Advances in AAV Gene Therapy (2021–2026): Engineering, Manufacturing, Safety, and the Ocular/Neuromuscular Treatment Landscape

1) Context: what changed most in the last five years

From 2021–2026, adeno-associated virus (AAV) gene therapy advanced along two parallel tracks: (1) higher-performing vectors (better targeting, lower dose needs, and improved immune evasion) and (2) more mature clinical/regulatory practice around safety monitoring, manufacturing comparability, and long-term follow-up. The period also reflects globalization of AAV approvals, exemplified by China’s first approved AAV gene therapy for hemophilia B, signaling expanding regulatory capacity and adoption beyond the US/EU frameworks 6. While hemophilia B is outside the ocular/neuromuscular focus here, it is an important marker of platform development.

2) Vector engineering and delivery advances

2.1 Capsid engineering: improved tropism and antibody evasion

A key technical theme is moving from “natural serotypes” to engineered capsids tailored for human tissues and immune environments. A representative example is structure-guided engineering of AAV8: cryo–electron microscopy (cryo-EM) mapping of antigenic footprints enabled targeted saturation mutagenesis and iterative selection in human hepatocytes, yielding AAVhum.8 1. In relevant models, AAVhum.8 showed markedly higher hepatocyte transduction and improved evasion from anti-AAV8 antibodies, including improved performance against pooled human immunoglobulin and sera from AAV8-exposed animals 1. Although liver-directed, this work is widely influential because it operationalizes a general strategy: co-optimizing potency + tropism + immune evasion rather than treating them as tradeoffs.

Broader capsid discovery remains relevant because many naturally occurring capsids are “underemployed.” Systematic characterization of less-studied capsids such as AAV7 is presented as a route to uncover differential potency/tropism/immunogenicity profiles that could be clinically advantageous, expanding the engineering design space beyond AAV8/AAV9/AAVrh.10 3.

2.2 Receptor biology: new levers to reduce dose and toxicity

A major 2025 advance was identification of carboxypeptidase D (CPD) as an alternate AAV receptor (“AAVR2”) 11. Clade E AAVs (including AAV8) use AAVR2 for transduction via specific binding interfaces mapped by cryo-EM, and receptor engagement could be engineered into capsids lacking AAVR2 binding 11. Because many AAV toxicities correlate with high systemic doses, receptor-informed engineering is positioned as a “clinically deployable” path to dose reduction and toxicity mitigation 11.

2.3 Genome-level engineering: promoters, control circuits, and durability

AAV improvement is not only capsid-based. Reviews emphasize genome-level innovation: synthetic enhancers/promoters for tissue specificity, engineered regulatory circuits (e.g., microRNA- or light-controlled expression), and structural insights into AAV DNA affecting stability and function 5. This complements the framing of AAV vectors as bioengineered therapeutic nanoparticles, integrating capsid and genome engineering to improve efficacy at lower doses and with better specificity 4.

Mechanistic work also highlights why durability can vary. Host factors can limit transduction: the SUMOylation pathway restricts intracellular AAV accumulation and transduction, with host proteins such as DAXX implicated; inhibiting SUMOylation increased reporter transduction in a model system 10. Separately, epigenetic silencing of recombinant AAV genomes via NP220 and the HUSH complex (H3K9 methylation–associated repression) provides a mechanistic basis for transgene silencing, with strong serotype dependence (capsid choice influencing the degree of silencing) 46. Together, these findings connect “vector choice” to long-term expression stability, not just initial transduction.

2.4 Delivery route realities (retina vs muscle)

Ocular and neuromuscular programs highlight route-specific biology:

• Ocular delivery relies mainly on subretinal or intravitreal injection. Intravitreal AAV can cause dose-dependent inflammation and still triggers systemic neutralizing antibody (NAb) formation, limiting redosing 7.
• Neuromuscular programs largely use systemic intravenous (IV) delivery (for body-wide muscle) or intrathecal (IT) dosing (for spinal targets), which magnifies systemic immune/toxicity risks and makes dose minimization central to safety 1728.

