Targeted DNA methylation editing in vivo

This study presents the development and validation of three Cre-dependent CRISPR-based mouse lines that enable inducible, locus-specific DNA methylation editing in vivo, successfully demonstrating causal links between methylation and gene expression at specific genomic loci in both ex vivo myeloid cells and in vivo neurons.

The Gist

The Big Picture: Editing the "Software" Without Changing the "Hardware"

Imagine your body's DNA as a massive library of instruction manuals (the hardware). Every cell has a copy of these manuals. Sometimes, a specific page in the manual gets covered in sticky notes or highlighted with a marker. This doesn't change the words on the page, but it tells the cell, "Ignore this page" or "Read this page loudly." This is called DNA methylation, and it's a form of epigenetics (the "software" running on the hardware).

Scientists have long known that these "sticky notes" are linked to diseases like cancer, autoimmune disorders, and brain conditions. But there's a huge problem: We don't know if the sticky notes cause the disease, or if the disease just happens to leave sticky notes behind. It's like seeing a puddle on the floor and not knowing if the leak caused the puddle, or if someone spilled water and the puddle is just a result.

To solve this, scientists need a tool to add or remove sticky notes on specific pages to see what happens. This paper describes how a team in Sweden built a new "sticky note robot" that works inside living mice.

---

The Tool: A "Cre-Dependent" Robot

The researchers built three new types of mice equipped with a special molecular machine. Think of this machine as a GPS-guided robotic arm that carries a marker pen.

1. The Robot (dCas9-DNMT3A): This is a modified version of the famous CRISPR gene-editing tool. Instead of cutting the DNA (which would break the manual), this robot is "dead" (dCas9). It can't cut, but it can still stick to a specific location. Attached to it is a marker pen (DNMT3A) that can write "silence" (methylation) on the DNA.
2. The Safety Switch (Cre): To make sure the robot doesn't go crazy and mark random pages, the researchers put a safety lock on it. The robot is hidden behind a "Stop Sign" (a genetic stop codon).
   • The Key (Cre): They created a key called "Cre." When they introduce the Cre key, it removes the Stop Sign, and the robot wakes up.
   • The Two Models:
     • The "Always-On" Mouse: In one model, the key is always present. The robot is active everywhere, all the time.
     • The "Remote-Control" Mouse: In the other model, the robot is asleep until the researchers inject a chemical (Tamoxifen) that acts as the key. This lets them turn the robot on at a specific time.

How They Tested It: The "Target Practice"

The team tested this robot in two different "neighborhoods" of the mouse body: the Immune System (blood cells) and the Brain.

1. The Immune System (Ex Vivo)

They took bone marrow cells from the mice and grew them in a dish (like a petri dish).

• The Target: They aimed the robot at two genes: one for immune defense (H2-Ab1) and one for inflammation (Il6).
• The Result: The robot successfully wrote "sticky notes" on the DNA, increasing methylation by a lot (up to 60% in some spots).
• The Surprise: Even though they successfully marked the pages, the genes didn't change their behavior. The immune cells didn't stop working or start screaming.
• The Lesson: Just because you can put a sticky note on a page doesn't mean the cell will listen. Some pages are "locked" by other factors, or the sticky note needs to be in a very specific spot to work. This proves that you can't assume a link between a sticky note and a disease just by looking at it; you have to test it.

2. The Brain (In Vivo)

This was the big test. They injected a virus into the brains of the mice. This virus carried the "key" (Cre) and a map (guide RNA) to tell the robot exactly where to go.

• The Target: They targeted the Cnr1 gene, which controls a receptor involved in pain, mood, and memory (the cannabinoid receptor).
• The Result: The robot went to the brain cells, found the Cnr1 gene, and put sticky notes on it.
• The Outcome: This time, it worked! The sticky notes successfully told the cell to turn down the volume. The production of the cannabinoid receptor dropped by 25%.
• The Lesson: In the brain, this tool can successfully silence a gene. This is huge for studying brain diseases where specific genes might be overactive.

The "Mosaic" Problem: A Patchwork Quilt

One interesting finding was that the robot didn't work on every cell equally.

