Intensification of rAAV Production Based on HEK293 Cell Transient Transfection

ABSTRACT

Recombinant adeno‐associated virus (rAAV) vectors are widely used in gene therapies, but the rapidly increasing global demand has created a significant challenge for rAAV manufacturing, where production capacity remains a critical bottleneck. To address this, strategies to enhance production yields are urgently needed. This study presents an innovative approach to rAAV production using high cell density (HCD) stirred tank perfusion culture. rAAV1 and rAAV9 vectors carrying GFP cargo were used as models, with triple‐plasmid transfection performed in suspension HEK293 cells at a high viable cell density of 50 million cells/mL in culture then maintained at ≥ 30 million cells/mL throughout production. Transfection and production parameters were first optimized in a 5 mL pseudo‐perfusion spin tube screening system at HCD. A proof‐of‐concept was then demonstrated by scaling up to a 200 mL stirred tank bioreactor (STR) in perfusion mode. This intensified process achieved rAAV9 production levels per cell comparable to those observed in reference shake flask cultures at 1 million cells/mL. By implementing transfection at very HCD in a perfusion‐based STR, this approach has the potential to significantly enhance rAAV volumetric production capacity, providing a promising solution to meet the growing demand for gene therapies.

Graphical Abstract and Lay Summary

1. Introduction

Recombinant adeno‐associated virus (rAAV) vectors are widely used for in vivo gene therapy [1, 2]. Facing an increasing demand for rAAV manufacturing, the cell culture yield remains a bottleneck [2, 3].

Transient transfection of HEK293 cells is commonly used due to the high rAAV infectivity, flexibility in producing different serotypes, and shorter time to market for drug candidates [4, 5]. This system involves cotransfection of three plasmids pRC, pHelper, and pAAV, and is considered safe since live helper virus is not needed [6, 7, 8, 9]. HEK293 cells express the E1 adenovirus gene required for rAAV replication. Stable packaging cell lines for rAAV production present challenges, including cytotoxicity, low empty/full capsid ratio, time for the generation of a cell line and low productivity [10, 11, 12, 13, 14].

Another major manufacturing platform is based on Sf9 insect cells using a baculovirus expression vector system (BEVS) [5, 15]. This method is favored for large‐scale production due to its robustness and scalability, achieving titers exceeding 1 × 10⁵ vector genomes per cell (vg/cell) or 1 × 10¹⁴ vg/L [15, 16]. However, the BEVS system has drawbacks, including low VP1 expression, reduced rAAV potency, infectivity, instability of recombinant baculoviruses [10, 17, 18–23]. Additionally, the susceptibility of insect cells to Rhabdovirus contamination necessitates extra purification steps, increasing complexity and potential risks to product quality [24, 25, 26, 27, 28].

For rAAV production via triple‐plasmid transfection in HEK293 cells, different transfection methods exist, including physical techniques like electroporation and chemical methods [29, 30]. Chemical transfection is widely used, employing reagents such as cationic polymers or lipids that form polyplexes with the DNA and facilitate its entry into cells via endocytosis/phagocytosis [31, 32, 33, 34, 35, 36]. Polyethylenimine (PEI), a stable cationic polymer, offering highly efficient transfection for suspension cells in serum‐free medium is cost‐effective compared to cationic lipids, making it a preferred choice for large‐scale rAAV manufacturing [37, 38, 39]. The most common PEI reagents include linear PEI, PEI MAX, and PEI PRO, with PEI MAX and PEI PRO exhibiting superior performance, however, also associated with high cost at scale [40, 41].

Current rAAV production via chemical transfection is typically conducted at cell densities below 5 million viable cells per milliliter (MVC/mL), more commonly between 1 and 2 MVC/mL. [41, 42, 43]. The resulting vector genome yields range from 10¹² to 10¹⁵ vg/L, depending on process conditions [10, 42, 44–49, 50]. Despite advancements in cell and vector engineering, achieving consistent product yield and quality at scale remains a challenge [44]. Large‐scale transfection processes require careful optimization of plasmid‐PEI complexation and control of raw material variability, including plasmid DNA, transfection reagent reagents, and media [50]. Some contract development and manufacturing organizations (CDMOs), such as Viralgen VC, have scaled rAAV production up to 2000 L, demonstrating feasibility but highlighting ongoing challenges [51].

Process intensification strategies focus on increasing productivity and/or reducing costs [42, 49, 52, 53]. Approaches include raising cell density, enhancing cell‐specific productivity, and shortening process duration. However, many studies report a decline in cell‐specific productivity beyond a certain culture density, known as the “cell density effect” [3, 15, 54–60]. This issue likely arises from byproduct accumulation and/or nutrient depletion in high cell density cultures (HCDC). One potential solution is perfusion culture, which continuously renews medium and removes waste [49, 61–63], allowing viable cell density (VCD) up to 10 MVC/mL for rAAV production [49].

This study aimed to intensify rAAV production by increasing the viable cell density in a transient triple‐plasmid transfection process using HEK293 cells. The target process involved transfection at a very high HEK293 cell density (≥50 MVC/mL) without subsequent dilution to lower VCD. The process intensification work was conducted using commercially available host cells, plasmids, transfection reagents, and media, with minimal optimization of factors such as plasmid ratio. The process was first established and optimized in a 5 mL screening system, operated in pseudo‐perfusion mode with frequent manual medium exchanges. The intensified process was then scaled up to a 200 mL perfusion culture in stirred tank bioreactor (STR).

