Simultaneous epigenetic perturbation and genome imaging reveal distinct roles of H3K9me3 in chromatin architecture and transcription

Introduction Despite the long-observed correlation between H3K9me3, chromatin architecture, and transcriptional repression, how H3K9me3 regulates genome higher-order organization and transcriptional activity in living cells remains unclear. Result Here, we develop EpiGo (Epigenetic perturbation induced Genome organization)-KRAB to introduce H3K9me3 at hundreds of loci spanning megabases on human chromosome 19 and simultaneously track genome organization. EpiGo-KRAB is sufficient to induce genomic clustering and de novo heterochromatin-like domain formation, which requires SETDB1, a methyltransferase of H3K9me3. Unexpectedly, EpiGo-KRAB-induced heterochromatin-like domain does not result in widespread gene repression except a small set of genes with concurrent loss of H3K4me3 and H3K27ac. Ectopic H3K9me3 appears to spread in inactive regions but is largely restricted from transcriptional initiation sites in active regions. Finally, Hi-C analysis showed that EpiGo-KRAB reshapes existing compartments mainly at compartment boundaries. Conclusions These results reveal the role of H3K9me3 in genome organization could be partially separated from its function in gene repression. Supplementary information Supplementary information accompanies this paper at 10.1186/s13059-020-02201-1.


Background
Human genome is organized in a hierarchy manner from kilobase to megabase scales such as nucleosome, loops, topologically associated domains (TADs), and A/B compartments [1][2][3][4]. It has been proposed that the loop extrusion drives TAD formation [5]. On the other hand, liquid-liquid phase separation is suggested to mediate genome compartmentalization [5,6]. For example, heterochromatin protein HP1α undergoes liquid-liquid demixing suggesting a role of phase separation in heterochromatin domain formation [7][8][9]. Heterochromatin drives compartmentalization in the inverted nuclei of rods in nocturnal mammals [10]. The co-segregated compartments often share similar chromatin states such as histone marks [11]. Despite the widely observed correlations [11], how the epigenetic modifications regulates genome organization, particularly in living cells, remains unclear. Direct visualization of chromatin structures in cell nucleus is still challenging. Multiplexed DNA FISH, OligoSTORM, Hi-M, or ORCA has been applied to trace DNA folding [12][13][14][15][16] in fixed cells. On the other hand, CRISPR-based imaging provides a versatile and powerful tool to track chromatin topology in live cells in real time [17,18] .
Here we develop a CRISPR-based EpiGo-KRAB system to investigate the effect of ectopic H3K9me3 on genome organization and transcription in living cells. We show EpiGo-KRAB of a large chromatin domain induces de novo heterochromatin-like domain formation and reshapes local compartmentalization. Surprisingly, EpiGo-KRABinduced heterochromatin-like domain does not coccurently result in global gene repression. We believe that this system should be applicable for other epigenetic perturbations.

