AI Redesigns Gene Editor Proteins to Boost Prime Editing Efficiency Nearly 3x
Researchers used AI protein design to rebuild prime editor enzymes, achieving up to 2.9-fold better genome editing in mice.
Summary
Prime editing is a precise gene-editing technology that can correct disease-causing DNA mutations without cutting both strands of DNA. Scientists at the Broad Institute used an AI tool called ProteinMPNN to redesign the reverse transcriptase component of prime editors — the molecular machinery that writes new genetic sequences. Previous lab-evolved versions of this enzyme were more active but unstable and poorly expressed inside cells. The AI-redesigned versions introduced dozens to over 150 amino acid changes while preserving the catalytic core, resulting in more stable, more abundant proteins. When tested in human primary cells and delivered into living mice, these redesigned PE8 editors achieved up to 2.9-fold higher editing efficiency than the current best-in-class prime editors. This advance offers a broadly applicable strategy for improving gene therapy tools.
Detailed Summary
Prime editing holds enormous promise for correcting genetic diseases by precisely rewriting DNA sequences without creating double-strand breaks. However, optimizing prime editors through traditional lab evolution has hit a wall: mutations that improve enzymatic activity often destabilize the protein and reduce how much functional enzyme gets made inside cells, undermining real-world performance.
Researchers from David Liu's lab at the Broad Institute tackled this problem by applying AI-guided protein design. Using ProteinMPNN, a structure-informed inverse-folding neural network, they systematically redesigned the reverse transcriptase (RT) domains of already-evolved prime editors. The goal was to find amino acid sequences that fold stably into the correct three-dimensional structure while preserving the catalytic regions essential for function. The redesigned enzymes carried between 30 and 163 amino acid substitutions compared to their evolved predecessors.
The results were striking. Redesigned RT variants showed improved folding stability and soluble expression, and delivered up to twofold higher intracellular prime editor protein levels when introduced via mRNA delivery. In multiple human primary cell types and across several delivery modalities, the new PE8 editors consistently outperformed prior versions.
In mouse studies, the redesigned PE8 editors achieved editing efficiencies up to 2.9-fold higher than state-of-the-art PE6, PE7, and PEmax editors — the current benchmarks in the field. This represents a meaningful leap for in vivo gene therapy applications.
The broader implication is that AI-guided protein redesign can serve as a generalizable second layer on top of laboratory evolution, rescuing stability and expression losses that otherwise limit performance. This pipeline could accelerate development of next-generation gene editors for treating inherited diseases. Caveats include limited public data beyond the abstract and unresolved questions about long-term safety and off-target editing profiles.
Key Findings
- AI-redesigned reverse transcriptases carried 30–163 amino acid substitutions with improved folding stability and soluble expression.
- Intracellular prime editor protein levels increased up to twofold after mRNA delivery with redesigned enzymes.
- PE8 editors achieved up to 2.9-fold higher editing efficiency in mice versus leading PE6, PE7, and PEmax editors.
- Enhanced performance was demonstrated across multiple human primary cell types and delivery methods.
- AI-guided redesign offers a generalizable strategy to rescue stability losses caused by laboratory evolution.
Methodology
The study used ProteinMPNN, a structure-informed inverse-folding AI network, to redesign RT domains of evolved prime editors while preserving catalytic residues. Performance was evaluated in human primary cells ex vivo across multiple delivery modalities and in mouse models in vivo. Comparisons were made against current benchmark prime editors PE6, PE7, and PEmax.
Study Limitations
This summary is based on the abstract only, as the full paper is not open access. Long-term safety, off-target editing profiles, and immune responses to the redesigned proteins have not been publicly detailed. Commercial conflicts of interest exist, as several authors are affiliated with genome editing companies.
Enjoyed this summary?
Get the latest longevity research delivered to your inbox every week.