Extracellular Vesicle-Encapsulated AAVs for Therapeutic Gene Delivery to the Heart

Abstract

Background:

Adeno-associated virus (AAV) has emerged as one of the best tools for cardiac gene delivery due to its cardiotropism, long-term expression, and safety. However, a significant challenge to its successful clinical use are pre-existing neutralizing antibodies (NAbs), which bind to free AAVs, prevent efficient gene transduction and reduce or negate therapeutic effects. Here we describe extracellular vesicle-encapsulated AAVs (EV-AAVs), secreted naturally by AAV-producing cells, as a superior cardiac gene delivery vector that delivers more genes and offers higher NAb resistance.

Methods:

We developed a two-step density-gradient ultracentrifugation method to isolate highly purified EV-AAVs. We compared the gene delivery and therapeutic efficacy of EV-AAVs with equal titer of free-AAVs in presence of NAbs, both in vitro and in vivo. In addition, we investigated the mechanism of EV-AAV uptake in human left ventricular and hiPSC cardiomyocytes in vitro and mouse models in vivo, using a combination of biochemical techniques, flow cytometry and immunofluorescence imaging.

Results:

Using cardiotropic AAV serotypes 6 and 9 and several reporter constructs, we demonstrated that EV-AAVs deliver significantly higher quantities of genes than AAVs in the presence of NAbs, both to human left ventricular and hiPSC cardiomyocytes in vitro, and to mouse hearts in vivo. Intramyocardial delivery of EV-AAV9-SERCA2a to infarcted hearts in pre-immunized mice significantly improved ejection fraction and fractional shortening compared to AAV9-SERCA2a delivery. These data validated NAb evasion by and therapeutic efficacy of EV-AAV9 vectors. Interestingly, trafficking studies using hiPSC derived cells in vitro and in mouse hearts in vivo showed significantly higher expression of EV-AAV6/9-delivered genes in cardiomyocytes compared to non-cardiomyocytes, even with comparable cellular uptake. Using cellular sub-fraction analyses and pH-sensitive dyes, we discovered that EV-AAVs were internalized into acidic endosomal compartments of cardiomyocytes for releasing and acidifying AAVs for their nuclear uptake.

Conclusion:

Together, using five different in vitro and in vivo model systems, we demonstrate significantly higher potency and therapeutic efficacy of EV-AAV vectors compared to free AAVs in presence of NAb. These results establish the potential of EV-AAV vectors as a gene delivery tool to treat heart failure.

INTRODUCTION

Gene therapy represents a major shift in medicine with treatments for more than 40 diseases undergoing clinical trials¹ and 5 FDA-approved viral-vector-based drugs are currently in use (https://www.fda.gov/vaccines-blood-biologics/cellular-gene-therapy-products/approved-cellular-and-gene-therapy-products). The gene therapy-based clinical trials not only include rare monogenic conditions such as inherited cardiomyopathies^2–4 but also more common ones like diabetes and heart failure (HF).^5–7 Available pharmacological drugs to treat heart failure control the symptoms and the disease but do not cure it. By contrast, gene therapy has the potential to correct underlying defects and pathologies and mediate long-lasting improvements in cardiomyocyte function.^8–10 Current strategies for myocardial gene transfer use adeno-associated viruses (AAVs), due to their non-pathogenic capacity;¹ rare integration;^11–13 and prolonged, high transgene expression level.^14–16 Recombinant AAV serotypes (termed AAVs) have been shown to persist largely as episomes for ~24 months after transduction,¹⁷ a result that shows AAVs are highly suitable for chronic HF therapy.

Although the concept of systemically delivering AAV to intervene directly within the genetic and molecular foundations of cardiac cells is simple and elegant, the path to clinical reality has been arduous. One major obstacle is the presence of neutralizing antibodies (NAbs), which form during natural exposure to AAV or following AAV administration. Data from clinical studies suggest that AAV NAbs, even at relatively low titers, block gene transduction and reduce or negate the effects of therapy.^18,19 NAbs against AAVs increase with age and appear to be common in all populations studied.²⁰ The high prevalence of pre-existing AAV NAbs in patients hinders the efficiency of AAV-based gene therapy and limits the number of patients eligible for enrollment in gene therapy trials.^18–23 In a recent clinical study of gene therapy with AAV1/sarcoplasmic reticulum calcium ATPase 2a (AAV1/SERCA2a),²⁴ AAV1 NAbs were found in 59.5% of a cohort of 1,552 heart failure patients. In a separate study, high-dose immunosuppression did not prevent AAV1 NAbs formation in minipigs.²⁴ Since patients may require more than one administration of viral-based gene therapy, strategies to circumvent NAbs are necessary.

Extracellular vesicles (EVs) are membrane-bound vesicles that actively shuttle biomolecules such as lipids, proteins and RNA released from different cell types. Recent studies have shown that EVs encapsulate several types of viruses including hepatitis, HIV, and AAV^25–32 and protect these viruses from antibody-mediated neutralization. Delivering AAVs protected by carrier EVs (EV-encapsulated AAVs or EV-AAVs) is a promising approach to circumvent NAb neutralization in AAV-based gene therapy. The robust EV membrane can shield AAVs from neutralizing antibodies, and EV-AAVs can therefore present higher resistance to NAbs.³³ Current research indicates that naturally secreted large and small EVs from AAV-producing HEK293T cells can carry and deliver intact AAVs³³ to the retina³², inner ear²⁶, liver², and nervous system²⁷ in mice.

Free AAVs, along with EV-AAVs are secreted out to the conditioned media of AAV-producing cells. Presence of free AAVs in EV-AAV preparations may reduce effective EV-AAV dosing, lower their NAb resistance, and cause unwanted side effects from higher levels of free AAVs. Efficiently removing free AAV contaminants from EV-AAVs is a critical first step in their clinical translation. Earlier attempts to purify EV-AAVs with neurotropic AAV serotypes 1, 8, and 6^31,34,35 resulted in limited NAb resistance, possibly due to insufficient separation from AAVs or their inadequate characterization. Here we evaluated the delivery and therapeutic efficacy of EV-AAV-mediated gene transfer to the heart as compared tomethods that use standard free AAV vectors, both with and without NAbs. We 1) optimized an ultracentrifugation-based EV-AAV purification strategy to isolate highly pure EV-AAVs from conditioned medium from AAV-producing HEK293T cells, with minimum free AAV contamination; 2) comprehensively characterized EV-AAV size, purity, concentration, morphology, presence of EV marker proteins, and molecular content; 3) evaluated the gene delivery efficacy and NAb resistance of EV-AAVs compared to standard free-AAV vectors, using cardiotropic AAV serotypes 6 and 9 and several reporter constructs, in both human left ventricular cardiomyocytes (CMs) and human induced pluripotent stem cells derived cardiomyocytes (hiPSC-CMs) in vitro and in murine hearts in vivo; 4) validated the therapeutic efficacy of EV-AAV-mediated delivery, versus AAV-mediated delivery, of SERCA2a in a preclinical model of myocardial infarction (MI) in pre-immunized mice; and 5) investigated EV-AAV intracellular trafficking and cardiotropic mechanisms in mice hearts in vivo and human cardiomyocytes in vitro. Together, using five different in vitro and in vivo model systems, our findings established a new strategy for clinical translation of AAV-based cardiac gene therapy.

MATERIALS AND METHODS

The data that support the findings of this study are available from the corresponding author on reasonable request. Expanded methods are provided in the Supplemental Material.

