Suspension Culture of HEK293 Cells for the Production of GMP-Grade AAV: Suspension Culture and Continuous Medium Harvesting Process for FIX and FLT1 Clinical Vectors

Aden-associated virus （ AAV ） has been widely used in numerous gene therapy clinical trials due to its advantages, including long-term transgenic expression, non-pathogenicity, the ability to transduce both quiescent and dividing cells, and the absence of the need for autonomous replication. However, the clinical application of AAV remains constrained by the lack of a scalable, GMP-grade, stable production system. The traditional adherent HEK293 triple-plasmid transfection system, whilst technically mature, is difficult to scale up; whilst HSV, baculovirus （ BEVS ） and early-generation suspension HEK293 platforms can improve scalability, they generally suffer from issues such as low yield per cell, a high proportion of empty particles and insufficient stability, making it difficult to meet the demands of large-scale clinical production. Through long-term cell domestication and process optimisation, this study has established an animal-free, high-yield suspension-culture HEK293 transient transfection production system. This system enables the stable, high-yield production of multiple AAV serotypes, reduces the proportion of empty capsids, is compatible with various bioreactors and supports linear process scale-up, whilst also enabling the continuous, long-term harvesting of supernatant from non-heparin-bound AAV. This significantly enhances overall production capacity and process stability, providing a reliable and optimised solution for the large-scale preparation of clinical-grade rAAV.

Establishment of a Suspension-Cultured HEK293 Cell Line

This study aimed to establish a scalable, high-titer, high-purity rAAV transient transfection production process. The study utilised high-transfection, high-yield adherent HEK293 cells from a GMP-compliant master cell bank, which were adapted through stepwise acclimatisation to a serum-free, animal-component-free and antibiotic-free suspension culture system. Following the screening and selection of a suitable serum-free suspension culture medium, the acclimatised suspension-cultured HEK293 cells were able to maintain high transfection efficiency and high rAAV production levels, thereby laying the foundation for subsequent large-scale production.

Optimisation of PEI:Max Transfection Conditions

In the past, this laboratory has used 25 kDa linear PEI for the transfection of adherent HEK293 cells. PEI exhibits low cytotoxicity towards HEK293 cells and high transfection efficiency; furthermore, there is no need to change the culture medium following transfection. As it offers significant cost advantages compared to lipid-based transfection reagents, this polymer was selected for suspension-based transfection in this study. For the transient transfection used in rAAV preparation, it is necessary to determine the optimal ratio of the three plasmids as well as the loading ratio of the transfection reagent to total DNA. On the day of transfection, 1 × 10⁶ viable cells/mL were seeded into a 125 mL shaking flask, with a working volume of 30 mL. Different ratios of the XX680:pXR2:TReGFP plasmids were set up, combined with PEIMax:DNA ratios of 2:1 and 4:1, with a total DNA concentration of 1 μg/mL; the mixture was incubated at room temperature for 10–15 min to complete complex formation. qPCR was used to determine the vector genome yield per cell in the cell lysates; the results in Figure 1a show that the highest vg/cell yield was achieved under a plasmid molar ratio of 2:1.5:1 and a PEIMax/DNA ratio of 2:1 yielded the highest vg/cell output. Among all plasmid ratio groups, the 4:1 PEIMax, whilst the increased amount of positively charged PEI Max/DNA complexes caused significant cell aggregation. All subsequent experiments were conducted under the optimised conditions of a plasmid ratio of 2:1.5:1 and a PEI Max/DNA ratio of 2:1. To investigate the effect of the volume of the complex system on PEIMax’s transfection efficiency and viral yield, this study established a volume gradient of the transfection complex, accounting for 2 per cent, 5 per cent and 10 per cent of the final culture volume, for comparison. Transfection efficiency was assessed via GFP flow cytometry, and viral yield per cell was determined by qPCR. The results in Figure 1b indicate that a transfection complex volume of 5 per cent yields the optimal transfection efficiency and the highest vg/cell output.

Figure 1 Optimisation of the ratio of the three plasmids and the volume of the transfection complex

（ a ） By adjusting the ratios of XX680, the pXR2 auxiliary plasmid and the TReGFP vector plasmid, the optimal particle-to-molecule ratio for rAAV production was identified. （ b ） Testing different transfection complex volume fractions under conditions of a PEI:Max/DNA ratio of 2:1 and an incubation time of 10–14 min; the complex volume fraction was calculated relative to the final cell culture volume. All the above parameters were optimised using SFM4Transfx for HeLa cells.

