Improving AAV Production Yield and Quality for Different Serotypes Using Distinct Processing Methods

Abstract

Adeno-associated viruses (AAVs) have emerged as a promising tool for gene therapy due to their excellent safety profile and efficient transduction in multiple target tissues. Currently generated AAV yields at lab scale are in the range of 10¹²–10¹⁴ vgc/L or between 10⁴ and 10⁵ vg/cell. Maximizing yields would significantly impact the overall production cost and accessibility. However, existing challenges to improving the overall yield from upstream and downstream processes persist. Using an adherent cell manufacturing process, we compared two different purification approaches to optimize the recovery from cell lysate (CsCl ultracentrifugation) and media supernatant (affinity column chromatography). We achieved combined purified yields of 2.25 × 10¹³ for AAV6 (1.00 × 10¹³ (44.44%) from ultracentrifugation and 1.25 × 10¹³ (55.56%) from affinity chromatography), 1.11 × 10¹⁴ for AAV8 (2.37 × 10¹³ (21.39%) from ultracentrifugation and 8.70 × 10¹³ (78.61%) from affinity chromatography), and 7.32 × 10¹³ for AAV9 (1.10 × 10¹³ (15.03%) from ultracentrifugation and 6.22 × 10¹³ (84.97%) from affinity chromatography), with AAV9 showing a ∼1.5-fold increase compared to previous reports. Interestingly, we found each serotype to elute at different pH values which may influence yields. The chromatography approach obtained higher percentage recoveries for AAV6 (88.35 ± 12.87%), AAV8 (96.67 ± 0.60%), and AAV9 (93.54 ± 1.08%) from the media supernatant and also demonstrated significantly higher transduction rates. Thus, we demonstrate the potential for further expanding the AAV production capacity while also improving the final product quality.

Introduction

Recombinant adeno-associated virus (AAV) vectors are one of the most promising delivery modalities for gene therapy. , In clinical trials, AAVs were successful in delivering various therapeutic treatments for variety of diseases including Leber congenital amaurosis type 2 (LCA2), Parkinson’s disease, Duchenne muscular dystrophy (DMD), hemophilia B, cystic fibrosis, and rheumatoid arthritis. , At present, numerous clinical trials are underway; however, an important challenge for clinical application lies in improving the efficiency and yields for large-scale production methods to obtain a substantial quantity of clinical-grade vectors. Due to their physicochemical characteristics and production process, the purification of AAVs is a lengthy process utilizing various types of technology in a multistep process. The large-scale production process of AAV vectors includes upstream and downstream processing. To facilitate the production of clinical-grade vectors at a large scale, it is necessary to retrieve AAVs from abundant volumes of cell lysate or culture medium. Thus, optimizing the purification method is essential to produce high-volume, clinical-grade vectors.

Cesium chloride (CsCl) density gradient ultracentrifugation is a widely utilized technique in laboratories for purifying various adeno-associated virus (AAV) serotypes. This method relies on the differences in density between the target AAV particles and impurities found in the solution, which primarily consist of cellular organelles, soluble proteins, nucleic acids, and empty vectors. It is crucial to remove these contaminants, as they can be harmful to cells, leading to interference with the transduction efficiency of the purified AAV particles. Furthermore, these impurities have the potential to induce an immune response, which can negatively impact the desired therapeutic or experimental outcomes. − Density gradient ultracentrifugation separates the virus particles from the contaminants based on isopycnic point of AAV virus. Generally, a fair number of impurities are reduced after the first round of CsCl ultracentrifugation, and further purity is obtained by two or three rounds of density gradient ultracentrifugation.

Chromatographic purification is generally found to be more suitable for scaling up vector purification compared to ultracentrifugation methods. Immunoaffinity chromatography (IAC) is a purification technique that utilizes monoclonal antibodies to separate AAVs from other contaminants. Each AAV serotype is defined by differences in their capsid protein sequences and have unique structural aspects that determine their tropism to specific tissues. , IAC therefore utilizes column immobilized antibodies against specific serotypes of AAV or antibodies that bind to conserved elements within the capsid structure that are shared across serotypes to bind to AAV particles in a complex mixture such as cell lysate or culture media. Due to the flow based nature of chromatography, scaling up of this technique is relatively simple. Ultracentrifugation based processes have a significant drawback in terms of scalability since larger volumes cannot be processed by ultracentrifugation, despite being a well standardized method. However, latest developments in continuous flow ultracentrifugation, Alfa Wassermann AW Promatix 1000 seems to be a promising alternative with above 55% of recovery rate in a single step and the linear scale up capacity with rotors up to 50 L illustrating the potential to meet industrial scale demands. Just like ultracentrifugation, IAC has its own limitations. First, use of monoclonal antibodies as ligands to separate AAVs from a large amount of cell lysate and supernatant medium is expensive. Second, the requirement of using strict elution conditions (low pH, denaturing agents, high salts, etc.) to break the strong antibody-protein interactions; extended exposure to such low pH may destabilize the AAV capsid. Lastly, the stability of the ligands over longer or repeated purification cycles needs to be considered. Therefore, we hypothesized that if the ultracentrifugation based process can be included in a workflow to improve overall AAV quality and yield, this may allow for reduction in net production times and cost.

The choice of capsid carrying therapeutic transgene in AAV is determined by the tissue-specific tropism of the desired target. , For example the LCA2 trial assessed AAV2 due to its preference for retinal pigmented epithelial (RPE) cells. For the hemophilia B trial, AAV8 was the serotype of choice as it could transduce hepatocytes , whereas, FDA approved gene therapies like Zolgensma used AAV9 serotype for spinal muscular atrophy (SMA). Hemgenix and Roctavian are among latest AAV5-based gene therapies for Hemophilia B and Hemophilia A respectively. , For body-wide gene delivery across various organ and tissue types, AAV9 is being increasingly preferred for clinical trial applications. Similarly, AAV8 targets multiple tissues with slight differences in distribution, while AAV6 shows a strong preference for the hematopoietic niche. Based on their relevance and potential for large-scale production, particularly for high-dose applications in diseases such as muscular dystrophy and blood disorders, we selected these three serotypes for this study. While the production processes during the early trials did use ultracentrifugation based purification, in combination with tangential flow filtration (TFF) and ion exchange chromatography (IEX), IAC has gained popularity for clinical-grade production more recently. Hence, we asked if the production processes having ultracentrifugation and IAC would have differential purity or physicochemical properties.

