Editing DNA methylation in vivo

Abstract

DNA methylation is a crucial epigenetic mechanism that regulates gene expression. Precise editing of DNA methylation has emerged as a promising tool for dissecting its biological function. However, challenges in delivery have limited most applications of DNA methylation editing to in vitro systems. Here, we develop two transgenic mouse lines harboring an inducible dCas9-DNMT3A or dCas9-TET1 editor to enable tissue-specific DNA methylation editing in vivo. We demonstrate that targeted methylation of the Psck9 promoter in the liver of dCas9-DNMT3A mice results in decreased Pcsk9 expression and a subsequent reduction in serum low-density lipoprotein cholesterol level. Targeted demethylation of the Mecp2 promoter in dCas9-TET1 mice reactivates Mecp2 expression from the inactive X chromosome and rescues neuronal nuclear size in Mecp2^+/- mice. Genome-wide sequencing analyses reveal minimal transcriptional off-targets, demonstrating the specificity of the system. These results demonstrate the feasibility and versatility of methylation editing, to functionally interrogate DNA methylation in vivo.

Precise editing of DNA methylation has emerged as a promising tool in disease biology but most applications are limited to in vitro systems. Here, we develop two transgenic mouse lines harboring an inducible dCas9-DNMT3A or dCas9-TET1 editor to enable tissue-specific DNA methylation editing in vivo.

Introduction

Applications of clustered regularly interspersed palindromic repeat (CRISPR) systems have greatly expanded our ability to interrogate and understand the genome. Most commonly employing an RNA-guided Cas9 nuclease to induce targeted double-strand breaks, CRISPR systems have enabled efficient disruption of gene function and precise insertion of exogenous DNA sequences¹. This technology has granted unprecedented control over genome sequence, advancing interrogation, and understanding the genomic sequence functions and therapeutic genomic editing². Building on the programmability of this technology, researchers have utilized fusion proteins consisting of a catalytically inactive Cas9 nuclease (dCas9) as a DNA targeting module and enzymatic regulators of the epigenome as effectors to introduce targeted epigenetic manipulations and thereby alter gene expression without changing the underlying DNA sequence or relying on DNA repair, offering a versatile, tunable, and reversible method to probe and modulate the regulation of gene expression^3–6.

DNA methylation is a chemical modification of DNA at the cytosine residue that serves as an important epigenetic regulator of gene expression in mammals⁷. DNA methylation studies have revealed its function in diverse biological processes, including the establishment of cellular identity, genomic imprinting, and X chromosome inactivation⁸. Methylation of DNA, particularly in the CpG dinucleotide sequence context, is catalyzed by the DNA methyltransferase family of enzymes (DNMTs), and demethylation of DNA occurs through a series of oxidative reactions catalyzed by the ten-eleven translocation (TET) dioxygenases^8,9. Our group and others have demonstrated that dCas9 fused to the catalytic domains of DNMT3A or TET1 can efficiently and specifically edit DNA methylation at precise genomic loci^10–14. These tools can silence gene expression, reactivate genes from heterochromatin, manipulate chromatin architecture, and reverse disease phenotypes in vitro by the targeted manipulation of DNA methylation^15–17.

DNA methylation editing has broad biological and therapeutic potential^6,18. However, extending these applications to in vivo systems remains a significant technological challenge due to the limitations of delivery methods for large transgene sequences of dCas9-DNMT3A and dCas9-TET1. Here, we sought to broadly enable DNA methylation editing in vivo by generating and characterizing two transgenic mouse lines in which Cre-inducible DNA methylation editors are encoded in the mouse genome. These lines enable conditional and tissue-specific DNA methylation or demethylation, requiring only the delivery of Cre recombinase and sgRNA. Using these lines, we demonstrate that targeted DNA methylation of the Pcsk9 promoter in the liver resulted in the durable repression of Pcsk9 and the lowering of LDL cholesterol level. We also show that targeted DNA demethylation of the Mecp2 promoter resulted in reactivation of Mecp2 from the inactive X chromosome in the brains of female Mecp2^+/− mice. These methylation editor mouse models allow for wider functional interrogation and therapeutic targeting of DNA methylation in a cell-type specific manner in vivo.

Results

Generation and characterization of two cell-type-specific DNA methylation editor mouse lines

To enable DNA methylation editing in vivo, we generated two transgenic lines of Cre recombination-dependent DNA methylation editor mice that carry a Cre-inducible transgene cassette expressing either dCas9-DNMT3A or dCas9-TET1 driven by a synthetic CAG promoter in the Rosa26 locus (see “Methods”). As illustrated in Fig. 1A, the first methylation editor mouse line carries a transgene consisting of pCAG-LoxP-stop-LoxP-dCas9-DNMT3A-P2A-eGFP (LSL-dC-D) to express a dCas9 fused with the catalytic domain of DNMT3A and an eGFP fluorescent reporter, separated by a P2A ribosomal skipping sequence. This mouse line enables targeted DNA methylation in a cell-type specific manner upon removal of the stop cassette by Cre-mediated recombination. To validate Cre-inducible expression of dCas9-DNMT3A transgene, we crossed this mouse line with an EIIa-Cre line in which Cre is expressed in early embryonic development¹⁹, resulting in a mosaic expression pattern across diverse tissues. As shown in Fig. 1B, we detected the expression of dCas9-DNMT3A protein in the brain lysate of dCas9-DNMT3A and EIIa-Cre double transgenic mice by western blot, but not in the lysate of dCas9-DNMT3A control mice in absence of Cre. We also observed a mosaic pattern of dCas9-DNMT3A expression within the hippocampus by immunofluorescent (IF) staining of dCas9-DNMT3A and Ella-Cre double transgenic mice, but not in the dCas9-DNMT3A alone mice as shown in Fig. 1C. These results confirmed the inducibility of dCas9-DNMT3A expression by Cre recombinase. To evaluate efficiency of dCas9-DNMT3A induction by viral delivery of cell-type specific Cre, we performed intracerebroventricular (ICV) injection of AAV9 expressing either mCherry alone as a control or mCherry and Cre driven by a neuron-specific human Synapsin promoter (hSyn) into dCas9-DNMT3A mice postnatally (P0). As shown in Fig. 1D and quantified in Fig. 1E, 96% of mCherry-positive cells expressed the GFP mark for dCas9-DNMT3A in the Cre-expressing animals (AAV9-hSyn-mCherry-Cre group) compared to 4% of mCherry-positive cells expressing GFP in non-Cre animals (AAV9-hSyn-mCherry group), demonstrating a high efficiency of cell-type specific expression of dCas9-DNMT3A methylation editor.

Fig. 1. Generation of a dCas9-DNMT3A methylation editor mouse line.

A Schematic representation of the Lox-Stop-Lox-dCas9-DNMT3A-P2A-GFP (LSL-dC9-D) transgene cassette inserted at the Rosa26 locus. pCAG cytomegalovirus enhancer fused with chicken beta-actin promoter and rabbit beta-globin splice acceptor, LSL Lox-stop-lox cassette, NLS nuclear localization sequence, P2A porcine teschivoris-1 2A self-cleaving sequence, eGFP enhanced green fluorescent protein, WPRE woodchuck hepatitis virus posttranscriptional regulatory element, bGHpA bovine growth hormone polyadenylation signal. B Western blot of DNMT3A, dCas9, and Tubulin expressions in brain tissue isolated from LSL-dCas9-DNMT3A-GFP mice and LSL-dCas9-DNMT3A-GFP; EIIa-Cre mice. C Immunofluorescent staining of GFP in the hippocampus of LSL-dCas9-DNMT3A-GFP and LSL-dCas9-DNMT3A-GFP; EIIa-Cre mice. Scale bar: 100 μm. D Immunofluorescent staining of DAPI, mCherry, GFP, dCas9 colocalization in mice injected contralaterally with either AAV9-mCherry or AAV9-mCherry-Cre. Scale bar: 100 μm. E Quantification of dCas9-DNMT3A induction efficiency in mCherry and mCherry-Cre labeled cells. (n = 3 mice per group, two-sided t test, P = 0.000010).