Establishment of EpiGo-KRAB
To investigate how H3K9me3 regulates genome architecture and gene expression in living cells, here we developed a CRISPR-based system, namely EpiGo (Epigenetic perturbation induced Genome organization)-KRAB (Fig. 1a). The EpiGo-KRAB system allows to epigenetic manipulation of defined regions and visualizing the subsequent spatiotemporal dynamics of these loci. First, we utilized dCas9-KRAB, which deposits loci-specific epigenetic modifications [19,20], and fluorescent guide RNAs from the CRISPRainbow system (sgRNA-2XPP7 and PCP-GFP) for DNA visualization [21]. We intended to alter the epigenetic states from kilobases to megabase scales, to study the role of epigenetic modification in genome organization at different scales. By mining the chromosome-specific repeats across megabases of human genome, we found a repeat class which consists of 836 copies of CRISPR target sites spanning~17 megabases at the q-arm of chromosome 19, and we dubbed it C19Q as EpiGo-KRAB targets for the following studies ( Fig. 1a and Table S1). C19Q can be visualized by co-expression of dCas9-KRAB, sgRNA-2XPP7, and PCP-GFP. We termed it as C19Q-KRAB. C19Q-KRAB could presumably recruit SETDB1 [22,23] which deposits loci-specific H3K9me3 [24]. HP1α could then interact with loci or regions with H3K9me3 [25,26]. Thus, we expect that EpiGo of C19Q will allow us to track the changes of genome organization upon epigenetic alterations in living cells. To further track the dynamic interaction between EpiGo-mediated H3K9me3 and HP1α condensates, HaloTag was knockedin at the C-terminus of HP1α by CRISPR-Cas9 system. This U2OS-HP1α-HaloTag cell line was then used to generate cell lines stably expressed C19Q-sgRNA-2XPP7, PCP-GFP and dCas9 or dCas9-KRAB, resulting in U2OS-EpiGo-Control or U2OS-EpiGo-KRAB cells for direct visualization of C19Q upon ectopic H3K9me3 modifications (Fig. 1b, c). Finally, ChIP-seq analysis confirmed that most target site of C19Q successfully acquired ectopic H3K9me3 ( Fig. S1 and Fig. 1d). These data demonstrate that we have successfully induced large scale epigenetic alterations using EpiGo-KRAB.
It was reported that dCas9-KRAB-mediated H3K9me3 deposition via histone methyltransferase SETDB1 [23,27]. To test whether SETDB1 is essential for EpiGo-KRAB- Fig. 1 A CRISPR-based EpiGo-KRAB system to track genome organization upon epigenetic modification. a A scheme of the EpiGo system. EpiGo (Epigenetic perturbation induced Genome organization) system consists of dCas9-KRAB, PCP-GFP, and sgRNA-2XPP7 targeting to C19Q repeats, which contain 836 copies of target sites on the q-arm of chromosome 19. C19Q-KRAB will recruit SETDB1 and induce H3K9 trimethylation at each target sites. H3K9me3 regions will recruit HP1α and induce genome organization. b CRISPR-based imaging of C19Q in presence of PCP-GFP, C19Q-sgRNA-2XPP7, and dCas9, marked with GFP-C19Q. The HP1α-HaloTag stable cell lines was used to examine the colocalization between C19Q (green) and HP1α (magenta). c CRISPR imaging with the same conditions as b except that dCas9 was replaced by dCas9-KRAB and marked with GFP-C19Q-KRAB. d H3K9me3 state of EpiGo-Control (dCas9) and EpiGo-KRAB (dCas9-KRAB) cell lines. ChIP-seq of H3K9me3 was performed in these two cell lines. C19Q-1 region (chr19: 48,800,000-52,300,000) is chosen to show the difference of H3K9me3 state between EpiGo-Control and EpiGo-KRAB cell lines mediated genomic clustering, SETDB1 was knocked down by siRNAs in EpiGo-KRAB (dCas9-KRAB) cell lines (Fig. 2d). The mRNA levels of SETDB1 decrease to 26% of the mock cells when transfected with SETDB1 siRNA. The percentage of C19Q clustering is 87% in cells transfected with control siRNA, but markedly decreased to 14% in cells transfected with SETDB1 siRNA (Fig. 2f). Finally, C19Q loci were highly dynamic in EpiGo-Control cells (Fig. 3a-c) but the mobility of C19Q loci dramatically decreased in EpiGo-KRAB cell lines ( Fig. 3d-f). These results support a role of H3K9me3 in mediating phase separation of HP1α droplets, large-scale genome organization, and chromatin dynamics.  Immunofluorescence showed extensive overlapping between H3K9me3 and HP1α by conventional widefiled microscopy ( Fig. 4a, b), while structured illumination microscopy (3D-SIM) showed that H3K9me3 localized nearby HP1α condensates with less degree of overlapping (Fig. 4c, d). Statistic analysis showed H3K9me3 and HP1α areas showed~52% overlap in images from widefield microscopy, which decrease to 25% when measured by 3D-SIM (Fig. 4e). Together with the finding that EpiGo-KRAB of C19Q loci adjacent to HP1α condensates, our data suggest that the role of chromatin We then sought to investigate the effect of ectopic H3K9me3 on transcription. The C19Q repeats are present in gene body (753 copies, 89%), promoter or distal regions (98 copies, 11%), and intergenic regions (158 copies, 18%) (Fig. 5a). The C19Q-1 region spans 3.5 Mb and contains 146 genes (Fig. 5b). This region contains clearly two sub-regions: regions 1 and 2 (Fig. 5b, c). Most genes (64%) in region 1 are active or modestly active, while the majority of genes (79%) in region 2 are inactive in EpiGo-Control (dCas9) cells. The peaks of EpiGo-induced ectopic H3K9me3 were negatively correlated with active transcription (Fig. 5b), as also observed for endogenous H3K9me3 genome wide (Fig. S2). We found that the expression of the majority (61 out of 73) of active genes in regions 1 and 2 was not affected by EpiGo-KRAB (Fig. 5b, c). As shown in Fig. 5d, EpiGo-KRAB does not affect the global expression levels neither. A close examination of EpiGo-KRAB targeted genes revealed that a small set of active genes are indeed silenced. For these genes, such as the top five silenced genes (KCNN4, KDELR1, HSD17B14, KLK6, and TNNT1), ectopic H3K9me3 primarily occurred at the promoter regions, which is accompanied by the loss of H3K4me3 and H3K27ac (Fig. 5e). We examined the relationship between the distance of CRISPR targets to gene promoters and their repression efficacy (Fig. S3). The result suggested that when closer the dCas9 target is to its nearby promoter, the more likely the Fig. 5 The influence of H3K9me3 on transcription. a Diagram of targeted positions by the EpiGo-KRAB system, which includes promoter, gene body (5′-UTR, exon, intron, 3′-UTR), and intergenic region. There are 98 copies of C19Q repeats on promoters, 158 copies on intergenic regions, and 753 copies on gene bodies. b RNA levels in EpiGo-Control (dCas9) and EpiGo-KRAB (dCas9-KRAB) cell lines. RNA-seq was performed in these two cell lines. Region 1 (chr19: 48,800,000-51,200,000) or region 2 (chr19: 51,200,001-52,300,000) of C19Q-1 with different transcriptional states was highlighted with cyan or pink respectively. c Histogram shown active, modestly active, and inactive genes in region 1 and region 2 in EpiGo-Control (dCas9) and EpiGo-KRAB (dCas9-KRAB) cell lines. Active genes were defined by applying a threshold of FPKM ≥ 5, modestly active genes were applied a threshold of 1 < FPKM < 5, and inactive genes were used a threshold of FPKM ≤ 1. d Global expression level in EpiGo-Control (dCas9) and EpiGo-KRAB (dCas9-KRAB) cell lines. e ChIP-seq of H3K9me3, H3K4me3, H3K27ac, and RNA-seq were performed in EpiGo-Control (dCas9) and EpiGo-KRAB (dCas9-KRAB) cell lines. KCNN4, HSD17B14, TNNT1 genes and KDELR1, KLK6 genes were highlighted with light orange or cyan respectively. All genes were marked location in C19Q repeats with red lines respective gene would be repressed. Such observation is also confirmed by the analyses of promoter H3K4me3 and H3K27ac changes (Fig. S3). By contrast, the orientation does not have apparent effects on gene repression. These data together suggest that EpiGo-KRABmediated global chromatin clustering does not result in widespread gene silencing beyond its direct promoter targets.