• Imagine a patchwork quilt where some squares are bright and others are dim.
• In the mice, some brain cells had the robot fully active (bright), while others had it barely active (dim).
• This "mosaic" pattern is a limitation. It means the researchers have to be careful when analyzing results because the effect is a mix of "edited" and "un-edited" cells.

Why Does This Matter?

This paper is like handing scientists a scalpel for the mind and immune system.

1. Proving Causation: Before this, we could only guess if a sticky note caused a disease. Now, we can add the note and see if the disease appears (or disappears).
2. Brain Research: Since the tool worked in the brain, it opens the door to treating neurological disorders (like Alzheimer's or addiction) by turning off specific "bad" genes without changing the DNA code itself.
3. Understanding Complexity: The fact that it worked in the brain but not the immune cells teaches us that biology is complex. A tool that works in one tissue might not work in another, and we need to study each case individually.

The Bottom Line

The researchers built a living mouse model that acts like a remote-controlled pen, allowing them to draw "silence" on specific genes inside a living animal. While it didn't work perfectly everywhere, it proved that we can now test the cause-and-effect relationship of DNA methylation in real-time. This is a major step toward understanding how our "software" (epigenetics) controls our health and how we might fix it when it goes wrong.

Technical Summary

1. Problem Statement

While Epigenome-Wide Association Studies (EWAS) have increasingly linked specific CpG DNA methylation sites to diseases, traits, and environmental exposures, establishing causality remains a significant challenge. Current epigenome editing tools (e.g., dCas9 fused to DNMT3A or TET1) are well-established in vitro, but generating robust in vivo models for targeted DNA methylation deposition is difficult.

• Limitations of existing methods: Viral delivery (AAV/Lentivirus) of large dCas9-effector constructs is limited by cargo size, variable transduction efficiency across tissues, and the inability to achieve stable, long-term expression required for maintaining methylation changes.
• Knowledge Gap: There is a lack of versatile, transgenic mouse lines that allow for locus-specific, inducible, or constitutive DNA methylation editing to test the functional impact of methylation on gene expression in complex physiological contexts (brain and immune systems).

2. Methodology

The authors developed and characterized three distinct Cre-dependent CRISPR-based mouse lines to enable targeted DNA methylation deposition.

• Mouse Line Generation:
  • Construct Design: A 10,669 bp cassette containing the human DNMT3A catalytic domain fused to dCas9 (dCas9-DNMT3A) and an EGFP reporter, separated by a T2A peptide, was inserted into the Rosa26 locus. The cassette is flanked by loxP sites and preceded by a double-floxed stop codon, making expression Cre-dependent.
  • Three Strains Created:
    1. Inducible (AAV-mediated): The original Rosa26-dCas9-DNMT3A line, activated by viral delivery of Cre (AAV1/AAV2 for CNS, AAV9 for systemic).
    2. Constitutive: Crossed with ACTB-Cre mice (human $\beta$-actin promoter) for ubiquitous, permanent expression.
    3. Inducible (Tamoxifen): Crossed with CAG-CreERTM mice for time-controlled, tamoxifen-inducible expression.
• Targeting Strategy:
  • In Vivo (Brain): Stereotaxic injection of AAVs expressing specific gRNAs (targeting Cnr1) and Cre (or Cre-mCherry) into the dorsolateral striatum of mice.
  • Ex Vivo (Immune): Bone marrow-derived macrophages (BMM) and dendritic cells (BMDC) were cultured and transduced with lentiviruses expressing gRNAs targeting H2-Ab1 (MHC Class II) and Il6.
• Validation Techniques:
  • Methylation Analysis: Pyrosequencing for locus-specific quantification; Illumina Infinium Mouse Methylation BeadChip (285k CpGs) for genome-wide off-target analysis.
  • Expression Analysis: RT-qPCR, RNAscope (single-cell resolution), and immunofluorescence.
  • Phenotyping: Flow cytometry for immune cell subsets; behavioral and weight monitoring for toxicity.