2. Materials and Methods

2.1. Cell Line and Media

HEK293F cells, denoted HEK293, and Viral Production Cell 2.0, VPC2, were obtained from Thermofisher Scientific. The culture media were: BalanCD HEK293(‐), a formulation without glucose, supplemented with glucose according to the Targeted feeding method (TAFE) [64, 65] and L‐glutamine to support the cell needs, BCD, (FUJIFILM); FreeStyle 293 Expression medium, FS293, (ThermoFisher Scientific); or Viral Production Medium, VPM, (ThermoFisher Scientific). The AAV‐MAX system kit (ThermoFisher Scientific) or some of its components were used. Cells cultured in Erlenmeyer flasks, SF or 50 mL spin tubes, ST, (Corning or TubeSpin bioreactor TPP, used interchangeably) were grown in an incubator at 37°C, 5% CO₂ with agitation at 120 rpm and 320 rpm, respectively.

2.2. Transfection Materials

Transient transfections were performed with PEI MAX (Polysciences) unless otherwise mentioned. rAAV was produced using plasmids pHelper, pGFP, and pRC1 (Cell Biolabs, cat#VPK‐401) for rAAV1 or pR2C9 (pAAV2/9n, Addgene plasmid #112865) for rAAV9, using GFP as GOI. The efficacy of the rAAV was evaluated in a transduction assay measuring GFP expression.

2.3. Experimental Set‐Up

ST operated in pseudo‐perfusion mode, where the medium was manually exchanged using centrifugation while all the cells were retained in the culture, was used as scale‐down system with a 5 mL working volume unless other specified. These were inoculated with cells in exponential growth phase, centrifuged, and resuspended in medium at the desired VCD, and transfected. The developed processes were applied to perfusion cultures in a DASBOX bioreactor (Eppendorf), at a 200 mL working volume, with an ATF‐2 (Repligen) at flow rate 0.3 L/min with a hollow fiber cartridge, HF (Spectrum, Repligen) of 88 cm² surface area and 0.2 µm pore size, as routinely used in industry for perfusion processes of CHO and HEK293 cells [62]. This bioreactor system in perfusion mode was established in‐house [62, 65–68]. Despite a size smaller than the filter cutoff, the rAAVs were retained by the HF, potentially due to interactions with the filter material or adherence of residual DNA and PEI MAX, or other components. Since the AAV was mainly present in the cell lysate and not secreted, its accumulation in the bioreactor instead of passage to the filter permeate was not disadvantageous in this proof‐of‐concept study, while future work could explore other filters. The bioreactor was harvested at 72 h posttransfection (hpT) by centrifugation (200 × g, 5 min), unless specified otherwise. The supernatant and cell pellet were stored at −80°C for analyses. To determine the production titers, cell pellet samples were resuspended in lysis buffer (50 mM Tris, 150 mM NaCl, 2 mM MgCl₂) followed by three rounds (10 min each) of freeze/thaw by alternating the tubes between an isopropanol bath (−80°C) and a water bath (37°C), centrifugation (3000 × g, 4°C, 10 min) and collection of the supernatant for analyses.

A standard protocol (shake flask cultures), as described by the reagent providers or in the literature [41, 42, 43], was used as reference, with the following steps: (1) a 20 mL cell suspension culture at 1 MVC/mL in a 125 mL SF was prepared with exponential growth phase cells. When the standard protocol was used as reference for transfection in bioreactors, cells were removed from the bioreactor to prepare satellite cultures, 20 million cells were removed from the bioreactor and diluted to 1 MVC/mL in 20 mL just before the transfection; (2) a plasmid DNA mix was prepared as 1 µg plasmid/1 MVC with a 1:1:2 ratio of pRC: pGFP: pHelper, diluted in transfection medium to 1 mL total volume; (3) PEI MAX was prepared as 2 µg PEI MAX per 1 MVC and diluted in transfection medium to 1 mL total volume; and (4) plasmid and PEI MAX solution was then mixed and incubated for 10–15 min at room temperature before addition to the cell suspension.

2.4. Analytical Methods

The VCD was measured by a Bioprofile FLEX (Nova Biomedical) or Norma XS (Iprasense). The concentrations of glucose, lactate, glutamine, and ammonium were measured by Cedex Bio (Roche). The transfection efficiency (GFP intensity) was measured by flow cytometry using a Gallios (Beckman Coulter) or a Guava EasyCyte (Luminex Corporation). The viral vector capsids were quantified using ELISA kits for AAV1 or AAV9 and technical duplicates were performed for each sample (Progen Biotechnik). The viral genome was analyzed as technical triplicates using an AAVpro titration kit for qPCR (TakaraBio) using a CFX96 real‐time qPCR detection system (Bio‐Rad). The percentage of full capsids was the ratio of the number of viral genomes to the number of capsids. Transduction assays were performed using 150 000 HEK293F cells/well in a 24‐well deep‐well plate (1.5 mL at 0.15 MVC/mL), with addition of 1 mL of AAV containing sample (cell lysate or centrifuged supernatant). After 72 h (37°C, 250 rpm, 5% CO₂), the GFP expression was measured by flow cytometry to determine the viral transfection efficiency.

3. Results

The process intensification efforts in this study aimed to establish a triple‐plasmid transient transfection process in HEK293 cell cultures at very high cell concentration while maintaining comparable rAAV production per cell to the standard protocol at 1 MVC/mL VCD. The goal was to sustain high cell density (HCD) throughout both the transfection and production phases without a dilution step posttransfection. To achieve this, medium renewal was carried out via perfusion during cell expansion and rAAV production.

