Abstract
Adeno-associated virus (AAV) is widely regarded as a leading vector for gene therapy, underscored by clinical successes such as Luxturna and Zolgensma. However, efficient gene delivery to hard-to-transduce tissues—including the retina, deep skeletal muscle, and the central nervous system—remains a significant challenge, limited by structural barriers, preexisting immunity, and dose-dependent toxicities. This review systematically outlines recent advances in overcoming these delivery bottlenecks. We delve into four key strategic areas: (i) capsid engineering (e.g., rational design, directed evolution, and computational approaches) to enhance tropism and evade immune detection; (ii) innovative delivery routes (local, systemic, and physical/chemical methods) to improve vector bioavailability; (iii) modulation of intracellular trafficking to boost nuclear delivery; and (iv) immunomodulatory strategies to mitigate both innate and adaptive immune responses. We further highlight translational progress in neuromuscular and retinal diseases and discuss persistent challenges. Looking forward, we envision that the convergence of next-generation capsids, smart vector systems, and integrated delivery platforms will be critical to expand the therapeutic landscape of AAVs from rare monogenic disorders to broader clinical applications.
Keywords
INTRODUCTION
Overview of adeno-associated virus vectors
Adeno-associated virus (AAV), classified within Parvoviridae, is a nonenveloped, ∼26 nm virus that circulates naturally in multiple vertebrate species, notably humans and nonhuman primates (NHPs). Traditionally, 13 AAV serotypes have been described, and more than 100 naturally occurring variants have been identified. In addition to the 13 classical AAV serotypes, several NHP-derived serotypes (e.g., AAVrh74) are widely used in clinical development and are therefore included in this review. The characteristics and clinical applications of these serotypes are summarized in Table 1. 20 The AAV genome consists of an ∼4.7-kb linear single-stranded DNA molecule with four established open reading frames encoding Rep, Cap, the assembly-activating protein, and the membrane-associated accessory protein, the latter identified only in recent years.21,22 At both ends, the genome bears inverted terminal repeats (ITRs) that form hairpin structures required for viral genome replication and persistence (Fig. 1).
Overview of canonical adeno-associated virus serotypes and selected noncanonical adeno-associated virus variants with translational relevance
This table summarizes the 13 canonical adeno-associated virus (AAV) serotypes (AAV1–AAV13) that have been historically identified from human or nonhuman primate isolates, highlighting their typical origin, primary attachment factors, known coreceptors, tissue tropism, and representative clinical application areas. Information regarding receptor usage reflects current consensus and may remain incomplete for certain serotypes.
In addition, AAVrh74 is included as a noncanonical, rhesus macaque-derived AAV variant due to its widespread use in translational and clinical studies, particularly for muscle-targeted gene therapy, despite not being classified as a distinct canonical serotype.
Food and Drug Administration-approved AAV-based gene therapies are listed where applicable to illustrate clinically validated capsid platforms rather than to denote additional natural serotypes.
Key original references correspond to the first reports describing serotype isolation, receptor usage, or capsid biology; review articles have been removed from the table. This table is adapted from Li and Samulski (2020) with updates incorporating recent advances and regulatory approvals.
AAT: Alpha-1 antitrypsin; AIP: Acute intermittent porphyria.
Schematic of the wild-type adeno-associated virus (AAV) genome structure. The linear single-stranded DNA genome of AAV, approximately 4.7 kb in length, is flanked by ITRs that form T-shaped hairpin structures, essential for replication, packaging, and genome integration. The genome contains two main open reading frames: the rep gene, encoding the nonstructural replication proteins (Rep78, Rep68, Rep52, Rep40), and the cap gene, encoding the structural capsid proteins (VP1, VP2, VP3). In addition, the genome encodes the AAP, which facilitates virion assembly, and the MAAP, which is involved in cellular egress. AAP, assembly-activating protein; ITR, inverted terminal repeat; MAAP, membrane-associated accessory protein.
AAV is notable for its strict dependence on helper functions supplied by coinfecting viruses, most commonly adenovirus or herpes simplex virus, to support productive replication and progeny virion formation. 23 In the presence of these helper viruses, AAV undergoes efficient genome replication, transcription, and capsid assembly. In contrast, in the absence of helper functions, AAV typically establishes a latent state in host cells, persisting either as nonreplicative extrachromosomal episomes or, less frequently, through integration into the host genome. This dual life cycle enables durable transgene persistence and distinguishes AAV from many other viral vectors, thereby contributing to its widespread use in both basic research and gene therapy applications.
AAV is nonpathogenic and exhibits low cytotoxicity, contributing to a favorable safety profile for gene therapy applications. After cellular transduction, recombinant AAV genomes are maintained primarily in an episomal form, existing as monomeric or concatemeric DNA species, with minimal integration into the host genome, 24 thereby supporting long-term and stable transgene expression and reducing the need for repeated dosing. Furthermore, the identification of classical AAV serotypes together with more than 100 naturally occurring variants has revealed distinct tissue tropisms, providing a versatile tool kit for targeted gene delivery. 20 For example, some serotypes preferentially transduce hepatocytes, whereas others display enhanced affinity for neurons, thereby enabling precise therapeutic targeting of specific tissues.
These unique properties establish AAV as an ideal vector for gene therapy. Compared with earlier viral vectors such as retroviruses, AAV reduces tumorigenic risks while overcoming issues of transient expression, thereby ensuring improved safety and durability. 25 Moreover, the serotypic diversity of AAV and its distinct tissue tropisms greatly broaden the scope of applications, enabling precise targeting of genetic disorders across different organs and systems. 25 Among current viral gene delivery systems, AAV vectors are extensively used in clinical trials, especially for monogenic diseases and neurological indications.
The clinical utility of AAV vectors is now well established. Luxturna (voretigene neparvovec-rzyl), the first AAV-based therapy approved by the U.S. Food and Drug Administration, delivers a functional RPE65 gene to retinal pigment epithelial cells in patients with RPE65-associated inherited retinal dystrophy, leading to sustained improvements in functional vision.6,26 Onasemnogene abeparvovec (Zolgensma) has shown pronounced clinical benefit in pediatric patients with spinal muscular atrophy (SMA) through AAV-mediated delivery of a functional SMN1 gene, leading to marked gains in survival and motor milestone achievement. 27 These pioneering clinical achievements underscore the therapeutic potential of AAV vectors and lay the groundwork for expanding AAV-based gene therapies to a wider spectrum of genetic diseases.
