Extracellular vesicles enhance pulmonary transduction of stably associated adeno‐associated virus following intratracheal administration

Abstract

Adeno‐associated virus (AAV) vector has shown multiple clinical breakthroughs, but its clinical implementation in inhaled gene therapy remains elusive due to difficulty in transducing lung airway cells. We demonstrate here AAV serotype 6 (AAV6) associated with extracellular vesicles (EVs) and secreted from vector‐producing HEK‐293 cells during vector preparation (EVAAV6) as a safe and highly efficacious gene delivery platform for inhaled gene therapy applications. Specifically, we discovered that EVAAV6 provided markedly enhanced reporter transgene expression in mucus‐covered air‐liquid interface (ALI) cultures of primary human bronchial and nasal epithelial cells as well as in mouse lung airways compared to standard preparations of AAV6 alone. Of note, AAV6 has been previously shown to outperform other clinically tested AAV serotypes, including those approved by the FDA for treating non‐lung diseases, in transducing ALI cultures of primary human airway cells. We provide compelling experimental evidence that the superior performance of EVAAV6 is attributed to the ability of EV to facilitate mucus penetration and cellular entry/transduction of AAV6. The tight and stable linkage between AAV6 and EVs appears essential to exploit the benefits of EVs given that a physical mixture of individually prepared EVs and AAV6 failed to mediate EV‐AAV6 interactions or to enhance gene transfer efficacy.

1. INTRODUCTION

Adeno‐associated virus (AAV) is the most widely explored virus‐based gene delivery platform experimentally validated for safety and/or therapeutic efficacy in treating various diseases affecting different target organs. Promising outcomes in preclinical studies have spawned numerous human clinical trials (High & Aubourg, 2011; Jacobson et al., 2015), and the FDA has recently approved AAV‐based gene therapy drugs for the treatments of rare eye diseases, spinal muscular atrophy, and hemophilia (Keeler & Flotte, 2019; Mendell et al., 2012; Mullard, 2023; Smalley, 2017). However, such clinical breakthroughs are yet to be realized for inhaled gene therapy of chronic lung diseases, due to the inability of clinically tested gene vectors to provide therapeutically relevant gene transfer efficacy in human lungs (Griesenbach et al., 2009; Guggino & Cebotaru, 2017; N. Kim et al., 2016). AAV serotype 2 (AAV2), the only AAV serotype tested in clinical trials of inhaled gene therapy to date, is inefficient in transducing airway cells in human lungs and is readily rejected by the host immune system upon initial and subsequent treatments (Guggino & Cebotaru, 2017; van Haasteren et al., 2018). Moreover, AAV2 and other clinically tested gene vectors cannot efficiently traverse the airway mucus (Boylan et al., 2011; Hida et al., 2011; Schuster et al., 2014; Suk et al., 2011) which obviates vector access to the underlying airway cells while promoting their removal from the lungs via physiological mucus clearance mechanisms (D. Chen et al., 2021). Airway mucus is now universally appreciated as one of the major hurdles that inhaled gene vectors must overcome to achieve clinically relevant gene delivery efficacy in the lung (Duncan et al., 2016).

Extracellular vesicles (EVs) are naturally occurring molecular carriers released from a variety of cells and play pivotal roles in intercellular communications (Maas et al., 2017). EVs have been intensively investigated over the past decade as a means to enhance therapeutic delivery, due to its inherent ability to shuttle various biological cargoes, including nucleic acids, proteins, and metabolites, through extracellular milieus and between cells (H. Kim et al., 2020; Nam et al., 2020). To our interest, EVs are found in human airway mucus and mediate crosstalk between lung‐resident parenchymal cells and/or immune cells (e.g., macrophages) (Kulshreshtha et al., 2013; Lee et al., 2016; McVey et al., 2019). Moreover, accumulated evidence has revealed that mucosal EVs play key pathological roles in various chronic respiratory diseases, such as cystic fibrosis (CF), chronic obstructive pulmonary disease and asthma (Genschmer et al., 2019; Lacedonia et al., 2016; Porro et al., 2013; Sanchez‐Vidaurre et al., 2017). These observations collectively suggest that EVs likely travel within the lung airway mucus in a relatively unhindered fashion. We thus hypothesized that EVs would facilitate the penetration of AAV vectors, if stably associated with EVs, through the airway mucus, thereby enhancing the transduction of lung airway cells following inhaled administration.

Unlike small molecule‐based drugs and short nucleic acids (e.g., siRNA), attempts to load and deliver therapeutic genes using EVs have been challenging due to their large cargo size (Lamichhane et al., 2015). On the other hand, cells transfected by AAV‐producing genes for vector preparation were shown to naturally release EVs harbouring DNA‐loaded AAVs via internal and/or external association (Maguire et al., 2012). The discovery was followed by studies demonstrating the enhanced ability of EV‐associated AAV (EVAAV) to transduce extrapulmonary cells in vitro and in vivo compared to standard preparations of the corresponding AAV alone (i.e., EV‐free AAV) (Gyorgy & Maguire, 2018; Gyorgy et al., 2014, 2017; Hudry et al., 2016; Wassmer et al., 2017). To this end, we investigated here the ability of EVAAV to overcome biological barriers present in the lung and transduce mucus‐covered airway cells in human primary air‐liquid interface (ALI) cultures and in mouse lungs in comparison to AAV. AAVs externally associated to EVs, if mucoadhesive, may restrict the motion of EVAAV in airway mucus. Thus, we prepared and explored EVAAV based on AAV6 (EVAAV6), leveraging on its superior ability to resist adhesive interactions with airway mucus (Duncan et al., 2018; Hida et al., 2011; Schuster et al., 2014) and to mediate airway cell transduction in vitro and in vivo (Kurosaki et al., 2017; Li et al., 2009, 2011; Limberis et al., 2009) compared to AAV2 and other widely studied AAV serotypes.

