Introduction

Recombinant adeno-associated virus (AAV) has attracted considerable attention for in vivo gene delivery owing to its distinctive attributes for establishing durable gene expression in various target tissues and demonstrated safety and efficacy as a leading gene delivery tool. In recent years, remarkable progress has been made in the clinical application of AAV gene therapy. To date, eight AAV-based gene therapy products have received regulatory approval, and over 300 clinical trials are currently underway [1, 2]. Gene therapy represents a groundbreaking approach to potentially curing previously untreatable genetic diseases. It also addresses unmet medical needs in chronic and prevalent human diseases by offering more tolerable treatment options. Clinical studies and the regulatory approval process for AAV gene therapies often utilize a fast-track or accelerated approach, in contrast to the traditional development lifecycle for biological drugs. This reduced timeline, combined with often limited AAV materials during development, can pose significant challenges to the Chemistry, Manufacturing, and Controls development of AAV products. Comprehensive studies to support the development of optimal drug substance and drug product manufacturing process, stable formulations, and robust analytics often fail to be carried out for AAV products, especially prior to First-in-human studies. To mitigate the risks associated with an accelerated development approach, acquiring a thorough understanding of the critical quality attributes (CQAs) along with the degradation pathways of AAV under the various stresses that may occur during manufacturing, transport, storage, and clinical handling is highly desirable. This is particularly important for AAV, given the complexity of its molecular structure. AAV is a non-enveloped virus containing an icosahedral capsid encapsulating a single-stranded DNA genome of 4.7 kb flanked by a 145 bp Inverted Terminal Repeat at each end. The capsid is assembled by 60 viral proteins (VP) with VP1, VP2, and VP3 at a molar ratio of 1:1:10. VP1 is the longest among the three VP proteins, extending 137 amino acids beyond VP2 [3]. This additional segment, known as the VP1 unique (VP1u) region, exhibits phospholipase A2 activity and is implicated in nuclear localization [4,5,6]. VP3 is composed of conserved β-strands linked by surface-exposed variable loops, which are known to be critical in determining the tropism of distinct AAV serotypes [6]. Furthermore, AAV capsids have unique structural features that are critical to their function. These include channels with pore-like openings at each five-fold axis, which are essential for genome packaging and the externalization of VP1u during endosomal trafficking [7,8,9]. Additionally, the capsids exhibit protrusions surrounding the three-fold axis that play a pivotal role in targeting cell surface [10]. These structural elements have been proposed to play a crucial role in the mechanisms by which AAV interacts with cellular surfaces and delivers transgenes into the nuclei of target tissues or cells. Similar to protein-based therapeutics, instability of AAVs can be mediated by either chemical or physical degradation pathways, or the combination of both [11]. The most commonly observed chemical degradation mechanisms include amino acid modifications such as deamidation, oxidation, and isomerization [11, 12]. In contrast, physical degradation often arises from conformational, colloidal, and interfacial mechanisms resulting in denaturation, aggregation, surface adsorption, and associated product loss [13]. One unique physical degradation pathway of AAV is vector genome DNA leakage, which has been observed under various stress conditions such as freeze-thaw and elevated temperatures [14,15,16]. It is presumably caused by conformational changes of AAV capsids leading to potential widening of the pore structure at the five-fold axis on the AAV capsids where DNA extrusion may occur [17]. AAVs are commonly stored under frozen conditions and may encounter many stress factors during drug substance and drug product manufacturing, transport, storage, and clinical handling [11, 18, 19]. These stress factors may include pH extremes, temperature variations, oxidation, multiple freeze-thaw cycles, and mechanical stress. In this study, we applied these different types of stress to AAV8 in an optimized formulation. The conditions selected for this study are more extreme than what are typically used during routine stability assessment. The goal of this study is to generate degradation products which help to determine the degradation pathways of AAV8 under each stress condition. The results from these studies will be valuable for determining the CQAs of AAV8, supporting the development of robust formulation and stability indicating assays, as well as contributing to the assessment of manufacturing comparability to ensure consistent manufacturing of safe and effective AAV products for clinical applications. In addition, these results will further help to ensure the success with regulatory submissions, especially as the regulatory framework for gene therapies continues to evolve significantly.

Materials and methods

Materials

AAV8 encoding a therapeutic transgene was produced using triple transfection approach in HEK293 cells by Virus Production Core at Regeneron Pharmaceuticals, Inc. (Tarrytown, NY, United States). All other excipients and reagents are reagent or compendial grades.