Next, we characterized the second methylation editor mouse line that carries a transgene of pCAG-LoxP-stop-LoxP-dCas9-TET1-P2A-eGFP (LSL-dC-T). Designed similarly, this second mouse line contains a Cre-inducible transgene consisting of a dCas9 fused with the catalytic domain of TET1 and an eGFP fluorescent reporter (Fig. 2A) to enable targeted DNA demethylation in a cell-type specific manner. By performing a similar set of experiments described in Fig. 1, we confirmed the Cre-dependent expression of dCas9-TET1 protein by western blot in Fig. 2B and by IF staining in Fig. 2C in mice carrying LSL-dC-T and Ella-Cre transgenes. Importantly, we demonstrated a robust and efficient induction of dCas9-TET1 via AAV9-mediated delivery of neuronal-specific Cre driven by the hSyn promoter in the injected LSL-dC-T mouse brain as shown in Fig. 2D, E. Neuronal specific induction of dCas9-TET1 by hSyn-Cre was validated by the 99% of NeuN positivity in the mCherry+ population (Fig. 2F). In summary, we generated and validated two transgenic mouse lines to enable targeted DNA methylation and demethylation in a cell-type specific manner in vivo.

Fig. 2. Generation of a dCas9-TET1 methylation editor mouse line.

A Schematic representation of the Lox-Stop-Lox-dCas9-TET1-P2A-GFP (LSL-dC-T) transgene cassette inserted at the Rosa26 locus. pCAG cytomegalovirus enhancer fused with chicken beta-actin promoter and rabbit beta-globin splice acceptor, LSL Lox-stop-lox cassette, NLS nuclear localization sequence, P2A porcine teschivoris-1 2A self-cleaving sequence, eGFP enhanced green fluorescent protein, WPRE woodchuck hepatitis virus posttranscriptional regulatory element, bGHpA bovine growth hormone polyadenylation signal. B Western blot of GFP, dCas9, and Tubulin expressions in brain tissue isolated from LSL-dCas9-TET1-GFP mice and LSL-dCas9-TET1-GFP; EIIa-Cre mice. C Immunofluorescent staining of GFP in the hippocampus of LSL-dCas9-TET1-GFP and LSL-dCas9-TET1-GFP; EIIa-Cre mice. Scale bar: 100 μm. D Immunofluorescent staining of DAPI, mCherry, eGFP, dCas9 colocalization in mice injected contralaterally with either mCherry or mCherry-Cre. Scale bar: 100 μm. E Quantification of dCas9-TET1 induction efficiency in mCherry-Cre and mCherry-labeled cells. (n = 3 mice per group, two-sided t test, P = 1.2 × 10⁻⁷). F Quantification of the percentage of NeuN+ cells in mCherry− and mCherry+ populations. (n = 8 mice per group, two-sided t test, P = 2.81 × 10⁻⁸).

Silencing hepatic Pcsk9 by DNA methylation in vivo

To test if targeted DNA methylation can repress gene expression in vivo, we targeted the promoter of Pcsk9, a known negative regulator for low-density lipoprotein (LDL) cholesterol²⁰, because silencing of Pcsk9 can reduce plasma LDL cholesterol level for the treatment of hypercholesterolemia. To identify effective sgRNAs targeting Pcsk9 for methylation, we designed three sgRNAs in the Pcsk9 promoter region and developed a pyrosequencing (Pyroseq) assay for the targeted region to measure DNA methylation level as illustrated in Fig. 3A. To test these sgRNAs, we chose AML-12 cells (a mouse hepatocyte cell line) and generated an AML-12 cell line that stably expresses a doxycycline (dox) inducible dCas9-DNMT3A cassette (Supplementary Fig. 1). After infecting these AML-12 cells with lentiviruses expressing target sgRNA-1, −2, −3, or scrambled (Scr) sgRNA as control, we found that the combination of sgRNA-1 and −2 resulted in the most robust decrease of Pcsk9 expression compared to the Scr control or each individual target sgRNA as shown in Fig. 3B. Combining sgRNA-1 and −2 significantly increased the DNA methylation level of each CpG within the Pcsk9 promoter compared to mock or cells transduced with lentivirus containing the Scr control sgRNA as shown in Fig. 3C. We proceeded with the combined sgRNA-1 and −2 for in vivo experiments.

Fig. 3. Silencing of Pcsk9 by DNA methylation editing in vivo.

A Schematic of designed sgRNAs targeting the Pcsk9 promoter and Pyro-seq assay area in yellow. B Pcsk9 expression in AML12 cells after DNA methylation editing by dCas9-DNMT3A with sgRNAs in (A). (n = 5 biological replicates per group, one way ANOVA (Dunnett’s test vs Scr), sgRNA-2 vs Scr P = 0.0058, sgRNA-1 + 2 vs Scr P = 0.00147). C DNA methylation status of the Pcsk9 promoter after DNA methylation editing by dCas9-DNMT3A in AML12 cells measured by pyrosequencing. Pcsk9 sgRNAs refers to editing with sgRNAs 1 + 2. (n = 3 biological replicates per group, one-way ANOVA with Tukey’s multiple comparisons per CpG, * P < 0.05, exact P-values provided in source data). D Scheme for in vivo repression of Pcsk9 by targeted methylation of the Pcsk9 promoter in the liver. Created in BioRender. Liu, S. (2025) https://BioRender.com/m5iikln. E DNA methylation status of the Pcsk9 promoter after 6 weeks from liver tissue isolated from dCas9-DNMT3A mice injected with AAV containing Cre and Scr or Pcsk9 targeting sgRNAs measured by pyrosequencing of area (A and B). (n = 3 (PBS), n = 9 (Scr), n = 10 (Pcsk9) mice, one-way ANOVA (Tukey’s test exact P values in source data). F Pcsk9 transcript expression in livers from (D). (n = 5 (Scr) and n = 6 (Pcsk9) mice, two-sided t test, P = 0.0292). G Western blot of LDLR and PCSK9 protein expression levels in liver tissue in (D). H Quantification of LDLR and PCSK9 protein levels from western blot in (G). (n = 6 (Scr) and n = 7 (Pcsk9) mice, two-sided t test, LDLR P = 0.0292, PCSK9 P = 0.040). I Representative immunofluorescent staining of LDLR in livers from Scr and Pcsk9 targeted livers. Scale bar: 50 μm. J Quantification of % area stained of LDLR in livers from Scr and Pcsk9 targeted livers. (n = 4 mice per group, two-sided t test, P = 0.0093). K Serum LDL cholesterol measured biweekly in dCas9-DNMT3A mice injected with Scr or Pcsk9 targeting sgRNAs. (n = 9 mice per group, two-way ANOVA repeated measures, P = 0.000021). L Quantification of serum PCSK9 levels at study endpoint from Scr and Pcsk9 targeted livers. (n = 5 (Scr) and n = 6 (Pcsk9) mice, two-sided t test, Absolute p = 0.0032, Relative p = 0.00078).

