- We established an N-1 perfusion process for HEK293 cells for the production of adeno-associated viruses 5 (AAV5) (Fig 1).
- Using a 50 L Xcellerex™ XDR-50 bioreactor connected to the Xcellerex™ Automated Perfusion System (APS), the cell culture achieved ~ 35 million viable cells/mL with high viability.
- The perfused cells were seeded in an Xcellerex™ XDR-10 bioreactor and transiently transfected to produce AAV5. The resulting AAV5 titer and full particles were above the acceptance criteria.
Fig 1. High-level workflow of the N-1 perfusion development and scale-up.
Why N‑1 perfusion for AAV manufacturing?
As AAV programs see increased demand in bioprocessing, upstream supply of cells for transfection are often a bottleneck on seed‑train time and scale. Introducing perfusion at N‑1 intensifies this step, and you can reach higher cell densities faster while continuously exchanging medium to remove by‑products. This improves cell fitness for transient transfection and enables higher production seeding densities and larger bioreactor seeding. Additionally, less seeding bioreactors are required.
Key results at a glance
- Peak viable cell density (VCD) at N‑1: 34.6–36.9 MVC/mL by Day 7; viability ~ 99%.
- Perfusion intensity: Constant cell-specific perfusion rate (CSPR) 80 pL/cell/d (scale-down model and large scale), translating from 0.23 to 3.3 vessel volumes/day (VVD) as biomass increased.
- Glucose control: Maintained ~ 4 g/L after perfusion start at large scale; Scale-down model stabilized via +4 g/L glucose and +3 mM L‑glutamine (Gln) supplementation.
- Recirculation shear: ~ 1000 s⁻¹ at 3 L/min across the hollow‑fiber cartridge (scale‑matched from scale-down model).
- Downstream impact: Xcellerex™ XDR‑10 bioreactor production with perfused seed delivered normalized viral genome (VG) and viral particle (VP) titers ≥ controls (VG 0.23 vs 0.21; VP 1.08 vs 1.00).
Introduction
As gene therapies expand beyond rare diseases into broader patient populations, there is growing demand for production of genomic medicines. This has increased the demand for viral vectors, such as adeno-associated viruses (AAVs), placing pressure on manufacturing processes to increase production.
Traditionally, AAVs are produced by transient transfection of HEK293 cells grown in batch mode. The cells are then transfected with plasmid DNA to insert the gene of interest. However, the cell expansion steps are long and labor-intensive, using multiple shake flasks and seed train bioreactors. These steps can slow production and introduce variability.
To increase the efficacy of viral vector production, perfusion can be introduced at the N-1 step for process intensification. This reduces the number of expansion steps by continuously supplying fresh medium while removing unwanted by-products. In addition, an optimized perfusion process allows higher cell densities and viability, further improving AAV productivity.
Here, we determined how N‑1 perfusion could be applied to HEK293 cells by characterizing the cells growth and metabolic behavior from small‑scale adaptation to large‑scale implementation. This allowed us to adapt to specific nutrient consumption rates, select a CSPR, and define when to initiate perfusion to maintain exponential growth. Ultimately, we confirmed that a high‑density perfused seed could reliably inoculate a production bioreactor and support robust transient transfection for rAAV5 manufacturing.
MATERIAL AND METHODS
The HEK293 cell line Expi293F (Thermo Fisher Scientific) was used in all experiments. Cell cultures were performed in shake flasks, TubeSpin bioreactors (pseudo‑perfusion), a 1.1 L scale‑down perfusion model, an Xcellerex™ XDR‑50 and APS N‑1 perfusion, and an Xcellerex™ XDR‑10 production run for transient rAAV5 expression.
Seed train in shake-flasks
Cells were thawed per standard operating procedure (SOP) and passaged in shake flasks every 3 to 4 d. To maintain exponential growth, maximum cell density was ≤ 6 million viable cells (MVC)/mL and the cells were re-seeded at each passage at 0.3 MVC/mL. The culture conditions are listed in Table 1.
Table 1. Culture conditions for shake flasks
| Parameter | Setting |
| Incubator agitation rate | 135 rpm |
| Orbital diameter | 50 mm |
| Target inoculation cell concentration | 0.3 × 106 viable cells/mL |
| Culture duration | 3—4 d |
| Temperature | 37°C |
| CO2 incubator concentration | 5% |
| Target viable cell concentration after 4 d | < 6 × 106 viable cells/mL |
| Target viability | > 95% |
| Flask volume | 250 mL |
| Flask working volume | 40 mL |
| Culture medium | HyClone™ peak expression |
rAAV5 transfection protocol
Transfection mix was prepared by pipetting plasmid DNA in unsupplemented HyClone™ peak expression medium with PEI MAX 40k transfection reagent, and incubated 15 min at room temperature (RT), before adding to cell cultures with gentle swirling. The transfection conditions are listed below (Table 2).