Stress conditions

AAV8 material was formulated with a base formulation buffer or dialyzed into the buffer of interest to achieve a target concentration of 2.0E + 13 vg/mL. The base formulation contains 10 mM Tris, 180 mM sodium chloride, 1.5% (w/v) sucrose, 0.005% (w/v) P188 at pH 7.3 [16]. Buffer exchange was performed using Slide-A-Lyzer dialysis cassettes with a 10 kDa molecular weight cutoff (Thermo Fisher, Waltham, MA). After formulation or dialysis, the AAV8 was filled into cyclic olefin polymer vials (West Pharmaceutical, Exton, PA). To evaluate the impact of oxidative stress on AAV8, the samples were exposed to hydrogen peroxide (H2O2) concentrations of 1 ppm and 1000 ppm, respectively, followed by incubation at 25 °C for 1 week. Untreated virus incubated under the same 25 °C condition was used as a control. Upon completion of the incubation, the samples were dialyzed into the base formulation using the 10 kDa molecular weight cutoff Slide-A-Lyzer dialysis cassettes to remove residual H2O2. For the pH stress study, AAV8 was dialyzed into two different buffers at pH 3.0 (Glycine-based buffer) and pH 9.0 (Bis-Tris Propane-based buffer). Following dialysis, the samples were incubated at 25 °C for 1 day and 1 week. To assess the effect of temperature on AAV8 stability, AAV8 was stored at 40 °C and 45 °C for 1, 2, and 4 weeks. The stability of AAV8 under freeze-thaw stress was assessed by subjecting the samples to 2 or 4 freeze-thaw cycles between −80 °C and room temperature. For the agitation stress study, the samples were vortexed at 1000 rpm for 60 min or 120 min on a benchtop vortexer.

In vitro protein expression assay

The in vitro protein expression assay utilized a proprietary engineered cell line to analyze the expression level of the protein encoded by the transgene encapsidated in AAV8. These cells were seeded and incubated with AAV8, along with a proprietary gene editing reagent. After transduction, the protein of interest was analyzed using an in-house ELISA-based method. The signal versus log dilution was evaluated using non-linear regression analysis with GraphPad Prism 10.1 (Dotmatics, Boston, MA). All critical reagents used in the assay are protected under proprietary rights.

Post-translational modification (PTM) analysis by liquid chromatography mass spectrometry (LC-MS/MS)

The AAV samples under different stress conditions were prepared for PTM analysis by LC-MS/MS based on the previously developed method with minor modifications [20]. Briefly, a total of ~7 µg of AAV capsids from each sample was digested with a trypsin/rLys-C mixture (1:1 w/w ratio) at an enzyme-to-substrate ratio of 1:5 (w/w) at 37 °C for 2 h after denaturation, reduction, and alkylation. After digestion, 3 µL of 10% formic acid was added to quench the digestion. The digests were transferred to sample vials, and ~2.7 µg of the digests were injected for LC-MS/MS analysis.

All the AAV8 samples were analyzed on an Orbitrap Exploris 480 mass spectrometer system coupled with a Vanquish Flex UPLC (Thermo Fisher). The digested peptides were separated on an ACQUITY Premier Peptide CSH C18 column (130 Å, 1.7 μm, 2.1 mm × 150 mm, Waters) at a flow rate of 0.25 mL/min using a 75-min gradient from 0.1% to 32% mobile phase B, then increasing to 45% buffer mobile phase B in 15 min. Mobile phase A and B were 0.1% (v/v) formic acid in water and 0.1% (v/v) formic acid in acetonitrile, respectively. The column temperature was set at 40 °C. The mass spectrometer was set up with a 3.5 kV source voltage, a capillary temperature of 320 °C, a sheath gas flow of 35, an auxiliary gas flow of 10, and an S-lens RF level of 50. MS1 data was collected in the range of 300–2000 m/z with a resolution of 120 K. The top 15 peptides were selected for MS/MS analysis at an HCD normalized collision energy of 30% and a resolution of 15 K.

All the raw data were searched with Byonic software Version 5.8.6 (Protein Metrics, Cupertino, CA). Skyline software (MacCoss Lab Software, Seattle, WA) was used for PTM quantification by importing the search data from Byonic. To align the modification locations in the different VP proteins, all the modification sites were mapped to the VP1 sequence unless otherwise noted.

Particle size distribution by dynamic light scattering (DLS)

Particle size distribution was assessed using a Zetasizer Ultra DLS instrument (Malvern, Westborough, MA) with a low-volume quartz cuvette (Malvern). Measurements were taken at a 90° angle, with each sample subjected to five replicate scans, each comprising 10 acquisitions of 10 s. Data was collected and analyzed with ZSXplorer software (Malvern).

Vector genome titer by droplet digital polymerase chain reaction (ddPCR)

The AAV8 samples were subjected to treatment with DNase I (Thermo Fisher) to eliminate un-encapsulated DNA, followed by digestion of the AAV8 capsids using proteinase K (Thermo Fisher). The digested samples were then diluted with dilution buffer (20 mM Tris, 0.05% (v/v) Pluronic F-68, 0.002 µg/µL sheared salmon sperm DNA, pH 8.0) to achieve a target concentration range of 200–8000 copies/µL, combined with ddPCR master mix (Bio-Rad, Hercules, CA) and transgene specific primer probe set, and partitioned on a AutoDG droplet generator (Bio-Rad). Polymerase Chain Reaction amplification was subsequently performed using a thermocycler (Bio-Rad), and the resulting droplets were analyzed with QxManager Software on a QX-200 droplet reader (Bio-Rad) to quantify the vector genome titer.

Capsid recovery analysis by size-exclusion chromatography (SEC)

Capsid recovery was assessed using SEC on an ACQUITY UPLC H-Class system equipped with PDA and FLR detectors (Waters, Milford, MA) [16]. The analysis employed a Sepax SRT SEC-1000 Column (Sepax, Newark, DE). Analyte separation was carried out using isocratic elution at a flow rate of 0.35 mL/min with a total runtime of 40 min. The mobile phase consists of 20 mM sodium phosphate, 300 mM sodium chloride, and 0.02% (w/v) poloxamer 188 at a pH of 7.3. Empower software (Waters) was used for peak integration and identification.