To evaluate if targeted DNA methylation represses Pcsk9 expression in vivo, we performed editing by retro-orbital injection of AAV9 virus carrying a U6 promoter driven expression of target or control sgRNA and a CBh promoter driven expression of Cre in dCas9-DNMT3A mice aged 8–10 weeks as illustrated in Fig. 3D. We harvested liver tissue 6 weeks after injection for downstream analysis. Pyroseq revealed a significant increase of DNA methylation of the Pcsk9 promoter in the Pcsk9 target sgRNA group, but not in the PBS control or Scr control sgRNA group (Fig. 3E). qPCR analysis confirmed decreased expression of Pcsk9 transcript in the Pcsk9 edited liver (Fig. 3F). We further validated the decrease of PCSK9 expression at the protein level. Western blot result showed a decrease of PCSK9 protein levels and an increase of its target LDLR protein levels in the livers of Pcsk9 target sgRNAs injected mice compared to Scr sgRNA injected mice (Fig. 3G); PCSK9 and LDLR protein quantification (Fig. 3H) confirmed this notion. In addition, we performed IF staining of LDLR using liver sections from edited mice. As shown in Fig. 3I and quantified in Fig. 3J, the expression of LDLR protein at the plasma membrane was significantly reduced in Pcsk9 edited mice. To assess the functional consequence of targeted Pcsk9 methylation, we collected serum from injected mice at baseline and after 2, 4, and 6 weeks. We observed a 33% decrease in LDL cholesterol from baseline as early as two weeks after injection, which was sustained to 59% after 6-weeks (Fig. 3K). To further assess the cholesterol-modulating effects by editing Pcsk9, we also measured HDL, VLDL, and total cholesterol levels following targeted methylation. While LDL cholesterol was significantly decreased at 2-, 4-, and 6-weeks post-injection, HDL, VLDL, and total cholesterol were not significantly different at the timepoints assessed (Supplementary Fig. S3A–D). Interestingly, VLDL and total cholesterol levels showed a trend of decrease at 2 weeks post-injection, but then stabilized and returned to similar levels by 4 weeks post-injection. These results suggest that the cholesterol-modulating effects of Pcsk9 editing are relatively specific to LDL cholesterol. Further, serum PCSK9 levels in Pcsk9 target sgRNAs injected mice was decreased by 60% from baseline after 6-weeks (Fig. 3L), demonstrating effective and sustained gene repression with a single dose of AAV-mediated methylation editing in vivo.

Off-target analysis of Pcsk9 methylation edited mice

It is essential to evaluate the off-target effects associated with DNA methylation editing in vivo. We performed systematic analysis by examining the genome-wide dCas9-DNMT3A binding, DNA methylation landscape, and gene expression in livers from mice injected with either Pcsk9 target sgRNAs or a Scr control sgRNA. ChIP-seq using an anti-Cas9 antibody showed specific binding of dCas9-DNMT3A at the targeted Pcsk9 promoter region in the methylation edited cells by target sgRNAs, but not in the mock cells or cells expressing dCas9-DNMT3A alone or with Scr control sgRNA (Fig. 4A). Among the 10 genome-wide binding sites for dCas9-DNMT3A detected by ChIP-seq (Supplementary Data 1), the targeted Pcsk9 promoter showed the most significant enrichment for dCas9-DNMT3A occupancy (Fig. 4B), highlighting the targeting specificity. To evaluate the potential off-target effects of dCas9-DNMT3A on genome-wide methylation, we performed whole-genome bisulfite sequencing (WGBS). Our WGBS experiment included three biological replicates for each of the experimental groups (livers from Pcsk9 target or Scr control sgRNA-injected LSL-dC-D mice), and covered 19,731,814 CpG sites in the mouse genome. The average CpG methylation in the Pcsk9 target group and Scr control group was the same (69%), suggesting no alteration in the overall DNA methylation by dCas9-DNMT3A. Among the 10 dCas9-DNMT3A binding sites, we found that the methylation level of the targeted Pcsk9 promoter region was significantly increased in the liver from the Pcsk9 sgRNA-injected mice (48%) compared to the Scr sgRNA-injected mice (0%), as shown in Fig. 4C, whereas methylation levels at 9 other dCas9-DNMT3A binding sites (labeled by black dots) were not altered compared to the Scr control (Fig. 4D and Supplementary Data 2), suggesting robust and specific DNA methylation editing of Pcsk9 in vivo. To evaluate potential off-target effects at the transcriptional level, we performed RNA sequencing (RNA-seq) using Pcsk9 target sgRNA and Scr control sgRNA edited livers (Supplementary Data 3). As shown in Fig. 4E and Supplementary Data 4, Pcsk9 was the only gene significantly repressed in Pcsk9 edited livers among the all the genes associated with dCas9-DNMT3A binding sites (labeled by black dots). In addition to Pcsk9, 15 other genes were identified as differentially expressed by RNA-Seq (P-adj <0.05, Fold-Change > 2, Supplementary Data 5). However, none of these 15 genes was bound by dCas9-DNMT3A as examined by ChIP-seq in Fig. 4B, suggesting that the expression changes in these 15 genes were unlikely to be caused by dCas9-DNMT3A.

Fig. 4. Off target analysis of Pcsk9 methylation editing.

A IGV browser track of anti-Cas9 ChIP-seq data demonstrating the specific binding of dCas9-DNMT3A to the Pcsk9 promoter in DNA methylation edited AML cells. B Representative IGV track of WGBS data of Scr and Pcsk9 sgRNA injected livers at the Pcsk9 locus. The differentially methylated region (DMR) at the Pcsk9 promoter is highlighted in red. C Quantification of average methylation at the DMR identified at the Pcsk9 promoter in Scr and Pcsk9 targeted livers. D DNA methylation of the dCas9-DNMT3A binding sites in Scr and Pcsk9 targeted livers measured by whole genome bisulfite sequencing (WGBS). Pcsk9 is labeled by a red dot and other sites are labeled by black dots. The dashed lines represent a 20% change of DNA methylation level. E RNA-seq of Scr sgRNA and Pcsk9 sgRNA injected livers. Pcsk9 is labeled by a red dot and other genes associated with dCas9-TET1 binding sites are labeled by black dots. The dashed lines represent a twofold change in expression between the conditions.

Furthermore, to evaluate off-target effects in non-hepatic tissues, we performed pyrosequencing and RNA-seq using heart and brain tissues from dCas9-DNMT3A mice edited with control Scr sgRNA or Pcsk9-targeting sgRNAs. We observed no significant differences at the DNA methylation level for the Pcsk9 promoter in the heart or brain (Supplementary Fig. 3E, F) and no alteration of the Pcsk9 expression level in the heart or brain (Supplementary Fig. 3G, H), suggesting that the effect of Pcsk9 editing was specific to the liver. RNA-seq identified one differentially expressed gene (DEG) in the brain, and 69 DEGs in the heart (Supplementary Fig. 3G, H). However, none of these DEGs overlapped with the dCas9-DNMT3A binding sites identified by ChIP-seq (Fig. 4B), suggesting these DEGs in brain and heart were independent of dCas9-DNMT3A. In summary, analyses of ChIP-seq, WGBS, and RNA-seq results did not detect off-target effects by dCas9-DNMT3A at the transcriptional level, supporting high specificity of this dCas9-DNMT3A mouse line for DNA methylation editing in vivo.

