Evolution of AAV Vectors: From Discovery to Therapeutic Applications

The review article “AAV Vector Development, Back to the Future” by Dr. Richard Jude Samulski and colleagues at M34, Inc.—a biotechnology company specializing in gene therapies that cross the blood-brain barrier to treat neurological diseases—provides a comprehensive overview of the historical development, molecular biology, engineering approaches, and clinical applications of adeno-associated virus (AAV) vectors in gene therapy. Published in Molecular Therapy, May 2025, the authors trace AAV’s evolution from its initial discovery as a contaminant to its current role as a premier gene delivery platform, highlighting pivotal scientific breakthroughs, technological innovations, and ongoing challenges.

Here, we extract and focus on selected sections of the paper that address key aspects of vector development, including:

• Historical milestones in AAV research
• Capsid biology, encompassing genome structure and capsid architecture
• Engineering strategies aimed at enhancing AAV vectors, such as directed evolution, rational design, and computational modeling
• Advances in AAV vector cassette design and strategies to improve transgene expression and therapeutic efficacy

This focused summary provides insight into the critical elements shaping the next generation of AAV-based gene therapies.

 1. Key Historical Milestones: From Contaminant to Clinical Success (Table 1)

• 1960s: AAV was first discovered as a contaminant in simian adenovirus preparations and was noted for its non-pathogenic, replication-deficient nature, which requires a helper virus to reproduce. This foundational discovery laid the groundwork for its future use as a safe vector.
• 1980s: This decade saw several breakthroughs, including the determination of the AAV2 ITR nucleotide sequence and the successful cloning of the wild-type AAV2 genome into a plasmid in 1982. This marked the beginning of AAV molecular biology and the ability to manipulate the viral genome for therapeutic purposes. In 1984, the first successful use of a vectorized AAV to deliver a gene into cultured mammalian cells was demonstrated, providing proof-of-concept for its potential in gene delivery.
• 1990s: The field moved toward in vivo applications and clinical trials.

  • In 1993, rAAV vectors were used for the first time in an in vivo therapeutic context for cystic fibrosis (CF), showing long-term gene expression in rabbits.
  • In 1994, AAV was used to deliver a gene to the mammalian brain, showing sustained expression and behavioral recovery in a rat model of Parkinson’s disease. This highlighted its potential for central nervous system gene therapy.
  • In 1995, the first human clinical trial using a rAAV vector for cystic fibrosis was conducted.
  • A major advancement for manufacturing occurred in 1998 with the development of a “helper-free” system for producing high-titer rAAV vectors, which improved scalability, reduced contamination risks, and enhanced clinical safety.

• 2000s and beyond: The focus shifted to improving vector efficiency and targeting.

• • • • In 2001, self-complementary AAV (scAAV) vectors were developed, which enhanced transduction efficiency by bypassing the rate-limiting step of second-strand DNA synthesis in the host cell.
      • In 2002, Rabinowitz et al. published a seminal study investigating the versatility of AAV vectors by packaging a single AAV genome into different capsid serotypes (e.g., AAV2, AAV5). This research highlighted the critical role of capsid serotypes in determining transduction efficiency and tissue specificity. By demonstrating that different serotypes could preferentially target specific tissues, such as the liver or CNS, this work opened new avenues for tailoring AAV vectors to specific therapeutic applications.
      • The mid-2000s saw the initiation of clinical trials using AAV vectors, building on preclinical successes. The review notes the first human AAV clinical trial for cystic fibrosis, which utilized AAV ITRs to drive transgene expression, demonstrating detectable CFTR mRNA and protein expression. Concurrently, researchers began exploring capsid engineering through mutagenesis strategies, such as degenerate oligonucleotide-based approaches, to generate diverse capsid libraries. These libraries, containing thousands to millions of variants, enabled the identification of novel capsids with enhanced transduction efficiency, tissue tropism, or immune evasion properties. The assembly of capsids, comprising VP1, VP2, and VP3 proteins in a 1:1:10 molar ratio to form 12 pentameric vertices, was also studied, revealing molecular heterogeneity influenced by serotype and production systems.
      • The 2010s marked a turning point with the FDA approval of AAV-based gene therapies, including Luxturna (2017) for inherited retinal dystrophy caused by RPE65 mutations and Zolgensma (2019) for spinal muscular atrophy caused by SMN1 mutations. These approvals validated AAV vectors as a robust platform for delivering therapeutic genes, demonstrating their ability to achieve clinically meaningful outcomes in rare genetic disorders. The success of these therapies underscored the importance of capsid design and promoter activity, as well as the safety of rAAV vectors in human applications.

