Prime editor with engineered mouse-derived reverse transcriptase achieves 1.88 times the efficiency at hard-to-edit targets
Background: Reverse transcriptase shapes prime-editing efficiency
A prime editor using an engineered mouse-derived reverse transcriptase achieved higher editing efficiency than existing tools in the human cells and rice protoplasts tested. The researchers’ enRERV-RT-based system achieved an average relative efficiency of 1.20 at the endogenous targets in the general comparison and 1.88 at four hard-to-edit endogenous targets. These results come from cell and plant experiments; the study did not evaluate therapeutic effects in patients.
Prime editing introduces desired base substitutions or short insertions and deletions without creating a DNA double-strand break. A Cas9 variant that nicks one DNA strand works together with a reverse transcriptase (RT) and a prime editing guide RNA (pegRNA) carrying the editing instructions. RT is the key component that copies the editing information from RNA into DNA. Because efficient enzyme options are limited, the researchers searched for alternatives to the widely used Moloney murine leukemia virus-derived RT (M-MLV-RT).
Key findings: From candidate discovery to validation at hard-to-edit targets
The researchers performed phylogenetic analysis of 558 unique RT sequences selected from databases and tested 54 representative candidates in rice protoplasts. Nineteen enzymes showed activity in a fluorescent reporter assay. In subsequent evaluations at endogenous targets, mouse endogenous retrovirus-derived RERV-RT showed the highest activity.
To engineer the enzyme, the team compared a RERV-RT structure predicted by AlphaFold 3 with a cryo-electron microscopy structure of M-MLV-RT. They first created RERV-m5 by introducing five mutations. They then added W149F, identified by deep mutational scanning, and the additional beneficial mutations Y68R and C413R to produce the final enzyme, enRERV-RT.
Editing efficiency averaged 1.17 times that of M-MLV-based editors across seven targets in human embryonic kidney-derived HEK293T cells and 1.22 times that of the comparator across six targets in rice protoplasts. The mean relative efficiency across all 13 targets in the two experimental systems was 1.20. At four hard-to-edit targets in HeLa cells and primary human dermal fibroblasts, efficiency averaged 1.88 times that of the comparator. These experiments included insertion of a 24-base-pair FLAG tag and an 18-base-pair 6×His tag into fibroblasts.
The researchers also built TRAP-seq-PE, an evaluation platform in which 475 pegRNA–target constructs were integrated into the HEK293T cell genome. It measures the proportion of intended edits by deep sequencing target DNA, allowing comparison of efficiencies across multiple substitutions, insertions, and deletions. In this library, the efficiency of enRERV-based editors averaged approximately 1.8 times that of existing tools overall and approximately 2.74 times that of existing tools in the hard-to-edit target group. These figures should be distinguished from the average relative efficiency of 1.88 at four endogenous targets.
Implications and outlook: More tool options, with validation needed for each application
Multiplex editing was demonstrated in rice experiments using the earlier variant RERV-m5. Of 60 regenerated rice plants, 28 had simultaneous edits in three genes, a rate of 46.7%. Separate experiments in regenerated rice using the final enRERV confirmed editing activity at two endogenous targets.
The researchers also conducted experiments introducing stop codons into four genes associated with chimeric antigen receptor T cells (CAR-T). However, they did not directly test improvements in CAR-T cell anticancer activity or persistence. The safety assessment also has a limited scope. The finding that no mutations were detected at 12 predicted off-target sites applies only to rice edited with RERV-m5. Across 17 editing events in plant and animal cells, no major difference in editing byproducts was observed between enRERV- and M-MLV-based systems.
The authors proposed differences in delivery and expression methods and chromatin environments as possible reasons for the differing efficiency gains between the library and endogenous targets. This is a possible explanation, not a directly demonstrated mechanism. Applications in therapy and breeding require further evaluation of efficiency, byproducts, and safety in the intended cells and at the intended targets. The study was published in Advanced Science; its experimental design and results are available in the full paper.
CRISPR-based prime editors (PEs) install precise edits into genomic DNA without generating double-strand breaks. Their editing efficiency is highly dependent on reverse transcriptases (RTs), but efficient RT candidates remain limited. Here, we identified 19 novel active RTs by screening 558 candidates. Among them, RERV-RT, derived from Rattus norvegicus, exhibited the highest activity. Through structure-guided engineering and deep mutational scanning, we developed an optimized variant, enRERV-RT, which outperforms conventional M-MLV-RT-based PE systems by 1.20-fold in mammalian and plant cells, and by 1.88-fold at hard-to-edit loci, while enabling precise multiplex editing of functionally relevant genes. Additionally, we developed a high-throughput platform, TRAP-seq-PE, to systematically evaluate prime editor performance. Across diverse mutation types, we found that PE systems based on enRERV-RT exhibited higher editing efficiencies than those based on M-MLV-RT. Collectively, our work establishes a versatile, high-efficiency PE system, thereby facilitating advances in clinical gene therapy and precise crop breeding.
Research teams encountering low prime-editing efficiency at particular targets now have another reverse transcriptase to compare. If insertion of a desired sequence is inefficient, they can design experiments comparing M-MLV-based and enRERV-based tools under the same cell, target, and delivery conditions, while measuring intended edits and byproducts together. TRAP-seq-PE offers a way to compare performance across editing types, but validation at actual endogenous targets must follow. In breeding research, editing results in regenerated rice can serve as a starting point for checking the desired trait and its inheritance in subsequent generations. In therapeutic research, the work provides a candidate tool, but it does not establish efficacy in patients, clinical safety, or a shorter time to regulatory approval.