pH-dependent DNA degradation pathways for adeno-associated virus gene therapy

Abstract

Adeno-associated virus (AAV) vector genome degradation mechanisms that reduce transduction efficiency and viral potency must be understood to define critical quality attributes to ensure safety and efficacy. Here, we report that under prolonged storage at 25°C with two capsid types from clade C and E and varying buffer conditions the observation of degradation of rAAV DNA. Studies revealed two main degradation mechanisms: chemical degradation of encapsidated DNA at acidic pH and ejection of DNA at basic pH, with both occurring at neutral pH. Moreover, storage of clade E capsid in acidic pH at 25°C for 7 days showed >50% potency loss, which strongly correlated with a drop in encapsidated single-stranded DNA (ssDNA) (>40%) but not with conventional attributes used to characterize AAV stability, such as vector genome titer or full percentage. Degradation of the encapsidated ssDNA could be potentially driven by low pH non-enzymatic depurination reactions, facilitated by the local DNA-protein interactions within the capsid. Therefore, DNA integrity is a product attribute that should be studied and controlled in the development of AAV-based gene therapies.

Graphical abstract

Adeno-associated virus (AAV) degradation reduces gene therapy efficacy, but mechanisms remain unclear. Researchers developed a capillary gel electrophoresis assay to quantify encapsidated DNA degradation, identifying acid-driven chemical breakdown and base-induced ejection. DNA degrades faster inside capsids, impacting AAV potency. Ensuring genome stability is crucial for manufacturing and efficacy.

Introduction

Recombinant adeno-associated virus (rAAV) is a leading vector for gene therapy.^1,2 AAV is a ∼25 nm non-enveloped virus with a single-stranded DNA (ssDNA) payload of typically 4.5 kilobases.^3,4 Despite significant efforts in improving rAAV production and analytics,^5,6,7 there are still gaps in understanding the relationship between rAAV’s biophysical characteristics, biological activity, and therapeutic efficiency (i.e., structure-function relationship). AAVs transduce cells via attachment, entry, endosomal escape, and then nuclear import.^8,9,10 Given the complex mechanisms involved in delivering genetic materials into cells, it is expected that the titer of virus with intact genome is a major determinant of potency. Less clear is the degree at which other attributes impact potency, such as the minimum required viral protein 1 (VP1) and 2 (VP2) copies for efficient cellular entry and endosomal escape.^11,12,13

Understanding the structure-function relationship of biologics is essential in early phase drug development. One method to study this relationship is through forced degradation studies, which subjects the biologic to extreme conditions (i.e., temperature and pH) to accelerate changes to physical properties and then probing key attributes, such as titer/capsid loss, aggregation, chemical modification, or change in activity.^14,15,16 For AAV, forced degradation studies have predominately reported on capsid titer and DNA ejection.^{17,18,19,20,21,22} For example, it was shown that DNA is ejected at exposure to even moderate temperatures of 35°C–50°C^17,18,21 or through undergoing multiple freeze-thaw cycles.¹⁹ Empty and full capsids were also reported to follow different aggregation kinetics, likely due to interactions with ejected DNA.¹⁷ Moreover, chemical modification of VP proteins was identified as a prominent chemical degradation in multiple AAV serotypes, with hot-spot residues on both the VP1 unique and the VP3 shared regions.^23,24,25 For example, rAAV9 stored at 40°C exhibited significant VP1 N56 deamidation and reduction in potency.¹⁶ Furthermore, the decreased thermal stability or ejection temperature of the cargo DNA was found to be dependent on both serotype and VP ratio.²⁶

Comparatively fewer studies have investigated degradation of DNA within the AAV capsids. Chemical degradation mechanisms of DNA^27,28,29,30 include cleavage of phosphate polymer to make DNA fragments^29,31 and cleavage of DNA bases from the phosphate backbone (i.e., depurination) followed by fragmentation. It is viable that such changes may happen to AAV DNA during liquid storage and manufacturing. Several reports have noted heterogeneity of packaged DNA^{32,33,34,35,36} but primarily associated with genome design. Furthermore, both the DNA structure (e.g., single-stranded or self-complementary) and size was shown to be key factors in AAV stability; for instance, ssDNA cargos >4.5 kb exhibited lower ejection temperatures and reduced encapsidation.¹⁸ To date, only a few studies have aimed to understand the chemical stability of the AAV cargo or analytical methods for determining its stability.^37,38,39 For example, Barnes et al. employed a charge detection mass spectrometry method to characterize DNA truncation,³⁸ while Takino et al. utilized orthogonal methods to investigate DNA stability during light exposure.³⁹ This is a key gap in understanding the impact of chemical and physical structure changes on efficacy or safety, as even minor damage to the DNA may completely disrupt the AAV’s function and therapeutic effectiveness.

