What Is an AAV Vector? The Delivery System Behind Most Gene Therapies

If gene therapy is the idea of fixing broken genes, then the most important question is: how do you actually get the fix inside a patient's cells? You can have the perfect therapeutic gene sitting in a test tube, but it's useless unless it can reach the right tissue, enter the right cells, and start working.

That is where AAV comes in. Adeno-associated virus, or AAV, is the most widely used delivery vehicle — or vector — in gene therapy today. It powers the majority of approved gene therapies and dominates clinical trial pipelines worldwide. Understanding AAV is essential to understanding modern gene therapy.

What Is AAV, Exactly?

AAV stands for adeno-associated virus. It is one of the smallest known viruses, measuring only about 25 nanometers in diameter — roughly 4,000 times smaller than the width of a human hair. It was first discovered in 1965 as a contaminant in preparations of adenovirus (the virus behind many common colds), which is how it got its name: it was "associated" with adenoviruses.

Here is the key fact that makes AAV so attractive for medicine: AAV is not known to cause any disease in humans. While most people have been exposed to wild-type AAV at some point in their lives, the virus has never been linked to any illness. It is what scientists call non-pathogenic. Compare that to other viruses used in gene therapy — lentivirus is derived from HIV, and adenovirus causes respiratory infections. AAV's clean safety record gives it a head start.

Think of AAV as an extremely small shipping container. In nature, it carries its own tiny genome (a single strand of DNA about 4,700 nucleotides long) wrapped inside a protein shell called a capsid. The capsid is made up of 60 protein subunits that assemble into an icosahedral shape — picture a soccer ball at the molecular scale. This capsid determines which cells AAV can enter and how efficiently it delivers its cargo.

How Scientists Engineer AAV for Gene Therapy

Wild-type AAV is not ready for clinical use straight out of nature. Scientists have to re-engineer it. The process involves three major steps:

Step 1: Gutting the Viral Genome

In its natural form, AAV's genome contains two genes: rep (which helps the virus replicate) and cap (which encodes the capsid proteins). For gene therapy, scientists remove both of these genes entirely. What remains are the inverted terminal repeats (ITRs) — short DNA sequences at each end of the genome that act as bookends. The ITRs are essential because they signal the cell's machinery to process and maintain the DNA, but they carry no viral instructions.

The analogy: imagine taking a postal envelope, throwing away the letter inside, and keeping only the envelope itself. The envelope (ITRs) still works perfectly for mailing — you just need to put a new letter (your therapeutic gene) inside.

Step 2: Inserting the Therapeutic Gene

Between those ITR bookends, scientists insert the therapeutic gene — the working copy of whatever gene is broken or missing in the patient. This therapeutic construct typically includes:

• A promoter — a DNA sequence that tells cells when and where to turn on the gene (some promoters are active everywhere; others are specific to liver cells, or muscle cells, or neurons)
• The therapeutic gene itself (or a shortened, optimized version of it)
• A polyadenylation signal — a sequence that tells the cell where the gene's message ends

This engineered DNA package is called the transgene cassette, and it replaces the original viral genome completely.

Step 3: Producing the Vector

Since the AAV's own replication and capsid genes have been removed, the engineered virus cannot replicate on its own. To produce AAV vectors at scale, scientists use a manufacturing system (usually in HEK293 cells — a human cell line) that provides the rep and cap genes on separate DNA molecules called helper plasmids. The cells assemble complete AAV particles: capsids on the outside, therapeutic transgene on the inside.

The result is a vector that looks like AAV on the outside and can enter cells just like the wild-type virus, but carries a therapeutic gene instead of viral genes. It cannot replicate. It cannot cause infection. It is, in essence, a molecular delivery truck with no engine of its own — a one-way trip.

AAV Serotypes: Different Addresses for Different Tissues

One of the most powerful features of AAV is that it comes in many natural variants, called serotypes. Each serotype has a slightly different capsid structure, which means each one binds to different receptors on cell surfaces and enters different tissues with different efficiencies. This property is called tissue tropism — the natural preference of a virus for certain cell types.