2. MATERIALS AND METHODS

2.1. Cell culture

Human embryonic kidney (HEK) 293 cells were obtained from the American Type Culture Collection (ATCC, Rockville, MD, USA). Human bronchial epithelial (HBE) cell line (16HBE14o‐) was a kind gift from Dr. Garry Cutting at Johns Hopkins University School of Medicine (Baltimore, MD, USA). These cells were cultured in Minimum Essential Medium (MEM; Thermo Fisher Scientific, Waltham, MA, USA) supplemented with 10% fetal bovine serum (FBS; Thermo Fisher Scientific) and 1% penicillin‐streptomycin (Thermo Fisher Scientific). Of note, 16HBE14o‐ cells were cultured in T75 flasks or 96‐well plates coated with 100 μg/mL bovine serum albumin (Millipore Sigma, Burlington, MA, USA), 10 μg/mL human fibronectin (Millipore Sigma), and 30 μg/mL collagen 1 rat tail (Thermo Fisher Scientific). The cells were passaged using 0.5% trypsin‐EDTA (Thermo Fisher Scientific) at 80% confluency.

2.2. Production of AAV6, EVs, and EVAAV6

Recombinant AAV6 vectors were prepared to carry a dual reporter cassette as previously described (Duncan et al., 2018). Briefly, HEK 293 cells were transfected with three AAV6‐producing plasmids, including AAV Cap‐Rep plasmid, adenoviral gene‐based helper plasmid, and internal terminal repeat‐flanked luciferase‐YFP dual reporter plasmid. We plated 1 × 10⁷ HEK 293 cells on a 150‐mm petri dish and incubated the cells for 24 h. Cells were then washed with serum‐free DMEM and replenished with 25 mL of DMEM supplemented with 5% FBS. In parallel, 60 μg of AAV6‐producing plasmids at an equimolar ratio of three plasmids were mixed with 125 μg of 10 kDa linear polyethyleneimine (PEI) in 1 mL of serum‐free DMEM. After a 5‐min incubation, the cells were treated with the plasmids compacted by PEI, harvested at 72‐h post‐transfection, and then lysed by with lysis buffer (150 mM NaCl, 50 mM Tris, pH 8.5) and two freeze‐and‐thaw cycles. The cell lysate was centrifuged in an iodixanol step gradient column (OptiPrep, Millipore Sigma) formed with fractions ranging 15%–60% iodixanol. The 40% iodixanol fraction was then further purified by ion‐exchange chromatography using HiTrap Q HP (Cytiva, Piscataway, NJ) and concentrated with spin concentrators (150 kDa MWCO; Orbital Biosciences, Topsfield, MA).

For the preparation of EVAAV6, HEK 293 cells were transfected with AAV‐producing plasmids using an identical manner employed for the AAV6 production except that cells were replenished with DMEM supplemented with 5% EV‐depleted FBS at 24‐h post‐transfection. Of note, EV‐depleted FBS was prepared by ultracentrifugation of FBS at 100,000 × g for 2 h at 4°C and subsequent collection of the supernatant. At 72‐h post‐transfection (i.e., 48 h after replenishing the media), the conditioned media was collected and centrifuged at 1000 × g for 15 min at 4°C, and then the supernatant was collected and centrifuged at 100,000 × g for 2 h at 4°C. The supernatant was discarded and the EVAAV6 pellet was resuspended in 500 μL sterile PBS. Subsequently, the resuspended EVAAV6 sample was centrifuged at 10,000 × g for 15 min and filtered through a 0.45‐μm centrifuge filter (Spin‐X, Thermo Fisher Scientific). Preparations of AAV‐free EVs were obtained by the identical procedure without the transfection of HEK 293 cells with AAV6‐producing plasmids.

AAV6 titers in the final AAV6 and EVAAV6 products were determined by quantitative polymerase chain reaction (qPCR) with SYBR Green qPCR Master Mix (Thermo Fisher Scientific) and the primer pair specific for the chicken β‐actin promoter region of expression cassette: forward 5′‐TCCCATAGTAACGCCAATAGG‐3′, reverse 5′‐CTTGGCATATGATACACTTGATG‐3′. AAV6 and EVAAV6 were then diluted with PBS to predetermined concentrations for their uses in subsequent studies.

2.3. Physicochemical characterization of AAV6, EVs, and EVAAV6

AAV6, EVs, and EVAAV6 equivalent to a 1‐μg protein content were diluted in 100 μL PBS at pH 7. EVAAV6 was subjected to probe sonication or nebulization to assess the stability of EV‐AAV association under physical agitation. Briefly, EVAAV6 was sonicated by VCX‐500 ultrasonic processor (Sonics & Materials, Newtown, CT, USA) at 20% power and six cycles by 4‐second pulse/2‐s pause for 2 min or was nebulized by a vibrating mesh nebulizer (Aerogen, Galway, Ireland). Hydrodynamic diameters were measured by dynamic light scattering (DLS) using a Zetasizer Nano ZS (Malvern Instruments, Malvern, UK). The morphology of AAV6, EVs, EVAAV6, and a physical mixture of individually prepared EVs and AAV6 (EV+AAV6) were visualized by transmission electron microscopy (TEM). Briefly, AAV6, EVs, EVAAV6, or EV+AAV6 equivalent to 1‐μg protein content were loaded on a carbon type‐B copper grid (Ted Pella Inc, Redding, CA, USA) and air‐dried for 3 min. After rinsing with deionized water for 1 min, the samples were stained with UranylLess EM STAIN (Electron Microscopy Sciences, Hatfield, PA, USA) for 1 min. The grid was then washed once with deionized water, dried overnight, and analyzed with Hitachi H7600 TEM (Hitachi High‑Technologies, Tokyo, Japan). To quantify the number and size distribution of EVs in the final AAV6‐free EV and EVAAV6 samples, the samples were diluted 10–100‐fold with PBS supplemented with 1% Tween‐20 (Millipore Sigma) and analyzed by microfluidic resistive pulse sensing (MRPS) technology using an nCS1™ nanoparticle analyzer (Spectrodyne, Signal Hill, CA, USA) and the C400 cartridge covering a size range of 65–400 nm. The measured concentrations of EV particles in AAV6‐free EV and EVAAV6 samples were then used to prepare dose‐matched EV+AAV6 physical mixture controls for individual experiments.