Study design

The aim of our study was to evaluate the efficacy of EV-AAVs for cardiac gene delivery and NAb resistance. First, we designed a two-step iodixanol density-gradient ultracentrifugation (DGUC) method to purify EV-AAVs, free of AAV contamination. We comprehensively characterized the isolated EV-AAVs for their morphology, presence of EV specific markers, size, concentration and AAV genome titer using Western blot, nano-flow cytometry, tunable resistive pulse sensing (TRPS) with qNano, qPCR, transmission electron microscopy (TEM), dynamic light scattering analysis (DLS), ExoView chip analysis and RNAseq. Next, we designed in vitro and in vivo studies using five different model systems, cardiotropic AAV serotypes 6 and 9 and several reporter constructs to evaluate the gene delivery efficacy of EV-AAVs in the presence of NAb compared it with free-AAVs. We used flow cytometry, confocal microscopy, and in vivo bioluminescence imaging to quantify gene delivery and gene expression. To assess the therapeutic potential of EV-AAVs, we intramyocardially injected EV-AAV9-SERCA2a to post-MI (ligation of left anterior descending artery, LAD) mouse hearts with or without NAb and evaluated % ejection fraction and % fractional shortening of the heart by echocardiography. To investigate the mechanism of EV-AAV uptake in cardiomyocytes, PKH67-fluorescent dye or pH sensitive (CypHer) dye-labeled EV-AAVs were used to quantify internalization of EV-AAVs in vivo or in vitro using flow cytometry and immunofluorescence. All animal procedures were approved by the Icahn School of Medicine at Mount Sinai Animal Care and Use Committee. Nude mice (Nu/J; ~8–10 weeks) were used in all our studies.

Statistical analysis

Groups were first assessed for normality to determine whether parametric or nonparametric tests would be used. Assumptions of normality were tested with Shapiro-Wilk test. Levene’s Test was used for equality of variances. Normally distributed data with equal variances were statistically analyzed with either two-tailed Student’s t-test (two groups) or one-way ANOVA (three or more groups with one experimental factor) with Tukey multiple comparison test. Normally distributed data with unequal variance were statistically tested with either Welch’s t-test (two groups) or Welch’s one-way ANOVA (three or more groups with one experimental factor) followed by Dunnett-T3 multiple comparison test. Normally distributed data with two experimental factors were analyzed with two-way ANOVA followed by pairwise comparisons with Bonferroni multiple comparison test or Dunnett multiple comparison test. Repeated measures data with two experimental factors were analyzed with two-way repeated-measures ANOVA followed by Bonferroni multiple comparison test. For not normally distributed data, Mann Whitney U test was used to compare two groups. All data are shown either as the mean ± SEM (normally distributed data) or median ± range or interquartile range (not normally distributed data). All data were graphed by Prism 9 (GraphPad software, version 9.5.0) and statistical analyses were performed in SPSS (IBM SPSS Statistics, version 29.0). The statistical details, sample size and significance levels for each experiment are specified in the figure legends.

RESULTS

Density gradient ultracentrifugation purification successfully enriched EV-AAVs

Our overall goals were to isolate and thoroughly characterize highly purified EV-AAVs and evaluate its cardiac gene delivery efficacy and NAb resistance compared to conventionally produced free AAVs (Figure 1A). FDA-approved HEK293T cells, widely used to generate AAVs, are known to secrete EVs.^36–38 Several previous studies have demonstrated that AAV-producing HEK293T cells naturally secrete EV-AAVs into cell culture media.^27,28,34 To determine the AAV-secretory potency of AAV-producing HEK293T cells, we quantified the total amounts of AAVs (EV-AAVs and free AAVs) either secreted into conditioned media or contained within the cells (Figure S1A). We found modest variations in AAV-secretory potency by serotype (AAV6 and AAV9; Figure S1B), in line with recently published literature.³⁸

Figure 1. Density-gradient ultracentrifugation-based purification process successfully enriched EV-AAVs.

(A) Schematic of EV-AAV isolation, characterization, function, tropism and uptake mechanism. (B) Density of 12 fractions obtained afteriodixanol density-gradient ultracentrifugation, n=3. (C) Western blot analysis of the 12 fractions. Equal volume of fractions was loaded on SDS-PAGE gels, and membranes were blotted with Alix, Flot1, Tsg101, Gm130, and Cyc1. (D) WB quantification for the percent of Alix relative expression in 12 fractions, n=3. (E) Determination of AAV vector genomes with qPCR in the 12 fractions, n=3. (F, G) Nano-flow cytometry analysis and quantification of surface tetraspanin proteins including CD63, CD81, and CD9 in the 12 fractions, n=3. Mean equivalent soluble fluorochromes (MESF) were used to calibrate the fluorescence scale in MESF units for each fraction. No MESF values were detected in fractions other than Fr3 and 4 where EV-AAVs were properly separated from other contaminants. (H) TRPS-qNano analysis of Fr3 and 4 measured EV-AAV size. (B, D, E, and F) are presented as mean ± SEM.

To optimize EV-AAV yield, and minimize free AAV presence, we meticulously designed a sequence of steps. Beginning with conditioned media from AAV-producing HEK293T cells (Figure S2), our step-1 involved sequential centrifugation and ultracentrifugation resulting in crude EV-AAV pellets potentially containing free-AAV capsids and other contaminants (Figure S2). A control experiment using standard free AAVs revealed that these crude pellets may pull down different quantities (> 80% of the starting amount) of free AAVs after step-1 (Figure S3A). Since presence of free AAV in EV-AAV preparations could decrease NAb resistance and vary dosing, we devised step-2 (Figure S2) to remove free AAVs via a modified iodixanol density gradient ultracentrifugation, with each fraction characterized thoroughly.

Density measurements of all 12 fractions (Fr) from step-2 revealed Fr3 and 4 had the buoyance density reported for EVs (1.07–1.13 g/ml; Figure 1B). Interestingly, Western blot analysis found Fr3 and 4 to be positive for known EV markers Alix, Flot1, and Tsg101 and negative for known EV negative controls Gm130 (Golgi marker) and Cyc1 (mitochondria marker) (Figure 1C and D; Figure S4A). Fr3 and 4 also contained the majority of AAV vector genomes, as determined by qPCR (Fr3 and 4, 28 ± 13%; Figure 1E). Meanwhile, quantification of EVs using nano-flow cytometry for surface tetraspanin markers (CD81, CD63, and CD9) and TRPS-qNano technique for total particle count, showed that the majority of pure EV-AAVs are enriched in Fr3 and 4 (Fr3 and 4, 87.3 ± 6.9% for tetraspanin-positive particles; Fr3 and 4, 99.7 ± 0.2% for total particles; Figure 1F and G; Figure S4B). To confirm EV-AAV presence in Fr3 and 4, we employed TRPS-qNano to measure particle size, which was found to be ~110 nm (Figure 1H; Figure S4C and D). In addition, we detected a weak unreliable signal measuring size and particle concentration from Fr8 (Figure S4E; no particles were detected in Fr1, 2, 5–7, and 9–12) suggesting presence of protein aggregations or contaminating membrane fragments/viral components in Fr8, which were heavier than EVs and were difficult to dissolve. To estimate the extent of free AAVs co-isolated with EV-AAVs in Fr3–4, we subjected different free AAVs doses (as a control) to step-1 and step-2 purification process identical to the EV-AAV isolation protocol. Majority (~60–70%) of free AAVs were found to be concentrated in Fr11 (as expected, between the 40% and 60% iodixanol solution) and in negligible amounts (~3–5%) in Fr3 and Fr4 (Figure S3B). These data are consistent with our particle size, concentration and EV-marker data, and confirm that our EV-AAV preparations from Fr3 and Fr4 are primarily AAV-encapsulated EVs, mostly free of contaminants such as protein aggregates, membrane fragments, and free-AAVs. Interestingly, the low AAV levels detected in Fr11 in step-2 (Figure 1E) suggests that the majority, if not all, of AAVs secreted by the AAV-producing HEK293T cells into the conditioned media could be EV-AAVs.

Taken together, these results demonstrate that our two-step iodixanol DGUC protocol successfully isolated EV-AAVs in Fr3 and 4, which are co-enriched for EV biomarkers, consistent with the density and size of EVs and contain viral vector genomes (vg). In our subsequent experiments, we further characterize and investigate EV-AAVs from Fr3 and 4 and compared their characteristics and function with free AAVs.