Challenges in hard-to-transduce tissues
Although AAV vectors offer significant advantages in gene delivery, their transduction efficiency is limited in certain tissues, particularly the central nervous system (CNS), which is hindered by the blood–brain barrier (BBB).21,28 The BBB, formed by brain microvascular endothelial cells with pericytic and astrocytic support, is defined by tight endothelial junctions that establish a highly selective physical barrier and limit the entry of macromolecules and pathogens into the brain. 29 Similarly, the retina represents another hard-to-transduce tissue. Its multilayered cellular organization, together with the extracellular matrix, markedly limits AAV penetration and impedes efficient targeting of photoreceptors and the retinal pigment epithelium (RPE). 30 Skeletal muscle, although receptive to certain AAV serotypes, also presents challenges in the context of deep or large muscle groups, where uniform and efficient transduction is difficult to achieve. 31 To systematically address these issues, Table 2 summarizes the core biological and physical barriers that restrict AAV delivery and their specific characteristics.
Key physical, biological, and dose-limiting barriers to efficient adeno-associated virus delivery in hard-to-transduce tissues
This table summarizes major physical, biological, and dose-related barriers that limit efficient AAV delivery across hard-to-transduce tissues, including the central nervous system, retina, skeletal muscle, kidney, and lung.
Necessity of efficient delivery
The physical, biological, and dose-related barriers encountered by AAV vectors in hard-to-transduce tissues not only restrict transduction efficiency but also limit the clinical applicability and therapeutic value of gene therapy. Overcoming these barriers is essential to broaden the indications of AAV therapies, enhance clinical outcomes, and lower both risk and cost.
Currently, most clinical AAV-based therapies are directed toward relatively permissive tissues, such as the liver and superficial muscle, whereas effective gene delivery to refractory targets—including the CNS, retina, and deep skeletal muscle—remains a major unmet challenge. 21 Neurodegenerative diseases, including Alzheimer’s and Parkinson’s diseases, are driven by dysfunction of discrete neuronal populations. However, efficient brain delivery of AAV vectors remains severely restricted by the BBB, particularly after systemic administration, which hampers the clinical translation of vectors validated in small-animal models. 28 To overcome this barrier, strategies, including BBB-penetrant capsid engineering and transient BBB disruption using focused ultrasound (FUS), have emerged as promising approaches to enhance AAV delivery to deep brain regions. FUS-mediated BBB opening has been shown to be feasible and reversible, and its safety has been demonstrated in NHP studies, thereby creating new opportunities for gene delivery to otherwise inaccessible brain nuclei.39,40 Similarly, for inherited retinal disorders such as retinitis pigmentosa, retinal barriers limit efficient transduction of photoreceptor cells by current AAV delivery methods. 32 In the eye, serotype optimization and local delivery strategies, including subretinal injection, can facilitate efficient transduction of target cells in the outer retina, such as retinal pigment epithelial cells, and have demonstrated potential for vision restoration in patients with inherited retinal blindness. 41
In addition, barriers in these tissues not only restrict the overall efficiency of AAV but also lead to uneven vector distribution and incomplete therapeutic responses. In muscle diseases, limited penetration and uneven distribution of systemically delivered AAV in large and deep muscle groups can leave subsets of diseased cells untreated, resulting in incomplete therapeutic benefit. 42 In CNS disorders, restricted BBB permeability may confine AAV expression to superficial brain regions while limiting delivery to deeper pathological sites. 43 By contrast, overcoming such barriers could enable uniform and efficient transduction across target tissues, ensuring stable and continuous expression of therapeutic genes at the site of pathology. Furthermore, engineering AAV capsids to bind receptors with higher specificity may improve targeting of diseased cells and reduce transduction of nontarget tissues. Likewise, mitigating immune barriers may prevent antibody neutralization and immune clearance of vectors, thereby prolonging expression, reducing repeat dosing, and enhancing durable efficacy.
To compensate for poor transduction efficiency, current clinical practice often relies on high AAV doses, which introduces additional challenges. High-dose administration may elicit strong immune responses and severe toxicities, including hepatotoxicity and neuroinflammation, which have raised major safety concerns in systemic AAV gene therapy. 38 Moreover, large-scale vector production is technically complex and costly, with some approved therapies associated with extremely high treatment costs. Effective strategies to overcome delivery barriers would enable therapeutic efficacy at lower doses, thereby reducing immune-related risks, minimizing vector requirements, and lowering production costs.
In summary, overcoming the barriers to AAV delivery in hard-to-transduce tissues represents a pivotal step toward expanding gene therapy from “niche indications” to “broad clinical applications.” Such breakthroughs would not only provide new therapeutic opportunities for refractory diseases but also enhance the clinical and societal value of gene therapy by improving efficacy, safety, and accessibility.
Purpose and scope of this review
Overcoming the barriers to AAV delivery in hard-to-transduce tissues holds critical significance. Such advances could extend AAV gene therapy to refractory disorders involving the CNS, the retina, and additional challenging tissues; enhance treatment efficacy by enabling uniform and efficient transduction, thereby achieving precise and durable therapeutic effects; and reduce the required vector dose, thereby potentially lowering immune-related side effects and manufacturing costs. Taken together, these innovations will accelerate the move from limited indications to broad clinical application of AAV gene therapy, thereby realizing greater clinical and societal benefit.
This review systematically surveys recent advances in AAV delivery for refractory tissues; showcases representative strategies in capsid engineering, dosing/administration, intracellular trafficking modulation, and immune regulation; and evaluates their underlying mechanisms, advantages, and constraints. We further discuss translational progress, current challenges, and future perspectives, with the aim of providing constructive insights to facilitate the efficient clinical translation and broad application of AAV-based therapies.
AAV VECTOR ENGINEERING STRATEGIES
Capsid engineering
The capsid serves as the critical structural component mediating interactions between AAV vectors and target cells. The capsid structure is crucial for determining AAV’s transduction efficiency, tissue specificity, and ability to overcome biological barriers. As a result, capsid engineering has become a key approach to optimizing AAV delivery, especially for hard-to-transduce tissues. Several approaches have been developed for capsid modification, which can be broadly categorized into five representative strategies (Fig. 2).
Representative strategies for AAV capsid engineering.
Rational design
Rational design generally refers to the purposeful modification of capsid proteins based on detailed structural and functional knowledge of the AAV capsid. 44 By identifying key regions involved in receptor binding, researchers can introduce site-directed mutations or insert specific amino acid sequences to alter capsid surface properties and enhance receptor affinity in hard-to-transduce tissues 45 (Fig. 2A). For example, in the case of AAV9, cell-penetrating peptides (CPPs) have been integrated into the capsid surface to generate the AAV.CPP.16 variant. Owing to the intrinsic ability of CPPs to traverse biological membranes, this modification enables AAV to partially overcome physical barriers such as the BBB. In murine and cynomolgus monkey models, intravenous AAV.CPP.16 traversed the BBB and enabled broad, efficient gene delivery across the CNS. 46 Another notable example is the work by Crosson et al., 47 who introduced point mutations (V387R, W502H, E530K, L583R) into the AAV2 capsid to alter capsid surface hydrophobicity. While these mutations reduced AAV2 transduction efficiency in retinal cells, additional point mutations (R585A, R588A) at the heparan sulfate proteoglycan binding site markedly improved retinal transduction in both mice and rhesus macaques. 47 These findings highlight the precision of rational design, which allows targeted enhancement of AAV tropism for specific tissues.
