Adding a Key Inhibitory Protein to Virus-Like Particles Improves Cytosine Base-Editing Efficiency in Mice

Cytosine base editing can rewrite C as T at specific sites without cutting both strands of DNA. Once inside an animal, however, the challenge is not only “how to reach the target,” but also whether every component required for editing can arrive simultaneously and remain active at sufficient levels. A team comprising researchers from ShanghaiTech University, Fudan University, and other institutions has now traced poor efficiency to an easily overlooked gap: insufficient uracil DNA glycosylase inhibitor protein carried by virus-like particles.

Cytosine editors first convert cytosine in DNA into uracil, after which cellular repair processes complete the C-to-T substitution. However, the cell’s own glycosylases, including UNG and SMUG1, remove uracil, not only counteracting the intended editing but also potentially increasing byproducts such as C-to-A or C-to-G conversions. The researchers found that simultaneously knocking out UNG and SMUG1 in human cells increased C-to-T editing and reduced other conversions, supporting the conclusion that these enzymes are a limiting factor in the virus-like particle delivery system.

Based on these findings, the team modified the editor and packaging method by adding extra aptamers to the RNA to recruit more of the uracil DNA glycosylase inhibitor protein UGI, while also increasing guide RNA loading. The final version, tBE-VLP4, worked in multiple human cell types and monkey kidney cells. It was also compatible with the Cas9 SpG variant, which has a broader recognition range, and with conventional cytosine editor architectures, indicating that the design may not be limited to a single editor.

Mouse experiments provided preliminary functional evidence. After a single tail-vein injection, an average of 46.0% C-to-T editing was achieved at the hepatic Pcsk9 site, followed by reductions in serum PCSK9 protein and total cholesterol. In a model of hereditary tyrosinemia type I, the reported editing rate at the Hpd site was 64.2%, with improvements in weight loss, liver injury, and mortality. Following subretinal injection, the average editing rate at the Vegfa site was 24.2%; laser-induced choroidal neovascular lesions were reduced, and retinal function was protected.

### Background

Virus-like particles do not contain a self-replicating viral genome and can transiently deliver editors as protein–RNA complexes or messenger RNA. Compared with adeno-associated viral vectors, which may continuously express editors, this characteristic of “gradually dissipating after delivery” could shorten the time window for off-target effects. The trade-off is that the amount of auxiliary components such as UGI that can be loaded at one time is limited. The key advance of this study was enabling transient delivery to carry sufficient inhibitory protein to complete the reaction.

The research team stated that no obvious off-target DNA or RNA editing was detected in the tested cell and mouse experiments, and that editing specificity was higher than with AAV or lipid nanoparticle delivery of mRNA. However, failure to detect events does not mean that no risk exists. The current results remain limited to specific sites, tissues, and animal models, and questions concerning immune responses, repeat dosing, long-term toxicity, distribution across different tissues, and consistency in large-scale manufacturing have yet to be answered.

The platform remains at the preclinical research stage. Zhengxu Biosciences, which participated in the study, stated that it is exploring in vivo hematopoietic stem cell editing and applications including thalassemia and sickle cell disease. Some authors are scientific co-founders or consultants of the company, and related patent applications have also been filed. These interests and subsequent development plans make independent replication, validation in large animals, and comprehensive safety assessment indispensable steps before proceeding to human trials.