Ectopic H3K9me3 spreads coutinously in inactive regions but restrictedly from transcriptional initiation sites in active regions
To reciprocally investigate the effect of transcriptional activity on H3K9me3 spreading, we chose two areas in region 1 and region 2. KDELR1 and CYTH2 located within a 150 kb region in region 1. KDELR1 acquired H3K9me3 in the promoter region, which was concurrently loss of H3K4me3 and H3K27ac, and lead to gene silencing ( Fig. 6a, b, d-f). CYTH2, which acquired ectopic H3K9me3 in the 3′-UTR region, did not show significant changes of promoter chromatin states and transcriptional level ( Fig. 6a, b, Fig. 6 EpiGo-KRAB-mediated H3K9me3 deposit does not result in widespread gene silencing. a H3K9me3 and RNA levels in EpiGo-Control (dCas9) and EpiGo-KRAB (dCas9-KRAB) cell lines. ChIP-seq of H3K9me3 and RNA-seq were performed in these two cell lines. Region 1 (chr19: 48,800,000-51,200,000, cyan) or region 2 (chr19: 51,200,001-52,300,000, pink) of C19Q-1 with different transcriptional states was highlighted with cyan or pink respectively. b H3K9me3, H3K4me3, H3K27ac states and RNA level in region 1. ChIP-seq of H3K9me3, H3K4me3, H3K27ac, and RNA-seq were performed in EpiGo-Control (dCas9) and EpiGo-KRAB (dCas9-KRAB) cell lines. KDELR1 gene or CYTH2 gene were highlighted with light orange or light green respectively. c H3K9me3, H3K4me3, H3K27ac states and RNA levelin region 2. KLK6 gene or CTU1 gene were highlighted with light orange or light green respectively. d Fold changes of mRNA level and H3K9me3 in the EpiGo-KRAB targeted genes. The active genes (FPKM ≥ 1) targeted by EpiGo-KRAB were chosen for the analysis. The five genes (KCNN4, KDELR1, HSD17B14, KLK6, and TNNT1) with most reduction in mRNA levels were marked in red. e Fold changes of mRNA level and H3K4me3 in the EpiGo-KRAB targeted genes. f Fold changes of mRNA level and H3K27ac in the EpiGo-KRAB targeted genes d-f). On the contrary, H3K9me3 spreading halted at the site of CYTH2 promoter (Fig. 6b, d). Interestingly, ectopic H3K9me3 appears to frequently exist as continuous domains in region 2 (Fig. 6a, c). For example, H3K9me3 spreads beyond C19Q repeats for hundreds of kilobases in region 2 until the proximity of an active gene CTU1 (Fig. 6c). By contrast, ectopic H3K9me3 in active region (region 1 in Fig. 6b) is intermittent and its spreading is frequently restricted, often stopping at the sites of active genes. These results raise a possibility that H3K9me3 spreads more efficiently in transcriptionally silenced regions but may be antagonized by active chromatin states.

EpiGo-KRAB induces de novo heterochromatin-like domain formation
To further examine genome organization of the active or silenced regions upon H3K9me3, Oligopaint FISH was used to visualize region 1, region 2, and region 3 ( Fig. 7a and Table S2-S3). Region 1 was transcriptionally active and showed intermittent H3K9me3 in EpiGo-KRAB cell lines. Region 2 was silenced and showed more continuous H3K9me3 in EpiGo-KRAB cell lines. We also included a neighbor region 3′ downstream of region 2 as a comparison (region 3). Region 3 showed strong and continuous H3K9me3 in both EpiGo-Control and EpiGo-KRAB cell lines, consistent with the fact that it is not targeted by EpiGo-KRAB. In EpiGo-Control cells, region 1 showed multiple discrete foci, which partially overlap with C19Q probe but with few loci associate with HP1α ( Fig. 7b-d). By contrast, region 2 exists as only one focal spot in the majority of cells. It overlaps with C19Q but, surprisingly, barely associated with HP1α ( Fig. 7b-d), suggesting that region 2 is compacted independent of HP1α association before EpiGo-KRAB induction and ectopic H3K9me3 acquisition. Region 3, which is extensively modified by H3K9me3 in EpiGo-Control cells, has one focal spot and is associated with HP1α ( Fig. 7b-d).
These results suggested that silenced regions (region 2 or 3) are more compacted than active regions (region 1), and their association with HP1α appears to be related to H3K9me3 strength. Indeed, in EpiGo-KRAB cells, region 1 collapsed into one focal spot and associated with HP1α ( Fig. 7e-g). Region 2 again had only one focal spot, but now became highly associated with HP1α ( Fig. 7e-g). These results indicated that EpiGo-KRAB promotes its association with HP1α condensates and further chromatin compaction perhaps in an H3K9me3 density-dependent manner.