Nevertheless, rational design is constrained by the requirement for detailed knowledge of AAV–receptor interactions, which remains incompletely understood for several hard-to-transduce tissues. Moreover, capsid modifications may inadvertently compromise viral stability or packaging efficiency, necessitating careful optimization to balance efficacy and manufacturability.
Directed evolution
By creating vast libraries of AAV capsid variants, directed evolution emulates natural selection, enabling the identification of variants with improved traits, such as enhanced transduction efficiency in tissues that are typically resistant to gene delivery. In contrast to rational design, it does not require prior knowledge of capsid structure or receptor interactions, 48 thereby enabling the discovery of novel capsid properties that are otherwise difficult to achieve. Directed evolution strategies can be broadly divided into the following three categories: in vitro screening, in vivo screening, and species-specific evolution (Fig. 2B).
In vitro screening typically uses methods such as DNA or peptide display libraries. For example, a Cre recombinase-dependent transduction system can be used in which Cre expression is linked to AAV transduction efficiency. By applying this system in reporter cell lines, AAV variants with enhanced transduction capacity for specific cell types can be selectively enriched.
In vivo screening is performed directly within target tissues, such as the brain or retina of mice and NHPs. Libraries of AAV variants are delivered into animals, and after sufficient expression time, capsids capable of successfully transducing the target tissue are isolated and amplified. Notable examples include AAV-PHP.eB, which demonstrates strong BBB penetration and robust CNS transduction in specific mouse strains, and AAV.CAP-B10, which exhibits efficient transduction of retinal tissue. 49
Species-specific evolution addresses the translational gap between animal models and humans. Because capsid performance can vary substantially across species, variants that are highly effective in rodents may not replicate these effects in humans. To overcome this limitation, libraries are screened in NHPs or other species with closer genetic and physiological similarity to humans. Variants derived from this strategy show greater translational potential, as their favorable transduction profiles in NHPs are more likely to be preserved in clinical applications. 50
Chimeric capsids
Chimeric capsids are artificially engineered AAV capsids generated by combining advantageous structural regions from different AAV serotypes. Each serotype displays unique strengths and limitations in terms of tissue tropism, transduction efficiency, and capacity to overcome biological barriers. By constructing hybrid capsids, the beneficial properties of multiple serotypes can be integrated to produce vectors with superior overall performance. 51
For example, certain AAV serotypes exhibit strong tissue specificity but relatively low transduction efficiency. AAV1, for instance, demonstrates robust tropism for muscle tissue but limited efficiency. Conversely, other serotypes, such as AAV2, show efficient cellular uptake and broad transduction profiles, yet lack distinct tissue specificity. For example, combining capsid regions associated with muscle tropism from one serotype with regions conferring high transduction efficiency from another serotype represents a conceptual strategy for generating chimeric variants with improved performance (Fig. 2C). This approach significantly broadens the potential for optimizing AAV vectors, particularly for applications involving hard-to-transduce tissues. By integrating complementary features from different serotypes, chimeric capsids provide a promising strategy for overcoming delivery barriers and improving the therapeutic performance of AAV-based gene therapies.
Pseudotyping
Pseudotyping describes a strategy in which an AAV genome flanked by ITRs from one serotype is packaged into the capsid of a different serotype, thereby conferring distinct biological properties to the resulting vector. 52 The core concept is to pair ITRs and capsids from distinct serotypes to combine their complementary strengths. For example, AAV2-derived ITRs are widely used due to their favorable replication and packaging characteristics, whereas the native AAV2 capsid exhibits limited transduction efficiency in many hard-to-transduce tissues. In contrast, AAV9 capsids display strong BBB penetration and efficient transduction of muscle tissue. By packaging an AAV2 ITR-flanked genome into an AAV9 capsid, the resulting pseudotyped AAV2/9 vector retains the advantageous ITR-associated properties of AAV2 while acquiring the superior tropism of AAV9, enabling efficient gene delivery to both the CNS and skeletal muscle 53 (Fig. 2D). This strategy is relatively straightforward to implement and allows rapid exploitation of well-characterized capsids, making pseudotyping a widely used approach when improved tissue tropism and delivery efficiency are required.
Computer-aided design
Computational design provides a powerful and precise strategy for generating AAV capsid variants with enhanced transduction efficiency (Fig. 2E). By leveraging sequence information and evolutionary relationships among different AAV serotypes, researchers can systematically identify functionally relevant sites and rationally design improved capsids. For instance, comparative analysis of DNA sequences and phylogenetic mapping reveals genetic variations that underlie the functional diversity of AAV capsids. These insights enable the construction of homologous capsid libraries enriched for variants with potential performance advantages. 22
Another application integrates machine learning with experimental screening. Extensive libraries of AAV capsid variants are first constructed and iteratively selected in vivo. High-throughput sequencing of input and output populations quantifies enrichment of specific variants. These datasets are then used to train predictive algorithms, such as convolutional neural networks, which in turn guide the prediction and synthesis of novel capsid proteins with desirable properties. The resulting variants are subsequently validated through in vitro assays. This integrative pipeline markedly accelerates the discovery of optimized AAV capsids, bridging computational prediction with experimental validation and significantly advancing the design of advanced AAV vectors. 54
Progress and representative capsid variants
In recent years, major advances have yielded next-generation AAV capsids tailored for hard-to-transduce tissues, including the CNS, retina, and skeletal muscle. These variants improve transduction efficiency, sharpen tissue specificity, and facilitate barrier crossing, ultimately expanding AAV’s therapeutic scope.
For the CNS, AAV-PHP.eB represents a landmark variant. Systemic administration of AAV-PHP.eB enables BBB passage and results in extensive transduction of both neurons and glial cells in mice, demonstrating substantially enhanced CNS delivery compared with conventional AAV serotypes.55,56 Moreover, optimized CNS-targeting capsids have shown promising performance in NHP models, underscoring their translational potential for neurological gene therapy. 56
In the retina, a prototypical hard-to-transduce tissue, the engineered capsid AAV2.7m8 (7m8) has emerged as a key variant for intravitreal gene delivery. Developed through in vivo directed evolution by David Schaffer’s group, AAV2.7m8 overcomes the inner limiting membrane barrier that restricts conventional AAV serotypes, enabling efficient transduction of photoreceptors and RPE cells following intravitreal injection. 57 Owing to these properties, AAV2.7m8 has been extensively used in preclinical models of inherited retinal diseases and has achieved efficient transduction of the outer retina in multiple large-animal species. 58
For skeletal muscle, while AAV1 and AAV6 already exhibit inherent muscle tropism, further capsid engineering has yielded variants with substantially enhanced efficiency. These optimized capsids not only enable more effective transduction of deep muscle fibers but also improve tissue specificity by minimizing off-target expression in nonmuscle tissues. 42
Together, these advances highlight the power of capsid engineering in tailoring AAV vectors to overcome tissue-specific barriers, laying the groundwork for more effective treatments of CNS, retinal, and muscle-related diseases. Representative engineered AAV variants illustrating these strategies are summarized in Supplementary Table S1.59–62
Promoter optimization
Promoters are critical regulatory elements that control transgene expression by dictating its tissue and cell specificity as well as expression intensity. The performance of AAV vectors depends heavily on promoter choice and optimization, with a pronounced impact in hard-to-transduce tissues. By fine-tuning promoter design, it is possible to achieve higher expression levels in target cells, minimize off-target activity, and reduce immune responses, thereby improving both delivery efficiency and therapeutic outcomes.
