CRISPR delivery explained: Why editing DNA is easier than getting the editor to the right cells

A human being is not a petri dish. It has immune defences, barriers between tissues, and a talent for sending injected particles to the liver whether you asked or not. That is why the hardest question in therapeutic gene editing is often not “can we make the edit?” but “can we deliver the editor safely, predictably, and at scale?” Reviews of in vivo CRISPR delivery make this point repeatedly: delivery is the gatekeeper for what is feasible.

How delivery works

In broad terms, delivery has to solve four jobs.

First, reach the right neighbourhood. After an intravenous injection, most carriers face filtration, immune clearance, and the physics of blood flow. Many lipid nanoparticles (LNPs) naturally accumulate in liver and spleen, which is helpful for liver targets but frustrating for almost everything else.

Second, enter the right cells. Viral vectors do this by exploiting evolved entry mechanisms. Non-viral systems must rely on particle chemistry, receptor binding, and cellular uptake that varies by tissue.

Third, escape the cell’s packaging department. LNP cargo usually enters via endosomes, and endosomal escape is a central bottleneck described in major reviews.

Fourth, deliver the right form of CRISPR. The payload could be DNA encoding Cas, mRNA encoding Cas, or Cas protein plus guide RNA (an RNP). The choice affects duration. DNA and some viral delivery can prolong expression; RNP delivery is transient, which can be safer but may require repeat dosing.

The delivery menu, and the trade-offs that matter

AAV: efficient, but boxed in

Adeno-associated virus (AAV) is popular because it is good at getting genetic instructions into cells. The problem is capacity. Empirical studies have shown that, in practice, packaged AAV genomes do not exceed about 5.2 kb, which constrains what you can fit into a single vector. That becomes painful when editors get larger or more complex.

Then there is immunity. Neutralising antibodies can block treatment or prevent redosing, which is why entire reviews are devoted to overcoming AAV neutralisation.

AAV can also mean long-lived expression. For gene replacement that is often a feature. For genome editing, prolonged Cas exposure can be an unwanted extension of risk.

Lentivirus: powerful ex vivo, complicated in vivo

Lentiviral vectors are a mainstay of cell engineering, especially outside-the-body workflows. They can be very effective, but integrating vectors raise insertional mutagenesis concerns, a risk regulators have discussed explicitly for lentivirus and related platforms.

Research is exploring ways to keep viral-like efficiency while avoiding integration, including lentivirus-derived nanoparticles for delivering editor RNPs.

LNPs: fast-moving, liver-friendly, hard to aim

LNPs are the leading non-viral platform for nucleic acid delivery. Their success in RNA medicines has pushed rapid innovation in formulation and manufacturing. Reviews lay out how formulation choices shape biodistribution and cellular uptake.

The recurring limitation is targeting. Many LNPs go to liver by default, which is a feature if the liver is your target and a bug if it is not. Tissue-selective formulations are an active research area, including demonstrations of lung and liver editing in animal models, but this remains a frontier rather than a solved capability.

Electroporation and ex vivo editing: the pragmatic path

If you can remove cells, edit them in a controlled facility, then return them, delivery becomes far more manageable. Ex vivo CRISPR reviews stress that you can edit the intended cell type, test the product, and reduce the risk of off-tissue editing.

The FDA’s Casgevy page describes an approved CRISPR-based therapy that sits in this ex vivo logic, illustrating why early clinical wins cluster in cell types you can collect and reinfuse.

Where it goes wrong

Delivery failures usually fall into a few bins.

Wrong place: the editor reaches the wrong organ or the wrong cell type. That is an off-tissue effect, and it can coexist with perfect sequence targeting. Ex vivo editing literature makes the distinction clearly: in vivo approaches must worry about unintended delivery to off-target cell types as well as unintended cutting at off-target genomic sites.

Wrong time: the editor hangs around longer than intended. Prolonged expression widens the window for unintended editing and immune recognition.

Immune pushback: neutralising antibodies and other immune responses can cap dosing, block redosing, or complicate safety monitoring.

Manufacturing reality: even brilliant delivery is useless if it cannot be made consistently. AAV manufacturing is widely described as challenging, with issues spanning upstream yields, downstream purification, and product quality attributes. LNPs bring their own scale-up and characterisation hurdles. Regulators expect robust CMC information for gene therapy investigational submissions.

Favourable vs difficult targets

Some tissues are simply friendlier.

More favourable: blood and immune cells (ex vivo), local sites such as the eye, and liver for LNP-heavy strategies.
More difficult: brain-wide diseases, many lung targets, and solid tissues that are hard to reach evenly and hard to monitor for biodistribution and unintended effects.

What breakthroughs would change the game

The most valuable breakthroughs are not “stronger scissors”. They are better couriers.

• Targeted non-viral delivery that can reach specific cell types beyond liver, without viral immunity constraints.
• Transient, protein-based delivery that limits exposure time while retaining efficiency, a promise seen in engineered virus-like particle approaches for editor RNP delivery.
• Better measurement of biodistribution and persistence, so “it worked in mice” becomes “we know where it went and how long it acted”.
• Manufacturing advances that make vectors and nanoparticles more consistent, because variability is a hidden driver of both cost and safety risk.

Credibility checklist for delivery claims

• Is the claim about cell entry, organ targeting, or clinical outcomes? These are not interchangeable.
• Are biodistribution and persistence measured directly, or inferred?
• Is immunity addressed, including neutralising antibodies for viral vectors?
• Are payload constraints acknowledged, especially for AAV?
• Is the manufacturing plan credible, with quality attributes and comparability strategy?

Fact-check list (claims, sources, confidence)

• CRISPR uses guide RNAs to direct Cas nucleases such as Cas9 for site-specific DNA cutting. High
• Reviews of in vivo CRISPR consistently identify delivery as a central remaining challenge. High
• AAV practical packaging is constrained, with data showing packaged genomes do not exceed ~5.2 kb in tested systems. High
• AAV packaging capacity is commonly described as ~4.7 kb and treated as a major development constraint. High
• Neutralising antibodies can limit AAV eligibility and redosing, and multiple reviews cover strategies to overcome AAV neutralisation. High
• Integrating vectors such as lentivirus carry insertional mutagenesis risk, explicitly discussed in EMA regulatory reflection. High
• LNP delivery performance depends on formulation and faces physiological barriers, with endosomal escape highlighted as a key bottleneck. High
• Many LNP formulations accumulate in liver and spleen, making liver an easier target than many other tissues. Medium-High (degree depends on formulation and route)
• Tissue-selective LNP formulations have demonstrated lung and liver editing in mouse models in peer-reviewed work. Medium-High (preclinical, model-dependent)
• Ex vivo CRISPR editing allows editing of the intended cell type and offers an opportunity to screen and characterise edited cells before infusion. High
• Casgevy is an FDA-listed approved CRISPR-based therapy for sickle cell disease and transfusion-dependent beta-thalassaemia in patients aged 12 years and older (as indicated on FDA page). High
• Engineered virus-like particles have been reported to deliver prime editing components as transient RNP complexes in vivo. High
• AAV manufacturing scale-up and product quality challenges are widely discussed in reviews and commentary. High
• FDA guidance outlines CMC expectations for human gene therapy IND submissions. High