Emerging Directions in AAV Capsid Engineering from ASGCT 2026

AAV capsid engineering emerged as a major theme at ASGCT 2026, reflecting the field’s need for gene delivery vectors that are more tissue-targeted, species-translatable, immune-compatible, and clinically scalable. Across the selected posters and presentations highlighted below, a clear trend is emerging: the field is moving beyond conventional serotype selection toward more deliberate, fit-for-purpose capsid design.

Researchers are increasingly using human-relevant tissues, non-human primate models, machine learning-guided library design, receptor-informed discovery, and route-specific screening to identify next-generation AAV capsids. The central question is no longer simply whether a capsid can transduce a target tissue, but whether it can do so at a clinically useful dose, with reduced off-target exposure, improved manufacturability, better immune compatibility, and stronger cross-species predictability.

Together, these representative ASGCT 2026 updates provide a snapshot of where AAV capsid engineering is heading and what developers should consider when designing the next generation of gene therapy vectors.

Identification of Natural Human-Derived AAV8 Variants with Enhanced Kidney-Tropic Properties

Kidney-directed gene therapy remains an important but difficult area for AAV development. Conventional AAV serotypes can show limited efficiency, heterogeneous renal distribution, and substantial off-target expression, especially in liver and other systemic tissues. In this study, Tapan Sharma and colleagues from UMass Chan Medical School / Horae Gene Therapy Center focused on identifying naturally occurring human-derived AAV8 variants with improved kidney tropism, aiming to develop capsids that could better support renal gene therapy applications.

The team isolated AAV variants from human tissue samples and used barcode-based library screening, sequencing, and in vivo evaluation to identify candidate capsids with enhanced kidney transduction. Selected variants were repackaged and compared against parental AAV8 and AAV9 controls using EGFP reporter expression, tissue biodistribution, histology, and molecular analysis of kidney and liver DNA/RNA.

Several AAV8-derived variants showed improved kidney signal compared with parental controls. One candidate, v806, appeared particularly promising, showing stronger kidney transduction and expression profiles than parental capsids. The poster also suggested that some variants may transduce renal tubule regions outside the glomeruli, which could expand their potential therapeutic relevance. Overall, this study supports the idea that naturally occurring human capsid diversity can be mined to identify organ-tropic AAV variants with improved translational potential.

Comparison of Viral Gene Therapy Vectors, AAV Capsids, and Delivery Routes on CNS and Systemic Organ Transduction

Efficient and safe CNS delivery remains one of the central challenges in AAV gene therapy. Rachel Searle and colleagues from the University of Edinburgh compared multiple AAV capsids and delivery routes, including intravenous, intracerebroventricular, and lentiviral vector delivery approaches, to better understand how route and capsid choice affect brain and systemic organ transduction.

The study evaluated AAV2, AAV9, AAVB1, AAVPHP.B, and AAVV66, with vector copy number and GFP expression assessed in CNS and peripheral tissues. By comparing direct CNS delivery with systemic delivery, the researchers examined how vector dose, route, and capsid identity shape transduction patterns in brain regions and peripheral organs.

The results suggested that direct ICV administration required substantially less vector to achieve CNS transduction than IV delivery, while systemic exposure varied by capsid. AAVPHP.B showed strong brain vector copy number following ICV delivery, whereas AAVV66 showed reduced liver affinity in some comparisons, potentially making it useful for systemic applications where lower hepatic exposure is desired. The study highlights that capsid performance cannot be separated from route of administration, and that delivery strategy remains a key determinant of both efficacy and safety.

Next-Generation AAV Capsids for Kidney Cell Transduction Validated in Human Kidney Explants, Renal Cell Systems, and Translational In Vivo Models

Kidney gene therapy has historically been limited by poor vector access to relevant renal cell types and by the weak predictive value of some animal models. Leszek Lisowski and colleagues from Nionyx Bio / University of Sydney presented NYX AAV capsids designed and validated through a translational workflow that includes human kidney explants, renal cell systems, and in vivo models.

The researchers administered candidate AAV capsids through renal artery delivery and screened them against benchmark vectors such as AAV-LK03 and AAV9 in healthy human kidney explants. Top candidates were further evaluated in RNA-isolated and intact-cell assays, disease-relevant podocyte models, and in vivo mouse and mini-pig studies.

