Engineering the Future of Gene Therapy: Directed Evolution of Liver-Detargeted AAV Vectors for Systemic Muscle and Heart Delivery

Adeno-associated virus (AAV) vectors have emerged as leading vehicles for gene therapy, offering the potential to treat a wide range of genetic disorders. However, their clinical application has been hampered by significant challenges, particularly when systemic administration is required. For muscular dystrophies such as Duchenne muscular dystrophy (DMD), effective therapy necessitates high vector doses (>1×10¹⁴ vg/kg) to achieve sufficient vector genomes per muscle cell. These high doses result in substantial vector accumulation in the liver, leading to hepatotoxicity risks that have manifested in both preclinical studies and clinical trials, including elevated liver enzymes, liver failure, hepatic necrosis, and even patient deaths. The critical need to overcome this limitation has driven extensive research into capsid engineering to create vectors with improved tissue specificity and safety profiles. A recent online publication in Molecular Therapy – Methods & Clinical Development by researchers from REGENXBIO and Rutgers, The State University of New Jersey, reports how modern capsid engineering, integrating serotype screening, rational mutagenesis, and directed evolution, was applied to develop AAV capsids that reduce liver tropism while enhancing transduction in muscle and heart tissues.

The Significance of Capsid Engineering

Over two decades, AAV capsid engineering has yielded major breakthroughs. Early rational mutagenesis in AAV9 (e.g., H527Y, R535S) reduced liver transduction by altering receptor interactions and clearance. Directed evolution then enabled library screening in vivo, producing capsids like PHP.eB, which crossed the blood-brain barrier and lowered liver transduction versus AAV9.

For muscle targeting, engineered myotropic capsids with RGD peptide insertions, such as AAVMYO, achieved up to 61-fold higher expression in skeletal muscle, 11-fold in heart, and reduced liver activity. However, translation to non-human primates (NHPs) proved inconsistent. In NHPs, evolved myotropic capsids with RGD insertions in AAV9 VR-VIII boosted muscle and heart mRNA expression, but showed limited liver detargeting, with effects confined primarily to the transcript level.

These milestones underscore progress in tissue-specific targeting but also highlight challenges in achieving consistent cross-species translation and simultaneous target enhancement with liver detargeting. 

The Combined Approach: Screening Natural Serotypes, Targeted Mutagenesis, and Directed Evolution

In this study, researchers developed a strategy combining natural serotype screening, targeted mutagenesis, and directed evolution to identify AAV capsids optimized for skeletal muscle and heart delivery while limiting liver transduction. Screening 118 capsids in NHPs highlighted AAVhu.32, a clade F variant with reduced liver tropism and sequence similarities to AAV2, while AAV5 showed high vector genome levels in muscle but lacked mRNA expression, remaining valuable due to high yield and low antibody prevalence. Building on this, targeted mutagenesis of AAV9 with the “AAA” mutation (N496A/N497A/N498A) produced strong liver detargeting, 48-fold in NHPs and 117-fold in mice, though muscle expression decreased; other mutations, such as W503R and Q474A, were weaker, confirming AAA as the most robust. The core innovation was NAVIGATE (Novel AAV Vector Intelligent Guided Adaptation Through Evolution), a directed evolution platform that introduced 7-mer peptide libraries into capsid sites: position 455 in AAV9.AAA and position 578 in AAV5. Careful library design, including stop codons at insertion sites and plasmid optimization, reduced wild-type contamination to <1% and minimized cross-packaging. Libraries were administered intravenously to NHPs, and after three weeks, tissue analysis and sequencing identified enriched peptides, with the top 2000 sequences advanced for further testing. Iterative rounds of selection revealed promising variants, and validation studies identified two lead capsids: AAV9.AAA.NVG7 (VQVGRTS insertion) and AAV5.NVG13 (VRSDRDQ insertion), both of which showed reduced liver activity and improved muscle and heart targeting compared with AAV9.

Key Findings of This Study

The key findings of this study include the identification of liver-detargeted myotropic capsids, structural insights into functional epistasis, an enhanced safety profile in preclinical models, and unexpected neuronal targeting in the brain.