Think of serotypes like different shipping companies. FedEx might be best for overnight packages, UPS for heavy freight, and the postal service for letters. Similarly, different AAV serotypes excel at reaching different organs:

AAV2 — The original workhorse. AAV2 was the first serotype used in clinical trials and remains the most extensively studied. It has broad tropism but is particularly effective in the eye and the liver. It was used in Luxturna, the gene therapy for inherited retinal dystrophy approved by the FDA in 2017. However, AAV2 is also the serotype that most people have pre-existing antibodies against, which limits its use for systemic (intravenous) delivery.

AAV5 — Shows strong tropism for the liver, lungs, and central nervous system. AAV5 is used in Hemgenix, the gene therapy for hemophilia B approved in 2022 (at the time, the most expensive drug ever at $3.5 million per dose). It has lower seroprevalence than AAV2, meaning fewer patients have pre-existing antibodies that would neutralize it.

AAV8 — Highly efficient at transducing (entering and delivering genes to) liver cells. AAV8 has become a favorite for liver-directed gene therapies because it reaches hepatocytes at high rates after intravenous injection. It is used in several clinical programs targeting hemophilia A, metabolic liver diseases, and other conditions.

AAV9 — The go-to serotype for reaching the central nervous system and muscles. AAV9 can cross the blood-brain barrier, which most AAV serotypes cannot do efficiently. This makes it invaluable for neurological conditions. AAV9 is the vector behind Zolgensma (onasemnogene abeparvovec), the gene therapy for spinal muscular atrophy (SMA) approved in 2019. It also transduces cardiac muscle and skeletal muscle effectively.

AAVrh74 — Originally isolated from rhesus monkeys. AAVrh74 shows excellent tropism for skeletal muscle and is the vector used in Elevidys (delandistrogene moxeparvovec), the gene therapy for Duchenne muscular dystrophy that received accelerated FDA approval in 2023. Its muscle tropism and relatively low seroprevalence make it attractive for neuromuscular diseases.

Researchers have now cataloged over 100 natural AAV serotypes and variants, and that number continues to grow as new ones are discovered in human and non-human primate tissues.

How AAV Delivers Genes to Cells

The journey of an AAV vector from injection to gene expression follows a precise series of biological steps:

1. Binding — The AAV capsid attaches to specific receptors on the surface of the target cell. Which receptors it recognizes depends on the serotype. For example, AAV2 binds to heparan sulfate proteoglycans, while AAV9 uses galactose as a primary receptor.

2. Internalization — The cell engulfs the AAV particle through a process called endocytosis, pulling it inside a small membrane-bound compartment (an endosome).

3. Endosomal escape — The AAV must escape the endosome before it gets routed to a lysosome and destroyed. Acidification of the endosome triggers conformational changes in the capsid that help it break free into the cytoplasm.

4. Nuclear entry — The AAV particle travels to the cell nucleus and enters through nuclear pores. This is a critical step — the therapeutic gene needs to reach the nucleus where gene expression occurs.

5. Uncoating — Inside the nucleus, the capsid disassembles and releases the single-stranded DNA transgene.

6. Second-strand synthesis — Because AAV carries single-stranded DNA, the cell must convert it to double-stranded DNA before it can be read. This step can take weeks and is one reason AAV gene expression has a slow onset. (Scientists have developed self-complementary AAV vectors that fold into double-stranded DNA immediately, bypassing this bottleneck — but at the cost of cutting the already small cargo capacity in half.)

7. Gene expression — The double-stranded transgene DNA forms circular structures called episomes that persist in the nucleus alongside the cell's chromosomes but typically do not integrate into them. The cell's machinery reads the transgene and produces the therapeutic protein.

Because the transgene exists as an episome rather than integrating into chromosomes, it will be lost when a cell divides — the episome goes to one daughter cell but not the other. This is why AAV works best in tissues with cells that rarely divide, such as the liver, brain, retina, and mature muscle.

