CRISPR’s next act: Editors that rewrite single letters without cutting the DNA

A gene-editing headline often begins with a satisfying verb: cut, snip, delete. That is classic clustered regularly interspaced short palindromic repeats (CRISPR) science, where CRISPR-associated protein 9 (Cas9) is guided to a DNA address and makes a double-strand break. The cell then repairs the break, either imperfectly, which can disrupt a gene, or, more rarely, by copying a supplied template, which can correct a sequence.

Base editing and prime editing keep the CRISPR “address label” but change the verb. Instead of cutting DNA clean through, they aim to rewrite it. In principle, that can reduce certain risks associated with double-strand breaks and expand the range of edits that are practical in living cells. In practice, it introduces new complexities and a new set of ways for things to go wrong.

From scissors to pencils: how the new editors differ from classic CRISPR

Classic CRISPR editing depends heavily on how a cell repairs DNA breaks. That can be powerful, particularly for “knockout” edits, but it can also create small insertions and deletions (indels) and other repair by-products.

Base editors attach a DNA-modifying enzyme to a Cas9 that does not fully cut both DNA strands. The first widely used cytosine base editors were designed to convert C to T at a target site, without making a double-strand break. Subsequent work produced adenine base editors that convert A to G.

Prime editors combine a Cas9 nickase with a reverse transcriptase and a specialised guide, a prime editing guide ribonucleic acid (pegRNA), which both directs the editor and provides a template for the desired edit. Prime editing can, in principle, install any of the 12 possible single-letter changes, as well as small insertions and deletions, again without a full double-strand break.

This is why you keep seeing “single-letter” in headlines. Many genetic diseases are caused by point mutations, changes in one DNA base. If you can reliably correct a letter rather than break and rebuild a region, you can sometimes aim for a more surgical fix.

Why single-letter changes matter

A large share of known disease-causing variants are single-letter changes. These mutations can alter a protein by one amino acid, create an early stop signal, or disrupt splicing so that a gene is misread.

Single-letter editing is also attractive because it can be framed as restoration rather than overhaul. In some diseases, replacing a gene is complicated by size limits of viral vectors, by the need for precise expression control, or by dominant mutations where simply adding a normal copy does not solve the problem. Reviews in inherited retinal disease, for example, note that editing approaches can be relevant when gene augmentation is not sufficient for certain mutation types.

The catch is that base editing is not a universal alphabet. Traditional cytosine and adenine base editors mainly perform “transition” changes, C to T and A to G. Newer “transversion” editors are being developed to broaden what can be corrected, but they remain earlier-stage. Prime editing is more flexible on paper, but it is more complex to deliver and optimise.

The two toolkits: base editing versus prime editing

Base editing, in brief
Base editors act within a small “editing window” around the target site. That makes them efficient for the right mutation, but it can also create bystander edits, unintended letter changes nearby if multiple editable bases sit within the window. Newer editor variants aim to narrow windows and reduce bystander activity.

Prime editing, in brief
Prime editing is closer to “search and replace”. The pegRNA tells the reverse transcriptase what to write. That flexibility is the headline advantage. The disadvantage is that the machinery is larger and the optimisation problem is harder, since efficiency can vary widely with the target site, cell type, and pegRNA design.

Where this is heading in the real world: three disease arenas

Blood disorders: ex vivo editing in stem cells

Blood is where gene editing has moved fastest because the workflow is established. Clinicians can remove a patient’s haematopoietic stem cells, edit them in a controlled setting, and return them after conditioning chemotherapy. That is ex vivo editing.

The UK has already authorised a CRISPR-based therapy, Casgevy (exagamglogene autotemcel), for sickle cell disease and transfusion-dependent beta thalassaemia, demonstrating the feasibility of gene editing as a medicine, albeit using classic cutting CRISPR.

Base editing is now following that path. Beam Therapeutics has reported clinical updates in sickle cell disease for its base-edited cell therapy, BEAM-101, describing durable induction of foetal haemoglobin in treated patients.

Prime editing has also entered humans. Prime Medicine has released clinical data and subsequently published initial results for a prime-editing therapy in chronic granulomatous disease, presenting it as early evidence that prime editing can work in people.

Blood disorders suit editors because measurement is straightforward. You can track edited cell fractions, protein outputs, and clinical events over time. They also illustrate the ethical and access tension. These are complex, resource-intensive treatments.

Inherited blindness: an eye-friendly target, with harder edits ahead

The eye is often described as a favourable organ for gene therapy because it is relatively accessible and can be treated locally. In 2024, Harvard Medical School highlighted results from a small proof-of-concept study suggesting that in vivo CRISPR editing for an inherited blindness condition was safe and improved vision in some participants, using a cutting approach rather than base or prime editing.

Base and prime editing are attractive in retinal diseases because many pathogenic variants are single-letter changes and some target genes exceed viral packaging limits for replacement strategies. A 2025 review focused on base and prime editing for inherited retinal diseases discusses delivery constraints and the promise of more precise correction strategies.

The crucial phrase here is “promise”. Much of the eye-related base and prime editing work remains preclinical. Researchers have reported prime editor delivery approaches that partially restored vision in blind mice, illustrating feasibility but not yet a clinical pathway.

Liver conditions: in vivo editing meets the delivery race

If blood is the ex vivo stronghold, the liver is the in vivo proving ground. Liver cells are reachable by intravenous delivery, and lipid nanoparticles (LNPs) have become a major platform for delivering nucleic acids, including gene editing components.