Free DNA analysis by fluorescence-based dye binding assay

Free DNA analysis was performed using the Qubit™ ssDNA Assay Kit (Thermo Fisher). Stressed AAV8 samples were diluted in AAV8 formulation buffer and mixed with the Qubit™ solution. Standards and test articles were loaded into a black 96-well microplate (Greiner Bio-One, Monroe, NC) and fluorescence was measured at 500 nm excitation and 525 nm emission using a Synergy Neo2 reader (Agilent, Santa Clara, CA). The percentage of free DNA was calculated by comparing to the free DNA level in a positive control AAV8 sample incubated at 75 °C for 40 min.

Determination of onset temperature for genome DNA release (Tonset) by differential scanning fluorimetry (DSF)

Tonset of untreated AAV8 sample was determined using DSF on an Uncle system (Unchained Labs, Pleasanton, CA). SYBR® Gold Nucleic Acid Gel Stain (Thermo Fisher) was diluted to the appropriate concentration using AAV8 formulation buffer and then combined with the AAV8 sample. A total of 9 µL of the prepared mixture was loaded into the Unis (Unchained Labs) for melting temperature (Tm) analysis using a thermal ramp from 20 °C to 95 °C with a heating rate of 0.5 °C per minute and a hold time of 60 s at each temperature step. SYBR® Gold was excited using a 473 nm laser. Tonset value of AAV8 was determined based on the fluorescence intensity area between 500 nm and 650 nm across the temperature range using the Uncle software (Unchained Labs).

VP ratio analysis by microfluidic electrophoresis

The VP ratio of AAV8 samples was analyzed using microfluidic electrophoresis on a LabChip® GXII Touch™ HT protein characterization system (Revvity, Waltham, MA) with the HT Protein Express LabChip (Revvity) and Pico Protein Express assay reagent kit (Revvity). Samples were diluted to 1.0E + 13 cp/mL in phosphate-buffered saline (Corning, NY, United States). Then, a total of 10 µL of reducing sample buffer (60 mM phosphate, 1.2% (w/v) lithium dodecyl sulfate, 80 mM dithiothreitol, pH 8.0) was added to 40 µL of the diluted sample. Following a 10-min denaturation at 75 °C, 5 µL of 16 µM dye was added to 5 µL of the sample mixture, incubated at 35 °C for 15 min, then further diluted with 105 µL dilution solution provided in the kit. The LabChip was prepared according to the manufacturer’s instructions. Data processing was performed using Empower software (Waters), and the calculation was conducted by integrating the signal from three major peak species observed in the electropherogram. To determine the VP ratio, the corrected peak area for each VP was divided by the total number of lysine residues in the respective VP, and the resulting values were used to calculate the VP ratio.

Results

The manufacturing of AAV involves a complex series of processes that can expose the virus to various stress factors such as oxidation, acidic and basic pH conditions, high and low temperatures, freeze-thaw and agitation. To identify potential degradation pathways, AAV8 was subjected to relevant stress conditions as described in Table 1. Under each stress condition, the impact on the biological activity of AAV8 was first assessed using an in vitro transgene expression assay. Subsequently, potential degradation pathways were examined by measuring changes in several key AAV attributes. Potential chemical degradation, indicated by the formation of various PTM, was studied using an extensive LC-MS-based reduced peptide mapping method. Meanwhile, potential physical degradation pathways such as capsid aggregation, genome DNA release, capsid fragmentation, and associated titer loss were evaluated using various techniques. Specifically, capsid aggregation was probed by measuring the particle size distribution with DLS and the recovery of monomeric AAV using SEC. Genome DNA release was studied through a fluorescence-based DNA dye binding assay. Potential capsid fragmentation was evaluated using a microfluidic electrophoresis method. Finally, changes in vector genome titer were analyzed via ddPCR.

Table 1 Summary of stress conditions utilized in AAV8 forced degradation study.
Full size table

