Advanced biomanufacturing of AAV

The process flow diagram (PFD) of an advanced AAV biomanufacturing was developed in this study, including suspensive production, bioproduction scale up, clarification, liquid chromatography purification and scale up, post-purification process, storage and evaluations (Fig. 1). The AAV production process development was performed in shaker flasks at scale of 30–100 mL and in spinner flasks with 60–100 mL of cultures. The production process in 1.2–2.0 L of stirred-tank bioreactors with process parameter control could be applied to pilot plant production and possible large-scale manufacturing production. As detailed later, the key production parameters identified in this study include host cell selection, transfection condition, and agitation speed. Two-step universal separation process using anionic exchange chromatography and ultrafiltration has been developed to purify multiple AAV serotypes. The post-purification desalting and concentration procedures have also been investigated. This study reported an advanced generic AAV biomanufacturing process of production, clarification and purification. Importantly, the developed platform is robust, scalable, and applicable to cover multiple (if not all) serotypes.

Fig. 1

Process flow diagram (PFD) of the advanced AAV biomanufacturing platform, including production, clarification and purification

Development of suspensive AAV production and clarification

We first compared two suspensive host cells, i.e. HEK 293F and VPC, in shaker flask production at 37 °C, 8% CO2 and 125 rpm. Both seed and production cultures showed that VPC cells had significantly lower cell clumping than HEK 293F. The VCD of VPC reached > 4.0 × 106 cells/mL on Day 1 post transfection, followed with VCD and viability dropping to ~ 3.4 × 106 cells/mL and 82% and AAV-DJ8 titer increased significantly from Day 2 (Fig. 2). AAV was harvested at 72 h post transfection with VCD of 2.8 × 106 cells/mL and cell viability of 70–80% in shaker flask. The dynamic production profile revealed a significant increase of AAV titer from Day 2 to Day 3. Similar cell growth and AAV productivity were observed in the productions of AAV2, AAV5 and AAV-DJ (cell culture profiles not shown). As summarized in Table 1, the volumetric productivity of AAV-DJ8 using the same triple plasmids, pAAV-NLuc-GOI (~ 3.9 kb), pAAV Rep-Cap and pHelper, was 0.50–0.53 ± 0.08 × 1010 vg/mL by HEK 293F and 2.40 ± 0.06 × 1010 vg/mL by VPC cells under respective optimal transfection conditions. It is obvious that VPC produced about 5-fold higher AAV in shaker flask than HEK 293F. Therefore, the process development and scale up in this study used VPC.

Fig. 2

Development of 3-day suspensive AAV production in small-scale shaker flask. A Evaluation of transfection VCD of 2.0, 3.0, and 4.0 × 106 cells/mL. B Evaluation of ratio of plasmid DNA and VPC cells including 0.4, 0.5 and 0.6 μg/1 × 106 cells. C Viable cell density and viability of VPC pre- and post-transfection with maximal VCD of 4.53 × 106 cells/mL and harvest viability of 71% with optimal transfection conditions. D Volumetric productivity of AAV with final titer of 3.13 × 10.10 vg/mL with optimal transfection conditions. VPC cells were cultivated in 30-mL viral production medium supplemented with 6 g/L of glucose and 4 mM of GlutaMax at 37 °C, 8% CO2, and 130 rpm. The production process could be applied to four serotypes (AAV2, 5, DJ, and DJ/8)

Table 1 Summary of AAV biomanufacturing development

Our previous study showed that the cell density at the time of transfection and amount of plasmid DNA are other two key transfection parameters to improve AAV production [25]. Therefore, we evaluated the effects of cell density at the time of transfection (2.0, 3.0, and 4.0 × 106 cells/mL) and ratio of total plasmid DNA/VPC cells (0.4, 0.5, and 0.6 μg/million cells) in shaker flask productions. As presented in Fig. 2A and B, the optimal transfection VCD is 3.0 × 106 cells/mL and plasmid DNA: VPC ratio is 0.5 µg: 106 cells, which generated final AAV titer of 5.6–10.0 × 1010 vg/mL. Therefore, our scaling up evaluation and purification development studies used the optimal transfection VCD and plasmid DNA amount identified here.