The lead NYX candidates showed strong transduction across multiple kidney cell types, with NYX-04 highlighted as a particularly promising capsid. The poster emphasized that the same lead candidates were validated across human tissue, in vitro disease models, and in vivo systems, reducing the risk of species mismatch. This work illustrates an increasingly important trend in capsid engineering: using human tissue-first discovery and validation pipelines to improve the translational relevance of lead vector selection.

AAV Capsid Engineering and Spatial Control Strategies to Reduce Hepatotoxicity

PackGene presented new work on AAV capsid engineering and spatial transgene control, highlighting a strategy to improve the therapeutic window of systemic AAV gene therapy. The study, titled “Decoupling Efficacy From Toxicity: Engineering Spatial Control in AAV-Mediated Gene Therapy to Mitigate Hepatotoxicity,” focused on how engineered capsids, tissue-specific promoters, microRNA-based regulation, and localized delivery can work together to reduce off-target toxicity while preserving therapeutic efficacy.

A key focus of the presentation was AAV.eM, a next-generation myotropic capsid designed to improve muscle-directed delivery while reducing liver off-targeting. Compared with AAV9, AAV.eM showed reduced hepatic transduction and stronger muscle tropism, supporting its potential as a safer and more efficient delivery vehicle for muscle-directed systemic therapies.

In a humanized lymphoma mouse model, intravenous delivery of AAV.eM-MHCK7-αCD19αCD3 BiTE achieved complete lymphoma clearance, while the AAV9-based comparator failed to reduce tumor burden. This finding suggests that capsid choice can directly affect both biodistribution and therapeutic outcome, especially when the therapeutic payload requires sustained expression from a specific tissue compartment.

The study also introduced a microRNA-mediated spatial control strategy. By inserting miR-208a binding sites into the transgene 3′UTR, PackGene created a post-transcriptional “fail-safe” that nearly eliminated transgene expression in cardiac tissue while preserving skeletal muscle expression. Importantly, this cardiac silencing strategy maintained full tumor clearance efficacy, showing that safety control can be added without compromising therapeutic activity.

Another important result was the ability of intramuscular AAV.eM delivery to maintain efficacy at a lower dose. At 5e11 vg/kg, AAV.eM-MHCK7-BiTE preserved potent tumor growth restriction and complete eradication, while AAV9 showed diminished therapeutic activity. This supports the idea that improved capsid tropism can reduce required dose, lower manufacturing burden, and potentially improve safety.

Overall, PackGene’s study provides a practical framework for the next generation of AAV therapy design: use capsid engineering to improve tissue targeting, combine it with promoter and microRNA regulation to refine expression, and apply dose-sparing delivery strategies to reduce toxicity risk. This integrated approach may be especially valuable for systemic AAV programs where liver exposure, immune response, and dose-related toxicity remain major development challenges.

Machine Learning-Guided Platform for Engineering High-Performing AAV Capsids Through Systematic Sequence Optimization

Machine learning is becoming a major tool for AAV capsid design because it can help search large sequence spaces that would be difficult to explore through conventional directed evolution alone. Youngseo Kim and colleagues from AAVATAR Therapeutics introduced a closed-loop, AI-guided capsid engineering platform designed to generate production-optimized and tissue-targeted AAV libraries.

The workflow combined AAV serotype diversity, mutant library construction, in-house AAV manufacturing, in vivo screening, next-generation sequencing, and predictive modeling. The platform used models such as convolutional neural networks, support vector machines, and ensemble approaches to predict capsid assembly viability and optimize sequence selection before downstream screening.

The platform generated large AAV capsid libraries with high diversity while maintaining predicted viability. The poster emphasized that ML-guided sequence design could improve the efficiency of identifying candidates with desirable production and targeting characteristics. This approach reflects a broader field-wide shift from random capsid diversification toward data-driven, production-aware, and function-guided AAV engineering.

Discovery of In Vivo Tissue-Targeted Candidates Through Machine Learning-Guided AAV Capsid Library Screening

Tissue-targeted delivery remains one of the biggest barriers to safe and effective AAV gene therapy. Hosung Yu and colleagues from AAVATAR Therapeutics expanded on a machine learning-guided approach by applying ML-designed capsid libraries to in vivo screening for tissue-enriched candidates, including brain- and heart-targeted AAV capsids.

