Repurposing the CRISPR-Cas9 system for targeted DNA methylation

Plasmid construction

Nickase activity of the plasmid pSpCas9n(BB)-2A-Puro (PX462) (Addgene plasmid # 48141, a gift from Feng Zhang) (2) encoding Cas9n D10A was abolished by introducing the H840A mutation into the Cas9 gene using the QuikChange Lightning Site-Directed Mutagenesis Kit (Agilent Technologies, Santa Clara, CA, USA). A short Gly₄Ser linker introducing the unique BamHI and FseI restriction sites was inserted at the C-terminus of the Cas9 gene.

Undesired BbsI restriction site within the catalytic domain of human DNMT3A from the plasmid pcDNA3/Myc-DNMT3A (Addgene plasmid # 35521, a gift from Arthur Riggs) (28) was removed by site-directed mutagenesis. The DNMT3A catalytic domain (amino acids P602-V912) was then amplified using the Herculase II Fusion DNA Polymerase (Agilent Technologies) with primers designed to enable insertion C–terminally from the Gly₄Ser linker, between the BamHI and FseI restriction sites. Catalytically inactive DNMT3A was generated by site-directed mutagenesis of the active site motif ENV to ANV (E756A) (28). EcoRI fragments encoding T2A-Puro and T2A-EGFP were cut from plasmids pSpCas9n(BB)-2A-Puro (PX462) (Addgene plasmid # 48141) and pSpCas9n(BB)-2A-GFP (PX461) (Addgene plasmid # 48140, a gift from Feng Zhang) (2), respectively and cloned at the C–terminus of the dCas9-DNMT3A fusion to allow for either puromycin selection or EGFP screening of transfected cells.

Guide sequences targeting the loci of interest were selected either manually or using the GT-Scan web-tool (http://gt-scan.braembl.org.au/gt-scan/) (35). Sequences of the non-targeting (NT) control guide RNAs, without a target in the human genome, were taken from the Human GeCKOv2 Library (36). All sgRNA genes (Supplementary Table S1) were synthesized by oligo annealing and cloned into expression vectors via BbsI sites essentially as described in (25) using BpiI/BbsI (Thermo Scientific, Waltham, Massachusetts, USA) and T4 DNA ligase (Takara Biotechnology, Dalian, China). Additionally, BACH2-sgRNA8 was cloned into H840A-mutated pSpCas9n(BB)-2A-Puro vector, expressing dCas9 without the DNMT3A catalytic domain, in order to analyze the sole effect of sgRNA-dCas9 binding on CpG methylation. The cloned guide RNAs were verified by sequencing using the U6–Fwd primer. Sequences of all oligonucleotides are listed in Supplementary Table S2.

Cell transfection

Human embryonic kidney cell line HEK293 was maintained in Dulbecco's Modified Eagle Medium (Lonza, Verviers, Belgium) supplemented with 10% heat inactivated fetal bovine serum (Biosera, Ringmer, UK), 4 mM L–glutamine (Lonza), 100 U/ml penicillin and 100 μg/ml streptomycin (Lonza). Cells were incubated at 37°C in a humidified 5% CO₂–containing atmosphere.

Cells were seeded into a 24–well plate and transfected the next day at 70–90% confluence using Lipofectamine 3000 (Life Technologies, Carlsbad, California, USA) according to the manufacturer's protocol. Transfections were done with 100 ng of plasmids co–expressing dCas9-DNMT3A and a chimeric sgRNA, in biological triplicates. Cells transfected with dCas9-DNMT3A-EGFP plasmid were fixed 24 h after transfection and the percentage of cells expressing GFP was assessed by fluorescent microscopy, to ensure consistent level of transfection efficiency in all experiments. One day after transfection, cells were passaged to 6–well plates and selected for 48 h with 1.5 μg/ml puromycin. Eight days after transfection (or at other defined time points for the time-course experiment), cells were harvested and incubated at 37°C overnight in lysis buffer (50 mM Tris pH 8.5, 1 mM EDTA, 0.5% Tween 20) with proteinase K (400 μg/ml) yielding crude cell lysates.