Optimisation of cell seeding density and plasmid DNA concentration

Suspended HEK293 cells were adjusted to 1 × 10⁶ and 2 × 10⁶ viable cells/mL, with a final culture volume of 30 mL; experiments were conducted using a gradient of total plasmid DNA concentrations of 1, 1.5 and 2 μg/mL. The results in Figure 2a show that rAAV vector yield was optimal under conditions of 1 × 10⁶ viable cells/mL and a DNA concentration of 1.5 μg/mL. The study further investigated the effect of the number of cell passages on the process by transfecting suspension-cultured HEK293 cells from low, medium and high passage numbers, respectively. Figure 2b indicates that the number of passages does not affect cell transfection efficiency; however, the rAAV yield per cell gradually decreases as the number of passages increases, suggesting that the maximum recommended number of passages for this suspension-cultured HEK293 clone is 30 to 40.

Figure 2: Optimisation of transfection cell density and the effect of cell passage number on vector yield

（ a ） Transfection experiments were conducted using seeding densities of 1 × 10⁶ and 2 × 10⁶ viable cells per millilitre, combined with total DNA doses of 1 μg/mL, 1.5 μg/mL and 2 μg/mL （ XX680:pXR:TR plasmid ratio 2:1.5:1 ）. （ b ） Transfection of low-, medium- and high-passage cells was carried out using the optimised process parameters to evaluate the effect of cell passage number on transfection efficiency and rAAV production capacity.

Ion-exchange chromatography for the purification of multi-serotype rAAV

Whilst optimising the rAAV production process using suspension-cultured HEK293 cells, this study undertook the development of an AAV purification process, with the aim of establishing a universal ion-exchange chromatography purification protocol suitable for all AAV serotypes and chimeric capsids. Existing literature confirms that different AAV serotypes can be purified using a variety of ion-exchange resins; this study proposes achieving universal purification for multiple serotypes using a single anion- or cation-exchange resin by adjusting the process parameters during the binding and elution stages. Previous reports have shown that a discontinuous iodoxanol gradient can be used to clarify rAAV lysates, and that iodoxanol does not affect the interaction between rAAV and the ion-exchange resin. Consequently, this study employed a modified discontinuous iodixanol gradient to clarify the lysate prior to ion-exchange chromatography; the buffer and pH of the iodixanol system were adjusted to allow direct coupling with the subsequent single-step ion-exchange chromatography. Figure 3a shows the preliminary experimental results of the combined modified iodixanol ion-exchange method; the components of the rAAV2 ¬CMVeGFP elution peaks were subjected to silver staining; transmission electron microscopy （ TEM ） negative staining revealed the presence of ring-shaped impurities measuring 5–10 μm in the latter two elution peaks, with subunit molecular weights of 20–25 kDa; this impurity was present in all AAV serotype preparation samples and was not specific to AAV2. Mass spectrometry identified this impurity as ferritin. By adjusting the ion-exchange chromatography conditions, effective separation of rAAV from ferritin was achieved. Figure 3b, combining the results of silver staining and negative transmission electron microscopy, demonstrates that the rAAV elution peak components have been separated from the ferritin impurity components. The modified discontinuous iodixanol gradient coupled with ion-exchange chromatography constitutes the downstream purification workflow for all rAAV serotypes and chimeric capsid vectors described in this paper; the entire purification process utilises the same set of buffers, resins, pH and chromatographic programmes, and is therefore universally applicable.

Figure 3 Characterisation of the components of the AAV elution peak following ion-exchange chromatography

（ a ） rAAV was analysed using non-optimised elution conditions, and the components of the AAV elution peak were characterised by silver staining and transmission electron microscopy （ TEM ） negative staining; the arrows indicate the major impurity, ferritin, in the silver-stained gel and the TEM image, respectively. （ b ） rAAV was eluted under optimised conditions, and the components of the AAV elution peak were analysed by silver staining and transmission electron microscopy （ TEM ） negative staining; the silver staining and TEM results demonstrate effective separation of the viral particles from the ferritin impurity.

