10 A defining characteristic of heterochromatin is its enrichment for 'repressive' 11 epigenetic features, such as Histone H3 lysine 9 di-and trimethylation (H3K9me2/3) and 12 its reader protein, Heterochromatin Protein 1a (HP1a) [18,19]. Interestingly, PCH 13 DNA/chromatin from different chromosomes coalesce into one or a few membraneless 14 PCH 'domains' (or chromocenters) in the 3D cell nucleus [20,21]. Recent studies have 15 shown that specific biophysical properties of HP1a and liquid-liquid phase separation 16 (LLPS) may mediate PCH domains formation [22,23]. This widely observed spatial 17 organization of PCH domains could significantly influence transcription and other genome 18 functions [24], such as silencing of euchromatic genes transposed near or in PCH genomic 19 regions [25][26][27]. Furthermore, PCH-PCH interactions have recently been proposed to drive 20 the global genome architecture [28]. 21 In addition to PCH and peritelomeric heterochromatin, regions of H3K9me2/3 22 enrichment are also present in the euchromatic genome [29][30][31]. Previous studies of a 23 Results 1 Hierarchical organizations of PCH domains 2 To decipher the 3D organization of PCH domains, we overcame technical limitations 3 inherent to analyzing repeated DNA sequences and developed a new method that includes 4 repetitive DNAs highly represented in PCH regions to analyze Hi-C data ( Figure 1A and 5 Figure S1). The Release 6 D. melanogaster genome is the most complete genome among all 6 multicellular eukaryotes, and includes a nearly full assembly of the non-satellite PCH DNA 7 [36,37]. The genomic boundaries between PCH and euchromatin have also been 8 epigenetically identified [31]. The annotated assembly allowed us to include three types 9 Hi-C reads that originate from PCH DNA ( Figure 1A): 1) unique single-copy sequences 10 within PCH (e.g. protein coding genes, "unique"), 2) simple repeats known to be enriched in 11 PCH ("repeat", Table S1), and 3) sequences that map to multiple sites in the PCH (i.e. non 12 single-locus mapping, "multi"). We used these sequence classifications to assess contact 13 frequencies between PCH regions, and between PCH and H3K9me2/3-enriched regions in 14 the euchromatic genome ( Figure 1B and below), using published Hi-C data from 16-18hr 15 D. melanogaster embryos [38]. 16 Analyses of the formation and function of 3D PCH domains generally assume they 17 are homogeneous, despite the fact that they contain coalesced PCH regions from different 18 chromosomes that have high sequence heterogeneity. To investigate potential 19 substructures within the PCH domains, we focused on Hi-C read pairs in which both ends 20 mapped uniquely to PCH genomic regions ("unique" PCH reads, Figure 1A) because of 21 their known chromosomal locations. In addition to PCH regions on the 2 nd , 3 rd , and X 22 chromosomes, the entire 4 th and Y chromosomes were included in the analysis because the 23 entirety of these two chromosomes are enriched with heterochromatic marks [31,39]. We 1 estimated the number of Hi-C read pairs coming from any two of the 100kb PCH regions. 2 Using a sequential exclusion approach (see Methods), we identified three types of 3 prevalent spatial interactions among PCH regions: within an arm (intra-arm), between 4 arms of the same chromosome (inter-arm), and between arms of different chromosomes 5 (inter-chromosome). The most frequent interactions were among PCH windows on the 6 same chromosomal arm, which accounts for 98.08% (replicate 1, Figure 2A) and 97.15% 7 (replicate 2, Figure S2; and see Figure S3) of parsed Hi-C read pairs (see Table S2 for the 8 number of read pairs supporting each interaction). Interactions among windows within 9 PCH arms are stronger than PCH-euchromatin interactions on the same arm ( Figure S4 10 and S5), suggesting that PCH arms (e.g. 2L PCH) are organized into distinct "territories." 11 Exclusion of intra-arm interactions revealed strong spatial interactions between PCH 12 regions flanking the centromeres (inter-arm, i.e. 2L-2R, 3L-3R), which accounted for 13 34.72% and 35.88% (replicate 1 and 2) of the remaining read pairs (0.67% and 1.02% of 14 total unique PCH-PCH read pairs respectively), and specific inter-chromosome interactions, 15 mainly 3L -4 (9.68% and 9.49% of non-intra-arm read pairs). To quantitatively investigate 16 whether these interactions are exceptional, we compared the observed percentage of read 17 pairs against expectations that are based on either theoretical mappability [40] or 18 empirically observed number of reads mapped to PCH on each chromosome arm (see 19 Methods, Figure 2B) We also performed permutation tests for the latter to evaluate the 20 statistical significance. Contact frequencies between 2L-2R, 3L-3R, and 3L-4 are indeed 21 significantly more than expected (compared to both expectations, permutation p-value < 22 0.0001). Finally, we excluded all intra-chromosome interactions to specifically study 23 8 contact frequencies between PCH regions on different chromosomes ( Figure 2B). The 1 relative frequencies of most inter-chromosome associations did not exceed expectations 2 (e.g. 2L-3L), suggesting random contacts across cell populations. However, frequencies of 3 3D contacts between 3 rd chromosome PCH and the 4 th chromosome (3L-4, 3R-4) were 4 exceptionally high (compared to both expectations, permutation p-value < 0.0001). Contact 5 frequencies between 2L-4, 2R-4, and 3R-Y were also significantly more than expected. . This allowed us to ask if chromosome-specific probes that label 12 simple repeats from PCH regions that displayed exceptional Hi-C spatial interactions (e.g.