The work presented here was designed to address the stability of AAV cargo by studying the degradation of encapsidated full length ssDNA (FL-ssDNA). We present an assay that extracts DNA cargo from AAV and quantifies both the size and abundance of extracted FL-ssDNA using capillary electrophoresis. Key solution conditions and clade E (capsid #1) and clade C (capsid #2) capsids were then tested in forced degradation studies. After extended storage at 25°C, the results show that the DNA undergoes chemical degradation within both AAV capsids at acidic pH, while ejection at basic pH occurs for capsid #1. Moreover, at acidic pH, the loss of encapsidated FL-ssDNA abundance correlates strongly to potency losses, with weak correlations between potency and more traditional AAV quality attributes such as vector genome (vg) titer and aggregation. Our findings show that genome integrity is a key AAV product characteristic that can change during AAV processing, formulation, and storage.

Results

Capillary-based ssDNA size and abundance assay to assess DNA stability

A capillary gel electrophoresis (cGE) method was developed to monitor genome integrity in release and stability studies. A diagram of the procedure workflow and example electropherograms for all FL-ssDNA purified from rAAVs used in this study are provided in Figure 1. The cGE method has several notable capabilities. First, it can differentiate between encapsidated and total FL-ssDNA (encapsidated + extra-viral) within a sample by comparing preparations with and without nuclease treatment (Figure 1A). Second, the method confirms the kilobase length of ssDNA that corresponds to each peak in the electropherograms by comparison to a ssDNA ladder. Finally, the relative abundance (to T0) of the FL-ssDNA genome can be calculated by integration of the main peak in each electropherogram. To do this, the total concentration of DNA is determined after purification by UV absorbance. Then, the total DNA concentration is normalized across all samples before separation. During separation, SYBR Green II binds to the ssDNA. Because SYBR Green II binds uniquely to each ssDNA based on its length and sequence, only peaks corresponding to the same sequence and length can be compared to each other. For each capsid studied below, the main peak corresponds to the FL-ssDNA genome, for which sequence is constant, and size is confirmed with the ssDNA ladder. If (DNA) is normalized before separation and the sequence of ssDNA corresponding to a peak is known, then the relative abundance of the FL-ssDNA can be calculated across samples by comparing the main peak area from a condition to the main peak area at time zero (T0).

Figure 1.

Determination of size and abundance of extracted AAV DNA by capillary gel electrophoresis

AAVs are a mix of empty capsids (empty hexagon), full capsids (hexagon + blue), partial capsids (hexagon + orange) and extra-viral DNA (purple) (A). Purification can be tailored with a nuclease step to measure the abundance of the encapsidated DNA inside the capsids (top arrow) or total DNA associated with the sample (bottom arrow). Purified DNA are analyzed with the same cGE separation, resulting in an electropherogram with peaks whose migration time and corrected peak areas were used to calculate size and relative abundance. Electropherograms (B) resulting from FL-ssDNA (main peak) purified from capsid #1 (4.7 kb ssDNA vg), capsid #2 (3.5 kb ssDNA vg), and capsid #1 (3.1 kb ssDNA vg). PhiX ssDNA is observed in all electropherograms and serves as an internal standard.

Three rAAVs were used in the studies presented here. Capsid #1 with a 4.7 kb ssDNA vg, capsid #2 with a 3.5 kb ssDNA vg, and capsid #1 with a 3.1 kb ssDNA vg. DNA was first extracted from each rAAV and separated according to the steps illustrated in Figure 1A. The resulting electropherograms are displayed in Figure 1B. PhiX ssDNA serves as an internal standard and is present in all three electropherograms. Both ssDNA genomes from capsid #1 are clean with a single main peak eluting at the appropriate migration time for its full-length size. The FL-ssDNA genome from capsid #2 also has a clean electropherogram and elutes as a doublet, likely due to ssDNA secondary structure. The relatively flat baselines on either side of the main peaks in each electropherogram allowed for uncomplicated integration and quantification across the studies presented further.

pH is a key factor in DNA degradation

The chemical stability of encapsidated FL-ssDNA in AAV was assessed by incubating capsid #1 and capsid #2 at 25°C for 30 days (TD30) in various formulation conditions (Table 1; Figure 2A). Higher temperatures were not evaluated to minimize DNA ejection, which is accelerated at ≥30°C and may potentially obscure impacts on the chemical stability of the encapsidated DNA. Key factors were varied to compare the impact of both capsid- and buffer-specific parameters. These factors included capsid type and solution parameters, including pH, NaCl concentration, and an antioxidant (methionine) (Table 1). For pH, a typical formulation development range of 5.8–9.5 was evaluated around the pI (∼6.8) of the full capsid. Post stress, the percentage full capsid (F%) of the rAAVs were measured via mass photometry (MP) and extracted encapsidated FL-ssDNA were quantified using the cGE method (Figure S1). The results were normalized to the T0 unstressed values (T_{x stress time point}/T₀) to compare the change in both attributes, which is used throughout. The cGE and MP results are summarized in Table S1.

Table 1.