Promoter optimization broadly comprises (i) tissue-/cell-specific promoters that restrict expression to target populations and (ii) potent or minimal promoters that increase transcription while remaining within packaging limits. Promoter optimization is central to improving AAV performance, particularly in hard-to-transduce tissues. Tissue-specific promoters (e.g., Syn1 for neurons, RPE65 for RPE, cardiac cTnT) confine transgene expression to intended cell populations and reduce off-target expression in non-target organs and professional antigen-presenting cells (APCs).63–65 Ubiquitous promoters combined with organ-selective micro RNA (miRNA) target sites (e.g., CAG plus miR-122 target sites to detarget the liver) further limit off-target expression while preserving efficacy in target tissues.66–68 Compact promoters help meet packaging constraints (notably for self-complementary AAV [scAAV]), whereas inducible systems (e.g., Tet-On/Off) have been implemented to enable temporal control when safety or disease biology requires dosing windows. 69 These designs entail trade-offs among strength, specificity, and size, but collectively provide a practical toolbox to increase on-target expression and mitigate immune liabilities without escalating dose. In practice, tissue-specific designs (e.g., Syn1, RPE65, cTnT) reduce off-target expression and antigen presentation, whereas compact or inducible promoters (e.g., miniCMV, hSyn; Tet-On/Off) balance expression strength with payload and temporal control.
More recently, machine learning-based approaches have been introduced to systematically optimize promoter selection and regulatory sequence design for tissue-specific transgene expression.70,71 By leveraging large-scale datasets encompassing promoter architectures, epigenomic features, and cell-type-resolved transcriptomic profiles, these models enable the identification of sequence motifs and regulatory combinations associated with restricted and robust expression in defined tissues. Such data-driven strategies have facilitated the rational design of synthetic and hybrid promoters with improved specificity and reduced off-target activity, offering a complementary pathway to traditional empirical screening. 71 As these approaches continue to evolve, machine learning-guided promoter design is expected to further enhance the precision and safety of AAV-mediated gene expression, particularly in hard-to-transduce tissues.
Genome optimization
The genomic architecture of AAV vectors critically shapes transduction efficiency, expression onset, and the durability of transgene activity. Optimizing the vector genome has proven effective for boosting performance, particularly in hard-to-transduce tissues where robust and sustained expression is required for therapeutic benefit.
Genome optimization typically uses scAAV to accelerate expression by bypassing second-strand DNA synthesis, 72 cDNA optimization (e.g., codon usage and transcript stability), and genome minimization to enhance packaging and delivery efficiency. Beyond these core measures, genome design can be further refined, in practice, by tuning 5′/3′ UTR elements (e.g., Kozak context, stability/translation motifs), incorporating functional introns to improve transcript processing, and selecting an appropriate polyadenylation signal (e.g., bGH polyA) to support messenger RNA maturation and persistence.53,73 Inclusion of post-transcriptional regulatory elements (such as WPRE or its engineered variants) may increase steady-state expression but should be used with explicit disclosure and regulatory consideration. 74 Where payload is limiting—particularly for scAAV—compact architectures and removal of nonessential sequences help maintain vector integrity and manufacturing robustness. 75 Finally, insulators/CTS can reduce enhancer cross talk and improve expression consistency, while barcoding/UMI schemes are valuable for discovery-phase screening but are typically omitted from clinical constructs. 20 Collectively, these genome-level strategies, together with advances in promoter engineering and data-driven design approaches, complement capsid optimization to achieve faster onset, higher peak output, and more durable expression in hard-to-transduce tissues.
DELIVERY TECHNOLOGIES AND METHODS
Local delivery/regional delivery
Local or regional delivery refers to the direct or targeted administration of AAV vectors into specific tissues or anatomical regions. This strategy enhances local transduction efficiency while minimizing systemic exposure. Six major delivery approaches are commonly used.
Intracerebral injection
Direct administration of AAV into the parenchyma, ventricular system, or cerebrospinal fluid (predominantly the subarachnoid space) constitutes a core approach for CNS gene therapy. Convection-enhanced delivery applies a continuous pressure gradient to achieve a broader, more uniform intraparenchymal vector distribution, thereby overcoming the limited diffusion of conventional local injections. For example, in glioblastoma studies, convection-enhanced delivery-mediated administration of therapeutic AAV vectors has achieved higher local concentrations and enhanced antitumor effects. 76
Intravitreal/subretinal injection
These routes are widely used in ocular gene therapy. Intravitreal injection is relatively straightforward, enabling AAV distribution within the vitreous cavity and efficient transduction of inner retinal cells. By contrast, subretinal injection delivers the vector directly adjacent to retinal pigment epithelial and photoreceptor cells, achieving superior transduction efficiency. For instance, subretinal AAV2–RPE65 therapy has achieved clinically meaningful improvements in vision in Leber congenital amaurosis. 41
Intrathecal/intraspinal injection
The method is suitable for spinal cord targets and specific disorders of the peripheral nervous system. By introducing AAV into the subarachnoid space or spinal canal, vectors can diffuse via the cerebrospinal fluid and transduce the spinal cord and nerve root tissues. In SMA research, intrathecal delivery of AAV9 has achieved robust expression in motor neurons, improving patient outcomes. 77
Intramuscular injection
Intramuscular delivery is suitable for localized muscle disorders and can also exploit skeletal muscle as a platform for systemic secretion of therapeutic proteins. In Duchenne muscular dystrophy (DMD), intramuscular AAV administration enables dystrophin expression within injected muscles; however, immune responses directed against dystrophin remain a major challenge. 78
Regional intravascular perfusion
Regional intravascular perfusion and systemic vascular delivery approaches, such as limb-isolation perfusion or intravenous administration, involve transient or targeted exposure of muscle vasculature to AAV vectors. This method enhances transduction efficiency in muscle tissues of the isolated limb while reducing systemic exposure. It holds the potential for treating localized muscle pathologies, ensuring concentrated delivery to diseased muscles. 79
Intra-airway delivery
Intra-airway delivery, including nebulization, intratracheal instillation, and intranasal administration, is primarily applied to pulmonary diseases. This delivery route enables localized exposure of AAV vectors to airway and alveolar epithelia and has been extensively investigated to overcome physical barriers such as the mucus layer and tight epithelial junctions. Previous studies have highlighted both the feasibility and the challenges of airway-directed AAV delivery, emphasizing the critical roles of vector design and delivery conditions in achieving efficient pulmonary transduction. 80 More recently, advances in capsid engineering have substantially improved the performance of intra-airway AAV delivery. Engineered AAV variants with enhanced respiratory tropism have demonstrated robust transduction of airway and lung epithelial cells following intranasal or intra-airway administration. Notably, the AAV.CPP.16 capsid derived from AAV9, originally developed to enhance systemic tissue penetration, has also demonstrated efficient and cross-species airway transduction following intranasal administration in rodent and NHP models. 46
Advantages and limitations
Local/regional delivery achieves high vector concentrations at target sites, reduces systemic toxicity and immune responses, and improves therapeutic precision. However, its invasive nature may lead to complications such as pain and infection. Moreover, the approach is inherently limited to localized or anatomically accessible diseases, and heterogeneity in vector diffusion may affect the consistency of therapeutic outcomes.
