AAV gene therapy at single-patient scale

The first published one-patient adeno-associated virus gene therapy trial described Michael Pirovolakis, a four-year-old boy in Toronto with hereditary spastic paraplegia type 50. The paper appeared in Nature Medicine in June 2024. Steven Gray's lab at UT Southwestern designed the vector. SickKids in Toronto dosed it. Michael's father, Terry Pirovolakis, raised the funding and held the investigational new drug application as the sponsor-investigator. The interval from diagnosis to first dose was less than three years. The follow-up at twelve months reported no serious adverse events, the disease appeared to be stabilizing, and Michael could stand with his heels on the ground for the first time.

The Pirovolakis trial was the proof of concept the AAV-for-one community had been waiting for, in the same way the milasen case in 2018 had been the proof of concept for n-of-1 antisense oligonucleotides. The two technologies are not interchangeable. They share a regulatory pathway and a manufacturing problem; they share almost nothing else.

What an AAV vector is

An adeno-associated virus is a small naturally-occurring virus that infects cells without causing disease. Wild-type AAV does not appear to cause any human pathology. Its small size and benign behavior made it an attractive candidate for engineered gene-therapy vectors in the 1990s, and it has since become the dominant platform for in vivo gene replacement.

A clinical AAV vector is a stripped-down particle. The viral protein shell, called the capsid, surrounds a single-stranded DNA payload. Engineers replace nearly all of the wild-type viral genome with a therapeutic cassette: a promoter, the coding sequence of the missing or defective gene, and regulatory elements to control expression. The cargo limit is approximately 4.7 kilobases. Genes longer than that, including the dystrophin gene that causes Duchenne muscular dystrophy at full length, do not fit in a single AAV particle.

The capsid serotype determines which tissues the vector reaches. AAV9 crosses the blood-brain barrier and reaches the central nervous system, which is why it has become the workhorse capsid for pediatric neurological gene therapy. AAV2 targets the retina. AAVrh10 reaches the central nervous system through different routes than AAV9. Newer engineered capsids, including the brain-targeted PHP.eB and the muscle-targeted MyoAAV variants, expand the tissue map further.

Once an AAV particle infects a cell, the cargo DNA enters the nucleus and exists primarily as an episome, a circular extrachromosomal element. A small fraction integrates into the genome. In non-dividing cells like neurons, the episome can persist and produce protein for years. In dividing cells, the episome is diluted with each division, and expression fades over time.

How the single-patient program shape works

The Pirovolakis program followed a sequence that other ultra-rare AAV programs have begun to mirror.

Michael was diagnosed with SPG50 at 18 months. The condition is caused by recessive mutations in AP4M1, which encodes a subunit of the AP-4 complex involved in protein trafficking. Approximately 80 children worldwide have been identified with the disease, with severe spasticity, intellectual disability, and seizures. There was no treatment.

Terry Pirovolakis, a Toronto IT executive, identified Steven Gray's UT Southwestern lab as the group most likely to be able to design an AAV9 vector for AP4M1. Gray's lab had already built a manufacturing template for AAV9 vectors targeting the central nervous system, and an AP4M1 program could reuse the capsid, the intrathecal route, and the dose-escalation framework from prior Gray-lab programs targeting CLN1 Batten disease, CLN5 Batten disease, CLN7 Batten disease, GM2 gangliosidosis, Rett syndrome, and giant axonal neuropathy. The development burden was to substitute the AP4M1 coding sequence into the existing vector design and validate the new construct.

Pirovolakis incorporated a non-profit, Cure SPG50, and raised approximately $3 million from family networks, friends, and small grants. The vector was manufactured at Children's GMP, an academic-affiliated good-manufacturing-practice facility at the University of North Carolina that produces small AAV batches for non-commercial trials. SickKids in Toronto, where Michael's clinicians had followed him from diagnosis, ran the dose. The FDA approved a parallel single-patient IND for treatment of Michael's siblings or other identified individuals if needed; in practice the dose was administered in Canada under Health Canada review.

The diagnosis-to-dose timeline of less than three years is itself the most important number in the case. It is approximately one third the timeline a small-pharmaceutical-company AAV gene therapy takes from program initiation to first dose, with much of the savings coming from skipping the company-formation, capital-raising, and intellectual-property steps that an academic-and-family program does not need.

Manufacturing is the harder problem than chemistry

AAV manufacturing is the central technical bottleneck of the field, and it is qualitatively different from antisense oligonucleotide manufacturing. An ASO is a small synthetic molecule produced by automated solid-phase chemistry; a clinical batch fits in a vial. An AAV vector is produced by transfecting human or insect cell lines with three plasmids (a transgene plasmid, a Rep/Cap plasmid encoding the capsid and replication proteins, and a helper plasmid), letting the cells assemble viral particles for several days, then harvesting and purifying the resulting mixture.