EV-AAVs share morphological and biochemical characteristics with EV-WT

To verify AAV presence and locations in isolated EV-AAVs, we processed ultrathin sections of EV-AAV pellets and imaged them using transmission electron microscopy (TEM). Wild type EVs (EV-WT, from untransfected HEK293T cells) and free AAVs were used as controls. TEM micrographs of EV-AAV sections show the majority of AAVs (measuring ~20–25 nm, consistent with known AAV sizes) in Fr3 and Fr4 EV-AAV preparations were encapsulated in membranous extracellular vesicles ~100 nm in size (white arrows) (Figure 2A; Figure S5). We used dynamic light scattering (DLS) analysis to confirm the sizes of EV-AAVs, EV-WT, and free-AAVs. We observed particles with sizes corresponding to EVs (~100 nm, in line with our TEM and qNano measurements) and free-AAVs (~20 nm, similar to TEM measurements) (Figure 2B). No significant size differences between EV-AAVs and EV-WT were detected. Nano-flow cytometry analysis for EV-surface marker proteins showed no difference in tetraspanin expression between EV-AAVs and EV-WT (Figure 2C). In accordance with nano-flow cytometry data, ExoView Chip, an interferometry-based protein expression profiling technique revealed canonical EV markers such as tetraspanin CD81, CD63, and CD9 on both EV-AAVs and EV-WT (Figure 2D). Interestingly, expression of CD9 in EV-AAVs was significantly decreased compared with EV-WT (Figure 2E). In addition, EV-AAVs and EV-WT carried similar EV-marker proteins: Alix, Tsg101, and Flot1 (Figure 2F). As expected, the viral capsid proteins VP1, 2, and 3 were detected in EV-AAVs and AAVs but not in EV-WT (Figure 2F). Interestingly, we detected stronger VP1/2/3 band intensity in EV-AAVs than free-AAVs, even when equal titer of vector genomes (quantified by qPCR) were loaded (Figure 2F). To address this discrepancy, and to examine viral genome integrity, we analyzed the DNA using alkaline agarose gel electrophoresis (Figure 2G) from equal titer of EV-AAVs and AAVs, and quantified the DNA band intensity. There was no significant difference between the DNA quantities between EV-AAVs and free AAVs (Figure 2H), confirming the accuracy of viral genome titer determination by qPCR. More VP1/2/3 protein in EV-AAV isolates may implicate presence of higher quantities of empty capsids compared to AAVs.³⁹

Figure 2. EV-AAVs share morphological and biochemical characteristics with EV-WT and are enriched for unique transcriptomes.

(A) Representative images of transmission electron microscopy (TEM) for EV-AAV (left), EV-WT (middle) and free AAV (right side); arrows indicate virus particles inside EVs. Scale: 100 nm. (B) EV-AAV, EV-WT, and free AAV sizes measured by dynamic light scattering (DLS) analysis. (C) Nano-flow cytometry analysis of EV-AAVs, EV-WT, and free AAVs for the presence of surface tetraspanin proteins including CD63, CD81, and CD9. Buffer plus tetraspanin was used as a control for gating. (D and E) Distribution and quantification of tetraspanin (CD63, CD81, and CD9)-positive particles in EV-AAV and EV-WT by NanoView chips. (F) Western blot of EV-AAVs, EV-WT, and free AAVs. Equal protein amounts (5 μg) for EV-AAV and EV-WT and equal AAV genomes (1e11 vector genome (vg)) for EV-AAV and free AAVs were loaded in the SDS page gel (n=2 technical replicates). Membranes were blotted with Alix, Flot1, Tsg101, Gm130, Cyc1, and VP1/VP2/VP3. (G) Image of alkaline agarose gel electrophoresis for equal vector genomes (1e11 and 5e10) of EV-AAVs and free AAVs. (H) Quantification of band intensity of (G). (I) Principal component analysis to identify the axis of variance showing that EV-AAV, EV-WT, and AAV each enriched for unique transcriptomic profiles. (J) Venn diagram showing an overlap among EV-AAV, EV-WT, and AAV and a set of common RNA present in EV-AAV and EV-WT. (K) Heatmap with top differentially expressed profiles for EV-AAV, EV-WT, and AAV. The scale is standardized (z-score) from the Log₂ Expression values showing the upregulated genes in red and the downregulated genes in blue. (E and H) were analyzed with two-way ANOVA with Tukey multiple comparison Test. Data are presented as mean ± SEM (n=3); n.s., not significant, *P < 0.05.

To identify distinct transcriptomic signatures in EV-AAVs and EV-WT, we sequenced the RNA from both using small RNA sequencing. Principal component analysis determined the axis of variance and revealed that EV-AAVs and EV-WT were enriched for both common and unique transcriptomic profiles (Figure 2I and J; Figure S6A to C). There were ~600 common transcripts between EV-WT and EV-AAVs while EV-AAVs had more than 200 unique transcripts, indicating that novel molecular signatures originate from AAV-transduced cells (Figure 2J). Heatmap analysis showed the top differentially expressed RNA between EV-AAV and EV-WT (Figure 2K; Figure S6D). Both EV-AAV and EV-WT contained coding and non-coding RNAs that modulated protein stability and metabolism (Figure S6E, Table S1 and S2). Yet EV-AAVs also contained unique RNA cargo that regulates cholesterol metabolism (e.g. MYLIP), myosin regulation (CFAP20), growth, and differentiation factors (NRG-2) which may exert beneficial functions (Figure S7A and B). The RNAs unique to EV-AAVs, may augment the benefits of EV-AAV gene therapy.^40,41 We also sequenced free AAVs and FBS (2% FBS that the HEK cells were cultured in) as control samples, and as expected, neither did not contain significant amounts of RNA and showed presence of identical contaminants (Figure S6B–D). Therefore, samples were not considered in subsequent analyses.

These results confirm that EV-AAVs used in our subsequent studies represent AAV-carrying vesicular fractions with morphology and antigen expression similar to those of exosomes (small EVs of endocytic origin) (Figure 1B-H). We also demonstrate that EV-AAVs and EV-WT share certain morphological and biochemical characteristics (Figure 2).

Together, our two-step DGUC protocol successfully isolated EV-AAVs, as determined from EV buoyance density; presence of EV surface marker proteins, viral capsid proteins and genome; size and morphology; presence of AAVs in the lumen of EVs; and EV-AAV protein and RNA profiles. Moreover, TEM pictures of EV-AAVs, along with parallel processing of a free AAV control established that free AAV contamination is negligible in our EV-AAV preparations. Our comprehensive characterization of EV-AAV is in line with MISEV2018 (Minimal Information for the Study of EVs) guidelines.⁴²

EV-AAV outperforms AAV in delivering genes to human cardiomyocytes in the presence of neutralizing antibodies in vitro

Pre-existing NAbs against AAV is prevalent in human serum.^23,43 NAbs bind to AAV, block its infection, and impair AAV-mediated gene delivery.^44–46 Cardiomyocytes are largely nondividing and are therefore promising targets for therapeutic AAV vectors. Recent reports have demonstrated that membrane-associated AAV vectors can very efficiently transduce various cell types and are less susceptible to antibody-mediated neutralization.^34,47,48 Thus, we hypothesized that EV-mediated AAV delivery would shield AAV vectors from pre-existing humoral immune responses and thereby facilitate improved cardiac gene transfer to cardiomyocytes in the presence of NAbs. To validate this, we first evaluated the transduction efficiency of EV-AAV6 encoding mCherry in human cardiomyocytes in vitro. AAV6 is known to be a muscle-tropic serotype and has superior transduction efficiency in vitro.^49,50 Infecting equal titers of AAV6/9 and EV-AAV6/9 resulted in significantly higher transduction efficiency of AAV6/EV-AAV6 compared to AAV9/EV-AAV9 (Figure S8). This also confirmed that EV-AAV6 can successfully deliver genes to hAC16-CMs in vitro.

Equal titers of EV-AAV6-mCherry and free AAV6-mCherry were pre-incubated with NAb (0–4 mg/ml human intravenous immunoglobulins, IVIG)⁵¹ for 30 minutes at 37°C and treated to human left ventricular (hAC16)-CMs or hiPSC-CMs (Figure 3). Vector transduction efficiency was assessed by quantifying mCherry expression using flow cytometry and confocal microscopy. Our results showed significantly higher transduction efficiency by EV-AAV6 in presence of NAb in both hAC16-CMs (Figure 3A to D) and in SIRPA-positive hiPSC-CMs (Figure 3E to H, Figure S9). Remarkably, EV-AAV6-mCherry retained significantly high transduction efficiency in both hAC16-CMs and hiPSC-CMs even at the highest concentration of NAb (79.2 ± 4.3% and 69.2 ± 13.9% respectively at 4 mg/ml), whereas free AAV6-mCherry transduction was almost completely blocked (0.18% ± 0.29% and 1.6± 1.0% respectively at 4 mg/ml; Figure 3C and F). These results confirm that EV-AAV vectors deliver genes more effectively to cardiomyocytes while also protecting the AAVs from neutralization by anti-capsid antibodies.