 Table 1. AAV Historical Timeline

 2. Capsid Biology: Structure and Function

AAV has a single-stranded DNA genome of about 4.7 kilobases, flanked by inverted terminal repeats (ITRs). The genome contains two main genes, REP and CAP. The REP gene is responsible for genome replication and packaging, while the CAP gene encodes the viral capsid proteins VP1, VP2, and VP3.

The ITRs are crucial for DNA replication and packaging the genome into the viral capsid. Additionally, the genome encodes accessory proteins that help with capsid assembly and viral particle release. AAV’s non-pathogenic nature makes it an ideal vector for therapeutic applications.

The AAV capsid is a highly symmetrical structure composed of 60 viral protein subunits (VP1, VP2, and VP3) in a molar ratio of roughly 1:1:10. This structure is critical for all stages of the viral life cycle, from host-cell attachment to genome delivery. They are detailed below:

• VP3, the most abundant subunit, forms the structural backbone of the capsid and is essential for its integrity.
• VP1 and VP2 are present in smaller quantities but play crucial, complementary roles in viral infectivity. The unique N-terminal region of VP1 (uVP1) contains a phospholipase A2 (PLA2) domain, which is vital for endosomal escape, allowing the viral genome to enter the cytoplasm. The absence or mutation of this domain leads to a significant loss of infectivity. VP2, while not essential for capsid formation, enhances infectivity and assists in stabilizing the capsid and facilitating nuclear transport.
• The capsid’s surface features protrusions and depressions that mediate interactions with host-cell receptors, determining the virus’s tissue tropism and transduction efficiency.
• Post-translational modifications (PTMs) are also highlighted as “Hidden Architects of AAV”. These dynamic modifications can significantly impact capsid stability, host-cell interactions, and immune evasion (Table 2).

Table 2. AAV Capsid and Key Components

3. AAV Capsid Engineering and Blueprint for Evolution

Natural AAV serotypes possess excellent safety profiles but are hampered by limitations such as broad tissue tropism, inefficient transduction of certain cell types, and vulnerability to neutralizing antibodies. The need for vectors with improved properties led to three main engineering strategies:

• Directed evolution
• Rational design
• Combinatorial & Computational approaches

Directed Evolution: Large AAV capsid libraries are created and screened for desirable traits using methodologies like peptide insertion, error-prone PCR, domain swapping, and DNA shuffling. High-throughput platforms (CREATE, M-CREATE, iTransduce, BRAVE, TRACER, TRADE, DELIVER, SEBIR, etc.) have produced numerous novel variants with enhanced tropism, immune evasion, and cross-species compatibility.

Rational Design: Driven by advances in structural and computational biology, rational design enables targeted modification of AAV capsids based on known structure-function relationships, detailed receptor footprints, and site-directed mutagenesis to optimize tropism, transduction, and immune escape. Innovations include engineered PTM profiles and the design of chimeric vectors for clinical applications.

Combinatorial & Computational Approaches: Methods such as virtual (in silico) family shuffling, SCHEMA-guided design, AI-driven modeling, and machine learning–assisted fitness profiling have accelerated vector optimization with unprecedented depth and predictive power, underscoring the value of robust, unbiased datasets and iterative experimental validation.

Below is the historical perspective on the evolution of engineering strategies (Table 3):

3.1.Mutagenesis Applied to AAV — Key Early Studies:

• 1999: Samulski’s lab used insertional mutagenesis to map the AAV2 capsid, identifying mutationally tolerant domains.
• Girod et al. (1999): Demonstrated the insertion of integrin-binding RGD peptides into AAV2 surface loops, creating the first rationally retargeted AAV capsids.
• These efforts set the foundation for capsid library creation and high-throughput screening.