Stability changes of AAV8 under oxidative stress

During biopharmaceutical drug product manufacturing process, vaporized hydrogen peroxide is commonly used to decontaminate isolators prior to aseptic filling. As a result, AAV products may be exposed to oxidative stress at about 1 ppm H2O2 from the residual vaporized hydrogen peroxide during manufacturing [21]. In this study, AAV8 at a concentration of 2E + 13 vg/mL were subjected to oxidative stress by incubation with 1 ppm and 1000 ppm H2O2 for one week at room temperature. Following the oxidative exposure, we found that under the highest oxidative stress condition (1000 ppm H2O2) assessed, a substantial reduction in AAV8 transgene expression was observed (Fig. 1A). In contrast, transgene expression remained stable under mild oxidative condition (1 ppm H2O2) when compared to control. Oxidation induced by exposure to hydrogen peroxide is known to affect the stability and biological activity of therapeutic proteins [22] by altering the protein primary structure through chemical modification of amino acid side chains. To characterize the changes which may be responsible for the reduced transgene expression under oxidative stress conditions, LC-MS-based reduced peptide mapping analysis was performed to achieve site-specific PTM quantitation. As depicted in Fig. 1B, treatment with 1000 ppm H₂O₂ resulted in over 90% oxidation at several methionine residues, including M204, M212, and M561. M204 is part of the extended sequences of VP1 and VP2, while M212 is in the N-terminal region of VP3. M561 is shared by all three capsid proteins, positioned on the surface of the three-fold protrusion, and exhibits limited solvent accessibility (Fig. 1C). Additionally, elevated methionine oxidation was observed at M374 (from 1.5% to 26%), M405 (from 0.2% to 13%), and M637 (from 0.4% to 16%) (Fig. 1B and Supplementary Table 1). These residues are either minimally solvent-accessible or buried within the native capsid structure [23] consistent with their relatively low oxidation levels under strong oxidative stress conditions. Mild oxidative stress (i.e., 1 ppm H2O2) induced similar changes in methionine oxidation, although to a much lesser extent (Fig. 1B). As the transgene expression remained stable under mild oxidative stress compared to the control, it was concluded that moderate oxidation (M204 from 1.3% to 10.1% and M212 from 1.3% to 8.9%) of these methionine residues (Supplementary Table 1) did not significantly impact AAV biological activity. Finally, no significant changes in other PTM, including asparagine deamidation, aspartic acid isomerization, VP protein N-terminal acetylation, and serine/threonine phosphorylation, were observed under either strong or mild oxidative stress conditions (data not shown). Previous studies have reported that oxidative modifications in therapeutic proteins often lead to alteration in higher-order structures, promote aggregate formation, and generate products with potentially reduced biological activity [24,25,26]. To understand whether physical degradation occurred alongside chemical modifications in the AAV8 samples under oxidative stress and contributed to the observed reduction in transgene expression various analytical techniques as described above were utilized. Despite the particle size remained consistent under strong oxidative stress conditions (Fig. 1D), SEC analysis revealed up to approximately 10% loss in monomeric AAV8 recovery, indicating that capsid titer loss may be attributed to the surface adsorption of denatured AAV8 products onto the container-closure system (Fig. 1E). Meanwhile, the increasing levels of oxidative stress were associated with a trend of decreasing vector genome titer, with up to ~15% reduction observed under 1000 ppm H₂O₂ treatment (Fig. 1F). This decline was accompanied by a corresponding increase in free DNA levels, rising from ~0.3% to 1.8% (Fig. 1G). These observations may be attributed to conformational instability of AAV8 capsid under oxidative stress, potentially driven by oxidation-induced conformational changes in structure, which in turn leads to the release of genomic DNA. The methionine oxidation profile further reveals capsid structural changes under oxidative stress, as maximum oxidation of M204 and M212 coincides with extensive oxidation of other methionine residues, including nearly complete oxidation of M561. Importantly, M561 is largely inaccessible to solvent in the native state (Fig. 1C), suggesting that additional structural rearrangements are necessary for it to become susceptible to oxidation. Interestingly, a prior study identified the critical role of M203 and M211 in capsid assembly. Point mutation of these two methionine residues to Leu led to failed capsid production. Therefore, it is reasonable to hypothesize that the near-complete oxidation of these two methionine residues under strong oxidative stress conditions may significantly destabilize the capsid structure, thereby facilitating the DNA release [27]. The stoichiometry of viral proteins was also analyzed, revealing that both the control and 1 ppm H₂O₂ treated samples exhibited a VP1:VP2:VP3 ratio of ~1:1.7:10.4 (Fig. 1H), consistent with the commonly reported VP molar ratio of ~1:1:10 for AAVs demonstrating good transduction efficiency [28, 29]. However, exposure of AAV8 to 1000 ppm H₂O₂ resulted in a noticeable change in viral protein stoichiometry, with the VP3 ratio increasing from 10.4 to 12.2 (Fig. 1H). This suggests fragmentation of capsids may occur under strong oxidative stress conditions, resulting in increased VP3 ratio. The data collectively suggests both chemical and physical degradation pathways occur under oxidative stress. Alterations in the transgene expression of AAV8 are likely to be attributed to the combination effect from methionine oxidation on capsid proteins, genomic DNA leakage caused by oxidation-induced capsid conformational change and instability, reduction of capsid titer due to associated surface adsorption, as well as increased VP3 ratio as a result of fragmentation of capsid under strong oxidative stress.

Fig. 1: Stability changes of AAV8 under oxidative stress.
figure 1

AAV8 samples were exposed to mild (1 ppm H2O2) and strong (1000 ppm H2O2) oxidative stress conditions and subsequently analyzed and compared to the control sample. Data are presented as mean ± standard deviation when multiple replicate measurements are performed. A In vitro protein expression analyzed by cell-based assay (n = 3). B Quantitation of Met oxidation showing the distribution of key Met sites identified by LC-MS/MS (n = 1). The asterisk indicates a residue in the unique N-terminal peptides of VP3. C M561 (PDB: 2QA0) is located on the capsid surface near the three-fold protrusion axis. It is highlighted in yellow, and the solvent accessibility analysis indicates limited solvent exposure of M561. D Average hydrodynamic diameter (Z-Average) measured by DLS (n = 5, technical replicates). E Monomeric capsid recovery analyzed by SE-UPLC (n = 1). F Vector genome titer determined by ddPCR and normalized to the vg titer of the control (n = 4, technical replicates). G Free DNA measured by fluorescence-based dye binding assay and normalized to the total genomic DNA amount (n = 4, technical replicates). H Determination of VP ratio by microfluidic electrophoresis under 1000 ppm H₂O₂ oxidative stress (n = 2, technical replicates).