Parameters studied to evaluate key AAV, DNA, and solution effects on DNA integrity

Parameter	Range	Justification
Clade	E: capsid #1 (pI 6.9)a C: capsid #2 (pI 6.8)a	Capsid protein-specific effects
pH	5.2, 6.2, 7.3, 8.7, 9.5	Low pH DNA hydrolysis mechanism40 High pH protein chemical modification23
[NaCl]	50, 150 mM	AAV colloidal stability41
[Met]	0, 5 mM	Oxidation inhibitor

^aCalculated pI value of the capsid were determined using pI values of the VP proteins and VP ratios.

Figure 2.

Impact of key AAV and solution parameters on abundance of encapsidated FL-ssDNA

Key effects on the abundance of encapsidated FL-ssDNA were studied on capsid #1 (4.7 kb vg) and capsid #2 (3.5 kb vg) in varying formulation conditions (A). The relative DNA abundance (blue circles) and F% capsids (red circles) vs. pH of stressed samples (B). Points were connected to highlight trends in the data. Filled and open circles indicate T30D samples and T30D + Met samples, respectively.

The change in relative abundance of FL-ssDNA and F% in both rAAVs at varying buffer conditions is shown in Figure 2B. When comparing the impact of the formulation conditions (pH, salt, and Met), the data demonstrates a strong pH-dependence for both attributes (Table S1). For capsid #1, abundance of FL-ssDNA was strongly pH dependent. Greater than 50% of encapsidated FL-ssDNA was lost at both pH ≤ 6.0 and pH ≥ 8.5. F% was unaffected between pH 5 and 7 but decreased >50% at pH ≥ 8.5. Likewise, capsid #2 also showed >50% encapsidated FL-ssDNA loss in acidic pH, however, at basic pH, both FL-ssDNA abundance and F% is maintained. Interestingly, for capsid #2, the sample with Met at pH 6.0 and 7.5 had decreased FL-ssDNA abundance but F% was unaffected.

The impact of factors on the abundance of FL-ssDNA is quantitatively summarized using analysis of variance (ANOVA) (Table S2). For both capsids, pH has a highly significant impact on the abundance of encapsidated FL-ssDNA (F-score >15, p ≤ 0.05). All other terms are not statistically significant (F < 2, p ≥ 0.05). This statistical analysis confirms that pH is the key factor affecting the abundance of the encapsidated FL-ssDNA, with no significant impact by NaCl and Met.

Single-molecule imaging of stressed AAV samples shows DNA ejection

Select capsid #1 and capsid #2 samples after T30D of stress at 25°C were imaged with low-voltage electron microscopy (LVEM) to further understand capsid and DNA morphological changes (Figure 3). Images were either negatively stained with thulium acetate to reveal capsid morphology or unstained for analysis of empty (EC), partially filled (PC), and filled (FC) capsid population percentages. For each population analysis, at least 1,500 capsids were analyzed across multiple representative images from the same grid using custom-written Mathematica-based image processing and neural-net categorization routines, as previously described.⁴²

Figure 3.

Representative low-voltage electron microscopy images

Capsid #1 (4.7 kb vg) pH 6.0 T30D sample EM: stained (A) and unstained (B). pH 8.7 TD30 sample EM: stained (C) and unstained (D). Arrows highlight sample species: in stained EMs yellow for intact capsids and orange for broken capsids, while in unstained EMs, green for full capsids, red for empty capsids, and blue for free DNA. All scale bars, 200 nm.

The electron micrographs (EMs) of stained and unstained capsid #1 T30D samples at pH 6.0 and pH 8.7 are shown in Figure 3. T30D samples at pH 6.0 consist of mostly intact capsids that were impermeable to stain (Figure 3A). The unstained EMs (Figure 3B) calculated an empty:partial:full capsid population as 11%:2%:86% while T0 samples are 14%:4%:81% (Figure S2A). In contrast, the T30D sample at pH 8.7 showed a significant number of broken and/or incomplete capsids in the stained EMs (Figure 3C) and visible ruptures in the periphery of the capsid boundary when observed at higher resolution (Figure S3). These samples also have decreased full (41%) and increased empty (55%) capsids (Figure S2A).

LVEM analysis was also performed on capsid #2 at pH 8.7 (Figure S2B). Unstained EMs demonstrated high full capsid population (88%) at T30D (Figure S2A) and mostly intact and full capsids were observed in stained EMs. This data corroborates the MP F% results in Figure 2B.

One notable aspect in the unstained EMs is the presence of free DNA in capsid #1 pH 8.7 T30D sample (Figure 3D). In contrast, free DNA was not observed in the unstained EMs of either capsid #1 at pH 6.0 T30D (Figure 3B) or capsid #2 at pH 8.7 T30D (Figure S2B), indicating free or ejected DNA were not present in these samples. The observed free DNA of capsid #1 at high pH are found both in solution and adjacent to one or more capsids (Figure 3D). Most of the DNA observed are longer than the contour length of the ssDNA cargo for this rAAV and are likely aggregates of DNA-DNA and DNA-capsid. Together, the LVEM analyses further supports ejection of encapsidated DNA from capsid #1 at a basic pH but not acidic pH condition.