Figure 3. EV-AAVs outperform AAVs in gene delivery efficiency to human cardiomyocytes in the presence of NAb in vitro.

(A and B) Representative images of flow cytometry and confocal microscopy analyses of hAC16-CMs at 3 days post infection with equal titers of EV-AAV6-mCherry and AAV6-mCherry. Before being added to cell cultures, AAV6 or EV-AAV6 preparations were incubated with NAb solution (0.125–4 mg/ml) or equal volume of PBS for 30 minutes at 37°C. Scale bar, 150 μm. (C) Flow cytometry quantification of mCherry+ hAC16-CMs after EV-AAV6 or AAV6 transduction at different NAb concentrations (n=4). (D) Confocal microscopy quantification of mCherry intensity in hAC16-CMs after EV-AAV6 or AAV6 transduction at different NAb concentrations (n=4). (C and D) are presented as median with range and were analyzed with Mann-Whitney U test. *P< 0.05. (E and F) Representative images of flow cytometry and confocal microscopy analyses of hiPSC-CMs at 3 days post infection with equal titers of EV-AAV6-mCherry and AAV6-mCherry. Before being added to cell cultures, AAV6 or EV-AAV6 preparations were incubated with NAb solution (0.5–4 mg/ml) or equal volume of PBS for 30 minutes at 37°C. Gating for mCherry+ population is derived from SIRPA+ population. Scale bar, 150 μm. (G) Flow cytometry quantification of mCherry+ hiPSC-CMs after EV-AAV6 or AAV6 transduction at different NAb concentrations (n=3). (H) Confocal microscopy analysis of mCherry intensity in hiPSC-CMs transduced with EV-AAV6 or AAV6s at different NAb concentrations (n=3). Values are present as Log₁₀ mCherry fluorescent intensity. (B and G) were adjusted with the same brightness and contrast. (F and H) are presented as mean ± SEM and were analyzed with Welch’s t-test. *P< 0.05, **P<0.01.

EV-AAV outperforms AAV in delivering genes to mouse hearts in the presence of NAb in vivo

Passive-immunity nude mouse model efficiently mimics naturally occuring immunity to AAV:

To evaluate EV-AAV-mediated gene delivery and NAb resistance in vivo, we used an animal model with pre-existing AAV immunity by intravenously injecting nude mice with IVIG (1 mg/mouse).²⁹ After 24 hours, we intravenously injected the mice with equal titers of either EV-AAV9 or AAV9 (Figure S10A). AAV9 serotype has been shown to have the highest cardiac gene-transduction efficacy in rodents with either systemic or direct cardiac injection.⁵² Sera isolated from IVIG pre-injected mice (NAb+) at 4 weeks significantly inhibited AAV9-FLuc (coding firefly luciferase) in vitro transduction compared to sera from saline-injected mice (NAb-)(Figure S10B). In addition, sera from NAb+ or NAb− mice injected with AAV or EV-AAV almost completely prevented AAV9-FLuc trandsuction in vitro, suggesting that administering AAV or EV-AAV stimuated the mice to produce natural antibodies which were functionally identical to the NAbs we injected (Figure S10B). In sum, we confirmed the neutralizing effect of the sera from NAb pre-injected mice.

Empty virus capsids overcome NAb-mediated immunity to AAV:

To confirm the neutralizing effect against AAV in a pre-existing NAb mouse model, we designed a NAb competition-binding assay with an empty AAV capsid. Empty AAV capsids compete with AAV particles for NAb binding, thereby increasing AAV transduction in the presence of NAb. Nude mice were passively immunized with NAb 24 hours before receiving AAV9-FLuc or empty AAV9 capsids followed by AAV9-FLuc (Figure S10C). After 2 or 4 weeks, we quantified FLuc expression in these mice by bioluminescent imaging (BLI) of the chest area in vivo and isolated hearts ex vivo. AAV-mediated FLuc expression was inhibited in NAb+ mice compared to NAb− (Figure S10D and E). Remarkably, mice injected with the same titers of both empty capsids and standard AAVs showed luciferase expression similar to NAb− mice (Figure S10D and E) suggesting that empty AAV capsids compete with standard AAV particles for NAb binding.

EV-AAVs deliver genes in the presence of NAb:

To test the gene delivery efficacy of EV-AAVs in vivo and to compare with free AAVs, nude mice were injected with NAb (NAb+ mice, 1mg/mouse) or saline (NAb− mice). After 24 hours, equal titers of EV-AAV9-FLuc or AAV9-FLuc were injected intramyocardially and FLuc expression was quantified using BLI of live mice and of their harvested organs after sacrifice at 4 weeks (Figure 4A). Remarkably, luciferase expression was higher in the EV-AAV9-FLuc-injected mice than the AAV9-FLuc-injected mice at 1 and 2 weeks (Figure S11) and both the NAb+ and NAb− mice at 4 weeks (Figure 4B and C, at least two-fold). In the AAV9-FLuc-NAb+ mice, luciferase expression was significantly lower than that in the EV-AAV9-FLuc-NAb+ mice; however, there was no difference between EV-AAV9-FLuc-NAb− and EV-AAV9-FLuc-NAb+ mice. These data demonstrate that EV-AAVs protect the encapsulated AAVs from NAb neutralization. Consistent with the in vivo BLI results, ex vivo imaging of isolated whole hearts showed significantly elevated luciferase expression in the EV-AAV9-FLuc-injected hearts than the AAV9-FLuc hearts in the presence of NAb (Figure 4D and E). We also observed similarly elevated luciferase expression in the livers of NAb+ and NAb− mice treated with EV-AAV9-FLuc compared with standard AAV9-FLuc (Figure S12A and B), though other organs, such as spleen, brain, and kidney, had almost no expression (Figure S12C). Transduction was also detected in the liver, but not other organs, and AAV9 vectors were consistent with the organ specificity previously described for this serotype.⁵³

Figure 4. EV-AAVs deliver genes more efficiently than AAVs in the presence of NAb in vivo.

(A) Study design. (B) Bioluminescent images of nude mice at 4 weeks post injection of equal titer (5e10 vg/mouse) of EV-AAV9-FLuc, AAV9-FLuc, or saline (as an imaging negative control) directly into myocardium. 24 hours before EV-AAV9/AAV9 administration, mice were intravenously injected with saline (NAb-) or NAb (NAb+, 1mg NAb/mouse). (C) Bioluminescent signal quantification of EV-AAV-FLuc, AAV-FLuc, and saline in the heart and liver regions of NAb− and NAb+ mice (n=4). (D) Ex vivo imaging of hearts from NAb− and NAb+ mice injected with EV-AAV9-FLuc, AAV9-FLuc, and saline. Mice were sacrificed at 4 weeks after EV-AAV9/AAV9 administration; organs were removed and imaged in the IVIS system. (E) Bioluminescent signal quantitation for EV-AAV, AAV, and saline in ex vivo hearts from NAb− and NAb+ mice (n=4). (C and E) were analyzed with two-way ANOVA and pairwise comparisons with Bonferroni multiple comparison test. Saline groups were used to assess background (negative controls) and are excluded from analysis. *P< 0.05. Data are presented as mean ± SEM.

These data provide compelling evidence that EV-AAVs, but not free AAVs, resist neutralization by NAb in vivo in a murine model. In addition, these results demonstrate that gene delivery by EV-AAVs is significantly more effective than by free-AAVs, both in the absence and presence of pre-existing antibodies.