Then the optimal suspensive production process was validated with four AAV serotypes, using pAAV2, 5, DJ and DJ8 Rep-Cap, expression vector with ~ 3.9 kb of inserted genes and pHelper, in shaker flask and/or bioreactor cultures at the developed conditions. The optimal transfection formulation, i.e. pAAV expression: pAAV Rep-Cap: pHelper ratio of 1:1:3, DNA: cell ratio of 0.5 µg:1 million cells, 10% viral-plex buffer and 0.6% AAV-MAX transfection reagent, and supplement of 0.3% booster and 1% enhancer, was applied. The qPCR titration of intracellular AAV showed similar range of productivity of 7.88 ± 0.39, 2.97 ± 0.13, 2.40 ± 0.06, and 5.60 ± 5.14 × 1010 vg/mL for AAV2, AAV5, AAV-DJ8, and AAV-DJ, respectively (Table 1). These results demonstrated that the suspensive AAV production process can be used to generate multiple serotypes.

Furthermore, we investigated and compared several raw AAV clarification strategies, including direct lysis of cell culture broth and lysis of cell pellets after centrifugation. The direct lysis by adding AAV-MAX lysis buffer and other supplements (MgCl2 and benzonase) and incubating the lysis mixture at 37 °C was time-consuming (2–6 h), and also had poor cell lysis efficiency in some batches which could be caused by culture variations. Then we tested the strategy of centrifugation to collect cell pellets followed with two lysis options as detailed in Section “AAV clarification”. Our results demonstrated that both strategies, i.e. incubation at 37 °C and repeated freeze–thaw cycles, achieved 95–100% VPC lysis. The lysis of culture broth enables direct collection of raw AAV from most productions tested in this study, but cell pellet lysis could achieve high AAV release efficiency (as confirmed with cell lysis rate), reduce lysis reagent amount and simplify clarification operation in bioreactor-based production.

Bioproduction scale-up

Before scaling up shaker flask production process to stirred-tank bioreactor, AAV production was evaluated in 250-mL spinner flask with working volume of 60–100 mL. The agitation speeds of 75, 100, 125 and 150 rpm were tested. The low agitation speed caused significant cell aggregation and shortened culture longevity. The AAV productions in spinner flask presented in Table 1 were performed at 37 °C, 210 rpm and 8% CO2. As compared to shaker flask, spinner flask production reached maximal VCD of 4.3–4.6 × 106 cells/mL on Day 2 and VPC cells containing AAV were harvested at viability of 70–80% (Fig. 3). Similar to shaker flask cultures, AAV titer was significantly increased from Day 2 to Day 3 in spinner flask. It was observed that spinner flask production was less than 20% of that in shaker flask, i.e., 0.41 vs 2.40 × 1010 vg/mL. These results suggested that the suspensive transfection and AAV production in stirred tank is feasible, but the process parameters need further optimization for high productivity.

Fig. 3

AAV production in spinner flask. A Kinetic profile of VPC cell growth with peak VCD of 4.51 × 106 cells/mL and harvest viability of 76%. B AAV production with final titer of 0.59 × 1010 vg/mL. Spinner flaks cultures were carried out at 37 °C, 8% CO2, and 230 rpm using AAV-DJ8 as model virus

Next, we investigated the process scale up to stirred-tank bioreactor using seed cultures from shaker flask or spinner flask. Both strategies showed similar cell growth and AAV production, so all bioreactor productions presented in this study used shaker flask seed cultures. As shown in Fig. 4, the maximal VCD reached 6.15 × 106 cells/mL (AAV-DJ8) or 7.60 × 106 cells/mL (AAV-DJ) and harvest viability was about 90–95% (AAV-DJ8) or 80–85% (AAV-DJ) at 72 h post triple-plasmid transfection, which had different cell growth kinetic profile from those in shaker flask and spinner flask. The production titers of 8.14 ± 1.91 × 1010 vg/mL for AAV-DJ and 7.52 ± 0.49 × 1010 vg/mL for AAV-DJ8 were obtained on Day 3 in 1.2–2-L bioreactor production at 37 °C, pH 7.0, 210 rpm and DO 40%. It is clear that VPC cell growth was enhanced by ~ 50% and AAV titer was improved by > 100% in stirred-tank bioreactor as compared to shaker flask (Table 1). These process-scaling up data demonstrated that our AAV production process was robust and scalable in bioreactors, which is important to future industrial productions to support clinical trials or potential clinical applications.