EpiGo-KRAB-induced large-scale rearrangement of chromatin compartmentalization
Genome is organized in 3D from sub-kilobase to megabase scales and segregated into domains in cellular nucleus [4]. Genome compartmentalization occurs at the hundreds of kilobases to megabase scales from Hi-C heatmap [11]. To understand how epigenetic manipulation affects compartmentalization in active and silenced regions, Hi-C was performed on U2OS-EpiGo-Control (dCas9) and U2OS-EpiGo-KRAB (dCas9-KRAB) cell lines. As a control, no drastic changes of global compartmentalization were observed in the Hi-C matrix of the genome upon EpiGo-KRAB induction (data not shown). Focused on the C19Q region, which showed substantial changes of H3K9me3 states, we found drastic rearrangement of local compartments (Fig. 8a, b). For instance, the entire region 2 and region 3 fused to become a large compartment in EpiGo-KRAB cells, which even erodes part of region 1 (1/2/3 compartment) (Fig. 8b). Throughout the regions, we have found a number of regions with C19Q repeats that show increased interactions with their neighbor regions, leading to compartment merging (red arrows, Fig. 8a). These results indicated that EpiGo-KRAB-induced extensive re-arrangement of chromatin compartments. We reasoned that this is possibly because H3K9me3 marked regions anchor chromatin to HP1α condensates, which breaks existing compartmentalization [28] and form new compartments. Therefore, the final compartments in the genome may be decided by the overall topology executed through a default compartmentalization (gene density and transcriptional state-correlated) altered by architectural proteins and epigenetic states. 48,800,000-51,200,000) representing an active region, region 2 (chr19: 51,200,001-52,300,000) representing a silenced region, and region 3 (chr19: 52,300,001-54,300,000) containing endogenous H3K9me3 was highlighted in cyan, pink, and yellow respectively. The target sites of C19Q repeats were shown in green. Oligopaint FISH probes were designed to detect region 1, region 2, and region 3 respectively. b Images of C19Q, HP1α and region 1, 2, or 3 in EpiGo-Control cell lines. C19Q (green) was labeled by GFP using CRISPR imaging, HP1α (magenta) was labeled by HaloTag, and region 1, 2, or 3 (blue) was labeled by in situ hybridization with DNA probes conjugated to the dye Alexa-647. The colocalization among C19Q, HP1α, and region 1, 2 or, 3 was shown in merged images. c Percentage of cells with one spot vs multiple spots in region 1, 2, or 3 of EpiGo-Control (dCas9) cell lines. n = 200 transfected cells in each experiment and the error bar is standard deviation (SD) from three independent experiments. d Percentage of foci association with HP1α in EpiGo-Control cell lines. The association or nonassociation is defined by overlapping (regardless of the degrees) or non-overlapping between foci and HP1α condensates. n = 16, 16, and 12 cells from left to right respectively. e Images of C19Q, HP1α and region 1, 2, or 3 in EpiGo-KRAB cell lines. All the conditions are the same as b except the dCas9 was replaced by dCas9-KRAB. The compartment A-to-B switch induced by EpiGo-KRAB is associated with multiple factors As EPIGO-KRAB would induce compartment re-organization, we first asked what would be the threshold of dcas9-KRAB target density to achieve a compartment A-to-B switch. A close examination revealed that out of 49 compartment A regions (100 kb bin) with target sequences in dCas9 cells, surprisingly only 4 of them were converted to compartment B in dCas9-KRAB cells (A-to-B), while the rest 45 remained as compartment A (A-to-A), raising the possibility that the target density may need to exeed a threshold to flip a compartment A to B. Indeed, we found that the average target densities (per 100 kb bin) in A-to-B regions (n = 10, 49,18, 20, respectively, see red box in Fig. 9a, b) are much higher than A-to-A regions (Fig. 9c), supporting the notion that high target density is needed to achieve chromatin organization alterations. Nevertheless, the dense targeting of dCas9-KRAB is clearly not a simple rule, as some A-to-A regions and A-to-B regions show high target density (Fig. 9c). One A-to-A example can be seen in Fig. 9a, b which show dense presence of dCas9-KRAB targets (green box). Notably, these regions appear to enrich for active genes (Fig. 9a, b), raising the possibility that additional factors may be also involved in compartment switch. Notably, our recent study also showed that during SCNT, many compartments A and B remain constant between mouse donor cells and SCNT embryos, and they preferentially have the highest and lowest gene densities respectively [29]. By contrast, switched compartment regions tend to have intermediate gene density [29]. We now confirmed such results in additional mouse cells (MEF vs. cumulus cells [29,30]) and human cells (hESC vs trophectoderm [31]) (Fig. 9d). Similar results were also reported for laminaassociated domains (LADs) which are closely associated with compartments A/B [32].
Regions with high and low gene density tend to be constant LADs (cLADs) and constant inter-LADs, while variable LADs (vLADs) preferentially show intermediate gene density and frequently reside at the boundary of LADs and iLADs [33]. Our analyses revealed that compared to the whole genome, chr19 has unusually high levels of gene density, including both A-to-B and A-to-A regions (Fig. 9e), suggesting that these regions may be naturally resistant to A-to-B switches. In addition, A-to-A regions show higher gene expression and histone acetylation levels than A-to-B regions, supporting that A-to-A regions are more likely to be stable compartment A (Fig. 9f). Finally, it was reported that variable LADs (vLADs) frequently reside at the boundary of LADs and iLADs [33]. Similarly, we found that A-to-B regions are preferentially located near A/B boundaries (Fig. 9g, see example in Fig. 9a, b, red box).
In sum, we show that compartment switch is likely regulated by multiple factors including dcas9-KRAB binding, closeness to A-B borders, and gene density. Compartment A with targeted sequences in C19Q show particularly high gene density in most regions (Fig. 9e), suggesting that these regions may be naturally more resistant to A-to-B switch, unless they have abundant sgRNA targets and/or are near compartment A-B boundaries.

