CRISPR/dCas9-Dnmt3a-mediated targeted DNA methylation of APP rescues brain pathology in a mouse model of Alzheimer’s disease

A single guide RNA (sgRNA) was cloned with a pLenti-sgRNA (Addgene, Cambridge, MA, #71,409) vector. The vector was linearized with the BsmbI (NEB) enzyme at 37 °C for 2 h. Next, guide RNA oligonucleotide pairs were annealed using a T4 polynucleotide kinase (NEB), and 30 ng of the digested pLenti-sgRNA vector was ligated with the annealed oligonucleotides at room temperature (RT) for 2 h. The ligated vectors were transformed, and transformation was confirmed through enzyme digestion and sequence analysis (GACTATCATATGCTTACCGT was used as the primer).

All animal experiments were approved by the Institutional Animal Care and Use Committee at Dongguk University and performed in accordance with institutional guidelines. The mice used in this study were maintained under controlled conditions (12-h light/dark cycle at 22–23 °C). APP-KI (App^{NL-G-F/NL-G-F}; Swedish [NL], Beyreuther/Iberian [F], and Arctic [G] mutation knock-in) mice were purchased from the RIKEN Brain Science Institute [23]. Three-month-old male APP-KI mice and B6C57 mice were used for the experiments. For stereotaxic injection, the male mice were anesthetized with 120 mg/kg Avertin (2,2,2-tribromoethanol; Sigma, St. Louis, MO), and 10 µl of dCas9-Dnmt3a and empty vector (control) or APP -189 sgRNA lentivirus were microinjected into the dentate gyrus (DG) region of each hemisphere at the following coordinates: AP − 2 mm, ML ± 1.1 mm, and DV − 2 mm. One month after injection, the mice were subjected to behavioral tests and biochemical analysis. For the behavioral tests, Y-maze, fear conditioning, and water maze tests were conducted, and observations were recorded and analyzed using Noldus Ethovision XT 13 (Noldus, Netherlands). In the Y-maze test, the number of alternations was determined using three open-arm chambers. The mouse behavior was recorded for 10 min and the total number of spontaneous visits to each arm position was calculated. Fear conditioning tests were conducted for 2 consecutive days. On day 1 of training, the mice were placed in a fear conditioning chamber and allowed to explore for 3 min. Then, they were given pairs of regular stimuli and aversive unconditioned stimuli (1 s, 0.7 mA). After exposure to stimuli, the mice stayed for further 1 min for measurement of freezing behavior. On day 2, freezing behavior tests were conducted to evaluate the conditioned fear, and freezing behavior was recorded for 2 min. The water maze test was performed in a circular water tank. In the visible platform trial, the mice were trained for 4 days, with three trials conducted per day. In the invisible platform test on day 4, the time spent in each quadrant was recorded for 1 min. After the behaviorial test, the 4-month-old male mice were sacrificed for biochemical analysis.

Mouse primary fibroblasts (MEFs) and NIH/3T3 cells were cultured in Dulbecco's Modified Eagle Medium (Gibco, Waltham, MA) supplemented with heat-inactivated fetal bovine serum (Gibco) and 1% penicillin/streptomycin (Gibco). Primary neurons were obtained from APP-KI mice on embryonic day 14 (E14) and maintained in a neurobasal medium (Gibco) supplemented with heat-inactivated fetal bovine serum (Gibco), glutamine (Gibco), B-27™ supplement (Gibco), L-glutamine (Gibco), P/S (Gibco), and laminin (Corning®, Corning, NY), for 2 weeks followed by biochemical analysis. All cells were incubated at 37 °C in 5% CO₂. Cell lines were authenticated by STR (Short Tandem Repeat) analysis (Kogene Biotech, Seoul, Korea), and mycoplasma was tested using a MycoSensor PCR assay kit (Agilent, Santa Clara, CA). For lentivirus production, HEK293T cells were passaged and cultured at 80% confluency and transfected with lentiviral vectors (psPAX2, pMD2.G, and pLenti-sgRNA [Addgene, Cambridge, MA, #71409] or Fuw-dCas9-Dnmt3a [Addgene, #84476]) using calcium phosphate (Sigma) and HEPES (Sigma). The medium was replaced 24 h after transfection, and the virus was harvested by centrifugation after 72 h.