Two promising capsid variants, AAV9.AAA.NVG7 and AAV5.NVG13, were identified as liver-detargeted yet myotropic. In NHPs, AAV5.NVG13 achieved a 4.8-fold increase in skeletal muscle mRNA expression compared with AAV5, maintained similar expression in the heart, and reduced liver expression 1.9-fold relative to AAV5 and 7.7-fold compared with AAV9. AAV9.AAA.NVG7 demonstrated even stronger performance, showing 10-fold and 17-fold increases in skeletal muscle and heart mRNA expression versus AAV9.AAA, with an additional 2-fold improvement compared to AAV9. In the liver, it reduced mRNA expression by 7-fold compared with AAV9 and had 4-fold fewer vector genomes, while its mRNA/DNA ratio in skeletal muscle and heart was about 10-fold higher than AAV9, suggesting enhanced intracellular processing or transcriptional efficiency rather than increased uptake. Structural analysis revealed a striking example of sign epistasis, as the VR-V mutation (AAA) and VR-IV insertion (VQVGRTS) that define AAV9.AAA.NVG7 showed non-additive effects. Cryo-EM analysis indicated that the AAA mutation straightens the VR-V loop, disrupting VR-IV packing, while insertion of the VQVGRTS peptide causes outward bulging in wild-type AAV9 but compacts against VR-V in the presence of the AAA mutation. These findings suggest VR-V acts as a structural “core” element modulating adjacent regions like VR-IV, with implications for future capsid design. In terms of safety, AAV9.AAA.NVG7 demonstrated reduced liver activity in mdx mice, 4-fold lower than AAVhu.32, while maintaining comparable microdystrophin expression in skeletal muscle and heart, with 1.7-fold less leaky liver expression. In NHPs at 1×10¹⁴ vg/kg, AAVhu.32 caused transient increases in liver enzymes, whereas AAV9.AAA.NVG7 showed no significant liver toxicity, highlighting an improved safety profile. Unexpectedly, AAV9.AAA.NVG7 also shifted brain tropism, showing selective neuronal transduction: in the cortex, 90% of tdTomato+ cells colocalized with NeuN versus only 31% for AAV9, with similar effects seen for AAVhu.32.AAA.NVG7. This shift from astrocytic to neuronal targeting suggests new opportunities for intravenous delivery of neurological gene therapies.

Potential Clinical Impact and Future Directions

The capsid variants described in this study hold significant promise for improving the safety and efficacy of gene therapies for muscular dystrophies and other disorders requiring systemic administration. Reduced liver transduction observed with AAVhu.32, AAV9.AAA.NVG7, and AAVhu.32.AAA.NVG7 in both NHPs and rodents suggests these vectors may provide enhanced safety profiles compared to current AAV capsids, potentially enabling higher therapeutic doses without hepatotoxicity risks. The study further emphasizes the value of iterative optimization and combinatorial approaches in capsid engineering. Incorporating the AAA mutation and NVG7 peptide into AAVhu.32 yielded superior expression in skeletal muscle and heart compared to either modification alone, while maintaining liver detargeting, underscoring the potential of strategically combining beneficial modifications. Moving forward, several directions merit exploration, including further optimization of the lead capsids through additional rounds of directed evolution or rational design, and evaluation of their therapeutic performance in relevant disease models, particularly for muscular dystrophies. Mechanistic studies to elucidate the intracellular basis of enhanced transduction, as suggested by increased mRNA expression without corresponding vector genome increases, will be essential. The unexpected neuronal tropism of NVG7-containing capsids also opens opportunities for neurological applications. Additionally, comprehensive immunogenicity assessments will be crucial for clinical translation, while scale-up and manufacturing process development will be necessary for advancing these capsids toward clinical application.

Conclusion

The study highlights a pivotal innovation in AAV capsid engineering. By combining directed evolution with selective detargeting, the researchers generated AAV vectors capable of systemic delivery to skeletal muscle and heart without significant hepatic sequestration. This breakthrough represents a crucial step toward safer and more effective gene therapies for neuromuscular and cardiac diseases.

Looking ahead, translation to larger animal models and eventual clinical testing will determine whether these liver-detargeted capsids can overcome the dose and toxicity barriers that currently limit systemic AAV gene therapy. If successful, this methodology may become a standard template for engineering vectors with bespoke biodistribution tailored to therapeutic needs.

References

1. Directed evolution of liver-detargeted AAV vectors for systemic genedelivery to skeletal muscle and heart, Molecular Therapy – Methods & Clinical Development (2025), doi: https://doi.org/10.1016/j.omtm.2025.101571.
2. A Rationally Engineered Capsid Variant of AAV9 for Systemic CNS-Directed and Peripheral Tissue-Detargeted Gene Delivery in Neonates. Mol Ther Methods Clin Dev. 2018; 9:234–246. doi: 1016/j.omtm.2018.03.004
3. CNS Transduction Benefits of AAV-PHP.eB over AAV9 Are Dependent on Administration Route and Mouse Strain. Mol Ther Methods Clin Dev. 2020; 19:447–458. doi: 10.1016/j.omtm.2020.10.011
4. Identification of a myotropic AAV by massively parallel in vivo evaluation of barcoded capsid variants. Nat Commun. 2020; 11, 5432. https://doi.org/10.1038/s41467-020-19230-w
5. Fantastic AAV Gene Therapy Vectors and How to Find Them-Random Diversification, Rational Design and Machine Learning. Pathogens2022, 11: 756; https://doi.org/10.3390/pathogens11070756
6. Directed evolution of a family of AAV capsid variants enabling potent muscle-directed gene delivery across species. Cell. 2021; 184:4919-4938.e22. https://doi.org/10.1016/j.cell.2021.08.028
7. Scratching the surface of RGD-directed AAV capsid engineering. Mol Ther. 2021; 29:3099–3100. doi: 1016/j.ymthe.2021.10.020