The team built focused AAV libraries using machine learning-guided sequence prioritization, screened them in vivo, and profiled tissue enrichment by NGS. Secondary screening of AAV9 mutant libraries was used to identify candidates enriched in brain and heart, followed by candidate refinement based on enrichment, reproducibility, and tissue selectivity.

The poster identified lead candidates with improved brain or heart enrichment compared with wild-type AAV9. Imaging and enrichment analyses supported prioritization of candidates for downstream validation. The work demonstrates how machine learning can be integrated with in vivo selection to move from sequence design to tissue-specific candidate nomination.

AI-Guided Capsid Engineering Enables In Vivo Selection of Neuron-Targeted AAV Vectors in Non-Human Primates

Translating CNS capsid performance from rodents to non-human primates remains a major challenge. Bumwhee Lee and colleagues from GenixCure focused on AI-guided AAV capsid engineering for neuron-targeted delivery in NHPs, with the goal of identifying candidates that achieve enhanced CNS transduction and improved translational relevance.

The researchers used directed diversification, AI-guided capsid design, and in vivo screening in non-human primates. The platform incorporated intracisternal magna delivery, comparative tissue enrichment analysis, and in vitro cross-species selection using receptor-targeting strategies. Candidate variants were evaluated for CNS enrichment, neuronal transduction, and peripheral distribution.

The poster reported enhanced CNS transduction after ICM delivery and identified BBB-penetrant or CNS-enriched variants through systemic and local delivery strategies. The study suggests that integrating AI design with NHP screening may improve the chance of identifying capsids that are more relevant to human CNS applications than rodent-only screening approaches.

Systematic Insertional Engineering of AAV5 Variable Regions Identifies a Permissive Capsid Engineering Site

AAV5 is attractive for capsid engineering because of its distinct phylogenetic and structural properties, but it has historically shown relatively low performance in some CNS applications. Lisa Melamed and colleagues from MIT / McGovern Institute investigated whether specific AAV5 variable regions could tolerate peptide insertion or loop engineering to improve capsid function.

The researchers systematically inserted peptide sequences into variable regions of AAV5 and related capsids, then evaluated the resulting variants using in vitro and in vivo assays. Functional loop deletion and loop swap experiments were used to identify regions that influence packaging and CNS transduction. The team also compared AAV5-derived engineering strategies with AAV9-related variants.

The poster identified VR4 and VR8 as permissive regions for AAV5 capsid engineering. Functional loop swap experiments suggested that transferring or modifying specific peptide regions could improve CNS transduction while preserving capsid viability. This work provides a useful structural framework for future AAV5-based capsid engineering and suggests that AAV5 may offer underexplored opportunities for targeted vector design.

Spatial Single-Cell Assessment of CapX Tropism, a Second-Generation Human Transferrin-Targeted AAV Capsid, at a Sub-Saturating Dose

AAV capsids targeting human transferrin receptor pathways are being explored for improved CNS delivery, but understanding their cell-type tropism at low dose is essential for therapeutic development. Irvin T. Garza and colleagues from UT Southwestern Medical Center examined CapX, a second-generation human transferrin-targeted AAV capsid, using spatial and single-cell readouts.

The study used multiplexed capsid validation, RNA in situ hybridization, MERFISH, biodistribution analysis, and single-cell spatial mapping to compare CapX with other capsids such as AAV9 and PHP.eB. The poster emphasized that multiplexing can reduce variability, cost, and time when comparing capsids in vivo.

CapX demonstrated CNS distribution at a sub-saturating dose, with analysis across multiple brain and peripheral tissues. The study used MERFISH and spatial analysis to assess cellular tropism and vector biodistribution at high resolution. This work highlights the growing importance of single-cell and spatial technologies for understanding not just where AAV goes, but which cell types it reaches.

Identification of a Novel Porcine AAV Capsid Supporting Efficient Retinal Gene Delivery

Retinal gene therapy has benefited greatly from AAV vectors, but improved capsids are still needed to enhance transduction efficiency, reduce dose, and expand targetable retinal cell populations. Emanuela Pone and colleagues from the University of Naples Federico II explored porcine tissues as a source of novel AAV capsids for retinal delivery.

The team screened porcine blood and liver samples using full-capsid PCR amplification and sequencing. Novel capsid sequences were used to produce rAAV-CMV-EGFP vectors, which were then evaluated for retinal transduction after subretinal and intravitreal delivery in mouse and pig retina models.