The mixture is heterogeneous in ways that ASO mixtures are not. A typical raw AAV preparation contains particles fully loaded with the therapeutic cassette, partially loaded particles, empty capsids with no DNA inside, and aggregates. Separating the full particles from the empty ones is a non-trivial purification step, typically involving cesium chloride density-gradient centrifugation or anion-exchange chromatography. Potency assays for AAV require infecting cultured cells and measuring transgene expression, a longer and more variable process than the analytical chemistry assays used for ASOs.

The cost is correspondingly higher. A small-batch academic AAV run for a single-patient trial typically runs in the low millions of dollars for the manufacturing alone, before toxicology, regulatory work, and clinical costs. The Pirovolakis program raised roughly $3 million; published commentary on other recent single-patient AAV programs reports total costs in the $3 million to $10 million range, several times higher than the $1.6 million Boston Children's reports for its individualized antisense oligonucleotide programs.

The platform and its safety database

Zolgensma, the AAV9-delivered gene replacement therapy for spinal muscular atrophy, was approved by the FDA in May 2019 and is the largest precedent for AAV9 in pediatric neurological disease. By 2024 more than 3,000 children had received Zolgensma. The safety profile is established: transient transaminase elevations are common, severe immune-mediated thrombotic microangiopathy is rare but documented, and a small number of children have died of acute liver failure following dosing. The class signal is real, monitoring protocols exist, and steroid pre-medication is now standard.

The Zolgensma database, combined with the Gray-lab clinical experience across multiple AAV9 intrathecal programs, gives sponsor-investigators a substantial safety reference for new programs in the same chemistry-and-route class. The FDA's draft guidance on individualized AAV gene therapies, which the agency has been developing since the analogous ASO guidance series of 2021 and 2022, leans on this precedent.

A specific engineering refinement Gray's lab has published is the miRARE element, a microRNA-responsive auto-regulatory feedback mechanism that limits transgene expression in cells that already produce sufficient endogenous protein. The miRARE design addresses a problem AAV gene therapy has that ASOs do not: overexpression toxicity. An ASO knocks down or modulates expression; an AAV adds a gene. If the added gene is too active in cells that already produce some baseline level of the protein, the result can be cytotoxicity. miRARE attempts to make the vector self-limiting.

Limits and risks

Pre-existing immunity is the first major limit. Approximately 30 to 70 percent of the general population, depending on age and geography, has neutralizing antibodies against AAV9 from prior natural exposure. Individuals with high antibody titers cannot receive AAV9 gene therapy because the antibodies neutralize the vector before it reaches its target tissue. The Pirovolakis trial screened for antibody titer and dosed only after confirming acceptable levels.

Re-dosing is essentially impossible. Once the cell has seen the AAV9 capsid, the immune response prevents a second dose from working. Programs requiring repeat dosing, including most chronic conditions, are poor AAV candidates unless durability of expression is high.

Durability is the second limit and is still partly unknown. Episomal expression in non-dividing cells should be long-lasting, but follow-up data on AAV9 patients beyond five to seven years is still being collected. Some loss of expression has been reported. Whether the loss is clinically meaningful depends on the disease.

Insertional mutagenesis is the third risk. The fraction of AAV cargo that integrates into the genome is small, but it is non-zero, and rare cases of integration-related cancer have been reported in animal models and in some adult AAV trials. Pediatric programs are watching this signal carefully.

The fourth limit is cost. The manufacturing complexity sets a floor on the cost of a clinical-grade AAV batch that small-batch academic facilities can lower somewhat but cannot eliminate. The Pirovolakis-class single-patient AAV program is a few million dollars more expensive than the milasen-class single-patient ASO program. For families raising the money themselves, the difference is material.

What the Pirovolakis case proved

What the SPG50 trial established, beyond the technical demonstration that an AAV9-AP4M1 vector could be designed and dosed in one child, is that the academic-plus-family program structure that worked for milasen also works for AAV. The same regulatory framework, the same single-patient IND, the same family-foundation funding model, the same academic vector-design partner. The difference is that the manufacturing burden is roughly twice as high and the durability question is open in ways that ASO programs do not face.

The Gray-lab platform now extends to several disorders that have one or a few affected families ready to fund a program. The pattern that ASOs established between 2018 and 2024, of academic groups treating one identified individual at a time and accumulating a safety database, is now visible in AAV programs as well. The infrastructure is younger and the cost per program is higher. The trajectory looks similar.