In standard triple‐plasmid transient transfection protocols for rAAV production, plasmid DNA is typically prepared at 1 g/L in Tris buffer (TE) or water, with 0.5 to 1.5 µg of plasmid per million cells. PEI MAX, a commonly used transfection reagent, is reconstituted as a 1 g/L stock solution in water [69]. In the standard protocol, plasmids and PEI MAX are diluted in equal volumes of transfection medium before transfection, and this reagent mix usually occupies 5%–10% of the total culture volume [42, 50, 70]. Scaling this standard protocol from 1 MVC/mL to HCD–where cell densities are increased 20‐ to 50‐fold poses a significant challenge due to the corresponding increase in transfection reagent and plasmid volumes, leading to very large volume additions. To address this, more concentrated solutions of transfection reagent and plasmids were explored, though higher concentrations risked precipitation. Experiments were conducted to determine conditions that could trigger precipitation based on concentration, volume ratio, buffer composition, or culture media. The findings revealed that PEI MAX stock concentration could be increased to 5 g/L in transfection medium without requiring pH titration postreconstitution. Additionally, a volume ratio of 3:1:1 for cell suspension, plasmid solution, and PEI MAX solution (ratio_{vol_cells:DNA:PEI}) did not generate obvious precipitation at cell densities up to 100 MVC/mL across various plasmid DNA stock solutions.

3.1. Development of a Small‐Scale Pseudo‐perfusion System Process–rAAV1

The development of a HCD rAAV production process using triple‐plasmid transient transfection required optimizing two key sequential phases: transfection at HCD, followed by production at HCD. As previously noted, VCD was expected to decrease in the production phase due to the added volumes of transfection reagent and plasmid. Additionally, each phase had distinct nutrient requirements.

To address these challenges, initial experiments were designed to optimize the transfection and production phases separately where possible. Preliminary results indicated that transfection at 100 million viable cells per milliliter (MVC/mL) was feasible; however, maintaining such a high density required intensive medium renewal. Spin tube pseudo‐perfusion (STPP) experiments were conducted to determine the optimal conditions for HCD cultivation, including volume ratio of cell suspension to plasmid DNA and PEI MAX solution (ratio_{vol_cells:DNA:PEI}), mass ratio of cell number to plasmid DNA and PEI MAX (ratio_{mass_cells:DNA:PEI}), reagent addition parameters, and VCD.

3.1.1. Selection of Medium Renewal Rate for Spin Tube Pseudo‐Perfusion Cultures

Building on our previous experience with glycoprotein production at 100 MVC/mL, an experiment in STPP was conducted to preliminarily assess rAAV1 production from transfection at 100 MVC/mL and medium renewal. Since maintaining a VCD of 100 MVC/mL in an STR is highly challenging, the transfection was carried out at this density, while the rAAV production phase was performed at reduced densities of 30 or 10 MVC/mL. The 30 MVC/mL density was chosen for its high but robust performance [68], whereas the lower density of 10 MVC/mL was selected due to its reduced nutrient and oxygenation demands.

Following the standard protocol, a plasmid cocktail (µg) to cell number (MVC) ratio of 1:1 and a plasmid (µg) to PEI MAX (µg) ratio of 1:2 were used. The transfection was performed with a ratio_{vol_cells:DNA:PEI} of 3:1:1 by adding a 2 mL mix of plasmid (1 mL) and PEI MAX (1 mL) to 3 mL of cell suspension. Immediately afterward, the VCD was adjusted to either 10 or 30 MVC/mL through dilution with culture medium. The cultures were harvested at 72 hpT in all screening experiments, a time point chosen based on standard industry harvest practices.

Four conditions for the production phase were tested, each with distinct medium exchange regimes (A, B, C), informed by our experience with STPP and CHO cell perfusion operations, where a daily medium renewal of one reactor volume (RV/day) was effective for 30 MVC/mL density [64]. Regime A involved medium exchange starting at 24 hpT for 10 and for 30 MVC/mL production phases. Regime B hypothesized that nutrient levels would suffice for 2 days at 10 MVC/mL, delaying medium exchange until 48 hpT. Regime C initiated medium exchange as soon as possible 1 day posttransfection in the 30 MVC/mL culture to ensure early nutrient replenishment, with the first exchange occurring at 18 hpT, followed by daily exchanges. As shown in Figure 1, rAAV production was higher in conditions 1 (1434 capsids/cell) and 2 (1077 capsids/cell) at 10 MVC/mL posttransfection compared to conditions 3 and 4 at 30 MVC/mL, in terms of rAAV concentration and production per cell. Higher medium exchange rates, such as daily (condition 1) vs. every second day (condition 2) or early initiation at 18 hpT (condition 3) versus later (condition 4), resulted in higher rAAV production. These findings highlight the critical role of medium renewal in viral vector production. Early initiation of medium exchange (at 18 hpT) was essential, and daily exchanges proved insufficient for maintaining 30 MVC/mL posttransfection, leading to the adoption of medium exchange of one reactor volume every 12 h, starting at 18 hpT, for further STPP process development.

FIGURE 1.

Effect of different regimes of medium exchange, A, B or C, on rAAV1 production, for transfection performed at 100 MVC/mL using ratio_{vol_cells:DNA:PEI} 3:1:1, followed by dilution to either 10 MVC/mL (conditions 1 and 2) or 30 MVC/mL (conditions 3 and 4) post transfection; production of rAAV1 from cell lysates harvested at 72 hpT in 5 mL STPP cultures (n = 1) using BCD medium showing the VCD at harvest 72 hpT (bars) and the capsids/mL at harvest (cross), and where the medium exchanges are detailed in the lower panel–error bars are calculated based on analytical duplicates of capsids/cell.