Reactivation of Mecp2 from the inactive X chromosome by DNA demethylation editing of Rett syndrome mice

To investigate whether dCas9-TET1 mice are capable of DNA demethylation editing in vivo, we targeted Mecp2, an X-linked gene in which heterozygous loss-of-function mutations cause the neurodevelopmental disorder Rett Syndrome (RTT)²¹. Random X chromosome inactivation in heterozygous female patients with RTT results in ~50% of neurons expressing the mutant MECP2 allele from the active X chromosome (Xa) without functional MeCP2 protein²². Thus, reactivation of the wild-type and healthy MECP2 allele from the inactive X chromosome (Xi) represents a potential therapeutic strategy for female RTT patients²³. We previously demonstrated that targeted DNA demethylation of the MECP2 promoter by dCas9-TET1 can reactivate MECP2 from the Xi in vitro and rescue disease-associated deficits in RTT neurons¹⁶. To target Mecp2 in vivo, we first identified effective sgRNAs for the demethylation of the mouse Mecp2 promoter. As illustrated in Fig. 5A, we designed 7 sgRNAs located across the mouse Mecp2 promoter and developed one Pyroseq assay for the targeted region to measure the methylation level after editing. We used a clonal Xi-linked Mecp2-Luciferase reporter fibroblast cell line²⁴ with XIST deletion as illustrated in Fig. 5B to test the effectiveness of these designed sgRNAs. At baseline, this reporter cell line features no detectable luciferase activity because the Mecp2-Luciferase reporter is silenced on the Xi, but DNA demethylation using small molecule DNMT1 inhibitor GSK3685032 (GSK) activates the reporter (Fig. 5C). We generated a Mecp2-Luciferase cell line that stably expresses a doxycycline (dox) inducible dCas9-TET1 cassette (Supplementary Fig. 2). Interestingly, we found that a combination of 7 target sgRNAs (Pool), but not individual target sgRNAs or Scr control sgRNA, activated the reporter to levels comparable to the positive control treated with a DNA methylation inhibitor (Fig. 5C). We confirmed robust demethylation of the Mecp2 promoter by this pool of target sgRNAs by pyrosequencing (Fig. 5D). These results demonstrate that targeted DNA demethylation can reactivate Mecp2 expression from the Xi in mouse cells. We then used this pool of Mecp2 targeting sgRNAs for in vivo methylation editing.

Fig. 5. Reactivation of Mecp2 from the Xi by demethylation editing in vivo.

A Design of 7 sgRNAs targeting the mouse Mecp2 promoter. sg–sgRNA, TSS–transcription start site. B Scheme of reporter cell line with Mecp2-luciferase on the Xi used to identify effective sgRNAs for Mecp2 reactivation. Luc luciferase, Hygro hygromycin resistance gene. C Luciferase assay results of Mecp2-luciferase reporter cells edited with dCas9-TET1 and an individual sgRNA or the pool of all 7 sgRNAs, or treated with GSK-3484862 (GSK). (n = 2 (Mock, sgRNA-1, sgRNA-3, sgRNA-6, Scr) and n = 3 sgRNA-2, sgRNA-4, sgRNA-5, sgRNA-7, sgRNA-Pool, GSK) biological replicates, one-way ANOVA (Dunnett’s test Pool vs Scr), P = 1.9 × 10⁻⁹) Scr Scrambled sgRNA. D DNA methylation status of the Mecp2 promoter measured by pyrosequencing of mock cells, and cells edited with dCas9-Tet1 with a non-targeting Scr sgRNA or the pool of 7 sgRNAs. (n = 2 (Mock, Scr) and n = 3 (Mecp2) biological replicates, one-way ANOVA with Tukey’s test, exact P-values found in source data). E Scheme for in vivo reactivation of Mecp2. ICV intracerebroventricular, IF immunofluorescence. Created in BioRender. Liu, S. (2025) https://BioRender.com/4dk6z6j. F Transduction efficiency of AAV9 in (D). G DNA methylation status of the Mecp2 promoter in the isolated nuclei (mCherry+) from mice edited by dCas9-TET1 with Scr or Mecp2 sgRNA pool. (n = 2 mice per group, one-way ANOVA with Tukey’s test, exact P-values found in source data). H Representative immunofluorescent staining of MeCP2, mCherry, GFP, and NeuN from Scr and Mecp2 targeted cortical sections. I Quantification of MeCP2+ cells in (H). (n = 3 mice per group, two-sided t test, P = 0.0309). J Flow cytometry analysis of neuronal nuclear size of mCherry- cells from dCas9-TET1 mice injected with Scr or Mecp2 sgRNA pool. (n = 5 (Scr) and n = 6 (Mecp2) mice, two-sided t test, P = 0.49). FSC–forward scatter. K Flow cytometry analysis of neuronal nuclear size of mCherry+ cells from dCas9-TET1 mice injected with Scr or Mecp2 sgRNA pool. (n = 5 (Scr) and n = 6 (Mecp2) mice, two-sided t test, P = 0.013). L Quantification of the relative neuronal nuclear size between mCherry+ and mCherry− cells in control (Scr) and edited (Mecp2) mice. (n = 5 (scr) and n = 6 (Mecp2) mice, two-sided t test, P = 0.0017).

To assess if targeted DNA demethylation can reactivate Mecp2 in vivo, we crossed the LSL-dC-T mouse line with a Mecp2^+/− heterozygous Rett syndrome mouse line²⁵ to generate female LSL-dC-T;Mecp2^+/− mice. Then, we generated and packaged AAV9 virus expressing a U6 promoter-driven Mecp2 target or Scr control sgRNAs and a hSyn promoter-driven Cre with mCherry fused to a KASH tag²⁶ that inserts into the nuclear membrane, enabling fluorescence-activated nucleus sorting (FACS) of mCherry-KASH positive cells or nuclei. We performed ICV injection of these AAVs into postnatal female LSL-dC-T/ Mecp2⁻ mice as illustrated in Fig. 5E. We quantified the efficiency of AAV transduction in the brain of injected mice and found that 55% of brain cells expressed GFP mark for dCas9-TET1 in AAV9 injected mice whereas <0.1% of brain cells expressed GFP in PBS injected control mice, suggesting efficient and robust AAV9 transduction to induce dCas9-TET1 editor expression in vivo (Fig. 5F). Then, we FACS isolated mCherry-KASH positive nuclei from the cortex and found significant demethylation of the Mecp2 promoter in mice injected with Mecp2 target sgRNAs compared to mice injected with a Scr control sgRNA (Fig. 5G), supporting in vivo demethylation editing of mouse Mecp2 in the targeted neurons.

To evaluate Mecp2 reactivation induced by demethylation editing in vivo, we performed IF staining of brain sections from Mecp2 target and Scr control sgRNA-injected mice (Fig. 5H) and quantified the percentage of MeCP2+ neurons in their brains. We found the percentage of MeCP2+ neurons increased from 50.7% in the cortex of Scr control mice to 67.5% in the cortex of Mecp2 edited mice (Fig. 5I), suggesting that targeted DNA demethylation reactivated Mecp2 on the Xi in vivo. This level of Mecp2 reactivation is consistent with our previous report on human MECP2 reactivation by DNA methylation editing of in vitro differentiated human Rett syndrome neurons¹⁶.

