The adeno-associated virus (AAV) has emerged as one of the most promising gene therapy delivery vectors, owing to its reduced immunogenicity, variable tissue tropism, and capacity for long-term transgene expression in post-mitotic cells. For decades, that promise ran headfirst into a hard biological wall: the ~4.7 kb cargo limit of AAV vectors. Some of the most clinically urgent genetic diseases are caused by genes that simply don't fit in the recombinant AAV genome. Dual AAV strategies emerged as the field's most direct engineering answer to that constraint. Here we break down how they work, where they stand clinically, and what they actually demand from manufacturers trying to bring them to scale.
The 4.7 kb Wall: Why Single AAV Falls Short
Recombinant AAVs have become the delivery vehicle of choice in gene therapy because they can target specific tissues by capsid serotype, demonstrate low immunogenicity relative to other viral vectors, provide long-term stable expression without integrating into the host genome, and have shown meaningful clinical efficacy across muscle, CNS, retinal, and liver indications.
There is, however, a major constraint: the AAV cargo limit of approximately 4.5 to 4.7 kb must accommodate not just the coding sequence of the gene of interest, but all associated regulatory elements like promoters, enhancers, and poly-A signals. The practical ceiling is tighter than it may appear. Genes like dystrophin (Duchenne muscular dystrophy), ABCA4 (Stargardt disease), and MYO7A (Usher syndrome type 1B) all exceed 5 kb and cannot be packaged in a single AAV without truncation. When the limit is breached, the capsid cleaves the DNA, producing an incomplete payload, unpredictable expression, and a sharp drop in packaging yield. Since the physical dimensions of the capsid cannot be altered (at least with current technology), the only path forward is to engineer how the genetic cargo is organized.
ALSO READ: Compact CRISPR Nucleases and the Future of In Vivo Gene Editing
The Dual AAV Concept: Split, Deliver, Reassemble
Dual AAV systems distribute a large transgene across two separate vectors, each carrying one half of the gene. When both vectors transduce the same cell, the full-length transgene is reconstituted, and a functional protein is produced. Four main strategies drive reconstitution, each with different mechanisms and tradeoffs.
Overlapping
This strategy relies on homologous recombination between a shared sequence region present on both vectors. It is the simplest in design and introduces minimal foreign DNA, but its efficiency is directly tied to the rate of homologous recombination in the target tissue — which is generally low in post-mitotic cells.
Trans-splicing
This includes ITR-mediated concatemerization of the two genomes, followed by RNA splicing at a donor/acceptor junction to produce a full-length mRNA. There is no overlapping transgene sequence between vectors. Efficiency depends on concatemerization rates and clean splicing, both influenced by capsid serotype and tissue biology.
Intein-mediated Protein Trans-splicing
This is the most common strategy and operates post-translationally. Each vector expresses a protein half fused to a split intein domain. After independent translation in the same cell, the intein domains associate, catalyze excision, and ligate the two protein halves via a native peptide bond. Because reconstitution occurs at the protein level, it does not require genome interaction, only a significant advantage in post-mitotic tissues like the retina or CNS. The tradeoff is a small amino acid scar at the junction that must be tolerated by the target protein's structure.
Hybrid Dual AAV system
This strategy combines overlapping and trans-splicing elements (using both homologous recombination sequences and ITR concatemerization) and generally achieves higher reconstitution efficiency than either approach alone, at the cost of introducing additional exogenous DNA that requires regulatory justification.
From Bench to Clinic: Where Dual AAV Stands Today
Retinal indications have emerged as the leading proving ground for dual AAV. Subretinal injection creates a confined, immune-privileged compartment where co-transduction efficiency is naturally high that acts as the most forgiving environment for a two-vector system. Relatively low total doses are required compared to systemic indications, easing capsid load and immunogenicity burden.
ABCA4 / Stargardt disease is the most advanced dual AAV clinical program. At ~6.8 kb, ABCA4 has no engineering shortcut to bring it within single-AAV range. Both overlapping and intein-based approaches have demonstrated preclinical proof of concept in large animal models. From a process standpoint, this program has stress-tested dual vector co-formulation strategies and defined early expectations around acceptable vector ratio ranges in the drug product.
MYO7A / Usher syndrome type 1B was one of the earliest dual AAV targets to reach clinical investigation. The slow progress reflects a narrow intervention window and the difficulty of translating co-transduction efficiency from preclinical models to human subretinal administration, but manufacturing has not been the rate-limiting factor.