13
3R-4) colocalized more often than probes from the same chromosomes with lower 14 frequency interactions (2R-3R and 2R-4). We measured the "relative distance," defined as 15 the distance between FISH signal centroids divided by the nuclear radius ( Figure 2C), to 16 account for variable cell size at late embryonic stages. The relative distance between 3R 17 (dodeca)-4 th chromosome (AATAT) is significantly shorter than 2R (AACAC)-3R or 2R-4 18 (Mann-Whitney test, p = 0.0001 (3R-4 vs 2R-3R) and <10 -6 (3R-4 vs 2R-4), Figure 2D). For 19 all three pairs of interactions, the distribution of relative distance is bimodal (Figure 2E), 20 with a sharp peak near zero. We defined two foci as 'overlapping' when their distances 21 were shorter than this natural threshold (denoted by arrow in Figure 2E). Consistent with 22 the Hi-C results, the proportion of nuclei with overlapping foci was higher for 3R-4 than for 23 9 2R-3R or 2R-4 (Fisher's Exact test, p = 0.22 and 0.0006 respectively, Figure 2E). Overall, 1 both Hi-C and FISH analyses demonstrate a hierarchical 3D organization of PCH domains. types of PCH-derived sequences were included in the Hi-C analysis: 1) reads mapped to 2 single-copy sequence in the epigenetically defined PCH regions ("unique" reads, 2.4% of 3 filtered Hi-C reads (see Figure S1)), 2) reads mapped to known heterochromatic simple 4 repeats ("repeat" reads, 6.44%), or 3) reads mapped to non-unique sequences (dark blue) 5 that are present within epigenetically defined PCH regions ("multi" reads, 3.0%). (B) 6 Methods for assessing if a H3K9me2-enriched euchromatic region displays exceptional 3D 7 contacts with PCH. The observed percentage of euchromatin-PCH read pairs for an 8 H3K9me2 enriched euchromatic region is compared to a null distribution generated using 9 randomly selected, non-H3K9me2 enriched euchromatic regions to estimate p-value.   HP1a liquid droplets both in vitro and in vivo [22,23], led us to predict that small regions 4 enriched for H3K9me2/3 and HP1a in the euchromatic genome could also spatially 5 associate with the main PCH domains. To test this hypothesis, we identified euchromatin-6 PCH Hi-C read pairs, which contain sequences from single-copy, euchromatic regions 7 paired with any PCH sequence (i.e. all three categories of PCH sequences, Figure 1A). We 8 then estimated, among Hi-C read pairs whose one end mapped uniquely to a specific 9 euchromatic region, the percentage of euchromatin-PCH read pairs ( Figure 1B). We

15
We identified by ChIP-seq 496 H3K9me2-enriched regions (defined as "H3K9me2 16 islands," 290bp -21.63Kb, with an average size of 3.84 kb) in the euchromatic genome 17 (>0.5 Mb distal from the epigenetically defined euchromatin-PCH boundaries) in embryos 18 of the same genotype and stage as the Hi-C data (see Methods). Of these H3K9me2 islands, 19 13.91% (n = 69) and 8.67% (n = 43) displayed significant spatial associations with PCH in 20 either or both Hi-C replicates, respectively ( Figure 3A). These numbers are significantly 21 higher than expected (i.e. 5% of the H3K9me2 islands would be significant under null 22 expectation; binomial test, p = 0.00059 (both) and 3.04x10 -14 (either)). Thus, we conclude 23 13 that H3K9me2 islands are more likely to spatially interact with PCH than euchromatic 1 regions without H3K9me2 enrichment in the euchromatic genome. For subsequent 2 analyses, we focused on H3K9me2 islands that significantly interacted with PCH in both Hi-3 C replicates (hereafter referred to as "EU-PCH" associations). 4 We found that H3K9me2 islands with PCH interactions have shorter linear distance 5 to PCH regions along the chromosome compared to H3K9me2 islands that lacked PCH 6 interactions (Mann-Whitney U test, p < 10 -4 , Figure S7), suggesting that proximity to PCH 7 on a linear chromosome is a strong defining feature for the tendency to spatially interact 8 with PCH. For each H3K9me2 island, we calculated the percentage of unique PCH reads 9 from each chromosome arm (e.g. percentage of EU-2L PCH read pairs). For PCH region on a 10 particular arm, H3K9me2 islands on the very same arm always have the highest such 11 percentage (e.g. 2L euchromatic regions have the highest percentage of EU-2L PCH read 12 pairs), followed by those on the other arm of the same chromosome ( Figure 3B and Figure   13 S8). This echoes the observed strong tendency of "intra-arm" PCH-PCH interactions, 14 followed by "inter-arm" PCH-PCH interactions (Figure 2A and 2B). 