Purification techniques are generally based on distinctive properties of the AAV capsid, emphasizing the optimization of capsid-specific purification methods, , thereby complicating the production process. Multiple chromatographic methods have been proposed to purify AAVs effectively. , Therefore, it remains of great interest to determine how media components and process differences impact the purification process. , Most serotype of AAV are secreted in the culture media except for some serotype like AAV2. , An appropriate purification method is essential for obtaining high-titer and pure AAV stocks; an inadequate purification can lead to low-quality AAV preparations. CsCl ultracentrifugation is known for efficiently removing contaminants from nucleic acid preparations. The density gradient helps to separate different components of the cell lysate, leaving behind pure nucleic acids. Chromatography often involves interactions between sample components and the stationary phase. CsCl ultracentrifugation does not rely on such interactions, which can be important in cases where the sample is sensitive to column matrices.

Previous studies have shown recoveries of 1.80 × 10¹⁴ for AAV6 produced in a 3 L HyPerforma glass stirred-tank bioreactor, 3.64 × 10¹⁴ for AAV8 produced in iCELLis Nano bioreactor 0.53 m² surface area, and 5.0 × 10¹³ for AAV9 produced in Hyperflask with 1720 cm² surface area. Based on this background, we investigated the effectiveness of ultracentrifugation and affinity chromatography to purify AAV6, AAV8, and AAV9 carrying the alkaline phosphatase (AP) gene produced in 15 × 150 mm Nunc EasYDish dishes. We tested a strategy of AAV purification to maximize the vector yield by combining CsCl ultracentrifugation and chromatography to purify the AAV from cell lysate and media supernatant, respectively, for three different AAV serotypes. We then assessed the purity, capsid protein integrity, and in vitro transduction efficiency of the vectors purified. TEM analysis was conducted to visualize the morphology and distribution of AAV among the purified fractions. The results shed light on the effect of sample source, AAV serotype, and sample processing on optimizing the purification yields, thereby enhancing the progress of our early phase biomanufacturing establishment.

Experimental Section

All of the experiments were done in triplicates.

Cell Culture

HEK293T cells were cultured in 15 × 150 mm Nunc EasYDish dishes (Thermo Fisher Scientific, cat. no. 150468) in Dulbecco’s modified Eagle’s medium (DMEM) (GIBCO, cat. no. #11965-092) with 10% fetal bovine serum (GIBCO, cat. no. 10270-106) and 1% antibiotic–antimycotic (GIBCO (Anti-Anti 100×) cat no. 15240-062). The cells were maintained in 5% CO₂ saturation at 37 °C.

Recombinant AAV6, AAV8, and AAV9 Production and Purification

AAV6, AAV8, and AAV9 stocks were generated using an adenoviral-free triple transfection protocol using the calcium phosphate method as described previously. Briefly, 70–80% confluent HEK293T cells were co-transfected with a cis-plasmid (pcisRSV.AP), a pRep2/Cap9 helper plasmid, and an adenoviral helper plasmid at a ratio of 1:3:3. Post 72 h, the supernatant media and crude cell lysate were collected. Both AAV containing samples were treated separately with 375 μL (1875 Kunitz units) of DNase I (Sigma, cat. no. D4513) and incubated at 37 °C for 1 h, and after 0.45 μm filtration (Millipore, cat. no. HAWP04700) AAVs were purified using affinity chromatography. Correspondingly, the cell lysate was purified through CsCl isopycnic ultracentrifugation (Beckman Coulter Optima XPN 100, USA). Purified AAV fractions were then dialyzed through multiple exchanges of N-(2-hydroxyethyl)­piperazine-N′-ethanesulfonic acid (HEPES) buffered saline at 4 °C and subsequently stored in a storage buffer at −80 °C until further use. Viral titer was determined through slot blot hybridization by using an AP probe. The viral genome (vg) copy number was determined according to plasmid copy number standards. The cis-plasmids (pcisRSV.AP) have been reported previously, pcisRSV.AP expresses the heat-resistant human placental AP gene under the transcriptional control of the RSV promoter and the SV40 polyadenylation signal (pA).

Ultracentrifugation Purification

Cell lysate was thawed and resuspended in 12 mL of 10 mM Tris buffer (pH 8.0) in a 15 mL tube. Freeze–thaw cycle was done using a dry ice-ethanol bath at −80 °C for 8 cycles to facilitate cell lysis. Sample volume was then adjusted to 18 mL using 10 mM Tris in a 50 mL tube. The lysate was then sonicated on ice using 30 s pulses (30 s on, 30 s off) for 12 cycles. 375 μL of DNase I was added and incubated at 37 °C for 45 min. An additional round of sonication was performed under the same conditions as those before. To further facilitate lysis, 1–5 mL of 10% sodium deoxycholate and 3 mL of 0.25% trypsin were added to the sample, followed by a 30 min incubation at 37 °C. 16.9 g of CsCl was added and incubated for 30 min at 37 °C. 300 μL of 10% sodium deoxycholate was added, and the final volume was increased to 32 mL with 10 mM Tris. Lysed sample was allowed to incubate at room temperature (RT) for 30 min before being centrifuged at 3100 rcf for 30 min at 4 °C to remove cellular debris. The resulting viral suspension was carefully transferred, and 4.9 mL aliquots were loaded into clear ultracentrifuge tubes for subsequent ultracentrifugation. Samples were spun at 194,432 rcf for 40 h at 4 °C. Post first round of ultracentrifugation, virus fractions were collected in 15 × 1.5 mL tubes. Slot blot was performed to estimate the initial titer, the high-titer fractions were pooled, and a second round of ultracentrifugation was conducted. Finally, purified AAV fractions were collected, and high-titer fractions were pooled then dialyzed and stored at −80 °C in appropriate storage buffer.

