Targeted DNA methylation in pericentromeres with genome editing-based artificial DNA methyltransferase

To study the impact of epigenetic changes on biological functions, the ability to manipulate the epigenetic status of certain genomic regions artificially could be an indispensable technology. “Epigenome editing” techniques have gradually emerged that apply TALE or CRISPR/Cas9 technologies with various effector domains isolated from epigenetic code writers or erasers such as DNA methyltransferase, 5-methylcytosine oxidase, and histone modification enzymes. Here we demonstrate that a TALE recognizing a major satellite, consisting of a repeated sequence in pericentromeres, could be fused with the bacterial CpG methyltransferase, SssI. ChIP-qPCR assays demonstrated that the fusion protein TALMaj-SssI preferentially bound to major chromosomal satellites in cultured cell lines. Then, TALMaj-SssI was expressed in fertilized mouse oocytes with hypomethylated major satellites (10–20% CpG islands). Bisulfite sequencing revealed that the DNA methylation status was increased specifically in major satellites (50–60%), but not in minor satellites or other repeat elements, such as Intracisternal A-particle (IAP) or long interspersed nuclear elements-1 (Line1) when the expression level of TALMaj-SssI is optimized in the cell. At a microscopic level, distal ends of chromosomes at the first mitotic stage were dramatically highlighted by the mCherry-tagged methyl CpG binding domain of human MBD1 (mCherry-MBD-NLS). Moreover, targeted DNA methylation to major satellites did not interfere with kinetochore function during early embryonic cleavages. Co-injection of dCas9 fused with SssI and guide RNA (gRNA) recognizing major satellite sequences enabled increment of the DNA methylation in the satellites, but a few off-target effects were also observed in minor satellites and retrotransposons. Although CRISPR can be applied instead of the TALE system, technical improvements to reduce off-target effects are required. We have demonstrated a new method of introducing DNA methylation without the need of other binding partners using the CpG methyltransferase, SssI.


Plasmid construction and transfection
Plasmids encoding major satellite recognition TALE [30], dCas9 [19], and guide RNA (gRNA) encoding vector pgRNA-humanized [19], were purchased from Addgene (plasmid # 47878, 44246, and 44248, respectively). TALMaj and dCas9 DNA fragments were amplified by PCR and ligated into pcDNA3.1 poly (A) vector [32], pTetOne vector (Takara Bio Inc., Japan) or PiggyBac cDNA expression vector (System Biosciences, Inc., CA, USA) together with a DNA fragment encoding SssI CpG methyltransferase. In 2007, we successfully cloned the SssI gene, which has four amber mutations in its open reading frame, and fixed these amber mutations into normal sense codons [33]. An enzymatically inactive SssI, T313D mutant, was prepared using site-direct mutagenesis by replacing the threonine located at position 313 of SssI with aspartic acid [34]. Enzymatically active (WT) and inactive (T313D) SssI were tagged with FLA-G-HA or 3× FLAG at the C-terminus. DNA sequences encoding a gRNA expression unit containing the T7 promoter, BbsI digestion site, dCas9 binding hairpin, including A-T flip and stem extension combined [19], and Streptococcus pyogenes terminator, was synthesized (Thermo Fisher Scientific Inc., MA, USA) and inserted between the BstXI and XhoI sites of a pgRNA-humanized vector. A DNA fragment encoding major satellite targeting gRNA was synthesized with oligonucleotides [20]. Two nucleotide fragments were annealed and ligated into the BbsI digested gRNA expression plasmid. A plasmid encoding mCherry-MBD-NLS, Histone H2B-EGFP and EGFP-CENPC were prepared previously [35]. Dnmt TKO ES cell line stably expressing both mCherry-MBD-NLS and TALMaj-SssI were obtained by introduction of PiggyBac cDNA expression vector and Super PiggyBac Transposase Expression vector (System Biosciences, Inc., CA, USA) with Lipofectamin3000. Antibiotic-resistant clones were selected and used in the study. TALMaj-SssI encoding pTetOne was transfected with Lipofec-tamin3000 (Thermo Fisher Scientific Inc., MA, USA) into Dnmt TKO ES cell and C3H10T1/2 cell. Doxycycline was added to transfected cells at 1 μg/mL concentration to induce TALMaj-SssI expression.