Full size image

Stability changes of AAV8 under pH stress

The pH-dependent degradation behavior of AAV8 was evaluated under extreme acidic and basic conditions to simulate potential exposures during the drug substance manufacturing process. Specifically, AAV8 was dialyzed into respective buffers at pH 3.0 and pH 9.0, followed by incubation at 25 °C for up to seven days. A buffer at neutral pH (7.3) was used as a control. Consistent with prior findings that AAV8 and AAV9 lose transduction efficiency under highly acidic conditions [30,31,32], our results revealed a substantial reduction in transgene expression in AAV8 following seven days of incubation at pH 3.0 (Fig. 2A). In contrast, transgene expression levels in the AAV8 samples incubated at pH 9.0 remained comparable to those observed in the control buffer at pH 7.3 (Fig. 2A). A comprehensive PTM analysis revealed that AAV8 exhibited increased asparagine deamidation at pH 9.0, with levels at N57, N94, and N263 reaching ~38%, 5%, and 29%, respectively (Fig. 2B). Notably, N57 and N94 are located at VP1u region, whereas N263 is surface-exposed (Fig. 2C) and situated in the region shared by all three capsid proteins. Although these asparagine deamidations were found to be critical for AAV8 transduction efficiency in a previous report using point mutation analysis [33], our results showed that moderate increases in deamidation at N57, N94 and N263, induced by pH 9.0 stress, did not significantly affect AAV8 transgene expression. This discrepancy is likely due to differences in deamidation levels: point mutations result in 100% change, whereas real-life stress typically induces only moderate levels. Therefore, AAV capsids may exhibit greater tolerance to deamidation than previously understood, or the impact of deamidation might vary based on the specific residues involved within the capsid structure. Furthermore, isomerization analysis revealed minor changes under acidic condition at D4, D220, and D628, with increases from ~3% to 9%, 5% to 7%, and 2.5% to 3.1%, respectively (Fig. 2D). The increased isomerization observed under pH 3.0 stress, compared to neutral pH and pH 9.0, aligns with the mechanism of this modification, as slightly acidic conditions facilitate the formation of succinimide, the intermediate product in Asp isomerization. No other notable PTM changes were detected under pH 9.0 and 3.0 conditions. These findings indicate that, although extreme pH conditions can induce chemical modifications in AAV capsid proteins, such modifications generally occur at low levels at room temperature over reasonably short durations (e.g., up to 7 days).

Fig. 2: Stability changes of AAV8 under pH stress.
figure 2

AAV8 samples were dialyzed into a glycine-based buffer at pH 3.0 and a Bis-Tris Propane-based buffer at pH 9.0, respectively. The samples were subsequently analyzed and compared to a control sample at pH 7.3. A In vitro protein expression (n = 3). B Quantitation of Asn deamidation. C N263 (PDB:2QA0), highlighted in red, is positioned on the capsid surface and is susceptible to deamidation. D Quantitation of Asp isomerization (n = 1). E Average hydrodynamic diameter (n = 5, technical replicates). F Monomeric capsid recovery (n = 1). G Normalized vector genome titer (n = 4, technical replicates). H Normalized free DNA (n = 4, technical replicates). I VP ratio (n = 2, technical replicates).

Full size image

Subsequently, we investigated whether physical degradation of AAV8 contributes to the observed reduction in AAV8 transgene expression under pH 3.0 stress condition. We observed that, at pH 3.0, the particle size of AAV8 increased from ~25 nm to 44 nm immediately after dialysis and further expanded to ~65 nm after 7 days of incubation at pH 3.0, indicating the formation of large AAV8 aggregates (Fig. 2E). In contrast, the particle size of AAV8 remained stable at pH 9.0 or 7.3 throughout the 7-day incubation period. Interestingly, SEC analysis revealed negligible changes in monomeric AAV recovery under both pH 3.0 and pH 9.0 conditions (Fig. 2F). This suggests the large aggregates detected at pH 3.0 by DLS may be reversible under the running condition of SEC. Thus, the stability-indicating capability of SEC for detecting and quantifying AAV aggregates remains to be further evaluated. Additionally, a slight reduction in vector genome titer was observed under both pH 3.0 and pH 9.0 conditions, with decreases of 13% and 9%, respectively (Fig. 2G). Free DNA levels in the AAV8 sample increased from 0.4% to 1.3% immediately after dialysis into the pH 3.0 buffer, with no further increase observed after the 7-day incubation at pH 3.0 (Fig. 2H). Similar genomic DNA leakage from viral particles at pH 4.0 has been reported for AAV2 and AAVv66 [34]. Viral protein stoichiometry analysis revealed that the VP3 ratio increased from 10.5 to 11.4 at pH 9.0 and to 11.9 at pH 3.0, while the VP2 ratio decreased from 1.7 to 1.5 and 1.3, respectively (Fig. 2I). This suggests high pH up to 9.0 and low pH near 3.0 may induce capsid fragmentation. Exposure of AAVs to these extreme pH conditions during the manufacturing process should be carefully controlled or minimized when possible. In summary, our findings indicate that during the drug substance manufacturing process, AAV8 may maintain acceptable stability and quality under strong basic conditions but is susceptible to degradation under strong acidic conditions. The degradations likely involve both chemical and physical pathways, with aggregation being the major contributor to the loss of AAV8 biological activity in strong acidic environments.

