Harvard, Broad team designs PE8 prime editors with AI boost

Harvard and Broad Institute researchers have pushed prime editing forward by redesigning the editor protein itself, not just the guide or target. David R. Liu’s group built PE8 variants from PE6 with structure-informed AI, including ProteinMPNN, and kept the catalytic core intact while overhauling reverse transcriptase domains to improve how the machine folds, expresses, and works in hard-to-edit settings.

The study, titled AI-guided redesign of laboratory-evolved reverse transcriptases enhances prime editing, appeared as a 2026 Nature Biotechnology paper from Harvard University and the Broad Institute of MIT and Harvard. The authors listed include Allen Tao, Holt Sakai, Allen Jiang, Nicholas Krasnow, Vasilii Vaganov, Brian Shim, Zachary Barsdale, Smriti Pandey, Nouraiz Ahmed, Man Na, Ting-Wei Liao, Keyede Oye, Ana Cristian, Emily Zhang, Joy Xu, Mattijs Bulcaen, and David Liu. The work extends the Broad group’s earlier PE6 advance, which had already been developed through directed evolution as a smaller and more efficient therapeutic prime editor.

What makes PE8 notable is how aggressively the protein was remodeled. The redesigned reverse transcriptases carry 30 to 163 amino acid substitutions, yet the team still preserved the regions needed for catalysis. Those changes improved folding stability and soluble expression, and after mRNA delivery they produced up to twofold higher intracellular prime editor protein levels. That matters because prime editors often live or die by how much usable protein actually reaches the cell after delivery, especially in primary cells where expression can be unforgiving.

In ex vivo tests, the redesigned PE8 editors improved editing across several human primary cell types and across multiple delivery modalities, signaling a broader operating window than earlier generations. In mice, the best PE8 results reached up to 2.9-fold higher editing efficiency than state-of-the-art PE6, PE7, and PEmax editors. For therapeutic development, that combination of higher activity, better protein behavior, and compatibility with mRNA delivery points to a more practical editor platform, one that could reduce the amount of material needed and widen the delivery options for future in vivo and ex vivo programs.

The Broad’s latest prime-editing push makes a clear argument: the next leap in editing may come from protein engineering as much as from guide design. By tuning the reverse transcriptase into a more stable, more soluble, and more productive enzyme, the Harvard and Broad team turned PE8 into a stronger candidate for real-world therapeutic development.