Discussion
Despite the long-observed correlation, whether epigenetic modifications can regulate 3D genome architecture in live cells remains poorly understood. The EpiGo-KRAB system provides a powerful tool to allow H3K9me3 at specific genes or regions and track their changes of location, structure, and dynamics. Our data show that EpiGo-KRAB is able to mediate de novo heterochromatin-like domain formation in a dynamic and These results also suggest neither non-promoter ectopic H3K9me3 per se nor H3K9me3-mediated chromatin compaction and phase separation with HP1α condensates is sufficient for gene silencing. In fact, active transcription can persist in EpiGo-KRAB-induced cells and appear to be able to antagonize H3K9me3 spreading. It would be worthy to further explore how transcription and H3K9me3 interplay in space and time with diverse biological context. Another interesting question is what degree of H3K9me3 is required for HP1α condensate association, chromatin local compaction, or genome compartmentalization. Here, by analyzing different regions with distinct transcriptional activities, we found that repressed regions can form compact DNA without associating with HP1α condensate, which is consistent to recent finding that compaction of mouse heterochromatin foci can be independent of HP1 [34]. However, high levels of ectopic H3K9me3 may be necessary to promote HP1α association and chromatin compaction in active regions.
Finally, drastic changes of epigenetic modifications such as H3K9me3, H3K27me3, and DNA methylation occur in stem cell differentiation, embryonic development, and many diseases [35][36][37]. With the EpiGo system, it will be interesting to explore how tissue-specific epigenetic modifications are established and maintained at kilobase to megabase scales and how they regulate chromatin architecture, gene expression, and cell fate decision.

Cell culture
The U2OS (human bone osteosarcoma epithelial, female) cell and HEK293T cells (human embryonic kidney epithelial, female) were cultured in DMEM (Life Technologies) with high glucose in 10% FBS (fetal bovine serum, Life Technologies). All cells were cultured at 37°C and 5% CO 2 in a humidified incubator.

Chromosome-specific repeats for the EpiGo system
Mining of chromosome-specific repeats was described previously with some modifications [38]. The human reference genome (assembly GRC h37/hg19) was downloaded from the UCSC Genome Browser (http://genome.ucsc.edu) to find target regions and design guide RNAs. The bioinformatics tool Jellyfish [39] was used to find all 15-mers (12-mers ending with NGG or starting with CCN) with at least 5 copies in any chromosome. The 15-mers with more than 100,000 targets were filtered out. Each candidate 15-mer was searched for off-targets in all other chromosomes. The candidate 15-mer was discarded if there was any cluster of 5 offtargets or more within any 50 kb region. The C19Q repeats (Table S1), which consists of 836 copies of CRISPR target sites (TCCCTCAGACCCnGG) spanning~17 megabases at the q-arm of chromosome 19, was chosen as EpiGo targets in this study. The overlapping genes including the intron/exon information were extracted from the GENCODE Genes track (version 27lift37), downloaded from the UCSC Genome Browser.

Plasmids construction
The expression plasmid pHAGE-TO-dCas9 has been described previously [17,21], in which HSA-P2A was inserted at the N-terminal of dCas9 resulting in pHAGE-TO-HSA-P2A-dCas9 and KRAB was then subcloned into the C-terminal of dCas9 resulting pHAGE-TO-HSA-P2A-dCas9-KRAB. The expression plasmid pHAGE-EFS-PCP-GFP has been described previously [21]. The expression vector for guide RNA was based on the pLKO.1 lentiviral expression system, in which TetR-BFP-P2A-2XPP7 was inserted right after the phosphoglycerate kinase (PGK) promoter, resulting pTetR-P2A-BFP-2XPP7. The expression plasmid for guide RNAs targeting to C19Q was made by using the rapid guide RNA construction protocol described previously [38].
The donor plasmid for knock-in of HaloTag at the C-terminal of HP1α was made by Golden Gate cloning. The donor consists of three fragments: a 400-bp fragment just upstream of stop codon of HP1α (left arm), HaloTag coding sequences, and a 400-bp fragment just downstream of stop codon of HP1α (right arm). Left arm and right arm fragments were amplified from U2OS genomic DNA isolated with Cell DNA isolation mini kit (Vazyme). The HP1α-HaloTag donor was assembled by Golden Gate cloning method into pDONOR vector [38], resulting in pDONOR-HP1α-HaloTag. The guide RNA targeting sequences AAACAGCAAAGAGCTAAAGG spanning the stop codon (TAA, underlined) of HP1α and cloned into guide RNA expression vector pLH-sgRNA2 [21], resulting in pLH-sgRNA2-HP1α.

Generation of U2OS-HP1α-HaloTag cell lines
For knock-in of HaloTag into HP1α locus, U2OS cells were co-transfected with 200 ng of pHAGE-TO-Cas9, 600 ng of pLH-sgRNA1-HP1α, and 600 ng pDONOR-HP1α-HaloTag using Lipofectamine 2000 (Life Technologies) for 6 h and then replaced culture media containing 2 nM HaloTag-JF549. The transfected cells were cultured for additional 24-48 h before examining the knock-in efficiency by fluorescent imaging or flow cytometry.
Fluorescence imaging was used to check the proper localization of HP1α-HaloTag. The HaloTag positive cells was sorted by BD FACS Aria III equipped with 561 nm excitation laser, and the emission signals were detected by using filter at 610/20 nm (wavelength/bandwidth) for HaloTag-JF549. The localization of HP1α-HaloTag was examined again under fluorescence microscope after cultured for additional two weeks. The resulting cell line was named U2OS-HP1α-HaloTag.