Analysis of DNA methylation

DNA from crude cell lysates was bisulfite converted and purified using EZ DNA Methylation-Gold Kit (Zymo Research Europe, Freiburg, Germany) according to the manufacturer's protocol. Genomic fragments encompassing sgRNA target sites were amplified from bisulfite converted DNA using the PyroMark PCR kit (Qiagen, Hilden, Germany) according to the manufacturer's instructions. Methylation of repetitive LINE-1 elements was analyzed as described previously (37). The cycling was performed as follows: initial denaturation for 15 min at 95°C; 50 cycles of 30 s at 95°C, 30 s at 48°C (BACH2–A and BACH2–B fragments), 52°C (IL6ST–A fragment) or 64°C (LINE-1) and 30 s at 72°C; final extension for 10 min at 72°C. In order to quantify methylation levels at individual consecutive CpG sites, PCR amplicons were sequenced using the PyroMark Q24 Advanced pyrosequencing system (Qiagen). Sequences of PCR primers and pyrosequencing primers are listed in Supplementary Table S2.

Dose-response experiment

Titration of dCas9-DNMT3A plasmid in HEK293 cells was performed in 24-well plates using 500, 250, 100 or 25 ng of plasmid DNA. Cells were transfected with the targeting BACH2-sgRNA3 constructs (wild-type dCas9-DNMT3A or ANV-mutant) or with two different non-targeting sgRNA constructs. After selection, CpG methylation in the BACH2-B gene fragment was assayed as described above.

Targeted promoter methylation and expression analysis

The promoter regions of the IL6ST and BACH2 genes were targeted using either individual or pooled sgRNAs co-expressed from dCas9-DNMT3A-PuroR plasmids. Eight different sgRNAs were used to target the BACH2 promoter, and four sgRNAs were used to target the IL6ST promoter (Supplementary Table S1). Additionally, the BACH2 promoter was targeted with the sgRNA8 co-expressed with dCas9 protein, in order to investigate if the sole binding of sgRNA-dCas9, in the absence of the DNMT3A domain, affects CpG methylation level.

Cells were transfected in 24-well plates using in total 100 ng of plasmid DNA per transfection, selected and harvested as described above. Experimental groups were cells transfected with either active or inactive constructs co–expressing targeting sgRNAs, while the control groups were mock-transfected cells (without plasmid DNA) and cells co-expressing the active dCas9-DNMT3A and the non-targeting sgRNA. Transfection was done in biological triplicates. Total RNA was isolated at the same time as DNA using RNeasy Mini Kit (Qiagen), according to the manufacturer's instructions. One microgram of RNA was used as a template for reverse transcription using the PrimeScript RTase (Takara) and random hexamer primers (Invitrogen, Paisley, UK). Quantitative real-time PCR (RT–qPCR) was performed on cDNA from cells transfected with pooled sgRNAs co–expressing either the active or the inactive construct and both controls (mock transfection and active non–targeting sgRNA). For RT–qPCR, we used hydrolysis probes in the 7500 Fast Real-Time PCR system (Applied Biosystems, Carlsbad, California, USA) in technical triplicates with TaqMan Gene Expression Master Mix and the following TaqMan Gene Expression Assays: Hs00174360_m1 (IL6ST), Hs00222364_m1 (BACH2) and Hs02800695_m1 (HPRT1) (Applied Biosystems). The expression of the target genes BACH2 and IL6ST was normalized to the reference gene HPRT1 and analyzed using the comparative C_q method (38). Expression was shown as fold change (FC) relative to mock-transfected cells.

Time course experiment

HEK293 cells were transfected in triplicates in 6-well plates with 500 ng of dCas9-DNMT3A-PuroR plasmid encoding BACH2-sgRNA8 or IL6ST-sgRNA3, or their DNMT3A-mutated counterparts. One day after transfection, triplicates were pooled and re-plated in 6-well plates. Cells were selected as described above. Aliquots of the transfected cells were collected daily until the 9th day and after that less frequently until the 42nd day. Those samples were used for CpG methylation analysis by pyrosequencing as well as for detection of plasmid DNA by qPCR. Methylation of the BACH2–A and IL6ST–A fragments was assayed as described above.