Affinity Column-Based Chromatography Purification

Media supernatant approximately 270 mL, for each AAV serotype (AAV6, AAV8, and AAV9) was collected, DNase treated, and filtered using 0.45 μm filter. Purification was performed on AKTA Pure 25 chromatography system (Cytiva, USA), and Thermo Scientific POROS CaptureSelect AAVX column (5 mL) was used for affinity separation. The AAV serotypes containing the AP reporter gene were processed under consistent conditions, maintaining a flow rate of 1 mL/min. We used 1× phosphate-buffered saline (PBS), pH 7.4, throughout the process (equilibration, sample loading, and column wash). 0.1 M citrate buffer with 300 mM sodium chloride (NaCl), pH 3.0, was used for the elution. For immediate neutralization, a 1 M Tris–HCl solution, pH 9.0, was used. Collected fractions were dialyzed in pre-autoclaved and refrigerated (4 °C) 1× PBS pH 7.4, and genomic titers were then estimated using qPCR and slot blot analysis. All of the experiments were conducted under ambient room temperature (∼25 °C) conditions.

Slot Blot Analysis

Slot blot was performed essentially as described previously. − Viral DNA was extracted and hybridization was performed using a DIG-labeled DNA probe. Briefly, 1 μL of virus obtained post dialysis was blotted to a positively charged nylon-6,6 transfer membrane from Pall life science (Cat no. 60207) by using a slot blot manifold, and then the membrane was UV cross-linked at 120 mJ. The blot was hybridized for a positive signal by using an AP probe. The vector genome was determined according to the plasmid copy number standards. The blot was developed according to the manufacturer’s protocol (DIG High Prime DNA Labeling and Detection Starter Kit I by Roche).

qPCR Analysis for Titer Estimation of AAV in Purified Fractions

A standard curve using the transgene plasmid (pcisRSV.AP) was prepared for every qPCR run. Standards were serially diluted from 1 × 10⁸ to 1 × 10³ vg/μL. AAV viral sample was diluted (1:10) in slot blot buffer (10 M NaOH, 0.5 M ethylenediaminetetraacetic acid (EDTA)) and heated at 95 °C for 10 min followed by 2 min of ice incubation. The sample was then serially diluted to 1:100 and 1:1000 in 10 mM Tris buffer. qPCR reaction was performed in a 96-well plate (Bio-Rad Hard Shell 96 well PCR Plates HSP9601) in a CFX Connect real-time PCR detection system (Bio-Rad, Philadelphia, PA, USA). 23 μL portion of master mix was distributed in each well with 2 μL of prediluted standards and sample. AP was the gene of interest, whose expression was analyzed to calculate the viral genome (vg) numbers.

Visualizing Protein Impurity and Capsid Integrity Analysis

Coomassie (CBB) staining and Western blot (WB) analysis were carried out for analyzing protein impurity as described previously. Briefly, an equal concentration of the AAV (1.0 × 10¹¹ vg) were separated on 10% sodium dodecyl sulfate (SDS-PAGE) and stained for CBB and transferred onto a poly­(vinylidene difluoride) (PVDF) membrane for WB analysis, 3 μL of prestained protein ladder (Abcam, Cat no # ab116028) was loaded for molecular weight comparison. The membrane was blocked in 5% skimmed milk in TBST for 1 h at room temperature (RT) followed by probing with the primary antibody Anti-AAV VP1/VP2/VP3; Clone B1 (1:1000) (Progen, catalog no. 61058) at 4 °C overnight. The membrane was washed with 0.25% TBST and probed with a secondary antibody (1:5000) for 2 h at RT. Protein bands were developed using an enhanced chemiluminescence detection kit (Pierce ECL Plus; Thermo Scientific) and analyzed with the Image Quant LAS 500 chemiluminescence detector (GE Healthcare Life Science, Uppsala, Sweden).

Silver Staining

Silver staining is known to be more sensitive below 10 ng concentration of proteins to visualize the impurities that cannot be resolved with CBB stain. An equal concentration of the AAV (1.0 × 10¹¹ vg) was separated by 10% SDS-PAGE. A fixation solution containing glacial acetic acid was added to the gel, and the mixture was incubated for 30 min. The gel was then rinsed with 20% ethanol, followed by a distilled water rinse. The gel was incubated in sensitization solution (10% Na₂S₂O₃) for 1 min followed by rinsing in distilled water. After this, AgNO₃ solution was added to the gel and was incubated for 30 min in the dark until a yellowish color is developed. The gel was rinsed with distilled water and developer solution was added. Quickly, after this, the stop solution was added and the gel was kept in it for 30 min after which the bands were visualized.

Agarose Gel Electrophoresis

2% agarose gel was prepared in 1× TAE buffer with 0.1% ethidium bromide (EtBr). The gel was immersed in a 1× TAE buffer and was run at 85 V for 2 h. An equal concentration of AAV (1.0 × 10¹¹ vg) was loaded into each well.

Transduction

3 × 10⁵ HEK293T cells were seeded in 6 well plates. Cells were seeded in DMEM supplemented with 10% FBS and 1% antibiotic. At 70% confluence, transduction was performed in a serum-free DMEM medium. Cells were transduced at different multiplicity of infection (MOIs) (5K, 10K, 25K, 50K, and 100K) with respective AAV serotypes purified by ultracentrifugation and affinity chromatography. Briefly, definite volumes of the viruses corresponding to the respective MOIs were mixed in 1 mL of serum-free DMEM media and were added to the wells gently. After 3 h, 1 mL of DMEM medium supplemented with 20% FBS and 1% antibiotic were added to each well. Transduction was carried out for 72 h.