Why AAV Is the Most Popular Gene Therapy Vector

AAV dominates the gene therapy landscape for several converging reasons:

Safety track record. Decades of clinical experience have established AAV as generally safe. It does not cause disease, and its transgene typically stays episomal (not integrated into the genome), which reduces the risk of insertional mutagenesis — the phenomenon where a gene inserts in the wrong spot and disrupts a tumor suppressor gene or activates an oncogene.

Long-lasting expression. In non-dividing cells, AAV-delivered transgenes can persist for years, potentially for the lifetime of the cell. This offers the tantalizing promise of a one-time treatment for chronic genetic diseases.

Broad tissue targeting. With multiple serotypes available, researchers can select — or engineer — capsids optimized for nearly any organ or cell type.

Established manufacturing and regulatory pathways. AAV vectors have been produced at clinical and commercial scale for years. Regulatory agencies (FDA, EMA) have extensive experience reviewing AAV-based therapies, which streamlines the approval process compared to entirely novel vector platforms.

Clinical validation. As of early 2026, AAV is the basis for multiple approved gene therapies: Luxturna (inherited blindness, AAV2), Zolgensma (SMA, AAV9), Hemgenix (hemophilia B, AAV5), Elevidys (DMD, AAVrh74), Roctavian (hemophilia A, AAV5), and Upstaza (AADC deficiency, AAV2). Hundreds more AAV-based therapies are in clinical trials. That breadth of real-world evidence gives physicians, regulators, and investors confidence in the platform.

The Limitations of AAV

For all its strengths, AAV has significant limitations that the field is actively working to overcome.

The 4.7 kb Packaging Limit

AAV can carry only about 4,700 base pairs of DNA — roughly 4.7 kilobases (kb). That sounds like a lot until you realize that many human genes are far larger. The dystrophin gene responsible for Duchenne muscular dystrophy, for instance, has a coding sequence of about 11,000 base pairs. It simply cannot fit inside an AAV capsid.

Scientists have developed workarounds. For dystrophin, they use micro-dystrophin — a shortened version of the gene that encodes a smaller but partially functional protein. This approach works (it is the basis of Elevidys), but the truncated protein may not provide the full function of the original. Other strategies include splitting a large gene across two AAV vectors (dual AAV systems) that recombine inside the cell, though this is less efficient.

For diseases caused by very large genes — such as certain forms of inherited deafness (otoferlin), some muscular dystrophies, and cystic fibrosis (CFTR) — the packaging limit remains a serious engineering challenge.

Immune Response

The human immune system is not fooled by AAV's benign nature. When trillions of AAV particles flood the bloodstream during intravenous delivery, the body mounts an immune response.

Innate immunity can trigger inflammation within hours of dosing. At high doses, this inflammatory response can damage the liver (hepatotoxicity) or cause systemic inflammatory reactions.

Adaptive immunity produces antibodies against the AAV capsid and activates T cells that can destroy cells harboring the vector. This immune response is typically managed with immunosuppressive drugs (corticosteroids) given around the time of treatment, but it complicates care and adds risk.

The Re-Dosing Problem

Once a patient receives an AAV vector, their immune system develops long-lasting neutralizing antibodies against that capsid. If the same serotype is administered again, those antibodies will bind to the AAV particles and destroy them before they can reach target cells. This means AAV gene therapy is essentially a one-shot treatment — if the first dose doesn't work well enough, or if gene expression fades over time, giving a second dose of the same vector is unlikely to succeed.

This is one of the biggest unsolved problems in the field. Researchers are exploring strategies to get around it: using different serotypes for the second dose, administering IgG-degrading enzymes (like imlifidase) to temporarily clear antibodies before re-dosing, or developing capsids that evade the immune system entirely.

Pre-Existing Antibodies

Because wild-type AAV circulates naturally in the human population, many people already carry antibodies against common serotypes before they ever receive gene therapy. Depending on the serotype and the population studied, 30% to 70% of people have pre-existing neutralizing antibodies that would inactivate the vector on contact.

