Early systemic “BBB-penetrant” rAAVs—most notably PhP.B and the higher-potency PhP.eB—produced striking brain-wide expression after intravenous dosing in mice (Deverman et al., 2016; Chan et al., 2017), but follow-up work showed the effect was strain-restricted and hinged on the LY6A receptor on mouse brain endothelium (Huang et al., 2019; Hordeaux et al., 2018; Hordeaux et al., 2019). That receptor dependence explained the uneven performance across mouse backgrounds and the failure to translate to primates that lack LY6A, crystallizing a central lesson for the field: without a defined, conserved receptor, rodent “hits” often stall at the species boundary. The next wave moved discovery into non-human primates (NHPs) to improve human relevance. Screening in marmoset and macaque produced CAP-Mac that outforms AAV9 by 6–9-fold in NHPs (Chuapoco et al., 2023), which supported brain-wide gene transfer across several primate species (marmoset, rhesus, green monkey), albeit with age and species nuances that matter for translation. Parallel Caltech efforts iterated on the PhP.eB scaffold to yield CAP-B10 and CAP-B22, variants that show BBB crossing in marmosets and relative liver detargeting by 6–12-fold (Goertsen et al., 2022); yet infant rhesus studies reported limited potency, underlining how NHP data can still diverge across models and developmental stages. The MaCPNS1/2 family broadened the lens further: evolved in rodents but validated across rodent and primate species, these capsids robustly transduced peripheral nervous system (PNS) and CNS of marmoset and rhesus after IV dosing (4–25-fold increase over AAV9)—useful examples of cross-species performance, though again with context-specific differences (Chen et al., 2022). Interestingly, capsids selected from NHP studies did not translate well in mice, either (Stanton et al., 2023). Together, these programs advanced the state of the art while keeping the spotlight on cross-species translatability as the core risk.

Other factors contributing to this cross-species translatability issue include the evolutionary path used to select the capsid—e.g., libraries selected in C57BL/6J mice can inadvertently enrich for mouse strain–restricted solutions. A parallel effort in capsid engineering for kidney tropism showed that cross-species cycling in mouse, pig, macaque could yield more broadly compatible variants such as AAV.cc47 (Gonzalez et al., 2022). Also, post-entry events, including uncoating, episome nuclear import, and chromatinization of the delivered genome, can be species-dependent and determine how much transgene is ultimately expressed in human cells even if the vector enters the same number of nuclei (Gonzalez-Sandoval et al., 2023; Loeb et al., 2024). These findings emphasize that “translatability” spans receptor biology, evolutionary selection pressure, and nucleus-level epigenetic fate of the rAAV genome.

The most straightforward way to reduce that risk is to identify the endothelial receptor that mediates transport and design capsids around it. Voyager used receptor-mapping to show that alkaline phosphatase (ALPL) is the primary receptor for its cross-species BBB capsid VCAP-102 (Moyer et al., 2025), with direct binding to human ALPL sufficient to drive receptor-mediated transcytosis in barrier models—an instructive template for mechanistically anchored translation. In addition, another study identified LRP6 as a potential receptor mediating BBB penetration of engineered capsids (Shay et al., 2024). VCAP-102 was reported to outform AAV9 by 20–400-fold across multiple brain regions in all animal species tested.

A complementary path flips the problem: start with a known human BBB receptor and build binding into the capsid. The leading example is transferrin receptor 1 (TfR1/CD71). In 2024, the Broad Institute reported BI-hTFR1, an engineered rAAV that binds human TfR1, actively traverses human brain endothelial models, and delivers ∼40–50× higher CNS expression than AAV9 after IV dosing in human TFRC knock-in mice (Huang et al., 2024). Beyond the performance itself, the work clarified the right preclinical models (humanized TFRC knock-ins rather than wild-type mice) and provided a clean, human-anchored mechanism—receptor-mediated transcytosis—to guide optimization. Translation momentum is building around this concept: pick a human-relevant door and engineer the capsid to use it (Apertura Gene Therapy, 2025).

In addition, several groups are pursuing dose-sparing distribution via local routes to complement systemic capsids. Decades of intrathecal/cisterna magna experience show that cerebrospinal fluid (CSF) delivery can provide broad brain coverage at lower doses, and modern techniques improve the safety/consistency of these procedures (Taghian et al., 2020; Hunter et al., 2025). Building on that foundation, Latus Bio recently unveiled AAV-Ep+ and AAV-DB-3, capsids tuned for ependymal engagement and CSF-driven spread, with preclinical data in mice and NHPs and an explicit goal of reducing clinical dose requirements (Business Wire, 2025). Together, these local-route/capsid innovations offer a practical alternative when systemic dosing is constrained by pre-existing antibodies, safety margins, or biodistribution needs.