Table 2. Transient rAAV5 transfection conditions
| Parameter | rAAV5 protocol |
| VCD at transfection | 2 × 106/mL |
| DNA concentration | 2 µg/mL |
| Molar DNA ratio a:b:c a = pAAV RC5 plasmid b = pHelper vector plasmid c = pAAV inverted terminal repeats (ITR) green fluorescent protein (GFP) vector plasmid |
1.5:2:0.6 |
| DNA:PEI ratio | 1:2 |
| Transfection volume | 10% of total volume |
| Incubation time | 15 min |
| Temperature transfection | 37°C |
| Harvest time | 72 h |
qPCR and ELISA
Cells were lysed by adding 10% (v/v) lysis buffer (1.65 M NaCl, 5.5% Tween-20, 11 mM MgCl2) and incubated for 20 min at 37°C (shaking). DENARASE enzyme (40 U/mL) was added for 4 h at 37°C to degrade host DNA. Lysates were clarified (300 × g, 10 min, RT), and supernatants aliquoted (0.5 mL) for quantitative PCR (qPCR) and ELISA.
Pseudo-perfusion cell culture in tube bioreactors
Cells were inoculated in triplicate at 0.25 MVC/mL, cultured for 3 d in batch to mid-log growth phase before pseudo-perfusion. Cell count and viability were assessed daily. Culture conditions are listed below (Table 3).
Table 3. Pseudo‑perfusion conditions
| Parameter | Setting |
| Agitation / orbit | 240 rpm / 50 mm |
| Temperature | 37°C |
| CO₂ concentration | 7.5% |
| Humidity | 70% |
| Working volume | 12 mL |
| Perfusion medium | HyClone™ peak expression + 0.1% poloxamer |
| Perfusion rate | 1 VVD |
| Sampling and analytics | 0.6 mL daily; VCD/viability, osmolality, metabolites |
| Targets | 30–35 MVC/mL, > 90% viability |
Scale-down perfusion process in 1.1-L scale
All media and feeds were prepared per manufacturer instructions. For the first perfusion run, basal peak expression and perfusion media were supplemented with 0.1% poloxamer before sterile filtration. For the second perfusion run, the perfusion media was additionally supplemented with 4 g/L glucose and 3 mM L-glutamine.
For the bioreactor assembly, we mounted two impellers on the agitator shaft. The microsparger was a 15 µm disk and the macrosparger was an L-sparger with seven 1 mm holes. The feed flow kit was installed by welding the two parts of the kit after the flowmeter but before the pressure sensor. The kit was then welded to the bioreactor outlet tubing by minimizing bending and the distance between the bioreactor and the recirculation pump. The feed flow kit and the hollow fiber (HF) cartridge were connected through their ReadyMate™ disposable aseptic connectors (DAC). After conditioning of the flow kit, the pump head was connected to the pump motor and flowmeter. The retentate flow kit was also connected through the ReadyMate™ DAC with the secondary flow path clamped, and the other end of the kit welded to the bioreactor inlet tubing. The permeate flow kit was connected through the ReadyMate™ DAC on the top of the hollow fiber cartridge, with the lower permeate lines on the hollow fiber cartridge clamped, and the other end of the permeate flow kit welded to the harvest container.
Fig 2. Benchtop scale down model flowpath
The flow kits and flow path were primed, the Levitronix flowmeter was calibrated to zero, and the pump started recirculation for 1 to 2 h.
The bioreactor was inoculated at 0.5 MVC/mL at 1.1 L working volume. The perfusion process started on Day 3 post-inoculation. Spent media were analyzed daily for amino acids and dipeptide concentration. The perfusion process parameters are listed below (Table 4).
Table 4. Perfusion process parameters
| Parameter | Setpoint |
| DO (DO control) | 40% air saturation (O₂ on demand, 15 µm sintered sparger) |
| pH (pH control) | 6.95 ± 0.15 (upward: 7.5% NaHCO3; downward: CO2) |
| Temperature | 37°C |
| Agitation (P/V) | 30 W/m3 (Run 1) 40 W/m3 (Run 2) |
| Perfusion rate (CSPR) | 120 pL/cell/d (Run 1) 80 pL/cell/d (Run 2) |
| Recirculation flow rate | 153 mL/min (~ 2000 s-1 over HF cartridge) |
| Perfusion media | Run 1: HyClone™ peak expression + 0.1% poloxamer Run 2: HyClone™ peak expression + 0.1% poloxamer + 4 g/L glucose + 3 mM L Gln |
| Working volume | 1.1 L |
| Inoculation cell concentration | 0.5 MVC/mL |
Large-scale perfusion process
All media and feeds were prepared per manufacturer instructions. Perfusion cell culture media was supplemented with 0.1% poloxamer, 4 g/L glucose, and 3 mM L-glutamine.
We performed the large-scale N-1 perfusion using an Xcellerex™ XDR-50 bioreactor connected to the Xcellerex™ APS using a 50L APS bag, following the Operating Instructions. An N-2 seed culture was run in a ReadyToProcess WAVE™ 25 bioreactor, as detailed below (Table 5).