This means that a significant fraction of patients who could benefit from an AAV gene therapy are ineligible simply because their immune systems have already seen a similar virus. Patients must be screened for anti-AAV antibodies before treatment, and those who test positive are excluded from clinical trials and, in many cases, from receiving the approved therapy.

Hepatotoxicity and Thrombotic Microangiopathy

Two serious safety signals have emerged in high-dose AAV programs:

Hepatotoxicity — At high systemic doses, AAV can cause severe liver inflammation. This has been observed particularly in clinical trials for Duchenne muscular dystrophy and other conditions requiring high vector doses. In rare cases, liver failure and death have occurred. The mechanism likely involves both direct cellular toxicity from overwhelming the liver with viral particles and immune-mediated destruction of transduced liver cells.

Thrombotic microangiopathy (TMA) — A condition involving blood clots in small blood vessels, low platelet counts, and organ damage. TMA has been reported in multiple high-dose AAV gene therapy trials, particularly those using intravenous delivery. It is thought to be related to complement activation triggered by the massive number of capsid proteins entering the bloodstream.

These safety concerns have led to FDA clinical holds on several programs and prompted the field to pursue lower-dose strategies, more targeted delivery routes (direct injection rather than IV), and capsids engineered for greater potency at lower doses.

Manufacturing Complexity and Cost

If the biology of AAV is challenging, the manufacturing is equally daunting.

The Production Process

AAV vectors are biological products, not small-molecule drugs. You cannot synthesize them with simple chemistry. Instead, they are grown in living cells — typically HEK293 cells in large bioreactors or insect cell systems using baculovirus. The process involves:

1. Growing billions of producer cells
2. Introducing the plasmids or viral constructs needed to assemble AAV
3. Harvesting the crude cell lysate containing AAV particles mixed with cellular debris and empty capsids
4. Purifying the full (transgene-containing) AAV particles away from empty capsids, host cell proteins, and DNA contaminants
5. Formulating, filling, and testing the final product under strict GMP (Good Manufacturing Practice) conditions

Each of these steps has failure modes. Yields can be inconsistent. The ratio of full-to-empty capsids varies batch to batch — and empty capsids contribute to immunogenicity without delivering any therapeutic benefit. Analytical characterization of AAV is notoriously difficult; unlike a chemical drug where you can verify the exact molecular structure, a biological product requires dozens of assays to assess potency, purity, identity, and safety.

The Cost Problem

The difficulty of manufacturing, combined with the small patient populations for most genetic diseases, results in staggering prices. Zolgensma costs approximately $2.1 million per patient. Hemgenix launched at $3.5 million. Elevidys is priced at $3.2 million.

These prices reflect several realities: the cost of building and operating GMP manufacturing facilities capable of producing AAV at scale; the decades of R&D investment; the small number of patients over which to spread fixed costs; and the one-time-treatment model, where the drug company must recoup lifetime value from a single administration.

The manufacturing bottleneck is arguably the biggest barrier to making gene therapy accessible. Producing enough high-quality AAV for even a few thousand patients per year stretches the capacity of most manufacturers. Companies like Catalent, Thermo Fisher (through its Brammer Bio acquisition), and FUJIFILM Diosynth Biotechnologies have invested billions in expanding AAV production capacity, but demand continues to outstrip supply.

Alternatives to AAV

AAV is not the only gene therapy vector. Several alternatives are gaining ground, each with distinct strengths:

Lipid Nanoparticles (LNPs)

LNPs are tiny fat bubbles that encapsulate nucleic acids — most famously, the mRNA in the Pfizer and Moderna COVID-19 vaccines. For gene therapy, LNPs can deliver mRNA, siRNA, or even gene-editing components (such as CRISPR-Cas9 mRNA and guide RNA).

Advantages over AAV: LNPs are non-viral, so they don't trigger anti-capsid immune responses. They can be re-dosed. Manufacturing is more straightforward and scalable. There is no packaging size limit in the same sense — you can encapsulate larger nucleic acid payloads.

