Accelerated Adeno Associated Virus Upstream Process Development From High‐Throughput Systems to Clinical Scale

ABSTRACT

Adeno‐associated virus (AAV) is one of the most common delivery systems used in gene therapy. Challenges in the development and manufacturing of AAVs include high cost of goods (COGs) per dose, process scalability, speed to market, and process‐related impurities such as empty capsids. This article presents a streamlined approach to developing and scaling AAV upstream production process via triple transfection from bench scale to commercial volumes exceeding 1,000 L. By leveraging high‐throughput technologies such as the AMBR®15 system, we achieved rapid upstream process development in under 2 months. These tools enabled optimization of productivity, impurity reduction, and COGs per dose. We also detail methodologies for direct scale‐up from AMBR®15 to a 2,000 L single‐use production bioreactor.

Accelerated, quality‐by‐design‐driven development and scale‐up of recombinant AAV production, from high‐throughput bench scale to commercial scale.

1. Introduction

AAV is a widely used gene delivery system, valued for its safety and efficacy in gene therapies. Its promise has driven numerous companies and research groups to develop AAV‐based drug products targeting chronic illnesses such as hemophilia and cancer (Jiang and Dalby 2023). As of 2022, over 200 clinical trials involving AAV gene therapies have been conducted (Aponte‐Ubillus et al. 2017). Despite ongoing innovations, significant challenges persist in scaling up AAV production. Transient transfection‐based manufacturing requires cell expansion and transfection with engineered plasmids, followed by extensive purification—steps that are both costly and time‐consuming (Technology Networks 2022). The increasing demand for AAV therapies continues to highlight industry‐wide hurdles, including the high cost of goods per dose (COGs), immune responses, complex and difficult‐to‐scale operations, and lengthy development timelines.

Although there are various challenges in AAV manufacturing, the high cost of goods per dose (COGs/dose) continues to be a key limiting factor. The cost to produce a single dose can reach over $1 million USD, driven by low product yields, expensive raw materials, and substantial overhead costs such as labor and facility usage (Salib 2023; Owens 2022; Thakur et al. 2024). A key contributor is the high proportion of empty capsids and other impurities, which necessitate extensive purification development—often resulting in low recovery and limited clearance of process‐related contaminants (Jiang and Dalby 2023; Dobrowsky et al. 2021). Additionally, long development and manufacturing timelines—ranging from 12 to 24 months (Holt 2021), add significant overhead cost and increase the complexity of bringing AAV‐based therapies to market. Raw material (such as plasmid, transfection reagent, and endonuclease) costs are a major contributor, with plasmid production being both expensive and complex, often involving long development timelines (Roughan 2024).

Scaling triple transfection‐based AAV production processes is particularly challenging because scale‐down models used for process development often lack scalability and maintaining critical transfection parameters (such as plasmid DNA – transfection reagent complexation time) during scale‐up poses significant operational challenges (Jiang and Dalby 2023; Shupe et al. 2022; Park et al. 2025). Differences in geometry and control settings between high‐throughput scale‐down models (e.g., AMBR®) and large‐scale production bioreactors lead to altered hydrodynamics and mass transfer, complicating scale‐up (Legrand et al. 2022; Sandner et al. 2019; Farsani et al. 2022).

Although quality by design (QbD) — a systematic, science‐ and risk‐based approach that leverages design of experiments (DoE) and historical data to ensure consistent product performance (ICH Expert Working Group 2009) — is increasingly applied in gene therapy, there is no comprehensive published research on using high‐throughput platforms (e.g., the Sartorius AMBR® system) to accelerate DoE execution and demonstrate successful AAV production scale‐up. Additionally, detailed methodologies and best practices for scaling up AAV upstream production process have yet to be widely shared. Thereby, in this article, we present a case study where AMBR®15 is leveraged for high‐throughput DoE development of AAV9 production process and scaling up to selected optimized conditions directly from AMBR® to 2000L Thermo Fisher HyPerforma^TM bioreactor. We also provide a comprehensive overview of the methodologies used to enable successful scale‐up, including strategies for maintaining transfection complex mixing, addition timing, agitation, and gassing (Figure 1).