Systemic delivery strategies
Systemic delivery strategies aim to achieve noninvasive administration and widespread vector distribution throughout the body, making them suitable for treating systemic or multiorgan diseases.
Intravenous injection
Intravenous injection is the most common systemic delivery approach. It can use strategies such as high-dose administration, vector shielding/stealth modification, and receptor-mediated transcytosis.
High-dose strategy
Transduction efficiency can be improved by administering high doses of AAV vectors intravenously. Despite its technical simplicity, the approach entails substantial disadvantages, including elevated production costs, enhanced off-target transduction—particularly in the liver—and increased toxicity. Notably, administration of high vector doses is associated with exaggerated innate and adaptive immune activation, which can intensify inflammatory responses and ultimately lead to dose-limiting toxicities. For instance, abnormal liver function has been reported in clinical studies following high-dose intravenous AAV infusion, highlighting dose-dependent safety concerns in systemic delivery. 38
Vector shielding/stealth modification
Chemical modification strategies, including polyethylene glycol (PEG)ylation, polymer encapsulation, and liposome coating, can reduce immune recognition and clearance of AAV vectors, thereby extending their systemic circulation time. 81 PEGylation, for example, forms a hydrophilic layer on the vector surface, masking immunogenic epitopes and enhancing delivery efficiency to target tissues.
Receptor-mediated transcytosis
Capsid engineering can introduce binding sites for endothelial cell receptors that mediate transcytosis, such as the human transferrin receptor 1. 82 This modification enables vectors to cross vascular barriers via receptor-mediated endocytosis and exocytosis, thereby improving delivery to otherwise inaccessible tissues such as the CNS.
In practice, however, the effectiveness of these approaches is often limited by suboptimal transduction efficiency and the emergence of off-target or immune-mediated adverse effects.
Overcoming preexisting neutralizing antibodies
Preexisting neutralizing antibodies (NAbs) markedly diminish the efficiency of systemic AAV delivery. An estimated 30–60% of individuals harbor NAbs, typically arising from natural childhood exposure to AAV. These antibodies neutralize therapeutic vectors, thereby limiting transduction and narrowing eligibility for gene therapy. 83 In response, multiple strategies have been developed.
Plasma exchange/immunoadsorption
Removing circulating antibodies before AAV infusion can transiently reduce neutralization, although this approach is complex and carries risks such as infection and electrolyte imbalance. 84
Using rare serotypes or evolved escape capsids
Rare AAV serotypes are associated with lower seroprevalence, while escape capsids generated via directed evolution can evade antibody recognition, both of which improve effective vector delivery. 85
Empty capsid competition
Coadministration of large amounts of empty AAV capsids can act as decoys, binding preexisting NAbs and preserving the activity of therapeutic vectors. However, this strategy increases the total capsid burden, which may further stimulate immune responses and raise safety concerns, particularly at clinically relevant doses.
Physical/chemical methods to enhance delivery
Beyond biological and immunological strategies, physical and chemical methods can transiently enhance AAV penetration across barriers or improve local availability.
FUS combined with microbubbles
FUS induces microbubble oscillation, which mechanically and reversibly opens barriers such as the BBB or blood–retinal barrier. This method enhances local AAV entry while maintaining spatiotemporal precision and minimal invasiveness. 86
Hypertonic mannitol solution
Intravenous infusion of hypertonic mannitol dehydrates and shrinks vascular endothelial cells, temporarily opening the BBB and allowing AAV entry into brain tissue. Despite its long clinical history, this approach remains controversial due to limited specificity, potential neurotoxicity, and poor controllability. 87
Sustained-release systems
Biomaterial-based delivery platforms have recently emerged as an advanced strategy to improve local AAV availability and transduction efficiency by providing controlled vector retention and release at target sites. Hydrogels and related biomaterials can encapsulate AAV vectors, enabling prolonged local exposure and improved spatial confinement of AAV vectors. 88
Recent studies demonstrate that such systems not only enhance transduction efficiency but also improve safety profiles. 89 For example, fibrin-based AAV-containing hydrogels have been shown to reduce injection-associated adverse effects while achieving superior retinal transduction compared with conventional bolus delivery methods. In parallel, alginate hydrogel formulations have enabled efficient and spatially confined delivery of vascular-targeted AAV vectors, highlighting the potential of biomaterials to improve tissue retention and targeting specificity.88,89
Beyond sustained release alone, next-generation biomaterial platforms are being developed to achieve spatiotemporally programmable viral delivery. Collagen-based scaffolds that integrate sustained and stimuli-responsive release enable sequential viral delivery, providing enhanced control over the kinetics of gene transfer. 90 Collectively, these advances position biomaterial-assisted AAV delivery as a promising and increasingly sophisticated approach for overcoming local delivery barriers in hard-to-transduce tissues.
In summary, systemic delivery approaches, particularly intravenous injection, provide a noninvasive route for vector distribution but are constrained by toxicity risks, off-target effects, and preexisting NAbs. Countermeasures such as rare serotypes, escape capsids, and empty capsid competition have shown promise in overcoming these barriers. Meanwhile, physical and chemical enhancement methods—including FUS, hypertonic mannitol, and sustained-release systems—offer innovative ways to transiently modulate biological barriers or extend vector persistence. Each approach has unique advantages but also notable limitations in safety, specificity, or scalability. Therefore, the rational integration of these strategies, combined with advances in vector engineering, will be essential to achieving safe, efficient, and clinically translatable systemic AAV delivery to hard-to-transduce tissues.