Microinjection of in vitro transcribed RNA into embryos
The gRNA expression plasmid was linearized by XhoI digestion, and gRNA was synthesized with a RiboMAX Large Scale RNA Production System-T7 (Promega Co., WI, USA). RNA encoding TALMaj-SssI (WT or T313D), dCas9-SssI (WT or T313D), mCherry-MBD-NLS, and Histone H2B-EGFP were synthesized with an in vitro transcription kit using linearized plasmids, as described previously [32,33,37]. Cumulus-intact Metaphase-II arrested oocytes were obtained from 8-12-week-old ICR mice and inseminated with capacitated sperm (50 cells/μL) in TYH medium and incubated for 60 min. Fertilized oocytes developed to Anaphase-II or Telophase-II stage were recovered and injected with RNA, and then incubated at 37˚C under 5% CO 2 in air. We expressed TALMaj-SssI at three different levels of expression (low, mid, and high) by injecting various concentrations (2 ng, 10 ng, and 50 ng/μL) of RNA encoding TALMaj-SssI. We also set three different levels of dCas9-SssI expression by injecting 12 ng, 60 ng, and 300 ng/μL of RNA mixed with a same concentration of major satellite targeting gRNA (12 ng, 60 ng, and 300 ng/μL). For imaging global levels of DNA methylation and chromatin, we injected RNA encoding mCherry-MBD-NLS (5 ng/μL) and Histone H2B-EGFP (10 ng/μL) with effector (TALMaj or dCas9)-SssI encoding RNA. This study was carried out in strict accordance with the recommendations in the Guide for the Care and Use of Laboratory Animals of the KINDAI University. The protocol was approved by the Committee on the Ethics of Animal Experiments of the KINDAI University (Permit Number: KABT-27-004). All mice were euthanized by cervical dislocation, and all efforts were made to minimize suffering.

Quantification of imaging data
Imaging data (Z-axis 3D images) obtained from time-lapse analysis were stacked into 2D images for each time-point using MetaMorph software (Molecular Devices LLC., CA, USA). Data were analyzed using MetaMorph software or ImageJ software (http://imagej.nih.gov/ij).
To construct the 3D images, we used Volocity software (PerkinElmer, MA, USA). To acquire the DNA methylation data from the first mitosis movie, multiple embryo images were assembled into one movie with the same dynamic range of fluorescence intensity. The intensity of mCherry-MBD-NLS in mitotic chromosomes was measured in the regions of interest (ROIs). The shapes of the chromosomes in the images were extracted using the threshold setting, and ROIs were placed over them manually. The mean fluorescence intensity was then measured at all times for each mitotic stage. To acquire the heterochromatin index from the 2-cell nucleus images, each round shape of a nucleus was defined as an ROI and extracted using the threshold setting, and then the fluorescent signal of the ROI was measured manually. The heterochromatin index [35] was calculated using the following formula: Heterochromatin index = standard deviation (SD) of the signal intensity / average signal intensity. For detecting abnormal chromosome segregation (ACS), the numbers of embryos showing misaligned chromosomes or lagging chromosomes labeled with H2B-EGFP were counted during the 1-cell to 8-cell stages [38]. Results were obtained from multiple experiments. For statistical analysis, we used Graph-Pad Prism version 6.02 (GraphPad Software, CA, USA). Two-way ANOVA, the Mann-Whitney U test or chi-square test was used for statistical analysis up to the composition of data sets; P<0.05 was considered to be significant.