Oligopaint FISH probe design and preparation
Oligopaint FISH probes targeting the q-arm of chromosome 19 were designed according to OligoMiner (https://github.com/brianbeliveau/OligoMiner) and probe datasets for Human chromosome 19 from hg19 with mining settings of Balance (https:// oligopaints.hms.harvard.edu/genome-files). Here we focused on the region of chr19: 48.8-54.3 Mb (hg19) including region A (chr19: 48.8-51.2 Mb), region B (chr19: 51.2-52.3 Mb), and region C (chr19: 52.3-54.3 Mb). The detail design of region A, region B, and region C probes were shown in Table S2 and Table S3. The oligo pool for Oligopaint FISH probes was ordered from Synbio Technologies Inc. The preparation of Oligopaint FISH probes has been described previously [13,40]. Briefly, the oligo pools for region A, region B, and region C were amplified via limited-cycle PCR (Vazyme), with forward primer and corresponding reverse readouts (one per 300 kb region) with T7 promoter sequence TAATACGACTCACTATAGGG appended to its 5′ end. Each pool was purified by AxyPrep DNA Gel Extraction Kit (Axygen). The oligo pools for region A, region B, or region C were used as templates for in vitro transcription by HiScribe T7 quick high yield RNA synthesis kit (NEB). The obtained RNA products were converted back to DNA oligo probes via reverse transcription (RT) with common RT primer (ATCGAC CCGGCATCAACGCCACGATCAGCT) conjugated with a 5′ AlexaFluor 647 using Maxima H Minus RT enzyme (Thermo Fisher). The intermediate RNA products were removed with NaOH-EDTA solution in 95°C for 10 min, and the single-strand DNA oligo probes were purified via DNA clean & concentrator-25 (D4033, Zymo Research). The ssDNA probes were either directly used for Oligopaint FISH or store in aliquots at − 80°C.

Immunofluorescence
Cells were cultured on #1 coverslip (Electron Microscopy Sciences), and then fixed with 4% PFA for 10 min, washed three times with 1XPBS and permeabilized by 0.1% Triton X-100 in 1XPBS for 10 min. Cells were incubated with rabbit polyclonal antibodies against H3K9me3 (1:1000; Active Motif) in the 1XPBS for 1 h at room temperature. After washed three times with 1XPBS and incubated with Goat anti-rabbit lgG (H+L) secondary antibody (1:500, Invitrogen) for 1 h. After washed three times with 1XPBS, coverslips were mounted with Prolong antifade reagents (Invitrogen) for imaging. ImmunoFISH 1 × 10 5 of U2OS-EpiGo cells were cultured on the #1 coverslip (Electron Microscopy Sciences) and added 500 ng/ml doxycycline for induction of dCas9 or dCas9-KRAB expression. After 72 h, cells were fixed with 4% PFA for 10 min, washed by 1XPBS for three times, and permeabilized with PBS-0.5%Triton X-100 for 10 min.
Cells were incubated with Rabbit anti-GFP polyclonal antibody (ab290, Abcam) in the 1XPBS for 1 h at room temperature. After washed three times with 1XPBS and incubated with Alexa Fluor 488 Goat anti-rabbit lgG (H+L) secondary antibody (1:500, Invitrogen) for 1 h. After washed three times with 1XPBS, cells were fixed with 4% PFA again for 20 min. The fixed cells were treated with 0.1 N HCl for 5 min and washed with 2XSSCT (300 mM NaCl, 30 mM sodium citrate, and 0.1% Tween-20), followed by incubation with 50% (vol/vol) formamide in 2XSSCT in 60°C for 20 min. Samples were denatured at 78°C for 3 min with 20 μl hybridization buffer (2XSSC, 50% formamide, 10% dextran sulfate, 0.2% Tween-20 mixed with~1.6 μM total DNA oligo probes) and incubated at 42°C for 16 h in a humidified chamber. Samples were washed by 2XSSCT for 4 times at 60°C and 2 times at room temperature. The cells were washed by 1XPBS with 0.5 M NaCl before mounted with Prolong antifade reagents (Invitrogen) for imaging.

Live cell imaging
All live cell imaging was carried out on a DeltaVision OMX™ V4 imaging system (GE Healthcare), equipped with a 60x (NA 1.42) Plan Apo oil-immersion objective (Olympus). The cells were cultured on No. 1.0 glass bottom dishes (MatTek). The microscope stage incubation chamber was maintained at 37°C and 5% CO 2 . HaloTag-JF549 was excited at 561 nm, and its emission was collected using filter at 609/37 nm (wavelength/ bandwidth); GFP was excited at 488 nm and collected using filter at 498/30 nm (wavelength/bandwidth). Imaging data were acquired by DeltaVision Elite imaging (GE Healthcare Inc.) software. Tracking C19Q cluster formation was performed for 4 h and images were captured every 30 min. For the representative images, the raw data were deconvoluted by softWoRx (GE Healthcare Inc.) software.

3D-SIM procedure
3D-SIM was performed on a DeltaVision OMX V4 system (GE Healthcare) equipped with a 60x (1.42 NA) Plan Apo oil-immersion objective (Olympus) and six laser (405, 445, 488, 514, 568, and 642 nm). Image stacks were captured and were reconstructed using softWoRx (GE Healthcare). Images were registered with alignment parameters obtained from calibration measurements with 100 nm diameter TetraSpeck Microspheres with four colors (Molecular Probes).