3) Manufacturing and analytical quality themes with clinical implications

3.1 Empty, full, and “intermediate” capsids: potency and inflammation implications

Manufacturing inherently creates mixtures of full, empty, and intermediate (partially packaged) capsids. Empty capsids are widely treated as impurities with potential to reduce efficacy and increase immunogenicity 31. In nonhuman primate intravitreal studies, reducing total capsid load by removing empty capsids reduced ocular inflammation and improved transduction, directly linking capsid composition to clinical-relevant tolerability and performance 7.

A 2024 study adds an important nuance: intermediate capsids can be infectious in vitro and contribute to genome titer (qPCR/dPCR) yet do not contribute to functional potency in vivo/functional assays, supporting their control as impurities to protect efficacy and reduce unnecessary capsid exposure 29.

3.2 Scalable purification and comparability expectations

Scalable anion-exchange chromatography (AEX) approaches were developed/validated for multiple serotypes (AAV5/6/8/9), achieving high enrichment of genome-containing capsids and reproducibility suitable for CGMP settings, with comparability demonstrated between monolith and packed-bed formats 39. These purification choices matter clinically because capsid load (including empty/intermediate particles) influences immunogenicity and inflammation risk.

Regulatory sensitivity to process changes is illustrated in the ELEVIDYS program: clinical development used two processes with non-comparable empty capsid residual profiles, emphasizing why analytical comparability and well-defined critical quality attributes are essential when scaling manufacturing 18.

3.3 Analytics: from “research assays” to QC-ready orthogonal toolkits

Multiple analytical methods advanced toward QC practicality while improving interpretability:

• Reviews summarize orthogonal approaches for full/empty quantification and genome characterization, including sv-AUC, ddPCR, SDS-PAGE, and alkaline agarose gels to evaluate packaging heterogeneity and genome integrity 31.
• QC-friendly methods include capillary isoelectric focusing (CIEF) for rapid full/empty determination with small sample volumes 14, and SEC with dual-wavelength detection validated against AUC and cryo-EM 43.
• More advanced characterization includes multiwavelength sv-AUC for compositional deconvolution of capsid vs encapsidated DNA 41 and native mass spectrometry for rapid empty:full assessment 45.
• Vector genome integrity tools expanded: a microfluidic capillary electrophoresis approach for genome sizing revealed workflow-dependent artifacts (ssDNA annealing creating dsDNA peaks), highlighting the need for standardized sample prep and dual-reference standards 34. PCR-free sequencing (SSV-Seq 2.0) improved detection/quantification of impurities and coverage biases in GC-rich/homopolymer regions 35.
• Manufacturing can also introduce genome rearrangements via recombination; abundance-biased codon diversification (ABCD) was proposed to prevent recombination and preserve in vivo functionality 38.

4) Immunology and safety learnings (and mitigation in practice)

4.1 Pre-existing NAbs and redosing constraints

Pre-existing anti-AAV NAbs remain a major eligibility barrier and a key reason redosing is difficult after initial administration. Reviews describe mitigation concepts: serotype switching, plasmapheresis, and enzymatic antibody degradation strategies, and broader immunosuppression targeting B- and T-cell responses 5758. In trials, baseline anti-capsid antibodies frequently function as exclusion criteria (e.g., AAV8- or AAV9-specific screening in neuromuscular programs) 20.

4.2 Complement activation and thrombotic microangiopathy (TMA)

Complement has emerged as a class-level safety concern. A mechanistic review describes both classical and alternative pathway contributions and emphasizes that patient-specific, vector-related, and environmental factors converge to determine outcomes from immune priming to severe toxicities such as TMA 2. A 2024 Journal of Clinical Investigation study further supports an antibody-dependent mechanism (classical pathway) amplified by the alternative pathway, and it highlights a critical early post-infusion window (days 4–10) for intensive monitoring; an immunomodulatory regimen (rituximab + sirolimus + steroids) attenuated antibody formation and complement activation compared with steroids alone in systemic AAV9 recipients 22.