Bisulfite sequencing
Embryos injected with RNA encoding effector (TALMaj or dCas9)-SssI, mCherry-MBD-NLS, and Histone H2B-EGFP were cultured for 72 h, and embryos developed to the morula/blast stage were recovered for a bisulfite reaction performed with an EZ DNA Methylation-Direct Kit (Zymo Research Co., CA, USA) according to the manufacturer's instructions. Briefly, cells were placed directly into PCR tubes containing lysis buffer and incubated at 50˚C for 20 min, and then subjected to a bisulfite reaction. After bisulfite treatment, eluted DNA was amplified by PCR using the following sets of primers: 5 0 -GGAATATGGTAAGAAAATTGAAAATTATGG-3 0 and 5 0 -CCATATTCCAAATCCTTCAATATACATTTC-3 0 for the major satellite; 5 0 -TAGAATATATTAGATGAGTGAGTTATATTG-3 0 and 5 0 -ATTATAACTCATTAATATACACTATTCTAC-3 0 for the minor satellite; 5 0 -TTGATAGTTGTGTTTTAAGTGGTAAATAAA-3 0 and 5 0 -AAAACACCACAAACCAAAATCTTCTAC-3 0 for the IAP; 5 0 -TAGGAAATTAGTTTGAATAGGTGAGAGGT-3 0 and 5 0 -TCAAACACTATATTACTTTAACAATTCCCA-3 0 for the Line1. The PCR reaction for amplifying major and minor satellites was 40 cycles of 95˚C for 30 s, 60˚C for 60 s, and 68˚C for 20 s. PCR conditions for amplifying IAP and the Line1 fragment were 35 cycles of 95˚C for 60 s, 60˚C for 60 s, 72˚C for 60 s, followed by 72˚C for 3 min. The amplified DNA fragments were subcloned into a pGEM-T Easy vector (Promega Co., WI, USA), and 10-20 clones derived from each group of bisulfite-treated genomic DNA were sequenced. Methylation sites were visualized using the web-based DNA methylation analysis tool "QUMA" (http://quma.cdb.riken.jp/) [39]. Statistical significance was quantified using the Mann-Whitney U test performed using QUMA.

Design and validation of major satellite-targeted DNA methyltransferase SssI, TALMaj-SssI
We designed fusion proteins composed of a TALE recognizing major satellites [30] and enzymatically active (WT) or inactive (T313D) SssI CpG methyltransferase (Fig 1A and 1B). We first confirmed the localization of TALMaj-SssI in cultured cells. Plasmids encoding 3×FLAGtagged TALMaj-SssI and EGFP-tagged CENPC were introduced into C3H10T1/2 cells. TAL-Maj-SssI was immunostained with anti-FLAG antibody, and 3D images were reconstructed. TALMaj-SssI was detectable in both the nucleus and cytoplasm. In the nucleus, DAPI-dense signals were well stained with anti-FLAG antibody, and CENPC signals were localized around these heterochromatin foci (Fig 1C). Although there was slight overlap of the TALMaj-SssI and CENPC signals, which indicated the kinetochore region comprising the minor satellite repeat (arrowheads in panel), TALMaj-SssI was enriched mainly in pericentromeres which are DAPI-dense regions comprising major satellite repeats [40].
Next, we introduced TALMaj-SssI into Dnmt (Dnmt1, Dnmt3a, and Dnmt3b) triple knock out ES cell (Dnmt TKO ES cell), which has global hypomethylation including major satellite and minor satellite and identify upregulation of DNA methylation in microscopic level [31] ( Fig 1D). To visualize the DNA methylation and centromeric regions, we co-transfected mCherry-MBD-NLS [35] and EGFP-CENPC. Accumulation of MBD signals was observed in DAPI-dense heterochromatin in TALMaj-SssI (WT)-expressing Dnmt TKO ES cells, whereas cells expressing the mutant (T313D) or mock control exhibited a uniform distribution of MBD signals in the nuclei except for the nucleoli. CENPC signals were localized around DAPI-dense heterochromatin. This result suggests that TALMaj-SssI expression upregulated pericentromeric DNA methylation.
