Abstract

Prime editing enables precise installation of genomic substitutions, insertions and deletions in living systems. Efficient in vitro and in vivo delivery of prime editing components, however, remains a challenge. Here we report prime editor engineered virus-like particles (PE-eVLPs) that deliver prime editor proteins, prime editing guide RNAs and nicking single guide RNAs as transient ribonucleoprotein complexes. We systematically engineered v3 and v3b PE-eVLPs with 65- to 170-fold higher editing efficiency in human cells compared to a PE-eVLP construct based on our previously reported base editor eVLP architecture. In two mouse models of genetic blindness, single injections of v3 PE-eVLPs resulted in therapeutically relevant levels of prime editing in the retina, protein expression restoration and partial visual function rescue. Optimized PE-eVLPs support transient in vivo delivery of prime editor ribonucleoproteins, enhancing the potential safety of prime editing by reducing off-target editing and obviating the possibility of oncogenic transgene integration.

Main

Among current genome editing systems that function in both dividing and nondividing mammalian cells in vitro and in vivo, prime editing¹ offers unusual versatility by enabling the replacement of a target DNA sequence with virtually any other specified sequence containing up to several hundred inserted, deleted or substituted base pairs^{2,3,4,5,6,7,8,9,10,11}. This versatility makes PE systems particularly promising for the treatment of a broad range of genetic diseases in humans. A prime editor (PE) is an engineered protein consisting of a catalytically impaired programmable nickase domain (such as a Cas9 nickase) fused to an engineered reverse transcriptase (RT) domain. The prime editing guide RNA (pegRNA) specifies the target protospacer sequence and simultaneously encodes the desired edits in the reverse transcription template in the 3′ extension of the pegRNA. The mechanism of prime editing requires three independent nucleic acid hybridization events before editing can take place and does not rely on double-strand DNA breaks or donor DNA templates. As a result of this mechanism, prime editing is inherently resistant to off-target editing or bystander editing, and can proceed with few indel byproducts or other undesired consequences of double-strand DNA breaks^{1,12,13,14,15,16,17,18,19,20,21}.

Fully realizing the potential of prime editing for research or therapeutic applications in mammals requires safe and efficient methods capable of delivering PEs into tissues in vivo. So far, several groups have reported the in vivo delivery of PE via viral delivery methods, including adenoviruses⁸ and adeno-associated viruses (AAV)^{8,9,10,11,12,22,23,24,25}. Viral delivery methods, however, require that the transgene be encoded directly in the viral gene expression cassette, limiting transgene size. The AAV genome has a cargo gene size limitation of ~4.7 kb (not including inverted terminal repeats)²⁶, requiring large cargoes such as PEs (6.4 kb in gene size for a first-generation PE) to be split into multiple AAVs²⁵, limiting editing efficiency especially at moderate or low vector doses²⁷. Viral delivery methods also pose potential safety risks including increased off-target editing from sustained transgene expression²⁸ and the possibility of unwanted cargo DNA integration into host cell genomes²⁹. Nonviral delivery methods, such as lipid nanoparticles, avoid some of these issues by packaging editors as transiently expressing messenger RNAs (mRNAs). In vivo nonviral targeting of tissues beyond the liver for efficient therapeutic gene editing remains a challenge^30,31, however, despite recent advances targeting hematopoietic stem cells³².

Virus-like particles (VLPs) are potentially promising delivery vehicles that in principle offer key benefits of both viral and nonviral delivery methods³³. VLPs are formed by spontaneous assembly and budding of retroviral polyproteins that encapsulate cargo molecules from producer cells. VLPs lack a packaged genome but retain the ability to transduce mammalian cells and release cargo^34,35. Previous studies explored VLPs for delivering Cas9 nuclease^{36,37,38,39,40,41,42,43}. We recently reported efficient in vivo delivery of adenine base editor (ABE):single guide RNA (sgRNA) ribonucleoproteins (RNPs) with iteratively engineered virus-like particles (eVLPs)⁴⁴ that overcame specific molecular bottlenecks in cargo packaging, release and localization.