The study identified 27 novel capsid variants, with AAVpo2.99 emerging as a promising retinal gene delivery vector. The poster reported efficient mouse and pig retina transduction after subretinal delivery, and AAVpo2.99-HITI achieved targeted photoreceptor integration with improved retinal function and structure in a disease model. This work suggests that non-human mammalian tissues can serve as valuable reservoirs for discovering capsids with retinal gene therapy potential.

Receptor-Guided AAV Capsid Discovery Enables Robust Cross-Species Intravitreal Photoreceptor Targeting and Therapeutic Benefit

Inherited retinal diseases are well suited for AAV gene therapy, but current clinical approaches often depend on subretinal injection, which provides localized delivery but is invasive and can limit retinal coverage. Jiang-Hui Wang, Mengtian Cui, and colleagues from the University of Melbourne, Centre for Eye Research Australia, and UMass Chan Medical School addressed this challenge by developing a receptor-guided capsid discovery strategy for intravitreal photoreceptor targeting.

The team began by examining receptor distribution across multiple species and found that heparan sulfate is enriched at the inner limiting membrane but weakly expressed on inner retinal cells and photoreceptors across mouse, rat, rabbit, dog, pig, rhesus macaque, and human retinas. In contrast, N-acetyl-glucosamine moieties were highly expressed on photoreceptor segments and conserved across species. Based on this receptor map, the researchers designed a dual-affinity discovery strategy that selected for retinal penetration with reduced HS affinity, followed by enrichment for enhanced N-GlcNAc engagement. This led to the identification of AAV2.IvtB.

After intravitreal administration in mice, AAV2.IvtB increased retinal GFP transcripts 6.5-fold compared with AAV2.7m8 and showed higher retinal vector genome abundance. Immunohistochemistry showed 4.1-fold greater photoreceptor transduction and 2.0-fold greater Müller glia transduction versus AAV2.7m8. The performance translated to pigs, where AAV2.IvtB produced a 2.0-fold increase in photoreceptor transduction and a 1.7-fold increase in Müller glia transduction. Therapeutically, AAV2.IvtB delivering anti-VEGF KH902 reduced lesion numbers in a laser-induced CNV model and achieved approximately 2.5-fold higher KH902 transcript levels than AAV2.7m8. In the rd10 model, AAV2.IvtB encoding Pde6b partially preserved outer nuclear layer thickness. Mechanistically, the study suggested that improved IVT performance was driven primarily by enhanced N-GlcNAc-dependent engagement rather than reduced HS binding alone. This work establishes a receptor-guided, non-animal screening framework for cross-species IVT photoreceptor delivery.

Advancing Capsid Engineering and Epigenetic Transcriptional Regulation Toward Decoding the Role of AAV2 VP1u

Most AAV capsid engineering efforts focus on changing external capsid surfaces to alter receptor binding or tissue tropism, but internal and less-characterized capsid regions may also shape vector performance. Mengtian Cui, Jiang-Hui Wang, Xiangru Huo, and colleagues from UMass Chan Medical School, the University of Melbourne, and Centre for Eye Research Australia investigated the VP1 unique region of AAV2, a region implicated in intracellular trafficking and transgene expression but still incompletely understood.

The researchers built a VP1u-focused mutagenesis library using error-prone PCR, generating approximately 4.9 × 10⁵ unique variants. The library was packaged into AAV and subjected to iterative in vivo selection after intravitreal injection in mice. Retinal transcripts were collected four weeks after injection, converted to cDNA, repackaged, and re-administered for a second round of IVT selection. Enriched VP1u variants were identified by NGS and individually validated in mouse retina. Cross-species activity was assessed in pig retina, and mechanistic studies included luciferase reporter assays, CUT&RUN, and RNA-seq.

NGS identified six VP1u candidates enriched above wild-type AAV2 VP1u. The top-performing variant, VP1u.N, achieved approximately 6-fold higher transduction than parental AAV2 and approximately 3-fold higher transduction than AAV2.7m8, including an additional 2-fold enhancement in photoreceptor transduction after IVT delivery. VP1u.N also produced broadly enhanced EGFP biodistribution in pig retina, supporting cross-species relevance. Mechanistically, AAV2 VP1u increased reporter activity in HEK293 and ARPE-19 cells, while VP1u.N further enhanced reporter expression. CUT&RUN showed that VP1u.N binding overlapped RNA polymerase II-associated genomic regions, and RNA-seq showed upregulation of multiple genes relative to AAV2. Together, these findings suggest that VP1u is not merely a structural or trafficking-associated element, but may also contribute to transcriptional activation and vector performance. This study positions VP1u as a tractable target for mechanism-guided capsid engineering.