We next used ChIP-qPCR analysis to confirm the binding specificity of TALMaj-SssI ( Fig  1E). Dnmt TKO ES cells stably expressing TALMaj-SssI (T313D) under the control of the EF1α promoter were established and analyzed. We found that TALMaj-SssI localized both to the major satellite and Line1 retrotransposon, but not to other repeats. This suggests that TAL-Maj-SssI localizes to the major satellite preferentially, but has some off-target binding under this experimental condition. Although the reason for the off-target binding of TALMaj-SssI is unclear, it is possible that it was caused by abundant expression. Moreover, Line1 has much more copy number compared with other repeat element we tested. These might be reasons why TALMaj-SssI shows off-target binding to Line1. Because TALMaj-SssI had some off-targets toward the Line1 retrotransposon in our ChIP-qPCR analysis (Fig 1E), we next used bisulfite sequencing to evaluate the DNA methylation of the major and minor satellites under the dose-dependent expression of TALMaj-SssI in mouse preimplantation embryos that were known to have hypomethylated centromeres and pericentromeres based on previous analysis [10]. DNA methylation of the major satellites was upregulated dose dependently in embryos expressing WT TALMaj-SssI, whereas few effects were observed with minor satellite methylation (Fig 2A). We evaluated Line1 and IAP DNA methylation further and found few effects in these sequences ( Fig 2B). Therefore, we conclude that TALMaj-SssI upregulates DNA methylation preferentially in the major satellites, but exerts little off-target DNA methylation toward other genomic regions.
We next capture the dynamics of DNA methylation induction, DNA methylation status and patterns in the nucleus were visualized using mCherry-MBD-NLS [35]. RNAs encoding TALMaj-SssI, mCherry-MBD-NLS, and Histone H2B-EGFP were co-injected into mouse fertilized embryos, and the fluorescent signals of mCherry and EGFP were observed during the 1-cell to morula/blast developmental stage. At the first mitotic stage, mCherry-MBD-NLS were strongly accumulated at the distal ends of chromosomes in embryos expressing WT TAL-Maj-SssI, but not in the embryos injected with the T313D mutant (Fig 3A and 3B and S1 and S2 Movies). Time-lapse analysis revealed that the induction of DNA methylation by TALMaj-SssI occurred mainly in late metaphase stages during the first mitosis (Fig 3B and 3C). It is unclear why this DNA methylation occurs strictly with this specific mitotic stage. In 2-cell nuclei, MBD signals also accumulated at the heterochromatin foci locating on the peripheral part of the nucleus (Fig 3D and S1 Movie). We next calculated the heterochromatin index with the MBD signals. The heterochromatin index is defined as the coefficient of variation of signal patterns in the nucleus [35]. The score fluctuates with fluorescent accumulation in heterochromatin or numbers of heterochromatin foci in the nucleus. Embryos expressing WT TALMaj-SssI indicated a higher heterochromatin index score (Fig 3E). This result suggested that heterochromatin organization is influenced by the upregulation of DNA methylation.
Taken together, these results suggested that DNA methylation was upregulated in pericentromeres of embryos expressing the WT TALMaj-SssI, and that this activity was dependent on the enzymatic activity of SssI. Moreover, this induction of DNA methylation was as drastic as it is able to capture by fluorescent microscope.