The experiment was repeated with two sets of samples taken on days 1, 3, 6, 7, 10 and 15 after transfection when cells were fixed on microscopy slides. The samples were used for direct detection of plasmid DNA in cells by fluorescent in situ hybridization (FISH) and for detection of Cas9 expression by immunofluorescence (IF).

Relative quantification of plasmid DNA by qPCR

qRT-PCR was performed on crude cell lysates from cells transfected in the time course experiment. We analyzed samples transfected with plasmids encoding BACH2-sgRNA8 (active or inactive construct), collected on days 1, 2, 3, 4, 7, 9 and 16 after transfection, and mock transfected cells.

The amount of plasmid encoding dCas9-DNMT3A was determined relative to the genomic RAG1 sequence, amplified using a pair of previously published primers (39). Sequences of qPCR primers are listed in Supplementary Table S2. Reactions were run in technical duplicates using Power SYBR Green PCR Master Mix (Applied Biosystems) and 50 nM primers for Cas9 or RAG1. To assess primer efficiency we prepared standard curves of known amounts of the linearized plasmid and purified genomic DNA isolated from HEK293 cells. Cycling conditions were 10 min at 95°C followed by 40 cycles at 95°C for 15 s and 62°C for 1 min, followed by melt curve analysis of amplified targets. PCR was performed in 7500 Fast Real-Time PCR System (Applied Biosystems) using standard method. The amount of plasmid DNA relative to genomic DNA was calculated essentially as described in (40).

EGFP fluorescence analysis

HEK293 cells were transfected with 500 ng of the plasmid co–expressing Cas9-DNMT3A-EGFP fusion with different guide RNAs targeting the BACH2 (sgRNA8) or IL6ST (sgRNA3) promoter regions. Constructs either co–expressing non–targeting guide RNA or having inactive DNMT3A domain were used as controls.

Images were acquired under the same conditions using Zeiss Axiovert 40CFL microscope with AxioCam MRm in the AxioVision software (Carl Zeiss Microscopy, LLC, USA) at specified time points (1, 4, 6, 7 and 10 days after transfection). EGFP fluorescence, number of EGFP positive cells per field and total number of cells per field were quantified using the ImageJ software (41).

Immunofluorescence

At each time point, expression of dCas9-DNMT3A protein was analyzed in cells transfected with BACH2-sgRNA8, BACH2-sgRNA8 ANV and mock-transfected cells. Fixed cells were washed with PBS, incubated with primary anti-tag antibody (MA1–91878, FLAG™ Epitope Tag (DYKDDDDK) Monoclonal Antibody, 1:5000, Thermo Fisher Scientific, Waltham, MA, USA) for 90 min and afterward with secondary antibody labeled with FITC (Goat polyclonal Secondary Antibody to Mouse IgG – H&L (FITC), 1:200, Abcam, Cambridge, UK) for 60 min. Cells were mounted using anti-fade Fluorescence Mounting Medium (Dako, Glostrup, Denmark) containing 8 μg/ml DAPI. Results were analyzed using Olympus BX51 microscope. Fluorescence intensities were measured using ImageJ software (41) and percentage of cells expressing Cas9 protein per sample was calculated.

Fluorescent in situ hybridization

The Cy3-labeled probe for the plasmid encoding dCas9-DNMT3A was prepared using Nick Translation Kit (Roche, Indianapolis, Indiana, USA) and subsequently purified with the ChargeSwitch PCR Clean-Up Kit (Invitrogen, Paisley, UK) according to the manufacturer's protocols. For detection of genomic DNA, the centromeric chromosome 7 enumeration probe was used (Vysis CEP7 SpectrumGreen; Abbott Molecular, Des Plaines, IL, USA). At each time point, mock-transfected cells and cells transfected with BACH2-sgRNA8 and BACH2-sgRNA8 ANV were analyzed. Chromosome slides containing the mixture of probes specific for the plasmid and the chromosome 7 were denatured together at 85°C for 5 min and hybridized at 37°C overnight. The next day, slides were washed three times in preheated 0.1x saline sodium citrate (SSC) buffer at 65°C and mounted with anti-fade Fluorescence Mounting Medium (Dako) containing 8 μg/ml DAPI. The images were acquired using Olympus BX51 microscope. Number of signals per cell representing plasmid DNA encoding dCas9-DNMT3A was determined on each slide and used as approximate value for evaluation of plasmid DNA stability.