Figure 1.

Rapid development of AAV9‐ The goal of this study is to perform rapid development and scale‐up right first time for rAAV9 production to address challenges with long manufacturing timelines and high cost of goods.

2. Materials and Methods

2.1. Cell Culture

Viral Production Cells 2.0 (Thermo Fisher Scientific, Waltham MA USA), a derivative of HEK293F cells, were thawed into a 125 mL shake flask (Corning, Corning NY USA) containing LV‐MAX™ Production Medium (Thermo Fisher Scientific, Waltham MA USA). The cells were sub‐cultured with a target seeding density of 3–6 × 10⁵ Vc/mL for 3–4 days. At the time of each passage, cells were counted using a Vi‐Cell XR (Beckman Coulter, Brea CA USA) while offline gas and metabolites were measured every day during the bioreactor stage on a Vi‐Cell MetaFLEX (Beckman Coulter, Brea CA USA). Shake flask cultures were incubated at 37°C with 5% CO2 and a rocking speed of 120 rpm on a 25 mm orbital throw shaker (Thermo Fisher Scientific, Waltham MA USA). In the AMBR®15, the cell culture for all conditions were controlled at 37°C temperature and a pH between 6.8 and 7.2 with CO2 sparge for acid control and 1 M sodium carbonate for base control. For dissolved oxygen (DO) control, the set point was set to 30% with a constant air sparge and O₂ as needed. For scale‐up, the culture was expanded into a 250 L HyPerforma™ Single Use Bioreactor (SUB) (Thermo Fisher Scientific, Waltham MA USA) and finally into a 2000 L HyPerforma™ SUB (Thermo Fisher Scientific, Waltham MA USA). The 250 L and 2000 L bioreactors were controlled at 37°C temperature, pH 6.8–7.2 with CO2 sparge for acid control and 1 M sodium carbonate for base control.

2.2. AAV9 Production

For the AMBR15 DoE, we assessed total DNA concentration and viable cell density at the time of transfection (Table 1). For total DNA, a range of 0.25–1.50 µg DNA per 10⁶ viable cell density (VCD) was chosen based on the requirements of total DNA for typical transfection reagent platforms (Jaqueline et al. 2022; Reese et al. 2023; Coplan et al. 2024). For VCD, a range of 1–5 × 10⁶ VC/mL was chosen to evaluate the effect of increasing cell density on vector genome titer. The cell density effect (CDE) is known to result in lower cell productivity at higher cell densities (Lavado‐García et al. 2022), therefore 5 × 10⁶ Vc/mL was the maximum cell density tested. Total transfection complex volume varied depending on target cell density with 5% complex volume per 2 × 10⁶ VC/mL. All other factors such as plasmid molar ratios (0.2:0.2:0.6 for helper: GOI: Rep/Cap), transfection reagent to total DNA ratio (1:1), and total complexation time (30 min, provided by the transfection reagent vendor) were kept the same across conditions and were selected based on prior experience with supporting in‐house data (Park et al. 2025). Plasmid DNA that was used consisted of a 6.9 kbp AAV9 Rep2Cap9 plasmid (Aldevron, South Fargo ND USA), 18.9 kbp pALD‐X80 helper plasmid (Aldevron, South Fargo ND USA), and a 7.17 kbp transgene plasmid. As shown in the table, each DOE design condition was repeated twice to assess reproducibility.

Table 1.

AAV9 transfection optimization DoE parameters.