2.4. Molecular analysis of AAV6, EVs, and EVAAV6

To conduct western blot analysis, AAV6, EV, and EVAAV6 samples were lysed with RIPA Lysis Buffer, 10X (Millipore Sigma) and the lysates were centrifuged at 12,000 × g for 20 min at 4°C to collect supernatants. The total protein content of the supernatant was quantified by Pierce™ BCA Protein Assay Kit (Thermo Fisher Scientific), and the samples comprising 1‐mg of protein content were mixed with 5X sodium dodecyl sulfate (SDS) gel‐loading buffer and annealed at 95°C for 5 min. The samples comprising 10‐μg of protein content were subjected to electrophoresis with 10% ‐ 20% SDS‐polyacrylamide gel (Thermo Fisher Scientific) and transferred onto a polyvinylidene fluoride PVDF membrane (Thermo Fisher Scientific). The membranes were then blocked at room temperature in TBS‐T solution (pH 7.4, 20 mM Tris, 150 mM NaCl, and 0.05% Tween 20) supplemented with 1% BSA. One hour later, the membranes were incubated with monoclonal antibodies against AAV capsid proteins (Cat no. 03–61058; Arp Inc, Waltham, MA, USA), human syntenin (Cat no. ab133267; Abcam, Cambridge, UK), and human CD9 (Cat no. 312102) (BioLegend, San Diego, CA, USA) at a 1:1000 dilution (v/v) for 24 h at 4°C. The membranes were then washed three times and incubated with secondary anti‐mouse or rabbit IgG antibody conjugated with HRP (Cat no. 31430 and 31460 for mouse and rabbit, respectively; Thermo Fisher Scientific) for an hour at room temperature. Subsequently, the membranes were washed three times, and protein bands were detected with the SuperSignal™ West Femto Maximum Sensitivity Substrate (Thermo Fisher Scientific) by Chemi‐Doc imaging system (Bio‐RAD, Hercules, CA, USA).

We next conducted single particle interferometric reflectance imaging (SP‐IRIS) for tetraspanin/syntenin‐1 phenotyping of EVs and EVAAV6. A 35 μL aliquot of the final EV or EVAAV6 samples was diluted in incubation solution II at a 1:1 dilution (v/v) and incubated at room temperature on ExoView Human Tetraspanin chips (Unchained Labs, Pleasanton, CA) printed with anti‐human CD81, anti‐human CD63, anti‐human CD9, and anti‐mouse IgG1. After a 16‐hour incubation, chips were washed with 1X Solution A I (Unchained Labs) four times for 3 min each and further incubated with a fluorescently labeled antibody cocktail of anti‐human CD81, anti‐human CD63, anti‐human CD9, and anti‐human syntenin‐1 monoclonal antibodies at a dilution of 1:1200 (v/v) in a 1:1 (v/v) mixture of 1X Solution A I and Blocking Solution II (Unchained Labs). Chips were then washed once with 1X Solution A I, 3 times with 1X Solution B I (Unchained Labs), and once with water (3 min each at 460 rpm). Chips were immersed twice in water for approximately 10 s each and tilted at a 45° angle to allow the liquid to vacate the chip. All chips were imaged in the ExoView scanner (Unchained Labs) by interferometric reflectance imaging and fluorescent detection. Data were analyzed using ExoView Analyzer 3.1 Software (Unchained Labs)

2.5. Multiple particle tracking analysis: diffusion of EVs and EVAAV6 in human airway mucus

For microscopic visualization, the vesicular membranes of EVs and EVAAV6 were fluorescently labeled using the PKH26 Red Fluorescent Cell Linker Kit (Millipore Sigma). Briefly, 100 μL of the final EV and EVAAV6 samples equivalent to 1 × 10¹⁰ EV counts were mixed with 100 μL of PKH26 dye solution containing 8 μM dye molecules and incubated at room temperature for 5 min. To stop the reaction, 100 μL of 1% BSA solution was added to the mixture and incubated for 1 min. The mixture was then placed on top of 2 mL 20% sucrose solution (Millipore Sigma) and centrifuged at 100,000 × g for 2 h at 4°C. The pellet was gently resuspended in 100 μL DPBS and washed with an Amicon® Ultra‐4 Centrifugal Filter Unit (100 kDa MWCO; Millipore Sigma). To conduct multiple particle tracking (MPT) analysis, 1 μL of the labeled EVs or EVAAV6 equivalent to 1 × 10⁹ of EV counts was added to and gently mixed with 30 μL human airway mucus samples spontaneously expectorated by CF patients (i.e., sputum) visiting the Johns Hopkins Adult CF Clinic. The motions of fluorescently labeled EV and EVAAV6 in CF sputum were recorded with fluorescent video microscopy (Axiovert, Carl Zeiss, Stuttgart, Germany) at a frame rate of 15 frames per second (i.e., 67 milliseconds per frame). The movies were then analyzed using software a custom‐written in MATLAB (MathWorks, Natick, MA, USA) to the extract mean square displacement (MSD) values.

2.6. In vitro transduction in mucus‐free human bronchial epithelial cells

16HBE41o‐ cells were plated in each well of 96‐well plates at a density of 1 × 10⁴ cells per well with 100 μL of the culture medium and incubated for 24 h. The cells were treated with AAV6, EVAAV6, or EV+AAV6 at multiplicity of infection (MOI) of 100, 1000, or 10,000. EV+AAV6 was prepared to comprise equivalent amounts of individual components as in EVAAV6 to conduct this and subsequent studies. Seventy‐two hours after the treatment, cells were lysed using reporter lysis buffer (Promega, Madison, WI, USA) and three freeze‐and‐thaw cycles, followed by the centrifugation of the cell lysate at 12,000 × g for 20 min at 4°C. The supernatant was subjected to cell homogenate‐based luciferase assay to record the luciferase activity in the relative light unit (RLU) using Luciferase Assay System (Promega) and a 20/20n luminometer (Turner Biosystems, Sunnyvale, CA). In parallel, the protein concentration in the supernatant was quantified by Pierce™ BCA Protein Assay Kit (Thermo Fisher Scientific). The RLU values were normalized by the total protein content.

2.7. In vivo transduction and safety in murine lungs

C57BL/6 mice (male, 5 weeks) were purchased from Charles River Laboratories (Wilmington, MA, USA). Animals were handled in accordance with the guidelines and policies of Johns Hopkins University Animal Care and Use Committee (animal protocol number: MO21M185).