Leveraging Artificial Intelligence to Design AAV Mutant Capsids Optimized for Antibody Evasion

Pre-existing neutralizing antibodies remain a major barrier to AAV gene therapy, excluding some patients from treatment and complicating redosing strategies. Doron K. Chan and colleagues from Voyager Therapeutics used artificial intelligence to design AAV mutant capsids with improved antibody evasion while maintaining production and transduction fitness.

The team built machine learning models using large mutant datasets and screened ML-based libraries for production fitness, brain transduction, and neutralizing antibody evasion. The workflow involved multiple modeling rounds, including recurrent neural networks and transformer-based approaches, followed by library screening in vivo and in vitro.

The poster showed that ML-based libraries could identify AAV9-derived mutants with improved antibody evasion and retained transduction potential. This is an important direction for the field because immune evasion must be balanced with manufacturability, potency, and biodistribution. AI-guided capsid redesign may help create vector candidates suitable for broader patient populations with pre-existing anti-AAV immunity.

ACE-502: An AAV5-Derived Capsid Engineered by In Vivo Directed Evolution for Robust CNS Transduction and Liver Detargeting

Systemic CNS delivery requires capsids that can achieve meaningful brain transduction while limiting peripheral exposure, especially liver transduction. Hye-Ju Kim and colleagues from Amyloid Solution described ACE-502, an AAV5-derived capsid engineered through in vivo directed evolution to improve CNS targeting and reduce liver exposure.

The team performed sequential in vivo screening for brain targeting and peripheral detargeting, including mouse and non-human primate evaluation. Candidate capsids were assessed by mRNA-based in vivo evaluation, biodistribution analysis, protein expression confirmation, and head-to-head comparison with benchmark CNS capsids.

ACE-502 showed enhanced CNS targeting and reduced peripheral exposure in NHPs. The poster reported broad brain enrichment across multiple regions, higher brain expression compared with reference capsids, and lower liver signal. These findings suggest that directed evolution can generate capsids with improved CNS-to-liver selectivity, a critical feature for safer systemic neurological gene therapy.

Identification of High-Performing Ocular AAV Capsids Through Directed Engineering Across Intravitreal and Suprachoroidal Delivery Routes

Ocular AAV delivery must overcome multiple anatomical and biological barriers, including the inner limiting membrane, retinal barriers, and route-dependent differences between intravitreal and suprachoroidal administration. Celeste Stephany and colleagues from Capsida Biotherapeutics focused on identifying engineered ocular capsids with improved performance across different routes of administration.

Capsida used directed capsid engineering across AAV2 and AAV8 backgrounds, screening libraries through intravitreal and suprachoroidal delivery routes. Candidate capsids were selected based on RNA and DNA enrichment, biodistribution, transgene expression, and target tissue performance. The poster compared engineered candidates with published and clinical benchmark capsids.

Several lead ocular AAV candidates showed improved RNA expression and biodistribution relative to conventional controls. The poster highlighted candidates for retinal indications, including diabetic retinopathy, geographic atrophy, retinitis pigmentosa, and Leber congenital amaurosis. The work demonstrates that route-specific screening can identify capsids better suited for the practical realities of ocular gene therapy delivery.

Cross-Species ATHENA-I Screening Identifies AAV-ShD Capsids with Preferential Gene Delivery to Resting-State Human and NHP T Cells

Efficient delivery to resting T cells remains a major challenge for in vivo immune-cell engineering. Danmeng Zhang and colleagues from Virginia Commonwealth University described a cross-species ATHENA-I screening strategy to identify AAV capsids capable of preferentially transducing resting-state human and non-human primate T cells.

The team used a barcoded AAV library screening approach across human and NHP CD4+ T cells. Candidate capsids were evaluated for transduction efficiency, cell-type specificity, capsid design, promoter selection, and compatibility with downstream CAR-T engineering applications.

