BGPT: Paper Review: Search-and-replace genome editing without double-strand breaks or donor DNA

Explore by Goal

Quick Explanation Copied

Prime editing (Anzalone et al., 2019) replaces DSB+donor “cut-and-paste” with a pegRNA-programmed “search-and-replace” write operation, enabling targeted insertions, deletions, and all 12 base-to-base substitutions in human cells—while reporting low indels (depending on PE variant) and reduced off-target activity at known Cas9 off-target sites.

Primary paper: 10.1038/s41586-019-1711-4.

Key claims and quantitative outcomes cited inline below.

Long Explanation

Paper Review (BGPT): Prime editing without DSBs or donor DNA

Anzalone et al., Nature (2019)

Core thesis: A Cas9 nickase fused to a reverse transcriptase (RT), guided by a prime editing guide RNA (pegRNA) that both targets and encodes the desired edit, can install edits directly at genomic loci without requiring double-strand breaks (DSBs) or donor DNA templates.

0) Visual map: what edits the system claims to support

Note: the figure is a binary “enabled by claim” schematic (not efficiency). The paper states prime editing performs targeted insertions, deletions, and all 12 types of point mutations, including combinations, in human cells.

1) Mechanistic logic (Deduced pathways vs demonstrated steps)

1A. Claimed pathway (paper’s model)

PAM-strand nicking: a Cas9 nickase (H840A) guided by pegRNA nicks the PAM-containing DNA strand.
pegRNA-templated synthesis: the nicked 3′ end hybridizes to the pegRNA PBS, and RT uses the pegRNA RT template to copy the intended edit into the genomic locus.
Flap resolution: the edited and unedited flaps form a branched intermediate; equilibration plus flap excision/ligation yields stably edited DNA.

Skeptical note: mechanistic claims (e.g., which flap is excised first, and how cellular repair pathway choice biases outcomes) are partially supported by biochemical reporter and sequencing outcomes described in vitro/yeast/human cells, but the paper also makes inferences about pathway preferences rather than directly imaging/interrupting every intermediate in vivo.

2) Quantitative performance: point mutations, indels, and precision trade-offs

2A. PE2 → PE3 → PE3b: efficiency vs indels

The paper reports that complementary-strand nicking (PE3) increases prime editing efficiencies, and PE3b further reduces indels by using an edit-specific nicking strategy that disfavors nicking until after the PAM strand is edited.

The plot uses only explicitly provided summary values: PE3 “up to as high as 55%” and the corresponding “6.8% average indels for the sgRNA with the highest editing efficiency,” plus the PE3b “0.74%” average indel.

2B. All 12 base substitutions at HEK3: mean efficiency and indels

The paper reports that 24 point mutations (covering all 12 substitution types) across HEK3 positions +1 to +8 (with a 10-nt RT template) achieved “average editing efficiencies (containing no indels) of 33 ± 7.9%” and “7.5 ± 1.8%” average indels.

2C. Disease-relevant edits: installation and correction (efficiency and indels)

Values are taken directly from the paper’s reported outcomes: HBB E6V installation 44% with 4.8% indels; HBB correction to WT with PAM-silent improved purity 58% correction with 1.4% indels.
HEXA 1278+TATC installation 31% with 0.8% indels; HEXA correction to WT best pegRNA via PE3b 33% correction with 0.32% indels.
PRNP protective G127V installation 53% with 1.7% indels; mouse primary cortical neurons DNMT1 edited (sorted nuclei) 7.1% prime editing with 0.58% indels.

Critical interpretation

Context dependence is real: the primary-neuron editing efficiency (7.1%) is substantially lower than optimized HEK293T contexts (often reported tens of percent), indicating strong locus/cell-type/delivery dependence.
“Indels” still matter: even when prime editing is “D(sb)-free,” the repair processes that resolve flaps still produce a measurable fraction of indels; PE3 vs PE3b illustrates a tunable precision-efficiency trade-off.

3) Specificity / off-targets: what’s tested and what isn’t

3A. Known Cas9 off-target loci (targeted-site set)

The paper tests prime editing off-target activity at known Cas9 off-target sites: 16 pegRNAs across 4 genomic loci, comparing Cas9 nuclease+sgRNA vs Cas9+pegRNA vs PE3/PE2 with the same pegRNA spacers.

The paper reports: PE2/PE3 detect off-target editing at only 3 out of 16 off-target sites, with only 1 of 16 showing off-target editing efficiency ≥1%.
Skeptical limitation: these are site-selected assays at previously characterized Cas9 off-targets, not genome-wide unbiased profiling; therefore, the “low off-target” conclusion is conditional on the tested site set and the sensitivity threshold. The paper itself flags the need for further genome-wide off-target assessment.

4) Editing “byproducts” beyond indels: pegRNA scaffold insertions

4A. Scaffold sequence insertion frequency

The paper analyzes 66 PE3 editing experiments and reports an average total insertion of any number of pegRNA scaffold nucleotides of 1.7 ± 1.5%.

Interpretation: pegRNA scaffold insertions are a distinct byproduct class from classic indels; “D SB-free” does not eliminate all unintended sequence incorporations. The reported average (1.7%) suggests these events are uncommon but measurable.

5) What the paper does especially well

Mechanism-to-outcome continuity: the study moves from in vitro biochemical assays (RT products on defined substrates) to yeast reporter systems, then into mammalian cell editing with targeted endogenous loci and sequencing-based quantification.
Scope claims supported by systematic editing sets: the paper reports 24 substitutions (covering all 12 substitution classes) across HEK3 and multiple insertion/deletion categories, plus combination edits.
Disease-relevant demonstrations: prime editing installs/corrects mutations in HBB (sickle cell), HEXA (Tay–Sachs), and PRNP (protective G127V), and is applied to insert epitope/tags.

6) Limitations, blind spots, and “what would disprove this”

6A. Editing efficiency depends strongly on cell type and delivery

The paper demonstrates prime editing in multiple human cell lines, but reports lower efficiencies outside HEK293T, and the neuron results show modest editing percentage in sorted nuclei.

6B. Off-target assessments are not genome-wide

The main off-target evidence is restricted to known Cas9 off-target sites and does not substitute for unbiased, genome-wide profiling.

6C. Byproduct biology: indels and scaffold insertions persist

Even with DSB-free operation, prime editing produces measurable indels (tuned by PE variant and nicking strategy) and measurable pegRNA scaffold insertions on average (1.7 ± 1.5% in the reported analysis).

Disproof-oriented checklist (derived from paper’s own claims)

Mechanistic falsification: demonstrate that pegRNA-programmed RT-templated synthesis does not generate stably edited outcomes in cells, or that claimed DSB independence fails under sensitive assays.
Scope falsification: show inability to install transversions/insertions/deletions broadly across many loci in multiple cell types.
Specificity falsification: show genome-wide off-target prime editing comparable to, or exceeding, Cas9 nuclease at many sites, and show that byproduct classes (indels, scaffold insertions) are not controllable via nicking strategies.

The paper explicitly calls out the need for genome-wide off-target assessment and additional mechanistic improvement studies.

7) Author reviews you can open next

Feedback:

Updated: April 15, 2026