The intravitreal (IVT) route remains attractive for its clinical simplicity, but two anatomical/biophysical hurdles dominate: the vitreous gel and the ILM. Canonical AAV2-like capsids engage heparan-sulfate proteoglycans (HSPGs) through a cluster of basic residues on the threefold spikes—R585/R588 most critically, with contributions from R484, R487, and K532—which promotes adsorption to the vitreous/ILM and limits penetration to deeper retinal layers (Kern et al., 2003; Opie et al., 2003). Structure–function studies and mutational mapping across two decades converge on this HSPG footprint, and several groups have shown that reducing—but not abolishing—HSPG binding improves ILM traversal and outer-retina access after IVT dosing (Perabo et al., 2006). Fine-tuning basic charge near 585–588 (and neighboring positions) is therefore a central design lever for IVT-optimized capsids (Ro et al., 2025). These efforts pointed a direction for capsid engineering for improved retinal tropism, unlike muscle- and CNS-tropic capsids where the mechanism is less elucidated and strategy is more diversified.

A key proof of principle is AAV2.7m8, derived by in vivo directed evolution with a 10 amino acid insertion in loop 4—within the heparin/HSPG-binding domain—to enhance outer-retina transduction after IVT (Bennett et al., 2020; Dalkara et al., 2013). The original report and subsequent analyses indicate that altered HSPG interactions and physical bypass of ILM constraints underlie its improved performance relative to parental AAV2, establishing the template for modern IVT discovery campaigns. R100, the intravitreal capsid was engineered to penetrate the ILM and drive pan-retinal expression at single, low IVT doses (Calton et al., 2024). Multi-study updates in wet age-related macular degeneration (AMD) and diabetic macular edema (DME) report durable activity and favorable ocular tolerability profiles, consistent with the intended ILM traversal and broad coverage. Also, AAV2.GL and AAV2.NN use an AAV2 backbone incorporating a rationally chosen peptide insertion within surface-exposed loop IV, at a position known to modulate receptor binding and ILM traversal (Pavlou et al., 2021). In mice, AAV2.7m8 achieves ∼5–6-fold higher photoreceptor transduction than AAV2. Rationally engineered AAV2.GL and AAV2.NN further improve outer retinal expression, delivering ∼12–13-fold higher transcript levels than AAV2 and significantly outperforming AAV2.7m8 in head-to-head analyses (Pavlou et al., 2021). Mechanistically, the peptide insertion perturbs the canonical HSPG patch, consistent with the idea that reduced—rather than abolished—HSPG engagement is optimal for IVT performance. In addition, P2-V1 combines 3–5-fold higher macular/outer-retina expression over AAV2 with better evasion of neutralizing antibodies in human vitreous; phosphorylation site tweaks (e.g., YF/TV) further enhance post-entry handling and far-peripheral photoreceptor transduction (Kellish et al., 2023).

A complementary body of work explores suprachoroidal delivery, which accesses the choroid/RPE complex and can produce wide posterior-segment spread with a minimally invasive microinjector. Recent phase 2 datasets with RGX-314 (AAV8, anti-VEGF) via suprachoroidal support route feasibility and clinically meaningful anti-VEGF durability signals (Regenxbio, 2024). Beyond single programs, reviews of ocular drug delivery highlight suprachoroidal as a platform route for retina-wide gene delivery and anti-VEGF suppression, while newer data continue to refine the technique and safety profile. On the capsid for suprachoroidal delivery, AAVv128 was recently reported to achieve robust transduction across species (mouse, rabbit, NHP) with broader layer coverage after intraocular delivery, and it has shown particularly strong performance with suprachoroidal injection (several folds higher transgene expression than AAV8). These results underscore that both capsid design and route can be co-optimized for maximal retinal access (Luo et al., 2024). Similarly, coAAV-SCS-01 exhibited up to 26-fold outperformance of other capsids in targeting RPE-choroid and retinal cells of NHPs (Coave Therapeutics, 2025), making it promising for further clinical development.

Finally, several groups are pursuing geometry-aware capsids to expand reach from local injections. Atsena Therapeutics’ laterally spreading capsid, AAV.SPR, was designed to spread beyond subretinal bleb margins, enabling central-retina (foveal) exposure without foveal detachment—a clinically meaningful advantage in indications like X-linked retinoschisis (XLRS) (Couto et al., 2022). A phase 1/2 clinical trial (NCT05878860) has been initiated to assess the safety and efficacy of rAAV gene therapy for X-linked retinoschisis via subretinal administration of AAV.SPR.