Disadvantages: LNP-delivered gene expression is transient (days to weeks for mRNA), so they're best suited for gene editing (where the edit is permanent even after the delivery vehicle is gone) rather than gene supplementation. LNPs currently have strong natural tropism for the liver, making it difficult to target other organs, though engineered LNPs with selective organ targeting (SORT) technology are advancing rapidly.

Intellia Therapeutics is pioneering LNP-based in vivo CRISPR gene editing, with clinical programs for transthyretin amyloidosis (NTLA-2001) and hereditary angioedema showing promising results.

Lentiviral Vectors

Derived from HIV but engineered to be replication-incompetent, lentiviral vectors integrate their payload directly into the host cell's chromosomes. This makes them ideal for ex vivo gene therapy, where cells are removed from the patient, modified in the lab, and returned.

Advantages over AAV: Larger packaging capacity (~8-10 kb), permanent gene integration (important for dividing cells like blood stem cells), no re-dosing problem for the modified cells themselves.

Disadvantages: Integration into chromosomes carries a small risk of insertional mutagenesis. Not practical for in vivo (direct injection) delivery due to immune responses and difficulty manufacturing at the needed scale. Ex vivo therapy requires complex cell processing and is expensive.

Lentiviral vectors power Casgevy and Lyfgenia (sickle cell/beta thalassemia), Skysona (cerebral adrenoleukodystrophy), and numerous CAR-T cancer therapies.

Virus-Like Particles (VLPs)

A newer approach uses engineered protein shells that resemble viral capsids but contain no viral genome — essentially empty shells loaded with gene-editing proteins (like Cas9) or ribonucleoproteins. Because the editing machinery is delivered as protein rather than DNA, expression is transient and self-limiting, which can improve safety.

Companies like Aera Therapeutics are developing VLP-based delivery for gene editing applications. The technology is still early-stage compared to AAV and LNPs.

Next-Generation Engineered Capsids

Perhaps the most exciting frontier in AAV research is the engineering of entirely new capsids that don't exist in nature — designed from the ground up to solve AAV's limitations.

Directed Evolution and Machine Learning

Traditional approaches to improving AAV capsids used directed evolution — creating vast libraries of capsid variants (millions or billions), packaging them into AAV particles, injecting them into animal models, and then recovering the variants that best reached the target tissue. Repeat for several rounds, and you enrich for capsids with improved properties.

This approach has yielded clinically relevant capsids like AAV-PHP.eB, which crosses the blood-brain barrier far more efficiently than AAV9 in mice (though translation to primates has been variable).

Now, companies are supercharging this approach with machine learning. Dyno Therapeutics, co-founded by George Church and Eric Kelsic, uses AI models trained on massive capsid sequence-function datasets to predict which mutations will improve tissue targeting, reduce immunogenicity, or increase packaging efficiency — without needing to test every variant in the lab. Their platform, CapsidMap, navigates the vast sequence space of possible AAV capsids far more efficiently than random mutagenesis. Dyno has partnerships with Novartis, Roche, Astellas, and Sarepta Therapeutics to develop next-generation capsids for specific disease programs.

4D Molecular Therapeutics takes a different approach with its Therapeutic Vector Evolution platform. 4D creates AAV capsid libraries using rational design guided by structural biology, then screens them in human tissue models and in vivo to identify variants optimized for specific clinical applications. Their pipeline includes 4D-150, an intravitreally delivered AAV vector for wet age-related macular degeneration, and 4D-710 for cystic fibrosis — a disease that has long been considered too difficult for AAV due to airway delivery challenges and the size of the CFTR gene.