EV-AAV-mediated SERCA2a gene delivery significantly improves cardiac function in the presence of NAb in a preclinical mouse model of MI

Having confirmed that EV-AAV9 vectors evade NAbs in mice, we investigated therapeutic delivery of genes such as SERCA2a to the myocardium in a preclinical setting. Earlier studies by our group and others have shown that SERCA2a plays a major role in regulating calcium level in cardiomyocytes, thereby affecting contractile function.^22,54–56 Reduced SERCA2a activity closely associates with contractile dysfunction and heart failure.^57,58 Specifically, our previously published studies showed that AAV-mediated SERCA2a gene delivery reversed contractile dysfunction and myocardial remodeling in both rodent^59,60 and large animal models.⁶¹ However, NAbs prevalence is still a significant obstacle to delivering genes,^20,24 such as SERCA2a, to patients.

To address this challenge, we compared the therapeutic effect on cardiac function of SERCA2a delivered using EV-AAV9 versus free AAV9 in a MI model in nude NAb− and NAb+ mice. MI was induced in mice by permanent ligation of LAD 24 hours after injection of either NAb or saline. Immediately after MI surgery, NAb+ and NAb− mice were intramyocardially injected with an equivalent vector dose of either EV-AAV9-SERCA2a or AAV9-SERCA2a (1E11 viral genomes per mouse). To test functional improvements, cardiac ejection fraction (EF) and fractional shortening (FS) were evaluated by echocardiography 2, 4, and 6 weeks post-surgery (Figure 5A). At the 6-week follow-up, EF was significantly higher in the NAb− mice that received EV-AAV9-SERCA2a than conventional free AAV9-SERCA2a or saline (52.2% in EV-AAV9 vs 41.2% in AAV9 and 27.0% in saline control; Figure 5B). In NAb+ mice, EF, FS, and cardiac wall motion in the free AAV9-administrated mice were significantly lower than in NAb− mice due to the neutralizing activity of anti-AAV antibodies (Figure 5B to D). However, SERCA2a delivery using EV-AAV9 outperformed the free AAV9 group in the presence of NAb, significantly enhancing cardiac function as demonstrated by elevated EF and FS values (EF, 55.1% vs. 27.3%; FS, 32.3% vs. 18%; Figure 5B to D). We obtained similar EF and FS results at 2- and 4-weeks post-surgery (Figure 5E and F; Figure S13A to D). Further, expression of SERCA2a and a related calcium ion regulator, Ryr2 mRNAs, quantified from left ventricular tissue, was significantly upregulated in the EV-AAV group compared with the AAV group in the presence of NAb at 6 weeks post-surgery (Figure S13E and F).

Figure 5. EV-AAV-mediated SERCA2a gene delivery significantly improves cardiac function in the presence of NAb in a preclinical mouse model of MI.

(A) Study design. (B) Echocardiographic assessments of left ventricular function showing EF (%) and (C) FS (%) at 6 weeks post MI in NAb− and NAb+ mice (1mg NAb/mouse) injected with equal titer (1e11 vg/mouse) of EV-AAV9-SERCA2a, AAV9-SERCA2a, or saline (n=6–10). Data were analyzed with two-way ANOVA and pairwise comparisons with Bonferroni multiple comparison test. Baseline and sham groups were negative controls and were excluded from the statistical comparison. (D) Representative M-Mode echocardiograms showing anterior and posterior left ventricle wall motion at 6 weeks post MI in NAb− and NAb+ mice. (E) Trendlines for EF and (F) FS, respectively, from 2 weeks to 6 weeks post MI. Data were analyzed with two-way repeated-measures ANOVA followed by Bonferroni multiple comparison test for NAb⁺EV-AAV-SERCA2a group compared with NAb⁺AAV9-SERCA2a group. *P< 0.05, **P< 0.01, and ***P< 0.001. Data are presented as mean ± SEM.

Overall, our in vivo functional data demonstrate that EV-AAV9-SERCA2a significantly improved cardiac function as compared to AAV9-SERCA2a in the presence of NAb, thereby indicating that EV-AAV vectors can increase the efficacy of AAV-mediated gene therapy in the heart and may potentially be used in humans with pre-existing antibodies to AAV.

EV-AAVs exhibit cardiotropism

Off-target transgene expression in liver remains a major challenge for AAV-mediated gene delivery. BLI images from our in vivo experiments showed both cardiac and liver delivery by EV-AAVs. To compare the gene delivery between the liver and the heart, we quantified both vector genome copy number and transgene expression from the heart and liver cells of the mice intramyocardially injected with AAV9-FLuc or EV-AAV9-FLuc (Figure S14). Interestingly, the vector genome copy number was significantly higher, but luciferase transgene expression was significantly lower in cells isolated from liver, compared to both CMs and non-myocytes (NMs) from the heart (Figure S14). This suggests a lack of correlation between vector genome DNA and transgene expression, an observation that has been reported before.⁶²

Our data also showed that EV-AAV-mediated delivery induced higher transgene expression in CMs compared to NMs or liver cells, suggesting a cardiotropic expression. We further investigated the cardiotropic mechanisms of EV-AAV9 in vivo by studying their uptake into cardiac cells by intramyocardially injecting PKH67-labeled EV-AAV9s into nude mice. Flow cytometry data from single cell suspensions revealed that both CMs and NMs internalized PKH67-EV-AAVs at comparable levels at both 1h and 24h (Figure S15). To identify the cell type expressing the transgene, we injected either EV-AAV9-FLuc or AAV9-FLuc intramyocardially, with (Figure 6A) and without NAb (Figure S16), and analyzed transgene expression and vector copies in the heart after 2 weeks. In line with our previous data (Figure 4), BLI images showed higher luciferase expression in the hearts delivered via EV-AAVs compared to free AAVs (Figure 6B and C; Figure S16B and C). Immunofluorescence imaging of the left ventricle (LV) revealed that the luciferase-positive cells were predominantly cTNT⁺ rod-shaped CMs (Figure 6D; Figure S16D), whereas luciferase-negative cells were largely the NMs, including the CD31⁺ endothelial cells or Vimentin⁺ cardiac fibroblasts (Figure 6D). We quantified luciferase expression in the CMs and NMs from the LVs (Figure 6A and E) using flow cytometry analysis for luciferase+ cells, luciferase expression assay and vector genome assessment by qPCR. Our data suggest that luciferase expression in CMs was significantly higher than NMs from both EV-AAV9 and AAV9 injected LVs (Figure 6G to I, Figure S17) implying cardiotropic delivery for both. Interestingly, EV-AAV9 delivered significantly higher luciferase to the CMs compared to AAV9 confirming its superior cardiotropism (Figure 6G to I; Figure S14 and S17).

Figure 6. EV-AAVs exhibit cardiotropism.

(A) Study design to examine cardiotropsim of EV-AAV9 in vivo. (B) In vivo bioluminescent images of nude mice at 2 weeks post intramyocardial injection (1e12 vg/mouse) of EV-AAV9-FLuc or AAV9-FLuc. 24 hours before EV-AAV9/AAV9 administration, each mouse was intravenously injected with 5mg NAb or saline. (C) Bioluminescent signal quantification of EV-AAV-FLuc or AAV-FLuc in the heart regions (n=4). Data were analyzed with the 2-tailed Student’s t-test. (D)Immunofluorescence images for cTNT, Luciferase, CD31 and Vimentin in the left ventricles of hearts from the nude mice at 2 weeks post intramyocardial injection of EV-AAV9-FLuc or AAV9-FLuc. DAPI-blue; cTNT-green; Luciferase-red; CD31/Vimentin-magenta. Scale bar, left, 200μm; right, 50 μm. (E) Representative brightfield images of cardiomyocytes (CMs) and non-cardiomyocytes (NMs). A langendorff perfusion method was used to isolate CMs and NMs from mice heart left ventricles. Scale bar, 50 μm. (F) Flow cytometry images and quantification (G) of Luciferase+ CMs and NMs from the left ventricles injected with EV-AAV9-FLuc or AAV9-FLuc (n=3). (H) Luciferase activity of CMs and NMs from the left ventricles injected with EV-AAV9-FLuc or AAV9-FLuc (n=3). Values are normalized to 1000 cells. (I) Vector genome copies in CMs and NMs from the left ventricle injected with EV-AAV9-FLuc or AAV9-FLuc (n=3). Values are normalized to 1000 cells. (J) Study design to examine cardiotropsim of EV-AAV6s in hiPSC-CMs and NMs. (K) Representativeimmunocytochemistry (ICC) images and quantification for luciferase expression on hiPSC- cTNT⁺-CMs and cTNT⁻NMs transduced with AAV6-FLuc or EV-AAV6-FLuc after 48 hours (MOI= 2e5 vg/cell). DAPI-blue; cTNT-green; luciferase-red. Scale bar, 100 μm. (G, H, I and K) were analyzed with two-way ANOVA and pairwise comparisons with Bonferroni multiple comparison test. (L) Representativeimmunocytochemistry images for PKH67, AAV6 and cTNI of cTNI⁺ hiPSC-CMs and cTNI⁻ hiPSC-NMs at 10 hours post-incubation with PKH67-labeled EV-AAV6 (MOI= 1e6 vg/cell). DAPI-blue; PKH67-green; cTNI-red; AAV6-magenta. Scale bar, left, 50 μm; right, 5 μm. (M) Vector genome copies in isolated nuclei of hiPSC-CMs and NMs at 24-hour post-incubation with EV-AAV6 (MOI=1e5 vg/cell; n=3). Data were analyzed with Welch’s t-test. *P<0.05, **P< 0.01, ***P< 0.001. Data are presented as mean ± SEM.