Condition	ug DNA/e6 cells	Target VCD
1	0.25	1
1	0.25	1
2	0.65	1
2	0.65	1
3	1.08	1
3	1.08	1
4	1.5	1
4	1.5	1
5	0.25	2.3
5	0.25	2.3
6	0.65	2.3
6	0.65	2.3
6	0.65	2.3
7	1.08	2.3
7	1.08	2.3
7	1.08	2.3
8	1.5	2.3
8	1.5	2.3
9	0.25	3.7
9	0.25	3.7
10	0.65	3.7
10	0.65	3.7
10	0.65	3.7
11	1.08	3.7
11	1.08	3.7
11	1.08	3.7
12	1.5	3.7
12	1.5	3.7
13	0.25	5
13	0.25	5
14	0.65	5
14	0.65	5
15	1.08	5
15	1.08	5
16	1.5	5
16	1.5	5

Note: A molar ratio of 0.2:0.2:0.6, pHelper: pGOI: pRepCap and transfection reagent:DNA ratio of 1 was used for all conditions.

Plasmid DNA was diluted in LV‐MAX™ production carrier media in 15 mL conical tubes (Corning, Corning NY USA) and following the dilution, FectoVir®‐AAV (Sartorius, Göttingen Germany) was added to each transfection mixture to begin the complexation. The transfection complexes were mixed by inversion for 30 s and incubated for a total of 30 min before addition into the bioreactor. For production in the 2000 L bioreactor, the culture was seeded at a viable cell density of 5.5 × 10⁵ VC/mL and incubated for 4 days. Following the 4‐day incubation, the culture was diluted with LV‐MAX™ medium to target cell density of 2.4 × 10⁶ VC/mL. 0.5 µg of total DNA per 2 × 10⁶ VC/mL was targeted for the 1000 L transfection, with a FectoVir®‐AAV to DNA ratio of 1:1 and total complexation volume target of 5.75%. Plasmid DNA was pooled and diluted into the carrier media, 20 L of LV‐MAX™, and filtered through a Sartopore 2 size 8 (Sartorius, Göttingen Germany) into a 50 L HyPerforma™ Single Use Mixer (SUM) (Thermo Fisher Scientific, Waltham MA) containing LV‐MAX™ carrier media. The plasmid filtration train was chased with additional media until the target volume in the SUM was achieved then mixed at 102 rpm for 5 min. The FectoVir®‐AAV transfection reagent was aliquoted into a 1 L transfer bottle, then welded to the SUM and pumped in via a balloon pump. Once the entire transfection reagent was added, the complex was mixed at 102 rpm for 3 min. Following the 30‐min incubation, the transfection complex was pumped into the bioreactor at 6.8 L/min through the bioreactor harvest line using a Quattroflow 1200 single use pump (High‐Purity New England Lab, Smithfield RI).

2.3. Harvest

On day 3 post transfection, pH and DO control were disabled in the bioreactors and the culture was treated with a lysis buffer mixture consisting of final concentrations of 2 mM MgCl₂ (Sigma‐Alrich, Burlington MA), 0.2% Polysorbate‐80 (Sigma‐Alrich, Burlington MA), and 50 U/mL Benzonase endonuclease (Sigma‐Alrich, Burlington MA) for 115–125 min. Following lysis, 5 M NaCl (Sigma‐Alrich, Burlington MA) was added to target concentration of 150 mM and treated for 30 min. Samples from the harvest were taken and centrifuged with only the supernatant collected and submitted for analytics.

2.4. Analytical Assays and Testing

Vector genome titer was quantified by targeting the CK8 promoter sequence of the vector. Samples were treated with DNAse I (New England BioLab, Ipswich MA), and Proteinase K (Thermo Fisher Scientific, Waltham MA) followed by serial dilutions and then ddPCR was performed using the QX200 Auto DG Droplet Digital PCR system (Bio‐Rad, Hercules, CA, USA).

Capsid titer was determined by immunotitration of intact AAV9. The assay is based on a sandwich ELISA technique, which uses a monoclonal antibody specific to a conformational epitope on the AAV9 capsid (Progen, Wayne, PA, USA).

Residual host cell DNA was determined by using a commercially available HEK293 DNA kit (Thermo Fisher, Waltham, MA) where samples go through a DNA extraction using magnetic bead technology and then qPCR is performed to endpoint.