3.2 Library Design and Evolution Platforms:

3.2.1. Saturation Mutagenesis and Capsid Libraries

• Systematically mutated libraries led to breakthrough AAV variants (AAV2.GL, AAV2.NN) with enhanced cross-species transduction and intravitreal delivery.
• Error-prone PCR expanded library diversity and uncovered mutants with improved antibody evasion or altered tropism.

3.2.2. RGD Motif: A Paradigm for Targeting

• Rational insertion and evolutionary selection of the RGD integrin-binding motif in the AAV2 and later the AAV9 backbone led to pronounced muscle tropism and translationally promising vectors (e.g., MYOAAV).

3.2.3. New Era Screening Platforms

• CREATE and M-CREATE: Cre/loxP-based in vivo evolution platforms that foster the discovery of high-performing AAV capsids (e.g., AAV-PHP.B and derivatives) for CNS delivery.
• TRACER: Uses RNA output in target cells as a selection marker, enabling the rapid development of CNS-optimized capsids.
• DELIVER: Transcript-based in vivo selection across species for muscle-tropic capsids.
• Modern screens are moving away from transgenic dependencies to broaden applicability.

3.3. The DNA Shuffling Approach:

• 2008: Grimm et al. established DNA shuffling, combining capsid sequences from multiple serotypes to generate highly diverse, chimeric capsid libraries.
• Resulted in canonical engineered vectors like AAV-DJ, AAV-LK03, and AAV-Olig001, now widely used in gene therapy.
• In silico “virtual family shuffling” further narrows design to capsid variable regions using sequence and structural data, producing high-performance vectors such as SCH9.

3.4. Rational Design and Advances in Structural Biology:

3.4.1. Receptor Biology and Structure-Guided Design

• Discoveries of cell surface receptors and co-receptors (e.g., HSPG, αVβ5 integrin, FGFR1) provided an empirical basis for targeted engineering.
• Tyrosine phosphorylation of capsids revealed by Srivastava’s group led to PTM-targeted mutations (e.g., triple Y-F mutants, now in clinical trials).

3.4.2. Homology and Ancestral Reconstruction

• Using homology and sequence alignments, hybrid vectors like AAV2i8 redirected tropism from liver to muscle and decreased antibody sensitivity, now in phase 2 trials.
• Ancestral sequence reconstruction (Vandenberghe, 2015): In silico prediction of ancient AAV capsids (Anc80L65) resulted in synthetic vectors with broad, robust tropism.

3.5. Integration of Machine Learning and AI:

• Deep machine learning (Kelsic, 2021): Created and predicted the functionality of >200,000 AAV2 variants, dramatically expanding viable sequence space.
• APPRAISE (Gradinaru, 2024): Uses AlphaFold-based modeling to predict and rank engineered protein interactions, accelerating design.
• Fit4Function (Deverman, 2024): Systematically designed multi-trait AAV capsids (e.g., enhanced liver targeting), validated across species.

3.6. Cross-Species Translation: Challenges and Human-centric Approaches:

3.6.1. From Animal Models to Patients

• Despite sophisticated in vivo evolution systems, species differences in protein structure, immune response, and molecular biology create challenges for translation to human therapeutics.
• Many high-performing capsids in rodents or non-human primates fail in human validation.

3.6.2. In-Patient and Decedent Screening

• Phage display libraries have been injected directly into live patients to map vascular targets—well-tolerated and successful, but underutilized for AAVs.
• Post-mortem human tissue-based selection now enables the amplification and re-testing of human-adapted AAV capsids in animal models, representing possible “reverse translation.”

3.6.3. Will New Capsids Meet Clinical Needs?

• Despite expanding capsid libraries (e.g., AAV2.GL, MYOAAV, AAV-BI-hTFR1, AAV.k13/k20), only a minority are likely to reach clinical approval or surpass first-generation vectors.

3.7. Looking Forward: The Path to Precision Gene Delivery:

• Synergy between computation and experimentation is key: AI and ML can nominate candidates, but empirical validation is irreplaceable.
• Increasing dataset quality, reducing model bias, and ensuring relevant human data are essential for progress.
• Expanding humanized preclinical models and direct patient/donated tissue screens could finally close the gap between vector engineering and clinical success.