Potential off-target sequences were validated using the off-target site prediction software Cas-OFFinder (http://​www.​rgenome.​net/​cas-offinder) [24]. For validation of potential off-target sites, the top six predicted off-target sites with two or less mismatches compared to the on-target sequence were analyzed using quantitative real-time PCR analysis.

Cells and hippocampal brain tissues were fixed in 4% paraformaldehyde (Sigma) and washed with PBS buffer. The samples were blocked with PBST supplemented with 1% BSA for 20 min and incubated with anti-NeuN (Invitrogen, PA5-78639), anti-Tuj1 (Sigma, T8578), anti-Map2 (Thermo Fisher Scientific, Waltham, MA, 13-1500), or anti-ab42 (Abcam, Cambridge, UK, ab201060) at 4 °C overnight. Then they were washed with PBST, incubated with appropriate secondary antibodies at RT for 2 h, and counterstained with DAPI (Invitrogen). The stained samples were visualized under an LSM 700 confocal microscope (Zeiss, Oberkochen, Germany). The percentage of Aβ42+/Map2+ cells was calculated as follows: number of Aβ42- and Map2-double positive cells/number of Map2-positive cells × 100%.

RNA was extracted using an eCube tissue RNA mini kit (Philekorea, Seoul, Korea) in accordance with the manufacturer’s instructions. In brief, 1 μg of the prepared RNA was reverse transcribed using AccuPower® CycleScript RT PreMix (Bioneer, Daejeon, Korea). qRT-PCR was conducted using a Rotor-Gene Q real-time PCR cycler (Qiagen, Hilden, Germany) with suitable primer sets and AccuPower® PCR PreMix (Bioneer, Daejeon, Korea).

Total DNA was extracted using an eCube tissue DNA mini kit (Philekorea, Seoul, Korea), in accordance with the manufacturer’s instructions. In brief, 2 μg of DNA was modified with an EpiTect bisulfite kit (Qiagen, Hilden, Germany) in accordance with the manufacturer’s instructions. The bisulfite-converted DNA was then amplified with PCR primer sets designed via the PrimerSuite website (www.​primer-suite.​com). The amplified products were extracted using NucleoSpin® gel and a PCR clean-up kit (Macherey–Nagel, Düren, Germany) and cloned using a TA Cloning™ kit (Thermo Fisher Scientific) for sequencing.

Samples (cells and brain tissues) were added with RIPA (Sigma) and  1× proteinase inhibitor cocktail (Sigma), and levels of Aβ40 and Aβ42 were determined using Aβ (1–40) and Aβ (1–42) (FL) kits (IBL International, Hamburg, Germany), respectively. Absorbance at 450 nm was measured using a VERSAmax tunable microplate reader (Molecular Devices, San Jose, CA).

Statistical analyses were conducted using IBM SPSS Statistics (IBM Corp, Armonk, NY). Differences between groups were evaluated using one-way ANOVA and Student’s t-test. P < 0.05 was considered as statistically significant.

