Ambryon Biosciences

Efficient prime editing in mouse brain, liver and heart with dual AAVs

Davis et al. (2023). Nat Biotechnol·Citations: 228Research

Prime editors exceed AAV packaging capacity, and earlier intein-split AAV delivery reached only 1.7–13.5% editing in bulk tissue. This paper optimized the split PE protein, epegRNA and AAV elements into v1em and v3em PE-AAV dual-AAV systems. These reached up to 42% prime editing in mouse cortex, 46% in liver and 11% in heart.

Key findings

  • Eight intein-split PE2 designs recover 37–96% of full-length PE2 editing in HEK293T cells; 1024-CFN and 844-CFN (Cys-Phe-Asn extein) are robust, and activity requires a catalytically active intein.

  • v3em PE3-AAV9 gives 46% (10^12 vg) and 14% (10^11 vg) liver editing versus 5.7% and 0.1% for v1em, consistent with higher Cbh-driven expression; 844-CFN and 1024-CFN splits perform similarly.

  • v1em and v3em PE3-AAV9 give similar bulk cortex editing (41% and 42%) in P0 brain; v1em is higher in GFP+ nuclei (81% versus 69%), and PE3 beats PE2 by 9.8-fold for +1 C-to-G.

  • Systemic PHP.eB PE3-AAV (1 × 10^12 vg) gives 14% bulk and 30% GFP+ cortex editing with v1em and 11% and 22% with v3em in adult mice, with 0.5% indels.

  • v3em PE3-AAV9 installs APOE3 R136S in humanized APOE3 mouse brain at 12% (neocortex) and 14% (hippocampus) after P1 injection, 8.2% and 7.1% after P3, with 9.4–11% in APOE cDNA.

  • v3em PE3-AAV9 installs Pcsk9 Q155H in 31–39% of adult mouse liver, with MMR-evading silent edits modestly higher (P = 0.06), lowering total cholesterol about 20% and LDL cholesterol 27%, persisting to 8 weeks.

  • v1 PE3-AAV9 with W3 in the 3′ UTR gives 2.8% (bulk) and 4.7% (GFP+ nuclei) Dnmt1 +5 G-to-T editing in P0 mouse cortex, versus ≤0.2% without W3.

  • v1 PE3-AAV9 given systemically at 1 × 10^12 vg yields ≤0.1–0.9% editing in liver and none in heart or muscle, and a longer 6-week expression time does not appear to significantly increase editing (≤1.1%).

  • Cbh promoter in triple-AAV v2e PE3-AAV9 gives 48% liver and 7.2% heart prime editing, 27- and 70-fold above EFS-driven v1e, with under 1% in muscle.

  • PEmax architecture (v2em) leaves liver editing similar but raises heart editing 2.4-fold (P = 0.017), and with epegRNA gives 41% editing in P0 cortex, 6.8-fold above PE3 with pegRNA.

  • Dual-AAV v3em PE3-AAV9 (Cbh, ΔRNaseH, SV40 late poly(A)) matches triple-AAV v2em in liver (35% versus 36%) and gives 2.1-fold more heart editing (9.1%, P = 0.026) and 4.9-fold more muscle editing (1.3%, not significant).

Author abstract

Realizing the promise of prime editing for the study and treatment of genetic disorders requires efficient methods for delivering prime editors (PEs) in vivo. Here we describe the identification of bottlenecks limiting adeno-associated virus (AAV)-mediated prime editing in vivo and the development of AAV-PE vectors with increased PE expression, prime editing guide RNA stability and modulation of DNA repair. The resulting dual-AAV systems, v1em and v3em PE-AAV, enable therapeutically relevant prime editing in mouse brain (up to 42% efficiency in cortex), liver (up to 46%) and heart (up to 11%). We apply these systems to install putative protective mutations in vivo for Alzheimer's disease in astrocytes and for coronary artery disease in hepatocytes. In vivo prime editing with v3em PE-AAV caused no detectable off-target effects or significant changes in liver enzymes or histology. Optimized PE-AAV systems support the highest unenriched levels of in vivo prime editing reported to date, facilitating the study and potential treatment of diseases with a genetic component.

Identifiers & revision