Engineered VLPs offer several advantages over other delivery methods as a candidate for in vivo PE delivery. First, eVLPs are not subject to stringent cargo size limitations, obviating the requirement of splitting PEs into multiple separate vectors. In addition, eVLPs can package RNPs, the most transient form of gene editing agents, thereby reducing frequency of off-target editing by minimizing the exposure duration of the genome to editing agents^44,45,46. Since eVLPs lack DNA^34,44, they avoid unwanted integration of viral genetic material into the genomes of transduced cells. Finally, eVLPs can be pseudotyped with different glycoproteins, enabling specific targeting of cell types of interest⁴² with envelope protein engineering efforts.

In this Article, we report the development of a PE-eVLP system that delivers complete PE systems including pegRNAs and nicking sgRNAs (ngRNAs) as RNPs. Simple replacement of base editors (BEs) with PEs in the optimized BE-eVLP system yielded very low functional delivery of prime editing systems (<1% editing efficiency in cultured mammalian cells). Through systematic identification of PE-eVLP delivery bottlenecks and engineering corresponding solutions, we developed third-generation v3 PE-eVLPs that offer a 79-fold improvement in prime editing efficiency compared to v1 PE-eVLPs in mouse Neuro-2A (N2A) cells and a 170-fold improvement in human HEK293T cells. A single subretinal injection of v3 PE-eVLPs demonstrated efficient in vivo prime editing in mouse models, correcting a 4-bp deletion in Mfrp in the rd6 mouse model of retinal degeneration (15% average efficiency) and correcting an Rpe65 substitution to partially rescue visual function in the rd12 model (7.2% average efficiency). Our study establishes PE-eVLPs as a virus-free method for the in vivo delivery of prime editing systems in RNP form.

Discussion

Through extensive engineering of each major component, we developed an all-in-one virus-like particle that delivers PE RNPs into mammalian cells in culture and in vivo. Recent improvements to prime editing systems, including epegRNAs⁴, the PEmax architecture² and MMR evasion², contributed to improved outcomes with PE-eVLPs. Identification of bottlenecks in cargo packaging yielded PE variants that promote delivery by PE-eVLPs, as well as optimized eVLP architectures that facilitate cargo release and cargo localization. Introducing an additional mechanism for guide RNA recruitment addressed guide RNA packaging limitations, and an alternative v3b PE-eVLP system eliminated the need for covalent fusion to the Gag polyprotein. Together, these improvements yielded 170-fold higher average prime editing efficiency compared to v1.1 PE-eVLPs at a benchmark HEK3 test edit in HEK293T cells.

The optimized v3 and v3b PE-eVLPs systems proved efficacious in vivo. Potent prime editing was achieved in the mouse CNS via neonatal ICV injection, marking the first demonstration of CNS editing with transient delivery of a PE RNP. In the mouse retina, a single injection of v3 PE3-eVLPs precisely corrected a pathogenic 4-bp deletion in the rd6 model of retinal degeneration, restoring production of full-length MFRP protein. In the rd12 mouse model of genetic blindness, v3 PE3-eVLPs achieved comparable prime editing levels to a recently reported triple-vector AAV–PE system²³, but using a nonviral, single-particle delivery vehicle, resulting in partial rescue of visual function. These findings demonstrate that v3 and v3b PE-eVLPs can achieve prime editing efficiency comparable to that attained using an AAV–PE delivery system, while avoiding drawbacks of viral delivery systems such as prolonged editor expression that increases off-target editing frequencies and the risk of oncogenic DNA integration^29,79. To our knowledge, these findings also represent the first use of PE RNPs to achieve phenotypic rescue of an animal model of genetic disease.

While v3 and v3b PE-eVLPs demonstrated therapeutically relevant editing levels, PE-eVLP systems would benefit from the continued engineering effort for the next-generation PEs and improved eVLP systems. Furthermore, tissue-specific envelope protein engineering could expand the scope of PE-eVLP applications to diverse tissues. The possibility that single-dose, transient delivery of PE RNPs by PE-eVLPs may mitigate clinically relevant immunogenicity⁸⁰ warrants further investigation. Lastly, future optimization in large-scale eVLP production will be necessary to fully realize the therapeutic potential of eVLPs. Nonetheless, the PE-eVLP system reported here offers unique advantages of nonviral, single-particle delivery of PEs in their most transient form as RNPs, presenting safety and target specificity advantages over DNA or mRNA delivery methods.