Since the highest titer of 1434 capsids/cell from transfection at 100 MVC/mL (Figure 1) was relatively low compared to the 5000–6000 capsids/cell achieved with the standard protocol at 1 MVC/mL, an experiment was conducted to determine the optimal VCD for further work. A cell density of 50 MVC/mL was identified as the highest viable concentration with satisfactory and robust performance across a range of 20–100 MVC/mL (see Figure S1). This selection was based on transfecting pGFP at multiple cell densities using a ratio_{vol_cells:DNA:PEI} of 3:1:1, followed by dilution of the culture to 1 MVC/mL for the production phase to minimize nutrient limitations and byproduct inhibition while measuring GFP expression at 72 hpT. For densities exceeding 20 MVC/mL, the transfection efficiency tended to decline as cell density increased, with reproducibility becoming less reliable at 80–100 MVC/mL, likely due to increased variability in transfection reagent complexation at high concentrations. Since densities of 40 and 60 MVC/mL demonstrated consistent transfection efficiencies above 60%, 50 MVC/mL was selected for further process optimization.

3.1.2. Selection of Ratio of Cell Number, Plasmid DNA, and PEI MAX

Several critical ratios are typically optimized in a transient transfection process, with the plasmid molar ratio and ratio_{mass_cells:DNA:PEI} being among the most important. The plasmid molar ratio is highly specific to the plasmids, cell type, and serotype, requiring GOI‐ and product‐specific optimization. In this study, a 1:1:2 ratio of pGFP: pRC:pHelper yielded the highest rAAV titer and was therefore adopted for all subsequent experiments, regardless of serotype, leaving further optimization outside the present scope.

The impact of ratio_{mass_cells:DNA:PEI} was assessed in HCD STPP at 50 MVC/mL, using a ratio_{vol_cells:DNA:PEI} of 3:1:1, with medium renewal every 12 h starting at 18 hpT. In this experiment (Figure 2), the ratio between plasmid cocktail (µg) and cell number (MVC) was varied at 1:1, 1:2, or 1:4, while the plasmid (µg) to PEI MAX (µg) ratio was tested at 1:1, 1:2, or 1:4. Capsid production from transfection at 1 MVC/mL in BCD medium, following the standard protocol in shake flasks (SF) with ratio_{mass_cells:DNA:PEI} 1:1:2, was used as reference.

FIGURE 2.

Effects of the ratio_{mass_cells:DNA:PEI}, ratio between the cell number (MVC), the plasmid DNA (µg), and PEI MAX (µg), on the rAAV1 capsid production per cell obtained by transient transfection at 72 hpT in BCD medium (n = 2) at 50 MVC/mL in spin tube pseudo‐perfusion system (filled bars) with medium renewal started at 18 hpT and performed every 12 h in comparison with transfection at 1 MVC/mL (empty bars) carried out in BCD medium with a ratio_{mass_cells:DNA:PEI} of 1:1:2 using the standard protocol in shake flasks, all using the same starting cells and performed in parallel—production titers quantified from cell lysates harvested at 72 hpT–error bars are calculated based on analytical duplicates.

Results showed that the 1:1:2 ratio_{mass_cells:DNA:PEI} produced the highest capsids/cell (5000–6000 capsids/cell), confirming its selection for all further experiments unless otherwise specified. It is important to note that this study aimed to develop a process intensification method applicable across different experimental conditions by increasing VCD while maintaining per‐cell production. Thus, maximizing cell‐specific productivity was outside the scope of this study, though further investigations to enhance per‐cell production are ongoing.

3.1.3. Optimization of PEI MAX/Plasmid DNA Complex Formation Time

Optimization of the incubation time of the mixed transfection reagents is a common practice in transient transfection process development. To evaluate the impact of incubation time, experiments were conducted in STPP using BCD medium, comparing incubation durations of 0‐, 5‐, 10‐, and 15‐min. The 0‐min condition involved quickly inverting the transfection reagent mixture a few times before adding it to the cell suspension, without incubation. Results showed that in comparison with shorter incubation times, 10‐ and 15‐min incubations yielded the highest viral genome production per cell, with no significant difference between them (Figure 3).

FIGURE 3.

Impact of the complex formation time and method on rAAV1 viral titers, showing the VCD and the viral genome rAAV titers measured in the cell lysates harvested at 72 hpT in STPP cultures transfected at 50 MVC/mL with BCD medium (n = 2); four incubation conditions of 0, 5, 10, and 15 min, of the mixed plasmids and PEI MAX were tested, where 0‐min condition refers to a quick mix of plasmids and PEI MAX by inverting the mixture a few times before addition to the cells without incubation, as well as an alternative condition “direct transfection”, where the plasmids were added to the cell culture, immediately followed by addition of PEI MAX, i.e., the pre‐mixing of plasmid with PEI MAX was completely eliminated–error bars indicate the standard deviation of analytical triplicates for the viral genome titer.

During the experiments, gradual opalescence (transition from a semitransparent to opaque liquid) was observed when mixing plasmids with PEI MAX, even with minimal agitation, including in the 0‐min condition. Additionally, an alternative “direct transfection” approach where plasmids were added directly to the cell suspension, followed immediately by PEI MAX was tested. While this method would simplify large‐scale operations by eliminating the need for premixing, it resulted in a 13% lower viral genome titer compared to the 10‐ and 15‐min incubations, and significantly reduced rAAV production in an STR (data not shown).

