Lipid nanoparticles are the dominant way gene editors reach the liver in people today, while viral vectors, mainly adeno-associated virus (AAV), remain the alternative when a longer-lasting or tissue-specific delivery is needed. Delivery, not editing chemistry, is often the real bottleneck: an editor that works perfectly in a dish does nothing if it cannot get into the right cells in the body.

This comparison covers how each vehicle works, where each one is stuck, and what recent papers and trials say about getting past the obstacles.

The delivery problem in one paragraph

A gene editor has to cross several barriers. It must survive the bloodstream, find the target tissue, enter the cell, escape the compartment it entered in, and reach the nucleus, all while avoiding the immune system. An editor made of Cas protein plus a guide RNA is large, and it only needs to function briefly. Those facts shape every design choice below.

Lipid nanoparticles: tiny fat bubbles carrying RNA

A lipid nanoparticle wraps nucleic acids in a shell of lipids. For gene editing the cargo is usually messenger RNA, which the cell translates into the editor, plus a guide RNA. The mRNA does not last, so the editor is present for a short time and then gone.

The strength of this approach is the liver. When lipid nanoparticles are injected into the bloodstream, they are taken up mostly by liver cells, and that matches many diseases with liver targets. Most of the in vivo programs reporting human data in 2026 use this route. Lilly’s base editor, for example, was given as a single intravenous infusion over about four hours (Eli Lilly, May 2026). Personalized therapy for baby KJ used the same logic, and he received three infusions over roughly two months, which shows that nanoparticles can be given more than once (Penn Medicine).

Beam’s alpha-1 antitrypsin trial also tested a two-dose cohort, and reported mild to moderate infusion reactions after single doses and, in the multi-dose group, a Grade 4 ALT elevation in one patient that the company called asymptomatic (Beam, March 2026).

Where nanoparticles struggle

  • The liver bias. Conventional particles accumulate in the liver, which is a gift for liver disease and a wall for everything else.
  • Reactions. Infusion reactions are common in the trials above, usually mild to moderate.
  • Liver stress. Enzyme elevations show up repeatedly, and Intellia’s paused nex-z trials involved a fatal liver injury. Whether the particle, the editor or the patient’s biology is responsible differs by case and is still under study. Intellia reported an association with one HLA allele in a genomic analysis of over 600 samples (Intellia, Q2 2026).

Recent advances in reaching other organs

A 2020 Nature Nanotechnology paper from Daniel Siegwart’s group described adding a supplemental “SORT molecule” to lipid nanoparticles to redirect them to the lung and spleen as well as the liver, with selective editing in several cell types (Cheng et al., 2020). In 2023, Li and colleagues in Nature Biotechnology screened a library of biodegradable lipids for inhalable particles. The lead formulations tolerated repeated dosing into the windpipe and edited lung epithelial cells in animals, suggesting a path for lung diseases (Li et al., 2023). Both are animal studies, not human results.

How a nanoparticle gets an editor into a cell

The sequence is worth spelling out, because each step is a place where things can fail.

  1. Packaging. The mRNA and guide RNA are mixed with ionizable lipids and other lipid components, which self-assemble into particles.
  2. Circulation. After infusion, the particles travel through the blood. Proteins attach to their surface and influence where they end up, which is one reason a small change to the lipid recipe can move the destination, as the SORT work showed.
  3. Uptake. Target cells take the particle in.
  4. Escape. Inside the cell the particle must release its cargo into the cell’s interior. Most of the chemistry in the lipids exists to make this step work.
  5. Translation and editing. The cell reads the mRNA, builds the editor, and the editor pairs with the guide and edits the DNA.
  6. Clearance. The mRNA and protein degrade within days, leaving the edit and nothing else.

Viral vectors: borrowing a virus’s delivery skills

AAV is a small virus stripped of its disease-causing genes and loaded with a therapeutic payload. It reaches many tissues, and some natural versions favor particular organs. AAV is the established workhorse of gene therapy, including the vector that carries opsin genes to the retina in optogenetics trials, discussed in our piece on the 2026 optogenetics Nobel Prize.

The central limit is cargo size. AAV can only package a small amount of DNA, and the common Cas9 from Streptococcus pyogenes is large. In 2015, Ran and colleagues reported a smaller Cas9 from Staphylococcus aureus that edits with similar efficiency while being more than 1 kilobase shorter. It fit with its guide cassette into a single AAV vector and, aimed at Pcsk9 in mouse liver, modified more than 40 percent of the gene within a week (Ran et al., Nature, 2015). In 2021, a research team went smaller still, redesigning the guide RNA of an ultra-compact nuclease, Cas12f1, to raise average editing about 867-fold and enable delivery from a single AAV (Kim et al., Nature Biotechnology, 2022).

The other limit is persistence. A virus-delivered gene can keep producing editor for a long time. For a therapy meant to cut a gene once, extended presence means extended opportunity for off-target cutting, a concern that regulators list among their first questions (FDA guidance, January 2024).

Two vehicles at once

Some groups combine them. The CHOP, Penn and Broad prime editing platform for urea cycle disorders uses a lipid nanoparticle for the editor’s mRNA and an engineered AAV for the guide RNAs. In preclinical liver work, giving the virus first and the nanoparticles two weeks later corrected about 30 to 40 percent of copies of one variant, against a roughly 10 percent level the team considers necessary (CHOP, March 31, 2026). The split exists because a prime editor plus its long guide is awkward to carry in one package.

Splitting a system across two vehicles has a price. Two products must be made, tested and released as a pair, and the order and spacing of doses become part of the treatment. Regulators will want evidence for each piece and for the combination. The CHOP team has described meeting the FDA about a trial across seven urea cycle genes, but no start date was given, so this remains a plan rather than a result.

The workaround: edit the cells outside the body

Delivery gets easier when the cells come to the editor. In ex vivo therapies, blood stem cells are collected, edited in a lab and returned. No vehicle has to find its way through the body. The cost is a heavier procedure. In Prime Medicine’s chronic granulomatous disease data, the company said all observed toxicities were consistent with busulfan-based conditioning, the chemotherapy step that clears space for the edited cells (Prime Medicine, December 2025). Casgevy, the approved ex vivo therapy, is delivered through a network of authorized treatment centers, and Vertex reported $43 million in first-quarter 2026 revenue (Vertex, Q1 2026). Better in vivo delivery is partly an attempt to make editing possible without that burden.

Side-by-side comparison

Question Lipid nanoparticles AAV vectors
Typical cargo mRNA and guide RNA DNA encoding the editor and guide
How long does the editor persist? Briefly Potentially much longer
Size limit Fairly forgiving Tight; needs small Cas proteins or split systems
Natural destination Liver, unless redesigned Depends on the virus type
Repeat dosing in humans Reported (three doses for baby KJ, repeat-dose cohort in the Beam trial) Generally harder, as a rule of thumb, but not verified here
Human editing data so far Most in vivo editing trials above Used in the CHOP preclinical platform and in optogenetics, not in the editing trials covered above

What to watch next

  1. Human data from nanoparticles that reach beyond the liver.
  2. Whether the liver enzyme signals in several programs turn out to be tied to a particular lipid, a particular editor, or specific patient genetics.
  3. Smaller editors, such as hypercompact nucleases, moving from lab papers to clinical programs.
  4. The prime editing urea cycle platform advancing from animals to a first-in-human study.

This is general scientific background, not medical advice.

Sources