15 Interestingly, H3K9me2 islands that show spatial interactions with PCH have higher 16 fractions of coding sequences when compared to H3K9me2 islands without PCH 17 interactions (Mann-Whitney U test, p = 0.0015, median: 70.1% (with) and 30.4% 18 (without)). In addition, these regions are more likely located within active Topologically 19 Associated Domains (TADs) identified at the same embryonic stage [8] than H3K9me2 20 islands without PCH interactions (Fisher's Exact Test, p = 0.0078, Table S3). Using 21 previously reported segmentations of the D. melanogaster genome into combinatorial 22 chromatin states [42,43], we also found that significant EU-PCH contacts are more likely to 23 involve euchromatic regions in active states: Red or Yellow chromatin (Fisher's Exact test, p 1 = 0.021), or modEncode State 1-4 (p < 10 -4 (S2) and =0.011 (BG3), Table S3). These 2 regions are also depleted for chromatin states that lack obvious enrichment for histone 3 modifications and/or protein binding: "null" TADS (Fisher's Exact test, p = 0.03), black 4 chromatin (p < 10 -3 ), and modEncode State 9 (p = 0.008 (S2), Table S3). It is currently 5 unclear why PCH associations would be enhanced for H3K9me2 islands containing coding 6 genes or active chromatin marks. It is worth noting that PCH associations were not 7 correlated with the following properties of H3K9me2 islands: autosome or sex  (Table S4). 12 To validate the EU-PCH 3D interactions identified by Hi-C analysis, we performed 13 FISH using Oligopaint probes [44-46] targeting 30.5-42.9kb euchromatic regions (Table   14 S5) and probes that broadly mark PCH (AAGAG, a satellite enriched in PCH regions of all 15 chromosomes, [47,48]). We focused on three 2R windows covering H3K9me2 islands that 16 spatially interact with PCH (EU1-3). Because we observed that the linear distance to PCH 17 genomic regions is a strong predictor for whether a H3K9me2 island interacts with PCH 18 (see above), for each of these regions, we chose a matching "control" window that is at a H3K9me2 enrichment level, and Figure 3D and Figure S10 for representative cell images). 22 Consistently, we observed that H3K9me2 islands displaying PCH interactions in the Hi-C 23 analysis are closer to PCH in 3D space than linearly equidistant euchromatic regions that 1 lack H3K9me2 enrichment (Mann-Whitney U test, p < 10 -6 (EU1 vs c.EU1), < 10 -13 (EU2 vs 2 c.EU2), and 0.0025 (EU3 vs c.EU3), Figure 3E), confirming the observations made by Hi-C 3 analysis. This difference is also reflected in the higher proportion of cells in which the two 4 foci overlap compared to the control regions ( Figure 3F). It is worth noting that the 5 comparatively lower frequency of overlapping foci for EU2 and EU3, when compared to 6 EU1, could result from the fact that these two regions are much farther from the PCH, and 7 thus less likely to spatially interact with PCH than EU1 (see above). This could potentially   spread. Among these TEs, 13.21% (n = 14) and 7.55% (n = 8) displayed significant spatial 17 interactions with PCH (p < 0.05) in either or both Hi-C replicates respectively (see Figure   18 S13 for their genomic distribution), which is significantly more than expected (binomial 19 test, p = 8.38x10 -4 (either) and 0.26 (both)). As a contrast, only 1.75% of TEs without 20 H3K9me2 enrichment (n = 1) display PCH interactions. We focused on analyzing the 14 TEs 21 showing significant PCH-contact in either replicate, while analyses restricted to eight TEs 22 significant for both replicates was qualitatively similar (Table S6). Similar to non-TE 23 induced H3K9me2 islands, TEs spatially interacting with PCH are closer to PCH genomic 1 regions on the linear chromosome than those that do not interact with PCH (Mann-Whitney 2 U test, p = 0.037, Figure S14). PCH-interacting TEs include those from roo, pogo, 17.6, 3 mdg3, FB, and S families. However, they were not significantly enriched for any specific TE 4 family (Fisher's Exact Test for individual TE family, p > 0.26), class, type, or sex-5 chromosome linkage (Table S6). 