Based on these findings, A 10‐ to 15‐min incubation was selected for further small‐scale and perfusion bioreactor experiments, aligning with literature recommendations. From the optimizations above, the following procedure was adopted for further development: plasmids and PEI MAX were mixed at a ratio_{mass_cells:DNA:PEI} of 1:1:2 and incubated for 10‐15 min. Transfection was performed at 50 MVC/mL, using a ratio_{vol_cells:DNA:PEI} of 3:1:1. After the transfection, the rAAV production phase began at 30 MVC/mL, reduced from 50 MVC/mL due to the DNA and PEI MAX mixture addition. Cultures were harvested at 72 hpT, unless otherwise specified.

3.2. Small‐Scale Evaluation of the HCD Process T50/30

The T50/30 process was aimed at providing a generic intensification strategy for rAAV production using commercially available host cells, plasmids, transfection reagents, and media. To evaluate its applicability across different serotypes, exploratory experiments were conducted in STPP, replacing the rAAV1 rep‐cap plasmid RC1 with RC5 (rAAV5) or RC9 (rAAV9), other components and operations being identical. Increases of 8‐ and 10‐fold in production per cell for rAAV5 and rAAV9 compared to rAAV1 (Figure S2), demonstrating the applicability of the T50/30 process across serotypes.

Given these promising results, the T50/30 process was implemented for rAAV9 production in both STPP and bioreactor systems without further serotype‐specific optimization, aiming to demonstrate a general applicability of the method, using only commercially available components. The production of rAAV5 was not pursued further to manage the high experimental workload.

To compare the performance of T50/30, the AAV‐MAX production system–a commercially available high‐titer platform–was used as a benchmark. The AAV‐MAX kit includes: Viral Production Cell 2.0 (VPC2); recommended for transfection at 3 MVC/mL; viral production medium (VPM); AAV‐MAX transfection reagent; AAV‐MAX booster; AAV‐MAX enhancer; Viral‐Plex Complexation Buffer; AAV‐MAX lysis buffer. It was valuable to assess the impact of the VPC2 cells in the T50/30 process outside of the AAV‐MAX protocol. For this, HCD transfection T50/30 process in STPP (ST‐a and ST‐b) using VPC2 was compared to 3 MVC/mL transfection in shake flasks (SF‐a and SF‐b).

The transfection and production were carried out with either (a) the AAV‐MAX system (ST‐a and SF‐a) according to the manufacturer protocol including recommended volumes (except for omission of the Viral‐Plex Complexation buffer to reduce the variation between the conditions and transfection reagents) or (b) PEI MAX transfection using FS293 medium (ST‐b and SF‐b). Notice that comparable AAV productions were observed between BCD and FS293 media in STPP (Figure S3), leading to the adoption of FS293 for logistical reasons.

Using VPC2 cells with AAV‐MAX transfection at 3 MVC/mL (SF‐a), the production per cell was higher than with PEI MAX transfection and FS293 medium (SF‐b), 2.2 × 10⁴ vs. 0.73 × 10⁴ vg/cell. However, when applied in the HCD STPP T50/30 process (ST‐a), the AAV‐MAX system resulted in significantly lower viral genome and capsid production per cell than the PEI MAX and FS293 medium‐based T50/30 process (ST‐b). This discrepancy may stem from suboptimal reagent proportions when scaled to VCDs 10 times higher than the manufacturer's recommended protocol–further adjustment being outside the scope of this study. Differences in transfection conditions also affected posttransfection growth profiles (ST‐a and ST‐b, Figure 4E).

FIGURE 4.

Comparison of rAAV9 production at 72 hpT (n = 2), summed from measurement in the cell lysate and in the supernatant, from transfection protocols differing by the cell density, the transfection reagent, the medium or the cell line, showing (A)–(B) the viral genome production per cell, capsid production per cell, percentage of filled capsids based on the viral genome production and the capsid production; (C)–(D) transfection efficiency (GFP expression) and transduction (GFP expression); (E)–(F) VCD, viability; with the following conditions: (A), (C), (E) comparison of rAAV production by VPC2 cells using either the AAV‐MAX system according to the manufacturer protocol (except the omission of the Viral‐Plex Complexation buffer) performed at 3 MVC/mL (shake flask SF‐a (n = 2)) or at 50 MVC/mL according to T50/30 process (spin tube ST‐a (n = 2)) vs. using PEI MAX transfection in FS293 medium performed at 3 MVC/mL (shake flask SF‐b (n = 2)) or at 50 MVC/mL according to T50/30 process (spin tube ST‐b (n = 2)); (B), (D), (F) comparison of rAAV9 production by VPC2 cells (spin tube ST‐b (n = 2)) vs. HEK293 cells (spin tube ST‐c (n = 2)) using PEI MAX transfection in FS293 medium performed at 50 MVC/mL according to T50/30 process–error bars are calculated based on analytical duplicates for the capsids titers and analytical triplicates for the viral genome titers.