We treated C57BL/6 mice with 40‐μL saline or suspension of AAV6, EVAAV6, or EV+AAV6 at an equivalent AAV6 dose of 4 × 10⁹ vg per mouse via intratracheal administration. Two weeks after the administration, the lungs were perfused with PBS via cardiac puncture and harvested. The lung tissues were frozen in optimal cutting temperature compound (Thermo Fisher Scientific) and sectioned at a 10‐μm thickness using Leica Cryostat (Leica, Wetzlar, Germany). The lung tissue slices were counterstained with ProLong™ Diamond Antifade Mountant with DAPI (Thermo Fisher Scientific) and observed with a confocal laser microscope (Zeiss LSM 710, Carl Zeiss). We conducted image‐based quantification of the coverage and intensity of YFP transgene expression in a blinded manner. The percentage of YFP coverage was defined as the YFP‐positive area divided by the total imaged area, and the YFP intensity was calculated by averaging the pixel fluorescence intensity, in each randomly selected image field. In parallel, lung tissue slices were stained with haematoxylin and eosin (H&E) and histopathologically scored by a board‐certified pathologist (K.M., D.V.M.) in a blinded manner for the assessment of local safety profiles of individual treatments. The scoring was conducted to detect the degree (i.e., percentage) of histological alteration in the lung fields (200×; 0: no detection, 1: <5%, 2: 5%–33%, 3: 33%–66%, 4: >66%).

Using a separate set of animals, we conducted a complementary assessment of in vivo reporter transgene expression based on lung homogenate‐based luciferase assay. C57BL/6 mice were treated as described above and PBS‐perfused lung tissues were harvested at 2‐week post‐administration. To determine the overall level of luciferase expression, the tissues were immersed in reporter lysis buffer and homogenized with a bead‐based TissueLyser (Qiagen, Hilden, Germany) at a 50/s oscillation rate for 30 min at 4°C. The lung tissues were then subjected to three freeze‐and‐thaw cycles and centrifuged at 12,000 × g for 20 min at 4°C. The RLU values and protein concentration of the supernatant were measured as described above, and the overall levels of luciferase transgene expression were compared as RLU values normalized by the protein content. Of note, we conducted this study with two independent EVAAV6 preparations to determine the batch‐to‐batch variation.

2.8. In vitro transduction in mucus‐covered primary airway‐liquid interface cultures

University of Alabama at Birmingham Institutional Review Board (IRB) approved the use of primary wild‐type (WT) and CF HBE cells harvested from a healthy individual and a CF patient harboring F508del homozygous genotype, respectively (Adewale et al., 2020). Johns Hopkins University IRB approved the use of primary human nasal epithelial (HNE) cells. To established ALI cultures, primary WT/CF HBE or HNE cells were conditionally reprogrammed using co‐culture with irradiated fibroblasts in a medium containing a rho‐associated protein kinase inhibitor‐reagent Y (StemCell Technologies, Vancouver, Canada). Primary WT/CF HBE cells were grown in 0.33 cm² polyester filter Transwell® inserts (Millipore Sigma), placed in a 24‐well plate, coated with NIH 3T3 fibroblast conditioned medium at 100% confluency and maintained in PneumaCult‐Ex‐Plus medium (StemCell Technologies) for at least 4–6 weeks to be fully differentiated. The primary HNE cells were transferred to 1.12 cm² Snapwell™ inserts (Corning, Glendale, AZ, USA), placed in a 6‐well plate, and grown in conditionally reprogrammed expansion medium at 100% confluency and maintained in PneumaCult‐Ex‐Plus medium for at least 3–4 weeks to be fully differentiated. The primary ALI cultures of mucus‐covered, polarized airway epithelium were maintained in PneumaCult‐ALI medium (StemCell Technologies) until use. A 10 or 34 μL suspension of AAV6, EVAAV6, or EV+AAV6 at an equivalent AAV6 dose of 1 × 10⁹ or 3.4 × 10⁹ vg was administered to the apical surface of the ALI cultures of primary WT/CF HBE or HNE cells, respectively. Two weeks after the administration, the filters were cut out, placed onto a glass slide, and counterstained with DAPI. Five randomly selected image fields were captured from each insert by confocal microscopy to be quantitatively analyzed. YFP coverage and intensity were quantified in a blinded manner as described above.

2.9. Statistical analysis

Statistical analyses were conducted using GraphPad Software (GraphPad Software Inc., La Jolla, CA). Two‐group comparisons were made using a two‐tailed Student's t‐test assuming unequal variances. If multiple‐group (>2 groups) comparisons were involved, one‐way analysis of variance (ANOVA), followed by an appropriate post‐hoc analysis, was employed. Differences were determined to be statistically significant at p < 0.05.

3. RESULTS

3.1. EVAAV6 exhibits stable internal and/or external association of AAV6 with EVs

We first conducted western blot analysis to confirm the presence of characteristic proteins in AAV6, EVs, and EVAAV6 produced in HEK293 cells. The analysis disclosed clear bands that corresponded to viral capsid proteins (VP1, VP2 and VP3), EV markers, and combination of both for AAV6, EVs, and EVAAV6, respectively (Figure 1a). Specifically, AAV6 and EVAAV6 comprised VP1, VP2, and VP3 at an expected ratio of 1:1:10 (Rose et al., 1971), and EV and EVAAV6 were shown to contain tetraspanin CD9 and syntenin‐1, which together indicate the presence of EVs as well as a subpopulation of exosomes (Kowal et al., 2016). Given the same cellular origin, we expected that EV phenotypes in the EVs and EVAAV6 would be comparable if not identical. To validate this hypothesis, we conducted SP‐IRIS analysis for tetraspanin/syntenin‐1 phenotyping where EVs and EVAAV6 captured by different tetraspanin‐specific antibodies against CD9, CD63, or CD81 exhibited similar levels and expression patterns for these tetraspanins and syntenin‐1 (Figure 1b).

FIGURE 1.

EVAAV6 prepared in HEK293 cells and harvested from the cell culture supernatant exhibits internal and/or external association of AAV6 with EV. (a) Western blot analysis of AAV6, EVs, and EVAAV6. Bands for VP1, 2 and 3 of AAV6 capsid and for syntenin‐1 (Syn‐1) and CD9 of EVs are shown. (b) ExoView analysis demonstrating that exosomes are a primary population similarly in EVs and EVAAV6. (c) Hydrodynamic diameters of AAV6, EVs, and EVAAV6 measured by DLS. (d) Representative transmission electron micrographs of AAV6, EVs, EVAAV6 and EV+AAV6. Scale bar = 50 nm.