Western blot analysis of total protein extracts

HEK293 cells were co–transfected with 100 ng of plasmid expressing BACH2-sgRNA8 constructs with puromycin selection marker (wild-type dCas9-DNMT3A or ANV-mutant) and 20 ng of plasmid expressing EGFP (pcDNA3-EGFP, Addgene plasmid # 13031, a gift from Doug Golenbock). The cells were selected with puromycin as described above and collected 1, 3, 7, 11 and 15 days following transfection and frozen in liquid nitrogen. Cell lysates were homogenized in RIPA buffer containing 150 mM NaCl, 1% NP–40, 0.5% sodium deoxycholate, 0.1% SDS, 50 mM Tris–HCl pH 8 and 2x cOmplete™ protease inhibitor cocktail (Roche, Basel, Switzerland). Subsequently, 10 μg of protein extract was run on 10% SDS-PAGE.

Proteins were transferred onto a nitrocellulose membrane for Western blot analysis. The membrane was cut into pieces that were handled separately. To detect Cas9, the membrane was blocked overnight in 5% milk-PBS (5% nonfat dry milk in 1x phosphate buffer saline: 137 mM NaCl, 2.7 mM KCl, 10 mM Na₂HPO₄, 2 mM KH₂PO₄, pH 7.4), then incubated overnight with anti-Cas9 antibody 1:2000 (ab204448; Abcam) in 5% milk-PBS containing 0.2% Tween 20, followed by incubation with anti-rabbit HRP-conjugated secondary antibody 1:2000 (Dako). Membranes used to detect tubulin (using the E7 antibody, 1:1000, Developmental Studies Hybridoma Bank, Iowa City, IA, USA; deposited by prof. Michael Klymkowsky, University of Colorado, Boulder, CO, USA) and EGFP (using the sc–9996 antibody, 1:1000, Santa Cruz Biotechnology, Dallas, TX, USA) were blocked for one hour at room temperature in 5% BSA–PBS, then incubated overnight with corresponding antibodies in 5% BSA–PBS (5% bovine serum albumin in 1x PBS) containing 0.2% Tween 20, followed by incubation with anti-mouse HRP-conjugated secondary antibody 1:3000 (Dako). The antibody-tagged protein bands were visualized by enhanced chemiluminescence (ECL) system (Pierce Biotechnology, Waltham, MA, USA) according to manufacturer's protocol. ECL signals captured on film were quantified using the ImageJ software (41).

Construction of the dCas9–DNMT3A fusion protein

The dCas9–DNMT3A fusion protein was constructed by addition of DNMT3A catalytic domain to the C–terminus of the inactive Cas9 (dCas9) via a short Gly₄Ser linker. The predicted structure of the fusion product based on the Cas9 crystal structure (42) with the DNMT3A catalytic domain protruding at the PAM–interacting (PI) domain side of Cas9 is depicted in Figure 1A. Four types of the construct were made: a construct with either active or inactive DNMT3A catalytic domain, each fused to either puromycin resistance or EGFP gene, as depicted in Figure 1B.

The final dCas9–DNMT3A expression plasmids contained sgRNA expression cassettes compatible with the oligo cloning protocol using BpiI/BbsI (25). The plasmids are deposited at Addgene (plasmid numbers #71667, #71684, #71666, #71685). The amino acid sequences of constructed fusion proteins are shown in Supplementary Figure S1.

