VLP Base Editing Powered Up: Enhanced UDG Inhibition Boosts In Vivo C-to-T Efficiency to 64%

Virus-like particles (VLPs) hold great promise for delivering genome editors; however, VLP-mediated cytosine base editing in vivo has historically suffered from limited efficacy.

Recently, researchers published a study online in Nature Biotechnology titled "Efficient in vivo cytosine base editing using virus-like particles with uracil DNA glycosylase inhibition." The study revealed that insufficient inhibition of uracil DNA glycosylase is a key underlying mechanism driving the low in vivo editing efficiency of cytosine base editors (CBEs). To address this, the authors engineered a previously reported CBE—the transformer base editor (tBE)—and developed a tailored VLP delivery system designed to enhance the recruitment of uracil DNA glycosylase inhibitor proteins.

Figure 1. In vitro and in vivo comparison of CBE and ABE efficiency using canonical VLP-mediated delivery. (Zhu J, et al., 2026)

The engineered tBE-VLPs achieved highly efficient in vivo C-to-T editing in the mouse liver and retina. Following a single injection, the average editing efficiency reached 46.0% at the mPcsk9 locus and 64.2% at the mHpd locus in the liver, while achieving 24.2% at the mVegfa locus in the retinal pigment epithelium (RPE), yielding profound therapeutic efficacy in mouse disease models.

Notably, tBE-VLP4 exhibited no detectable off-target editing either in vitro or in vivo, demonstrating superior specificity compared with adeno-associated virus (AAV) or lipid nanoparticle (LNP)-mRNA delivery platforms. The authors' work firmly establishes tBE-VLP4 as a precise and robust system for in vivo cytosine base editing.

Adeno-associated viruses (AAVs) are widely used to deliver genome editors but suffer from several notable limitations, including restricted cargo capacity, persistent expression of editing enzymes, and potential risks of genomic integration. Lipid nanoparticles (LNPs) offer an alternative strategy by enabling transient delivery of genome editors in the form of RNA (such as gRNA and mRNA); however, their editing efficiency remains largely confined to the liver, restricting broader application in extrahepatic tissues. In contrast, virus-like particles (VLPs) have emerged as a highly promising therapeutic delivery system capable of transiently delivering genome editors as ribonucleoprotein (RNP) complexes with broader tissue tropism.

The vast majority of hereditary human diseases are caused by base substitutions, particularly A-to-G (or T-to-C on the complementary strand) and C-to-T (or G-to-A on the complementary strand) transitions. To correct these point mutations, researchers developed cytosine base editors (CBEs) and adenine base editors (ABEs), which accomplish precise C-to-T and A-to-G transitions, respectively, without inducing double-strand DNA breaks (DSBs).

Prime editors (PEs) have further expanded the scope of genome editing by enabling all possible base substitutions as well as small insertions and deletions, though their efficiency remains suboptimal at certain target loci. Previous studies have demonstrated the feasibility of VLP-mediated in vivo delivery for ABEs and PEs. While VLP-mediated CBE editing was recently reported in vitro, whether VLP-based CBE delivery could achieve clinically meaningful editing in vivo remained unclear, posing a bottleneck to unlocking the full therapeutic potential of base editing.

Overcoming the UDG Barrier with Modular Engineering

Previously, the authors developed a high-precision CBE known as the transformer base editor (tBE) by harnessing a deoxycytidine deaminase inhibitor (dCDI) protein. Unlike classic base editors that typically exist as a single fusion protein, tBE is a modular system composed of four distinct protein components:

1. A Cas9 nickase D10A (nCas9);
2. An N-terminal fragment of the Tobacco Etch Virus (TEV) protease (TEVn);
3. A fusion protein combining an RNA aptamer-binding protein (N22p) with the C-terminal fragment of TEV protease (TEVc);
4. A multifunctional fusion protein integrating a cytidine deaminase (mouse APOBEC3 cytidine deaminase domain 1, mA3CDA1), a dCDI derived from mouse APOBEC3 (mA3dCDI), a uracil DNA glycosylase inhibitor (UGI), and another RNA aptamer-binding protein (phage MS2 coat protein, MCP).

Additionally, the system utilizes two guide RNAs: a primary sgRNA containing a boxB aptamer (msgRNA_boxB) and a helper sgRNA containing an MS2 aptamer (hsgRNA_MS2). This modular architecture tightly controls deaminase activity: mA3dCDI suppresses mA3CDA1 and is specifically cleaved by TEV protease only at the target site. Consequently, tBE achieves highly efficient on-target C-to-T editing while minimizing both gRNA-dependent and gRNA-independent off-target events.

In the present study, the authors systematically compared the in vivo editing efficiencies of ABE and tBE delivered via VLPs. While VLP-mediated ABE delivery yielded robust A-to-G editing, tBE delivery resulted in undetectable C-to-T editing in vivo. To elucidate the underlying cause, the authors used VLPs to deliver tBE into 293FT double-knockout (DKO) cells lacking the two major uracil-specific DNA glycosylases: human uracil DNA glycosylase (hUNG) and single-strand selective monofunctional uracil DNA glycosylase 1 (hSMUG1). The resulting editing profiles revealed that insufficient inhibition of uracil DNA glycosylases represents the primary barrier to achieving robust C-to-T base editing.

To resolve this limitation, the authors re-engineered tBE and developed a murine VLP platform optimized for tBE delivery—both specifically tailored to enhance UGI recruitment. Next, they packaged the engineered tBE into VLPs and evaluated the in vivo editing efficiency, specificity, and therapeutic potential of the tBE-VLP system in mice. This evaluation encompassed generating protective variants in wild-type mice, performing therapeutic editing in a metabolic disorder model, and treating an ocular disease model.

Potent In Vivo Efficacy and Unmatched Specificity

A single injection of the optimized system (tBE-VLP4) achieved high-efficiency C-to-T editing in vivo, including an average efficiency of 46.0% at the mouse proprotein convertase subtilisin/kexin type 9 (mPcsk9) locus, 42.5–64.2% at the mouse 4-hydroxyphenylpyruvate dioxygenase (mHpd) locus in the liver, and 24.2% at the mouse vascular endothelial growth factor A (mVegfa) locus in the RPE.

Remarkably, tBE-VLP4-mediated editing conferred clear therapeutic benefits across all treated mouse models without any detectable off-target editing in vivo. These findings underscore that VLP-mediated CBE delivery offers significantly higher precision and safety compared to conventional AAV- or LNP-mRNA-based delivery systems.