Although there was leakage of DNA methylation to minor satellites when an excess amount of TALMaj-SssI was expressed, moderate TALMaj-SssI expression ensured target DNA methylation to major satellites without mis/upregulation of DNA methylation in minor satellites (Fig 2A). Furthermore, to evaluate whether TALMaj-SssI expression had some effect on minor satellites or kinetochores under these experimental conditions we analyzed chromosome segregation errors with embryos expressing moderate levels of WT or T313D TALMaj-SssI (Fig 4). Chromosomes labeled with H2B-EGFP were monitored using live-cell imaging from the 1-cell  Induction of DNA methylation with TAL-SssI and dCas9-SssI to 8-cell stages and the numbers of embryos with misaligned or lagging chromosomes were counted (Fig 4A). This showed that there were no significant differences in ACS between the embryos expressing TALMaj-SssI WT or T313D (Fig 4B). We conclude that targeted DNA methylation to major satellites did not interfere with kinetochore function during early embryonic cleavage stages. Introduction of pericentromere DNA methylation with CRISPR/Cas9 system Compared with the TALE technology, it is easier to design and construct targeting vectors with the CRISPR/Cas9 system. We expressed the fusion protein dCas9-SssI with gRNA recognizing major satellite sequences [20] in preimplantation embryos. In 2-cell stage embryos, MBD signals accumulating at pericentromeric heterochromatin were slightly higher in embryos expressing WT dCas9-SssI compared with the T313D and gRNA combination ( Fig  5A). The heterochromatin index during the 2-cell stage was upregulated in the WT dCas9-SssI expressing embryos, but not in T313D expressing embryos (Fig 5B). Induction of DNA methylation with TAL-SssI and dCas9-SssI Finally, we evaluated the specificity of DNA methylation with bisulfite sequencing (Fig 6). Although unexpected DNA methylation was also observed in embryos expressing a high dose of the T313D mutant, dose and gRNA dependent upregulation of major satellite DNA methylation was demonstrated by WT dCas9-SssI expression (Fig 6A). Line1 DNA methylation was higher than that of TALMaj-SssI expressed embryos (Figs 2B and 6B). As these results demonstrate, fewer off-target effects were shown with the TALE-based system compared with CRISPR/Cas9 when SssI methyltransferase was used as the effector domain of epigenome editing enzymes.

Discussion
We demonstrated that it is possible to use the TALE or CRISPR/Cas9 systems to induce DNA methylation in mouse preimplantation embryos with the bacterial CpG methyltransferase, SssI. Successful induction of DNA methylation was performed in major satellite repeats located in mouse pericentromeres with the epigenome editing enzymes TALMaj-SssI or dCas9-SssI. Recently, there have been reports of inducing DNA methylation by TALE-Dnmt3a or dCas9-Dnmt3a in single copy sequences such as Snrpn, p16, IL6ST, or BACH2 promoters [22,25,26]. This is the first report where DNA methylation has been induced in tandem copy sequence, and the upregulated DNA methylation was detectable at the fluorescent microscopic level with a mCherry-MBD-NLS probe. It is known that Dnmt3L stimulates the catalytic activity of Dnmt3a [27,28], and it is possible that the methylation activity of Dnmt3a is dependent on the amount of Dnmt3L in host cells. By contrast, SssI does not require coupling factor to induce DNA methylation. This is the advantage of using SssI as a methylation effector in various types of cells including DNMT3L downregulated cells such as teratoma or some of the embryonal carcinoma-derived cell lines [41,42].
In this study, embryos expressing dCas9-SssI with gRNA induced DNA methylation in major satellites, target DNA methylation was reduced when dCas9-SssI was introduced in Induction of DNA methylation with TAL-SssI and dCas9-SssI embryos without gRNA (Fig 6A). This suggests that gRNA has an important function in tethering (d) Cas9 (fusion) proteins to the desired genomic region. This gRNA dependent DNA methylation was also observed in the dCas9-Dnmt3a system [25,26].