Using PEI MAX transfection, the capsid production per cell in HCD STPP (ST‐b) was comparable to low VCD (3 MVC/mL) transfection in SF (SF‐b), 5.3 × 10⁴ vs. 7.7 × 10⁴ capsids/cell, despite a 4.84‐fold higher harvest VCD in STPP (Figure 4A,E). The transfection efficiency was 84% in STPP vs. 72% in SF, with transduction rates of 80% and 65%, respectively. The viral genome production per cell was slightly lower in HCD STPP, possibly due to higher nutrient demands or accumulation of inhibitory byproducts, factors explored in Figure 1 but not yet fully optimized. These results confirmed that the HCDC T50/30 process, originally developed for rAAV1, was successfully applicable to rAAV9, achieving production levels comparable to low‐VCD transfections.

Significantly higher rAAV production per cell using PEI MAX transfection was achieved with VPC2 cells (ST‐b) compared to HEK293 cells (ST‐c) in the T50/30 process (Figure 4B,D,F). However, since this study focused on a widely accessible process using common components without restrictive licensing, HEK293 cells were selected for perfusion bioreactor applications in the following section.

3.3. HCDC Transfection Processes in Perfusion Bioreactor Runs

The T50/30 process developed in STPP was evaluated in perfusion STR for the production of rAAV1 and rAAV9 using HEK293 cells expanded to a density of 50 MVC/mL. Once the VCD reached at least 50 MVC/mL, a portion of the culture was discarded, and fresh medium was added to achieve a total of 120 mL of cell suspension at 50 MVC/mL in the bioreactor. To apply the ratio_{vol_cells:DNA:PEI} 3:1:1, the 120 mL cell suspension (i.e., 60% of the final volume) was transfected with 40 mL of plasmid mix and 40 mL of PEI MAX solution, each constituting 20% of the final volume. Additionally, a plasmid concentration of 0.5 µg plasmid/MVC was assessed in small‐scale experiments and subsequently adopted for the bioreactor process to minimize DNA usage, aligning with findings published by others [50, 70].

A series of experiments conducted with this protocol in a perfusion STR with ATF restarted immediately after transfection yielded significantly lower rAAV1 titers compared to the STPP system (data not shown). A critical distinction between this process and the STPP system was the repeated transit of the cell suspension in the lumens of the filter hollow fibers and deep tubes during cell separation by ATF, which may have adversely affected transfection outcomes from mechanical damage. To mitigate this mechanical impact in the present proof‐of‐concept study, we introduced, a resting step of 30 min in a large‐culture area container, specifically a nontissue culture treated surface area of 300 cm², in an incubator at 37°C with 5% CO₂, allowing the high gas exchange necessary for the very high VCD. Preliminary studies indicated that this resting step did not negatively affect outcomes when varying the culture volumes from 0.5 to 5 mL (Figure S4). The resting step was added immediately after the transfection and was followed by transfer of the transfected cell suspension back into the bioreactor, while the perfusion was resumed. A graphical representation of this process is given in Figure S5. The intention was to minimize the mechanical effects immediately posttransfection to evaluate the potential of the T50/30 process.

The T50/30 process was employed for rAAV1 production in BCD medium, chosen for its compatibility with both cell growth and transfection. Posttransfection, the VCD and cell viability remained relatively stable (Figure 5A). On average, a production of 3 × 10⁹ vg/mL was achieved from 1 to 3 dpT, indicating that rAAV1 production is feasible at this VCD. This yield was lower than that observed in the STPP system, which produced approximately 2 × 10¹⁰ vg/mL (Figure 3), and the transfection efficiency was also lower. It is important to note that a significant difference between the perfusion culture in the bioreactor and the STPP system was the total medium renewal prior to transfection in the latter. In perfusion mode, while medium renewal occurred, it was not complete, and the culture had been initiated 6 days before transfection, potentially the proportion of conditioned medium in the bioreactor could have impacted the transfection efficiency.

FIGURE 5.

Feasibility of rAAV production by HCDC transfection process T50/30R in perfusion in a 200 mL STR with HEK293 cells using PEI MAX transfection in BCD medium; (A) rAAV1 production showing the VCD, cell viability, transfection efficiency (GFP), viral genome/mL in the cell lysate samples; (B) rAAV9 production showing the VCD, cell viability, transfection efficiency (GFP), viral genome/mL in the cell lysate samples, and including control culture control_1 (n = 1) transfected at 1 MVC/mL and control culture control_15 (n = 1) transfected at 15 MVC/mL using the same cells and reagents as the HCDC run (see text); (C) transduction assay (n = 2), measured at 72 h post transduction, performed with rAAV9 supernatant samples collected from the T50/30R bioreactor run of panel (B) at 3, 6, and 7 dpT (filled bars) and from control_1 culture with supernatant collected at 3 dpT (empty bar).

Encouraged by the promising rAAV1 titers obtained in the HCD bioreactor, it was decided to adopt the resting step in the present proof‐of‐concept study, designating the obtained process as ‘T50/30R,’ and to plan exploration of alternative approaches for this step in future studies.