Degradation kinetics of full length encapsidated ssDNA loss

With solution pH identified as the key factor impacting the stability of the encapsidated FL-ssDNA, the rate of loss was further studied as a function of pH for capsid #1. Here, capsid #1 was formulated into pH 6.0, 7.5, and 8.7 in buffered solutions with 175 mM NaCl. These samples were also stressed at 25°C and then sampled across T30D of storage. The cGE assay was used to evaluate the stability of the DNA and MP for AAV F%. The relative impact of both DNA degradation and DNA ejection on DNA loss was assessed by measuring the abundance of encapsidated (with nuclease) and total FL-ssDNA (without nuclease) for each sample. The electropherograms and raw data are summarized in Figure S4 and Table S3, respectively.

Three distinct mechanisms of FL-ssDNA genome loss were observed. At pH 6.0, the abundance of encapsidated ssDNA decreased by 30% by TD7 and >50% by TD30 (Figure 4A). The total ssDNA displays a similar trend of FL-ssDNA loss, which suggests DNA degradation occurred inside the capsid. Consistent with this, F% remained unchanged when compared to T0 (Figure S5). At pH 8.7, rapid loss of encapsidated FL-ssDNA is observed with ∼30% of the abundance remaining at TD7 and 0% at TD30 (Figure 4A). However, the abundance of total FL-ssDNA (no nuclease pretreatment) was unchanged, demonstrating that the vg loss in the encapsidated samples was due to ejection. Consistent with ejection, by TD30, the F% decreased to 18% from 82% (Figure S5). Finally, by TD30 at pH 7.5, ∼30% of encapsidated FL-ssDNA remained and ∼75% of the total FL-ssDNA (Figure 4A), while F% only decreased by ∼30%, (Figure S5). Together these data suggest that at neutral pH both mechanisms of DNA internal degradation and ejection are occurring simultaneously but to a lesser extent than either pH 6.0 or pH 8.7 (Figure 4B).

Figure 4.

Kinetics of FL-ssDNA loss at pH 6.0, 7.5, and 8.7

Capsid #1 (4.7 kb vg) normalized cGE FL-ssDNA abundance graphed vs. time (A) for all three pH values tested and illustration of DNA degradation (B). Encapsidated FL-ssDNA is depicted by the red trace (with nuclease treatment) and total FL-ssDNA is depicted by the blue trace (without nuclease treatment). Data points are connected to highlight data trends. Tx = days incubated at 25°C.

DNA chemical degradation is accelerated within the AAV capsid

The role of the AAV capsid in ssDNA degradation was further assessed by comparison of the loss in abundance of encapsidated ssDNA to free ssDNA in solution. FL-ssDNA was extracted from capsid #1 and incubated either alone (pH 6.0, 7.5) or with empty capsid#1 (pH 6.0 only) at 25°C for T30D (Figure 5A). The normalized cGE results for the stressed extracted samples are plotted in Figure 5B with the encapsidated results from Figure 4A to compare the impact of the capsid. The cGE results over T30D for the extracted ssDNA are summarized in Table S4 and the electropherograms compared in Figure S6.

Figure 5.

Comparison of DNA abundance loss of extracted and encapsidated FL-ssDNA

Schematic of ssDNA sources (A) of encapsidated DNA from full capsid #1 with 4.7 kb vg (green), extracted DNA in solution (blue) and extracted DNA with 1:1 empty capsid (red), and total ssDNA loss over time normalized to T0 (B) Tx = days incubated at 25°C. Data points are connected to emphasize trends. The encapsidated DNA data is from Figure 4.

For all samples in this experiment, FL-ssDNA quantification focused on the total amount of DNA in solution (no nuclease pre-treatment) so that the DNA loss only represents chemical degradation. Extracted FL-ssDNA incubated in the buffer alone shows minimal loss at pH 6.0 (Figures 5B and S6). When empty capsid #1 was incubated with extracted DNA at pH 6.0, mild loss of the FL-ssDNA was also observed and to the same extent as extracted ssDNA alone. These data are in sharp contrast to the accelerated loss of encapsidated FL-ssDNA, with >25% DNA loss by T7D. At pH 7.5, minimal loss is observed for extracted and encapsidated FL-ssDNA (Figure 5B).

The accelerated degradation for encapsidated FL-ssDNA was quantitatively estimated by fitting the rate of ssDNA loss to Arrhenius kinetics to determine first-order rate constants (Figure S7A). At pH 6.0, the rate constant for encapsidated FL-ssDNA loss was approximately 5-fold higher than free FL-ssDNA only and with empty capsid (Figure S7B). At pH 7.5, the rate constant for encapsidated FL-ssDNA loss is also ∼5-fold lower than pH 6.0 but still non-zero, while the rate of free ssDNA loss in solution is effectively zero (Figure S7B). For each pH, this increased rate of DNA chemical degradation is higher for encapsidated ssDNA than the extracted ssDNA in formulation alone. As such, the increased loss of ssDNA from inside capsid #1 compared to extracted DNA with or without empty capsid again suggests that the capsid microenvironment accelerates the chemical degradation of the AAV genome at low pH.