Stability changes of AAV8 under 40 °C and 45 °C

The impact of thermal stress on AAV8 was assessed by exposing AAV8 to elevated temperatures for durations of up to four weeks. The onset temperature for vector genome DNA release (Tonset) of AAV8 in the optimized AAV8 formulation, determined using DSF, was ~50 °C (Fig. 3A), and was used to guide the selection of stress temperatures. To ensure sufficient generation of degradation products and to understand the differential effect of temperature on AAV8, two stress temperatures of 40 °C and 45 °C were assessed in this study following the general rule of thumb that the selected thermal stress temperature should stay below Tonset and at least 10 °C to 20 °C below Tm [35]. As illustrated in Fig. 3B, subjecting AAV8 to up to 4 weeks of incubation at 40 °C and 45 °C resulted in a drastic decrease in transgene expression, demonstrating a temperature-dependent effect on impacting AAV stability. After 4 weeks, the transgene expression in the AAV8 sample was greatly reduced at 40 °C and completely abolished at 45 °C. High-temperature incubation also led to extensive chemical modifications on AAV8 capsid proteins. After incubating AAV8 at 40 °C and 45 °C for 4 weeks, increased aspartic acid isomerization was detected at D4 (from 3% to 24% and 43%), D69 (from 0.3% to 4% and 11%), and D220 (from 5% to 19% and 27%) (Fig. 3C). Elevated asparagine deamidation was observed at N57 (from 18% to 50% and 82%), N94 (from 2% to 21% and 56%), and N263 (from 14% to 35% and 53%) (Fig. 3D). Additionally, methionine oxidation increased notably at M204 (from ~8% to 15%), M212 (from ~9% to 17%), and M212* (from ~5% to 12%) (Fig. 3E). Consistent with previous findings, the VP1u, VP1/2, and VP3 N-terminus regions were identified as hotspots for PTM formation under thermal stress [20]. The VP1u externalization under heat shock, along with its unstructured conformation, may provide a plausible explanation for the accelerated deamidation at N57 and N94 and isomerization at D4 and D69 [7]. Additionally, a recent hydrogen/deuterium exchange mass spectrometry study analyzed AAV8 incubated at 60 °C and revealed increased deuterium uptake in the five-fold axis region, specifically in a polypeptide sequence spanning residues A205 to W229 [36]. The increased structure dynamics of this region under thermal stress correlated with the accelerated formation of oxidation at M204 and M212, as well as isomerization at D220 (Fig. 3F). Finally, the extended levels of deamidation at N263 under thermal stress are attributed to its high solvent accessibility and favorable Asn-Gly motif, consistent with previous reports [33]. Moreover, our results showed that the rate of PTM formation in AAV8 accelerates at 45 °C compared to 40 °C, correlating with reduced transgene expression at 45 °C. Interestingly, prior studies conducted under 1 ppm H2O2 oxidative stress and pH 9.0 stress revealed that moderate levels of Met oxidation and Asn deamidation did not significantly affect biological activity of AAV8, indicating that other degradation pathways may also play a role.

Fig. 3: Stability changes of AAV8 under thermal stress at 40°C and 45°C.
figure 3

A Determination of Tonset and Tm by DSF using SYBR® gold dye. B In vitro protein expression (n = 3). Quantitation of C Asp isomerization, D Asn deamidation, and E Met oxidation (n = 1). F D220 (PDB:2QA0), highlighted in red, is oriented towards the interior of the capsid, near the five-fold axis pore. G Average hydrodynamic diameter (n = 5, technical replicates). H Monomeric capsid recovery (n = 1). I Normalized vector genome titer (n = 4, technical replicates). J Normalized free DNA (n = 4, technical replicates). K VP ratio (n = 2, technical replicates).