4.3 Liver injury, thrombocytopenia, DRG toxicity, myocarditis: “class signals” in systemic dosing

Product labels and trial experience show recurring systemic risks:

• Hepatotoxicity is central enough to justify boxed warnings and mandated steroid regimens (e.g., ZOLGENSMA, ITVISMA, ELEVIDYS) with structured lab monitoring and taper schedules 271728.
• Thrombocytopenia is commonly monitored early post-dose (weekly in some regimens) 172728.
• DRG (dorsal root ganglion) toxicity is a recognized nonclinical and clinical concern. A 2022 translational study supports circulating neurofilament light chain (NF-L) as a sensitive nonclinical biomarker of DRG injury, with dose-dependent early rises (days ~8–14) and better correlation in serum/plasma than CSF 60.
• Myocarditis/troponin elevations appear in neuromuscular labels and trials; ELEVIDYS labeling includes myocarditis risk and weekly troponin-I monitoring in the first month 28.

Ocular immune reactions are more localized but still meaningful: intravitreal AAV dosing can trigger measurable inflammation and systemic NAb induction, and empty capsid load contributes to inflammatory burden 7.

5) Clinical evidence and endpoints (ocular and neuromuscular)

5.1 Ocular monogenic disease programs: endpoints and mixed outcomes

RPE65 retinal dystrophy (Luxturna; voretigene neparvovec, AAV2, subretinal) remains a benchmark for functional endpoints: the Phase 3 trial showed significant improvement in multi-luminance mobility testing (MLMT) at 1 year, with many patients reaching maximal improvement at the lowest luminance and no product-related serious adverse events 13. Prescribing information reinforces dosing (1.5×10^11 vg/eye) and perioperative steroid regimens, plus warnings (endophthalmitis, retinal tears, cataract, intraocular pressure increases) 23.

Other ocular programs demonstrate both promise and failure modes:

• Choroideremia (REP1): Phase III BIIB111 failed its primary endpoint (BCVA ≥15-letter improvement) and key secondary endpoints; earlier RG6367 (SPK-7001) showed no consistent effect in later-stage patients 20.
• LCA1 (GUCY2D; ATSN-101, subretinal): high-dose cohorts showed substantial retinal sensitivity improvements (FST) and some mobility improvements with manageable inflammation 20.
• LCA10 (CEP290; EDIT-101, subretinal CRISPR): a subset achieved meaningful BCVA improvements; program paused for independent development partly due to small patient population 20.
• X-linked retinoschisis (RS1): intravitreal BIIB087 showed no clinical activity in interim follow-up, while subretinal ATSN-201 showed structural closure of schisis in most treated eyes plus microperimetry/visual acuity improvements, with dose-dependent retinal findings managed by steroids and a move to a mid-dose cohort; FDA agreed to expand ATSN-201’s study toward a pivotal path with a BLA targeted for 2028 per company update 2052.
• Retinitis pigmentosa (modifier gene therapy; OCU400, subretinal): 2-year data reported improvement/preservation vs untreated eyes with statistically significant findings, with high-dose safety events including BCVA loss in some cases 20.
• RPGR (X-linked RP) remains a major Phase III target class, with Phase III programs including botaretigene sparoparvovec (AAV5-RPGR) and AGTC-501 (AAV2tYF-RPGR) listed in Phase III development datasets 48 and supported by disease-focused reviews describing RPGR genetics and development challenges 25.
• LHON (ND4; GS010/LUMEVOQ): company-reported REFLECT Phase III follow-up described sustained visual acuity improvements through 3 years after bilateral intravitreal injection and favorable safety; a French-authorized Phase II dose-ranging study began in 2026 5149.