Fig. 4

Scaled up AAV production in stirred-tank bioreactor. A VPC cell growth profile with peak VCD of 6.15 × 106 cells/mL (AAV-DJ8) or 7.60 × 106 cells/mL (AAV-DJ). VPC had better and healthier cell growth in bioreactor. B AAV concentration reached 7.17 × 1010 vg/mL (AAV-DJ8) or 8.14 × 1010 vg/mL (AAV-DJ). The 1.2–2.0 L of production cultures were performed in 2.5-L bioreactor with automatically controlled process parameters of 37 °C, pH 7.0, 210 rpm and DO 40%

Purification development and scale-up

Multiple commercial columns for AAV purification have been evaluated in this study (Table 1), including Cytiva HiTrap Q Sepharose XL strong anion exchange column, Cytiva Sepharose Fast Flow anion exchange column, Cytiva HiTrap AVB Sepharose column, Bio-Rad Foresight Nuvia HPQ column, and Bio-Rad EconoFit Nuria aPrime 4A. The primary purification method using NGC liquid chromatography equipped with these columns were developed.

As shown in Fig. 5, aPrime 4A column achieved purification recovery of > 85% using equilibration buffer A (25 mM Tris–HCl, 20 mM NaCl, pH 9.0) and elution buffers of A and B (25 mM Tris–HCl, 1 M NaCl, pH 9.0). Linear elution (0 ➔ 100% increase of buffer B) in aPrime 4A column did not well separate AAV from other peaks (data not shown). The stepwise elution (0, 15, 25, 70, 85 and 100% of buffer B) at flow rate of 1.0 mL/min well separated AAV peak from other impurities, with high binding rate of 85–95% and elution rate of ~ 100%, using 1-mL aPrime 4A column and pellet lysate from 20-mL culture (Fig. 5A). The binding rate was calculated by titrating raw AAV samples and flow through collection. We further increased the loading amount of raw AAV by using pellet lysate from 100-mL culture in 1-mL aPrime 4A column, which showed that the AAV binding rate was reduced to < 80% although the binding amount was significantly increased (Fig. 5B). The representative chromatography profile of AAV-DJ8 was described in Fig. 5, but four serotypes of AAV2, 5, DJ and DJ8 were tested using the same column, loading and elution conditions, which did not show obvious difference in binding and elution. These results confirmed the robustness and scalability of our primary AAV purification using IEX. Small amount of AAV was detected in flow through and other elution peaks from aPrime 4A column. Further optimization of sample loading and elution conditions (e.g. flow rate and stepwise strategy of buffer B) might be able to increase the overall purification recovery rate.

Fig. 5

Development and optimization of anion exchange purification using liquid chromatography (LC). The 80–140 mL of cell lysis from 20–100 mL of VPC pellet was loaded to the 1-mL or 5-mL columns. The representative LC profile of AAV-DJ8 was described here but four serotypes of AAV2, 5, DJ and DJ8 were tested using the developed purification strategy. Equilibration buffer: 25 mM Tris–HCl, pH 9.0. Elution Buffer A: 25 mM Tris–HCl, 20 mM NaCl, pH 9.0. Elution buffer B: 25 mM Tris–HCl, 1 M NaCl, pH 9.0. Flow rate: 1.0 mL/min. A Stepwise elution of AAV-DJ8, 80-mL AAV lysis from 20-mL VPC pellet, in 1-mL EconoFit Nuvia aPrime 4A column. B Stepwise elution of AAV-DJ8, 140-mL AAV lysis from 100-mL VPC pellet, in 1-mL EconoFit Nuvia aPrime 4A column. C Stepwise elution of AAV-DJ8, 100-mL AAV lysis from 50-mL VPC pellet, using 1-mL Foresight Nuvia HPQ column, which can be scaled up from 1-mL column to 5-mL and 25-mL columns. D Stepwise elution of AAV-DJ8, 100-mL AAV lysis from 50-mL VPC pellet, using 5-mL Foresight Nuvia HPQ column

The stepwise elution (0, 50, 65 and 100% of buffer B) of raw AAV lysis from 20-mL pellet using 1-mL HPQ column showed lower binding and overall recovery rate of 40–60% (Fig. 5C) than aPrime 4A. Furthermore, we scaled up the purification process to a 5-mL pre-packed commercial HPQ column, loaded with AAV lysis from 50-mL pellet (Fig. 5D), and to an in house packed 25-mL column using the same Nuvia HPQ media. The similar binding rate, elution profile and recovery rate were observed in both 5-mL and 25-mL HPQ columns while significant (~ 50%) amount of AAV was detected in flow through and other elution peaks using HPQ column.