2.5. Statistical Analysis

For DoE design and analysis, JMP statistical software (SAS Institute, Cary NC USA) was used in conjunction with COGs/dose calculator to determine optimal transfection and bioreactor conditions. A response surface design was made for optimization of the transfection conditions, and all data were analyzed with RSM and third‐order polynomial approach.

3. Results and Discussion

3.1. HEK293F Triple Transfection Optimization in AMBR®15

Cell density and viability were monitored daily following transfection for the duplicates and triplicates of all 16 conditions (Figure 2A). Conditions with higher total DNA and transfection reagent levels exhibited lower viabilities compared to those with reduced amounts. This decrease in viability could be caused by increased cytotoxicity from elevated DNA and transfection reagent concentration (Carreno et al. 2024; Chopra et al. 2020; Liu et al. 2008), along with potential nutrient depletion and the accumulation of toxic byproducts, such as lactate and ammonia, at higher cell densities (Jang et al. 2022; Abaandou et al. 2021). An increase in viable cell density is observed until post transfection day 2, then viability begins to drop, and cell growth is stalled. This may be due to the depletion of important nutrients needed for cell growth, such as glucose, amino acids, and other metabolites (Zhu and Thompson 2019), as well as cytotoxicity due to rAAV production (Zhu and Thompson 2019).

Figure 2.

Optimization of triple transfection of HEK293 cells for AAV9 production‐(A) Average viable cell density and viability trends per DNA and cell density in AMBR15 during AAV9 production. The error bars represent vg titer assay variability of ± 20%. (B) AAV9 vector genome titer results per condition. Circle size indicates % full capsids estimate based on vg:capsid ratio. Color indicates cost of goods per dose with green being up to 60% reduction and red being 0%.

Viral genome (vg) titer, percent full capsid (estimated by viral genome to capsid titer ratio‐ vg:cp), and percent reduction in cost of goods per dose was assessed across various total DNA amounts and viable cell densities (VCDs) during the AMBR®15 DOE study (Figure 2B). Transfection at 1 × 10⁶ VC/mL resulted in the lowest productivity and percent full capsid ratio, most likely due to limited cell‐to‐cell contacts and lower overall number of cells available for transfection (Dash et al. 2021; Xue et al. 2024). An increase in vg titer was observed as cell density was increased for conditions transfected with ≥ 0.65 µg DNA/e6 cells with highest titer achieved being 2.3 × 10¹¹ vg/mL, although this resulted in lower percent full capsid. The overall low productivity and percent full capsid may be a result of the large transgene ( > 4.7 kbp) used for the experiment (Ibreljic et al. 2024). This data demonstrates that increasing cell density at the time of transfection leads to higher vg titers, specifically for conditions transfected with ≥ 0.65 µg DNA per 10⁶ cells. Further analysis is warranted to determine the maximum cell density in batch processes beyond which additional increases no longer result in higher vg/mL productivity. It is also important to investigate whether this trend persists at higher cell densities ( < 15 × 10⁶ vc/mL) in perfusion‐based processes.

Despite the increase in vg titer, the condition with the highest overall genome was not chosen for scale up. The selection of the optimal condition was based on achieving a minimum upstream full capsid percentage of 20%, together with the highest specific productivity (defined as viral genome titer, normalized to viable cell density at the time of transfection). According to these criteria, the optimized condition was identified as the run with 0.25 µg DNA per 10⁶ cells at a viable cell density of 2.3 × 10⁶ VC/mL, determined through AMBR®15 Design of Experiments (DoE) studies.

The upstream target of > 20% full capsids was established to meet the overall recommendation of exceeding 50% full capsids in the bulk drug substance. Although anion exchange chromatography is implemented in the downstream to enrich full capsids, this step typically achieves only a 2.5‐fold increase in full capsid percentage with minimal development (Park et al. 2025). Therefore, conditions yielding less than 20% full capsids at harvest were excluded from consideration.

