Gene Therapy
AAV Vector
A 25-nanometer icosahedral capsid, 60 subunits, 4.7 kb of single-stranded DNA — and the dominant delivery vehicle of gene therapy
AAV is a small DNA parvovirus engineered into the workhorse gene-therapy vector. Icosahedral capsid of 60 subunits, ~25 nm diameter, packs 4.7 kb ssDNA. Serotypes give tissue-specific delivery: AAV2 to eye, AAV8 to liver, AAV9 across the blood-brain barrier.
- Diameter~25 nm (passes through nuclear pore)
- Subunits60 capsid proteins (VP1:VP2:VP3 ≈ 5:5:50)
- Cargo limit~4.7 kb ssDNA (incl. 145 nt ITRs)
- SymmetryIcosahedral T=1
- PersistenceEpisomal (mostly); concatemers in non-dividing cells
- Pre-existing immunity~30-70% of adults positive for AAV2 NAbs
Interactive visualization
Press play, or step through manually. The visualization is yours to drive — try it before reading on.
Watch the 60-second explainer
A condensed visual walkthrough — narrated, captioned, under a minute.
Worked example — packaging an SMN1 transgene in AAV9
The goal is a clinical-grade AAV9 batch encoding human SMN1 cDNA — the basis of Zolgensma. Here is what happens in the manufacturing run.
Plasmid 1 (transgene). A bacterial plasmid carries the ITR-flanked therapeutic cassette: 145 nt 5' ITR — chicken-beta-actin/CMV hybrid promoter (~1.7 kb) — human SMN1 cDNA (~885 bp coding + UTRs) — SV40 polyA (~225 bp) — 145 nt 3' ITR. Total ITR-to-ITR length: ~3.4 kb. This fits comfortably under the 4.7 kb limit.
Plasmid 2 (rep/cap). Carries the AAV2 rep genes (replication) and AAV9 cap genes (capsid). Pseudotyping by mixing rep and cap from different serotypes is how researchers separate genome production from capsid tropism.
Plasmid 3 (helper). Provides adenoviral helper functions — E2A, E4, VA RNA — which the wild-type AAV needs from a co-infecting adenovirus to replicate.
Production. 200 L of HEK293 cell suspension culture in WAVE bioreactors. Triple plasmid transfection by PEI complexes. Cells produce virions over ~72 hours. Lysis with detergent or freeze-thaw, treatment with nuclease to digest non-encapsidated DNA, clarification by depth filtration, capture by affinity chromatography (POROS AAVX), polish by ion exchange to separate full from empty capsids.
Fill and finish. Buffer exchange into formulation buffer (typically PBS + Pluronic + sucrose). Sterile filtration, fill into single-use vials. Final QC: vector genome titer by ddPCR, capsid titer by ELISA, full:empty ratio by analytical ultracentrifugation (AUC) or charge detection mass spectrometry, residual host cell DNA, residual helper plasmid, sterility, endotoxin, potency assay.
Dose math. Zolgensma label dose is 1.1 × 1014 vg/kg. A 7 kg infant needs 7.7 × 1014 vg — about 1015 capsids when accounting for empty capsid burden. At ~1014 vg/L typical yield, that is several hundred liters per patient — explaining the cost pressure.
From discovery to drug
- 1965. Atchison, Casto, Hammon discover AAV as a contaminant in adenovirus preparations — too small to replicate on its own.
- 1972. Hoggan, Berns characterize AAV as a defective parvovirus requiring helper.
- 1982. Samulski clones AAV2 into a bacterial plasmid — recombinant AAV (rAAV) technology born.
- 1996. First human AAV trial — AAV-CFTR for cystic fibrosis (lung, modest results).
- 2008-2010. Maguire and Bennett's Leber congenital amaurosis trial — AAV2-RPE65 subretinal injection restores vision; later becomes Luxturna.
- 2012. Glybera (AAV1-LPL) — first AAV product approved (EU; withdrawn 2017).
- 2017. Luxturna FDA approval — first US AAV gene therapy.
- 2019. Zolgensma — first systemic IV AAV gene therapy; the AAV9 + SMN1 product.