The evaluations of other commercial columns showed that the Q Sepharose IEX column had low AAV binding rate (< 5%) and AVB Sepharose affinity column showed weak binding rate (< 5%) of AAV2 and DJ8 using the manufacturer provided purification parameters as detailed in Section “AAV purification”. Taken together, the IEX purification using aPrime 4A column with stepwise elution was identified as the optimal primary purification in this study although further development and optimization is needed in future.

The secondary purification using ultrafiltration and other strategy such as G25 column or dialysis was tested to concentrate and desalt (i.e. buffer exchange) the purified AAV. The AAV2, 5 and DJ8, which were filtered, concentrated and washed with PBS using 100 kDa MWCO PES column following manufacture procedure, showed high recovery rate (> 90%). However, the AAV-DJ elute from IEX column blocked PES column, and 100 kDa MWCO regenerated cellulose column was identified as a suitable column to ultrafiltrate AAV-DJ with high recovery rate of 90%. The alternative strategies are to combine desalting operation using HiTrap G25 column equipped in liquid chromatography system following the manufacture protocol or 20 kDa dialysis cassette with additional ultrafiltration concentration or refrigerated vacuum concentrator. The purified AAVs were aliquoted in formulation buffer of 1 × PBS, 5% Sorbitol and 350 mmol/L NaCl, and stored at -80 °C for long term.

Quality evaluations of produced AAV

Although the developed biomanufacturing process was validated using four serotypes of AAV, the AAV-DJ8 was applied in the following characterizations or evaluations. To characterize the AAV-DJ8 produced from our developed bioprocess, SDS-PAGE was performed with silver staining and detected three capsid proteins, 87-kDa VP1, 73-kDa VP2 and 62-kDa VP3 (Fig. 6A). Western blotting was carried out to analyze the purified AAV, which confirmed the integrity and expression of all three capsid proteins (Fig. 6B). Moreover, TEM image confirmed the right size and morphology of AAV (Fig. 6C). In addition to high productivity and recovery, transduction capability of functional AAV was also evaluated using live-cell imaging. As described in Fig. 7, glioblastoma U251 cells (green color, GFP labelled) were transduced with Cy5.5-labelled AAV-DJ8 (red color), and confocal microscope imaging demonstrated that AAV accumulated around the DAPI-stained nucleaus (blue color) within 24 h post incubation. These images revealed that our AAV could effectively transduce cells in vitro.

Fig. 6

Characterizations of produced AAV. A SDS-PAGE of AAV pre-purification and post anion exchange purification. M: marker, and C: negative control protein. B Western blot confirmed three AAV capsid proteins: VP1 (87 kDa), VP2 (73 kDa) and VP3 (62 kDa). C TEM image of purified AAV. Scale bar: 200 nm

Fig. 7

Confocal microscope demonstrating high transduction of AAV, revealed by co-localization of green GFP (U251 cells), blue DAPI (nucleus), and red Sulfo-cyanine 5.5 (AAV). MOI = 5,000

The in vivo AAV induction and functional expression of AAV-delivered gene were tested by intracranially injecting 1 × 1010 vg of AAV-DJ8 into the glioblastoma U251 xenograft NSG mouse models. As described in Fig. 8A, the NLuc gene was delivered to glioblastoma tumor and functionally expressed to generate bioluminescence in vivo with induction of ViviRen (37 µg, intravenous injection), as detected by live-animal IVIS imaging. This result also confirmed the gene expression in tumor only facilitated with the tumor-specific promoter in AAV expression vector [25]. It was observed that the in vivo NLuc bioluminecence lasted 1–2 h post injection of substrate ViviRen.

Fig. 8

Evaluations of functional gene expression. A Live-animal IVIS imaging showed high in vivo expression of AAV-delivered gene. About 0.5 × 106 U251 cells were intracranially injected to NSG mice using stereotactic instrument to develop glioblastoma xenografted models. AAV (1 × 1011 vg) and ViviRen (3.7 μg) were injected. B In vitro AAV gene expression is dosage (multiplicity of infection, MOI)-dependent. C AAV gene expression correlates to MOI, as measured by i3x plate reader

Furthermore, we transduced 5 × 104 cells of U251 that were seeded in 96-well plates with AAV-DJ8 at MOIs of 0, 1,000, 2,500, 5,000, and 7,500. Neither MOI of 0 (25 µM of ViviRen only) nor 1,000 generated bioluminescence signal while MOIs of 5,000 and 7,500 had strong bioluminescence (Fig. 8B). Higher MOI of AAV generated stronger bioluminescence than lower MOI in 6-well plate cultures. The dynamic SpectraMax iD3 profiles showed that the bioluminescence signal decreased to minimal levels within 25 min post induction in vitro (Fig. 8C). All these characterization and evaluation data demonstrated that our new biomanfuacturing process generated high-quality AAVs.