Encapsidated ssDNA loss explains AAV potency loss at acidic pH

Given that the AAV DNA cargo chemically degrades at acidic, and, to a lesser extent neutral pH, we next assessed how the loss of FL-ssDNA at low pH corresponds to in-vitro potency. Samples from two lots of capsid #1 with a 3.1 kb transgene were prepared in a pH 5.2 solution and stressed at 25°C. Samples were tested across 45 days to assess changes in key product attributes (Table S5): in-vitro potency, encapsidated full length ssDNA abundance (Figure S8), and genome titer (vg) by ddPCR, as well as characterization commonly used in AAV manufacturing such as F% by analytical ultracentrifugation (AUC), aggregate species by AUC high molecular weight (HMW), capsid titer by ELISA, and the fraction copy of VP1 and VP2 proteins (Table S6) by SDS capillary electrophoresis. Furthermore, long read next generation sequencing (LR-NGS) was performed on T0 and stressed samples (Figure S9), serving as a direct orthogonal method to the cGE.

Relative changes in potency, which is measured by the transfection and subsequent measurement of enzymatic activity of the expressed α-Gal A protein, and other key attributes of stressed capsid #1 is compared in Figure 6A. After T14D, in-vitro potency exhibited >50% loss compared to the initial T0 sample and full potency loss after T30D. There were minimal changes in vg titer and F%, as well as in capsid titer and HMW species (Table S5). The relative fraction of VP1 decreased by one copy number after T30D, which suggests capsid changes due to peptide cleavage, however, those minor changes do not trend with significant potency loss. In contrast, the drop in encapsidated FL-ssDNA abundance as measured by cGE was the only attribute that strongly trends with potency (Figure 6A).

Figure 6.

DNA-integrity directly correlates with potency loss at acidic pH

Normalized change in cGE FL-ssDNA abundance, potency, and key quality attributes (A) and averaged mapped read lengths of the FL-ssDNA (B) for capsid #1 with a 3.1 kb ssDNA in pH 5.2 stored at 25°C over 45 days. Data points in panel A are connected to highlight data trends and error bars in both panels represent standard deviation. Tx = days incubated at 25°C.

A correlation fit analysis was performed to further corroborate the changes between potency loss and the key quality attributes (Figure S10). Here, the relative change (vs. T0) of each attribute was fitted to the relative change in potency, and the linear fit statistics are summarized using R² for correlation and slope for response strength. Potency and ssDNA abundance showed the strongest correlation with an R² = 0.96 and a slope of 1.1. There is modest correlation between potency and the VP1 protein fraction, with 0.7 R². All other attributes, including F%, v titer, and capsid titer, have weak R² below 0.5.

The most notable observation was again the abundance of FL-ssDNA decreased rapidly with potency, while the vg by ddPCR is relatively unaffected. This apparent contradiction can be explained by differences between orthogonal methods for genome analysis. The capillary-based ssDNA abundance assay measures the DNA length and abundance via fluorophore-labeled intensity. DNA damage will reduce the fluorescence signal relative to an intact sequence. In contrast, dPCR measures vg by amplifying a short ∼100 base region of the vg. Any chemical damage outside of the amplification region will not affect the amplification efficiency, and therefore, will not reduce the apparent vg titer. The LR-NGS analysis further corroborated the degradation of encapsidated FL-ssDNA, revealing a 50% drop in the mapped read length (MRL) of the full 3.1 kb vg at T7 and a 99% at T45 (Figure 6B and S9). Therefore, the cGE DNA integrity and LR-NGS method collectively indicate non-specific degradation of the DNA cargo, which impaired AAV function and was not apparent in standard PCR or biophysical characterization of AAVs.

Discussion

A capillary-based ssDNA assay was developed and applied to study the impact of key formulation factors on AAV degradation and to further understand the role of DNA integrity in AAV activity. We found pH had the strongest impact on the stability of the encapsidated FL-ssDNA for two different serotypes from clade E and C (Figure 2). In basic pH, capsid #1 (but not capsid #2) experienced significant encapsidated ssDNA loss due to DNA ejection outside of the capsid (Figures 3 and 4). This ejection in stressed conditions is likely due to change in the capsid structure at the DNA and capsid interface via chemical modification (e.g., deamidation).²³

The degree of stabilization is also likely dependent on the effective pI of the capsid interior, and especially the VP1-unique region which interacts strongly with the DNA cargo.²⁴ A recent publication that investigated the structural role of VP1 in two different serotypes showed vg release is pH sensitive and dependent on the VP make-up of the capsid.²⁶ Interactions of the DNA cargo with the VP1-unique region may also contribute as capsid #2 has approximately 2-fold greater VP1 per capsid relative to capsid #1, and therefore, more of the potentially stabilizing effects that can mitigate DNA ejection at basic pH and effects from local electrostatic interactions even with similar overall capsid pI (6.8–6.9). In addition to the capsid microenvironments, the size of the vg has been shown to drive DNA ejection from the capsid. Horowitz et al. compared the ejection of three ssDNAs (3.4 kb, 4.1 kb, and 4.7 kb) and showed that the larger ssDNA had a higher propensity for ejection at lower temperature by 10°C–15°C, potentially due to increased internal pressure.¹⁸ Overall, the impact of the local structural environment (e.g., VP1-DNA interaction) and cargo size likely both contribute to the DNA ejection mechanism between capsid #1 and capsid #2 at basic pH. Additional work is needed to further understand the impact of these factors (e.g., capsid type and DNA length) on AAV degradation.