Next, we sought to evaluate the phenotypic rescue of Mecp2 reactivation in vivo. Here, we assessed neuronal nuclear size—a known cellular defect of MeCP2-null neurons^27–29—in Mecp2-targeted and Scr control sgRNA injected mice using an established flow cytometry approach³⁰. We analyzed cortical neuronal nuclei and observed no significant difference in the mean neuronal nuclear size of unedited (mCherry−) neurons between the two groups (Fig. 5J). In contrast, we observed a significant increase in the neuronal nuclear size of edited neurons (mCherry+) in Mecp2-target sgRNA-injected mice compared to Scr sgRNA-injected mice (Fig. 5K). To control for variation between individual mice, we also calculated the ratio of neuronal nuclear size between mCherry+ and mCherry− neurons within each mouse, and observed a significant increase in Mecp2 edited mice compared to control mice (Fig. 5L). These results demonstrate phenotypic rescue at the cellular level following Mecp2 reactivation by targeted demethylation of Mecp2 in vivo.

Off-target analysis of Mecp2 methylation edited mice

To evaluate off-target effects of DNA methylation editing of Mecp2 in vivo, we performed systematic analysis by examining genome-wide dCas9-TET1 binding, DNA methylation status, and gene expression in neurons from Mecp2 target sgRNAs and Scr control sgRNA-injected mice. ChIP-seq using an anti-Cas9 antibody showed specific binding of dCas9-TET1 at the targeted Mecp2 promoter region only in the methylation edited cells by Mecp2 target sgRNAs, but not in the mock cells or cells expressing dCas9-TET1 alone or with Scr control sgRNA (Fig. 6A). Among the 80 genome-wide binding sites by dCas9-TET1 detected by ChIP-seq (Supplementary Data 6), the targeted Mecp2 promoter showed the most significant enrichment for dCas9-TET1 occupancy (Fig. 6B), suggesting high targeting specificity. To further evaluate potential off-target effects of dCas9-TET1 on genome-wide methylation, we performed WGBS using Mecp2 edited or Scr control nuclei (mCherry-KASH+) isolated from mice. Our WGBS experiment included three biological replicates for each experimental group and covered 35,422,595 CpG sites in the mouse genome. Among the 80 dCas9-TET1 binding sites, we found that the methylation level of the targeted Mecp2 promoter region was significantly decreased in the brain from the Mecp2 sgRNAs injected mice (13%) compared to the Scr sgRNA injected mice (42%) as shown in Fig. 6C. In addition to Mecp2, 23 of the 80 total binding sites showed significant changes of DNA methylation level for at least one cytosine (q < 0.01, methylation difference > 15%) as labeled by black dots in Fig. 6D and listed in Supplementary Data 7, which likely resulted from the use of 7 Mecp2 target sgRNAs. To evaluate off-target effects at the transcriptional level, we performed RNA-seq using Mecp2 target sgRNAs and Scr control sgRNA edited cells (expressions of all detected genes listed in Supplementary Data 8). As shown in Fig. 6E and Supplementary Data 9, the Mecp2 reporter was the only gene significantly upregulated in Mecp2 edited cells among all the genes associated with 80 dCas9-TET1 binding sites, supporting high transcriptional specificity by DNA demethylation editing. Interestingly, three other genes (Fndc1, Plac8, and Rspo2) associated with dCas9-TET1 binding were downregulated. As the dCas9-TET1 binding sites are at least 4 kb away from the transcriptional start sites of these genes, their downregulation is unlikely to be caused by dCas9-TET1 binding. However, we were not able to completely rule out the possibility that dCas9 binding may influence regulatory elements that might alter gene expression. Besides Mecp2, our RNA-seq identified other 1371 DEGs (P-adj < 0.05, Fold-Change > 2) as listed in Supplementary Data 10. Overlapping these DEGs and the genes with a change of DNA methylation level in Fig. 6D only identified the target gene Mecp2 and another gene Plac8 (Fig. S4E). Since the expression of Plac8 was downregulated and the associated dCas9-TET1 binding site is 34 kb upstream of the transcriptional start site of Plac8, the downregulation of Plac8 expression is unlikely to be caused by dCas9-mediated demethylation or binding. Furthermore, none of the other 1370 DEGs were associated with dCas9-TET1 binding and DNA methylation changes, indicating that the expression changes in these genes were unlikely to be caused by dCas9-TET1. In summary, integrated analyses of ChIP-seq, WGBS, and RNA-seq results did not detect off-target effects by dCas9-TET1 at the transcriptional level, suggesting a high specificity of this dCas9-TET1 mouse line for DNA demethylation editing in vivo.

Fig. 6. Off target analysis of Mecp2 demethylation editing.

A IGV browser track of dCas9 ChIP-seq data demonstrating binding of dCas9-TET1 to the Mecp2 promoter. ChIP – chromatin immunoprecipitation. B Manhattan plot of genome-wide binding sites of dCas9-TET1 with the pool of 7 sgRNAs. C Representative IGV track of WGBS data at the Mecp2 promoter of nuclei isolated from Scr sgRNA and Mecp2 pool sgRNA injected mice. The differentially methylated region (DMR) at the Mecp2 promoter is highlighted in red. WGBS whole genome bisulfite sequencing. D DNA methylation levels of dCas9-TET1 binding sites in mCherry+ nuclei isolated from Scr sgRNA or Mecp2 pool sgRNAs. Average methylation was calculated by averaging methylation of differentially methylated cytosines (DMCs) within binding sites if DMCs were identified within, or an average of all CpGs within the binding site if no DMCs were identified. Mecp2 is labeled by a red dot and other sites are labeled by black dots. The dashed lines represent a 15% change of DNA methylation level. E RNA-seq of Mecp2-luciferase reporter cells edited with dCas9-TET1 + a Scr sgRNA or the pool of Mecp2 sgRNAs. Genes associated with dCas9-TET1 binding sites are labeled in black. Mecp2 reporter gene expression is labeled in red. The dashed lines represent a fourfold change in expression between the conditions. CPM counts per million.

To further evaluate off-target effects in non-targeted tissues other than the brain, we performed pyrosequencing of the Mecp2 promoter in the liver and heart tissues isolated from dCas9-TET1 mice injected with Mecp2-target sgRNAs and control Scr sgRNA. We detected no changes in DNA methylation at the Mecp2 promoter in the liver and heart (Supplementary Fig. 4A, B), suggesting that the targeted demethylation of Mecp2 was brain-specific. To investigate potential transcriptional off-target effects in non-targeted tissues, we also performed RNA-seq using the liver and heart tissue from these mice (Supplementary Fig. 4C, D). Notably, Mecp2 expression was not altered in either tissue, confirming the editing effect was specific to the brain. RNA-seq identified 24 DEGs in the heart and 196 DEGs in the liver, but none of these DEGs overlapped with dCas9-TET1 binding sites identified in Fig. 6B, suggesting that these transcriptional changes were independent of dCas9-TET1 activity. In summary, these results confirm the tissue specificity of Mecp2 editing in the brain.