- 2022-2024. Six more AAV products (hemophilia A&B, Duchenne, RPE65, etc.) join the market; engineered capsids enter trials.
AAV serotypes commonly used in therapy
| Serotype | Primary receptor | Best tissue tropism | Used in | Notes |
|---|---|---|---|---|
| AAV2 | Heparan sulfate proteoglycan | Eye, CNS | Luxturna (subretinal RPE65) | Most studied; high pre-existing immunity |
| AAV5 | Sialic acid + PDGFR | Lung, CNS, liver | Hemgenix, Roctavian | Lower pre-existing immunity than AAV2 |
| AAV8 | Laminin receptor | Liver (strong) | Preclinical and clinical hepatic targets | Highly hepatotropic in primates |
| AAV9 | Galactose + LamR | CNS (BBB-crossing in young), heart, muscle, liver | Zolgensma (IV SMN1) | Broadest tissue distribution; BBB key |
| AAVrh10 | Multiple | CNS | Preclinical CNS targets | Rhesus isolate; engineered variants in dev |
| AAVrh74 | Multiple | Skeletal muscle | Elevidys (Duchenne) | Strong muscle tropism |
| AAV-PHP.B / .eB | Engineered | Mouse CNS (very high BBB crossing) | Research (limited primate translation) | Capsid evolution success story |
Why AAV dominates in vivo gene therapy
- Low immunogenicity. Unlike adenovirus, AAV does not trigger overwhelming acute immune responses at clinical doses.
- Durable expression. Episomal persistence in non-dividing cells lasts years to a lifetime.
- Tissue targeting. Serotype tropism + engineered capsids + tissue-specific promoters give multiplicative selectivity.
- Non-integrating. Reduced risk of insertional mutagenesis vs lentivirus or retrovirus.
- Established manufacturing. Scalable HEK293 and Sf9 baculovirus platforms; affinity capture by AAVX resin.
- Regulatory precedent. Six+ in vivo AAV products are now FDA-approved — agencies know how to review them.
- Smallness. ~25 nm capsid passes through tissue barriers and even nuclear pores intact.
Common misconceptions
- AAV integrates into the genome. Recombinant AAV is overwhelmingly episomal; integration is rare and at low copy number compared with retro/lentivirus.
- AAV cargo is unlimited. Hard 4.7 kb ssDNA ceiling. Large transgenes (Duchenne dystrophin 11 kb) require micro-dystrophin redesigns or dual-AAV strategies.
- One serotype fits all. Tropism is serotype-specific; choosing the wrong serotype wastes vector on off-target cells.
- You can redose freely. Anti-capsid neutralizing antibodies form within weeks of first dose; redosing the same serotype is generally foreclosed.
- Empty capsids are inert. Empty capsids carry the same antigenic load and contribute to T-cell responses; full:empty ratios are critical to product quality.
- AAV is harmless wild-type. No known wild-type AAV disease, but recombinant gene therapy products have caused serious adverse events at high doses.
Frequently asked questions
What is the structure of an AAV capsid?
AAV capsids are icosahedral with T=1 symmetry — 60 capsid protein subunits in 12 pentamers at the 5-fold vertices. Three capsid proteins (VP1, VP2, VP3) share an overlapping C-terminus from alternative translation start sites in one mRNA. Stoichiometry is ~5:5:50 (VP1:VP2:VP3). The diameter is ~25 nm — small enough to pass through nuclear pore complexes (the channel limit ~39 nm) intact. Surface features include the 3-fold protrusions (often the strongest determinants of tropism and antibody recognition) and 2-fold dimple. Inside, the capsid holds a single-stranded DNA genome of ~4.7 kb. The capsid mass is ~3.7 MDa with the DNA.
What is the cargo limit?
AAV's packaging limit is roughly 4.7 kb of single-stranded DNA, set by the physical volume inside the capsid. The therapeutic genome includes two ~145 nt inverted terminal repeats (ITRs) — the minimum cis element required for replication and packaging — leaving about 4.4 kb for the promoter, transgene, polyA, and any regulatory elements. Larger payloads must be split across two AAVs that reassemble after co-transduction (dual-AAV intein splicing, dual-AAV trans-splicing) — works in some tissues but reduces efficiency. Self-complementary AAV (scAAV) packages a folded dsDNA that doesn't require second-strand synthesis (faster expression) but halves the cargo capacity to ~2.4 kb.
