Engineering viral vectors for acoustically targeted gene delivery

Gene therapy using AAV vectors is a powerful approach for treating disease, but delivery to the brain is challenging. Standard intraparenchymal injections are invasive and spatially limited, while systemic or intrathecal routes lack regional specificity. Focused ultrasound–mediated blood–brain barrier opening (FUS-BBBO) can noninvasively and regionally permit AAV entry into targeted brain sites. Yet, when paired with natural AAV serotypes (e.g., AAV9), FUS-BBBO yields modest on-target transduction and significant peripheral organ transduction, raising toxicity concerns and necessitating higher systemic doses. The authors hypothesized that wild-type AAVs are not optimized to cross physically loosened BBB barriers and that engineering capsids specifically for FUS-BBBO would increase on-target neuronal transduction while reducing off-target peripheral expression. The study’s purpose was to evolve AAV capsids optimized for FUS-BBBO delivery and validate their efficiency, neuronal tropism, and specificity across brain regions and mouse strains.

Prior work shows FUS-BBBO can noninvasively open the BBB at millimeter-scale foci to enable AAV entry and has been used to deliver reporters, growth factors, opto/chemogenetic tools, and modulate memory in rodents. However, natural AAVs (notably AAV9) require higher IV doses than direct injections and exhibit high peripheral (liver) transduction and potential toxicity. Typical IV AAV9 doses for FUS-BBBO studies have been 5×10^9–1.67×10^10 vg/g; lowering dose could reduce toxicity and cost. Capsid engineering via peptide insertions or directed evolution has previously improved tissue specificity, immune evasion, and tracing, but has not been applied to optimize vectors for a specific physical delivery mechanism like FUS-BBBO. The study builds on Cre-dependent in vivo selection approaches to enrich variants with desired neuronal transduction properties.

Screening and selection: An AAV9-based capsid library with 7-mer randomized peptide insertions between residues 587–588/588–589 was constructed. Round 1: A large library (1.3×10^9 sequences; dose 6.7×10^9 vg/g IV) was delivered to transgenic mice expressing Cre in neurons. FUS-BBBO targeted four MRI-guided sites in one hemisphere (0.33 MPa at 1.5 MHz, 1% duty, 1 Hz, 120 pulses; DEFINITY microbubbles 1.5×10^9/g and gadolinium contrast). After 14 days, targeted hemispheres were harvested; viral genomes modified by Cre were selectively amplified by Cre-dependent PCR and sequenced, and the 2098 most abundant sequences were selected. Round 2: These 2098 variants (with codon-synonymous versions) were repackaged and injected IV at 1.3×10^9 vg/g total (≈1.5–3×10^7 vg per clone per mouse). FUS-BBBO was performed in one hemisphere; after 14 days, both hemispheres were dissected, Cre-dependent PCR performed, and NGS quantified variant enrichment. Selection criteria: variants ≥100-fold enriched in the FUS-targeted vs untargeted hemisphere in both mice and supported by multiple codon sequences; 35 met criteria. The five most abundant were synthesized as AAV.FUS.1–5 (inserted between 587–588 of AAV9). Validation studies: Co-injection comparisons of each AAV.FUS candidate with AAV9 were performed in the same mice to control for targeting and dose. Reporters: AAV9-mCherry and AAV.FUS-EGFP under CaG promoter; IV dose typically 1×10^10 vg/g for regional tests and 1×10^6 vg/g for low-dose tests. FUS-BBBO parameters as above; MRI guidance for main experiments. Quantification: Two weeks post-delivery, brains and livers were sectioned. Brain: counted GFP+ and mCherry+ cells at FUS sites; off-target brain expression quantified outside insonated regions. Neuronal tropism: NeuN immunostaining; fraction of transduced cells positive for NeuN; additional analysis of astrocyte, microglia/macrophage, and oligodendrocyte markers. Liver: counted transduced cells for each fluorophore. Targeting efficiency metric: ratio of brain to liver transduction normalized to AAV9 in the same animal. Regional performance: Separate cohorts with single-region targeting (cortex, striatum, thalamus, hippocampus, midbrain) using independently titered batches; dose 1×10^10 vg/g. Low-dose study: 1×10^6 vg/g using alternate FUS instrument (RK50) with 1.5 MHz; empirical pressure matching to achieve BBB opening without damage; bregma–lambda targeting. Intraparenchymal injections: Co-injection into hippocampal CA1 (AP −1.94 mm; ML +1.0 mm; DV −1.3 mm) of AAV9 and AAV.FUS.3 at 4×10^8 vg/g each, 200 nL/min, with 5 min dwell. Cross-strain validation: BALB/cJ mice tested for brain and liver transduction and neuronal tropism. Statistical analyses: Two-way ANOVA with Sidak’s test for brain/liver comparisons; one-way ANOVA with Tukey HSD for targeting efficiency; paired t tests where appropriate; 95% CI reported; sample sizes detailed per figure and text. Animals and ethics: 10–14 week-old C57BL/6J, BALB/cJ, Syn-1-Cre mice; both sexes; IACUC-approved protocols at Caltech and Rice. Production: AAV libraries produced via PEI transfection and iodixanol purification; candidates packaged and titered by a commercial service (Vigene) with re-titering in-house. NGS processing filtered and consolidated variable regions; sequences normalized and selected per criteria outlined.

