Repurposing the CRISPR-Cas9 system for targeted DNA methylation

Abstract

Epigenetic studies relied so far on correlations between epigenetic marks and gene expression pattern. Technologies developed for epigenome editing now enable direct study of functional relevance of precise epigenetic modifications and gene regulation. The reversible nature of epigenetic modifications, including DNA methylation, has been already exploited in cancer therapy for remodeling the aberrant epigenetic landscape. However, this was achieved non-selectively using epigenetic inhibitors. Epigenetic editing at specific loci represents a novel approach that might selectively and heritably alter gene expression. Here, we developed a CRISPR-Cas9-based tool for specific DNA methylation consisting of deactivated Cas9 (dCas9) nuclease and catalytic domain of the DNA methyltransferase DNMT3A targeted by co-expression of a guide RNA to any 20 bp DNA sequence followed by the NGG trinucleotide. We demonstrated targeted CpG methylation in a ∼35 bp wide region by the fusion protein. We also showed that multiple guide RNAs could target the dCas9-DNMT3A construct to multiple adjacent sites, which enabled methylation of a larger part of the promoter. DNA methylation activity was specific for the targeted region and heritable across mitotic divisions. Finally, we demonstrated that directed DNA methylation of a wider promoter region of the target loci IL6ST and BACH2 decreased their expression.

Figure 1.

(A) Schematic representation of the dCas9-DNMT3A fusion protein in complex with sgRNA and its target DNA sequence. The sgRNA is bound in a cleft between the recognition lobe (RecI, II and III domains) and the nuclease lobe (HNH, RuvC and PI domains) of Cas9 protein. The C–terminus of Cas9 is located on the PAM–interacting (PI) domain and faces the side where the bound genomic DNA protrudes with its 3′ end relative to the sgRNA sequence. The sgRNA is a synthetic fusion between bacterial crRNA and tracrRNA, with guide sequence and tracrRNA part shown in different colors. The catalytic domain of DNMT3A recruits its partner for dimerization and DNMT3L proteins in vivo (dashed lightened symbols). NLS, nuclear localization signal; GS, Gly₄Ser peptide linker. (B) Domain structure of the dCas9–DNMT3A fusion protein. The nuclease-inactivating mutations D10A and H840A of Streptococcus pyogenes Cas9 are indicated. Deactivated Cas9 was fused to the catalytic domain of the human de novo DNA methyltransferase 3A (DNMT3A CD) using a short Gly₄Ser peptide (GS). The dCas9–DNMT3A is expressed as a bicistronic mRNA, along with puromycin resistance gene (PuroR, shown) or EGFP gene, thus enabling selection of transfected cells. The PuroR (or EGFP) moiety is separated during translation by action of the T2A self-cleaving peptide. The inactive fusion methyltransferase (dCas9-DNMT3A-ANV) for use as a negative control contains an additional substitution (E155A*) in the active site of DNMT3A. 3x FLAG, epitope tag; NLS, nuclear localization signal.

Figure 2.

Targeting the dCas9-DNMT3A tool to the BACH2 and IL6ST regulatory regions. The human (A) BACH2 and (B) IL6ST loci are shown with positions of the pyrosequencing assays (BACH2-A assay, BACH2-B assay and IL6ST-A assay) indicated by blue rectangles. Magnified insets show individual CpG sites analyzed by pyrosequencing, with arrows (aligned to the magnified regions) indicating 20 bp binding sites of sgRNAs used to guide the dCas9-DNMT3A construct. Arrows point toward the PAM sequence.

Figure 3.

(A) Activity of the dCas9-DNMT3A tool guided by sgRNA8 or sgRNA3 was quantified using the BACH2-A and BACH2-B assays, respectively. The graphs show increase in CpG methylation level relative to the mock–transfected cells. Increase in the methylation level for active constructs could be observed some distance downstream from the PAM sequence (e.g. sgRNA8) or at both sides of the binding site (e.g. sgRNA3) but not at the binding site itself (shaded region indicated with arrows labeled with sgRNA names). Inactive (dCas9–DNMT3A–ANV) or non-targeting (sgRNA with no binding site in the human genome) did not show any significant increase in methylation. (B) Summary profile of the dCas9–DNMT3A activity is shown as absolute methylation fraction increase (compared to mock–transfected cells) relative to the distance of a CpG site from the PAM sequence. The summary activity profile is based on the activity of all sgRNAs targeting the BACH2 promoter and the IL6ST promoter. Vertical solid red lines represent the binding region complementary to the sgRNA. Results of all experiments were integrated by orienting the sgRNAs in the same direction and aligning the PAM sequence to position zero. Different relative positions of binding sites and pyrosequencing assays enabled construction of an activity profile covering a wide region when all available experimental data were used. The peak of CpG methylation activity extends over about 25–30 nucleotide pairs centered at the 27th nucleotide (vertical dashed purple line) downstream from the targeted PAM sequence. Another much smaller peak was consistently observed at the approximately same distance upstream from the sgRNA binding site. Gray bars represent the CpG methylation level increase observed within a single experiment (error bars show standard deviation). Each gray bar summarizes data for one CpG site position and also serves as a visual guide showing the density of coverage with experimental points. The blue curve shows LOESS smoothing of the data from multiple experiments. The brown dotted curve shows smoothed data for CpG methylation level increase by the inactive construct.