SETDB1 knockdown
SETDB1 siRNA (GGCAAGAAGAGAACUAAGA) and Control siRNA were purchased from GenePharma. EpiGo-KRAB cell lines were transfected with control siRNA or SETDB1 siRNA by SiRNA-Mate (GenePharma) according to the manufacture's instruction. Cells were collected 48 h after transfection. For RT-qPCR, total RNAs were isolated using quick-RNA MiniPrep kit (Zymo Research Inc.) according to manufacturer's instruction. RNA was converted to cDNA using HifairTM II 1st Strand cDNA Synthesis SuperMix (Yeasen). Quantitative PCR was carried out on StepOnePlusTM by using Hieff UNICON® Power qPCR SYBR Green Master Mix (Yeasen). All qPCR data is represented as the mean +/− standard deviation of three replicates. Primer sequences for SETDB1 and GAPDH are as follows: SETDB1 forward primer (AGGAACTTCGGCAT TTCATCG) and SETDB1 reverse primer (TGTCCCGGTATTGTAGTCCCA); GAPD H forward primer (CGACCACTTTGTCAAGCTCA) and GAPDH reverse primer (AGGGGTCTACATGGCAACTG).

ChIP-seq
Cells were collected and crosslinked by 1% formaldehyde. Cells were suspended in cell lysis buffer B (50 mM Tris-HCl PH 8.0, 20 mM EDTA, 0.3% SDS, freshly added protease inhibitor) for 10 min on ice before sonication. One milligram beads (Invitrogen) were washed with PBS+5 mg/mL BSA for three times and incubated with 5 μg antibody at 4°C for 6-8 h before adding chromatin. Sonicated chromatin was diluted in dilution buffer (16.7 mM Tris-HCl PH 8.0, 1.1% Triton X-100, 1.2 mM EDTA, 167 mM NaCl, freshly added protease inhibitor) and was incubate with pre-treated beads and antibodies by rotation at 4°C overnight. Then, beads were washed with washing buffer (50 mM Hepes PH 8.0, 1% NP-40, 0.7% DOC, 0.5 M LiCl, freshly added protease inhibitor) for 5 times followed by wash with TE once. One hundred microliters of elution buffer (50 mM Tris-HCl pH 8.0, 1 mM EDTA, 1% SDS) was added and beads were incubated in Thermo mixer at 65°C for 30 min with max speed. Supernatant was collected and treated with protease K at 55°C for 2 h, then purified with DNA purification kit (Tiangen). RNA was removed by treatment of RNase at 37°C for 1 h. DNA was then purified by AMPure Beads (Beckman) and subjected to DNA library preparation as described below.

RNA-seq library preparation and sequencing
Five milligrams of RNA was extracted using quick-RNA MiniPrep kit (Zymo) and then treated with DNase I (Fermentas) at 37°C for 1 h. RNA was then purified using AMPure beads. Poly-A tailed mRNA was collected using Dynabeads™ mRNA purification kit (Invitrogen). Purified RNA was fragmented with RNA Fragmentation Buffer (NEB) at 95°C for 5 min. Reaction was stopped and RNA was purified by AMPure beads. First-strand cDNA was synthesized with a commercial kit using both oligo dT and random primers (Invitrogen). Second-strand cDNA was synthesized with secondstrand synthesis buffer (Invitrogen), MgCl 2 , DTT, dNTP, dUTP, RNase H (Fermentas), E. coli DNA ligase (NEB), and DNA polymerase I (NEB). DNA was purified after 2 h incubation on thermomixer at 16°C. Synthesized cDNA was subjected to DNA library preparation as described below.

Library construction
Purified DNA or cDNA was subjected to NEBNext Ultra II DNA Library Prep Kit for Illumina (NEB). The DNA was resuspended in 25 μl ddH2O, followed by end-repair/A-tailing with 3.5 μl End Prep buffer and 1.5 μl enzyme mix according to manufactory instruction. The ligation reaction was then performed by adding diluted 1.25 μl adaptors (NEB), 15 μl Ligation master mix, 0.5 μl Ligation enhancer, at 4°C for overnight. The ligation reaction was treated with 1.5 μl USERTM enzyme according to the instruction and was purified by AMPure beads. The 1st round PCR was performed by adding 25 μl 2x KAPA HiFi Hot-Start Ready Mix (KAPA biosystems) with primers of NEB Oligos kit, with the program of 98°C for 45 s, (98°C for 15 s and 60°C for 10 s) with 8-9 cycles and 72°C for 1 min. DNA was purified using 1x AMPure beads and a 2nd round PCR was performed with 25 μl 2x KAPA HiFi HotStart Ready Mix with Illumina universal primers, and same PCR program for 8 cycles. The final libraries were purified by 1x AMPure beads and subjected to nextgeneration sequencing. All libraries were sequenced on Illumina HiSeq 2500 or HiSeq XTen according to the manufacturer's instruction.