Preparation and purification of single-stranded and self-complementary rAAV （ serotypes 1, 6, 8, 9 ）

Using optimised transfection parameters, transfection was carried out in a 1 L culture system （ approximately 1×10⁹ viable cells ） to prepare single-stranded and self-complementary rAAV serotypes 1ที่6, 8 and 9 carrying the CMV }\eGFP expression cassette. The yield of all single-stranded genomic serotypes per cell reached 1×10⁵ vg/cell or higher; Excluding AAV4, which exhibited low yield, the remaining single-stranded rAAVs were cultured in 1 L of medium and purified, yielding a total vector yield ranging from 8.2 × 10¹² to 3.3 × 10¹³. The silver staining results in Figure 4a indicate that the samples contained few impurities, with the main bands corresponding solely to the VP1, VP2 and VP3 capsid proteins. The overall yield of self-complementary AAV is generally lower than that of single-stranded AAV, with the product comprising both dimeric self-complementary genomes and monomeric single-stranded genomes. In this study, the total vector yield for each serotype of self-complementary rAAV, following 1 L culture and purification, ranged from 4.0×10¹² to 1.3×10¹³. The results of alkaline gel Southern blotting （ Figure 4b ） confirmed that the self-complementary genome constituted a high proportion of the genomes packaged into viral particles; when self-complementary AAV was prepared using other target gene expression cassettes, packaging of the self-complementary genome remained predominant, indicating that this phenomenon is not specific to the eGFP expression cassette. Currently, there is no universal cell line compatible with the vast majority of AAV serotypes. HeLaRC32 cells integrate the AAV2 rep and cap genes; following adenoviral infection, they can replicate a vector sequence flanked by AAV2 ITRs. An increase in eGFP copies enhances the fluorescent signal, making it suitable for fluorescent observation and quantification. Infection experiments were conducted using HeLaRC32 cells. As there are as yet no standardised cell lines or methods for assessing AAV infectivity, such transduction experiments can only serve as a reference indicator for batch release. Data on infectivity reported in different literature sources should be compared with caution, as different assay methods may lead to discrepancies in results.

Transmission electron microscopy analysis of rAAV serotypes using negative staining

To assess the overall purity of single-stranded and self-complementary rAAV, as well as the ratio of intact particles to empty particles, this study utilised transmission electron microscopy with negative staining to image samples of various serotypes. All existing AAV production processes result in the formation of empty capsid particles; there is therefore a practical need to develop purification processes capable of effectively separating intact particles from empty capsid particles. Empty capsid particles and intact particles carrying a genome exhibit different uptake behaviours with respect to the 2% uranium acetate negative staining dye; based on this, the proportion of intact particles to empty capsid particles in a sample can be quantified. Figure 4c shows negative-stained TEM images of rAAV1, rAAV2 and an AAV2 empty-cap control sample; the TEM results for rAAV1 and rAAV2 are representative of the sample states for the other serotypes tested; the AAV2 empty capsid prepared using a CsCl density gradient serves as the negative-staining reference for TEM.

Figure 4: Preparation of single-stranded and self-complementary rAAV serotypes 1 ต้องการ 6, 8 and 9 using optimised production and purification conditions

（ a ） Silver-stained electrophoresis images of purified single-stranded rAAV serotypes 1 ต้องการ 6, 8 and 9. （ b ） Alkaline agarose gel Southern blot results for self-complementary rAAV1 serotypes 6, 8 and 9. （ c ） Negative-stained transmission electron microscopy images, showing rAAV1 （ left ）, rAAV2 （ centre ） and AAV2 empty capsids （ right ）, in that order; electron micrographs were also obtained for serotype 3 serotypes 6, 8 and 9 （ images not shown ）. Arrows indicate empty capsid particles; intact particles can be distinguished from empty capsid particles by the electron-dense centres within the particles.

rAAV preparation using WAVE bioreactors

Production parameters for rAAV based on serum-free suspension medium in a shake-flask system have been optimised. The next step is to transfer the process to WAVE bioreactors to validate process scalability at different culture volumes. Multi-batch production runs were conducted in the WAVE bioreactor, with culture volumes ranging from 4.3 to 20 L, to prepare various AAV serotype vectors. Shaker flask transfection was set up concurrently as a control to compare the transfection efficiency of the GFP vector and the purified vector yield per litre of culture medium. The results showed that transfection efficiency in shake flasks was broadly comparable to that in WAVE bioreactor bags; in most WAVE batches, the purified rAAV yield per litre of culture medium reached 1×10¹³ or higher, demonstrating that the optimised parameters for cell culture, transfection and virusculture obtained from shake flasks could be successfully transferred to the WAVE bioreactor. For the majority of 10 L batches and all 20 L WAVE production batches, the total purified yield of single-stranded and self-complementary rAAV exceeded 1×10¹⁴.