6 The polymorphic nature of TEs offers a rare opportunity to compare the 3D 7 conformations of homologous sequences with and without TE-induced H3K9me2/3 8 enrichment. To validate the Hi-C results, we performed FISH analysis focusing on two TEs 9 that are present in the Hi-C strain (ORw1118) but absent in another wildtype strain. These 10 two TEs also induced ORw1118-specific enrichment of H3K9me2 ( Figure S12) and 11 spatially interact with PCH (TE1-2, Figure 3C). As controls, we included two additional 12 ORw1118-specific TEs that did not interact with PCH and do not have H3K9me2 13 enrichment (c.TE1-2, Figure 3C and Figure S12). Our FISH used Oligopaint probes that 14 target unique regions flanking the selected euchromatic TE insertions (Table S5) and 15 probes that broadly mark PCH (see Figure S10 for representative cell images). For TE1 and 16 TE2, the relative 3D distance to PCH signals is shorter in ORw1118 than in wildtype (Mann-17 Whitney U test, p = 0.0004 (TE1) and p = 0.015 (TE2), Figure 4A). Interestingly, the 18 distribution of relative distance between TE1/TE2 and PCH is bimodal for ORw1118 nuclei 19 but unimodal for wildtype, which lacks the peaks around zero, or nuclei with overlapping 20 foci ( Figure 4B). Indeed, there are more nuclei with overlapping foci in ORw1118 than in 21 the wildtype (Fisher's Exact Test, p = 0.0003 (TE1) and 0.070 (TE2)). Importantly, these 22 between-strain differences were not observed for control TEs that lacked PCH interactions 23 (Mann-Whitney U test, p = 0.55 (c.TE1) and 0.91 (c.TE2), Fisher's Exact test, p = 0.49 (c.TE1) 1 and 1 (c.TE2), Figure 4A and 4B). This comparison of homologous regions with and 2 without euchromatic TEs suggests that H3K9me2 enrichment is required for spatial 3 contacts between euchromatic regions and PCH domains.  between PCH and H3K9me2 islands that displayed significant PCH interactions (see above) 8 in permeabilized embryos with and without 1,6-hexanediol treatment (see Methods). We 9 focused on TE1 because it is ORw1118-specific and leads to strain-specific H3K9me2 10 enrichment. This allows comparisons between genotypes with and without TEs to 11 investigate whether the sensitivity to 1,6-hexandiol treatment is H3K9me2-enrichment 12 dependent (see Methods, Figure S16). We observed significantly longer TE1-PCH relative 13 3D distance (orange in Figure 4C, Mann-Whitney test, p < 10 -4 ) and fewer nuclei with 14 overlapping foci (orange in Figure 4D, Fisher's Exact test, p = 0.02) in ORw1118 embryos 15 treated with 1,6-hexanediol compared to untreated controls. In contrast, no such difference 16 was observed in wildtype embryos, which do not have the TE insertion and thus no 17 frequent TE1-PCH 3D contacts (green in Figure 4C and 4D, Mann-Whitney test, p = 0.74, 18 and Fisher's Exact test, p = 1). Importantly, the significant difference in TE1-PCH 3D 19 distance between genotypes with and without TE insertion is only observed for embryos 20 without 1,6-hexanediol treatments (Mann-Whitney test, p = 0.0037, Fisher's Exact test, p = 21 0.057), but not for those with the treatment (Mann-Whitney test, p = 0.77 and Fisher's Exact 22 test, p = 0.55, Figure 4C and 4D). The sensitivity of TE-PCH 3D contacts to 1,6-hexanediol 23 is consistent with the spatial interactions between H3K9me2 islands and PCH domains 1 being mediated by liquid fusions, an emergent property of liquid-liquid phase separation.  (Table S6) 19 or local recombination rate (Mann-Whitney U test, p = 0.40). On the other hand, we did 20 observe that TEs with PCH interactions tend to be closer to genes than TEs without such   (Figure 3A and 4A). Importantly, quantitative FISH analysis 4 provides cytogenetic support for the Hi-C results. The bimodal distributions of PCH-PCH or 5 EU-PCH distances in nuclei (Figure 2F, 3G, 3E) also demonstrate that these 3D contacts  It is important to note that TEs comprise an appreciable fraction of euchromatic 7 genomes in virtually all eukaryotes [77]. For instance, more than 50% of assembled human    The entirety of 4 th and Y chromosomes are enriched with heterochromatic marks [31,39] 16 and are considered to be entirely heterochromatic.

