Simultaneous Epigenetic Perturbation and Genome Imaging Reveal Distinct Roles of H3K9me3 in Chromatin Architecture and Transcription [preprint]

Repository Citation Feng Y, Wang Y, Wang X, He X, Yang C, Naseri A, Pederson T, Zheng J, Zhang S, Xiao X, Xie W, Ma H. (2020). Simultaneous Epigenetic Perturbation and Genome Imaging Reveal Distinct Roles of H3K9me3 in Chromatin Architecture and Transcription [preprint]. University of Massachusetts Medical School Faculty Publications. https://doi.org/10.1101/2020.07.15.204719. Retrieved from https://escholarship.umassmed.edu/faculty_pubs/1741


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 [21,22], and fluorescent guide RNAs from the CRISPRainbow system (sgRNA-2XPP7 and PCP-GFP) for DNA visualization [23]. 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 [24,25] which deposits loci-specific H3K9me3 [26]. HP1α could then interact with loci or regions with H3K9me3 [27,28]. 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 knocked-in 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-. CC-BY-NC-ND 4.0 International license was not certified by peer review) is the author/funder. It is made available under a The copyright holder for this preprint (which this version posted July 16, 2020. . https://doi.org/10.1101/2020.07.15.204719 doi: bioRxiv preprint

EpiGo-Control or U2OS-EpiGo-KRAB cells for direct visualization of C19Q upon ectopic
H3K9me3 modifications (Figure 1B and 1C). 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.

EpiGo-KRAB induces SETDB1-dependent genomic clustering and phase separation
We then sought to examine the spatiotemporal dynamics of C19Q regions upon EpiGo-KRAB mediated H3K9me3. As shown in Fig. 2A, discrete foci of C19Q (green) were visible by CRISPR labeling in U2OS-EpiGo-Control (dCas9) cells, which barely colocalize with HP1α (red) over time. However, In EpiGo-KRAB cell lines (dCas9-KRAB), C19Q loci dynamically interacted with HP1α condensates and clustered at the surface of HP1α condensates. Eventually, adjacent HP1α condensates coalesced together. To investigate how HP1α mediates local compaction of genomic regions upon EpiGo-KRAB induction (Fig. 2B), structured illumination microscopy (3D-SIM) was used to acquire high resolution imaging of C19Q and HP1α. We found that C19Q foci decorated on the surface of HP1α condensates in EpiGo-KRAB (dCas9-KRAB) cell lines (Fig. 2B). Quantitative analysis confirmed that C19Q foci were not clustered in EpiGo-Control cells, while 90% of C19Q foci showed clustering in EpiGo-KRAB cells (Fig.   2C). In sum, these results suggest that EpiGo-KRAB can induce genome reorganization through a dynamic and stepwise process.
Immunofluorescence also showed that H3K9me3 surrounded the HP1α condensates (Fig.   S2). It was reported that dCas9-KRAB mediated H3K9me3 deposition via histone methyltransferase SETDB1 [25,29]. To test whether SETDB1 is essential for EpiGo-KRAB . CC-BY-NC-ND 4.0 International license was not certified by peer review) is the author/funder. It is made available under a The copyright holder for this preprint (which this version posted July 16, 2020. . https://doi.org/10.1101/2020.07.15.204719 doi: bioRxiv preprint 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 but the mobility of C19Q loci dramatically decreased in EpiGo-KRAB cell lines (Fig. S3). These results support a role of H3K9me3 in mediating phase separation of HP1α droplets and largescale genome organization.

EpiGo-KRAB mediated H3K9me3 and genomic clustering does not result in widespread gene silencing
We then sought to investigate the relationship of ectopic H3K9me3 and transcription. The C19Q repeats are present in gene body, promoter or distal regions (Fig. S4A). The C19Q-1 region spans 3.5 Mb and contains 146 genes (Fig. 3A). This region contains clearly two sub-regions: regions 1 and 2 ( Fig. 3A and Fig. S4B-4C). 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. As shown in Fig. S4D, , EpiGo-KRAB does not affect the global expression levels. 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. 3B-3C and Fig.   S4E). This observation was confirmed by global analysis of all genes targeted by EpiGo-KRAB ( Fig. 3D-3F). Unexpectedly, we found about the expression of the majority (61 out of 73) of . CC-BY-NC-ND 4.0 International license was not certified by peer review) is the author/funder. It is made available under a The copyright holder for this preprint (which this version posted July 16, 2020. . https://doi.org/10.1101/2020.07.15.204719 doi: bioRxiv preprint active genes in Region 1 and 2 was not affected by EpiGo-KRAB ( Fig. 3A and Fig. S4B-4C).
For example, CYTH2, which acquired ectopic H3K9me3 in the 3'-UTR region, did not show significant changes of promoter chromatin states and transcriptional level (Fig. 3B, 3D-3F).
These results suggest that EpiGo-KRAB mediated global chromatin clustering does not result in widespread gene silencing beyond its direct promoter targets.
Notably, the peaks of EpiGo-induced ectopic H3K9me3 were negatively correlated with active transcription (Fig. 3B), as also observed for endogenous H3K9me3 genome wide (Fig.   S5). Interestingly, ectopic H3K9me3 appears to frequently exist as continuous domains in Region 2 (Fig. 3C). For example, H3K9me3 spreads beyond C19Q repeats for hundreds of kilobases in Region 2 until the proximity of an active gene CTU1 (Fig. 3C). By contrast, ectopic H3K9me3 in active region (Region 1 in Fig. 3B) 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 ( Fig. 4A  showed multiple discrete foci, which partially overlap with C19Q probe but barely associate with HP1α (Fig. 4B). By contrast, Region 2 exists as only one focal spot in the majority of cells ( Fig.   4B-4C). It overlaps with C19Q but, surprisingly, barely associated with HP1α, 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-KRAB cells, has one focal spot and is associated with HP1α ( Fig. 4B-4C). 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. 4D-4E). Region 2 again had only one focal spot, but now became associated with HP1α. 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. 5A and 5B). For instance, the entire Region 2 and Region 3 fused to . CC-BY-NC-ND 4.0 International license was not certified by peer review) is the author/funder. It is made available under a The copyright holder for this preprint (which this version posted July 16, 2020. . https://doi.org/10.1101/2020.07.15.204719 doi: bioRxiv preprint become a large compartment in EpiGo-KRAB cells, which even erodes part of Region 1 (1/2/3 compartment) (Fig. 5B). 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. 5A). These results indicated that EpiGo-KRAB induced extensive rearrangement of chromatin compartments. We reasoned that this is possibly because H3K9me3 marked regions anchor chromatin to HP1α condensates, which breaks existing compartmentalization [30] 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.