For retina, the most successful IVT strategies modulate—not abolish—HSPG binding to traverse the ILM (e.g., AAV2.7m8, R100), while suprachoroidal delivery and laterally spreading subretinal vectors (AAV.SPR) offer complementary paths to pan-retinal coverage at dose-sparing exposures. Future programs should treat capsid chemistry and route engineering as a coupled system, validated across species and routes with metrics that emphasize coverage, durability, and ocular safety.

Systemic rAAV therapies for muscle diseases have demonstrated consistent biological activity—including transgene expression and restoration of missing proteins—but clinical benefit has been variable. In Duchenne muscular dystrophy (DMD), Elevidys received accelerated approval based on micro-dystrophin expression, yet the EMBARK Phase 3 trial did not meet its primary functional endpoint, and secondary outcomes showed only modest, non-significant trends. Thus, while biological activity is clear, meaningful clinical improvement remains uncertain, and high systemic doses continue to present safety challenges. In June–July 2025, the U.S. FDA disclosed investigations of fatal acute liver failure following Sarepta’s AAVrh74 platform (two deaths in non-ambulatory DMD treated with ELEVIDYS and one death in an adult with LGMD treated with SRP-9004), asked the company to suspend ELEVIDYS distribution, and placed related trials on hold while causality and risk-mitigation measures were reviewed (U.S. Food and Drug Administration, 2025). For limb-girdle muscular dystrophy (LGMD2E/R4), SRP-9003 has shown more consistent biomarker and early functional gains, and a recent Phase 3 readout reported meeting its primary endpoint in ambulatory patients. Even so, questions regarding durability and dose-related safety remain. These experiences underscore the need for next-generation muscle-tropic capsids that can achieve therapeutic benefit at lower systemic capsid burdens, reduce liver uptake, and improve safety—motivating ongoing efforts in rational engineering, directed evolution, and machine-learning-guided capsid optimization.

A strong preclinical foundation for dose reduction comes from muscle-tropic capsids discovered by in vivo selections: AAVMYO emerged from massively parallel in vivo screening as a highly myotropic AAV9 mutant that outperformed AAV9 by 11–61-fold in skeletal muscle, diaphragm, and heart of mice (Weinmann et al., 2020), while relatively sparing the liver—data that helped catalyze today’s “muscle-biased, liver-detargeted” design goals. In parallel, Tabebordbar et al. (2021) identified RGD-containing MyoAAV capsids with markedly superior skeletal muscle and cardiac transduction after IV dosing in mice and NHPs (10–50-fold higher mRNA levels than AAVrh74 in NHPs), establishing modern benchmarks for cross-species myotropism. Building directly on these efforts, Kate Therapeutics reported MyoAAV-LD variants (e.g., KT809 class) at ASGCT 2024 (PR Newswire, 2024), showing robust, uniform expression in skeletal and cardiac muscle at lower IV doses in NHPs together with reduced liver transduction, positioning the platform for clinically meaningful dose cuts versus legacy capsids.

Translation to patients is underway with proprietary clinical vectors. Solid Biosciences’ next-generation DMD program SGT-003 uses the AAV-SLB101 capsid; early 2025 readouts reported that the first six participants had detectable microdystrophin expression with an initial tolerability profile supportive of continued development, and the company outlined plans for regulatory dialogue on potential accelerated pathways (Solid Biosciences, 2025). While peer-reviewed NHP head-to-heads remain limited publicly, the data are directionally consistent with the dose reduction and liver-detargeting objective.

Across all these engineered capsids, the consensus is clear: achieving strong myotropism together with explicit liver and DRG detargeting represents the most direct route to widening the therapeutic index for systemic muscle gene therapy. For clinical translation, programs should prioritize NHP evidence of (i) uniform skeletal and cardiac muscle expression at lower total capsid burden, (ii) quantitative liver and DRG detargeting by biodistribution and histopathology, and (iii) immunosuppression strategies that remain feasible and safe in vulnerable neuromuscular populations. These considerations have become increasingly important in light of recent safety communications involving high-dose systemic administration of AAVrh74, which emphasized that serious adverse events likely arise from the combined effects of large capsid loads, strong hepatic uptake typical of AAV9-like vectors, and underlying disease fragility—rather than from any single capsid alone.

High-dose systemic rAAV toxicity arises from a combination of vector quality, delivery route, organ tropism, and patient susceptibility—rather than solely from limited potency of natural capsids. First, rAAV lots often contain a high proportion of empty or partially filled capsids, increasing vp/kg exposure relative to vg/kg and amplifying complement activation, receptor competition, and innate immune sensing. Second, systemic intravenous administration exposes vector to hepatic and splenic clearance pathways, whereas locoregional delivery (e.g., IVT, subretinal, intrathecal, intracisterna magna) achieves therapeutic expression at much lower total capsid loads. Third, natural serotypes such as AAV8 and AAV9 show strong hepatotropism, causing the liver to sequester a large fraction of the infused vector and increasing hepatotoxicity risk. Finally, patients with severe neuromuscular disease—including X-linked myotubular myopathy (XLMTM) and advanced DMD—often have baseline hepatic dysfunction, inflammation, and reduced physiologic reserve, making them more vulnerable to systemic stress and immune activation at doses tolerated by healthier individuals. Therefore, capsid engineering and improved manufacturing must aim to reduce capsid burden (vp/kg), detarget the liver, enhance packaging fidelity, and improve tissue specificity to achieve meaningful dose reductions and safer systemic administration.