APP, the precursor of Aβ, plays a pivotal role in the development of AD [2]. Aβ is generated from APP through sequential cleavage by β-secretase, followed by the γ-secretase complex [1, 2]. In the brains of AD patients, aberrant CpG methylation of APP occurs, possibly contributing to the aberrant gene expression of APP. Targeted DNA methylation editing can be performed by fusing Tet1 or Dnmt3a with dCas9 [15]. Therefore, we evaluated whether the dCas9-Dnmt3a system could be used to alter the methylation level of the APP promoter locus (Fig. 1a). Initially, we designed three different sgRNAs targeting the APP promoter region at the most efficient binding sites for APP methylation (Fig. 1b and Additional file 1: Table S1), and tested the targeting efficacy of these sgRNAs with dCas9-Dnmt3a in NIH/3T3 cells. We found that the APP expression in NIH/3T3 cells treated with APP sgRNAs and dCas9-Dnmt3a was significantly decreased (Fig. 1c). Particularly, the transduction of the − 189 sgRNA targeting − 189 bp from the start codon of APP, with dCas9-Dnmt3a, led to the most effective reduction in APP expression (Fig. 1c–e). Thus, dCas9-Dnmt3a coupled with the − 189 sgRNA was used to edit APP DNA methylation in the following experiments.

We further confirmed the decrease in APP expression in the MEF cell line. The dCas9-Dnmt3a combined with the − 189 sgRNA efficiently decreased the mRNA expression of APP in MEF and NIH/3T3 cells (Fig. 1f). Moreover, we validated CpG methylation in the APP promoter region by bisulfite sequencing analysis. We observed a significant increase in CpG methylation of the APP promoter region in NIH/3T3 cells treated with − 189 sgRNA and dCas9-Dnmt3a (Fig. 1g, h). Additionally, we examined the off-target effects of dCas9-DNMT in treated cells. We analyzed six predicted off-target sites in the NIH/3T3 cells, but none of them were affected by sgRNA and dCas9-Dnmt3a (Fig. 1i and Additional file 1: Table S2). These results indicate that dCas9-Dnmt3a can efficiently edit the methylation of the APP promoter and thereby decrease APP expression in cultured mouse cells, without any obvious impact on other genes.

We examined whether the dCas9-Dnmt3a-mediated CpG methylation of APP could reduce Aβ generation and exert neuroprotection on APP-KI mouse primary neurons in vitro. Seven days after transduction of − 189 APP sgRNA and dCas9-Dnmt3a into the APP-KI mouse primary neurons, we observed a significant decrease in APP mRNA expression in the wild-type (WT) and APP-KI mouse primary neurons (Fig. 2a). We further verified that transduction of dCas9-Dnmt3a resulted in a significant reduction of APP protein level in the APP-KI mouse primary neurons (Fig. 2b, c). In addition, bisulfite sequencing of the promoter region of APP revealed that the cytosines of the CpG site at − 189 bp were highly methylated in the dCas9-Dnmt3a-treated primary neurons in which APP expression was suppressed (Fig. 2d, e).

Given that AD is characterized by deterioration of neuronal connectivity or the loss of dendritic spines of pyramidal cells [25, 26], we investigated the effects of dCas9-Dnmt3a-mediated APP methylation on neuronal connectivity and dendritic spines. The number of neurons and neurite length decreased in APP-KI mouse primary neurons, but these decreases were reversed by dCas9-Dnmt3a transduction (Fig. 2f–h). These results demonstrated that APP methylation by dCas9-Dnmt3a and APP sgRNA reduced the apoptosis and neurite dystrophy of APP-KI mouse primary neurons. Aβ peptide is a marker of AD, promoting the formation of amyloid fibrils that accumulate in senile plaques [1, 27]. Therefore, we determined the Aβ42 levels in primary neurons after transduction with dCas9-Dnmt3a. Results showed that the number of Aβ42+/Map2+ double-positive neurons significantly decreased in APP-KI mouse primary neurons transduced with dCas9-Dnmt3a (Fig. 2i, j). We also validated the decrease in the Aβ42/Aβ40 ratio in the dCas9-Dnmt3a-infected APP-KI mouse primary neurons (Fig. 2k). Our results suggest that APP methylation via dCas9-Dnmt3a can alleviate neuronal deficits in APP-KI mouse primary neurons.

To evaluate whether dCas9-Dnmt3a-mediated methylation editing could be used for in vivo APP methylation, we injected the lentivirus containing dCas9-Dnmt3a with APP sgRNA directly into the DG of mouse brain (Fig. 3a). The transduction of dCas9-Dnmt3a significantly decreased the mRNA expression and protein level of APP in the WT mouse brain (Fig. 3b, c).