The HCDC T50/30R process was also evaluated for the production of rAAV9 (Figure 5B). rAAV9 production was monitored until no further increase in the GFP level was observed at 10 dpT to assess production beyond 3 dpT. Results demonstrated that rAAV production was amenable to process intensification through increased VCD in the perfusion STR. Following transfection, the VCD slowly increased to 120 MVC/mL by 7 dpT and was sustained for an additional 3 days, with high viability (>93%). The transfection efficiency ranged from 51% to 62%, indicating sustained long‐term production. In parallel with this bioreactor run, two control cultures at lower densities were conducted using the same plasmids and PEI MAX stock solutions. Cells taken from the bioreactor before transfection were transferred into spin tubes and diluted to either 1 MVC/mL (control_1) or 15 MVC/mL (control_15) using BCD medium, and transfected. For control_1 the standard protocol was followed, while the control_15 culture used the same ratio_{vol_cells:DNA:PEI} 3:1:1 as used in the perfusion STR, with daily medium exchanges performed posttransfection until harvest. At 3 dpT, the perfusion process yielded a 32.4‐fold higher VCD (64.8 vs. 2 MVC/mL) and a 30‐fold higher viral genome production per mL (8.4 × 10⁹ vs. 2.8 × 10⁸ vg/mL) than control_1. At 3 dpT, compared to control_15, the perfusion process resulted in a 4.4‐fold higher VCD (64.8 vs. 14.8 MVC/mL) and 8.4‐fold higher viral genome production per mL (8.4 × 10⁹ vs. 1 × 10⁹ vg/mL). Thus, the perfusion process achieved comparable production per cell (130 vg/cell) as control_1 (140 vg/cell) using the standard protocol, while nearly doubling the production yield per cell compared to control_15 (68 vg/cell). The rAAV9 was partially secreted into the culture, allowing sample supernatants to be used to assess the transduction efficiency from sample supernatants collected between 3 and 7 dpT from STR as well as from control_1 (Figure 5C). The transduction efficiency from STR samples increased from 3 to 6 dpT, stabilized at 7 dpT, and was higher than that of control_1 culture. This indicated that at 7 dpT, the HCDC process had generated rAAV9 that was more infectious for human cells, demonstrating significantly higher transduction efficiency compared to control_1, which was transfected at 1 MVC/mL using the same cells and transfection reagents.

4. Discussion

Gene therapy holds immense potential for patient treatment; however, viral vector manufacturing remains a significant bottleneck. Due to their intrinsic complexity, increasing the viable cell density to enhance viral vector production is challenging. This study aimed to explore strategies to overcome these limitations.

4.1. HCD Effect

The so‐called “cell density effect” has been described as a challenge for process intensification by increasing the VCD in viral vector production [71]. This phenomenon may result from altered cellular metabolism, physiological state, microenvironmental changes, or the accumulation of inhibitory components [72]. While some studies have mitigated this effect through medium exchange or nutrient supplementation [49], others have observed reduced production per cell at high VCD despite complete medium renewal, suggesting additional underlying factors beyond nutrient limitation [57]. High production per cell has been reported in perfusion process at 10 MVC/mL using a 5 µm pore size TFDF hollow fiber filter [49]. However, this filter was too large for the 200 mL bioreactor system used in this work.

In the present study, the STPP system demonstrated high reproducibility across experiments, independent of variations in cell lines, working volumes, transfection media, or cell densities, see Figure S4 for transfections performed at 50 and 80 MVC/mL at different scales followed by dilution to 1 MVC/mL. The T50/30 process in STPP was successfully applied to HEK293F for rAAV1, rAAV5, and rAAV9 production, as well as to VPC2 cells for rAAV9 production, without specific optimization. In STPP, cells were centrifuged and resuspended at the target density in fresh medium before transfection, creating an environment free of byproducts or impurities while achieving near 100% viability. This approach was similar to the method described by Backliwal [73], where high density cell suspension at 20 MVC/mL was obtained via centrifugation and resuspension in fresh medium. Conversely, in perfusion bioreactor experiments, cell expansion required at least 6 days to reach ≥ 50 MVC/mL, depending on inoculation density. Standard hollow fiber filters (0.2 µm pore size) were used for the perfusion bioprocesses, and these filters are known to retain impurities, potentially exacerbated by gradual filter fouling. However, fouling typically takes about a week to occur. Given the consistent success of spin tube transfections, the variability observed in T50/30 bioreactor experiments may be attributed to culture impurities interfering with transfection rather than high VCD itself. Key differences between the perfusion bioreactor and STPP system included the extent and method of medium renewal before transfection, both of which affected the cellular microenvironment and impurity accumulation. Another factor was the delivery method for the transfection reagents: pipetting in spin tubes versus liquid transfer through tubing in the bioreactor system.

The T50/30R process in perfusion bioreactors yielded rAAV9 production per cell comparable to STPP, but results lacked consistency without a resting step. Electrostatic and hydrophobic interactions of the polyplexes with the cell membrane are crucial for the polyplex endocytosis [31, 33, 35, 36]. Interfering components produced by cells, such as extracellular vesicles, have been reported to affect the net positive charge of the polyplexes and thus impact transfection [72]. These inhibitory components are present in the perfusion bioreactor and are potentially retained by the 0.2 µm pore size hollow fiber filter ATF, partially or completely. However, they are probably not present in the STPP where centrifuged cells have been resuspended in fresh medium. It is highly likely that the presence of these interfering components weakens the net positive charge of the polyplexes and thus affects the electrostatic and hydrophobic interactions of the polyplexes with the cell membrane in the perfusion STR. One hypothesis is that the reduced net positive charge of the polyplexes might still be sufficient for the endocytosis in the case where a resting step is applied, due to the increased proximity between sedimented cells and polyplexes, and the elimination of any agitation from the impeller or the circulation in the lumen of the hollow fiber filter. Future studies should investigate the polyplex formation interference hypothesis in HCDC perfusion cultures. In the work of Mendes [49] at 10 MVC/mL VCD, the transfection was efficient with a high rAAV yield, which could be due to the use of a larger pore size hollow fiber filter for the perfusion compared to the present study. Additionally, rAAV production itself may induce cellular stress due to metabolic shifts or antiviral responses. However, it remains unclear why such stress would be more pronounced in the bioreactor than in STPP, aside from the complete medium renewal in the latter compared to partial retention in perfusion bioreactors. This points towards higher presence of inhibitory compounds as developed above, but does not exclude the potential impact of a lower level of component(s) favorable for the transfection in the perfusion STR compared to STPP.

