Researchers at Seoul National University College of Medicine have developed a prime editing-based system capable of replacing genomic sequences with DNA payloads up to 6.5 kilobases in size and excising genomic regions spanning up to one megabase without introducing double-strand breaks (DSBs), according to a study published August 28 in Nature Biotechnology. The work, led by corresponding author Sangsu Bae, addresses a persistent limitation of precision genome editing: efficiently replacing large genomic segments without relying on DSB-mediated repair.
Standard prime editing, which uses a Cas9 nickase fused to a reverse transcriptase to write short sequences into the genome, is well suited to point mutations and small insertions but becomes less efficient as payload size increases. Homology-directed repair (HDR)-based approaches can accommodate larger inserts but require DSBs, carry risks including unwanted indels and chromosomal rearrangements, and are generally less effective in non-dividing cells. The new system, which the authors call prime assembly (PA), extends the prime editing architecture to kilobase-scale gene replacement.
Mechanism
PA generates complementary single-stranded 3′ flaps on both the genomic target and donor DNA using prime editing machinery directed by engineered prime editing guide RNAs (pegRNAs). These genomic and donor flaps anneal to one another in a Gibson assembly-like process, followed by strand invasion and DNA synthesis.
Mechanistic experiments identified RAD51-mediated strand invasion as an important dependency: pharmacological RAD51 inhibition reduced PA efficiency in a dose-dependent manner. By contrast, inhibition of DNA-PKcs-mediated canonical non-homologous end joining modestly reduced the efficiency of some PA configurations, suggesting that PA does not simply operate through conventional DSB-dependent integration pathways.
Three configurations were developed: single-flap PA (SF-PA), which uses a homology arm; single-flap PA with an additional nicking guide RNA (SFn-PA), designed to improve strand invasion; and dual-flap PA (DF-PA), which uses two pairs of complementary flaps to bracket a genomic region, enabling simultaneous excision and donor insertion without conventional homology arms.
Key findings
The researchers demonstrated PA activity across six cell lines — HEK293T, HEK293, HeLa, K562, HuH7, and Hepa-1c1c7 — and at multiple endogenous loci. PA accepted circular plasmid and linear double-stranded DNA donors ranging from 1.0 to 6.5 kb.
The system achieved replacement efficiency of up to 57.8% at endogenous loci. Long-read sequencing showed precise donor incorporation without major alterations in 93.7%, 84.3%, and 77.5% of donor-integrated reads for SF-PA, SFn-PA, and DF-PA, respectively.
The authors also conducted a genome-wide off-target integration analysis using a modified GUIDE-seq approach. No notable off-target integration event accounting for more than 0.06% of total reads was detected for any of the three PA formats.
DF-PA was additionally able to excise genomic regions spanning 100 kb and approximately 1 Mb while inserting a 2.9-kb donor sequence.
In disease-relevant models, PA replaced exon 1 of HTT containing expanded CAG repeats with a synthesized sequence containing ten repeats in HEK293T cells. The researchers also modeled correction strategies for pathogenic variants in GSDME and SLC39A14 and integrated an F9 cassette at the ALB locus in HuH7 hepatocyte cells as a model for hemophilia B gene therapy.
In primary human T cells, PA-mediated insertion of a CD19 CAR cassette generated CAR-positive populations of 8.3% with SF-PA, 19.1% with SFn-PA, and 28.1% with DF-PA. Edited cells expanded approximately 20-fold over seven days and demonstrated antigen-dependent killing of CD19-positive Raji cells in vitro. PA-engineered CAR-T cells also suppressed tumor growth in a Raji xenograft mouse model.
In vivo delivery using hydrodynamic tail-vein injection in C57BL/6J mice produced an average integration efficiency of 4.3% among GFP-positive hepatocytes.