What Engineered Capsids Could Solve

If successful, next-generation capsids could address many of AAV's current weaknesses:

• Lower doses needed: A capsid that targets the right tissue with 10x or 100x greater efficiency means you can inject far fewer total particles, reducing immune reactions and liver toxicity.
• Immune evasion: Capsids engineered to avoid recognition by pre-existing antibodies could open AAV therapy to the 30-70% of patients currently excluded.
• New tissue targets: Engineered capsids could unlock AAV delivery to tissues that are currently hard to reach, such as the lung epithelium, kidney, and specific brain regions.
• Re-dosing capability: If each dose uses a capsid with a distinct immunological profile, sequential dosing might become possible.

The Bigger Picture

AAV occupies a peculiar position in medicine. It is simultaneously the most proven, most trusted gene therapy platform and one with frustrating constraints that limit its reach. The 4.7 kb packaging limit, the one-shot dosing paradigm, the manufacturing costs, and the immune barriers are all real and significant.

Yet the trajectory is clear. Fifteen years ago, there was not a single approved AAV gene therapy. Today there are more than half a dozen, with hundreds more in clinical trials. The capsid engineering revolution — powered by AI, structural biology, and high-throughput screening — is producing vectors that are measurably better than their predecessors, generation after generation.

For patients with rare genetic diseases, AAV-based gene therapy is already life-changing. For the broader field, AAV is the platform on which the first wave of genetic medicine was built — and its evolution will shape the next wave.

---

Key Takeaways

• AAV (adeno-associated virus) is a small, non-pathogenic virus engineered to deliver therapeutic genes into patients' cells.
• Its viral genome is replaced with a therapeutic gene cassette flanked by ITR sequences.
• Different serotypes (AAV2, AAV5, AAV8, AAV9, AAVrh74) target different tissues — eye, liver, brain, muscle.
• AAV powers most approved gene therapies, including Luxturna, Zolgensma, Hemgenix, and Elevidys.
• Key limitations include a 4.7 kb packaging limit, immune responses preventing re-dosing, pre-existing antibodies in 30-70% of people, and safety risks at high doses.
• Manufacturing is complex and expensive, contributing to multi-million-dollar price tags.
• Alternatives include LNPs, lentiviral vectors, and virus-like particles.
• Next-generation engineered capsids from companies like Dyno Therapeutics and 4D Molecular Therapeutics use AI and directed evolution to build better AAV vectors.

---

Sources

1. Naso, M.F., et al. "Adeno-Associated Virus (AAV) as a Vector for Gene Therapy." BioDrugs, 31(4), 317-334, 2017.
2. Wang, D., Tai, P.W.L., & Gao, G. "Adeno-associated virus vector as a platform for gene therapy delivery." Nature Reviews Drug Discovery, 18(5), 358-378, 2019.
3. Mendell, J.R., et al. "Single-dose gene-replacement therapy for spinal muscular atrophy." New England Journal of Medicine, 377(18), 1713-1722, 2017.
4. Pipe, S.W., et al. "Gene Therapy with Etranacogene Dezaparvovec for Hemophilia B." New England Journal of Medicine, 388(8), 706-718, 2023.
5. Kelsic, E., et al. "Machine-guided capsid engineering for AAV gene therapy." Annual Review of Biomedical Data Science, 2024.
6. Paulk, N. "Gene Therapy: It's Time to Talk about High-Dose AAV." Genetic Engineering & Biotechnology News, 2020.
7. Verdera, H.C., Kuranda, K., & Mingozzi, F. "AAV Vector Immunogenicity in Humans: A Long Journey to Successful Gene Transfer." Molecular Therapy, 28(3), 723-746, 2020.
8. U.S. Food and Drug Administration. "Approved Cellular and Gene Therapy Products." FDA.gov, accessed March 2026.
9. Tabebordbar, M., et al. "Directed evolution of a family of AAV capsid variants enabling potent muscle-directed gene delivery across species." Cell, 184(19), 4919-4938, 2021.
10. Dyno Therapeutics. "CapsidMap Platform." dynotherapeutics.com, accessed March 2026.
11. 4D Molecular Therapeutics. "Therapeutic Vector Evolution." 4dmoleculartherapeutics.com, accessed March 2026.