Saturated mutagenesis screen of M-MLV reverse transcriptase identifies variants enhancing prime editing efficiency

Prime editing offers a versatile "search-and-replace" tool for genome engineering. It performs precise modifications without requiring double-strand breaks (cutting both strands of the DNA helix) or external donor templates. However, the technology faces a ceiling. Editing efficiency—the percentage of successfully modified cells—remains low and inconsistent across different genetic changes. Researchers seek ways to boost this efficiency. The fundamental bottleneck often lies in the reverse transcriptase (RT), the enzyme that writes the new genetic code.

The bottleneck of enzymatic precision

Current prime editing systems use a fusion protein. This protein combines a Cas9 nickase (a protein that cuts only one DNA strand) with a Moloney murine leukemia virus reverse transcriptase (MMLV-RT). The enzyme must bind to a genomic site and recognize a prime editing guide RNA (pegRNA). It then synthesizes new DNA using that RNA as a template. This step is precarious. The interaction between the enzyme and the growing RNA/DNA hybrid is often unstable. If the enzyme fails to maintain a tight grip, the process stalls. This leads to poor editing rates or unintended mutations.

Existing attempts to fix this involve increasing enzyme concentration or stabilizing the guide RNA. But the study notes that editing efficiency remains a significant bottleneck. There is a deeper mechanical problem. The physical interaction strength between the reverse transcriptase and the template-primer complex is insufficient. This limits processivity (the ability of an enzyme to catalyze consecutive reactions without releasing its substrate).

Engineering a tighter grip through mutagenesis

The researchers used a two-stage strategy to shrink the editor and supercharge its catalytic core. First, they developed a compact editor called PE2ΔR. They created this by deleting the RNase H domain of the MMLV-RT. This domain normally degrades the RNA strand in an RNA/DNA hybrid. Its removal reduced the enzyme's size by 26.4%. This is critical for delivering editors into cells via viral vectors. The deletion also improved efficiency by enhancing the stability of the DNA:RNA complex [Figure 1a].

Second, the authors used an unbiased, high-throughput search for better mutations. They targeted two polypeptide stretches within the "fingers domain" of the RT. These regions interact directly with the DNA template [Figure 2a]. They used saturated mutagenesis (a technique testing every possible amino acid substitution at every position). They constructed large libraries of variants. They screened these in an eGFP reporter cell line. In this line, successful editing restores green fluorescence. This allowed them to use flow cytometry (a method to sort cells by light signals) to identify the "winners" [Figure 2b, 2c].

Among 1,045 variants, they identified a potent triple mutant: I61R, V101R, and S67W. When these mutations were added to an optimized editor (PEmaxΔRM3), the results changed significantly.

Synergistic gains across the genome

The impact of these mutations is synergistic. The authors report that the triple mutant PE2ΔRM3 increased editing efficiency by 86% compared to the parental PE2 in reporter cells [Figure 3a]. When applied to the PEmax architecture, the PEmaxΔRM3 variant achieved a 91% increase in efficiency [Figure 3b].

This improvement works across many genetic edit types. The researchers tested PEmaxΔRM3 across twelve pegRNA/nicking sgRNA pairs at seven endogenous genomic loci. Base substitutions and small insertions saw improvements of 12% to 37%. However, the effect on small deletions was much larger. It reached up to an 87% increase in efficiency [Figure 3c-3j]. This suggests the engineered enzyme handles the structural challenges of removing DNA segments more effectively.

The authors used AlphaFold 3 to model the molecular architecture. The modeling shows the mutations act as a specialized "clamp" [Figure 4a]. The I61R and V101R mutations introduce positively charged arginine side chains. These form salt bridges (electrostatic attractions) with the negative phosphate backbone of the RNA template [Figure 4b]. Meanwhile, the S67W mutation places a bulky indole ring that stacks against the template. This prevents the template from "fraying" or peeling away. Together, these contacts stabilize the duplex and guide it into a better trajectory for catalysis [Figure 4d].

Limitations and the road to clinical utility

The study leaves several technical gaps. First, the proposed mechanism relies on structural modeling. AlphaFold 3 provides a plausible blueprint. However, the authors admit the exact structural mechanics are not fully understood. They suggest molecular dynamics simulations (computer models of atomic movement) to confirm how these residues move.

Second, validation occurred mainly in HEK293T cells. These are common laboratory cell lines. The performance of these editors in primary human cells remains unknown. Primary cells are cells taken directly from living tissue. The cellular environment, such as chromatin state (how tightly DNA is packed), might change how enzymes access targets. This could diminish the advantages seen in the lab.

The verdict: A new framework for editor optimization

The evidence supports the utility of this approach. By focusing on the mechanics of the reverse transcriptase, the researchers provided a blueprint for improving prime editing. The discovery proves we do not have to choose between a small footprint and high efficiency. A compact, RNase H-deficient editor can be further optimized.

For practitioners, the PEmaxΔRM3 variant is a significant upgrade. It is especially useful for applications involving small deletions. For researchers, this work shows that unbiased mutagenesis is a robust way to tune complex genome editors. The next step is testing these variants in the more demanding environments of living tissues.

Figures from the paper