RNA Extraction and qPCR to Check the Expression of the AP Gene

RNA extraction was performed using the TRIZOL reagent (Ambion, Carlsbad, CA, USA). RNA was quantified using Nanodrop spectrophotometer 1000 (Thermo Scientific, Wilmington, USA), and complementary DNA was made using a high-capacity cDNA Reverse Transcription Kit, (Life Technologies, CA, USA). Quantitative real-time PCR was done by using KAPA SYBR FAST qPCR Master Mix (Kapa Biosystems, Wilmington, MA, USA), and analyzed on a Bio-Rad CFX Connect (Bio-Rad, Hercules, CA, USA). For AP gene expression, actin was used as a housekeeping gene, and mRNA fold change was calculated by using the 2^–ΔΔCT method. The primers used are available in “Supporting information: Table S1List of primers, includes forward and reverse primer sequence details (PDF).”

Alkaline Phosphatase (AP) Activity Assay

The alkaline phosphatase activity assay was performed according to the manufacturer’s protocol (StemTAG AP activity assay kit; Cell Biolabs Inc., cat no. CBA-301). The transduced HEK293T cells were lysed in 100 μL of cell lysis buffer, as provided by the kit. The cell lysate was centrifuged at 17,530 rcf for 10 min, and the supernatant was collected. The protein concentration was determined using the Bradford method. 20 μg of protein supernatant was diluted with ddH₂O so that 50 μL of cell lysate was added to each well of a 96-well plate. In addition, blank wells were prepared by adding 50 μL of a cell lysis buffer. The reaction was initiated by adding 50 μL of StemTAG AP activity assay substrate and was incubated for 30 min at 37 °C. The reaction was stopped by adding 50 μL of 1× Stop Solution and the plate was mixed by placing it on an orbital plate shaker for 30 s. The absorbance of each well was read at 405 nm. AP activity was determined according to a para-nitrophenol standard curve. The AP activity is defined as μM para-nitrophenol being generated per μg of total protein in the tissue lysate (one molecule of para-nitrophenol is generated when one molecule of phosphate is released from the substrate).

Transmission Electron Microscopy (TEM)

Sample Preparation

10 μL sample of purified fractions of adeno-associated virus (AAV) was used, with an approximate concentration of 1.0 × 10¹³ vg/mL. These fractions were subjected to negative stain and TEM analysis was performed at the Nanomaterials and Energy Lab, IISER, Mohali, India. The samples were added and allowed to adsorb onto the surface of a glow-discharged 300 mesh Formvar/carbon-coated grid (Sigma, cat no. 930253). It was subsequently blotted off using Kimwipes and stained with 1–2% uranyl acetate. After 50–60 s, the excess stain was blotted off, and the sample was air-dried before imaging in a transmission electron microscope.

Image Acquisition

JEM-F200 (JEOL Ltd., Japan) was used for recording high resolution images for the analysis with Schottky FEG operated at an accelerating voltage of 200 kV. A minimum of 5 images were taken per sample at different magnifications between 30K× and 80K×. To confidently differentiate between full, empty, and other capsid subtypes from the electron micrographs, the criteria described by Fu et al. were used for counting the virions. Two researchers, not informed of the identity of the samples, systematically counted the AAV particles in each image. Percentage of empty, full, and other subtypes were averaged based on the resulting counts. During the counting process, any regions within the images where individual particles could not be confidently distinguished were excluded from the total count. Additionally, areas that contributed to image noise or artifacts, such as dark spots that hindered clear classification of the virion type, were not considered in the analysis.

Capsid Ratio Calculation

The ratios of full, empty, broken, and multimeric forms of AAV capsids were annotated and counted from the micrographs captured. The percentage of each category was derived against the total number of identified capsids in the frame from three separate images and plotted using GraphPad Prism 9.

Statistical Analysis

GraphPad Prism 9 software was used for the statistical analysis. Student t test, one-way analysis of variance (ANOVA) with Kruskal–Wallis test and, 2-way ANOVA with Tukey’s multiple comparison test were performed; all of the experiments were repeated in triplicates. Data are represented as mean ± standard error of the mean (SEM) and significant differences are indicated as *p ≤ 0.05, **p ≤ 0.01, ***p ≤ 0.001, and ****p ≤ 0.0001.

Results

Our results indicate successful recoveries of AAV using crude cell lysate through ultracentrifugation across three replicates. The recovery ranges expressed as yield per cell were 3.20 × 10⁴ to 3.47 × 10⁴ (4% CV) for AAV6, 6.67 × 10⁴ to 1.00 × 10⁵ (23.27% CV) for AAV8, and 3.33 × 10⁴ to 4.00 × 10⁴ (9.09% CV) for AAV9. In comparison to this, we achieved >88% recovery using the affinity chromatography method with estimated yields per cell ranges from 3.33 × 10⁴ to 5.07 × 10⁴ for AAV6, 2.00 × 10⁵ to 3.70 × 10⁵ for AAV8, and 1.83 × 10⁵ to 2.52 × 10⁵ for AAV9, as shown in Figure C. We observed high absorbance (∼600 mAU) values in the flow-through indicating the presence of media-related protein impurities that could add to sample complexity and may possibly hinder AAV binding (Figure A). Despite this, we achieved higher recovery of AAV vectors in the elution fractions with moderate variability in recovery, as measured by qPCR and slot blot analysis (Figure B). The average loading titers of AAV6, AAV8, and AAV9 were 1.21 × 10¹⁴, 2.63 × 10¹⁴, and 4.15 × 10¹⁴, respectively, reflective of the efficacy of packaging and produced yield being inherently different between the different serotypes. Using 270 mL of media supernatant, we recovered 88.35% (27.07% CV) for AAV6, 96.67% (1.06% CV) for AAV8, and 93.54% (2% CV) for AAV9 (Table S4). In our experiments, we got a slightly lower recovery for AAV6 as compared to the other two serotypes. This trend is also reflected in the elution peak profiles in Figure C, where AAV8 showed the highest elution peak followed by AAV9 and AAV6. The genomic titer estimation revealed affinity purification was able to capture higher amounts of AAV (Figure B) with superior titers for AAV8 8.70 × 10¹³ and AAV9 6.22 × 10¹³ compared to AAV6 1.25 × 10¹³ (Figure B). The ultracentrifugation purification for cell lysate revealed genomic titer for AAV6 1.0 × 10¹³, AAV8 2.36 × 10¹³ and AAV9 1.1 × 10¹³. Serotypes AAV8 and AAV9 exhibited higher yields per cell when compared to AAV6 (Figure C).