Reducing the off-target effects of (epi) genome editing technology to increase the quality of experiments is a subject often discussed. With the dCas9-SssI and gRNA combination, off-target DNA methylation was induced in Line1 sequences when compared with TALE-SssI (Figs 2  and 6). These results indicate that TALE-SssI has the advantage of fewer off-target effects compared with dCas9-SssI. Optimizing the gRNA design, expression ratio of dCas9-SssI and gRNA, and moreover, applying split enzyme methods [43] and optimization of SssI amino acid residues [44], may reduce off-target effects, and these are required in our CRISPR/Cas9 system. In contrast to our study, Liu et al. reported that the CRISPR-based system has advantages over the TALE-based system in terms of editing DNA methylation [26]. They used Dnmt3a to induce DNA methylation, whereas we applied SssI as the effector. This difference Low, mid and high expression of dCas9-SssI was achieved by injection of various concentration (12ng, 60ng and 300ng/μL) of RNA. Embryos were recovered 3 days after RNA injection. In high expression condition, Embryos expressing dCas9-SssI without gRNA were also evaluated. Asterisk indicates significant difference between low and other test groups by Mann-Whitney U test. (B) Bisulfite sequence of IAP and Line1 CpGs of dCas9-SssI expressing and major satellite gRNA introduced embryos. Mid-level expression of dCas9-SssI was achieved by injection of 60ng/μL of RNA. Data are represented as % of methylated CpG sites per total CpG sites. Asterisk indicates significant difference between WT and T313D by Mann-Whitney U test. https://doi.org/10.1371/journal.pone.0177764.g006 Induction of DNA methylation with TAL-SssI and dCas9-SssI between studies suggests that the efficiency of editing the epigenome using a TALE-or CRISPR-based system depends of the characteristics of fused enzyme, and that it is important to assess the target specificity of the designed enzyme. By confirming the target specificity of these enzymes, whole-genome bisulfite sequence (WGBS) is one of the best ways to assess the target specificity of DNA methylation clearly and quantitatively [45]. TALMaj-SssI bound the major satellite and Line1 retrotransposon in our ChIP-qPCR analysis (Fig 1E). This result was unexpected because TALMaj-Ty1 specifically bound the major satellite sequence in a previous ChIP-qPCR analysis [30]. We consider a possibility to explain this off-target binding of TAL-Maj-SssI. High expression level may induce off-target localization of TALMaj-SssI in chromatin. We used the Ef1α promoter to drive the TALMaj-SssI gene stably in Dnmt TKO ES cells. This may have been sufficient for the expression of the TALMaj-SssI gene, but too strong to drive TALMaj-SssI in cultured cells. Therefore, we evaluated DNA methylation in embryos expressing TALMaj-SssI and dCas9-SssI at various levels (Figs 2 and 6). Bisulfite sequence analysis indicated that DNA methylation was upregulated in the major satellite in TALMaj-SssI-expressing embryos, whereas Line1 DNA methylation was not when it was expressed with moderate level (10ng/μL). This result suggests that optimization of the expression level of the epigenome editing enzyme is important for reducing the off-target effect in cells. Chromosome segregation analysis also indicated that there were no significant differences in ACS frequency between the WT and T313D forms of TALMaj-SssI-expressing embryos (Fig 4). This implies that DNA methylation in pericentromeres has few effects on kinetochore or chromosome segregation during preimplantation development. Because ACS occurs in 54.1% of IVF embryos [38], the frequencies of ACS seen here (38% in the WT and 25% in the T313D forms) appears to be normal. These data also suggest that little if any leakage to minor satellites occurs when appropriate amounts of the enzyme are expressed.
In this study, the heterochromatin index was upregulated in TALMaj-SssI expressing embryos (Fig 3). This suggests that DNA methylation in pericentromeres affects heterochromatin formation in the 2-cell embryonic stage. Because loss of DNA methylation brings a slight reduction of H3K9me3 and H4K20me3 and increment of H3K27me3 in major satellites [46], it is possible that induction of DNA methylation in major satellites might lead to some repressive histone modifications such as H3K9me3 and H4K20me3 and affect chromatin configuration. There are several reports that centromeres or pericentromeres have roles in biological functions such as forming functional kinetochores [47] and genome stability in cancerous cells [48][49][50][51]. In gametogenesis, global hypomethylation, including imprinting control regions, centromeres, and pericentromeres, is observed in 13.5 dpc PGCs, whereas 10.5 dpc PGCs reveal hypermethylation in these regions [10,52]. This demethylation, which is ensured by both passive DNA demethylation [52,53] and replication independent DNA demethylation [54], has roles in resetting somatic memory such as genomic imprinting and establishing germ-cell specific epigenomes during gametogenesis. We focused on DNA methylation in centromeres and pericentromeres. Currently, few studies have investigated these functions because of difficulties in manipulating locus-specific DNA methylation. Our study may lead to advances in ensuring locus-specific induction of DNA methylation and may reveal the consequence and biological function of the DNA methylation status in pericentromeres in future studies.