3. Key Contributions

• Novel Transgenic Resources: The generation of three new mouse lines (Constitutive, Tamoxifen-inducible, and AAV-inducible) that enable precise, locus-specific DNA methylation editing in vivo and ex vivo.
• Demonstration of Locus-Dependency: The study provides empirical evidence that the causal effect of DNA methylation on gene expression is locus-dependent. Methylation did not universally silence genes; its impact varied significantly between H2-Ab1, Il6, and Cnr1.
• Technical Optimization: Successful adaptation of the dCas9-DNMT3A system for use in primary immune cells and specific neuronal populations, overcoming previous hurdles related to transgene size and delivery.

4. Key Results

A. Mouse Line Characterization

• Induction Efficiency: Viral delivery of Cre (AAV1/2 in brain, AAV9 systemically) successfully induced dCas9-DNMT3A expression. AAV9 showed high induction in liver and heart but failed to induce expression in the brain or spleen.
• Mosaic Expression: Both constitutive (ACTB-Cre) and inducible (CAG-CreERTM) lines exhibited mosaic expression patterns.
  • In the brain, expression was highest in neurons (NeuN+), with EGFP-high and EGFP-dim populations.
  • In the immune system, expression varied widely across myeloid and lymphoid subsets.
  • Phenotype: Constitutive lines showed breeding difficulties, hydrocephalus, and reduced weight gain, suggesting potential toxicity of continuous dCas9-DNMT3A expression during development.

B. Genome-Wide Off-Target Effects

• Minimal Off-Target Methylation: Whole-genome methylation profiling (Illumina arrays) in brain and spleen revealed very few differentially methylated positions (DMPs) between edited and control mice.
  • Brain: 28 DMPs (dCas9 vs. WT) and 15 DMPs (Inducible vs. Control).
  • Spleen: 7 DMPs (dCas9 vs. WT) and 31 DMPs (Inducible vs. Control).
• Specificity: Most off-target changes were hypermethylated in the inducible lines, mapping to specific gene promoters, but the overall effect size was small, indicating high specificity of the editing tool.

C. Locus-Specific Editing Outcomes

1. Immune Genes (Ex Vivo):

   • H2-Ab1: Targeted methylation of promoter CpGs resulted in a robust increase in methylation (up to 56% at specific CpGs) but no significant change in gene expression.
   • Il6: Targeted methylation of an exonic CpG (+286) increased methylation from ~8% to ~60% but did not alter Il6 expression levels, even after LPS stimulation.
   • Conclusion: In primary myeloid cells, methylation at these specific sites is not sufficient to drive transcriptional silencing, highlighting the complexity of immune gene regulation.

2. Neuronal Gene (In Vivo):

   • Cnr1 (Cannabinoid Receptor 1): Targeted methylation of the Cnr1 promoter in the striatum using AAV-delivered gRNAs.
   • Bulk Analysis: Failed to show significant changes (likely due to heterogeneity in transduction efficiency).
   • Single-Cell Analysis (RNAscope): Revealed a significant 25% reduction in Cnr1 expression specifically in mCherry-positive (transduced) striatal neurons.
   • Conclusion: Successful causal link established between promoter methylation and gene downregulation in a specific neuronal context.

5. Significance and Implications

• Causality in Epigenetics: The study underscores that DNA methylation is not a universal "off switch." Its regulatory impact is highly context-dependent (cell type, genomic location, and chromatin state). Tools like these mouse lines are essential for distinguishing causal methylation sites from passive markers in disease associations.
• Therapeutic Potential: The ability to downregulate Cnr1 in specific neurons suggests a potential pathway for treating neuropsychiatric disorders where CNR1 dysregulation is implicated. Similarly, the ability to edit immune genes ex vivo opens doors for engineering immune cells for autoimmune therapies.
• Refinement of Tools: The identification of mosaic expression and developmental toxicity in constitutive lines highlights the need for inducible systems (like the Tamoxifen or AAV-Cre models) to minimize off-target effects and cytotoxicity.
• Future Directions: The authors propose combining these mouse lines with gRNA library screens and optimized delivery (e.g., doxycycline-inducible Cre drivers) to systematically map functional CpGs across the genome in disease models.

In summary, this paper provides a critical technical advancement in epigenome editing by validating a versatile mouse platform that successfully demonstrates that targeted DNA methylation can causally regulate gene expression, but only at specific loci and in specific cellular contexts.