Figure 4.

Targeted CpG methylation and transcriptional silencing of the IL6ST gene by the dCas9–DNMT3A tool. (A) Increase in CpG methylation level relative to the mock–transfected cells in the IL6ST promoter region targeted by either individual sgRNAs (1–4) or by pooled sgRNAs 1–4. (B) Expression level of the IL6ST gene as measured using RT–qPCR revealed a statistically significant decrease (P < 0.05) in the transcript level following transfection with pooled sgRNAs 1–4. Fold change is relative to mock-transfected cells. Error bars represent standard deviation. Non-targeting sgRNA served as negative control.

Figure 5.

Targeted CpG methylation and transcriptional silencing of the BACH2 gene using the dCas9–DNMT3A tool. (A) CpG methylation level increased relative to mock-transfected cells in the BACH2 promoter region when targeted by either individual sgRNAs (6–8) or pooled sgRNAs 1–8. The region is covered by the pyrosequencing assay BACH2–A. Note that sgRNAs 6–8 bind close to the region covered by the BACH2–A assay, while sgRNAs 1–5 bind further downstream (see also Figure 2 for reference). (B) CpG methylation level increased relative to mock-transfected cells in the BACH2 promoter region (covered by the pyrosequencing assay BACH2–B) when targeted by either individual sgRNAs (1–5) or pooled sgRNAs 1–8. Note that sgRNAs 1–5 bind close to the region covered by the BACH2–B assay, while sgRNAs 6–8 bind further upstream (see also Figure 2 for reference). (C) Expression level of the BACH2 gene as measured using RT–qPCR revealed a statistically significant (P < 0.05) decrease in the transcript level following transfection with dCas9–DNMT3A and pooled sgRNAs 1–8. A lower but statistically significant decrease in expression was observed with matching pooled sgRNAs 1–8 used with inactive dCas9–DNMT3A constructs, which is consistent with CRISPR interference. Fold change is relative to mock-transfected cells. Error bars represent standard deviation. Non-targeting sgRNA served as negative control.

Figure 6.

Methylation level of LINE-1 elements reflecting genome-wide methylation. There was no difference in LINE-1 methylation between the mock-transfected cells (M) and those transfected with non–targeting (NT) sgRNA or any of the sgRNAs targeting (A) the IL6ST promoter region or (B) the BACH2 promoter region. Same results were obtained for either individual or pooled sgRNAs. For targeted constructs, the lighter shaded columns represent the active construct, while the darker columns correspond to controls without methyltransferase activity. Columns represent the mean methylation level across 6 CpG sites. Error bars show standard deviation.

Figure 7.

(A) Time-course evaluation of targeted CpG methylation in cells transfected with active or inactive dCas9-DNMT3A construct co-expressed with individual targeting sgRNAs. CpG methylation level increase was measured at different time points after transfection over a time period of 40 days. We analyzed CpG sites falling within the peak of methylation activity 18–42 bp downstream from the PAM sequence (positions of CpG sites relative to the PAM sequence are given in brackets). The time point when DNA methylation reaches its maximum is marked with a purple dashed vertical line. Left panel: Methylation of the BACH2-A fragment in cells co-expressing BACH2-sgRNA8. Right panel: Methylation of the IL6ST-A fragment in cells co-expressing IL6ST-sgRNA3. (B) Relative amount of plasmid DNA per cell, as determined by qPCR, decreases sharply in the 10 days following transfection. (C) Even without puromycin selection (using the construct co–expressing EGFP instead of puromycin resistance gene), the detected expression of the Cas9-DNMT3A construct (measured via co-expression of EGFP) decreases within 10 days after transfection. Arbitrary units (AU) represent fluorescence intensity per field of view normalized to the total number of cells counted under visible light.