The optimized condition achieved the highest cost of goods per dose reduction, with a 60% decrease compared to the industry benchmark (Lyle et al. 2023). The cost of goods (COGs) per dose for production at the 2000 L scale was calculated for each AMBR®15 run using the methodology described by Lyle et al. (2023). COGs are primarily influenced by the total amount of plasmid DNA and transfection reagent used, as well as the upstream production genome titer, which significantly impacts the final cost.

3.2. Agitation and Gas Input Bridging Study

For the DoE runs in AMBR®15, the system was operated at a power per volume (P/V) of 81 W/m³ and an air flow rate of 0.02 VVM. The AMBR®15 system is typically operated at agitation speeds 10–12 times higher, as a result of the physical differences that it has compared to a traditional bioreactor (Legrand et al. 2022; Sandner et al. 2019; Cytiva 2024). Therefore, using a target power/volume (P/V) ratio constant is not used when scaling from an AMBR®15 system. A combination of lower agitation speeds and adjustment of DO setpoints is typically done to match the DO profile in the AMBR®15 to a larger scale bioreactor (Sandner et al. 2019), making the system difficult to scale directly. To address this challenge, a study was conducted to evaluate bioreactor parameters, bridge the differences in power input and gas flow rates, and confirm the suitability of the selected target operating conditions between the AMBR®15 and the 2000 L bioreactor.

Agitation rates ranging from 8 to 160 W/m³ and air VVMs from 0.004 to 0.02 were tested in the AMBR®15 to assess the scalability of agitation based on power per volume combined with constant air sparging. While P/V values below 15 W/m³ are not typically used for AAV production (Decaria et al. 2023)—or above 30 W/m³ based on internal experience—a maximum of 160 W/m³ was tested in the AMBR®15 to determine the system's response to agitation.

As expected, due to their direct influence on mass transfer (kLa) (Chaudhry 2024; Bandaiphet and Prasertsan 2006), air flow rate times agitation rate term correlated with viral genome titer with p < 0.05 (Figure 3). The resulting correlation was used to define the operating ranges in the AMBR®15 which resulted in similar productivity to our top condition ‐‐ P/V of 8–81 W/m³ combined with constant air sparge of 0.004–0.013 and 0.018–0.02 VVM. Target P/V of 20 W/m³ with a minimal amount of constant air sparge flow rate of 0.004 VVM was used at 2000 L production stage. 20 W/m³ was chosen for scale up because it is within the historical operating range successfully used for AAV production (Park et al. 2025) and it is within the identified range in the AMBR15 where productivity remains consistent. These findings demonstrate agitation and gas flow rate can be scaled up by applying a constant P/V of 20 W/m³ at both scales with 0.004 VVM gasing. These findings are consistent with recommended operating ranges for large‐scale bioreactors, where the 2000 L system is typically operated at a P/V of 17–23 W/m³ with an air flow rate of 0.004–0.008 VVM with HEK293 cells (Park et al. 2025).

Figure 3.

Scale up of AAV9 process into 2000L bioreactor‐ significant interactions (p < 0.05) between air, agitation, and a combination of air and agitation rates are found. P/V range of 8–160 w/m³ with air VVM of 0.0038–0.02 were assessed in the AMBR15.

3.3. Scale‐up From AMBR®15 to 2000 L Under Accelerated Timeline

The timeline for process development and scale‐up was accelerated by maintaining a continuous seed train for up to 16 passages in which culture for development studies was split off as needed and the seed train was scaled‐up once final parameters were determined.

Stable production was observed for up to 16 passages and viability of the seed train remained > 90% throughout the process (not shown), reducing the need to thaw multiple vials for numerous experiments and waiting for recovery out of thaw. Viral Production Cells 2.0 (HEK293F cells) can be passaged up to 25 times post thaw per the manufacturer (Thermo Fisher Scientific 2021a) and based off internal data generated in‐house showing no impact on AAV quality (% full capsid, and viral genome titer) for up to 20 passages post thaw. Further data needs to be generated to continue production past 25 passages.

