Precision single-nucleotide genome editing without a double-strand DNA break — direct chemical base conversion and reverse-transcriptase sequence rewriting
Classic CRISPR-Cas9 genome editing works by guiding the Cas9 nuclease to a target sequence, where it cuts both strands of the DNA double helix. This double-strand break (DSB) is a cellular emergency: the cell must repair it quickly, and the repair pathways it uses — non-homologous end joining (NHEJ) or homology-directed repair (HDR) — are inherently variable, often introducing small insertions, deletions, or other unintended changes exactly where a clean, single-base correction was intended.
Cas9 is directed to its target by a single guide RNA (sgRNA) that base-pairs with a ~20-nucleotide protospacer sequence adjacent to a PAM motif. Once bound, Cas9's two nuclease domains (RuvC and HNH) each cleave one strand of the DNA, producing a blunt double-strand break a few bases upstream of the PAM.
This break is a powerful editing lever — it forces the cell to engage its DNA repair machinery — but it is also the source of most of the unpredictability in traditional CRISPR editing outcomes.
Non-homologous end joining (NHEJ), the dominant repair pathway in most cell types, rejoins the two broken ends directly, often trimming or adding a few nucleotides in the process. This is useful for gene disruption (knockouts) but is fundamentally imprecise: the exact indel produced at any given break is largely unpredictable in advance.
Homology-directed repair (HDR) can achieve precise edits by copying a supplied donor template, but it is only active in dividing cells during S/G2 phase and is typically far less efficient than NHEJ — meaning most repair events at a DSB remain error-prone even when a template is provided.
Because every double-strand break is repaired through pathways that can introduce unintended edits, traditional Cas9 editing carries an inherent DSB risk profile — a risk that base and prime editing were specifically engineered to avoid by never cutting both strands in the first place.
Base editors combine a catalytically impaired ("dead" or nickase) Cas9 with a chemical-conversion enzyme — a deaminase — fused directly to it. Guided to the target locus by an sgRNA exactly as in traditional CRISPR, the complex does not cut both strands. Instead, the deaminase chemically converts one base to another within a small editing window, while the DNA backbone stays intact throughout.
A base editor fuses three components onto a single scaffold: (1) a Cas9 nickase (nCas9) that binds the target sequence via sgRNA pairing but only nicks — rather than fully severing — one strand; (2) a deaminase enzyme (cytidine deaminase for C-to-T editors, or an engineered adenine deaminase for A-to-G editors) that acts on the exposed single-stranded DNA within the R-loop; and (3) often a uracil glycosylase inhibitor (UGI) that suppresses the cell's base-excision repair machinery long enough for the edit to be fixed into the genome.
Because only one strand is nicked — and never both strands cut simultaneously — the double helix remains structurally continuous throughout the entire editing process.
The single-strand nick on the non-edited strand biases the cell's repair machinery to use the edited strand as the template when resolving the mismatch, effectively "locking in" the new base pair. Because there is no double-strand break, the cell never engages the error-prone NHEJ pathway that produces indels in traditional editing.
The trade-off is scope: base editors can only perform the specific chemical conversions their deaminase enzyme is capable of (C→T or A→G, and their complementary strand equivalents G→A or T→C). They cannot insert or delete bases, and they cannot perform transversions (e.g., C→A) or arbitrary substitutions.
Base editing trades broad flexibility for exceptional precision: within its narrow chemical scope of base-to-base conversions, it achieves editing outcomes with far fewer unintended indels than traditional double-strand-break editing, because the DNA backbone is never fully severed.
Prime editing goes a step further than base editing. It fuses a Cas9 nickase to an engineered reverse transcriptase and pairs it with a prime editing guide RNA (pegRNA) — an extended guide RNA that both specifies the target site and carries a template encoding the desired new sequence. The reverse transcriptase reads this template and directly writes the edited sequence into the genome, all without ever creating a double-strand break.
The pegRNA extends a standard guide RNA with two additional elements: a primer binding site (PBS) that anneals to the nicked genomic strand, and an RT template that encodes the new, edited sequence. After the Cas9 nickase nicks the non-target strand near the target site, the exposed 3' end of that nicked strand hybridizes to the PBS.
The fused reverse transcriptase then extends this primer using the RT template as its guide, synthesizing a new DNA flap that carries the desired edit — a substitution, a small insertion, or a small deletion — directly onto the existing strand.