The hydrodynamic diameters of AAV6, EVs, and EVAAV6 were measured by DLS to be 22.3 ± 2.3, 115.1 ± 5.2, and 136.7 ± 6.1 nm, respectively (Figure 1c). In parallel, the MRPS‐based nanoparticle size analysis revealed that while the average particle diameters of EV and EVAAV6 were comparable (∼ 65 nm), the sizes distributed up to 160 and 200 nm for the former and the latter, respectively (Supporting Figure 1), corresponding to the shift in size distributions measured by DLS (Figure 1c). Collectively, the size analyses indicated the presence of externally associated AAV6 on EVAAV6, in agreement with previous observations (Maguire et al., 2012). We note that probe sonication at a condition enabling intravesicular loading of a large protein cargo (i.e., catalase tetramer; 240 kDa) (Haney et al., 2015) and 40–45 nm nanoparticles (Sancho‐Albero et al., 2019) did not perturb the hydrodynamic diameters of EVAAV6 (139.2 ± 4.8 nm; Supporting Figure 2 and Supporting Table 1). Likewise, nebulization with a vibrating mesh nebulization did not significantly change the hydrodynamic diameters of EVAAV6 (142.3 ± 5.4 nm; Supporting Figure 2 and Supporting Table 1). We next conducted TEM of AAV6, EVs, EVAAV6, and a physical mixture of individually prepared EVs and AAV6 (i.e., EV+AAV6) to visually confirm EV‐AAV association. We observed that EVAAV6 exhibited internal and/or external association of AAV6 with EVs, whereas EV+AAV6 showed no direct association between the EVs and AAV6 (Figure 1d).

3.2. EVs and EVAAV6 diffuse in human pathological airway mucus in a relatively unhindered manner

We next tested our hypothesis that EVs would efficiently penetrate human airway mucus and AAV6 association would not compromise the mucus‐penetrating properties of EVs. Specifically, PKH26‐labeled EVs and EVAAV6 were added to sputum samples collected from three independent CF patients by minimally diluting the samples (∼3% v/v), and MPT analysis was conducted to quantify their MSD values. Of note, MSD is the square of distance traveled by an individual particulate matter at a given time interval (i.e., time scale) and thus is directly proportional to the particle diffusion rate (Duncan et al., 2018). We found that the median MSD values of EVs and EVAAV6 in these sputum samples were virtually identical and were over three‐fold greater than those of the mucus‐permeable AAV6 (Figure 2a) that we previously measured in 8 independent sputum samples (Duncan et al., 2018).

FIGURE 2.

EVAAV6 exhibits efficient penetration through human airway mucus and enhanced transduction of human bronchial epithelial (HBE) cell line. (a) Median MSD values of EVs and EVAAV6 in sputum samples spontaneously expectorated by CF patients. MSD is a square of distance traveled by an individual particulate matter within a predetermined time interval (i.e., time scale; τ = 1 s) and thus is directly proportional to the particle diffusion rate. The red dashed line indicates the MSD value of AAV6 previously measured in CF sputum (Gyorgy et al., 2014). respectively. n.s.: no significance (two‐tailed Student's t‐test) (b) Luciferase activity measured in lysates of HBE cells (16HBE14o‐) treated with EVs, EVAAV6 or EV+AAV6. n.s.: no significance, ****p < 0.0001 (one‐way ANOVA).

3.3. EVAAV6, but not EV+AAV6, mediates markedly greater reporter transgene expression compared to AAV6 in mucus‐free human bronchial epithelial cells

We compared in vitro transduction efficacy of AAV6, EVAAV6, and EV+AAV6 in conventional mucus‐free cultures of HBE cell line (i.e., 16HBE14o‐ cells) over a range of MOI values. Of note, the EV+AAV6 physical mixture was prepared to comprise the equivalent viral load and EV count as EVAAV6 to conduct this and subsequent studies. We assessed the luciferase activity as a measure of reporter transgene expression 72 h after the treatment where EVAAV6 exhibited more than an order of magnitude greater luciferase transgene expression compared to AAV6 regardless of MOI (Figure 2b). In contrast, the physical mixing the EVs and AAV6 failed to enhance the transduction efficacy as indicated by the virtually identical levels of luciferase activity observed with AAV6 and EV+AAV6 at all MOI values (Figure 2b)

3.4. EVAAV6, but not EV+AAV6, provided enhanced transduction compared to AAV6 in mouse lungs following intratracheal administration while exhibiting excellent local safety profile

We next went on to test whether enhanced mucus penetration and cellular transduction observed with EVAAV6 resulted in greater reporter transgene expression in mouse lungs in vivo following localized administration. Specifically, C57BL/6 mice were treated with saline or either of AAV6, EVAAV6, or EV+AAV6 at an equivalent AAV6 dose of 4 × 10⁹ vg per mouse (i.e., 40 μL of 1 × 10¹¹ vg/mL AAV6) via intratracheal administration and the whole lung tissues were harvested 2‐week post‐administration for the assessment of reporter transgene expression. Of note, a YFP/luciferase dual reporter cassette was packaged into the AAV6 virions for comprehensive and complementary assessment of reporter transgene expression. The representative whole lung images showed that all treatment groups inclusive of AAV6 mediated uniformed YFP transgene expression, but EVAAV6 provided significantly greater fluorescence intensity compared to the other groups (Figure 3a‐c). Quantitatively, the mean YFP intensity of EVAAV6 group was more than double compared to either AAV6 or the EV+AAV6 physical mixture, which exhibited similar levels of YFP intensity (Figure 3c). Likewise, tissue homogenate‐based luciferase assay revealed that EVAAV6, but not EV+AAV6, provided significantly greater luciferase activity in the lung lysates compared to AAV6 (p < 0.01; Figures 3d). We repeated this experiment with an independently prepared EVAAV6 batch. In accordance with the outcome of the first study (Figure 3d), we found that EVAAV6, but not EV+AAV6, exhibited significantly greater luciferase activity compared to AAV6 (p < 0.01; Figure 3e). Of note, the EV concentrations in the first and second EVAAV6 batches were 1.43 × 10¹¹ and 1.85 × 10¹² EV particles/mL, respectively, and EV+AAV6 physical mixture in each study was prepared to match the EV concentration in the respective EVAAV6. We note that the difference in the luciferase activity of EVAAV6 in these two studies was not statistically significant despite the approximately 13‐fold difference in the EV concentration (Supporting Figure 2).