To investigate whether EV-AAVs also exhibit cardiotropism in human CMs in vitro, we characterized the transgene expression and uptake of AAV6 and EV-AAV6 in hiPSC-CMs and hiPSC-NMs (Figure 6J). hiPSC-dervied cells were treated with equal titer of EV-AAV6-FLuc or AAV6-FLuc and immunostained to quantify luciferease⁺ cells. Similar to the cardiotropic expression in mice LVs, luciferase was primarily expressed in cTNT⁺ CMs in both AAV6 and EV-AAV6 infected hiPSC-derived cells (Figure 6K). Particularly, EV-AAV6 transduced the hiPSC-CMs with significantly greater efficiency compared to AAV6 (Figure 6K). In the same line, fluorescence-activated cell sorting (FACS)-sorted SIPRA⁺CD90⁻ hiPSC-CMs had significantly higher luciferase expression compared to SIPRA⁻CD90⁺ NMs transduced with equal titer of EV-AAV6-FLuc or AAV6-FLuc (Figure S18), confirming cardiotropic delivery of EV-AAVs to hiPSC-CMs.

In line with the in vivo uptake of EV-AAVs, uptake of PKH67-labeled EV-AAV6 in hiPSC-derived cells showed internalization by both cTNI⁺CMs and cTNI⁻NMs (Figure 6L, Figure S19). However, in hiPSC-CMs, the uptaken EV-AAVs were detected around the perinuclear region (stronger green and magenta staining, at 10h), whereas, in hiPSC-NMs, they appeared diffused in the cytoplasm suggesting a distinct intracellular trafficking between CMs and NMs (right panel, Figure 6L). Nuclear entry of AAV is a critical step for successful transgene expression.⁶³ To compare the nuclear entry and genome release of AAV particles between hiPSC-CMs and NMs, we infected each with equal amounts of EV-AAV6 and quantified the vector genomes by qPCR in the extracted nuclei after 24h. Interestingly, EV-AAVs delivered significantly higher quantity of vector genome into the nucleus of hiPSC-CMs compared with NMs (Figure 6M). This suggests a distinction in the post-lysosomal nuclear entry between CMs and NMs, which may play a role in the cardiotropic mechanism of EV-AAV. In sum, these data demonstrate superior cardiotropism of EV-AAV9 in vivo and EV-AAV6 in vitro.

EV-AAV transduction is dependent on AAVR

To gain mechanistic insights into EV-AAV-mediated cardiac gene delivery, we studied its uptake and transduction in cardiomyocytes. AAV transduction is a muti-step process including several endocytic internalization routes, sorting and acidification in the cytoplasm, nuclear entry and transgene expression.^63–65 In contrast, the mechanisms of EV-AAV uptake, intracellular trafficking, and AAV release from EV-AAVs have not been studied yet. Uptake of certain AAV serotypes (e.g. AAV2) have been shown to be mediated by the well-characterized clathrin-independent carriers (CLICs) in glycosylphosphatidylinositol (GPI)-anchored protein-enriched endosomal compartments (GEECs), CLIC/GEEC.^64,66 Knocking down RhoGTPase-activating protein GRAF1, a critical regulator of GLIC/GEEC endocytosis, in hAC16-CMs (Figure S20A and B) did not inhibit AAV6/9 or EV-AAV6/9 transduction (Figure 7A and B, Figure S20C and D), on the contrary, significantly increased it. This suggests that GRAF1-mediated CLIC/GEEC pathway did not facilitate the internalization of EV-AAV6/9 and AAV6/9, but negatively regulated their uptake and expression in hAC16-CMs. Next, we tested another whether essential AAV receptor (AAVR, a recycling receptor trafficking AAVs to the trans-Golgi network, TGN)^67,68, shown to facilitate AAV6/9 infection, is required for EV-AAV transduction. Knocking down AAVR (Figure S20E and F) in hAC16-CMs resulted in significant inhibition in the transduction efficiency of both AAV6/9 and EV-AAV6/9 (Figure 7A and C, Figure S20G and H), suggesting infection of both AAV6/9 and EV-AAV6/9 was dependent on AAVR.

Figure 7. EV-AAV uptake involves trafficking via endocytic/acidic compartments.

(A) Study design to examine involvement endocytic pathway and AAV receptors (AAVR). (B) Relative transduction efficiency of EV-AAV6-FLuc and AAV6-FLuc in hAC16-CMs following GRAF1 or (C) AAVR knockdown by siRNA. Relative luciferase activities are normalized to each siRNA-Control. Data were analyzed with two-way ANOVA followed by Dunnett multiple comparison test (n=3). (D) Flow cytometry analysis showing PKH67-labeled EV-AAV6 uptake in hAC16-CMs at different time points. Scale bar, 20 μm. (E) Quantification of PKH67+ hAC16-CMs at 1, 4, 10, 24 hours (n=3). (F) Confocal microscopy analysis showing the internalization of PKH67-labeled EV-AAV6 in hAC16-CMs at 4 hours and 10 hours. Scale bar, 5 μm. (G) Immunostaining for LAMP1 on hAC16-CMs at 10 hours post incubation with PKH67-labeled EV-AAV6 (blue, DAPI; green, PKH67-labeled EV-AAV; red, LAMP1; yellow, colocalization of PKH67 and LAMP1). Scale bar, 2 μm. (H) Three-dimensional (3D) visualization showing the colocalization of PKH67-labeled EV-AAV6 and LAMP1+ sub-compartments, performed with Imaris 9.8. Scale bar, 2 μm. (I) Immunostaining for LAMP1 on hiPSC-CMs at 10 hours post incubation with PKH67-labeled EV-AAV6. DAPI-blue; PKH67-labeled EV-AAV-green; LAMP1-red; cTNI-magenta; white arrows showed colocalization of PKH67 and LAMP1. Scale bar, 5 μm. (J) Immunostaining for intact AAV6 particles on hAC16 cells at 10 hours post incubation with PKH67-labeled EV-AAV. Scale bar, 2 μm. (K) 3D visualization showing the colocalization or delocalization of PKH67-labeled EV-AAV6 and AAV6, performed with an Imaris 9.8. White arrows point to released free AAVs labeled red. Scale bar, left, 5 μm; right, 2 μm. (L) Quantification of the PKH67 and AAV colocalization in (K): G; green, PKH67; R, red, AAV6; RG, red and green (n=5). (M) Flow cytometry analysis and quantification (N) showing PKH67 and CypHer-labeled EV-AAV6 uptake in hAC16-CMs at different time points (n=3). Scale bar, 20 μm. Data were analyzed with Welch’s one-way ANOVA followed by Dunnett-T3 multiple comparison test. (O) Flow cytometry analysis and quantification (P) showing PKH67 and CypHer-labeled EV-AAV6 uptake in hAC16-CMs at 10 hours after treatment with bafilomycin A1 (n=3). Scale bar, 20 μm. Data were analyzed with one-way ANOVA followed by Tukey multiple comparison test. (Q) Luciferase activity of hAC16-CMs transduced with EV-AAV6 after 48 hours. hAC16-CMs were treated with DMSO or Baf-1A for 2 hours prior to transduction with EV-AAV6 (n=4). Data are present as median ± interquartile range and were analyzed with Mann-Whitney U test. (B, C, E, L, N and P) are presented as mean ± SEM. *P<0.05, **P< 0.01, ***P < 0.001.