18
Generation and analysis of H3K9me2 ChIP-seq data 19 We performed ChIP-seq using antibody targeting H3K9me2 (Abcam 1220) on 16-18hr paired-end reads. Each sample has two ChIP replicates (biological replicates) with 2 matching inputs. 3 Raw reads were processed with trim_galore [88] to remove adaptors, low quality 4 bases, and single-end reads. Processed reads were mapped to release 6 D. melanogaster 5 genome with bwa mem with default parameters. Reads with mapping quality lower than 6 30 were removed using samtools [89]. To have enough noise for the IDR analysis (see 7 below), we ran Macs2 [90] using broad-peak and pair-end mode, and a liberal p-value 8 threshold (0.5). This was followed by performing Irreproducible Rate (IDR) analysis [91] to 9 identify H3K9me2 enriched regions that are consistent between replicates. We defined 10 H3K9me2-enriched regions as those with low IDR (IDR < 0.01). IDR plots for replicates for 11 three ChIP-seq samples can be found in Figure S19-21. 12 13

Identification and analysis of TE insertions
14 TEs in wildtype strains: All potential TE insertions in RAL315 and RAL360 strains were 15 previously identified using TIDAL [83]. We used the recommended coverage ratio (read 16 number supporting TE presence/TE absence, coverage ratio at least three) to identify TEs 17 with high confidence in these two wildtype strains. TEs in wildtype strains are used to 18 identify ORw1118-specific TEs (see below). 19 20 Identification of TEs in ORw1118: To identify TEs in the ORw1118 strain, we performed 21 genomic sequencing. Genomic DNA was prepared from 100 ORw1118 adult female flies for 22 each biological replicate (three biological replicates in total) with Gentra Puregene Cell kit 23 (Qiagen cat#158388) according to the manufacturer's instructions. Whole genome 1 sequencing was done with overlapping 165bp pair-end Illumina sequencing on 230-240bp 2 size genomic fragments. 3 We combined all three replicates of ORw1118 genomic sequencing to call TEs and 4 quality filtered reads with Trim_galore. We identified TEs in ORw1118 also using TIDAL 5 [83], which calls TEs with split-read methods and requires input reads to have the same 6 length. Accordingly, we used two approaches to generate single-end reads from the original TEs in ORw1118 were identified with these criteria. 16 To identify TE-induced local enrichment of H3K9me2, we used methods described in 17 [35], which leverages between strain differences to identify TE-induced H3K9me2 18 enrichment regions with any shape, which oftentimes do not resemble peaks (e.g. Figure   19 S12). This approach is more sensitive than other custom pipelines, which look for 20 enrichment with "peak" shape, followed by ad hoc merging of sharp peaks to generate 21 "broad peak" calls (reviewed in [54,55]). We compared the enrichment of H3K9me2 in 22 euchromatic TE neighborhoods in ORw1118 against wildtypes strains to estimate (1) the 23 extent of TE-induced H3K9me2 enrichment (in kb) and (2) % of increase of H3K9me2 1 enrichment. We identified 106 ORw1118 TEs leading to at least 1kb spread of H3K9me2, 2 with only 13 of them overlap with H3K9me2 enriched regions identified by Macs2. 3 We used the same approach as in [35] to estimate the population frequencies of  which genomic compartment a TE-mapping read is from. Accordingly, we filtered reads 19 that mapped to canonical TEs using bwa [93] and samtools [89]. Filtered reads were then 20 mapped to release 6 D. melanogaster reference genome using bwa with default parameters.