FIGURE 3.

EVAAV6 provides widespread and enhanced reporter transgene expression in mouse lungs following intratracheal administration. (a) Representative confocal images demonstrating reporter transgene expression throughout the whole left lung lobes of mice intratracheally treated with normal saline, AAV6, EVAAV6 or EV+AAV6. Green: YFP, Blue: nuclei. Image‐based quantification of (b) coverage and (c) intensity of YFP transgene expression in whole lung lobes treated with normal saline, AAV6, EVAAV6 or EV+AAV6 (n = 5 mice per group). (d and e) Luciferase activity measured in lysates of the whole mouse lungs treated with normal saline, AAV6, EVAAV6 or EV+AAV6. Mice treated with EVAAV6 or EV+AAV6 with (d) lower and (e) higher EV content (n = 5 mice per group). n.s.: no significance, **p < 0.01, ***p < 0.001, ****p < 0.0001 (one‐way ANOVA).

For a local safety assessment, we harvested lungs of mice intratracheally received saline or either of AAV6, EVAAV6, or EV+AAV6 at an equivalent AAV6 dose of 4 × 10⁹ vg per mouse at 2‐week post‐administration. The lung tissues were fixed and stained with H&E, followed by a blinded scoring by a board‐certified pathologist (K.M., D.V.M.). Representative histological images of lung tissues harvested from animals received any treatment inclusive of AAV6 were virtually identical to the saline‐treated lungs (Figure 4a). The histopathological scoring revealed no significant inflammation and structural alteration featured with mononuclear infiltrates, presence of neutrophils, and necrosis in the lungs of animals treated with AAV6, EVAAV6, or EV+AAV6, similar to the saline‐treated lungs (Figure 4b).

FIGURE 4.

EVAAV6 does not induce local toxicity following intratracheal administration. (a) Representative histological images and (b) a blinded histopathological scoring (n ≥ 3 mice per group) of H&E‐stained lung tissues from mice treated with normal saline, AAV6, EVAAV6, or EV+AAV6.

3.5. EVAAV6, but not EV+AAV6, mediated markedly greater reporter transgene expression compared to AAV6 in mucus‐covered ALI cultures of primary HBE or HNE cells

To establish clinical relevance of EVAAV6 for inhaled gene therapy, we next assessed the ability of EVAAV6, in comparison to AAV6 and EV+AAV6, to transduce the mucus‐covered ALI cultures of primary HBE cells from a healthy individual or a CF patient harboring F508del homozygous genotype. The model uniquely emulates the physiological human lung airways and associated biological barriers (i.e., mucus), thereby serving as an excellent testbed for predicting the performance of gene delivery platforms administered via the inhaled route (Keegan & Brewington, 2021; Zaidman et al., 2016). Representative confocal images showed that EVAAV6 mediated markedly more uniform and greater YFP reporter transgene expression compared to other treatment groups in primary WT and CF HBE ALI cultures (Figure 5a). Image‐based quantitative analysis revealed significantly greater coverage and level (i.e., YFP intensity) of transgene expression achieved by EVAAV6 over AAV6 and EV+AAV6 regardless of the disease state of donor cells (Figure 5b and c). In agreement with our in vivo observation (Figure 3), EV+AAV6 failed to provide enhanced transgene expression compared to AAV6. We note that the coverage and level of transgene expression mediated by EVAAV6 were markedly greater in primary ALI cultures of WT HBE cells than of CF HBE cells characterized by mucus dehydration and thickening (Adewale et al., 2020). In a similar manner, EVAAV6, but not EV+AAV6, significantly increased the coverage and level of transgene expression compared to AAV6 in primary HNE ALI cultures (Supporting Figure 3).

FIGURE 5.

EVAAV6 provides enhanced reporter transgene expression in mucus‐secreting ALI cultures of primary wild‐type (WT) or cystic fibrosis (CF) HBE cells following apical administration. (a) Representative confocal images demonstrating reporter transgene expression in ALI cultures of primary WT or CF HBE cells treated with normal saline, AAV6, EVAAV6, or EV+AAV6. Green: YFP, Blue: nuclei. Image‐based quantification of (b) coverage and (c) intensity of YFP transgene expression in the ALI cultures of primary WT (black) and CF (red) HBE cells treated with normal saline, AAV6, EVAAV6, or EV+AAV6. n.s.: no significance, ***p < 0.001, ****p < 0.0001 (two‐way ANOVA).

4. DISCUSSION

In this study, we demonstrate that our AAV‐piggybacking‐on‐EVs approach provides markedly enhanced lung airway cell transduction by AAV6 in vitro and in vivo due to the ability of EVs to carry AAV6 efficiently through biological barriers in the lung encountered following inhaled administration. We found that a physical mixture of individually prepared EVs and AAV6 (i.e., EV+AAV6) failed to establish direct EV‐AAV association or to enhance gene transfer efficacy. We also note that EV phenotypes of EVAAV6 and the EVs used to prepare the physical mixture of EVs and AAV6 were virtually identical as revealed by the SP‐IRIS analysis. These findings underscore that the association between EVs and AAV6 naturally formed during the standard vector production process or an equivalent is likely required to exploit the benefits of EVs on AAV6 gene transfer in the lung. The membrane‐associated accessory protein (MAAP) is expressed from a frameshifted open reading frame in the AAV cap gene (Ogden et al., 2019) and has been recently identified as a novel viral egress factor that appears to hijack EV secretory pathway (Elmore et al., 2021). Thus, MAAP may have contributed to the stable EV‐AAV association, resistant to physical agitation, observed in this study, which warrants further investigation. EVAAV based on AAV8 or AAV9 has been shown to evade pre‐existing anti‐AAV humoral immunity following systemic administration (Gyorgy et al., 2014; Meliani et al., 2017), and thus EVs are expected to serve as an immunological Trojan horse for the associated AAV6. This is an important translational aspect since significant fractions of healthy and CF populations are seropositive to various AAV serotypes (Boutin et al., 2010; Halbert et al., 2006) and repeated treatment is likely needed for lifelong therapeutic effects unless gene editing is implemented to progenitor cells. Further, EVAAV enables simultaneous delivery of multiple AAVs carrying two complementary halves of a large gene payload beyond the AAV capacity of 4.7 kb (Tornabene & Trapani, 2020) into individual target cells for reconstitution. Collectively, EVAAV6 holds tremendous potential as a gene delivery platform capable of addressing multiple challenges to the AAV‐based inhaled gene therapy.