EV-AAV uptake involves trafficking via acidic compartments

To investigate the mechanisms of EV-AAV intracellular trafficking, we examined the uptake of PKH67-labeled EV-AAV6s to different cellular compartments of hAC16-CMs in vitro. Flow cytometry analysis of PKH67-EV-AAV6 at different time points showed partial internalization of EV-AAV6 by the treated cells at 1 hour and almost complete internalization at 4–24 hours (Figure 7D and E; Figure S21A). High-resolution confocal microscopy analysis detected PKH67-EV-AAV6 with markers of early endosomes (Rab5, modest presence at 4 hours), late endosomes (Rab7, at 10 hours), and lysosomes (LAMP1, at 10 hours), suggesting these cytoplasmic compartments had internalized EV-AAV (Figure 7F to H; Figure S21B and C). Consistent with hAC16-CMs uptake, hiPSC-CMs incubated with PKH67-EV-AA6 showed EV-AAVs colocalized with LAMP1+ lysosomes at 10 hours (Figure 7I).

To address whether EV-AAVs were intact (i.e. the EVs still enveloped the AAVs), we co-stained the PKH67-EV-AAV6s with anti-AAV6 antibody (that detects intact AAV6 particles) in hAC16-CMs and hiPSC-CMs. Interestingly, we observed partial co-localization of PKH67-EV-AAV6 with AAV6 in punctate structures within the cytoplasm at 4–10 hours, the point at which the PKH67-EV-AAVs co-localized with late endosome/lysosome (Figure 7J; Figure S22A). These data suggest that at 4–10 hours of uptake, EV-AAVs are in the process of releasing AAVs in both hAC16-CMs and hiPSC-CMs. Interestingly, intracellular trafficking of AAVs may follow a different timeline, as they are found in the nucleus in hiPSC-CMs at 10h (Figure S22B). Our 3D model reconstruction and co-localization analysis of the same images, obtained at 10 hours, confirmed that ~50% of the vesicles were still intact (stained with both PKH67 and AAV6) and ~50% were separated (stained with either PKH67 or AAV6) (Figure 7K and L). These results indicate that EV-AAVs might be internalized by late endosomes/lysosomes, possibly undergo acidification, and release AAVs before they enter the nucleus.

AAV acidification is an important step in their delivery to the nucleus. To investigate whether EV-AAVs were taken up by acidic sub-compartments, we labeled EV-AAV6s with both PHK67 dye, to track EV-AAV trafficking in the cytoplasm, and CypHer dye (a pH sensitive dye that is fluorescent at an acidic pH), to track EV-AAV uptake into acidic cytoplasmic vesicles, and incubated with hAC16-CMs for 1, 4, 10, and 24 hours before flow cytometry and imaging analysis. We observed that the percentage of PKH67+CypHer+ cells maximized at 10 hours (98.28 ± 0.12%), demonstrating that EV-AAVs sorted into sub-cellular compartments with acidic pH (Figure 7M and N; Figure S23). This finding is in line with our previous observation of EV-AAV uptake to LAMP1+ compartments at 10 hours. To confirm this, we pre-treated hAC16-CMs with bafilomycin A1 (Baf-A1), a V-ATPases inhibitor that blocks the acidification of endosomal/lysosomal compartments⁶⁹, followed by incubation with PKH67-CypHer-EV-AAV6s. We found that the majority of the PKH67-EV-AAV6s were still internalized (~98% PKH67+ cells at 10 hours) but not in the acidic compartments (~10% CypHer+ cells at 10 h; Figure 7O and P; Figure S23). Lack of acidification of endosomal/lysosomal compartments (Baf-A1 pre-treated cells), significantly decreased the transduction efficiency of EV-AAV (Figure 7Q). Together, these data suggest that EV-AAVs are internalized to the acidic compartments of endosomes/lysosomes, where the EV-AAVs might be released and AAVs may be acidified for further transport into the nucleus.

DISCUSSION

In this study, we have demonstrated the functional benefits and mechanisms of EV-AAVs, a superior vector that offers NAb resistance and delivers therapeutic genes to ischemic hearts. We optimized an EV-AAV purification strategy, comprehensively characterized EV-AAVs, and used five different in vitro and in vivo model systems to demonstrate significantly higher potency and therapeutic efficacy of EV-AAVs, and compared that with free-AAVs. We also have reported a mechanism of uptake and cardiotropism of EV-AAV vectors.

AAV-mediated transgene expression in the myocardium has low efficiency (0.5–2.6%).⁷⁰ Enhancing the transduction efficiency of viral genome delivered to the heart is significant for developing AAV-based cardiac gene therapy. Our study has uncovered that EV-AAVs consistently deliver more genes to cardiomyocytes compared with free AAVs both in vitro and in vivo, even in the absence of NAbs (although in some cases it was not statistically significant; Figure 3F and, Figure 4C, Figure S16C). Prior studies have shown that fetal bovine serum (FBS) can inhibit AAV6 transduction in cell culture.⁷¹ In our experiments, presence of 10% FBS in cell culture of hAC16-CMs significantly inhibited the AAV6 transduction, but not the EV-AAV6 transduction (Figure S24). Interestingly, pre-existing antibodies against the cardiotropic AAV serotypes AAV1, AAV6, and AAV9 and against AAV2 in mice from commercial vendors¹⁸, as well as murine⁷² and bovine AAVs⁷³ has also been reported before. It is possible that presence of pre-existing NAbs in mice and FBS in our experiments neutralizes free-AAVs to a certain extent, and benefits EV-AAVs. Interestingly, EV-AAVs demonstrate significantly higher transduction efficiency, compared with free-AAVs, when infected to hiPSC-CMs, which were cultured without FBS. Besides, in both hiPSCs-derived cells and mice, EV-AAVs expressed more transgenes in CMs compared to NMs, even with equivalent cellular uptake. Interestingly, higher transgene expression of EV-AAVs in hiPSC-CMs was consistent with significantly higher nuclear entry of vector genome, compared to NMs (Figure 6M). The cardiotropic expression of EV-AAVs could be due to one or more of the following: 1) because NMs, including fibroblasts and endothelial cells, proliferate at higher rates than CMs⁷⁴, the delivered gene can be lost during proliferation (the recombinant AAV genome rarely undergoes site-specific integration in host DNA and persists largely as an episome);^11–13,17 2) uptaken EV-AAVs have differential cytoplasmic processing⁷⁵ and endosomal/lysosomal degradation and release⁷⁶ between CMs and NMs (Figure 6L and M); 3) Intracellular trafficking, conformational changes of the AAV capsid in the cytoplasm, interaction between capsid and nuclear pore complexes and post-lysosomal nuclear entry of AAVs⁶³ and EV-AAVs could be different in different cell types. Deciphering the mechanism of EV-AAV uptake and transgene expression in CMs will provide new insights into augmenting AAV transduction therein.

A detailed picture of AAV cellular entry and uptake mechanisms are now emerging.^63,77 Different AAV serotypes may have similar or distinct uptake pathways, which may also be recipient cell-type dependent. AAV capsids bind to receptors on the host cell surface and are internalized via multiple endocytic routes, undergo conformational changes via endosome/Golgi retrograde transport or late endosome/lysosome pathway, followed by cytoplasmic escape and nuclear import.^63,64 The EV-AAV uptake mechanisms have not been studied yet. Our study revealed that infection of both AAV6/9 and EV-AAV6/9 was dependent on AAVR, but not on GRAF1-dependent CLIC/GEEC. This data suggests that AAVR could be a common receptor for AAVs and EV-AAVs and may have a broader function than previously known.