ENHANCING AAV VECTOR INTRACELLULAR TRAFFICKING EFFICIENCY
After entering target cells, AAV vectors undergo a series of intracellular processes, including endosomal trafficking, endosomal escape, cytoplasmic transport, and nuclear entry. Among these steps, inefficient endosomal escape and nonproductive intracellular routing represent major bottlenecks that critically limit the overall transduction efficiency, particularly in hard-to-transduce tissues. 26 Although AAV vectors are favored for their safety profile and sustained transgene expression, only a small fraction of internalized particles ultimately reach the nucleus and establish productive infection. Accordingly, improving intracellular trafficking efficiency has emerged as a key strategy to enhance AAV performance without escalating vector dose.
Enhancing endosomal escape
Following receptor-mediated endocytosis, AAV particles are sequestered within endosomal compartments. Failure to escape from endosomes in a timely manner leads to trafficking toward lysosomal degradation, resulting in substantial loss of vector genomes.31,91 Therefore, facilitating efficient endosomal escape is a central determinant of successful AAV transduction.
Capsid engineering with membrane-disrupting peptides
One major approach to promote endosomal escape involves capsid engineering to enhance membrane-disrupting activity. Efficient endosomal escape of AAV relies on the VP1 unique region (VP1u), which contains a functional phospholipase A2 domain. Upon endosomal acidification, structural rearrangements of the capsid promote VP1u exposure, thereby facilitating endosomal membrane destabilization and cytoplasmic release. Recent structural and functional studies have provided direct evidence that this acid-triggered exposure of VP1u is a key mechanistic event enabling productive endosomal escape and downstream transduction. 92
Building on this principle, additional membrane-active sequences have been engineered into AAV capsids to further enhance endosomal escape. Viral fusion peptides or synthetic CPPs can respond to the acidic endosomal environment and increase membrane permeability, thereby improving cytoplasmic release of the vector. While such modifications can significantly enhance transduction efficiency, careful optimization is required to avoid compromising capsid stability, packaging efficiency, or infectivity. 31
Combined application of endosomal escape agents
An alternative strategy involves coadministration of AAV vectors with agents that modulate endosomal conditions. Chloroquine, for example, inhibits endosomal acidification and has been shown to enhance AAV transduction in vitro by promoting endosomal membrane destabilization. However, its clinical utility is limited by cytotoxicity and lack of specificity. To overcome these limitations, recent studies have explored biocompatible polymers and nanomaterials capable of inducing endosomal disruption in a more controlled and less toxic manner, offering a complementary approach to capsid engineering. 93
Regulating intracellular trafficking pathways
Following endosomal escape, AAV particles are transported through the cytoplasm toward the perinuclear region and subsequently access the nucleus to initiate transgene expression. Increasing evidence suggests that this process does not occur via passive diffusion but instead depends on regulated intracellular trafficking pathways (Fig. 3).
Intracellular trafficking pathway of AAV vectors from endosomal escape to nuclear entry. After receptor-mediated endocytosis, AAV particles are internalized into endosomal compartments, where endosomal acidification induces conformational rearrangements of the capsid and externalization of the VP1 unique region (VP1u), exposing its phospholipase A2 activity to facilitate endosomal membrane disruption and cytoplasmic escape. Following release into the cytoplasm, productive AAV infection relies on regulated intracellular trafficking rather than passive diffusion. Emerging evidence indicates that AAV particles undergo coordinated retrograde transport through Golgi-associated compartments, including the trans-Golgi network, which serves as a critical intracellular checkpoint influencing capsid processing and downstream nuclear delivery. Cytoskeletal transport along microtubules, modulated by post-translational modifications such as detyrosination, promotes perinuclear accumulation and efficient nuclear targeting. Ultimately, AAV genomes enter the nucleus through the nuclear pore complex, where uncoating and subsequent transcription initiate transgene expression.
Recent studies suggest that productive AAV infection involves coordinated retrograde trafficking through Golgi-associated compartments, including the trans-Golgi network, which functions as a critical intracellular checkpoint influencing capsid processing and nuclear entry. Disruption of these pathways has been shown to impair transduction efficiency, highlighting their functional importance in the AAV life cycle.22,91
Cytoskeletal transport further plays a central role in directing AAV particles toward the nucleus. AAV capsids interact with microtubule networks and associated motor proteins to achieve efficient perinuclear accumulation. Experimental studies demonstrate that perturbation of microtubule dynamics or motor protein function significantly reduces nuclear delivery and transgene expression. More recent work has further shown that microtubule posttranslational modifications, such as detyrosination, regulate motor protein engagement and significantly modulate AAV nuclear entry. 94
Following perinuclear accumulation, AAV genomes enter the nucleus through the nuclear pore complex, where uncoating and second-strand synthesis enable transcription. 26 Collectively, optimization of both endosomal escape and intracellular trafficking pathways provides a powerful means to enhance AAV transduction efficiency, particularly in hard-to-transduce tissues, without increasing systemic vector exposure.
IMMUNE MODULATION STRATEGIES
When delivered systemically, AAV vectors are capable of eliciting innate and adaptive immune responses that promote inflammation, clearance of transduced cells, and ultimately compromise gene transfer efficiency.36,84,95,96 Therefore, rational immune-modulation strategies are essential to sustain therapeutic transgene expression and improve safety profiles.
Reducing innate immune responses
As the earliest defensive barrier, innate immunity is promptly triggered upon administration of AAV vectors. These responses are primarily triggered by immune recognition of vector-associated components, including capsid proteins and vector DNA. Current strategies to attenuate innate immunity focus on vector purification, capsid and genome engineering, and pharmacological modulation.