Full size image

It is well known that therapeutic proteins are prone to extended exposure at high temperatures, which may induce conformational changes resulting in physical instability and/or reduced biological activity [37]. Aligning with previous reports [38, 39], our results showed that the particle size of AAV8, as measured by DLS, increased from 26 nm to 31 nm and 62 nm after 4 weeks at 40 °C and 45 °C, respectively (Fig. 3G). This was accompanied by a significant loss in monomeric AAV recovery, as revealed by SEC analysis, which showed reductions of ~75% and 26% after the 4-week incubation at 40 °C and 45 °C, respectively (Fig. 3H). These results indicate, unlike the AAV aggregates formed under pH 3.0 condition, the large AAV aggregates formed after extended exposure at high temperatures might be due to profound and extensive conformational changes. It is likely these aggregates remain irreversible under the SEC running conditions and thus cannot enter the SEC column. Concurrently, a significant reduction in vector genome titer (Fig. 3I) and an increase in free DNA (Fig. 3J) were also observed, especially at 45 °C. To assess the impact of vector genome DNA release on genome titer loss, genome titers of the AAV8 samples stressed at 40 °C and 45 °C were also measured without DNase I digestion prior to capsid denaturation. We found omitting DNase I digestion improved genome titer recovery from 40% to 75% for AAV8 incubated at 45 °C for 4 weeks (Supplemental Fig. 1A, B) supporting the presence of externalized DNA. Free DNA-mediated capsid aggregation, along with the subsequent sample loss due to either precipitation or surface adsorption, likely accounts for the remaining genome titer loss. In addition, thermal stress resulted in substantial alterations in viral protein ratios. The VP3 ratio increased from 10.3 to 12.1 at 40 °C and 16.3 at 45 °C, while the VP2 ratio rose from 1.6 to 1.8 and 2.5, respectively (Fig. 3K). This suggests fragmentation of capsids occurs under these thermal stress conditions, especially at 45 °C. Altogether, our study revealed that several physical degradation pathways of AAV8 including capsid aggregation, genome DNA release, viral protein fragmentation, and associated surface adsorption, are triggered by high-temperature stress. Additionally, we observed a significant difference in the extent of physical degradation at 40 °C versus 45 °C, indicating that AAV8 capsids might lose structural integrity at temperatures approaching the Tonset temperature of 50 °C in this formulation. It is clear a combination effect of both chemical and physical degradations of AAV8 accounts for the reduction of biological activity under thermal stress.

Stability changes of AAV8 under freeze-thaw and agitation stress

AAV8 was subjected to two and four freeze-thaw cycles between −80 °C and room temperature, or vortexing at 1000 rpm for durations of 60 min and 120 min, to assess its degradation profile under freeze-thaw and agitation stress conditions. Our results showed that no meaningful changes to the AAV8 transgene expression, particle size distribution, vector genome titer, recovery of monomeric AAV8 capsids, and viral protein ratio were detected in the freeze-thaw or agitation-stressed samples (Fig. 4A–E). These results indicate that neither chemical nor physical degradations of AAV8 occurred in the base formulation under the freeze-thaw and agitation stress conditions assessed in this study.

Fig. 4: Stability changes of AAV8 under freeze-thaw and agitation stress.
figure 4

A In vitro protein expression (n = 3). B Average hydrodynamic diameter (n = 5, technical replicates). C Recovery of monomeric AAV8 measured by SE-UPLC (n = 1). D Normalized vector genome titer (n = 4, technical replicates). E The VP ratios remained unchanged after freeze-thaw cycles and agitation stress (n = 2, technical replicates).

Full size image

Discussion

In this work, we performed a comprehensive forced degradation study of AAV8 by applying various types of stresses which might be encountered during AAV manufacturing, storage, shipping and handling, but exceeding the conditions commonly used for stability assessment. Our goal is to gain an in-depth understanding of the major degradation pathways under each type of stress and recommend CQAs of AAV to monitor and control during manufacturing, product release, and stability testing. Performing such a study is particularly important in the case of AAV gene therapy drug development due to the complexity of their molecular structure and the limited prior knowledge about the potential influence of these stresses on the product attributes of various AAV serotypes. Although the degradation products generated from forced degradation studies may not be formed under relevant manufacturing, storage, and in-use conditions, the result of the study provides critical insight into developing stability-indicating assays, designing effective strategies for manufacturing, formulation, and drug product processes, as well as establishing the product degradation profiles for comparability assessment.

In this study, we employed various analytical techniques capable of detecting and quantifying the different types of degradation products that may alter the sequence, structure, concentration, purity, or biological activity of the stressed AAV8 samples. Comprehensive and appropriate characterization of the biochemical, biophysical, and biological properties of these samples is an essential part for a well-designed forced degradation study for achieving a thorough understanding of the chemical and physical degradation mechanisms of the product. While LC-MS/MS was found sufficient for characterizing the site-specific chemical modifications in stressed AAV8 samples, the complexity of physical degradation mechanisms for AAVs and our evolving understanding of preliminary critical quality attributes necessitated the use of a large and diverse set of assays to monitor the stability-indicating preliminary critical quality attributes. These assays include vector genome titer by ddPCR, aggregation or particle size distribution by DLS, monomeric AAV capsid recovery by SEC, free DNA by dye-binding assay, viral protein ratio by microfluidic electrophoresis, and transgene protein expression by cell-based assay. Indeed, these assays allow us to form a much more complete picture of the degradation pathways of AAV8, which occur under the various modes of stress. Many of the degradation pathways identified in this study either failed to be revealed or analyzed to the same extent in detail in previously reported forced degradation studies of other AAV serotypes [13, 31, 39]. Furthermore, our results showed the assays used in this study demonstrated sufficient sensitivity and good performance for detecting and quantifying the AAV8 degradation products and have great potential to be established in a quality control setting as stability-indicating release assays to support AAV product release and shelf-life stability testing.