2.

Viral genome titer estimation of AAV6, AAV8, and AAV9 after purification through ultracentrifugation and chromatography. (A) Schematic representation of the experiment performed. (B) Comparison of estimated genomic titer (vgc/mL) and the combined titer yields (vgc/mL) for AAV6, AAV8, and AAV9. Statistical analysis was conducted using 2-way ANOVA with Tukey’s multiple comparison test, with significant differences indicated as ***p ≤ 0.0005, ****p ≤ 0.0001, and ns for nonsignificant results. For combined genomic titer yields, one-way ANOVA with Kruskal–Wallis test was performed, revealing significant differences marked as *p ≤ 0.0107, along with ns for nonsignificant results. (C) Graphical representation of yield per cell for the lysate and media supernatant across the three serotypes. Statistical comparisons was conducted using one-way ANOVA, and the Kruskal–Wallis test revealed significant differences indicated by *p ≤ 0.0405 for the lysate and *p ≤ 0.0338 for the supernatant, with ns denoting nonsignificant results.

1.

Chromatography run profiles from AAVX column at various steps of purification. (A) Flow-through, (B) wash, and (C) elution, triplicate runs are compiled here and represent the binding, washing, and recovery patterns for all three serotypes used (AAV6, AAV8, and AAV9).

We then calculated average yields recovered per cell for AAV6, AAV8 and AAV9 vectors shown in “Supporting information: Figure S1bar graph represents the obtained yield per cell post purification from cell lysate and supernatant of the three AAV serotypes used and Table S3average titer of AAV vectors and calculated yield per cell obtained using cell lysate and media supernatant samples post purification (PDF)”. Furthermore, we noticed that peak release point for each serotype arrives at different pH AAV6 at 3.52 ± 0.0521, AAV8 at 4.51 ± 0.0120, and AAV9 at 4.93 ± 0.0561, as outlined in Table , indicating their inherent structural differences in the epitopes influencing their binding. This variation in the binding strength results in differences in the peak release for each serotype. To achieve optimal retrieval of a specific serotype, the elution conditions must be optimized before scaling up the process.

1. Different Parameters Recorded during the Affinity Column Chromatography-Based Purification of AAV6, AAV8, and AAV9 Serotypes.

serotype	pH (initial)	pH (peak elution)	conductivity (at peak)
AAV6	6.36 ± 0.0694	3.52 ± 0.0521	13.07
AAV8	6.45 ± 0.1506	4.51 ± 0.0120	13.71
AAV9	6.62 ± 0.0586	4.93 ± 0.0561	13.07

However, our recovery yields and percentage efficiency were comparatively higher than the previously published reports where the authors reported between 65 and 80% of purification efficiency for multiple serotypes; − thus, our findings support that this affinity column can be effectively utilized for the recovery of many other AAV serotypes.

Assessment of Capsid Integrity and Genomic DNA Impurity in Purified Fractions

To check the capsid integrity, the purified viruses by ultracentrifugation and affinity column chromatography were evaluated for the presence of VP1, VP2, and VP3 proteins by CBB staining, WB, and silver staining (Figure B–D). Based on the densitometry analysis of the WB, the estimated capsid proteins VP1, VP2, and VP3 bands are present in the ratio of 1:1:6.54. WB analysis showed that ultracentrifuged purified viruses had no nonspecific bands (impurities) as compared to chromatography-purified vectors (Figure C) indicating the absence of broken capsids, and other partially filled capsids that coelute in chromatography fractions. We then quantified the nonspecific bands through densitometric analysis, and the impurity ratio was found to be 9.36% for AAV6, 4.61% for AAV8 and 8.25% for AAV9 (Figure D). Subsequently, the gel was stained with CBB where only the VP bands were prominently seen and did not show any impurities. However, when the gel was exposed to silver staining, the affinity-purified AAV’s showed prominent VP bands with few other impurities which can be observed above and below VP3 bands (Figure B–D). It was noted that no DNA bands were detectable on the agarose gel; this observation indicates the absence of DNA impurity in the purified vector fractions through both methods (Figure E).

3.

Capsid protein integrity and host cell DNA impurity in purified AAV vectors. (A) Overview of experimental design of figures (B), (C), and (D). (B) Coomassie-stained SDS-PAGE showing the VP1, VP2, and VP3 capsid proteins of AAV vectors. (C) Immunoblot showing the VP1, VP2, and VP3 capsid proteins of AAV vectors. (D) Silver-stained SDS-PAGE showing the VP1, VP2, and VP3 capsid proteins of AAV vectors. Equal titers (∼1 × 10¹¹ vg/lane) of the AAV vectors were loaded in each well. (E) Analysis of 2% agarose gel confirms the absence of genomic DNA contamination in AAV vectors purified through ultracentrifugation and chromatography. Each well was loaded with AAV vectors of equal titers (∼1 × 10¹¹ vg/lane).