How does AAV enter cells?
Receptor binding is serotype-specific. AAV2 binds heparan sulfate proteoglycan (HSPG) as primary receptor, αVβ5 integrin and FGFR1 as co-receptors. AAV5 uses N-linked sialic acid + PDGFR. AAV8 uses laminin receptor (LamR). AAV9 binds galactose (N-linked) + LamR and uses the AAVR (KIAA0319L) universal AAV receptor for endocytic uptake. After endocytosis, the capsid traffics through endosomes; phospholipase A2 activity in the VP1 N-terminal region helps endosomal escape at low pH. Capsids traffic to the nucleus, dock at the nuclear pore complex, and uncoat — slowly. Single-stranded DNA is released, converted to dsDNA by host machinery, and either persists as concatemerized episomes or, very rarely, integrates.
How do serotypes target different tissues?
Tropism arises from capsid surface residues that determine receptor binding and intracellular trafficking. AAV2 — eye (subretinal Luxturna), CNS (intracerebral). AAV5 — liver (Hemgenix, Roctavian), CNS (some), lung. AAV6 — muscle, hematopoietic. AAV8 — strong liver (preclinical and clinical hepatic targets). AAV9 — broad: CNS (crosses BBB in infants — used in Zolgensma), heart, skeletal muscle, liver. AAVrh10, AAVrh74 — engineered/rhesus serotypes used in Duchenne (Elevidys uses AAVrh74). Engineered capsids: Anc80 (ancestral reconstruction), AAV-PHP.B (mouse-CNS-tropic), AAV-PHP.eB. Directed evolution generates novel capsids with bespoke tropism.
How is AAV manufactured at clinical scale?
Triple-plasmid transfection of HEK293 cells. One plasmid carries the ITR-flanked transgene; one carries AAV rep/cap genes for the desired serotype; one provides adenoviral helper functions (E2A, E4, VA RNA). Cells produce virions over 48-72 hours; lysis releases capsids. Downstream purification: clarification, iodixanol gradient ultracentrifugation or CsCl, ion-exchange chromatography, affinity chromatography (POROS AAVX with camelid-derived antibodies), tangential flow filtration for concentration and buffer exchange. Yield: 10¹³-10¹⁵ vector genomes per liter of suspension culture. A single Zolgensma dose (~10¹⁴ vg/kg, ~10¹⁵ vg total) requires hundreds of liters of bioreactor culture. Sf9 insect cell baculovirus systems are an alternative scalable platform.
What are the safety risks?
Pre-existing neutralizing antibodies (30-70% of adults positive for AAV2; lower for AAV5/8/9) exclude patients and prevent redosing. Capsid T-cell immunity can cause transaminitis weeks after dosing — managed with steroid prophylaxis (Roctavian, Hemgenix). High systemic doses can trigger complement activation and thrombotic microangiopathy (Zolgensma >2×10¹⁴ vg/kg). Liver injury (some hepatic AAV products have boxed warnings). Off-target tissue expression. Rare integrations — orders of magnitude less than retroviral but not zero; FDA monitors for malignancy signals. Manufacturing variability — empty capsids (no genome inside) co-purify and increase immune burden; ratios of full:empty are critical quality attributes.
How does AAV compare to other delivery methods?
AAV vs lentivirus: AAV is non-integrating (episomal), smaller cargo (4.7 vs 8 kb), better for in vivo systemic dosing in non-dividing cells; lentivirus integrates, larger cargo, used for ex vivo HSC/T-cell modification. AAV vs adenovirus: AAV is far less immunogenic; adenovirus packs 36 kb but triggers strong T-cell responses and clears quickly. AAV vs LNP-mRNA: AAV gives years-long expression; LNP-mRNA is transient (days) but allows repeated dosing and integrates nothing. AAV vs naked DNA / minicircle: AAV is more efficient and tissue-targeted. For most in vivo applications targeting non-dividing tissue (eye, CNS, liver, muscle) requiring durable expression, AAV is the current state of the art.