4.2. Scale‐Up

Process scalability is critical for mammalian expression platforms, and it is not unusual that conditions optimized at small‐scale fail to translate to large‐scale manufacturing [43, 74] A key challenge in transient transfection is the need for complexation of DNA and transfection reagents before transfection. Transient transfection for recombinant protein production at 20 MVC/mL using 25‐kDa linear PEI without preformation of polyplexes prior to transfection has been reported [73], and eliminating this step would simplify large‐scale processing.

In this study, a “direct transfection” approach–where DNA and PEI MAX were added sequentially to the cell suspension–was tested in both STPP and perfusion STR. While successful in STPP, this method failed to generate rAAV in the bioreactor, regardless of reagent addition order (data not shown). A 15‐min preincubation of mixed reagents before transfection resulted in satisfactory rAAV production as shown for rAAV1 and rAAV9 in HCDC bioreactor scale. The failure of direct transfection may be linked to interfering components accumulating in HCDC perfusion cultures, where medium renewal is partial due to the use of 0.2 µm hollow fiber filters. Additionally, the increased DNA and PEI MAX concentrations at high VCD may have led to precipitation, which could be further exacerbated at large scale due to mixing dynamics challenges.

This proof‐of‐concept study demonstrated the feasibility of transfection at 50 MVC/mL and proposed a viable approach. The method included a resting step, which could be implemented at larger scales using a large sterile inflated bag with a 5% CO₂ atmosphere and kept at room temperature. However, this additional step may limit large‐scale practicality, and future studies will explore alternative solutions. Another promising avenue for process intensification is continuous transfection, which has been demonstrated for rAAV9 production and holds significant potential for further development [75].

5. Conclusions

This study proposed a transient transfection process for rAAV production at high cell densities (≥50 MVC/mL) without posttransfection dilution, except for the dilution introduced by transfection reagent addition. This intensified approach was developed and established in a 5 mL STPP system, was successfully applied across various rAAV serotypes, culture systems, transfection media, and cell lines (HEK293F and VPC2). A transfection density of 50 MVC/mL was identified as the highest recommended concentration for a robust and efficient STR process using PEI MAX, as higher densities led to precipitation during transfection reagent mixing. The required PEI MAX and plasmid amounts increased proportionally with cell density, impacting associated production costs. While the scope of this study did not include enhancing the cell specific rAAV productivity, future work will explore alternative transfection reagents, mitigate potential inhibitory effects on transfection, and eliminate the need for a resting step. The process was validated in a 200 mL perfusion STR, demonstrating that rAAV9 production per cell transfected at 50 MVC/mL in HCD STR was comparable to that in spin tubes at 15 and 1 MVC/mL, confirming the feasibility of this intensification strategy. This approach provided a 17‐fold intensification compared to transfection at 3 MVC/mL, a typical production density at scale, highlighting the benefit of significantly increased volumetric productivity at very HCD.

Supporting Information

Figure S1_suppInfo: Comparison of the efficiency of pGFP transfection carried out in duplicate cultures at different cell densities in FS293 medium, where the cultures were diluted to 1 MVC/mL just after the transfection for the production phase. FigureS2_suppInfo: Comparison of production of rAAV1, rAAV5 and rAAV9 using the same T50/10 process with transfection at 50 MVC/mL, followed by dilution at 10 MVC/mL and daily medium exchange in spin tube system. Figure S3_suppInfo: Comparison of rAAV9 production at 72 hpT (n = 2), summed from measurement in the cell lysate and in the supernatant, from transfection performed at 50 MVC/mL according to T50/30 process in STPP and differing by the medium. Figure S4_suppInfo: Transfections at 50 MVC/mL and 80 MVC/mL performed at 500 µL, 2 mL and 5 mL scales including evaluation of the resting step performed at 50 MVC/mL in 5 mL culture. Figure S5_suppInfo: Graphical representation of the HCDC T50/30R transfection process in a perfusion bioreactor.

Author Contributions

Conceptualization: Ye Zhang, Emil Sundäng Peters, and Véronique Chotteau. Methodology: Ye Zhang, Emil Sundäng Peters, and Véronique Chotteau. Investigation: Ye Zhang, Emil Sundäng Peters, Olalekan Daramola, Trudy Ann Tucker, and Véronique Chotteau. Formal analysis: Ye Zhang, Emil Sundäng Peters, Trudy Ann Tucker, and Véronique Chotteau. Resources: Véronique Chotteau, Johan Rockberg, and Diane Hatton. Data curation: Ye Zhang, Emil Sundäng Peters, Olalekan Daramola, Trudy Ann Tucker, and Véronique Chotteau. Writing–original draft preparation: Ye Zhang and Véronique Chotteau. Writing–review and editing: Ye Zhang, Olalekan Daramola, Trudy Ann Tucker, Diane Hatton, and Véronique Chotteau. Supervision Véronique Chotteau. Project administration: Véronique Chotteau, Johan Rockberg, and Diane Hatton. Funding acquisition: Johan Rockberg, Diane Hatton, and Véronique Chotteau.

Conflicts of Interest

The funders had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript; or in the decision to publish the results.

Supporting information

Supporting information

Data Availability Statement

The authors have nothing to report.