In acidic pH, ensemble of measurements shows that the FL-ssDNA cargo degrades within the capsid (Figure 4 and S5). LVEM imaging supports this conclusion by revealing that the sample primarily consists of intact capsids without evidence of ejected DNA (Figure 3). The rate of chemical degradation was calculated to be approximately 5-fold faster within a capsid compared to extracted DNA in solution (Figure 5), with rate influenced by the buffer condition (i.e., pH) but not the capsid exterior. Corroborated by the LR-NGS analysis, the cGE DNA method provided a clear explanation for potency loss in the forced degradation at acidic pH. The change in the encapsidated FL-ssDNA had the strongest correlation with potency loss compared to typical attributes (e.g., vg, and capsid titer, F%, and HMW) used in AAV manufacturing (Figure 6). A decrease in the VP1 per capsid also occurred, which indicates a change in the capsid structure and stability, but not to the same magnitude as potency, and occurring after the potency loss. Nonetheless, we cannot isolate the individual contributions of DNA degradation and VP1 region cleavage to potency loss. It’s possible that the degradation of the encapsidated DNA in capsid #1 removes stabilizing interactions between the VP1 peptide and the DNA. This could expose VP1/VP2 regions to non-enzymatic cleavage and post-translational modification (PTM, data not shown), further decreasing efficacy. The complex interactions within the capsid environment and with the cargo DNA are the subject of our subsequent work.

Two likely mechanisms contribute to the loss of AAV genome integrity and thus potency observed in this study: depurination and fragmentation.^27,30 Both mechanisms would be observed in the cGE method as a loss in encapsidated FL-ssDNA signal. Depurination naturally occurs at low pH through hydrolysis, followed by cleavage of DNA bases from the phosphate backbone sugar. Fragmentation of the phosphate backbone, which is typically enzyme catalyzed, can occur at both 5′ and 3′ side of any given phosphate group. Given that both cGE and NGS DNA calculations are based on abundance, we expect the FL-ssDNA fragmentation is non-specific and random. Further supporting this observation, we found no evidence of a co-packaged DNase in this study, but we can’t rule out the possibility that internal amino acids are acting as DNases. The AAV capsid’s microenvironment is highly charged, with a high local concentration of DNA bases, positively charged amino acids, and counterions for the DNA phosphate. This high local concentration could alter the local pH and ionic strength, affecting inter- and intra-molecular interactions like electrostatics and hydrogen-bonding. Ultimately, this might accelerate chemical changes to the DNA and stabilize reactive intermediates that drive chemical degradation.

Depurination of ssDNA is known to be accelerated in acidic pH and increased temperature, as well as by DNA sequence.^35,40 While subsequent phosphate backbone breakage is more likely, the rate of phosphodiester bond cleavage after depurination is slow in solution. However, this process could be accelerated by the protein environment and the sterically constrained interior of the AAV capsid. Consistent with this, encapsidated ssDNA showed accelerated signal loss at pH 6.0, while extracted (non-encapsidated) ssDNA remained unaffected in the same buffer conditions (Figure 5). Moreover, conformational strain on polymers and nanotubes has been shown to enhance chemical decomposition of polymers and carbon nanotubes.⁴³ Given that DNA is highly compressed within the AAV capsid,¹⁸ the resulting mechanical stress may destabilize the DNA relative to its more mechanically relaxed and favored conformation in solution.

Conclusions

This work studied the chemical degradation of encapsidated FL-ssDNA cargo within two rAAV capsids. A capillary-based assay determined the integrity of encapsidated FL-ssDNA after extraction by determining size and relative abundance. When varying both solution conditions and AAV attributes, pH was determined to be the major driver for encapsidated ssDNA loss. DNA was shown to chemically degrade inside the capsid at low pH for clade E and C capsids, while one the clade E capsid demonstrated ssDNA ejection at high pH. Both mechanisms occur at neutral pH but with slower overall rates.

Encapsidated ssDNA loss at low pH was due to chemical degradation within both capsids studied. Comparison of degradation rates between encapsidated clade E and extracted DNA in solution showed that the DNA degradation was accelerated within the AAV capsid. For samples at pH 5.2, DNA degradation with clade E capsid was the only attribute studied that correlated with potency loss, with a nearly 1:1 ratio. Encapsidated ssDNA damage could likely occur via depurination of the DNA bases followed by fragmentation.