A direct scale‐up from AMBR®15 vessels to a 2000 L HyPerforma™ bioreactor was successfully executed, leveraging the continuous seed train described earlier. The optimized transfection condition—0.25 µg DNA per 10⁶ cells and VCD of 2.3 × 10⁶ VC/mL—was selected based on AMBR®15 DoE studies.

Given the geometric differences between the AMBR®15 and the 2000 L HyPerforma™ bioreactor, special consideration was given to scaling agitation rates and gas inputs as previously described. A more moderate P/V of 20 W/m³ was selected to better simulate large‐scale operating conditions (Sandner et al. 2019; Cytiva 2024).

Transfection scale‐up challenges were addressed by introducing a 50 L HyPerforma™ SUM to enable controlled mixing of the transfection complex (Figure 4), along with the use of a shear‐friendly pump for transferring the complex into the bioreactor. A low‐shear quattroflow single‐use diaphragm pump was selected to deliver the transfection complex through ¾″ tubing from the SUM into the bioreactor in under 10 min. Pump‐based addition was preferred over gravity feeding at the 2000 L scale, as gravity‐driven transfer is impractical due to safety concerns associated with lifting large transfection cocktail bags, as well as variability introduced during scale‐up or scale‐down.

Figure 4.

Process flow diagram of the transfection process‐ large scale transfection process flow diagram that shows the fluid pathway of the transfection complex components and the transfection complex into the bioreactor. A peristaltic pump was used to filter diluted DNA into a Thermo Fisher Hyperforma SUM. A quattroflow single use pump was used to pump the transfection complex into the bioreactor.

To mitigate the risk of contamination from pooling numerous small plasmid DNA aliquots, a plasmid filtration step was implemented, allowing DNA to be filtered directly into the SUM. The plasmid DNA and transfection reagent were then gently mixed for 2–3 min at a low agitation rate to ensure homogeneity (Thermo Fisher Scientific 2021b). The flow rate for bioreactor addition was optimized by matching the shear rate experienced during gravity feeding, ensuring consistency and minimizing potential impact on complex integrity.

The time‐series data for pH and dissolved oxygen (DO) in both the AMBR®15 and 2000 L bioreactors (Figure 5), demonstrate precise pH and DO control as the trends across both systems aligned within the target range of 6.7–6.9 for pH and 25%–35% DO. This indicates successful scale‐up of bioreactor parameters, such as agitation and gassing flow rates, using the methodology previously described.

Figure 5.

Comparison of AMBR®15 to 2000L BRX‐comparison of on‐line bioreactor pH and DO trends between AMBR15 and 2000L bioreactor. DO was controlled at 20%–40% while pH was controlled at 6.8 + /− 0.1 during production.

Viral genome (vg) titers of 7.9 × 10¹⁰ and 6.2 × 10¹⁰ vg/mL were obtained for AMBR®15 and the 2000 L bioreactor, respectively (Table 2). Considering assay variability of up to 20%, these titers are statistically comparable, demonstrating successful scalability of productivity from 15 mL to 2000 L—a scale‐up of over 100,000‐fold. Additionally, the percent full capsids measured in the 2000 L harvest was 22%, based off vg:cp ratio, which is consistent with results obtained from the AMBR®15. It is worth noting that a parallel study conducted in shake flasks using the same conditions resulted in a higher full capsid percentage of 46%, which is not representative of the 2000 L scale. This highlights that shake flasks may not serve as a universally representative scale‐down model for AAV production development and characterization. Therefore, data generated from shake flask studies should be interpreted with caution. Although shake flask and AMBR systems both serve as scale down models for AAV production, these systems have different mass transfer and process control characteristics. As an example, in shake flasks, gas liquid mass transfer only occurs at the liquid‐gas surface layer in the culture vessel, whereas in the AMBR system, sparging not only increases the oxygen exchange surface area but also improves the mass transfer coefficient. Additionally, the AMBR system controls dissolved oxygen and pH levels while shake flasks do not (Nienow et al. 2013). These differences are hypothesized to be the primary reason for the difference in AAV product quality. Finally, residual host cell DNA levels in the harvest material were 836 ng/mL for AMBR®15 and 848 ng/mL for the 2000 L bioreactor, further demonstrating consistent product quality across scales. The scale up principles applied in this study, constant P/V and air flow rate, are not limited to the Thermo Fisher HyPerforma bioreactor platform and can be applied for scaling into other bioreactor systems based on historical data (Park et al. 2025).