AAV-ShD emerged as a top-performing capsid, outperforming AAV6 in human resting T-cell transduction. The poster also showed preferential transduction of PBMCs and enrichment in CD3+ and CD4+ populations. This work is highly relevant to in vivo CAR-T and immune-cell gene therapy, where selective delivery to target immune populations could reduce the need for ex vivo manufacturing.

Ex Vivo Selection of Tumor-Targeting AAV Vectors by Perfusion of Cancer Patient Tissue

Tumor-targeted AAV delivery remains difficult because cancer tissues are heterogeneous and because capsids selected in standard cell lines may not perform well in patient-derived tissues. M. Rothe and colleagues from ENDomics Lab, University Medical Center Hamburg-Eppendorf, presented an ex vivo perfusion-based approach to select tumor-targeting AAV vectors directly in human cancer patient tissue.

The team generated peptide-display AAV libraries based on AAV9 and AAV2 backbones and perfused them through resected human colon and kidney cancer tissues. After multiple rounds of selection, viral DNA was isolated, amplified, and analyzed by NGS to identify enriched peptide variants and serotype distributions.

The poster showed enrichment of specific peptide variants during selection and changes in library composition across rounds. The approach provides a patient-tissue-based discovery system that could help identify capsids with tumor-selective properties. This strategy is especially important for oncology applications, where ex vivo human tissue screening may improve relevance compared with conventional in vitro models.

Engineered AAV Capsid Achieves Robust Transduction in Non-Human Primate Central Nervous System After Low-Dose Systemic Administration

Systemic CNS gene therapy requires capsids that can cross or interact with the blood-brain barrier and achieve broad neuronal transduction at clinically feasible doses. Kyle Chamberlain and colleagues from Affinia Therapeutics evaluated ATC-134, an engineered AAV capsid designed for CNS delivery after low-dose systemic administration.

The study used adult NHPs dosed intravenously with ATC-134-H2B-HA at 3E13 vg/kg. Vector genome biodistribution, RNA expression, and immunohistochemistry were used to assess transduction across multiple CNS regions, including motor cortex, caudate, putamen, hippocampus, thalamus, temporal lobe, pons, substantia nigra, occipital cortex, cerebellum, and spinal cord.

ATC-134 showed robust and widespread neuronal transduction across multiple CNS regions, with reported near-complete neuronal staining in several areas and increased RNA expression compared with AAV9. The poster also suggested lower off-target liver distribution relative to AAV9 in NHPs. These data support the potential of engineered BBB-penetrant AAV capsids for systemic neurological gene therapy.

Synthetic Divergent AAV Clades Open New Routes for NAb-Resistant, Cross-Species Capsid Design

Pre-existing neutralizing antibodies, species-dependent tropism, liver exposure, and insufficient tissue selectivity remain major barriers for systemic AAV gene therapy. Anusha Sairavi and colleagues from Oregon Health & Science University presented an unconventional ancestral sequence reconstruction strategy designed to generate synthetic AAV capsids that are highly divergent from naturally occurring serotypes while preserving useful transduction properties.

Instead of reconstructing ancestral sequences from wild-type AAVs, the team inferred a virtual ancestral backbone from phenotype-verified DNA-shuffled capsids, including AAV-DJ, LK03, and KP1. Ambiguous amino acid positions were diversified to create a large library of approximately 524,000 variants. This library was then subjected to iterative in vivo and in vitro selection across mice, human cell systems, and nonhuman primates, followed by SMRT sequencing, barcode validation, and neutralization testing against pooled human IVIG and anti-AAV9 antibodies.

The resulting synthetic “AAVε” clade demonstrated several notable properties. Multiple variants showed systemic transduction, enhanced CNS tropism, renal tropism, pancreatic tropism, and high resistance to pooled human IVIG and anti-AAV9 antibodies. Some variants showed particularly distinct tissue behaviors, including brain enrichment, podocyte transduction, pancreatic β cell selectivity, and liver detargeting. The study also showed that selected AAVε capsids could be further modified with BBB-penetrant peptides to strengthen cell-type-specific delivery. Overall, this work suggests that synthetic ancestral capsid design can open new AAV sequence space beyond natural serotypes and may provide a powerful route for developing cross-species, NAb-resistant vectors for translational gene therapy.