Overall, this study reveals key degradation mechanisms of rAAV DNA cargo that impacts the viral therapeutic effectiveness. Significant loss of encapsidated FL-ssDNA and potency can occur within days at 25°C, at even mildly acidic and basic pH, so early formulation development and control strategies are crucial to avoid extended storage, handling, and manufacturing at these conditions. Products with extended liquid storage without mechanism-of-action or similar potency testing to support the ssDNA stability will be of especially high risk as genome damage is not observable by more traditional AAV analytics (e.g., qPCR and AUC). In addition to supporting process development and storage activities, this work identifies the integrity of the DNA cargo as a critical quality attribute that must be characterized during manufacturing and storage of AAV therapies.

Materials and methods

AAV capsid production and formulation

This study used clade E and clade C engineered rAAV capsids #1 and #2, respectively. Two capsid #1 constructs were used with a gene of interest (GOI)/vg length of 3.1 or 4.7 kb and one capsid #2 construct with 3.5 kb vg. At >10 L scales, vectors were expressed in an adapted suspension HEK293 cell line via a triple-transfection method. Cells were cultured and agitated in gas (air, CO₂, and O₂)-controlled bioreactors at 37°C using Expi293F expression medium. The medium also contained sodium bicarbonate (J.T. Baker) for pH control, dextrose (J.T. Baker) for glucose regulation, and antifoam (Cytiva) to manage excessive foaming. After reaching the target cell density, the cells were transfected with a DNA reagent solution containing an optimized Rep/Cap:Helper:Transgene ratio mix, polyethylenimine (Polysciences), and sodium valproate (Spectrum Chemical). Following 72 h of incubation, cells were chemically lysed with a surfactant and then clarified using depth filters. The filtrate from the clade C and E capsids was subsequently purified via affinity chromatography using AVB Sepharose (Cytiva) and POROS CaptureSelect AAVX (Thermo Fisher Scientific) resins, respectively. Finally, affinity eluates were polished with anion-exchange chromatography, utilizing POROS XQ resin (Thermo Fisher Scientific), to remove cellular contaminants and enrich the full AAV fraction. For capsid #1 (4.7 kb vg) and capsid #2 (3.5 kb vg), single drug substance (DS) lots were generated. Two DS lots were also produced for capsid #1 (3.1 kb vg), employing the same production process previously described. These lots were used in the experiments detailed herein.

Tris, phosphate, and acetate salts were sourced from Thermo Fisher Scientific. Methionine and NaCl were purchased from Sigma. High concentration buffer and solution stocks were prepared and filtered through a 0.2 μm filter before use. Tris and phosphate buffers were prepared for pH conditions ≥7.2 and acetate for pH conditions ≤6.2. AAV were prepared in target solutions either by performing buffer exchange using a 100 kDa cellulose acetate membrane (Unchained Labs) or diluting a high-concentration rAAV stock into the target solution at ≥10-fold dilution. All rAAV samples were filled in crystal zenith (West Pharmaceuticals) vials or polypropylene tubes (Fisher Scientific) and stored in a temperature-controlled chamber with 60% relative humidity. To induce mild forced AAV degradation, capsid #1 and capsid #2 were stored at 25°C for 30–45 days and sampled at select time points. Control un-stressed samples (T0) were also prepared in parallel and frozen in −80°C freezer prior to testing with the stressed samples.

Analytical methods

Analytical methods consist of minor product-specific adaptations from standard protocols within the field, and for brevity, only key details are provided and referred to as standard protocols.

qPCR

Encapsidated vector genome titer of rAAV (vg titer) was measured by digital PCR.⁴⁴ Samples were pre-treated with nuclease to digest DNA and RNA in solution, then treated with proteinase K to degrade the capsid and disrupt any residual nuclease activity, followed by heat inactivation of proteinase K. Vector genome DNA concentration was then measured with QIAcuity8 dPCR system using appropriate primer-probe sets for each transgene, with an amplicon size of typically of 80–150 bp.

ELISA

Capsid titer was measured using an ELISA method with capsid-specific capture antibodies.⁴⁴

Mass photometry and analytical ultracentrifugation

Empty, partial, and full capsid percentages were quantified using MP (Refeyn)⁴⁵ for high-throughput low-resolution analytics and sedimentation velocity AUC (SV-AUC)⁴⁶ for low-throughput high-resolution analytics. For MP, AAV samples were diluted into a phosphate buffer. SV-AUC was performed using two-channel Epon charcoal-filled centerpieces equipped with sapphire windows in an Optima AUC (Beckman Coulter, Brea, CA) instrument with an An50 Ti rotor. Sedimentation boundaries were collected at 12,000 rpm using UV detection at 230, 260, and 280 nm and interference (IF). All data were collected at 20°C. SV-AUC data were analyzed using a continuous sedimentation coefficient [c(s)] distribution model in Sedfit (v.16.p36).