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 stepwise process (  Finally, drastic changes of epigenetic modifications such as H3K9me3, H3K27me3 and DNA methylation occur in stem cell differentiation, embryonic development and many diseases [32][33][34]. 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.
. CC-BY-NC-ND 4.0 International license was not certified by peer review) is the author/funder. It is made available under a The copyright holder for this preprint (which this version posted July 16, 2020. . https://doi.org/10.1101/2020.07.15.204719 doi: bioRxiv preprint 1 1

Chromosome-specific repeats for the EpiGo system
Mining of chromosome-specific repeats was described previously with some modifications [35].
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 [36] 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 off-targets or more within any 50 kb region. The C19Q repeats (Table S1), which consists of 836 copies of CRISPR target sites 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 [19,23], 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 [23]. 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 . CC-BY-NC-ND 4.0 International license was not certified by peer review) is the author/funder. It is made available under a The copyright holder for this preprint (which this version posted July 16, 2020. . https://doi.org/10.1101/2020.07.15.204719 doi: bioRxiv preprint targeting to C19Q was made by using the rapid guide RNA construction protocol described previously [35]. 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 [35], 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 [23], resulting in pLH-sgRNA2-HP1α.   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 [14,37]. 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).  and subjected to DNA library preparation as described below.

RNA-seq library preparation and sequencing
5 mg RNA was extracted using quick-RNA MiniPrep kit (Zymo) and then treated with DNase I (Fermentas) at 37°C for 1 hour. RNA was then purified using AMPure beads. Poly-A tailed mRNA was collected using Dynabeads TM 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 second strand 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 hours incubation on thermomixer at 16°C. Synthesized cDNA was subjected to DNA library preparation as described below.

sisHi-C library generation and sequencing
The sisHi-C library generation was performed as described previously [38]. Briefly, spermatogenetic cells were fixed with 1% formaldehyde at room temperature (RT) for 10 min.

Imaging data analysis
The maximum projection of Z stack was used for the quantification of cell numbers and

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 [40]. Briefly, sequencing reads were first independently aligned to the human reference genome (hg19) using the bowtie2 end-to-end 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 . CC-BY-NC-ND 4.0 International license was not certified by peer review) is the author/funder. It is made available under a The copyright holder for this preprint (which this version posted July 16, 2020. . https://doi.org/10.1101/2020.07.15.204719 doi: bioRxiv preprint 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 [40,41] to correct the biases such as GC content, mappability and effective fragment length in Hi-C data.

Identification of conventional chromatin compartments and refined-A/B
Conventional chromatin compartments A and B were identified with a method described C19Q-KRAB will recruit SETDB1 and induce H3K9 trimethylation at each target sites.
H3K9me3 regions will recruit HP1α and induce genome organization.
(C) CRISPR imaging with the same conditions as (B) except that dCas9 was replaced by dCas9-KRAB and marked with GFP-C19Q-KRAB.
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.
(B) Hi-C heatmap shown the compartmentalization of the C19Q region in EpiGo-KRAB (dCas9) cells. All the data processing and display are the same as (A).
. CC-BY-NC-ND 4.0 International license was not certified by peer review) is the author/funder. It is made available under a The copyright holder for this preprint (which this version posted July 16, 2020.             The H3K9me3 state and RNA levels of 30-40 Mb regions in chromosome 1-6 were shown in A-F respectively. The ChIP-seq of H3K9me3 and RNA-seq were performed on U2OS-EpiGo-Control cell lines. Table S1. Target sites of C19Q on the q-arm of human chromosome 19.

Supplementary Movies
Movie S1. Tracking the movement of C19Q loci in U2OS-EpiGo-Control cells.
Images were cropped to 50X50 pixels and each video includes 300 frames (a total time of 30 seconds). The imaging rate is 100 milliseconds per frame and the play rate is 30 frames per second. The individual locus movements were corrected for the possible movements of microscope stage.

Movie S2. Tracking the movement of C19Q loci in U2OS-EpiGo-KRAB cells.
The image processing details are the same as described in the Video S1.
. CC-BY-NC-ND 4.0 International license was not certified by peer review) is the author/funder. It is made available under a The copyright holder for this preprint (which this version posted July 16, 2020. . https://doi.org/10.1101/2020.07.15.204719 doi: bioRxiv preprint