RGD-Free Myocyte-Tropic AAV Capsids May Help Decouple Muscle Delivery from Liver Burden

Systemic delivery to cardiac and skeletal muscle is a major goal for many neuromuscular and cardiovascular gene therapy programs, but high liver exposure remains a persistent safety concern. Lei Zhao and colleagues from Taxell Therapeutics Co. Ltd. presented a directed evolution strategy in nonhuman primates to identify a heart- and muscle-tropic AAV capsid with liver detargeting, without relying on the RGD motif commonly associated with some muscle-tropic engineered capsids.

The team used capsid DNA shuffling, natural AAV library diversity, and NHP-based selection to improve translational relevance. After two rounds of screening in nonhuman primates, they identified a lead capsid, ID-31. Unlike MyoAAV-like approaches that depend on RGD-mediated integrin interactions, ID-31 is an RGD-free chimeric capsid. Mechanistic studies suggested that its tropism may involve a distinct interaction pattern with glycans and AAVR, supported by CRISPR screening and Cryo-EM analysis of the ID-31:AAVR complex.

ID-31 showed approximately 10-fold higher mRNA expression in heart and 20-fold higher expression in skeletal muscle compared with AAV9 after intravenous administration in NHPs. At the same time, liver expression was reduced to approximately 0.05-fold of AAV9 levels. The capsid also demonstrated cross-species activity in mice and NHPs and supported delivery of SERCA2a or FGF21 in a transverse aortic constriction model. In translational head-to-head comparison, ID-31-mediated SERCA2a delivery increased cardiac transgene expression while reducing liver expression compared with a clinical-stage AAV1-SERCA2a strategy. These findings suggest that strong myocyte tropism can be achieved through non-RGD-dependent mechanisms and may help improve the therapeutic window for systemic heart and muscle gene therapy.

AAV9-Derived Muscle-Tropic Capsids Can Be Engineered for Human NAb Evasion

Immune exclusion remains one of the most important limitations for AAV gene therapy, especially for systemic programs where many patients may carry pre-existing neutralizing antibodies. Damien Maura and colleagues from Voyager Therapeutics Inc. presented a strategy to engineer AAV9-derived muscle-tropic capsids that better evade human neutralizing antibodies while retaining muscle-directed activity.

The team used Voyager’s TRACER platform and iterative machine-learning-assisted directed evolution to identify capsid surface mutations that reduce recognition by pooled human IVIG and pre-existing anti-AAV antibodies. The workflow focused on generating “stealth” mutations that could be layered onto an AAV9-derived muscle-tropic capsid without eliminating its desired tissue tropism.

The study showed that selected surface mutations improved human IVIG evasion and reduced neutralizing antibody seroprevalence. Importantly, these mutations could be transferred into an AAV9-derived muscle capsid background while preserving robust muscle expression. The poster also compared engineered stealth capsids with naturally NAb-evasive variants, suggesting that rationally layered modifications can improve immune evasion without fully sacrificing tropism or functional performance. This work is highly relevant for systemic muscle-directed programs, where immune evasion could expand patient eligibility, reduce exclusion due to pre-existing antibodies, and potentially support future redosing strategies.

AAVεB2 Demonstrates Selective Pancreatic β Cell Tropism with Liver Detargeting

Pancreatic islet gene delivery is an important but difficult goal for diabetes-related gene therapy. Existing AAV serotypes often show strong liver transduction after systemic administration and may preferentially target pancreatic acinar or ductal cells rather than endocrine β cells. Anusha Sairavi and colleagues from Oregon Health & Science University used the synthetic AAVε clade platform to identify a liver-detargeted capsid with selective pancreatic β cell tropism.

The AAVε capsid library was generated using ancestral reconstruction-guided design based on DNA-shuffled capsids, followed by in vitro and in vivo selection and SMRT sequencing. Enriched variants were validated using barcode-based approaches and single-capsid studies. For validation, AAVεB2, AAV9, and AAV-KP1 were packaged with a CAG-tdTomato reporter and administered systemically to mice. Major organs, including pancreas and liver, were evaluated by immunofluorescence and cell-type-specific marker analysis.

AAVεB2 showed striking liver detargeting compared with AAV9 and AAV-KP1. In the pancreas, AAV9 produced broad pancreatic transduction dominated by acinar cells, while AAV-KP1 showed limited islet expression. In contrast, AAVεB2 selectively transduced pancreatic islet cells, with co-staining confirming expression in insulin-positive β cells even under a ubiquitous CAG promoter. No detectable acinar cell transduction was observed. This finding suggests that capsid engineering alone can achieve highly selective endocrine pancreas delivery without relying exclusively on transcriptional targeting or invasive local administration. AAVεB2 therefore represents a promising platform for pancreatic β cell-directed gene therapy, including future diabetes applications.

