Delivery: Protein-delivered base editor preserves embryo development

Delivery format emerged as the study’s central determinant of editing performance and embryo development. When ABE8.8-m was injected as mRNA into two-pronuclear human zygotes, 38 of 44 embryos arrested at the one- to four-cell stage. Arrest also occurred without guide RNA but was reduced when the deaminase was inactivated, implicating prolonged, guide-independent enzyme activity rather than on-target editing.

By contrast, ABEmax or ABE8e-V106W delivered as preassembled protein-guide RNA complexes supported blastocyst development at rates similar to controls. This format provides immediate but transient activity, thereby limiting exposure to the deaminase. ABE8e-V106W delivered during fertilisation edited all 32 PCSK9 target alleles examined across five embryos, which produced homozygously edited embryonic stem-cell lines lacking detectable PCSK9 protein.

No short or large on-target deletions were detected, although one of 55 base-edited embryos showed target-associated chromosome breakage, and bystander and off-target edits remained mosaic. Protein delivery therefore largely separated efficient on-target editing from developmental toxicity but did not resolve specificity or safety concerns. A caveat is that the human experiments compared different editor variants, although matched mouse experiments also found protein delivery less harmful than mRNA delivery.

The study was led by Stepan Jerabek and Dieter Egli at Columbia University. It was published in Nature on 9 September 2026.