Discussion

Previous genetic and pharmacologic studies of DNA methylation have demonstrated its importance on a genome-wide scale³¹, however, the functional significance of methylation at specific genomic loci has traditionally been difficult to dissect. DNA methylation editing tools can efficiently edit DNA methylation status at specific genomic loci, enabling site-specific functional dissection of DNA methylation. Studies applying these tools have further revealed the utility of rewriting DNA methylation to ameliorate certain disease-associated phenotypes^{15–17,32,33}, highlighting their therapeutic potential. Here, we describe two transgenic mouse models that enable a cell-type specific DNA methylation editing in vivo. We chose AAV9 as a delivery vector for cell-type specific Cre to activate dCas9-DNMT3A or dCas9-TET1 editor and for target sgRNA to achieve versatile editing in different tissues. Using this in vivo editing model, we demonstrate that DNA methylation editing in the liver and brain is efficient and specific, resulting in stable gene expression changes and downstream physiological effects.

We targeted Pcsk9 as a proof-of-concept for methylation editing as silencing of Pcsk9 can lower plasma LDL cholesterol levels and thereby reduce the risk for cardiovascular disease^20,34,35. Our single-dose AAV9 retro-orbital injection significantly methylated the Pcsk9 promoter in liver, leading to sustained repression of Pcsk9 and a corresponding decrease in LDL cholesterol levels. This result is consistent with recent studies reporting a similar silencing of Pcsk9 by epigenome editing³², suggesting its valuable therapeutic potential. We recognize the limitation of our Pcsk9 editing experiments that were done with AAV-Cbh-Cre, but not by a liver-specific Cre. Editing of hepatic Pcsk9 by AAV-Cbh-Cre was largely attributed to the liver-enrichment of systematic delivery of AAV as reported previously³⁶.

Loss-of-function mutations of MECP2, an X-linked gene, causes the severe neurodevelopmental disorder Rett syndrome²¹. Most female patients with RTT carry heterozygous mutations, resulting in about half of their neurons expressing the mutant MECP2 allele from the active X chromosome and leaving the wild-type MECP2 silenced on the inactive X chromosome²². Given that restoring Mecp2 expression rescues mouse models of Mecp2 deficiency³⁷, reactivation of the silenced wild-type MECP2 allele from the inactive X presents a unique opportunity for epigenome editing-mediated gene activation. This approach also offers a potential advantage over gene replacement strategies, as it is unlikely to cause overexpression of MECP2 and result in MECP2 duplication syndrome, which can complicate gene replacement strategies. Using dCas9-TET1 mice, we demonstrate that targeted DNA demethylation of the Mecp2 promoter results in its reactivation from the inactive X chromosome and rescue of neuronal nuclear size in Rett syndrome mice, consistent with our previous study in which we reactivate MECP2 in Rett syndrome hESCs and in vitro derived neurons by DNA methylation editing¹⁶. The level of Mecp2 reactivation by DNA methylation editing is moderate, from ~50% MeCP2 positivity in control heterozygous mice to 67% in edited mice. Thus, further studies are needed to quantify the extent of Mecp2 reactivation in a Mecp2-null background as well as the degree of phenotypic rescue at the physiological level to evaluate this therapeutic strategy for Rett syndrome.

Off-target effects are an important consideration with genome and epigenome editing². Our genome-wide sequencing analysis revealed that DNA methylation editing in vivo using these two mouse models is highly specific, without significant transcriptional off-target effects. At the DNA methylation level, we detected more off-target sites for targeted Mecp2 demethylation compared to targeted Pcsk9 methylation, which likely resulted from the number of targeting sgRNAs used for these two loci (7 sgRNAs for Mecp2 versus 2 sgRNAs for Pcsk9). Nevertheless, no detectable off-target effect at the transcriptional expression level was found for both methylation and demethylation editing, highlighting the specific outcome by these tools in vivo. Furthermore, methylation and transcriptomic analyses revealed little to no detectable off-target effects in non-targeted tissues, highlighting the tissue specificity of editing enabled by these transgenic mouse lines.

Another consideration with DNA methylation editing is the maintenance of the edit without the expression of editors. Due to the genetic design of the Cre recombination-dependent transgenes in these mice, methylation editor expression is permanently induced following excision of the stop cassette. Thus, we were unable to examine the maintenance of methylation editing in vivo. Interestingly, a recent study using lipid nanoparticles (LNPs) delivering an epigenome editor of dCas9 fused with DNMT3A, DNMT3L, and KRAB showed a durably silenced Pcsk9 in vivo, suggesting a promising durability for epigenetically edited locus. Our previous studies on the methylation editing of FMR1 in Fragile X syndrome^15,38 and MECP2 in Rett syndrome¹⁶ support the notion that the duration of methylation editing is influenced by the local chromatin status. Further studies using transient delivery of editor and sgRNA (e.g., LNPs) can help determine the duration of DNA methylation editing at different loci in vivo.

In addition to demonstrating the therapeutic potential of DNA methylation editing, these mice may help reveal the functional significance of DNA methylation in physiological and pathological processes. For example, global hypermethylation in certain cancers containing isocitrate dehydrogenase mutations disrupt the boundaries of topologically associated domains, leading to aberrant enhancer activity and oncogene upregulation³⁹. These mice may be useful in determining if specific methylation events drive oncogenesis in vivo. Further, DNA methylation may play a critical role in the mechanisms underlying memory consolidation, stabilizing long-term transcriptional changes underlying memory encoding⁴⁰. These mice can be used to test the role of specific methylation events to form and consolidate memories in vivo. In summary, these mice allow for the targeted manipulation of DNA methylation in vivo, enabling the further understanding of epigenetic processes in health and disease.

Methods

Mouse lines and breeding strategies

LSL-dCas9-DNMT3A and LSL-dCas9-TET1 mouse lines were generated by blastocyst injection of the genetically engineered KV-1 hybrid mouse embryonic stem cells (mESC), embryo transplantation, chimeric mouse production, and germline transmission by crossing with C57BL/6 mouse line in Genetically Modified Mouse Models/GMMMSR (HICCC) core facility at Columbia University Irving Medical Center. The targeted KV-1 mESC clones were generated by electroporation of the modified Rosa26 targeting vector construct Ai9 (Addgene 22799) containing a transgenic LSL-dCas9-DNMT3A or LSL-dCas9-TET1 cassette and identification of positive clones by genotyping PCR. Mice were handled in accordance with institutional guidelines and approved by the Institutional Animal Care and Use Committee (IUCAC) and ICM of Columbia University. LSL-dCas9-DNMT3A mice with C57BL/6 background aged 6 to 8 weeks were used for Pcsk9 editing experiments (9 mice Scr and 10 mice Pcsk9), sex was not considered in study design. For Mecp2 experiments, female postnatal mice with LSL-dCas9-TET1 and Mecp2^+/− background were used (5 Scr control and 6 Mecp2). Only female mice were used for reactivation from the inactive X chromosome. Mice were housed with 12-h light-dark cycles at 65–75 °F and 40–60% humidity.

Plasmid design and construction

PCR amplified dCas9-TET1-P2A-eGFP or dCas9-DNMT3A-P2A-eGFP fragment was cloned into the Ai-9 construct (Addgene plasmid: 22799) with FseI restriction site to generate target vectors for mESC electroporation. The AAV construct expressing Pcsk9 target sgRNA was cloned by inserting an mCherry-KASH fragment into the backbone construct (Addgene plasmid: 60229), and the AAV construct expressing Mecp2 target sgRNA was cloned by replacing the EGFP with mCherry in the backbone construct (Addgene plasmid: 60231)⁴¹. For cell culture experiments, sgRNAs were cloned into the pgRNA backbone (Addgene 84477)¹¹ using an AarI cut site. For in vivo experiments, sgRNAs were cloned into the modified AAV constructs using a SapI cut site. Oligonucleotides were ordered from Eton Bioscience Inc. and Integrated DNA Technologies. All constructs were sequenced before transfection. sgRNA sequences are listed in Supplementary Table 2. Requests for reagents may be directed to and will be fulfilled by the corresponding author X. Shawn Liu (sl4738@cumc.columbia.edu).