By exposing AAV8 to commonly used forced degradation conditions including oxidation, high and low pH, high temperature, freeze-thaw, and agitation stress, we have identified the major chemical degradation pathways of AAV8 are methionine oxidation, asparagine deamidation, and aspartic acid isomerization (Table 2). The major physical degradation pathways of AAV8 include aggregation, genome DNA leakage, viral protein fragmentation, and associated surface adsorption potentially due to conformational changes in capsids and/or aggregation (Table 2). Specifically, strong oxidative stress causes significant methionine oxidation on AAV8 capsid proteins. This likely induces conformational change in the capsids, which subsequently induces genome DNA leakage, fragmentation of viral proteins, and associated surface adsorption. As a result, a substantial reduction of AAV8 transgene expression was observed under strong oxidative stress. In contrast, chemical degradations are not the major degradation pathway for AAV8 under pH stress. We found that AAV8 maintains acceptable stability and quality in our previously developed formulation under strong basic conditions but is susceptible to physical degradation under strong acidic conditions. Extreme low pH causes formation of large AAV8 aggregates and accelerates viral protein fragmentation, which are likely to be the major degradation pathways underlying the loss of AAV8 transgene expression at pH 3.0. Our study also revealed that high temperature stress accelerates various chemical and physical degradation pathways. Exposure of AAV8 at 40 °C and 45 °C after 4 weeks led to significantly elevated aspartic acid isomerization and asparagine deamidation, as well as a notable increase in methionine oxidation. Extensive aggregation, viral protein fragmentation, and genome DNA leakage were also observed under high temperature stress. The combination effect from these changes caused drastic abolishment of AAV8 transgene expression in a temperature-dependent manner. In addition, we observed a surprisingly large difference in the extent of chemical and physical degradation at 40 °C versus 45 °C, indicating AAV8 capsids might lose structural integrity at temperatures approaching the Tonset temperature of 50 °C in this formulation. Under the freeze-thaw and agitation stress conditions evaluated in this study, minimal degradation of AAV8 such as aggregation or particulate formation was observed, indicating the formulation factors we selected for AAV8 including pH, ionic strength, composition and concentration of cryoprotection, and surfactant are sufficient to protect AAV8 against the interfacial stress and phase separation occurred under the conditions evaluated in this study, thus maintaining good stability.

Table 2 Summary of AAV8 degradation pathways under stress conditions.
Full size table

It is worthwhile to mention that the formulation conditions selected for forced degradation studies could influence the study results because pH, ionic strength, as well as the composition and concentration of each excipient in the formulation may impact chemical stability and/or physical stability of a protein, with the latter mainly attributed to influence on colloidal stability of the protein [40,41,42]. This protective effect may contribute to the mild changes observed in oxidation and transgene expression of AAV8 when exposed to 1 ppm H2O2, or it is possible that the degree of methionine oxidation did not reach the threshold required to compromise capsid stability. Similarly as reported previously by Bee J et al. [42], the optimized AAV8 formulation used in this study, containing a surfactant (0.005% w/v P188), cryoprotectant (1.5% w/v sucrose), and low concentration of a buffer which does not cause pH shift under freezing condition (10 mM Tris), was designed to enhance the stability of AAV as evident that no meaningful changes were observed in all the quality attributes monitored after four freeze-thaw cycles or vortexing at 1000 rpm for 120 min. It is worth noting that AAV8 at 2E + 13 vg/mL was evaluated in all stress conditions. However, lower concentration formulation could have shown greater impact on some of the stresses that may have been masked with higher concentration formulations (e.g., titer loss due to adsorption, oxidation). In addition, a standardized AAV manufacturing process has yet to be developed so far due to many practical challenges such as the complexity of the molecular structure of AAV, the fact that variations in AAV capsid structure and serotype may demand changes in the production and/or purification process, and most importantly, the intrinsic heterogenous nature of AAV rendering high-resolution separation of empty, partial, and full capsids extremely difficult especially during large-scale clinical manufacturing process. Consequently, AAVs from different resources not only vary in AAV serotypes or capsid structure, transgene sequence, format, or length, they could also have different levels of chemical modifications, purity, or impurity profiles. These differences may have an impact on the quality and/or stability of the AAV products. Given the absence of well-established standardized analytics for AAV characterization and definition of CQAs for AAV, caution should be taken when analyzing and interpreting results from different forced degradation studies reported in literature.

Recently, Plegaria et al. reported a pH-dependent DNA degradation pathway for two undisclosed AAV serotypes [43]. By using a DNA capillary gel electrophoresis (cGE) method combined with SYBR Green II-based fluorescence detection, the authors reported significant loss of encapsidated ssDNA and potency which occurred within days of storage at 25 °C, at even mildly acidic and basic pH. They further hypothesized that encapsidated ssDNA damage could likely occur via depurination of the DNA bases followed by fragmentation under acidic condition, although no direct evidence was provided in this report. These observations seem to deviate from the findings in our study, which probably could be explained largely by the factors described above such as variations in AAV formulation compositions, manufacturing process, capsid structure or serotypes, as well as transgene sequence, format, or length. Nonetheless, their work provides new directions towards characterizing the degradation pathways of the encapsidated DNA in addition to the AAV capsid proteins. The continuous improvement and expansion of the AAV stability knowledge and analytics toolbox will be critical to make this happen.

In summary, our study has elucidated how AAV8 degrades under various stress conditions. These findings are likely applicable to other AAV serotypes as the capsid protein sequences and structure are highly conserved among certain serotypes. In addition, the knowledge gained from this study can accelerate analytical, formulation, and process development efforts, supporting the establishment of robust manufacturing process and the production of safe and efficacious AAV gene therapy products.