6.

Transmission electron microscopy images of AAV vectors. (A) Purified AAV vectors from ultracentrifuge AAV6, AAV8, and AAV9; (B) chromatography-purified AAV vectors AAV6, AAV8, and AAV9. (C) Percentage distribution of AAV capsids from ultracentrifuge and chromatography, depicted here as AAV Ultra and AAV Chrom. (D) Comparison of the ratio of impurity percentages identified using TEM, Western blot (WB), and silver stain (SS).

Transduction Efficiency of AAV6, AAV8, and AAV9 Purified through Ultracentrifugation and Chromatography

Since AP was the transgene present in AAV6, AAV8, and AAV9, we checked the transduction efficiency of the purified AAV vectors for which HEK293T cells were transduced with different MOIs of AAV6, AAV8, and AAV9 purified through ultracentrifugation and chromatography. Through AP activity assay, it was observed that AAV vectors purified through chromatography using affinity column displayed significantly higher AP enzymatic activity for all of the AAV serotypes across all of the MOIs compared to the ultracentrifuged purified AAV vectors (Figure B–D). In the case of AAV8, for 5K MOI, no significant difference was noticed between ultracentrifuge and chromatography-purified AAV8 (Figure C). Similarly, AP gene expression analysis through qPCR showed significantly higher AP gene expression in cells transduced with AAV vectors purified through chromatography compared with vectors purified through ultracentrifugation across different MOIs (Figure B–D). For 5K MOI, no significant difference in AP gene expression was observed for all of the AAV serotypes purified through chromatography and ultracentrifugation (Figure B–D). Therefore, through this result, it was inferred that AAV vectors purified through chromatography have a better capability of transduction compared to AAV vectors purified through ultracentrifugation.

4.

Alkaline phosphatase (AP) enzymatic activity. (A) Schematic representation of the protocol for the AP enzymatic assay. (B–D) Graphical representation of quantification of AP enzymatic activity in HEK293T cells transduced with different MOIs of ultracentrifugation and chromatography-purified AAV6 vector (B), AAV8 vector (C), and AAV9 vector (D). Student t test was performed. All of the experiments were repeated in biological triplicate. Data are represented as mean ± SEM and significant differences are indicated as *p ≤ 0.05, **p ≤ 0.01, ***p ≤ 0.001, and ****p ≤ 0.0001 (UTuntreated, K = 1000, MOImultiplicity of infection).

5.

Relative Alkaline phosphatase gene expression. (A) Schematic representation of the protocol for the AP gene expression. (B–D) Graphical representation of Alkaline phosphatase gene expression in HEK293T cells transduced with different MOIs of ultracentrifugation and chromatography-purified AAV6 vector (B), AAV8 vector (C), and AAV9 vector (D). Student t test was performed. All of the experiments were repeated in biological triplicate. Data are represented as mean ± SEM and significant differences are indicated as *p ≤ 0.05, **p ≤ 0.01, ***p ≤ 0.001, and ****p ≤ 0.0001 (UTuntreated, K = 1000, MOImultiplicity of infection).

Transmission Electron Microscopy (TEM) Analysis

TEM images were analyzed from the purified fractions obtained through two methods, ultracentrifugation and chromatography, revealing the structural morphology (Figure A,B) and percentage distribution (Figure C) of the AAV capsids. The distribution of full, empty, broken, and multimeric forms of AAV capsids was determined by analyzing a minimum of 3 different TEM images. Based on our visual classification, empty AAV capsids gets stained with uranyl acetate and can be identified by their electron-dense centers. In contrast, full capsids remain unstained due to the exclusion of the dye by the encapsidated DNA, appearing as bright hexagonal particles. Partially filled capsids exhibit a dimmer center compared to full capsids with a dark hollow ring surrounding the genetic material. , Broken capsids are easier to distinguish by their incomplete boundary and appear as bisected capsid portions. Multimeric capsids appeared like aggregates that shared common boundaries; the adjacent particles may contain genetic material or be devoid of it. Each category was counted separately, and the percentages were calculated relative to the total number of identified capsids (Figure C).

In the ultracentrifuge purified fractions, the AAV6 sample was found to contain 84.6% full, 7.7% empty, 3.8% partially filled, 1.3% broken, and 2.6% multimeric forms. AAV8 comprised 78.1% full, 6.8% empty, 5.5% partially filled, 4.1% broken, and 5.5% multimeric forms. AAV9 consisted of 81.3% full, 10.7% empty, 5.3% partially filled, 1.3% broken, and 1.3% multimeric forms (such as dimers or oligomers).

For the affinity chromatography method, the percentages of AAV6 were as follows: 87.2% full, 3.8% empty, 5.1% partially filled, 2.6% broken, and 1.3% multimeric forms. AAV8 had 88% full, 4% empty, 3.3% partially filled, 2% broken, and 2.7% multimeric forms. AAV9 showed 87.6% full, 2.6% empty, 5.2% partially filled, 2.6% broken, and 2.1% multimeric forms (such as dimers or oligomers). “Supporting information: Table S2Percentage of capsid subtypes visualized and quantified using TEM images (PDF).”

Based on the full to empty ratio, we can compare the purification methods in terms of their efficacy in obtaining a higher proportion of intact (full) AAV capsids. For AAV6, the ultracentrifuge purification method resulted in a full/empty ratio of approximately 11:1 (84.6% full to 7.7% empty). On the other hand, AAV6 purified using the affinity chromatography method had a higher full to empty ratio of approximately 23:1 (87.2% full to 3.8% empty). Therefore, the affinity chromatography method appears to be more effective in obtaining samples with intact AAV capsids. For AAV8, the ultracentrifuge purification method yielded a full to empty ratio of approximately 11.4:1 (78.1% full to 6.8% empty). Similarly, the affinity chromatography method resulted in a full/empty ratio of approximately 22:1 (88% full to 4% empty). Once again, the affinity chromatography method demonstrated a higher full to empty ratio, indicating its superior proportions of full AAV capsids. For AAV9, the ultracentrifuge purification method gave a full to empty ratio of approximately 7.6:1 (81.3% full to 10.7% empty). The affinity chromatography method for AAV9 resulted in a full:empty ratio of approximately 33.6:1 (87.6% full to 2.6% empty).