Dysferlin / LGMD2B illustrates the ceiling systemic delivery imposes. Preclinical muscle studies showed protein reconstitution, but co-transduction rates via intravenous delivery were substantially lower than in retinal models. The doses required to compensate are pushed into ranges that raise serious safety concerns. No program has cleared this bar in the clinic, and systemic dual AAV for muscle remains an open question.
The pattern is consistent: dual AAV performs best when anatomy and injection route conspire to maximize the probability that both vectors reach the same cell. When those conditions are met, manufacturing can deliver. When they are not, no amount of process optimization compensates.
What Dual AAV Actually Demands from Manufacturers
The scientific logic of dual AAV is sound, but the operational reality is that you have doubled your manufacturing problem — and the two halves are not independent.
Two productions, not one
Each vector requires its own upstream process development, downstream purification, and full analytical characterization. Every release test, including capsid titering, full/empty ratio, residual host cell protein and DNA clearance, must be performed twice on two distinct constructs before you have a releasable drug product. For organizations accustomed to single-AAV programs, this resource implication is consistently underestimated in early planning.
Vector ratio as a Critical Quality Attribute
The relative abundance of the two vectors in the final drug product directly drives reconstitution efficiency and therapeutic outcome. Too much of one vector relative to the other increases the probability of single-vector transduction, leading to half-product accumulation rather than functional protein. Defining an acceptable ratio specification, demonstrating process consistency, and building lot-release analytical methods to verify it adds a layer of CMC complexity with no equivalent in single-AAV development. Regulators will expect a scientific rationale for the chosen ratio, supported by dose-ratio studies conducted early enough to inform process design.
Co-production vs. separate production
Producing both vectors simultaneously in the same bioreactor via dual plasmid transfection offers the appeal of a single upstream run, but sacrifices independent control over each vector's yield and quality. Separate production preserves flexibility and simplifies troubleshooting but requires coordinating two campaigns and introduces scheduling complexity at fill-finish. The right answer depends on your platform, facility constraints, and how tightly the ratio specification must be held.
Half-product characterization
Cells receiving only one vector will express truncated protein fragments. Whether those fragments are inert, immunogenic, or dominant-negative depends on the gene and split point. Regulators will expect characterization and risk assessment. Building analytical tools capable of distinguishing full-length reconstituted protein from truncated half-products before IND-enabling studies avoids discovering a half-product liability during clinical manufacturing which is not the best time.
The organizations best positioned to advance dual AAV programs are those that treat manufacturing strategy as a scientific question from day one, not an operational detail to resolve after the biology is settled.
ALSO READ: Rare Disease Gene Therapy Economics: When Small Populations Make Business Sense
Dual AAV in a Crowded Field: Platform or Stepping Stone?
Dual AAV is one of the most ambitious genetic engineering solutions to a fundamental biological constraint in gene therapy. Early clinical programs in Stargardt and Usher 1B have demonstrated feasibility and a manageable safety profile, offering genuine hope to patients with no current treatment options.
However, the platform has entered development at a competitive moment. Base editing, prime editing, and LNP-mediated mRNA delivery are all maturing rapidly, each offering a potential route to the same patients through simpler mechanisms. The field has also shown that creative protein engineering can sometimes sidestep the problem entirely. B-domain deleted Factor VIII (BDD-FVIII) at ~4.4 kb is the canonical example, fitting within single-AAV capacity after removal of a functionally dispensable domain.
Whether dual AAV secures an approved product before alternative technologies converge on the same indications will largely determine its place in the gene therapy canon. If current trials yield clear, durable efficacy signals in the near term, dual AAV will have earned its platform status. If they do not, it risks being remembered as an important proof of concept: a transitional technology whose greatest contribution was revealing both the potential and the limits of AAV-based large-gene delivery and accelerating the search for something better.
For manufacturers evaluating whether to invest in dual AAV process capabilities, the retinal programs offer the clearest signal: technically tractable, clinically urgent, and advancing. That is a reasonable foundation on which to build.
ALSO READ: Making Base Editors Smaller for Gene Therapy
References
- Trapani I. (2019). Adeno-Associated Viral Vectors as a Tool for Large Gene Delivery to the Retina. Genes, 10(4), 287.
- Sweich et al. (2015). In vivo interrogation of gene function in the mammalian brain using CRISPR-Cas9. Nat. Biotechnol. 33, 102–106.
