Genome engineering
Prime editing
How a nickase-reverse transcriptase fusion writes a specified sequence without a double-strand break or a donor, and why primer binding site length, flap equilibrium and mismatch repair decide whether it works.
Prime editing answers the limitation that both preceding chemistries leave open. Nuclease editing can install an arbitrary sequence but needs a double-strand break and a donor template, and therefore needs a dividing cell; base editing needs neither but can only make transitions within a narrow window. Prime editing writes an arbitrary short sequence — all base substitutions, and small insertions and deletions — with no double-strand break and no separate donor.
The guide carries the answer
The mechanism turns on a single design move: the new sequence travels with the guide. A pegRNA is an ordinary guide extended at its 3’ end by two segments. The first, the primer binding site, is complementary to the genomic strand the editor has just nicked. The second is a reverse transcription template encoding what the edit should read.
The protein is a Cas9 nickase fused to an engineered reverse transcriptase. It nicks the protospacer-adjacent strand, freeing a 3’ end; the primer binding site anneals to that end; the reverse transcriptase extends it, copying the RT template into DNA. The cell now has, at that position, a newly synthesised 3’ flap carrying the edit and an original 5’ flap carrying the old sequence. Flap resolution is an equilibrium, and the edit is only installed if the new flap wins and the strand is ligated in. Nicking the opposite, unedited strand — the PE3 arrangement — biases repair toward copying the edit into it, at the cost of introducing a second nick and the indels that come with two nicks close together.
Why design is the work
Prime editing has no separate delivery of a donor to optimise, so essentially all of its efficiency lives in the pegRNA. Primer binding site length trades priming stability against competition with the intact genomic strand; the reverse transcription template must be long enough to include the edit with homology beyond it, but a long single-stranded extension is degraded and can fold. Structured motifs added at the pegRNA 3’ end to resist exonuclease trimming were a substantial efficiency gain on their own. Every target position has its own optimum, which is why prime editing is screened empirically rather than predicted.
Mismatch repair is the hidden opponent
A newly installed edit sits in the duplex as a mismatch, and DNA mismatch repair is a highly active surveillance system whose job is to remove exactly that. It preferentially excises the nicked, newly synthesised strand — that is, the edited one — reverting the result. Transiently suppressing this pathway during editing, the PE4 and PE5 designs, raised efficiencies severalfold and also reduced indel byproducts. The effect is strongest for single-base edits, which look most like ordinary replication errors; longer insertions are less efficiently recognised and repaired.
The trade is stated honestly: prime editing avoids the break and its structural consequences, and pays for it in a longer, more delivery-constrained construct and target-dependent efficiency. Nuclease and deaminase chemistries have their own pages; delivery is not treated here.