Removal of empty capsids
During AAV production, a considerable proportion of capsids lacking vector genomes (“empty particles”) are inevitably generated. Although nontherapeutic, these particles can trigger innate immune responses primarily through complement activation and increased antigenic load, with secondary effects on toll-like receptor (TLR)-mediated sensing. Refining purification methods such as density-gradient ultracentrifugation, ion-exchange, or affinity chromatography significantly decreases empty-capsid load and reduces immunogenic burden. 97 High-purity AAV preparations with minimal empty-capsid ratios induce weaker cytokine responses and achieve superior transduction efficiency.98,99 Recent industrial-scale advances integrating affinity chromatography with multiangle light scattering enable quantitative control of full-to-empty capsid ratios, achieving <2% empty particles and establishing a new benchmark for clinical-grade AAV manufacturing. 100
Using immune-silent capsids and genome engineering
Capsid engineering allows the generation of AAV variants with reduced immunogenicity, thereby decreasing their detection by innate immune sensors. One common strategy involves mutating or masking immunodominant regions on the capsid surface to limit interactions with pattern-recognition receptors, including TLRs. 81 For instance, the AAV.rh10 serotype naturally exhibits low innate immunogenicity. After further engineering modification, its interaction with TLRs on dendritic cells is markedly attenuated, leading to reduced activation of immune-related organs such as the spleen following intravenous administration in mice. In addition to capsid-level engineering, genome-level CpG nucleotide depletion has emerged as an effective strategy to limit TLR-mediated innate immune activation. Unmethylated CpG sequences present in the AAV vector genome can be sensed by endosomal TLR9, leading to activation of downstream proinflammatory signaling pathways. Rational removal or minimization of CpG motifs within the transgene cassette can significantly reduce TLR9 activation, dampen innate immune responses, and improve transgene expression durability without compromising vector potency. 101 Importantly, recent studies have expanded the innate immune paradigm beyond endosomal TLR9 sensing. After endosomal escape, cytoplasmic AAV genomes may be sensed by cyclic GMP–AMP synthase (cGAS), thereby activating the STING signaling axis and inducing type I interferon production. This cGAS–STING axis has emerged as a central cytosolic immune checkpoint that critically limits hepatic and systemic AAV transduction. Pharmacological inhibition of STING signaling (e.g., H-151) or genetic disruption of cGAS–STING components significantly enhances transgene expression while attenuating inflammatory cytokine production, highlighting this pathway as a next-generation target for innate immune modulation.36,102
Pharmacological pretreatment (e.g., glucocorticoids)
Glucocorticoids such as prednisone and dexamethasone mitigate AAV-induced inflammation by suppressing NF-κB activation and reducing cytokine secretion from macrophages and dendritic cells. Clinical studies in hemophilia B have demonstrated that prophylactic prednisone, initiated before and continued after intravenous AAV infusion, significantly reduces serum transaminase elevation without compromising vector transduction efficiency.103,104 Nevertheless, extended corticosteroid exposure may increase susceptibility to infection and metabolic complications, underscoring the need for individualized dosing and duration. Collectively, modern innate immune modulation strategies are transitioning from broad immunosuppression toward pathway-specific interventions that target both endosomal (TLR9) and cytosolic (cGAS–STING) sensing mechanisms, thereby preserving efficacy while minimizing systemic toxicity. 36
In summary, enhanced vector purification, immune-silent capsid and genome design, and pharmacological modulation such as glucocorticoid pretreatment effectively attenuate AAV-triggered innate immunity. These strategies alleviate early inflammatory responses and preserve transduction efficiency, thereby improving gene therapy safety and efficacy. However, innate immunity represents only the initial barrier—adaptive immune responses, including capsid-specific T cells and NAbs, remain major obstacles to long-term expression and readministration.
Reducing adaptive immune responses
Following the early innate phase, adaptive immune responses mediated by T and B cells constitute a principal barrier. NAbs and capsid-specific cytotoxic T lymphocytes directed at capsid proteins or transgene products can rapidly eliminate circulating vectors and kill successfully transduced cells, ultimately compromising durability and efficacy.84,105 Therefore, strategies aimed at mitigating adaptive immunity focus on minimizing immunogen exposure and finely regulating immune cell activation.
Using tissue-specific promoters and miRNA regulation: One strategy to reduce adaptive immunity is to confine transgene expression to target cells using tissue-specific promoters, thereby minimizing antigen presentation by APCs. In the CNS, for example, the Syn1 promoter drives expression predominantly in neurons while sparing immune organs such as the liver and spleen, resulting in reduced T cell priming relative to ubiquitous promoters under comparable dosing conditions. In mice, systemic delivery of AAV-PHP.eB under Syn1 control elicited approximately one-third the CD8+ T cell response observed with a CMV promoter.56,106 In addition, incorporation of microRNA target sites (e.g., miR-142-3p) into the 3′ UTR of the transgene cassette actively suppresses expression in hematopoietic APCs, further reducing antigen presentation and adaptive immune activation.107,108 This layered regulatory design is increasingly regarded as a standard component of immune-aware AAV vector engineering.
Optimizing transgenes to minimize novel epitopes: Therapeutic transgenes containing nonself sequences may generate novel epitopes recognized by cytotoxic T lymphocytes. Codon optimization and site-directed mutagenesis can eliminate such epitopes while maintaining protein structure. 109 For example, synonymous mutations introduced into predicted HLA-A*02:01 epitopes of the coagulation factor IX (FIX) gene reduced antigen recognition by ≈70% in vitro. In HLA-humanized mice, the modified FIX gene evoked significantly weaker CD8+ T cell responses, supporting the feasibility of rational epitope silencing.
Transient immunosuppression: Short-term treatment with calcineurin inhibitors, including cyclosporine A and tacrolimus, attenuates T cell activation through inhibition of the calcineurin–NFAT signaling axis, thereby extending transgene expression in preclinical studies.110,111 However, broad immunosuppression carries inherent safety risks. Recent advances emphasize a paradigm shift from nonspecific immunosuppression toward active immune tolerance induction. Synthetic vaccine particles encapsulating rapamycin selectively promote regulatory T cell expansion and have demonstrated long-term tolerance to AAV capsids in preclinical and translational models. 112 In parallel, IgG-degrading enzymes such as IdeS enable transient cleavage of preexisting NAbs immediately before vector administration, thereby rescuing AAV transduction in seropositive subjects and enabling potential redosing. 113 These strategies represent a new generation of adaptive immune management that aligns with clinical safety and durability requirements.
Together, advances in innate and adaptive immune modulation reveal a clear evolution in AAV gene therapy—from broad immunosuppression toward precision immune engineering that integrates vector design, genome optimization, and pathway-specific immune interventions. Such strategies are expected to be essential for expanding the therapeutic window, enabling readministration, and supporting durable expression in hard-to-transduce tissues.
CLINICAL TRANSLATION AND CHALLENGES
Representative clinical progress
Clinical translation of AAV-mediated gene therapies has advanced rapidly, with several landmark approvals that demonstrate durable efficacy yet expose new biological and regulatory hurdles. Currently approved AAV-based gene therapies comprise onasemnogene abeparvovec-xioi (Zolgensma; AAV9) for SMA, voretigene neparvovec-rzyl (Luxturna; AAV2) for biallelic RPE65-associated retinal dystrophy, and delandistrogene moxeparvovec (Elevidys; AAVrh74) for DMD.
Spinal muscular atrophy
Long-term follow-up of Zolgensma recipients continues to show sustained survival and long-term maintenance of clinically meaningful motor function for up to 8 years after a single intravenous infusion.114,115 Notably, presymptomatic infants demonstrate the greatest benefit, while outcomes are less robust in patients with established motor neuron loss. 116 Despite transformative efficacy, hepatotoxicity and transient thrombocytopenia remain dose-limiting toxicities, and redosing remains infeasible because of NAbs. 117
Leber congenital amaurosis
The AAV2-based therapy Luxturna demonstrates long-term restoration of RPE65 function, with most recipients achieving clinically meaningful improvements in full-field light sensitivity and navigation performance for 3–4 years post-treatment. 118 However, structural retinal degeneration and the invasiveness of subretinal delivery limit efficacy in advanced disease stages and necessitate refinement toward less invasive intravitreal or suprachoroidal approaches.