sisHi-C library generation and sequencing
The sisHi-C library generation was performed as described previously [41]. Briefly, spermatogenetic cells were fixed with 1% formaldehyde at room temperature (RT) for 10 min. Formaldehyde was quenched with glycine for 10 min at RT. Cells were washed with 1XPBS for two times and then lysed in 50 μl lysis buffer (10 mM Tris-HCl pH 7.4, 10 mM NaCl, 0.1 mM EDTA, 0.5% NP-40 and proteinase inhibitor) on ice for 50 min. After spinning at 3000 rpm/min in 4°C for 5 min, the supernatant was discarded with a pipette carefully. Chromatin was solubilized in 0.5% SDS and incubated at 62°C for 10 min. SDS was quenched by 10% Triton X-100 at 37°C for 30 min. Then, the nuclei were digested with 50 U Mbo I at 37°C overnight with rotation. Mbo I was then inactivated at 62°C for 20 min. To fill in the biotin to the DNA, dATP, dGTP, dTTP, biotin-14-dCTP, and Klenow were added to the solution and the reaction was carried out at 37°C for 1.5 h with rotation. The fragments were ligated at RT for 6 h with rotation. This was followed by reversal of crosslink and DNA purification. DNA was sheared to 300-500 bp with Covaris M220. The biotin-labeled DNA was then pulled down with 10ul Dynabeads MyOne Streptavidin C1 (Life Technology). Sequencing library preparation was performed on beads, including end-repair, dATP tailing and adaptor-ligation. DNA was eluted twice by adding 20ul water to the tube and incubation at 66°C for 20 min. Nine to 15 cycles of PCR amplification were performed with Extaq (Takara). Finally, size selection was done with AMPure XP beads and fragments ranging from 200 bp to 1000 bp were selected. All the libraries were sequenced on Illumina HiSeq2500 or HiSeq Xten according to the manufacturer's instruction.

Imaging data analysis
The maximum projection of Z stack was used for the quantification of cell numbers. The Imaris 9.0.1platfrom was used to process the HP1apha and H3K9me3 channel intensity simultaneously and to measure the degree of colocolization of two channel using Coloc2. Analysis trajectory of live cell loci site using MTrack2 Plugins. Fluorescent puncta were identified as local maxima satisfying the minimum intensity and min/ max peak with threshold by Gaussian fitting. Data are represented as mean ± SEM. The exact number, n, of data points and the representation of n (cells, independent experiments) are indicated in the respective figure legends and in the "Results" section.

ChIP data processing
The paired-end ChIP reads were aligned with the parameters: -t -q -N 1 -L 25 -X 1000 --no-mixed --no-discordant. All unmapped reads, non-uniquely mapped reads, reads with low mapping quality (MAPQ < 20), and PCR duplicates were removed. For downstream analysis, we normalized the read counts by computing the numbers of reads per kilobase of bin per million of reads sequenced (RPKM) for 100-bp bins of the genome. To minimize the batch and cell type variation, RPKM values across whole genome were further Z-score normalized. To visualize the ChIP signals in the UCSC genome browser, we generated the RPKM values on a 100 bp-window base. H3K9me3 peaks were called using MACS2 [42] with the parameters -g hs -broad -nomodelnolambda --broad and noisy peaks with very weak signals (RPKM < 5) were removed from further analyses. Adjacent peaks with very close distance (< 5 kb) were merged for downstream analyses.

RNA-seq data processing
All RNA-seq data were mapped to hg19 reference genome by Tophat. The gene expression levels were calculated by Cufflinks (version 2.2.1) using the refFlat database from the UCSC genome browser.

Hi-C data mapping
Paired end raw reads of Hi-C libraries were aligned, processed, and iteratively corrected using HiC-Pro (version 2.7.1b) as described [43]. Briefly, sequencing reads were first independently aligned to the human reference genome (hg19) using the bowtie2 end-toend algorithm and "-very-sensitive" option. To rescue the chimeric fragments spanning the ligation junction, the ligation site was detected and the 5′ fraction of the reads was aligned back to the reference genome. Unmapped reads, multiple mapped reads, and singletons were then discarded. Pairs of aligned reads were then assigned to Mbo I restriction fragments. Read pairs from uncut DNA, self-circle ligation, and PCR artifacts were filtered out and the valid read pairs involving two different restriction fragments were used to build the contact matrix. Valid read pairs were then binned at a specific resolution by dividing the genome into bins of equal size. We chose 100-kb bin size for examination of global interaction patterns of the whole chromosome and 40-kb bin size to show local interactions and to perform TAD calling. Then, the binned interaction matrices were normalized using the iterative correction method [43,44] to correct the biases such as GC content, mappability, and effective fragment length in Hi-C data.

Identification of chromatin compartments A/B
Conventional chromatin compartments A and B were identified with a method described previously [45] with some modifications. The normalized 100 kb interaction matrices for each stage were used in this analysis. Firstly, the bins that have no interactions with any other bins were removed before expected interaction matrices were calculated. Observed/ expected matrices were generated using a sliding window approach [31] with the bin size of 400 kb and the step size of 100 kb. Principal component analysis was performed on the correlation matrices generated from the observed/expected matrices. The first principal component of the correlation matrices together with gene density were used to identify A/B compartments. In this analysis, the correlation matrices were calculated according to the interaction matrices separated according to the location of predicted centromeres for each chromatin, as the principal component often reflects the partitioning of chromosome arms [45]. Juicebox was used to generate all chromosome-wide as well as the zoom-in views of interaction frequency heatmap in this study.

Analysis of the promoter H3K4me3 and H3K27ac enrichment
Annotated promoters (TSS± 2.5 kb) of the hg19 refFlat annotation database from the USCS genome browser were used. Whole genome Z-score normalized RPKM scores were used for calculation. The nearest genes to the sgRNA target were found by bedtools (-closest).