Thus, the affinity chromatography method shows a significantly higher full to empty ratio, indicating its effectiveness in purifying a larger percentage of intact AAV capsids. Based on the full to empty ratio, it is shown that the affinity chromatography method consistently outperforms the ultracentrifugation method in terms of obtaining a higher proportion of full AAV capsids across all three serotypes.

Discussion

Substantial developments in upstream production of viral vectors have added more challenges in the downstream process of manufacturing AAV vectors. With rise in number of clinical trials using AAV vectors for gene therapy, there is an utmost need to develop methods for production and purification of large volumes while maintaining high purity. The major bottleneck is to recover vectors from larger volumes of cell lysate or medium, due to their complex nature of composition as starting material for purification. Apart from producing vectors of high titer, high potency, and high purity, maintaining the biological activity, it is essential to remove active impurities and contaminants present in the crude sample. Several methods to purify viral vectors have been reported, but they either lack scalability or require multiple purification steps that ultimately lead to low efficiency in terms of final yield. Multiple filtration and other purification steps are known to have some amount of loss while processing. Traditional CsCl density gradient centrifugation alone is not suitable for large-scale production. Therefore, it is imperative to optimize techniques apt for large-scale AAV vector production. The CsCl ultracentrifugation for purification of AAV vectors is time-consuming with the entire method taking approximately 3–4 days while purification through chromatography using the affinity column reduces time significantly. Setting up an effective downstream platform for AAV vectors becomes important to enhance productivity and meet industrial requirements. The use of affinity chromatography can be a scalable technique that can lead to a high-fold purification of AAV vectors whereas scaling ultracentrifugation, while feasible with recent advancements, is yet to become a mainstream application. Therefore, using a pan-serotype column for the purification of AAV serotypes could be a promising approach. AAV lysate typically contains a complex mixture of viral particles, host cell proteins, nucleic acids, and other debris related impurities. This complexity can pose difficulties in achieving effective separation and purification using chromatographic techniques. To ensure the maximum recovery in the chromatography step, biodeburdening of the sample plays a crucial role. This can be achieved at a large scale, by utilizing size cutoff membranes and depth filters of various pore sizes ranging from 2 to 50 μm; these filters effectively reduce host cell proteins, free floating plasmids, cellular debris, etc. In continuation, a tangential flow filtration (TFF) step is applied to concentrate and perform buffer exchange, which allows for effective reduction in the higher volumes produced in manufacturing. Apart from this, chromatography matrices also have a finite binding capacity for target molecules. AAV lysate contains a high concentration of viral particles along with the impurities, and the limited binding capacity of the column can lead to incomplete purification or loss of target AAVs. Some of these challenges can be overcome by density gradient ultracentrifugation since it physically separates the AAV capsids from all other debris. Considering several challenges in the purification process, the purpose of this study was to improve AAV recovery yields for the three serotypes (AAV6, AAV8, and AAV9) by combining multiple approaches across sample types for purification.

Affinity column-based purification gave us the best elution for AAV8 followed by AAV9 and AAV6. Elution pH varies for each serotype as AAV vectors showed differences in their elution behavior. In agreement with previous reports, for elution, reducing the pH to 2.5–3.0 is generally successful but may vary with AAV serotypes necessitating serotype specific optimizations. Different capsid serotypes have slight differences in the overall structure and electrostatic net charges, resulting in differences in their isoelectric points. As per earlier reports, , we used a pH of 3.0 using 0.1 M citrate buffer containing 0.3 M NaCl for the elution, with obtained recovery percentages of 88.35 ± 12.87, 96.67 ± 0.60, and 93.54 ± 1.08% for AAV6, AAV8, and AAV9 respectively. Through slot blot and qPCR analysis, it was observed that the media supernatant of AAV serotypes purified through affinity column chromatography displayed significantly better genomic titers compared to those obtained by the ultracentrifugation method. The total combined yield for three serotypes purified through both methods was calculated to be 2.25 × 10¹³ (12.66% CV) for AAV6, 1.11 × 10¹⁴ (24.60% CV) for AAV8, and 7.32 × 10¹³ (15.91% CV) for AAV9. However, we urge careful consideration before pooling the vectors produced by the two processing methods. Mixing the two purified fractions of differing vector content or quality may lead to unfavorable outcomes, regarding the final quality. Furthermore, since both fractions contribute to the overall volume, this could dilute the total yield, although an additional concentration step may rescue this issue.