The dCas9-DNMT3A construct induced targeted CpG methylation at the BACH2 locus

We modified the CRISPR-Cas9 tool previously developed for genome editing (2) to be used for targeted DNA methylation of specific loci in mammalian cells. We fused the DNMT3A catalytic domain to the C-terminus of dCas9 via a short flexible linker to enable methylation of DNA sequences adjacent to the sgRNA binding site. In order to determine the optimal amount of dCas9-DNMT3A-encoding plasmid for HEK293 transfection, we performed a dose-response experiment. The plasmid amount eliciting a saturation level of on-target activity and acceptable non-specific activity arising from DNA methyltransferase overexpression (100 ng for transfection in a 24–well plate or 500 ng for a 6–well plate) was used in all subsequent experiments (Supplementary Figure S2).

Two genomic fragments within the BACH2 promoter, unmethylated in HEK293 cell line, were targeted with several different sgRNAs (Figure 2), which bind in different orientations and at distinct positions relative to the 13 or 14 CpG sites covered either by the BACH2-A or the BACH2-B pyrosequencing assay, respectively. All five sgRNAs targeting the intronic BACH2-B fragment induced a strong increase of CpG methylation relative to the mock-transfected cells (Figure 3A, Supplementary Figure S3). Among them, sgRNA3 showed the highest activity, increasing the methylation level of certain CpG sites upstream and downstream of its binding site up to 34%. Two out of the three sgRNAs targeting the upstream BACH2-A fragment induced similar or even higher increase of the methylation level relative to the mock-transfected cells (Supplementary Figure S3). Methylation level of CpG sites located downstream from the PAM sequence was increased up to 33% (sgRNA7) or 59% (sgRNA8). Methylation level of CpGs within the sgRNA binding sites remained unchanged. Non-targeting dCas9-DNMT3A and targeted inactive constructs (dCas9-DNMT3A-ANV), used as negative controls, did not induce any significant change in methylation level, confirming that both sgRNA-mediated binding and intact methyltransferase activity of dCas9-DNMT3A are required for specific DNA methylation of targeted loci. We observed only a slight increase in CpG methylation level with the catalytically inactive BACH2-sgRNA8 construct (up to ∼10% with the inactive versus ∼55% with the active construct; Figure 3A left panel, Supplementary Figure S3B) mirroring the activity pattern of the active construct, there was no increase in methylation when we targeted the dCas9-only construct (i.e. without the DNMT3A domain) using the BACH2-sgRNA8, thus ruling out any source of CpG methylation activity other than the DNMT3A catalytic domain (Supplementary Figure S3C).

The dCas9–DNMT3A construct specifically methylates a ∼35 bp region centered at 27 bp downstream from the PAM sequence

Various individual sgRNAs targeted the dCas9–DNMT3A construct to different CpG sites within the promoter regions of the BACH2 and IL6ST genes. Changes in CpG methylation level in cells transfected with the dCas9-DNMT3A/sgRNA-expressing plasmids were assessed by pyrosequencing of the targeted regions. Targeting the construct to different positions and in different orientations relative to the pyrosequencing assays (and the CpG sites covered by those assays, Figure 2) enabled us to characterize the summary DNA methyltransferase activity profile of the dCas9–DNMT3A construct (Figure 3). We used results from multiple experiments with different combinations of binding sites and the ‘windows’ where methylation levels were assessed, thus covering regions about 100 bp upstream and downstream from the binding site in an overlapping and redundant manner. The genomic regions were initially unmethylated and no significant increase in methylation level was detected with either of the two negative controls, represented by non–targeting sgRNA combined with the active dCas9–DNMT3A or by sequence-specific sgRNAs with inactive dCas9–DNMT3A–ANV constructs (Figure 3A, Supplementary Figure S3). The experiments using non–targeting sgRNA showed a non–significant increase in methylation level of 1.71 ± 3.28% compared to mock-transfected cells. The summary of the CpG methylation level increase for all sgRNAs aligned relative to their binding sites and orientations (Figure 3B) showed peak activity of about 35% on average, reaching up to 60% for some individual targeted CpG sites. We consistently found a 25–35 bp wide peak of methylation activity centered at 27 bp downstream from the protospacer adjacent motif (PAM) sequence. A much lower peak of methylation activity could be detected approximately at the same distance upstream from the sgRNA binding site. Inactive (dCas9–DNMT3A–ANV) constructs did not induce any statistically significant differences in CpG methylation levels compared to mock–transfected cells (Student's t–test, α = 0.95 corrected for multiple testing), although a slight increase paralleling the peak of activity for the active constructs can be observed on the graph (Figure 3B, dotted curve).

