AAV Gene Therapy and Capsid Engineering: Challenges, Strategies, and Directed Evolution

Gene therapy has emerged as one of the most closely watched therapeutic modalities in modern medicine and is widely expected to enable disease-modifying or potentially curative treatments for a broad range of conditions driven by genetic abnormalities, including inherited genetic disorders, cancer, and certain chronic diseases.

In gene therapy, gene delivery vectors carrying therapeutic genetic material are introduced into patients to correct, replace, or compensate for defective or dysregulated genes, thereby restoring normal biological function.

Among available delivery platforms, recombinant adeno-associated virus (rAAV) has become one of the most widely used gene therapy vectors worldwide. Native AAV is non-pathogenic in humans and exhibits relatively low immunogenicity, making rAAV an attractive and clinically validated vector system. Recombinant AAV vectors can be engineered into multiple serotypes with distinct tissue tropisms, determined primarily by their capsid composition. Importantly, rAAV vectors can support long-term transgene expression in transduced cells while remaining predominantly episomal, with minimal integration into the host genome, positioning AAV as an ideal vehicle for exogenous gene delivery.

Key Challenges in AAV Gene Delivery

The central challenge in AAV-based gene therapy is efficient and specific gene delivery, which depends heavily on selecting an appropriate AAV serotype that matches the target tissue tropism. Despite clinical success, current AAV gene therapy approaches face several major limitations:

• Off-target tropism of natural AAV serotypes
  Most AAV serotypes used in research and clinical development are naturally occurring. For example, AAV9 is widely used for central nervous system (CNS) gene delivery due to its ability to cross the blood–brain barrier, but natural AAV serotypes often exhibit broad and imperfect tissue specificity, leading to unwanted transduction of non-target tissues.

• Pre-existing and induced anti-AAV immunity
  Natural AAV infections are common in the human population, resulting in pre-existing neutralizing antibodies (NAbs) that can significantly reduce AAV transduction efficiency. Even in AAV-naïve individuals, repeat dosing of AAV vectors typically induces strong humoral immune responses, limiting redosing potential and long-term therapeutic efficacy.

• Manufacturing and scalability limitations
  Not all AAV serotypes are equally amenable to large-scale, GMP-compliant manufacturing. Certain capsids exhibit low packaging efficiency, reduced vector yields, or batch-to-batch variability, making it difficult to consistently produce high-quality AAV at clinical or commercial scale. These challenges restrict broader clinical adoption of diverse AAV serotypes.

Engineering Next-Generation AAV Capsids

To overcome the inherent limitations of natural AAV serotypes, engineered AAV capsids can be generated by modifying native capsid sequences and subjecting them to systematic selection. The goal is to develop synthetic AAV serotypes with:

• Enhanced tissue specificity

• Improved transduction efficiency

• Reduced susceptibility to neutralizing antibodies

• Superior manufacturability and scalability

Directed Evolution of AAV Capsids

Directed evolution is a powerful strategy that mimics natural selection while dramatically accelerating evolutionary timelines to identify high-performance AAV capsids. Unlike rational design, directed evolution does not require prior structural or mechanistic knowledge of capsid–function relationships, making it particularly effective for discovering novel AAV variants with optimized properties.

Construction of AAV Capsid Libraries

The first and most critical step in AAV capsid directed evolution is the creation of a high-diversity AAV transfer plasmid library, in which each plasmid contains:

• A functional rep gene

• A mutated cap gene encoding variant AAV capsids

Several complementary strategies are commonly used to generate cap gene diversity:

1. Error-Prone PCR

Error-prone PCR introduces random mutations into the AAV cap gene by using suboptimal amplification conditions, such as low-fidelity polymerases, extended elongation times, elevated Mg²⁺ concentrations, Mn²⁺ supplementation, and imbalanced dNTP ratios.

2. Random Peptide Display

This method inserts random 7–9 amino acid peptides into defined, surface-exposed regions of the AAV capsid. These insertions alter natural interactions between AAV and cellular receptors, potentially redirecting tissue tropism or enabling immune evasion.

3. DNA Family Shuffling

DNA family shuffling recombines cap gene fragments from multiple parental AAV serotypes using primerless PCR based on sequence homology. This approach can be combined with rational design, where structural and sequence analyses guide the synthesis of mutated cap fragments that are reassembled into full-length genes.

4. In Silico and Machine Learning–Assisted Design

Computational approaches are increasingly used to predict capsid sequences with enhanced performance. Ancestral capsid reconstruction generates libraries of inferred ancestral AAV capsids for experimental validation. Machine learning models trained on large datasets linking capsid structure and function can also predict novel AAV variants with improved assembly, tropism, or manufacturability.

Packaging of AAV Capsid Libraries

One-Step Packaging

In the traditional one-step method, AAV transfer plasmid libraries and adenoviral helper plasmids are co-transfected into packaging cells. While widely used, this approach can lead to:

• Cross-packaging, where genomes and capsids are mismatched

• Capsid mosaicism, where capsids are composed of subunits from multiple variants

To mitigate these issues, extremely low plasmid-to-cell ratios are often used to ensure that each cell receives no more than one library plasmid.

Two-Step Packaging

In a two-step strategy, the capsid library is first packaged using a helper plasmid encoding wild-type capsid proteins but lacking ITRs, generating mosaic AAV “transfer shuttles.” These particles are then used to infect packaging cells at a low multiplicity of infection, followed by adenoviral superinfection to produce high-titer, genetically linked AAV capsid libraries.

Screening and Selection of AAV Capsid Libraries

Following library construction, iterative screening is required to enrich AAV variants with desired properties.

In Vitro Screening

Cell-based screening using established cell lines is commonly used to identify AAV variants with altered receptor usage or improved transduction efficiency. While rapid and technically straightforward, in vitro screening often fails to predict in vivo performance, including tissue specificity and immune interactions.

In Vivo Screening

Animal models provide a more physiologically relevant platform for AAV selection, especially for targeting complex tissues or hard-to-culture cell types. Both rodents and non-human primates (NHPs) are widely used. Due to their close genetic and physiological similarity to humans, NHP models—such as cynomolgus macaques (Macaca fascicularis) and rhesus macaques (Macaca mulatta)—are considered the gold standard for identifying clinically translatable AAV capsids. All in vivo studies are typically conducted in AAALAC-accredited facilities under internationally recognized ethical standards.

Conclusion

Advances in AAV capsid engineering, directed evolution, and in vivo screening are enabling the development of next-generation AAV gene therapy vectors with improved specificity, efficiency, immune evasion, and manufacturability. These innovations are critical to expanding the therapeutic reach of AAV-based gene therapy and addressing the limitations of natural AAV serotypes in clinical applications.