Regulatory expectations for engineered rAAV capsids have evolved significantly. FDA now routinely requires (i) quantitative biodistribution of novel capsids across rodents and NHPs; (ii) mechanistic justification for any receptor-targeting strategy (e.g., TfR1, ALPL) including binding affinity, cross-reactivity, and saturability; (iii) manufacturability data, because engineered capsids may alter empty:full ratios, particle stability, or genome integrity; and (iv) liver, DRG, and complement-related safety packages enriched with biomarkers (e.g., aspartate aminotransferase, alanine aminotransferase, neurofilament light chain) and dose-response curves. For systemic indications, FDA increasingly asks sponsors to propose strategies for dose minimization, including capsid potency metrics, route optimization, and predefined halting rules for liver or complement activation. These expectations place quantitative performance benchmarks—rather than qualitative tropism alone—at the center of regulatory evaluation.

Ethical considerations now prominently influence rAAV trial design. High-dose systemic rAAV administration (>1 × 10¹⁴ vg/kg) has been linked to acute liver failure, complement-mediated shock, thrombotic microangiopathy, and multiple patient deaths (Table 3). Ethically, this creates an obligation to (i) justify dose selection with quantitative potency data, including lower-dose efficacy in NHPs; (ii) provide a clear rationale for why alternative routes (intrathecal, intracisterna magna, regional limb infusion) or lower-dose engineered capsids cannot achieve similar outcomes; (iii) implement independent safety monitoring, real-time complement/liver biomarker surveillance, and stringent stopping rules; and (iv) design trials that minimize pediatric exposure when adult mechanistic and biodistribution data are insufficient. As engineered capsids achieve multi-fold dose reductions, their use becomes not only a scientific improvement but an ethical requirement to mitigate preventable risk.

The field is converging on a pragmatic, clinically driven blueprint. First, future CNS programs are expected to favor human-anchored mechanisms (e.g., TfR1- or ALPL-mediated receptor transcytosis) over rodent-only capsids, with humanized knock-in models plus NHP biodistribution serving as the translational gatekeepers (Shay et al., 2024; Opie et al., 2003; Perabo et al., 2006; Huang et al., 2023). Second, capsid optimization is no longer single-objective; potency in the target tissue, detargeting of high-liability organs, manufacturability/packaging quality, and resistance to prevalent neutralizing antibodies are being co-optimized from the very first design cycle (Tabebordbar et al., 2021; PR Newswire, 2024; Solid Biosciences, 2025; Zhong et al., 2008; Boye et al., 2016). Third, dose itself is now the most important parameter: the most important deliverable for next-generation capsids is multi-fold dose reduction at equal or better efficacy, because that directly widens the therapeutic window, lowers manufacturing cost per patient, and reduces the likelihood of catastrophic systemic toxicities (Shay et al., 2024; Huang et al., 2024; Apertura Gene Therapy, 2025; Taghian et al., 2020; Weinmann et al., 2020; Tabebordbar et al., 2021; PR Newswire, 2024; Solid Biosciences, 2025; Zhong et al., 2008; Boye et al., 2016). Fourth, patient access and durability will increasingly hinge on immune management and re-dosing: IgG- and IgM-cleaving biologics and complementary immunomodulatory regimens are being advanced precisely to (i) open trials to patients who are currently excluded by pre-existing anti-AAV antibodies, (ii) permit follow-on dosing to boost expression as children grow, and (iii) improve safety by reducing complement-driven acute reactions (Leborgne et al., 2020; Elmore et al., 2020; Smith et al., 2024). Finally, looking 5–10 years ahead, we anticipate three transformative shifts: (i) receptor-defined BBB shuttles making systemic neurology gene therapy viable at doses an order of magnitude below today’s benchmarks; (ii) office-based intravitreal or suprachoroidal ocular gene therapy that no longer requires subretinal surgery; and (iii) muscle-predominant, liver-detargeted capsids that unlock safe systemic treatment for neuromuscular and cardiomyopathic diseases. Together, these advances point toward an rAAV gene therapy landscape that is more precise, lower dose, more redosable, and ultimately more accessible to patients.