Verve Therapeutics has presented human proof-of-concept for in vivo base editing of PCSK9 in the liver, aiming for durable low-density lipoprotein (LDL) cholesterol reductions after a single infusion, while also discussing delivery-related laboratory abnormalities.

Beam Therapeutics has reported initial clinical data for BEAM-302, a liver-targeting LNP formulation of base editing reagents designed to correct the PiZ mutation in alpha-1 antitrypsin deficiency, framing it as in vivo genetic correction.

These programmes underline the trade-off. In vivo editing avoids stem cell collection and transplant-style conditioning, but it raises the stakes on delivery safety, tissue targeting, and long-term monitoring.

Delivery: ex vivo versus in vivo, and how the editors get there

Delivery is the field’s recurring constraint.

Ex vivo approaches typically deliver editors to cells in culture using electroporation of messenger RNA (mRNA) and guide RNAs, or viral vectors in some settings, before reinfusing the edited cells.

In vivo approaches must package editors for systemic or local administration. Two broad classes dominate current work:

• Viral vectors, especially adeno-associated virus (AAV), which can deliver genetic instructions for editors. The limitation is payload size, which is challenging for large editors such as prime editors, often requiring split systems. Reviews in retinal editing emphasise that many active trials still rely heavily on AAV-based delivery.
• Lipid nanoparticles, which can deliver mRNA and guide RNAs to organs such as the liver. LNP delivery has reached clinical application in several genetic medicine contexts.

A pragmatic way to read the next five years is as a delivery contest. Editors improve on the bench. Medicines succeed or fail in the body.

Safety fears: off-target edits, mosaicism, and the long tail of monitoring

Precision is the selling point, so safety is the scrutiny point.

Off-target editing can occur when the guide RNA leads the editor to similar sequences elsewhere in the genome, or when the editing enzyme has its own off-target activity. Reviews describe methods for detecting and mitigating off-target effects in CRISPR therapeutics, and base editing adds additional layers such as bystander editing and deaminase-driven off-target risks.

In 2026, a Nature Biotechnology study reported genome-wide profiling of base editor off-target activity and found substantial unintended editing at some nominated sites in a haematopoietic stem and progenitor cell setting. It is the kind of paper that should make any serious developer uncomfortable, and any serious reader grateful.

Mosaicism is another concern, particularly for in vivo editing. Not every cell receives the editor, and not every edited cell receives the same edit. That can matter for efficacy and for safety, especially if rare off-target events occur in long-lived cell populations.

Long-term monitoring is not optional for permanent changes. The FDA’s guidance on long term follow-up after administration of human gene therapy products sets expectations for extended safety observation because delayed adverse events may occur after durable biological changes.

When people say “cure”, regulators tend to hear “follow-up”.

What does “curing” mean, in practice?

A useful discipline is to replace “cure” with a more specific claim.

For blood disorders, a “functional cure” might mean no severe pain crises for sickle cell disease, or transfusion independence for beta thalassaemia, over a sustained period, with durable edited cell populations.

For liver editing, it might mean a stable reduction in a disease-driving protein, such as PCSK9, with acceptable safety and durability.

For retinal disease, it might mean a measurable improvement in visual function that persists, in a defined subgroup, without harming other retinal cells.

None of these remove uncertainty. They narrow it. They also highlight why headlines often run ahead of timelines. Prime editing, in particular, is earlier in clinical development than base editing in several major indications, and delivery constraints remain central.

Ethics and access: who gets edited, and who gets left behind?

Editors raise familiar gene therapy ethics, plus new twists.

Cost and infrastructure: Casgevy’s NHS rollout has been described as a one-off treatment costing around £1.65 million, illustrating the mismatch between transformative potential and budgetary reality.

Global equity: diseases such as sickle cell disproportionately affect populations that have historically faced healthcare inequities. If editing therapies remain expensive and specialised, the benefits risk concentrating in wealthy health systems.

Consent and uncertainty: permanent edits and long follow-up create a consent landscape that is closer to organ transplantation than to routine prescribing.

Language discipline: “precision” is not a synonym for “risk-free”. Editors change the risk profile, they do not eliminate it.

Questions to ask when you read a headline about gene editing

1. Is this classic CRISPR cutting, base editing, or prime editing?
2. Is the edit ex vivo or in vivo?
3. What exactly is being changed, a gene knockout, a single-letter substitution, or a small insertion or deletion?
4. How many cells are edited, and how stable is that fraction over time?
5. What delivery method is used, AAV, LNP, or something else, and what are the known safety issues of that delivery platform?
6. What are the main off-target risks in this system, including bystander edits for base editors?
7. How were off-target edits assessed, and with which detection methods?
8. What is the duration of follow-up, and is long-term monitoring planned, often in years rather than months?
9. What does success mean here, a biomarker change, a clinical outcome, or both?
10. Who can access it if it works, and what does the delivery route imply about cost and scalability?

The timeline, plainly stated

Base editing is already in human trials across multiple areas, including ex vivo blood disorders and in vivo liver programmes. Prime editing has entered the clinic and has early human data, but it is still earlier in its therapeutic arc, with delivery and efficiency challenges likely to shape the pace of expansion.

The next act of CRISPR is not one technology replacing another. It is a growing toolbox, matched to the kind of mutation, the target tissue, and the acceptable risk. The most trustworthy stories in gene editing will be the ones that specify which tool is being used, what is being measured, and what remains unknown.