5.2 Neuromuscular monogenic disease programs: endpoints, durability, and safety constraints

Spinal muscular atrophy (SMA; SMN1):

• IV onasemnogene abeparvovec (ZOLGENSMA, AAV9) Phase III studies show strong survival and motor milestone gains compared with natural history, with particularly striking results in presymptomatic infants where many achieved sitting/standing/walking within expected developmental windows 20. Labeling codifies steroid prophylaxis, liver/platelet monitoring, and TMA warnings, and emphasizes that repeat dosing has not been evaluated 27.
• Intrathecal AAV9 dosing has expanded into older children (e.g., STEER), with statistically significant improvement on HFMSE vs sham, plus ongoing long-term follow-up indicating milestone maintenance and additional milestone acquisition in many patients 20. The 2025 FDA approval of ITVISMA (intrathecal scAAV9) for SMA ≥2 years includes boxed warnings and detailed monitoring guidance (liver injury, thrombocytopenia, TMA, sensory neuropathy) and shedding kinetics 17.

Duchenne muscular dystrophy (DMD; micro-dystrophin):

• ELEVIDYS (AAVrh74, IV) received accelerated approval in 2023 based on micro-dystrophin expression, with later regulatory updates describing expansion (including non-ambulatory pathways) and confirmatory trial requirements 1824. In EMBARK, the 52-week primary endpoint (NSAA change) was not met overall, while a younger subgroup showed benefit; longer-term data showed improvements vs placebo on NSAA and timed function tests, and micro-dystrophin expression was sustained over time in reported analyses 20. Labeling highlights acute liver failure risk (boxed warning), infection risk under steroids, myocarditis/troponin monitoring, immune-mediated myositis (including genotype-linked risk), and anti-AAVrh74 antibody screening thresholds 28.
• Other DMD programs show heterogeneous outcomes: Pfizer’s AAV9 microdystrophin program failed its Phase III endpoint and was paused in related development due to a fatal SAE in a separate trial, per retrieved trial summaries 20. In contrast, early/interim data from RGX-202 (AAV8) and SGT-003 (AAV9) described micro-dystrophin expression and functional/biomarker improvements with favorable short-term tolerability in small cohorts 20.

Limb-girdle muscular dystrophy (LGMD):

• SRP-9003 (beta-sarcoglycan; IV) showed meaningful beta-sarcoglycan expression and functional improvements vs natural history in early cohorts, with no new drug-related safety signals reported in the summarized interim data 20. Pipeline datasets also list Phase III and Phase II AAV programs across LGMD subtypes (e.g., Sarepta’s AAVrh74-based programs and FKRP-targeting programs) 21.

X-linked myotubular myopathy (XLMTM; MTM1, AAV8): resamirigene bilparvovec (AT132) produced ventilatory support reductions and milestone gains in surviving patients but was associated with multiple deaths from hepatic/hepatobiliary failure in both dose cohorts, underscoring the dose-limited safety envelope in systemic high-dose AAV 20.

6) Treatment landscape summary (approved + late-stage ocular and neuromuscular AAV gene therapies)

Table includes approved and late-stage (Phase II/III or Phase III) programs explicitly found in the retrieved materials; where capsid/serotype or outcomes were not provided in the tool results, the table states that limitation.

Closing synthesis: what the platform has learned (2021–2026)

Across ocular and neuromuscular diseases, the most consistent lesson is that AAV success is now bounded less by “can we deliver a gene?” and more by dose, immunology, and manufacturing-defined product consistency. Engineering advances (immune-evasive capsids, receptor-informed targeting, and promoter/genome control) are increasingly aimed at lowering required dose and improving durability 1115. In parallel, manufacturing analytics have become clinically relevant: empty/intermediate capsids and genome integrity heterogeneity are not abstract impurities—they influence inflammation, immune priming, potency interpretation, and comparability when processes change 7293139. Clinically, ocular programs demonstrate that meaningful functional endpoints (e.g., MLMT) can be achieved with manageable local risks in the eye 1323, while systemic neuromuscular programs highlight the continuing need for intensive immunosuppression and monitoring to manage hepatotoxicity, complement/TMA risk, and other class effects 2222728.