Table 3. Evolution of AAV Engineering Strategies

 4. Strategies and Advancement in AAV Vector Design

The paper also outlines the advancement of AAV vector design, focusing on strategies to enhance transgene expression and therapeutic efficacy, while addressing challenges such as limited packaging capacity and immune responses. The evolution of rAAV cassette design and engineering innovations is shown in Figure 1.

Key Strategies in AAV Vector Design:

• Cis-Acting Elements: Engineering inverted terminal repeats (ITRs), promoters, enhancers, splice elements, and polyA sequences to optimize gene expression. Short synthetic promoters and polyA sequences address AAV’s ~4.7 kb packaging limit.
• Transgene Cassette Design: Using cDNAs (0.5–2.5 kb) instead of large genomic sequences (average 62 kb) for compatibility with AAV. Codon optimization, CpG removal, and protein modifications (e.g., B-domain-deleted FVIII, Factor IX Padua) enhance activity.
• Promoter Development:

  • Constitutive Promoters: Early vectors used strong viral promoters like CMV (e.g., Glybera, Upstaza) and chimeric promoters like CBA and CAG (e.g., Luxturna, Zolgensma). These are effective but lack specificity and are prone to methylation, potentially reducing expression.
  • Tissue-Specific Promoters: Promoters like MHCK7 (muscle-specific, used in Elevidys) and LP1/HLP (liver-specific, used in Hemgenix, Roctavian) minimize off-target expression. Examples include CK8, hSYN1, and TBG for muscle, CNS, and liver targeting. Representative tissue- and cell-specific promoters are shown in Table 4.
  • Regulatable Promoters: Inducible systems (e.g., Tet-rtTA, epigenetic chemical control) offer controlled expression but are not yet in clinical trials due to safety concerns like immune responses from bacterial transactivators. A novel liver-specific inducible promoter showed >5,000-fold induction.

• ITR as Enhancer/Promoter: AAV2 ITRs exhibit promoter activity, driving transgene expression (e.g., CFTR in early trials). ITRs from AAV1–6 and AAV5 also show promoter function, but bidirectional activity may trigger immune responses via dsRNA pathways.
• Capsid-Genome Interaction: Recent studies reveal that capsid proteins (VP1/VP2) and promoters (e.g., CBA, CBh) jointly influence cell-specific transgene expression. For example, AAV9 with CBA promotes neuronal expression, while CBh shifts to oligodendrocytes, highlighting the role of capsid-promoter interplay.

Figure 1. Strategies and innovations in rAAV cassette design and engineering

Table 4. Tissue- and cell-specific promoters utilized in AAV vectors across preclinical research (ˆ), clinical trials, and approved drugs (#)

AAV vector design faces several significant challenges. The limited packaging capacity of AAV restricts the delivery of large transgenes, such as those for Duchenne muscular dystrophy (DMD) and factor VIII (FVIII), necessitating the use of minimal promoters and ITRs to fit within the size constraints. Constitutive promoters commonly employed in vectors carry the risk of off-target gene expression and are susceptible to transcriptional silencing through methylation. Additionally, the bidirectional promoter activity inherent to AAV inverted terminal repeats (ITRs) may provoke immune responses, which have been implicated in severe adverse events, including fatalities associated with high-dose AAV therapies. Furthermore, the interactions between capsid proteins and promoter elements often exhibit species-specific differences, complicating the translation of findings from animal models to human clinical applications.

Looking ahead, future advancements aim to address these challenges by developing compact, regulatable promoters alongside engineered ITRs designed to minimize immune activation. Investigating the intricate interplay between capsid-promoter elements and ITR-host factor interactions holds promise for enhancing vector specificity and safety. The integration of artificial intelligence, structural biology, and humanized preclinical models is expected to facilitate the rational design of vectors with optimized transgene expression and reduced immunogenicity. Finally, improving the quality of datasets and emphasizing empirical validation will be crucial for bridging the gap between preclinical successes and clinical efficacy.

References:

https://pubmed.ncbi.nlm.nih.gov/40186350/