Low voltage electron microscopy

LVEM, a form of transmission electron microscopy, was employed to assess capsid morphology, free DNA content, and the population distribution of empty, partially filled, and full capsids. Briefly, samples were diluted, deposited onto TEM grids (Ted Pella 01822-F IC-A 400 mesh Cu), washed, and in some cases negative-stained with thulium acetate. The resulting grids were imaged in an LVEM25 instrument (Delong Instruments) in TEM mode. Population analyses were performed using custom-written Mathematica-based image processing and neural-net categorization routines.⁴²

Potency assays

In vitro mechanism-of-action potency was assessed by the expression and enzymatic activity of the α-Gal A protein. Human hepatoma 7 cells were transduced with specific dilutions of AAV, leading to expression of α-Gal A. Enzymatic activity of α-Gal A is measured by incubating with a substrate that upon cleavage by α-Gal A forms a fluorogenic product. Fluorescence for reference standard and test article(s) dilutions are plotted against vg concentration, and dose-response curves are fitted by linear regression. The test article is compared to the reference standard using parallel-line analysis (slope ratio), and the relative potency of the test article is reported as a percentage of the reference standard.

CE-SDS for viral protein

Viral protein capillary electrophoresis sodium dodecyl sulfate (CE-SDS) was used to quantify the size and relative abundance and capsid proteins.¹³ Capsids were denatured in the presence of SDS, reducing agent, and heat. Denatured proteins were then separated using the IgG Heterogeneity and Purity Kit (SciEx) and the PA 800 Plus Capillary Electrophoresis System. Proteins were detected using UV absorbance at 220 nm.

CGE for single-stranded DNA

Capillary gel electrophoresis (cGE) method for ssDNA to size and abundance was developed to assess the vg integrity of the AAV cargo. Encapsidated nucleic acids were purified from AAV samples by first incubating with nuclease to degrade extra-vector nucleic acids. Capsid proteins were then degraded by proteinase K treatment, followed by vector DNA purification with the QIAquick PCR clean up kit. The concentration and quality of purified DNA was determined by obtaining an absorbance spectrum on the Nanodrop2000. Samples were normalized in concentration using sample loading solution (SciEx) then heated at 95°C for 5 min followed by placing them on ice. Separation of DNA samples was done using the RNA9000 purity and integrity kit (SciEx) and the PA 800 plus capillary electrophoresis system (SciEx) with laser-induced fluorescence detection. Instrument methods were modified to use pressure injection. Both the 7K ssDNA ladder and all samples are spiked with PhiX ssDNA (NEB) to serve as an internal standard. Relative migration times for all DNA peaks were determined. A graph of relative migration time vs. kb length was fit with a logarithmic curve and used to calculate the size of all DNA peaks in samples. Relative percent of full-length ssDNA species were calculated using corrected peak areas.

Long read next generation sequencing

Long read next generation sequencing (LR-NGS) was performed to assess the size and relative abundance of the encapsidated vector genome of capsid #1 with a 3.1 kb cargo DNA. Pacific Biosciences SMRT sequencing libraries were prepared following the Pacific Biosciences protocol: Preparing multiplexed AAV SMRTbell libraries using SMRTbell prep kit 3.0 version 2 (102-182-700). Viral DNA was purified from one of the DS lots of capsid #1 (with 3.1 kb vg) using the MagMAX Viral/Pathogen Nucleic Acid Isolation Kit (Thermo Scientific Cat. A42352). For each sample 1000 ng of purified viral DNA was used to start the library preparation with the following protocol modification: the incubation time of the “Repair and A-Tailing” step was reduced from 30 min at 37°C to 4 min at 37°C. Data analysis on duplicate runs was performed with open-source packages, and custom Bash and Python scripts.

Additional analysis

Multivariate and other statistical analysis were conducted using JMP18. An ANOVA was performed to determine significance of key parameters impacting DNA integrity. F scores of >10 were statistically significant, with p < 0.05, at a 95% confidence level. R² values (coefficient of determination) were determined using a linear fit analysis to determine correlations between attributes.

To compare degradation kinetics of extracted free FL-ssDNA and encapsidated FL-ssDNA, the data points in Figure 5 were fitted to the first-order rate constant equation (see Figure S7): $\mathit{ln}\mathit{ln}\left[Tx\right]=-kt+\mathit{ln}\left[T0\right]$, where Tx = FL-ssDA abundance at a time point, T0 = FL-ssDNA abundance at zero time point, t = time in days, and k = rate constant in d⁻¹.

Data availability

The data shown in this article are available upon request.

Acknowledgments

The authors give sincere thanks to Spark Therapeutics Analytical Development members, Jason Penera and Dr. Jin Wen, for testing and characterization, Pilot Scale Operations members, Pat Valerio and Cecilia McMahon, for vector production, and Dr. Bashkim Kokona and Karen Baker for their scientific input.

Author contributions

J.S.P. and K.A.M.: first-coauthors, conceptualized, experimented, written, edited, and visualized. Q.L, K.A., A.V., and A.L.D.L.: experimented and edited. M.P. and D.B.: co-corresponding authors, conceptualized, supervised, administrated the project, written, edited, and visualized.

Declaration of interests

J.S.P., K.A.M., Q.L., K.A., A.V., M.P., and D.B. are employees of Spark Therapeutics, a subsidiary of Hoffman-La Roche.