The structural integrity of viral capsids was determined by studying the expression of the capsid proteins VP1, VP2, and VP3 by immunoblot analysis. , On checking the expression of VP1, VP2, and VP3 proteins was checked, we noticed that vectors purified through both methods had good capsid protein expression, with chromatographic fractions showing higher expression of the VP proteins. For orthogonal comparison through WB with TEM images, we observed that AAV6 has approximately 5.13% partially filled capsids and 2.56% broken capsids, AAV8 has approximately 3.33% partially filled capsids and 2.0% broken capsids, and AAV9 has approximately 5.18% partially filled capsids and 2.6% broken capsids. The extra nonspecific bands visible in the chromatography-purified lanes of AAV6 and AAV9 than AAV8, suggest the presence of partially filled, broken or unformed capsid polypeptides in the chromatography-purified fractions (Figure C). As reported previously, these nonspecific bands in the chromatography fractions which could be due to presence of broken fragments or from partially filled capsids, etc. On comparing the impurity ratios obtained from TEM, CBB, and silver stain data by densitometric analysis, we also observed a consistent impurity in the affinity-purified fractions across serotypes, characterized by an extra band present above VP3 (Figure C,D). This band may represent broken fragments or partially filled capsids and can be detected in both silver stain and WB. However, the other bands below VP3 in Figure C are also the impurities generated from fragmented capsids, identifiable solely by WB. Moreover, these bands were not visible in CBB or silver-stained gels. This finding confirms that these impurities arise from the unintended capture of capsids that coelute with the desired full AAV capsids. We further quantified the percentage of capsid impurities observed in WB and compared it with the ratios obtained from TEM and silver staining (Figure D), finding that the ratios identified by these three methods were relatively similar. Broken capsids can be caused as a byproduct of vector production processes, improper storage conditions, freeze–thaw cycles, inefficient neutralization, etc., which have the potential to contribute toward anticapsid immune response in patients. Previously, for AAV6, AAV8 and AAV9 vectors full capsid ratios were reported between 11 and 61% which were further polished using methods like size exclusion chromatography (SEC), ion exchange chromatography (IEX), etc. ,, However, in our process we recovered significantly higher (above 87.6%) full to empty capsid ratios across all three serotypes used, thus enabling us to exclude polishing steps and reduce production times. Although our experimental production process did not have it, we recommend to incorporate a polishing step such as ion exchange or size exclusion chromatography, which can significantly improve the final product quality for clinical applications by removing the trace impurities present. Further, final product purity may be verified using additional orthogonal approaches like enzyme-linked immunosorbent assay (ELISA), high-performance liquid chromatography-multi-angle light scattering (HPLC-MALS), capillary electrophoresis, and mass spectrometry when considering clinical applications.

The transduction efficiency of a given AAV serotype is decided by the effectiveness of each of the different steps in the AAV life cycle. For the development of an optimal gene transfer vector, a good transduction efficiency of the AAV product is important. Interestingly, alkaline phosphatase enzymatic assays showed that affinity chromatography-purified vectors had significantly better transduction efficiency compared to ultracentrifugation, with AAV6 being the most efficient. The same was also confirmed by a gene expression analysis. This unique observation brings forth important considerations regarding the potency of the final product. For 5K MOI, no significant difference in AP gene expression was observed for all of the AAV serotypes purified through chromatography and ultracentrifugation. A successful transduction depends on the concentration of the virus and not the overall number of virions present. The vector mediated transduction process determines the number of viable vector particles (or vector titer) in a given vector stock and can be influenced by various factors which are poorly understood. Thus, the results suggest that pooling the lysate and supernate-derived vectors, while increasing yield, may negatively impact potency. Limitations of this observation are that we do not understand the cause, and we do not know if such observations hold true in vivo. We hope to further explore these aspects in future studies. It is important to note that the original processes used for AAV purification for Luxturna used ultracentrifugation, which provided functional visual recovery and did not cause severe adverse events. While many laboratory experiments continue to use both ultracentrifugation or chromatography-based processes for AAV gene therapy, comparisons between intracellular and secreted forms of AAV in vivo have not yet been described. Therefore, in the event our observation is not as pronounced in vivo in tissues, improving the yield by combining the lysate and supernatant will remain important. Nonetheless, our observations open up a new discussion regarding the capsid structure or surface modifications that may have led to the transduction differences despite being of similar quality.

Conclusions

Our novel findings suggest that using the combination of lysate and supernatant samples purified through ultracentrifugation and affinity chromatography respectively, can lead to higher AAV yields reaching >3.0 × 10⁵ vg/cell in serotypes 8 and 9. Parameters like high genomic titer, absence of host DNA impurities, absence of host protein impurities, AAV capsid structural integrity and distribution of expected capsid protein ratio in WB indicate high quality of purified AAV vectors in media supernatant. Our production process demonstrated consistently high filled capsid ratios >87% across serotypes. IAC recovery of >88% illustrates that the purification process and serotype specific elution conditions yield higher recoveries. While combining the AAV capsids purified from both cell lysate and culture supernatant improves net yields, differences in potency suggests caution and re-examination of the quality parameters. Our findings pave the way for new considerations and improvements in scaling up of AAV production process.

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acsomega.4c10900.

• List of primers used, including forward and reverse primer; percentage of capsid subtypes visualized and quantified using TEM images; average titer of AAV vectors and calculated yield per cell obtained using cell lysate and media supernatant samples post purification; data represent the average of total viral genome copies obtained for the mentioned volume of samples and percentage recovery of AAV6, AAV8, and AAV9 serotypes obtained after purification through affinity chromatography, based on triplicate experiments conducted (±represents SEM values); bar graph represent the obtained yield per cell post purification from cell lysate and supernatant of the three AAV serotypes used; (A) slot blot image and table representing genomic titer estimation (titer/μL) of post dialyzed samples of AAV6, AAV8, and AAV9 purified using affinity chromatography; (B) Slot blot image and table representing genomic titer estimation (titer/μL) of post dialyzed samples of AAV6, AAV8, and AAV9 purified using ultracentrifugation; percentage contribution of AAV recovery by ultracentrifuge and affinity chromatography method (PDF)

§.

A.K., R.P., and S.R. share equal authorship. A.K. conceptualized, drafted and edited the manuscript, planned and conducted the experiments, performed data analysis, analytical experiments, and TEM analysis. R.P. was responsible for AAV production, ultracentrifugation, and in vitro experiments. S.R. carried out gene expression studies, AP assays, as well as contributed to the initial manuscript draft. S.B.G.R. edited the manuscript, A.G. conceptualized and edited the manuscript.

The study was supported by an unrestricted institutional grant from NNF (GT-PD) and from India Alliance (IA/TSG/20/1/600029) to A.G.

The authors declare no competing financial interest.