Table 2.

Quality attribute summary at harvest for 2000 L HyPerforma™ SUB, AMBR®15, and Shake Flask.

Quality attribute	Shake flask	AMBR®15	2000 L
Viral genome titer (vg/mL)	6.5 ± 1.3 ×1010	7.9 ± 1.6 ×1010	6.2 ± 1.2 ×1010
Capsid titer (cp/mL)	1.4 ± 0.3 ×1011	3.6 ± 0.7 ×1011	3.0 ± 0.6 ×1011
Vg‐to‐capsid ratio (%)	46 ± 4a	22	21
Residual host cell DNA (ng/mL)	N/A	836	848

^aAverage of triplicate flasks.

Multiple runs have been conducted in the AMBR®15 system to assess reproducibility (Figure 6). Across nine independent repeats, a total variability of 37% was observed in viral genome titer, capturing inherent process variability associated with cell culture, AAV production, assay performance, and system‐level factors such as AMBR®15 control and manual operations (e.g., pipetting, mixing). Notably, the titer obtained at harvest from the 2000 L HyPerforma™ SUB falls within this observed variability range, further supporting the scalability and consistency of the process. Note that, in addition to viral genome titer, capsid titer, and % full, other quality attributes such as infectious titer, capsid purity, and molecular identity should also be evaluated to ensure comparability across batches and scales (Park et al. 2025).

Figure 6.

Summary of AMBR15 and 2000L bioreactor data ‐ 9 runs in AMBR15 bioreactor demonstrate reproducibility of viral genome titer to titer obtained in 2000 L bioreactor. Error bars represent 1 standard deviation.

4. Conclusion

In this paper, we addressed critical challenges in AAV manufacturing via transient triple transfection —including high COGs, scalability, speed to market, and process impurities such as empty capsids. Leveraging high‐throughput screening in the AMBR®15 system allowed rapid screening for the optimized transfection condition, resulting in the highest productivity with lowest COGs in under 2 months. Results from the AMBR®15 studies demonstrated that balancing DNA/transfection reagent amounts and cell density is crucial for robust productivity and minimal cytotoxicity. A study was conducted to evaluate agitation and gas flow parameters in the AMBR®15 to enable scale up to large scale bioreactors, accounting for the system's distinct physical characteristics. Viral genome titer correlated statistically with the product of agitation and air flow rate and the resulting correlation was used to identify optimal AMBR®15 agitation and airflow rate conditions for robust scale up. Scale up from AMBR®15 to a 2000 L production vessel was achieved by introducing a single use mixer and a low‐shear pump to control complexation mixing of plasmid DNA with the transfection reagent and shear. The optimized condition was successfully scaled up to the 2000 L HyPerforma™ bioreactor with comparable viral genome titers, percent full capsids, and residual host cell DNA levels. This combined strategy of high‐throughput process development and thoughtful engineering controls for robust scale up not only accelerated timelines but also supported consistent product quality and reduced COGs at commercial manufacturing scales.

Author Contributions

Angela Andaluz: conceptualization, experiments, writing – original draft preparation. Brittany Monteverde: conceptualization, experiments, writing – original draft preparation. Kevin Vera: experiments. Brandon Tse: experiments. Ivan Gajic: conceptualization. Clifford Froelich: conceptualization. Seyed Pouria Motevalian: conceptualization, writing – review and editing, supervision.

Conflicts of Interest

The authors declare no conflicts of interest. Consumables from Thermo Fisher Scientific and other suppliers were used as required by process and testing needs.

Data Availability Statement

The authors have nothing to report.