• High-throughput in vivo selection under FUS-BBBO identified 35 AAV9-derived capsid variants enriched ≥100× in FUS-targeted vs untargeted hemispheres; five top sequences (AAV.FUS.1–5) were selected for validation.
• Brain transduction at FUS sites: Four of five candidates (AAV.FUS.1, .2, .3, .5) significantly outperformed AAV9; AAV.FUS.4 showed no improvement. Reported p-values: 0.0274, 0.0003, 0.0052, 0.0087 for .1, .2, .3, .5 respectively (two-way ANOVA with Sidak’s, F(4,24)=59.49).
• Off-target brain expression outside insonated regions remained low and comparable to AAV9 (e.g., AAV9 0.29–0.4% vs AAV.FUS.3 ~0.17–0.19%; not significant).
• Liver transduction: All AAV.FUS candidates showed markedly reduced liver transduction vs co-injected AAV9 (two-way ANOVA F(1,24)=375.9, p<0.0001; AAV.FUS.1 p=0.0058; others p<0.0001).
• Targeting efficiency (brain:liver transduction ratio normalized to AAV9): AAV.FUS.3 was the top performer and significantly exceeded AAV.FUS.1, .2, .4, .5 (all p<0.0001), yielding an overall 12.1-fold improvement vs AAV9.
• Neuronal tropism at FUS sites increased for all candidates: AAV9 neurons 44.7% ±1.5% vs AAV.FUS.1–5 64.6%–69.8%; AAV.FUS.3 highest at 69.8% (all p<0.0001). AAV9 transduced microglia/macrophages and oligodendrocytes more than AAV.FUS (microglia 3.5% vs 0.7%, p=0.0174; oligodendrocytes 74.3% vs 3.4%, p<0.0001).
• Regional performance (dose 1×10^10 vg/g): AAV.FUS.3 significantly outperformed AAV9 in cortex, striatum, thalamus, hippocampus, midbrain, with fold-improvement from 2.4±0.08 (cortex) to 4.3±0.08 (hippocampus). Two-way ANOVA F(1,20)=141.2; region-wise p values significant.
• Low-dose performance (1×10^6 vg/g): AAV.FUS.3 yielded 2.2±0.6× more brain-transduced cells (p=0.0004) and 5.2±1.6× lower liver transduction (p=0.0004) than AAV9; brain:liver ratio 11.6±3.7, comparable to high dose (12.1 vs 11.6; p=0.798). Neuronal transduction across regions: AAV9 12.6%±3.7% vs AAV.FUS.3 54.4%±8.8% (4.6×; p<0.0001).
• Intraparenchymal hippocampal injection: AAV.FUS.3 showed 2.29× higher transduction than AAV9, similar to FUS-BBBO improvement (≈2.56×), indicating enhanced parenchymal transduction efficiency.
• Cross-strain validation (BALB/cJ): AAV.FUS.3 improved brain transduction 3.9±0.1× and reduced liver transduction 4.1±0.3× vs AAV9; brain:liver ratio 16.1±0.9× (greater than 12.1× in C57BL/6J; p=0.000376). Neuronal tropism 73%±2.2% (p<0.0001).

Engineering AAV capsids specifically for FUS-BBBO addressed key limitations of noninvasive brain gene delivery: on-target transduction efficiency increased, peripheral (liver) transduction decreased, and neuronal tropism improved. The internal co-injection design isolates capsid performance from procedural variability, showing that modified capsids, particularly AAV.FUS.3, better exploit FUS-BBBO to deliver genes to targeted neurons across multiple brain regions and mouse strains at both standard and low doses. Comparable gains after intraparenchymal injection suggest that part of the improvement stems from enhanced cellular entry/transduction within brain parenchyma, not solely from differential BBB passage. Potential mechanisms include reduced peripheral uptake, altered extracellular matrix interactions, or FUS-induced endothelial changes affecting vector binding/transport. These findings support a paradigm of co-designing viral vectors with physical delivery methods to achieve higher specificity and safety, with translational promise given prior demonstrations of FUS-BBBO AAV delivery in nonhuman primates. Further mechanistic insights could guide next-generation capsids tailored to regional tropism and optimal FUS parameters.

This work demonstrates that AAV capsids can be evolved to synergize with a physical delivery modality, yielding AAV.FUS.3—the first viral vector engineered for focused ultrasound–mediated BBB opening. Across brain regions and mouse strains, AAV.FUS.3 increases neuronal transduction, reduces liver transduction, and improves brain-targeting specificity by ~12-fold over AAV9, with efficacy preserved at low doses and after direct intraparenchymal delivery. These advances can lower required viral doses, reduce off-target risks, and broaden the utility of noninvasive, region-specific gene delivery. Future research should elucidate the mechanisms of BBB transit and parenchymal transduction for engineered capsids, optimize region-specific tropism, evaluate safety and biodistribution (including DRG and other peripheral tissues) in large animals, and assess immunogenicity and efficacy in humans.

• Species and strain scope: Primary data in mice (C57BL/6J, BALB/cJ); translation to large animals and humans remains untested.
• Peripheral assessment focused on liver; comprehensive biodistribution and toxicity (including DRG, other organs, and long-term effects) were not fully evaluated.
• Short-term evaluation (≈2 weeks) without long-term expression, safety, or immunogenicity profiling.
• Selection used neuronal Cre expression, potentially biasing toward neuronal tropism and not assessing other clinically relevant cell types.
• FUS parameter and equipment variations (especially in low-dose experiments) may influence generalizability; precise pressure calibration differences noted.
• Pre-existing anti-AAV immunity and neutralizing antibodies were not assessed, which can impact clinical translation.