21
Three types of reads are defined as from heterochromatin. (1) Reads that uniquely mapped 22 (mapping quality at least 30) within epigenetically defined PCH regions. (2) Reads mapped 23 to known heterochromatic repeats (Table S1). (3) Reads that mapped to epigenetically 1 defined PCH but have mapping quality equals zero, which bwa assigns to multiple-mapped 2 reads. Mapping locations of unique PCH reads are recorded and used for both PCH-PCH and 3 PCH-EU analysis. Other two types of PCH reads were only used for PCH-EU analysis and 4 their mapping locations, which are multiple in the genome, are not used. 5 All the reads parsing were done with samtools. Figure S1 shows the flow chart for the 6 filtering, mapping, and identification of PCH Hi-C reads, and the number of reads at each 7 step. Genome-wide contact maps for both PCH and euchromatic regions (Figure S4   Spatial interaction between PCH regions: Hi-C read pairs whose both ends mapped 11 uniquely to epigenetically defined PCH were included in the analysis. Read pairs whose 12 mapping locations are within 10kb to each other were removed, as our analysis focuses on mapping uniquely to the PCH on each chromosome arm, ignoring read pair information. 3 Because the Hi-C data were generated using unsexed embryos, we assumed equal sex ratio 4 when estimating expectations. To assess whether the observed percentage is more than the 5 empirical expectation, we randomly permuted 10,000 times read pair labels, generated an 6 empirical distribution of the percentage, and calculated one-sided p-values.

Spatial interaction between euchromatic regions and heterochromatin:
We used 9 samtools to parse out read pairs whose one end mapped uniquely (with mapping quality at 10 least 30) within the focused euchromatin regions, and estimated the percentage of PCH 11 reads at the other end. All three categories of heterochromatic reads were included.

12
Regions with less than 1,000 Hi-C read pairs were excluded from the analysis. We found   Figure S24). Also, smaller euchromatic regions have fewer Hi-C read pairs included in the 3 analysis, which translates into smaller sample size and thus larger variance of the 4 estimated percentage (Figure S25), leading to the estimates more likely to hit the 5 boundary condition (i.e. no euchromatin-PCH read pairs, Figure S25, red circles).     Bridge oligos and fluorescence labeled secondary Oligopaint probe were ordered from IDT. 8 9 Embryo collections, treatments, and fixations 10 Embryo collections: Flies laid eggs on fresh apple juice plate for 1hr (pre-lay), followed by 11 2hr egg-laying on new apple juice plates. Collected embryos were incubated at 25°C for 12 16hr to harvest 16-18hr embryos, which were then fixed immediately.  Permeabilized embryos were either fixed immediately or incubated in 10% 1,6-hexanediol 20 (dissolved in PBS) for 4 min, followed by a quick wash with PBS and fixed immediately.  AATAT probe, embryos were incubated at 37°C for three hours, then 25°C overnight. 20 Embryos were washed with hybridization buffer twice at 37/25°C, followed by sequential 21 transition into PBT, two PBT washes at room temperature, DAPI staining, two PBS washes, 1 resuspended in Prolong Gold Antifade (Life Technologies), and mounted on slides. 2 We used AATAT to mark 4 th chromosome heterochromatin. Because this repeat is also 3 abundant on the Y [41], embryos were also stained with Y-specific repeat, AATAGAC and 4 female embryos were analyzed.  Imaging and data analysis 11 Images of embryos were collected on Zeiss LSM710 confocal florescence microscope, using 12 a 1.4NA 63X oil objective (Zeiss), and analyzed in Fiji [101]. Distances between foci were 13 measured by Fiji linetool, and divided by the radius of the nucleus to get relative distance.
14 In cases where the nuclei are not perfectly round, we used radius on the longest axis. There 15 are usually one to two AAGAG foci in a nucleus and the distance was measured between 16 Oligopaint focus and the nearest AAGAG focus. At least 70 nuclei were counted for each 17 treatment/genotype.

19
Data availability 20 Genome sequence (raw genome data and called transposable elements) and ChIP-seq data 21 (raw data and processed tracks) have been deposited to GEO (GSE125307 -private token 22 for reviewers: gzavmcqizxmjxil; GSE125031 -private token for reviewers: 1 mrmhquucztcdlwv).