We discovered that EVs efficiently penetrated human airway mucus collected from CF patients (i.e., sputum) and that such mucus‐penetrating property of EVs was not compromised by the externally associated AAV6. The diffusion behaviour of EV in human airway mucus has not been investigated thus far but the finding is in accordance with a previous study demonstrating that milk exosomes penetrate a layer of native porcine intestinal mucus with significantly greater diffusion rates compared to small molecules (i.e., FITC) (Warren et al., 2021). Encouragingly, diffusion rates of EVAAV6, despite its markedly larger size, were greater than those of standard preparations of AAV6 alone which was shown to possess superior ability to resist mucoadhesion compared to other AAV serotypes, including AAV1, 2 and 5 (Duncan et al., 2018; Hida et al., 2011; Schuster et al., 2014). EVAAV6 mediated markedly enhanced transduction compared to AAV6 in mouse lungs that naturally secrete mucus and in mucus‐covered ALI cultures of primary HBE cells irrespectively of the disease state. On the other hand, EVAAV6 exhibited significantly greater transduction in ALI cultures of primary healthy WT HBE cells than in those of primary CF HBE cells. The difference is primarily attributed to the reinforced mucus barrier in the ALI culture of primary CF HBE cells resulted from the dysfunctional CFTR and aftermath pathological events in CF (D. Chen et al., 2021; Suk, 2016). The findings here well align with previous studies demonstrating potential abilities of EVAAV to overcome key extracellular delivery barriers encountered en route to other target organs, including blood‐brain barrier (BBB) and vitreous gel (Hudry et al., 2016; Wassmer et al., 2017). It is yet to be determined whether the EV‐mediated enhancement in the inhaled gene transfer efficacy would be applicable to other mucoadhesive AAV serotypes. However, EVAAV based on BBB‐impermeable AAV8 has been shown to traverse the BBB in vitro and in vivo albeit to a lesser extent compared to EVAAV based on BBB‐permeating AAV9 (Hudry et al., 2016). The tolerability of mucus‐impermeable AAV2 has been confirmed in clinical trials of inhaled gene therapy (Moss et al., 2004). Thus, implementation of the AAV‐piggybacking‐on‐EVs strategy on this serotype may expedite its clinical translation if the enhanced delivery efficacy is established in a similar manner as EVAAV6. It has been shown that conjugation of polyethylene glycol polymers on EVs significantly improves the penetration of EVs through intestinal mucus (Warren et al., 2021). This strategy, however, should be employed with caution not to compromise other merits of this hybrid delivery platform described below.

In addition to improved mucus penetration, EVs promoted the transduction by AAV6 of a conventional 2D culture of HBE cell line (i.e., 16HBE14o‐ cells) largely devoid of the mucus barrier, underscoring the ability of EVs to enhance cellular entry and transduction. Likewise, EVAAVs based on various AAV serotypes showed enhanced in vitro transduction efficacy compared to respective AAVs in 2D cultures of explanted mouse hair cells (Gyorgy et al., 2017), human liver (Meliani et al., 2017) and retinal (Maurya & Jayandharan, 2020) cells, and an array of cancer cell lines (Gyorgy et al., 2014; B. Liu et al., 2021; Maguire et al., 2012). A consensus is yet to be established for the cellular uptake or cargo delivery mechanism of EVs, but various entry pathways have been proposed and/or experimentally validated, which include receptor‐mediated endocytosis, non‐specific macro‐ and micropinocytosis and membrane fusion (Mathieu et al., 2019; Reclusa et al., 2020). Thus, universal enhancement of in vitro cellular transduction observed in our and previous studies may indicate that EVAAV primarily utilizes one or more of the EV entry pathway(s) rather than internalizes into cells via receptors specific to individual AAV serotypes. The ability to transduce various cells may serve as an advantageous aspect for certain therapeutic applications (e.g., production and secretion of therapeutic proteins), but may not be desired if a cell‐specific therapeutic action is preferred. Strategies to address the latter scenario include bestowing EVs with an ability to target specific cells, as implemented for achieving EV‐mediated cell‐specific delivery of therapeutic payloads (Liang et al., 2021). Alternatively, a therapeutic gene controlled by a cell‐ or disease‐specific promoter can be packaged into the AAV (Powell et al., 2015) associated with EVs to promote preferential transgene expression in a designated target cell. We note that the performances of EVAAV can be synergistically enhanced by modulating the AAV component to improve intracellular trafficking (Aslanidi et al., 2013; Pandya et al., 2014). Triple mutations conferred on the viral capsid to preclude ubiquitin‐mediated degradation were shown to significantly improve transduction efficacy of AAV6 in mouse lungs (van Lieshout et al., 2018), and such mutations could be implemented on the EVAAV platform.