Advantages of our AAV biomanufacturing process

This study developed a novel AAV biomanufacturing procedure with multiple advantages as compared to previously reported production processes. First, high productivity can be achieved in the stirred-tank bioreactor-based production. Second, the developed process is robust and scalable to large-scale biomanufacturing for future pre-clinical and clinical trials. Third, good-purity AAV was generated using the identified ion-exchange columns and developed purification protocols. Fourth, the good-quality AAV produced from the developed process can be used in vitro and in vivo without detected side effects such as fever or immune toxicity. Most importantly, the developed universal biomanufacturing process can be applied to produce and purify different serotypes of AAV (AAV2, 5, DJ and DJ8 in this study).

Prospective AAV biomanufacturing

This study developed a scalable suspensive AAV production process by evaluating host cell and transfection parameters. The Viral Production Cell (VPC) 2.0 engineered from parental HEK 293F cells by Gibco, which has larger cell size, faster cell growth, and minimal cell clumping at optimal shaking or agitation condition, was applied and enhanced AAV titer by 5 folds as compared to HEK 293F [25]. Compared to adherent HEK 293AAV or 293A, the VPC-based AAV production process is robust and easy to scale up in bioreactor. Moreover, this host cell showed high resistance to shear force and could directly inoculate the production medium in bioreactor using the seed cultures from shaker flask without any adaptation.

One of the key parameters in AAV production using VPC was the agitation speed. For instance, low agitation could increase cell clumping, reduce cell growth, and decrease AAV production significantly. The high agitation speed of 210 rpm enabled high AAV production and minimal cell aggregation.

Another important parameter is the high consumption of glucose and GlutMAX due to the fast cell growth and high AAV productivity. In the biomanufacturing process developed in this study, the same basal medium was used from Day -1 when seeding the production bioreactor until the end of AAV harvest without medium exchange or culture dilution. The batch culture of AAV production showed significant cell viability dropping on Day 2 (data not shown). The booster and enhancer added during transfection could extend the culture longevity and maintain high viability. However, it was found that more than 8.2 g/L of glucose was consumed from Day -1 to Day 3. Lack of assay to titrate GlutaMAX in culture broth, we assumed the 1:1 consumption rate of glucose and GlutaMAX, and fed 3.5 mM of GlutaMAX together with ~ 3.5 g/L of glucose between Day 1 and Day 2 to avoid nutrient depletion in this study. To further optimize AAV production, a full extracellular and intracellular metabolite analysis is needed to monitor glucose and GlutaMAX consumption and correlate cellular metabolism to cell growth and AAV production.

In the presented AAV bioproduction, we stopped culture at 72 h post triple-plasmid transfection, but AAV-DJ did not reach maximal value at the harvest viability. Therefore, we suspected that AAV titer could be further improved by optimizing the endpoint of production process via evaluating different harvest viabilities.

Further optimization of AAV purification

The challenge in the purification of engineered AAV capsids is the lack of a high-specificity of binding resin with high capture rate. The generic IEX column separation developed in this study can be applied to multiple AAV serotypes, but the purity could be lower than the affinity column purification. To further improve the purity of AAV, affinity-based primary capture and purification followed with secondary or polishing strategies could be developed in future to benefit the recovery and purity of multiple AAV serotypes.

In addition, the primary purification using IEX aPrime 4A liquid chromatography column captured 85–95% AAV in one round of sample loading. To achieve higher capture rate, the loading capacity, flow rate of loading buffer, and packing strategy of purification resin should be further optimized. Another strategy is to run serial purification using both aPrime 4A and HPQ columns to improve the binding rate of AAV.

Ultrafiltration could further purify the AAV post IEX purification by removing the impurities with molecular weight of < 100 kDa, and also combine desalting, buffer exchange and sample concentration into one step. However, we observed that AAV-DJ had high retention rate in PES membrane but showed high recovery in regenerated cellulose column. Evaluation and selection of suitable ultrafiltration material might be needed for different serotypes. An alternative strategy is to use G25 desalting column or dialysis in combination with vacuum concentration to process the purified AAV, but the multi-step operation could reduce the recovery rate of AAV.