Cell culture and lentivirus production

AML12 mouse hepatocytes were purchased from ATCC (CRL-2254) and were cultured in Dulbecco’s modified Eagle medium (DMEM) with 10% Cosmic Calf Serum (HyClone), 1% penicillin/streptomycin (Gibco), 1% Insulin-Trasferrin-Selenium (Gibco), and 40 ng/uL dexamethasone (MilliporeSigma). Mecp2-Luciferase-Hygromycin mouse tail fibroblasts were kindly provided by Dr. Antonio Bedalov and were cultured in DMEM with 10% Cosmic Calf Serum (Hyclone), 1% pencillin/streptomycin (Gibco), 2 mM glutamine (Invitrogen), and 1% non-essential amino acids (Invitrogen). Cells were cultured at 37 °C with 5% CO2. The doxycycline-inducible dCas9-DNMT3A AML12 cell line was constructed by transfecting wild-type AML12 cells with a piggyBac transposase constructs 137¹¹ and 138-dCas9-DNMT3a (Addgene 84570) and selecting with hygromycin for 7 days. The doxycycline-inducible dCas9-TET1 tail fibroblast line was generated similarly following transfection of a piggyBac transposase constructs 137¹¹ and a modified 138-dCas9-TET1 from Addgene plasmid 84570 and 7 days of hygromycin selection.

Lentiviruses expressing sgRNAs were generated by transfecting HEK293T cells with the pgRNA constructs together with the lentiviral packaging plasmids pCMV-dR8.74 and pCMV-VSVG. Viral supernatants were collected, filtered, and concentrated by ultracentrifugation.

Immunohistochemistry, microscopy, and image analysis

Mice were perfused with 4% paraformaldehyde and then brain and livers were harvested for sectioning. 40 μm sections were cut using a vibratome (Leica) for immunofluorescence staining. Antigen retrieval was performed by incubating sections in 10 mM sodium citrate buffer at 80 °C for 30 min prior to staining. Sections were then washed three times in PBS and blocked/permeabilized with 1× PBS with 0.3% Triton ×-100 and 3% normal donkey serum. Sections were then washed three times at room temperature with PBS and stained with primary antibody solution in 1× PBS with 0.3% Triton ×-100 and 1% normal donkey serum overnight at 4 °C. Sections were then washed three times in PBS and then stained with secondary antibody solution in 1× PBS with 0.3% Triton ×-100 and 1% normal donkey serum for 2 h at room temperature. Sections were then washed three times in PBS before being mounted and coverslipped. The following antibodies were used in this study: rabbit anti-Cas9 (1:1000, Encor Biotech, Cat# RPCA-CAS9-Sp), chicken anti-GFP (1:2500, Aves Labs, Cat#GFP-1020), mouse anti-dsRed2 (1:500, Santa Cruz Biotech Cat#sc-101529), guinea pig anti-NeuN (1:1000, Synatpic Systems, Cat#266004), rabbit anti-LDLR (1:1000, Abcam, Cat# ab52818). Images were acquired on a Nikon Ti Eclipse inverted confocal microscope, and Nikon spinning disk confocal microscope at the Confocal and Microscopy Core at the Herbert Irving Comprehensive Cancer Center. Images were processed with ImageJ/Fiji and Adobe Photoshop. For imaging-based quantification, unless otherwise specified, 3–5 representative images were quantified and data were plotted as mean ± SD.

Flow cytometry

Transduced mCherry-positive cells or mCherry-KASH nuclei were isolated by fluorescence-activated cell sorting (FACS). Cultured cells were lifted with 0.25% Trypsin (Gibco), washed with 1× PBS and then resuspended in staining buffer (PBS, 1% BSA, 1 mM EDTA). Samples were filtered through a 35 μM strainer before analysis. For flow cytometry and analysis of neuronal nuclei, cortical tissue was dissected from isolated brain tissue and processed via the Nuclei EZ Prep Isolation kit (Sigma). Tissue samples were homogenized in a Dounce homogenizer and then washed twice in ice-cold Nuclei EZ Lysis Buffer. Samples were then washed in staining buffer (PBS, 1% BSA, 1 mM EDTA). and filtered through a 35 μM cell strainer. Samples were then stained with NeuN-488 antibody (Abcam EPR12763, Cat#ab190195, 1:100) at 4 °C for 30 min. Samples were then washed three times and resuspended in staining buffer before analysis. Flow sorting was done on a BD Influx Cell Sorter. Flow analysis was performed on an Agilent NovoCyte Penteon.

Western blot

Lysate from cells and livers was extracted in RIPA buffer with proteinase inhibitor (Invitrogen) and subject to standard immunoblotting analysis. Lysate from brain tissue was isolated using N-PER—Neuronal Protein Extraction Reagent (Thermo-scientific 87792) according to the manufacturer’s instructions. In short, 1 × g of tissue was homogenized in 10 mL of N-PER in a dounce homogenizer and then centrifuged at 10,000 × g for 10 min at 4 °C. The supernatant was then quantified for downstream immunoblotting. 20 μg of protein was loaded onto NuPAGE 12% Bis-Tris gel and then transferred onto a PVDF membrane. Blocking with 5% milk was performed for 30 min at room temperature, followed by overnight staining with primary antibody at 4 °C. Membranes were then washed with PBST three times and then stained with secondary HRP-conjugated antibody at room temperature for one hour. Membranes were then washed with PBST three times and then washed in Pierce ECL Western Blotting Substrate (Thermo Scientific) and then developed on film. Mouse anti-Cas9 (1:1000, Active Motif), mouse α-Tubulin (1:1000, Sigma), rabbit anti-GAPDH (1:1000, Cell Signaling), rabbit anti-LDLR (1:1000, Abcam), rabbit anti-Pcsk9 (1:1000, Abcam) antibodies were used.

RNA Isolation and RT-qPCR

RNA was harvested from cultured cells and from tissue using Trizol. Samples were homogenized in Trizol and then mixed with 100% ethanol and purified with the Direct-zol RNA Miniprep kit (Zymo). RNA was converted to cDNA using First-strand cDNA synthesis (Invitrogen SuperScript III). Quantitative PCR reactions were prepared with SYBR Green (Invitrogen), and performed in QuantStudio 3 (Life Technology). Primer information for RT-qPCR is listed in Supplementary Table 3.

RNA Sequencing

RNA from all cell lines and tissue was harvested by Trizol followed by Direct-zol RNA kit (Zymo). Library prep and sequencing were performed at Novogene. Approximately 30 M reads were obtained for each sample. Quality of reads was assessed using FastQC, and then adaptor sequences and low-quality reads were trimmed using Trim Galore (0.6.4). Trimmed reads were aligned to mm10 using STAR (2.5.0a)⁴² with default settings. DEGs were determined using DESeq2 (1.44) with cut offs of p-adjusted value < 0.05 and a fold change greater than 2. RNA-sequencing data was visualized as scatterplot of the Log2 counts per million or fragments per kilobase million between scrambled sgRNA samples and targeting sgRNA samples. 0.1 was added to each gene count to prevent the log transformation of a value of 0.