The newly synthesized 3' flap (carrying the edit) competes with the original, unedited 5' flap for incorporation. Cellular repair enzymes preferentially resolve this in favor of the edited flap, and DNA repair/replication seals the edit into one strand. A second nick on the complementary strand (in prime editing systems such as PE3) encourages the cell to copy the edited strand as the template for correcting the other strand, further embedding the new sequence.
At no point in this process are both strands ever cut simultaneously — the mechanism relies entirely on sequential single-strand nicking and flap resolution, which is why prime editing retains the low double-strand-break risk profile of base editing while dramatically expanding what kinds of edits are possible.
Because the pegRNA directly specifies the new sequence rather than relying on a fixed deaminase chemistry, prime editing can, in principle, install any of the 12 possible base-to-base conversions as well as small insertions and deletions — all without a double-strand break.
The central practical advantage of base and prime editing over traditional Cas9 editing is outcome predictability. Because neither approach creates a double-strand break, neither engages the error-prone NHEJ repair pathway that generates unpredictable insertions and deletions at the target site. This translates directly into a much cleaner distribution of editing outcomes, which is especially valuable when the therapeutic or research goal is a single, specific point correction rather than gene disruption.
Neither base nor prime editing is entirely free of imperfection. Base editors can produce "bystander edits" — converting additional susceptible bases within the editing window beyond the intended target — and can generate a low rate of unwanted RNA or DNA off-target deamination. Prime editing can occasionally produce small indels as byproducts of imperfect flap resolution, particularly at the nick sites.
However, these residual error modes are categorically different from — and generally far less frequent than — the large, unpredictable indels produced when a cell repairs a fully severed double-strand break.
Many inherited genetic diseases are caused by a single-nucleotide substitution (e.g., sickle cell disease, many forms of inherited blindness, some metabolic disorders). For these applications, the therapeutic goal is a precise, minimal edit — exactly one base changed, nothing else disturbed nearby.
Using traditional Cas9 for such a correction means every repaired allele carries a real chance of an unintended local indel, potentially disrupting the very gene being corrected. Base or prime editing, by directly converting or rewriting the base without a double-strand break, removes that indel-risk pathway entirely — making the editing outcome far more predictable and far better matched to a single-point-mutation correction goal.
The absence of a double-strand break is not merely a mechanistic curiosity — it removes the dominant source of unpredictable editing outcomes, making base and prime editing substantially better suited than traditional Cas9 editing for correcting single, well-defined point mutations.
Base editing and prime editing are not interchangeable — they differ meaningfully in what kinds of genetic changes they can address. Base editors are restricted to the specific chemical conversions their deaminase can perform, while prime editing's reverse-transcriptase mechanism can, in principle, write any short sequence change. Choosing the right tool starts with correctly classifying the target mutation.
Cytosine base editors (CBEs) convert C•G base pairs to T•A; adenine base editors (ABEs) convert A•T base pairs to G•C. Because these are the only two chemistries currently in routine use, base editing can correct roughly half of known pathogenic single-nucleotide substitutions — those matching a C→T, G→A, A→G, or T→C change — but cannot address transversion mutations (e.g., C→A, G→T) or any insertion or deletion, however small.
When the target mutation matches the available base-editor chemistry, base editing is usually the simplest and cleanest tool: it uses a single, well-characterized enzyme with predictable behavior and minimal machinery beyond the Cas9-deaminase fusion.
Prime editing's pegRNA-encoded, reverse-transcriptase-written approach is not limited to a fixed deaminase chemistry. Because the new sequence is directly specified in the pegRNA template, prime editing can in principle install any of the 12 possible base-to-base conversions, as well as small (typically under ~40–50 bp) insertions and deletions.
This makes prime editing the appropriate choice whenever the target mutation is a transversion, a small insertion or deletion, or otherwise falls outside base-editor chemistry — at the cost of somewhat greater molecular machinery complexity (a longer pegRNA and reverse transcriptase) and a modestly higher (though still DSB-free) unintended-outcome rate than base editing.
A practical decision rule: if the correction is a simple C→T/G→A or A→G/T→C substitution, base editing is usually the simplest fit. If the correction involves a transversion, an insertion, or a deletion, prime editing is required — and if the correction needs a large sequence replacement or the DSB-driven efficiency of HDR is acceptable, traditional Cas9 editing remains an option, at the cost of double-strand-break risk.