For successful transduction, AAV must escape from the sub-cellular compartments to the acidic pH in the endosome and lysosome prior to entering the nucleus. Like AAVs, EVs undergo endocytosis and release cargos in acidic compartments. Endosomal acidification has been shown to facilitate EV cargo function.⁷⁶ In line with other reports on EV uptake,^76,78 here we showed that EV-AAVs colocalize with bafilomycin-sensitive low-pH compartments in late endosomes and lysosomes, from which they release AAVs (Figure 7 and 8). This work reveals two important steps in the expression of genes carried via EV-AAVs: 1) delivering AAV cargo and 2) preparing AAV for its nuclear translocation. It is worth noting that AAV particle degradation before nuclear entry leads to low transduction. Interestingly, our data show EV-AAVs have significantly higher transduction efficiency than free-AAVs even without NAbs, thereby indicating that EV-AAVs might protect AAV from degradation before it enters the nucleus, as observed in EV-AAV membrane separation from the AAV cargo that it delivers. An endosomal pathway to the trans-Golgi network exists for AAV transduction;⁷⁹ however, whether EVs undergo retrograde trafficking to the TGN is poorly understood.⁸⁰ AAVR may play a role in cell-surface attachment of AAVs, trafficking AAV particles to TGN, or the escape of AAVs out of TGN.^67,68,81 The inhibition of EV-AAV transduction in AAVR-knockdown cells (Figure S20E–H) suggest that either the AAVR-mediated intracellular trafficking of AAVs (which are released from EV-AAVs) to TGN is impaired, or AAVR directly regulates EV-AAV uptake and intracellular processing independent of its regulation of free-AAV capsid trafficking, thereby affecting the internalization and transduction of EV-AAVs. Uncovering the underlying mechanism for EV-AAV trafficking requires deeper investigation. Distinct mechanisms between AAVs and EV-AAVs could have important therapeutic implications and need further investigation. Taken together, our findings on EV-AAV trafficking provide novel insights into EV-AAV content delivery to and function in recipient cardiomyocytes.

Figure 8. Schematic of EV-AAV uptake in cardiomyocytes.

Free AAVs bind with neutralizing antibodies that block AAV delivery into cells. EV-AAVs protect the AAV from this neutralizing effect to improve gene delivery to cardiomyocytes. In the cytoplasm, EV-AAVs are taken up into acidic sub-cellular compartments such as late endosomes and lysosomes, which may release the AAVs from EV-AAVs, thereby enabling nuclear entry and gene expression.

Our study demonstrates the NAb evasion and preclinical benefits of EV-AAV vectors for therapeutic gene delivery to the heart after myocardial ischemia and addresses knowledge gaps regarding their biology and cellular trafficking mechanisms. The benefits of EV-AAV-mediated gene delivery are multifold: 1) because their robustexosomalmembrane can protect AAVs from being accessed by neutralizing antibodies, EV-AAVshave higher resistance to NAb³³, and EV-AAV-mediated delivery will facilitate repeated dosing in patients. Moreover, since exosomes are structurally sturdier than other extracellular vesicles,^41,82 they may be good therapeutic vehicles; 2)extracellular vesicles and EV-AAVs can be rapidly taken up by target cells^83,84, such ascardiomyocytes and thus can efficiently deliver genes to the dynamic cardiac environment heart; 3)extracellular vesicles outperform synthetic nanovesiclesin stability, uptake, and gene delivery efficiency.⁸⁵

By deciphering the molecular uptake mechanisms ofEV-AAV,compartmentalized cellular signaling, and enhanced expression in cardiomyocytes, our findings may provide important insights into AAV-mediated gene delivery and cardioprotection.While our investigation established the proof-of-concept for successfully using EV-AAVs to deliver genes to the heart in the presence of NAbs, ultracentrifugation-based methods are time-consuming, inefficient, and result in low yield, making them unsuitable for the high-throughput production needed for clinical use. Therefore, newer methods need to be tested to efficiently separate AAV-contaminants from EV-AAVs to accurately evaluate their purity for futurepreclinical and clinical applications.

Our results offer a thorough overview of the gene delivery potential of EV-AAVs and their suitability as a vector for AAV delivery, in the presence of NAbs, to the myocardium. EVs, in general, do not exhibit cell and tissue tropism, yet show different rates of uptake by different cells.⁸⁰ Therefore, to improve their expression in the heart, we delivered EV-AAVs intramyocardially. Local delivery methods were preferred routes as well in earlier studies with EV-AAV-mediated delivery to other organs such as the retina and brain.^{27,28,32,33,86} Nonetheless, genes delivered to the heart via EV-AAVs, similar to those delivered via AAVs,are associated with off-target expression such as in the liver, even with local intramyocardial delivery. Intramyocardial delivery of EV-AAVs is an invasive procedure and could be a limitation to the EV-AAV-mediated gene delivery. Thus, a cardiomyocyte-targeting strategy for EV-AAVs using engineered AAVs⁸⁷ with improved potency for cardiac or muscle delivery⁶² may confine gene expression to the heart and may hold a key to their successful clinical translation. All previous studies, including ours, observed good cell tolerance of EV-AAVs with no side effects.Nevertheless, future investigationsinto their long-termefficiency, safety, dosing and biodistribution are necessary. There are a few functional studies on EV-mediated gene delivery to the brain, retina, and cancerous cells, but the uptake and delivery mechanisms of EV-AAVs had not been investigated prior to our study. Advancing EV-AAVs therapies to the clinic will require quantitative, mechanism-driven analyses of EV-mediated delivery.

Collectively, our work here confirms that EV-AAVs are fully infectious, resist antibody-mediated neutralization, and can efficiently transduce the myocardium. Our investigation provides advanced strategies to improve gene therapy by circumventingNAbneutralization, which has been a roadblock to prior efforts in the field. These findings open a new avenue for clinically translating AAV-based cardiac gene therapies to treat a broader population of patients with HF. In summary, our study will positively impact future preclinical studies and clinical trials in gene therapy and has the potential to advance current AAV-based therapeutic applications for treating cardiovascular diseases.

CLINICAL PERSPECTIVE.

What is New:

• Extracellular vesicle-encapsulated AAVs (EV-AAVs) shield AAVs from neutralizing antibodies (NAbs) and can overcome a key roadblock associated with AAV-mediated gene therapy.

• EV-AAVs are a superior cardiac gene delivery vector that offers higher NAb resistance and delivers more genes in presence of NAbs, compared to AAVs.

• EV-AAVs are cardiotropic and efficiently transduce genes to cardiomyocytes in the left ventricle for a long-term gene delivery to the heart.

What are the clinical Implications:

• SERCA2a gene delivered using EV-AAVs improves cardiac remodeling and function in mice with infarcted hearts even in presence of neutralizing antibodies.

• EV-AAVs improve existing cardiac gene therapy by circumventingNAbneutralization, which can expand the patient population receiving gene therapy by including NAb+ patients with heart failure and can enable redosing in already treated patients (with AAV gene therapy).

• EV-AAVs are clinically compatible, next generation cardiac delivery vector which can advance current AAV-based therapeutic applications for treating cardiovascular diseases.

Nonstandard Abbreviations and Acronyms

AAV

Adeno-associated virus

AAVR

AAV receptor

CMs

Cardiomyocytes

DGUC

Density gradient ultracentrifugation

DLS

Dynamic light scattering

EF

Ejection fraction

EV

Extracellular vesicle

EV-AAV

Extracellular vesicle-encapsulated AAV

FS

Fractional shortening

HF

Heart failure

hiPSC

Human induced pluripotent stem cell

IVIG

Intravenous immunoglobulins

LV

Left ventricle

MI

Myocardial infarction

LAD

Left anterior descending artery

NAb

Neutralizing antibody

NMs

Non-myocytes

SERCA2a

Sarcoplasmic reticulum calcium ATPase 2a

TEM

Transmission electron microscopy

TGN

Trans-Golgi network

TRPS

Tunable resistive pulse sensing

Vg

Vector genomes

WT

Wild type