Relative positions and orientations of individual sgRNAs in reference to the pyrosequencing assays within the BACH2 and IL6ST genes are depicted in Figure 2. The profile did not critically depend on any single experiment or a particular sgRNA, which we validated by omitting each single sgRNA in turn and constructing the activity profiles, which were similar to each other and showed only a minimal deviation from the summary profile (Supplementary Figure S4).

Targeted CpG methylation of the IL6ST and BACH2 promoters using the dCas9-DNMT3A tool reduces gene expression

The locus IL6ST, located on a chromosome different from the one containing the BACH2 locus, was targeted with dCas9-DNMT3A construct in HEK293 cells where the IL6ST gene promoter is unmethylated. When cells were transfected with four different sgRNAs independently, the methylation level at the targeted CpG sites in the IL6ST promoter region was elevated up to 35% as confirmed by pyrosequencing of this region using the assay IL6ST–A (Figure 4A). Co–transfection of pooled sgRNAs 1–4 resulted in markedly larger increase in methylation level compared to the individual sgRNAs reaching up to 55% at some CpG sites. Inactive constructs (used either individually or pooled) had no detectable impact on CpG methylation (Figure 4A). Pooled sgRNAs 1–4 targeting the IL6ST promoter region paired with the active dCas9–DNMT3A construct elicited statistically significant decrease in IL6ST mRNA level compared to mock-transfected cells, as detected by qPCR (Student's t–test, P = 0.028, Figure 4B). No statistically significant decrease in mRNA level was detected with pooled sgRNAs 1–4 paired with the inactive construct (P = 0.209). Similarly, active constructs with non–targeting sgRNA did not decrease IL6ST gene expression (P = 0.230).

Similar results were obtained when we targeted the BACH2 promoter with pooled sgRNAs 1–8. As outlined in Figure 2A, sgRNAs 6–8 bind in the region covered by the pyrosequencing assay BACH2–A, while sgRNAs 1–5 bind in the region covered by the BACH2–B pyrosequencing assay. Pooled sgRNAs showed increased total CpG methylation activity (up to 64% at some CpG sites) in both assayed regions when compared with individual sgRNAs (Figure 5A and B). Catalytically inactive and non-targeting constructs did not show any increase in methylation levels. When we tested expression level of the BACH2 gene following transfection with pooled sgRNAs, we found a statistically significant decrease in expression level (compared to mock-transfected cells) with the active construct (P = 0.018) and some decrease in expression level (still statistically significant, P = 0.032) with the inactive construct (Figure 5C). No statistically significant change in expression level was observed with the non-targeting control (P = 0.217, Figure 5C).

Methylation level of LINE-1 elements, which was used for estimation of the global genome methylation, showed no change when we transfected cells with either individual or pooled sgRNAs (Figure 6), indicating absence of unspecific genome-wide CpG methylation activity.

Methylation level at the targeted loci peaks at 25–55% on days 6 or 7 in transfected HEK293 cells

Changes in methylation level were followed during 42 days after transfection of HEK293 cells with the dCas9-DNMT3A construct. The profile of the methylation activity is summarized in Figure 7A. The initial increase in methylation level was rapid, reaching its maximum at days 7 (for the BACH2 gene, Figure 7A, left panel) or 6 (for the IL6ST gene, Figure 7A, right panel), then slowly decreasing in the following 10 to 15 days. In the days 20 to 42 after the transfection with the dCas9-DNMT3A construct, a low basically unchanging level of increased methylation could be detected.

Number of HEK293 cells containing the dCas9-DNMT3A construct decreased within 10 days after transfection