Duchenne muscular dystrophy
Micro-dystrophin replacement using AAVrh74 (Elevidys) has achieved consistent biomarker expression but limited functional improvement in randomized phase 3 trials.119,120 Reports of fatal hepatic failure cases in 2025 underscored the critical need for improved dose safety and immune monitoring. 121
Main challenges
Species-specific translation gaps
Efficacious dosing and tropism observed in rodents or NHPs frequently fail to replicate in humans because of interspecies differences in receptor expression, capsid trafficking, and innate immune sensing. 121
Immunogenicity and vector-related toxicity
Innate sensing of vector-associated nucleic acids—via pathways such as TLR9 and cytosolic DNA sensors—as well as complement activation, together with adaptive immune responses against the AAV capsid and transgene, continues to represent major clinical safety liabilities. High systemic doses provoke hepatic inflammation, thrombocytopenia, and in rare cases fulminant liver failure. 120 Strategies such as CpG-depleted genomes, immune-silent capsids, and transient immunosuppression (e.g., corticosteroids or IL-6 blockade) are being refined to expand the therapeutic window.
Dose and tissue-distribution limitations
For CNS or muscular diseases, achieving broad and uniform transgene expression at clinically tolerable doses remains a key barrier. Large-transgene disorders requiring delivery of >5 kb coding sequences are further constrained by AAV’s limited packaging capacity. 45
Manufacturing, scalability, and access
Industrial-scale AAV production still faces bottlenecks in capsid assembly efficiency, full-to-empty ratio control, and downstream purification yield. 122 Emerging continuous bioprocessing and AI-driven analytics show promise in enhancing consistency and reducing cost but are not yet standardized across manufacturers.
Ethical and regulatory considerations
Postmarketing surveillance of Zolgensma and Elevidys illustrates the need for global pharmacovigilance frameworks and harmonized long-term registries to detect delayed toxicities or insertional events. Ethical discourse now centers on equitable access, early-life intervention thresholds, and retreatment eligibility. 117
In summary, while AAV gene therapy has achieved landmark clinical milestones with treatments such as Zolgensma, Luxturna, and Elevidys, its broader translation remains constrained by dose-related toxicity, immune barriers, manufacturing limitations, and species-specific variability. Continued progress will rely on the convergence of next-generation capsid engineering, precise immune modulation, scalable bioprocessing, and improved preclinical modeling to ensure both safety and accessibility. The field now stands at a defining juncture—transitioning from rare-disease success toward sustainable, mainstream genetic medicine through innovation grounded in scientific rigor and clinical responsibility.
CONCLUSION AND FUTURE OUTLOOK
A primary obstacle impeding the widespread medical adoption of AAV gene therapies is the difficulty of achieving effective gene transfer in refractory tissues. Physical barriers such as the BBB and inner limiting membrane, together with cellular entry constraints, intracellular trafficking bottlenecks, and host immune responses, collectively restrict vector performance even for the most advanced serotypes. Overcoming these multilayered barriers therefore requires integrated solutions rather than reliance on any single optimization strategy.
The development of novel vectors with superior tissue tropism and barrier-crossing capabilities has been greatly accelerated by modern engineering techniques, such as rational design, directed evolution, and computational modeling. However, capsid engineering alone is rarely sufficient. Complementary optimization at the genome level, including promoter selection, regulatory element design, and genome architecture refinement, plays an equally critical role in determining expression specificity, durability, and safety. Importantly, intracellular trafficking has emerged as a decisive yet often underappreciated determinant of transduction efficiency. Strategies that enhance endosomal escape, avoid degradative pathways, and promote efficient nuclear transport can markedly amplify the functional gains achieved by capsid and genome engineering.
A significant barrier to the clinical translation of systemic, high-dose AAV therapies is the host immune response. The activation of innate immunity against vector components, combined with adaptive responses involving NAbs and cytotoxic T cells, not only reduces transduction efficacy but also poses safety risks and prevents vector readministration. Genome-level strategies such as CpG depletion, improved vector purification to minimize empty capsids, and rational immunomodulation are therefore essential components of next-generation AAV delivery paradigms. Such considerations are increasingly shaping clinical trial design and regulatory expectations.
Looking forward, the future of AAV gene therapy for hard-to-transduce tissues is likely to be defined by combinatorial and data-driven approaches. The integration of capsid engineering with optimized regulatory architectures, intracellular trafficking enhancement, and precisely controlled delivery routes holds promise for achieving robust efficacy at clinically acceptable doses. Advances in machine learning, high-throughput screening, and NHP models are expected to accelerate the discovery of translatable vectors and regulatory elements. Ultimately, the convergence of smart vector design, mechanistic insight, and rational clinical strategies will be critical for extending AAV-based gene therapy from rare indications to a broader spectrum of neurological, ocular, muscular, and systemic diseases.
AUTHORS’ CONTRIBUTIONS
W.G.: Funding acquisition, conceptualization, and writing—review and editing. Y.W.: Funding acquisition, writing—review and editing. X.N.: Writing—original draft, investigation, writing—review and editing, and visualization. J.S.: Investigation and writing—review and editing. Y.J.: Investigation and writing—review and editing. W.L.: Investigation and writing—review and editing. Y.C.: Funding acquisition and resources. Y.L.: Funding acquisition and writing—review and editing. B.X.: Funding acquisition and writing—review and editing. L.Y.: Funding acquisition and writing—review and editing.
Footnotes
ACKNOWLEDGMENTS
The authors are deeply grateful to Associate Professor W.G. from Hunan University of Technology for his invaluable guidance and intellectual support throughout the conceptualization and writing of this review. They also extend their sincere thanks to Associate Research Fellow Y.W. from the Innovation Academy for Precision Measurement Science and Technology, Chinese Academy of Sciences, for his critical insights and constructive feedback on the article. The authors also extend their appreciation to the colleagues who provided valuable suggestions during the preparation of this article.
AUTHOR DISCLOSURE
The authors declare that they have no competing financial interests or personal relationships that could have appeared to influence the work reported in this article.
FUNDING INFORMATION
This work was supported by the National Natural Science Foundation of China (Grant No. 82301383), the Natural Science Foundation of Hunan Province (Grant No. 2023JJ30211), the Scientific Research Project of Hunan Provincial Administration of Traditional Chinese Medicine (Grant No. D2024009), and the General Project of Hunan Provincial Education Department (Grant No. 24C0273).
Supplemental Material
References
Supplementary Material
Please find the following supplemental material available below.
For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.
For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.