Continuous Harvesting of rAAV from Suspension Culture Medium

It has been reported in the literature that adherent HEK293 cells secrete AAV into the culture medium via an unknown mechanism, and the virus can be collected from the supernatant during the late stages of transfection. Building upon a suspension-culture HEK293 production system, this study employed an intermittent medium exchange and collection protocol to carry out batch harvesting, with the aim of maximising vector yield per transfection. A 30 mL suspension culture of HEK293 cells was established in a shake flask at a seeding density of 1×10⁶ cells/mL, and transfection was performed to produce rAAV8, rAAV9 and CMVeGFP. Forty-eight hours after transfection, the supernatant was collected by low-speed centrifugation, and the titre of DNase-resistant particles was determined by dot blot hybridisation and qPCR; the cell pellet was resuspended in fresh medium and returned to the shake flask for continued virus production. Subsequently, samples were taken every 24 hours, and the medium was replaced every 48 hours. At 120 h, the rAAV8 and rAAV9 titres in the supernatant and cell pellet were measured respectively. The results in Figure 5 indicate that rAAV8 and rAAV9 could be detected in the supernatant as early as 48 h post-transfection; as time progressed and the medium was replaced, the vector titres in the supernatant continued to rise; Compared with the control group, in which cell pellets were lysed at 48 h, the intermittent harvest process increased the total yields of rAAV8 and rAAV9 by 6.5 times and 4.8 times, respectively. This process was further adapted to the perfusion mode of a WAVE bioreactor: every 24 h, a portion of the culture medium was removed for vector purification, and fresh medium was added to maintain cell viability and sustained virusculture. A 1 L-scale culture was conducted in a 2 L perfusion-type WAVE bag; 80 % of the culture medium was harvested via the perfusion filter every 24 hours after 48 hours of transfection, and fresh medium was replenished. The harvested rAAV8-containing CMVeGFP supernatant was concentrated via TFF prior to purification. For AAV serotypes with low affinity for production cells, the perfusion process, based on an animal-free suspension system, enables continuous harvesting and purification of the supernatant following a single transfection, eliminating the need for cell lysis and reducing the introduction of impurities such as host proteins and host genomic DNA. Compared with the traditional process involving termination of culture after 48 h and lysis for virus recovery, higher rAAV yields can be achieved with the same plasmid input. This study reports for the first time the continuous harvesting of rAAV supernatant at multiple time points post-transfection, based on a scalable, animal-free, suspended HEK293 bioreactor system.

Figure 5: Results of sustained production and harvest of rAAV8 and rAAV9 in suspension culture of HEK293 cells

The total yield is the sum of the supernatant harvest at each time point and the rAAV yield from the 120-hour cell pellet; The 48-hour cell pellet control group represents the vector yield obtained from the conventional 48-hour single-harvest process following transfection, as described in this study.

Discussion

Globally, there have been numerous Phase I and II clinical trials involving AAV for both hereditary and acquired diseases. With the number of clinical trials continuing to grow, the industry urgently requires the development of scalable production technologies that are compatible with multiple serotypes, compliant with GMP regulations, and easy to operate. This paper provides a comprehensive description of a large-scale production process for rAAV via transient transfection that exhibits excellent stability （ Figure 6 ）. This process utilises animal-free, antibiotic-free, and passaged HEK293 suspension cells; the rAAV yield per cell is comparable to that of the HSV dual-infection system and the BEVS baculovirus system, and outperforms other reported HEK293 suspension production platforms.

Figure 6: Schematic diagram of the complete rAAV production process

（ 1 ） Transfection of suspended HEK293 cells in a WAVE bioreactor; （ 2 ） Approximately 48 hours after transfection, the cells are transferred to a centrifuge flask and centrifuged at low speed to collect the cell pellet; （ 3 ） Resuspend the cell pellet and sonicate the cells; （ 4 ） Add Benzonase nuclease to the lysate and incubate at 37 °C for 45 minutes; （ 5 ） Clarify the lysate by low-speed centrifugation; （ 6 ） Load the clarified lysate onto an iodixanol gradient and perform ultracentrifugation; （ 7 ） Collect the rAAV fraction at the 40 %/60 % iodixanol interface, load it onto an FPLC system, and purify it by anion-exchange chromatography; （ 8 ） Dialyse and replace the rAAV chromatographic elution peak fraction with the final formulation buffer; （ 9 ） Perform sterile filtration through a 0.2 μm filter.