We also tested whether dCas9-Dnmt3a could increase the level of APP methylation in the brain. Bisulfite sequencing showed that the demethylated CpG sites in the APP promoter were extensively methylated by dCas9-Dnmt3a (Fig. 3d, e). To further investigate the potential in vivo off-target effects of dCas9-Dnmt3a in the brain, predicted off-target sites were tested (Fig. 3f). The results indicated that the − 189 APP sgRNA had no off-target effects (Fig. 3f). Additionally, we verified the effect on APP expression levels in APP-KI mice. After injection of dCas9-Dnmt3a with APP sgRNA into the APP-KI mouse brain, APP expression diminished to a similar level as that in the WT mouse brain (Fig. 3g, h). These findings indicate that dCas9-Dnmt3a induces efficient DNA methylation editing in the brain.

Aβ accumulation is associated with neurodegeneration and cognitive decline [2]. Thus, we tested whether APP methylation via dCas9-Dnmt3a in the mouse brain could decrease Aβ (1–42) and reduce cognitive and memory impairments in the APP-KI mouse model. We injected dCas9-Dnmt3a into the DG region of the APP-KI mouse brain and conducted biochemical and behavioral analyses 4 weeks after injection (Fig. 4a). Results showed decreased APP expression in the DG of dCas9-Dnmt3a-injected APP-KI mice compared with that in the DG of age-matched APP-KI control mice (Fig. 3g, h). We validated the methylation range at the APP promoter locus in the brain through bisulfite sequencing, and found that the demethylated CpG sites in the APP promoter were extensively methylated by dCas9-Dnmt3a compared with those in age-matched APP-KI control mice (Additional file 1: Fig. S1a, b).

To investigate the effects of APP methylation via dCas9-Dnmt3a on AD pathologies, we examined Aβ formation in the APP-KI mouse model. Four weeks after injection, dCas9-Dnmt3a significantly reduced the level of full-length APP compared with that in the age-matched APP-KI control mice. In addition, the level of β-secretase-derived APP C-terminal fragment, specifically C99, which generates Aβ peptides, was decreased in the dCas9-Dnmt3a-treated APP-KI mice (Fig. 4b, c; Additional file 1: Fig. S2). Moreover, dCas9-Dnmt3a significantly reduced the Aβ42/40 ratio in the hippocampus of APP-KI mice (Fig. 4d). Consistent with these results, Aβ42 plaque accumulation in the hippocampus and cortex of APP-KI mice was significantly decreased by dCas9-Dnmt3a (Fig. 4e, f). These results indicate that the dCas9-Dnmt3a-mediated APP methylation attenuated APP expression, resulting in reduced Aβ levels in APP-KI mice.

Next, behavioral tests were performed to determine whether cognitive behaviors were affected by APP methylation via dCas9-Dnmt3a. First, the Y-maze test was performed to evaluate the spatial working memory of APP-KI mice. The percentage of alternation decreased in the APP-KI mice compared to the WT mice,  but there was no significant difference between the dCas9-Dnmt3a-injected APP-KI mice and the age-matched APP-KI control mice (Fig. 4g). Next, short-term memory was examined using the contextual fear conditioning test. Interestingly, after 24 h of training, the freezing levels of the dCas9-Dnmt3a-injected APP-KI mice were significantly increased (Fig. 4h). To further investigate the spatial reference memory and working memory, the Morris water maze test was performed. The escape latency and distance of movement showed no significant difference between the control and dCas9-Dnmt3a-injected groups in the visible platform trials (Fig. 4i). In the test without the platform, the dCas9-Dnmt3a-injected APP-KI mice spent significantly more time in the target quadrant than the age-matched APP-KI control mice (Fig. 4j). These findings showed that dCas9-Dnmt3a prevented decline in learning ability and improved memory retention in APP-KI mice.