ChIP sequencing sample preparation

For ChIP-seq experiments, 10⁷ cells were harvested and cross-linked with 1% formaldehyde (10 min, RT). The fixation reaction was then quenched with 0.125 M glycine for 5 min. Samples were then washed twice with PBS and sonicated using a Covaris M220 Focused-Ultrasonicator with the parameters Peak Power: 75.0, Duty Factor: 10.0, Cycle/Burst: 200, Average Power: 7.5, and Duration: 20 min. Lysates were centrifuged for 10 min at 20,000 × g at 4 °C and used for downstream processing. 5% of the sonicated lysate was taken as input. For immunoprecipitation reactions, 1 μL of anti-Cas9 antibody (Active Motif 8C1-F10, 1 μg/μL) was added to lysates and incubated overnight at 4 °C. Protein G dynabeads were then added and incubated for 4 h at 4 °C and then sequentially washed with low salt buffer (150 mM NaCl, 0.1% SDS, 1% Triton-X, 1 mM EDTA, 50 mM Tris-HCl), high salt buffer (500 mM NaCl, 0.1% SDS, 1% Triton-X, 1 mM EDTA, 50 mM Tris-HCl), LiCl buffer (150 mM LiCl, 0.1% SDS, 1% Triton-X, 1 mM EDTA, 50 mM Tris-HCl), and then three times with TE + 50 mM NaCl. Immunoprecipitated samples and inputs were reverse crosslinked at 65 °C overnight, and then digested with RNase A (Qiagen) for 1 h at 37 °C and then Proteinase K for 1 h at 55 °C. DNA was then purified with Qiagen QIAquick PCR Purification kit. Library preparation was performed using the NEBNext Ultra II DNA Library Prep Kit (NEB) and sequenced at Novogene.

ChIP-seq analysis

Sequencing data was analyzed with a previously reported method⁴³ with modifications. In brief, adaptor sequences and low-quality reads were removed using Trim Galore. The remaining reads were mapped to mm10 reference genome using Bowtie2 (2.3.5.1) with default parameters⁴⁴. Sequencing reads were normalized across samples and peaks were called by MACS2(2.1.2)⁴⁵ with a q-value cutoff of 0.05. To identify binding sites for dCas9-DNMT3A and dCas9-TET1, samples with dCas9-DNMT3A/TET1 alone, dCas9-TET1/DNMT3A with scrambled sgRNA, and the mock sample were used as controls to call peaks. Peaks unique in the dCas9-DNMT3A/TET1 with target sgRNAs sample were then designated as candidate binding sites. Peaks were annotated to their nearest gene with the ChIPSeeker (1.46.1) package⁴⁶. Data were visualized with Integrated Genomics Viewer (IGV 2.17.4).

Whole genome bisulfite sequencing and analysis

Genomic DNA was extracted from liver tissue from dCas9-DNMT3A mice injected with Scr and Pcsk9 sgRNAs or mCherry+ nuclei sorted from dCas9-TET1 mice injected with Scr and Mecp2 sgRNAs and subject to sequencing library preparation using the Pico Methyl-Seq Library Prep Kit (Zymo Research, D5455). Three biological replicates were included for each group. Sequencing was performed by Novogene. Adaptor sequences and low-quality reads were removed by Trim Galore. The remaining high-quality reads were mapped to the mouse mm10 reference genome by Bismark (0.24.2) with the parameters “—non_directional –local –un –ambiguous –hisat2 -N 1 -p 2 -score_min L, −6, −0.3.” The differentially methylated CpG (DMC) was called by Methylkit (3.19) (49). DMCs were filtered according to the following criteria: each CpG site had a read coverage of at least 1, q value < 0.01 (adjusted P value for the False Discovery Rate), and a difference in DNA methylation percentage larger than 25%. A cutoff of 15% DNA methylation percentage difference was used for determining DMCs between Mecp2 and Scr edited nuclei as the presence of a unmethylated promoters on the Xa results in an average baseline methylation of X-linked gene promoters at approximately 50%. Based on the DMCs, differentially methylated regions (DMR) were determined by identifying regions that included at least 3 DMCs with a methylation change in the same direction and a distance between the DMCs being less than 250 bp. The genes annotated in the differentially methylated regions were defined as differentially methylated genes.

Bisulfite conversion, PyroPCR, and pyrosequencing

Genomic DNA was bisulfilte-converted using the EZ DNA Methylation-Gold Kit (Zymo). The resulting bisulfite-converted DNA was amplified using the Pyromark PCR kit (Qiagen). Pyrosequencing and analysis of all bisulfite-converted genomic DNA samples were performed with PyroMark Q48 Autoprep (Qiagen). Primer information for Pyroseq is listed in Supplementary Table 1.

Intracerebroventricular injection of AAVs expressing sgRNA and Cre

ICV injections were performed on female LSL-dC-T; Mecp2^+/− mice at post-natal day 0–3. Hypothermia induced anesthesia was induced through placement on ice for 30 s. Absence of motor activity was assessed to confirm anesthesia. 10¹² viral genomes (vg) of AAV9-U6-sgRNA-Cre-mCherry-KASH were then injected bilaterally via a Hamilton syringe (Hamilton Company, 65460-06).

Statistics and reproducibility

Statistical analyses were performed with GraphPad Prism 10.3.1. Two-sided Student’s t-test was used for comparisons of means between two groups. One-way ANOVA with Dunnett’s or Tukey’s multiple comparisons test was used for comparisons of means across multiple groups. Two-way repeated measures ANOVA was used for comparisons of cholesterol levels over time. No statistical method was used to predetermine sample size. Data were excluded if they did not meet pre-existing technical quality control metrics. The experimenters were blinded to allocation during experiments and outcome assessment. Representative micrographs and blots were performed at least twice with similar results.

Ethics

All animal experiments were performed in accordance with institutional guidelines and approved by the Institutional Animal Care and Use Committee (IUCAC) and Institute of Comparative Medicine (ICM) of Columbia University.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Author contributions

R.P. and X.S.L. conceived the idea for this project. R.P. and X.S.L. designed the experiments. R.P., J.R., X.C., L.F.F., R.V.L.G., A.A.A., and X.S.L. performed the experiments. B.L. cultured and provided the Mecp2 reporter cells. J.W. and J.H. packaged AAV9 virus. R.P., O.D., K.F., and X.S.L. interpreted the data, and wrote the manuscript with input from all the other authors.

Peer review

Peer review information

Nature Communications thanks Takuro Horii and the other anonymous, reviewer(s) for their contribution to the peer review of this work. A peer review file is available.

Data availability

The RNA-seq, ChIP-seq, and whole genome bisulfilte genome sequencing data generated as part of this study have been deposited in the Gene Expression Omnibus database under GSE280072, GSE280073, GSE280074. All other data can be found within the paper, supplementary materials, and source data. Source data are provided with this paper.

Competing interests

O.D. has equity in Epitor Therapeutics, Regel Therapeutics, Ajna Biosciences, Blackrock Neurotech, ConnectRN, Tevard Therapeutics, and PhiFund Ventures. K.F. is a SAB member for Epitor Therapeutics. X.S.L. is a co-founder of Epitor Therapeutics. The remaining authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s41467-025-67222-5.