  

Cross-Poster Insights: What These Findings Suggest About the Future of AAV Capsid Engineering

Across these ASGCT posters and presentations, a major trend is the movement away from generic AAV serotype selection and toward disease-, tissue-, route-, immune-, and species-specific capsid design. Kidney, CNS, retina, pancreas, muscle, heart, tumor tissue, and T-cell applications each require different engineering priorities, and these studies show that no single capsid is likely to solve every delivery challenge. Instead, the field is moving toward fit-for-purpose vector design, where capsids are engineered for a defined biological target, delivery route, therapeutic context, and safety requirement.

A second key insight is that translational screening systems are becoming more human- and species-relevant. Human kidney explants, primary human astrocytes, iPSC-derived neurons, patient tumor tissues, human and NHP T cells, pig retina, and nonhuman primate screening models are increasingly being used earlier in the capsid discovery process. The addition of NHP-directed muscle capsid evolution and cross-species AAVε validation further emphasizes that rodent-only screening is no longer sufficient for many translational programs. These approaches may reduce the risk of advancing capsids that perform well in mice but fail to translate to human biology.

A third theme is the convergence of computational design, machine learning, and experimental selection. AI-guided library design, ML-based prediction of capsid viability, closed-loop screening platforms, and immune-evasion modeling are becoming practical tools for identifying higher-performing capsids while reducing the size and cost of experimental screening campaigns. The synthetic AAVε clade work further expands this concept by showing that computational phylogeny-guided ancestral reconstruction can generate entirely new capsid sequence space, rather than simply optimizing known natural serotypes.

A fourth important direction is receptor- and mechanism-guided capsid design. The AAV2.IvtB study shows that mapping conserved receptor distributions can guide non-animal capsid discovery for intravitreal photoreceptor targeting, while the VP1u.N study suggests that previously underappreciated capsid domains may influence transcriptional output and not only cell entry or trafficking. The ID-31 muscle-tropic capsid also highlights how non-RGD-dependent mechanisms, glycan interactions, and AAVR engagement can create strong myocyte tropism with liver detargeting. Together, these findings point to a more mechanistic future for AAV engineering, where capsid design is informed by receptor biology, intracellular processing, transcriptional regulation, immune recognition, pharmacokinetics, and tissue architecture.

Another emerging priority is immune evasion as a core capsid engineering goal. Several studies directly addressed pre-existing neutralizing antibodies, including AI-designed AAV mutant capsids, stealth-engineered AAV9-derived muscle capsids, and synthetic AAVε variants with resistance to pooled human IVIG and anti-AAV9 antibodies. These findings suggest that next-generation capsids will increasingly need to combine tissue tropism with reduced antibody recognition, especially for systemic delivery programs where pre-existing immunity can limit patient eligibility and complicate redosing.

Finally, capsid engineering is increasingly being evaluated through multidimensional success criteria. A promising capsid must not only transduce a target tissue, but also maintain manufacturability, avoid excessive liver exposure, minimize immunogenicity, support clinically relevant routes of administration, perform across disease-relevant models, and ideally show cross-species activity. The field is moving toward a more integrated model in which potency, specificity, safety, immune compatibility, and translational relevance are engineered together from the beginning.

 

How PackGene Can Support AAV Capsid Engineering and Translational Development

PackGene provides integrated AAV capabilities to support capsid discovery, screening, validation, and translational development, from research-grade vector production to process development and GMP manufacturing. With experience across AAV plasmid design, AAV packaging, capsid library production and screening, analytical characterization, genome integrity analysis, and scalable manufacturing, PackGene can help researchers move promising capsid candidates from early discovery into robust preclinical and clinical development workflows.

Through flexible AAV production services, advanced analytical support, and end-to-end gene therapy development capabilities, PackGene helps accelerate the evaluation of novel capsids for CNS, ocular, renal, metabolic, oncology, and immune-cell applications. As the field moves toward more precise and tissue-targeted gene delivery, integrated vector design, production, and characterization platforms will be critical for translating next-generation AAV capsids into therapeutic candidates.

Author: Jin Qiu