The ability of AAV vectors to transduce lung airway cells has been explored by independent groups using diverse experimental settings, and thus it is difficult to make direct comparisons. However, mucus‐secreting ALI cultures of primary human airway epithelial (HAE; inclusive of HBE) may be arguably the most relevant testbed to investigate the efficacy of gene delivery platforms seeking for clinical translation of inhaled gene therapy. In early studies, AAV1 (X. Liu, Luo, Guo, et al., 2007; X. Liu, Luo, Trygg, et al., 2007) and chimeric AAV mutant identified via directed evolution (i.e., AAV2.5T) (Excoffon et al., 2009) were shown to mediate up to ∼100‐fold greater reporter luciferase expression compared to AAV2 and AAV5 in ALI cultures of primary HAE cells. Of note, AAV2.5T was recently shown to enable robust CFTR transgene expression in ALI cultures of primary CF HAE cells and ferret lungs in vivo (Tang et al., 2020). On the other hand, Li et al. reported superior transduction efficacy of AAV6 in ALI cultures of primary WT HAE cells in comparison to AAV1, AAV5 and AAV9, demonstrating roughly double the luciferase expression compared to AAV1 (Li et al., 2009). More recently, we showed that AAV6 mediated ∼3‐fold greater reporter transgene expression compared to AAV1 in ALI cultures of primary CF HBE cells while mucus removal abrogated the difference (Duncan et al., 2018), which highlights the airway mucus as a key delivery barrier. Remarkably, we demonstrate in the present study that EVAAV6 provides over 12‐ and 6‐fold greater reporter transgene expression compared to AAV6 in ALI cultures of WT and CF HBE cells, respectively. We note that this comparative analysis should be interpreted with caution due to the variations in cell donor and culture condition as well as AAV dose and incubation time. Nevertheless, EVAAV6 appears to outperform the clinically tested AAV2 on human airway cell transduction at least by three orders of magnitude.

We found that the EV concentrations in two independently prepared EVAAV6 batches varied at a fixed AAV6 titer. EV and AAV concentrations were quantified to yield the relative EV‐to‐AAV ratios of 1.4 and 18.5 on average for two different batches. However, we and other demonstrated with TEM that multiple AAV capsids were associated with individual EVs (Maguire et al., 2012; Maurya & Jayandharan, 2020; Wassmer et al., 2017). Quantitatively, a previous TEM‐based analysis revealed the relative AAV‐to‐EV ratios to be 8.2 ± 8.2 and 1.8 ± 1.0 for AAV1 and AAV2, respectively (Maguire et al., 2012). We also note that EVAAV‐producing HEK 293 cells secret both AAV‐free EVs and EVAAV, which are collectively harvested as a final product (Maurya & Jayandharan, 2020; Wassmer et al., 2017). To this end, the relative abundance of EVs observed in this study most likely reflects the presence of AAV‐free EVs, and its fraction may vary with individual preparations. These AAV‐free EVs appeared to have negligible impact on the transduction capacity of EVAAV6 in the lung, as evidenced by the comparable levels of in vivo reporter transgene expression observed with two different batches exhibited over an order of magnitude difference in the EV content.

Alongside with therapeutic efficacy, safety is another essential pillar that upholds the clinical development of therapeutic strategies. We show here that the lungs of mice intratracheally receiving AAV6, EVAAV6, or a physical mixture of EVs and AAV6 (i.e., EV+AAV6) are virtually identical to the saline‐treated mouse lungs, underscoring their excellent local safety profiles. The safety of AAV6 was expected a prior given that various AAV serotypes have been validated for tolerability in previous clinical trials, including AAV2 in inhaled gene therapy studies (Flotte et al., 2003; Moss et al., 2007, 2004). EVs are on the threshold of entering the clinical realm with numerous clinical trials recently endorsed by the FDA (Claridge et al., 2021). These trials focus on EVs harvested from naturally occurring cells, including bone marrow mesenchymal stem cells, adipose tissue‐derived stem cells, and monocyte‐derived dendritic cells (Claridge et al., 2021). However, EVs collected from transformed cell lines, such as AAV‐ and EVAAV‐producing HEK 293 cells, may require an additional layer of safety measure. Interestingly, HEK 293 cells have been recently applied in good manufacturing practice to prepare clinical‐grade EVs (Y. S. Chen & Zhi, 2020; Watson et al., 2018), which points to the possibility of their clinical validation in the future. Imminent clinical studies will shed light on the safety landscape of EVs, thereby providing a stepping‐stone towards the clinical development of EVAAV platform.

5. CONCLUSION

In this study, we demonstrated EVAAV6 as a safe and highly efficacious gene delivery platform for inhaled gene therapy applications. We found that EVAAV6 provided markedly greater transduction of human lung airway cells in primary ALI cultures and in mouse lungs compared to standard AAV6, a serotype previously shown to outperform other clinically tested AAV serotypes. In contrast, a physical mixture of individually prepared EVs and AAV6 was unable to establish direct EV‐AAV6 association or to enhance lung airway transduction in vitro and in vivo. Mechanistically, we showed here that EVs improved the mucus penetration and cellular entry/transduction of the stably associated AAV6, underscoring the ability of EVs to carry AAV6 through multiple biological delivery barriers. Importantly, lung tissues from EVAAV6‐treated animals were indistinguishable with those from saline‐treated animals, reflecting excellent local safety profile of this hybrid gene delivery platform. The AAV‐piggybacking‐on‐EVs strategy may open a new avenue for unprecedented clinical translation of AAV‐based inhaled gene therapy.

AUTHOR CONTRIBUTIONS

Gijung Kwak: Data curation; Formal analysis; Investigation; Validation; Writing—original draft; Writing—review & editing. Olesia Gololobova: Formal analysis; Investigation; Writing—review & editing. Neeraj Sharma: Investigation; Writing – review & editing. Colin Caine: Investigation; Resources. Marina Mazur: Investigation; Resources. Kathleen R Mulka: Investigation; Validation. Natalie E. West: Resources; Writing‐review & editing. George M. Solomon: Investigation; Writing—review & editing|Supporting. Garry R. Cutting: Investigation; Writing—review & editing. Kenneth W. Witwer: Investigation; Writing—review & editing. Steven M. Rowe: Investigation; Writing—review & editing. Michael Paulaitis: Investigation; Writing—review & editing. George Aslanidi: Conceptualization; Investigation; Writing—review & editing. Jung Soo Suk: Conceptualization; Funding acquisition; Investigation; Supervision; Validation; Writing—original draft; Writing—review & editing.

CONFLICT OF INTEREST STATEMENT

There is no conflict of interest to be declared by the authors.

Supporting information

supplementary information

Kwak, G. , Gololobova, O. , Sharma, N. , Caine, C. , Mazur, M. , Mulka, K. , West, N. E. , Solomon, G. M. , Cutting, G. R. , Witwer, K. W. , Rowe, S. M. , Paulaitis, M. , Aslanidi, G. , & Suk, J. S. (2023). Extracellular vesicles enhance pulmonary transduction of stably associated adeno‐associated virus following intratracheal administration. Journal of Extracellular Vesicles, 12, e12324. 10.1002/jev2.12324