Table 5. ReadyToProcess WAVE™ 25 (N‑2) settings
| Parameter | Setting |
| Medium (supplements) | HyClone™ peak expression (0.1% poloxamer) |
| pH control (setpoint) | CO₂ and 7.5% NaHCO3 (7.1) |
| DO control (setpoint) | Cascade control using air and O₂ (40%) |
| Temperature | 37°C |
| Agitation | 22 rpm, 6° rocking angle |
| Gas flow | 0.25 LPM; O₂ under DO regulation |
| Inoculation | 0.3 × 106 cells/mL |
| Culture duration | 4 d |
| Target viability | > 96% (from inoculation to end) |
| Sampling / analytics | Daily: VCD, viability, pH, gases |
For the N‑1 seeding, we inoculated 40 L from the N-2 culture at 0.3 MVC/mL, according to the parameters below (Table 6).
Table 6. Parameters for N‑1 perfusion culture
| Parameter | Setting |
| Working volume | 40 L |
| Agitation (P/V) | 80 W/m3 |
| Perfusion strategy | Perfusion start: ~ 2 MVC/mL Perfusion rate (CSPR): 80 pL/cell/d |
| Temperature | 37°C |
| pH | 6.95 ± 0.15 |
| DO | 40% |
| Target inoculation | 0.3 MVC/mL |
| Target VCD | 35–40 MVC/mL |
| Target viability | > 90% |
| Culture duration | 6–7 d |
| Culture media | Expansion: HyClone™ peak expression + 0.1% poloxamer 188 Perfusion: HyClone™ peak expression + 0.1% poloxamer 188 + 4 g/L glucose + 3 mM L Gln |
| Recirculation flow rate | 3 L/min (~ 1000 s-1 shear across HF) |
| Sampling / analytics | Daily: VCD, viability, osmolality, metabolites (20 mL waste + 10 mL cells) |
Transient transfection in XDR-10
We inoculated an Xcellerex™ XDR-10 bioreactor with perfused N-1 cells at 0.5 MVC/mL, for batch growth and transient transfection as described earlier (Table 2). rAAV5 was harvested 72 h post-transfection. The operating parameters for the Xcellerex™ XDR-10 bioreactor are presented below (Table 7).
Table 7. XDR‑10 bioreactor settings
| Parameter | Setting |
| Medium (supplements) | HyClone™ peak expression (0.1% poloxamer) |
| pH control (setpoint) | CO₂ and 7.5% NaHCO3 7.1) |
| DO control (setpoint) | Cascade control using air and O₂ (40%). DO 100% before inoculation. |
| Culture volume | Inoculation at 5 L. Top up to 10 L prior to transfection. |
| Temperature | 37°C |
| Agitation | 111 rpm at 5 L; 140 rpm at 10 L (P/V ≈ 80 W/m³) |
| Spargers | 20 µm (DO) and 1 mm (pH and stripping) |
| Stripping strategy | Start air 0.05 LPM; increase 0.05 LPM steps if pCO₂ > 15 kPa |
| Inoculation | 0.5 × 106 cells/mL |
| Culture duration | 7 d |
| Sampling / analytics | Daily: VCD, viability, metabolites, osmolality, pH, pCO₂ |
| Transfection mix | See Table 2 |
RESULTS
Pseudo-perfusion cell culture in tube bioreactors
To determine a perfusion rate for the cells, we analyzed their growth profile and metabolite patterns. We performed 1 VVD of media and calculated the specific consumption rates of some of the most critical nutrients and determined the growth plateau due to nutrient limitation. Then, we established a suitable perfusion rate or supplemented with additional nutrients to sustain further growth. The average growth profile is shown in Figure 3.
Fig 3. Average viable cell concentration and cell viability for the pseudo-perfusion cell culture.
Nutrient analysis showed that the cell culture was depleted of glucose and lactate on Day 7 and 11, respectively (Fig 4).
Fig 4. Average glucose and lactate concentration during cell culture.
The average specific glucose consumption rate (qGlc) was 650 pg/cell/d during the exponential growth phase. To supply enough glucose, we determined 90 pL/cell/d as the minimum CSPR required.
Scale-down perfusion process (1.1 L scale)
We ran two scale-down models to determine cell growth. The perfusion cultures reached a viable cell concentration of 36.4 MVC/mL (Run 1) and 36.9 MVC/mL (Run 2) on Day 7 (Fig 5). In addition, cell viability remained above 98% in both cultures.
We also analyzed the spent media to determine the concentration of all amino acids and the Gln-Ala (alanine) dipeptide during the perfusion cell culture processes (results not shown). The first run required a perfusion rate of 120 pL/cell/d due to higher glucose and glutamine consumption (Fig 6). Based on this, for the second run we added 4 g/L glucose and 3 mM glutamine to the perfusion media to keep the residual glucose and glutamine concentration more stable and we were able to sustain a CSPR of 80 pL/cell/d.
Fig 5. Viable cell concentration and cell viability for the two scale-down perfusion cell culture runs.
Fig 6. Glucose concentration in the two scale-down perfusion cell culture runs.
Large-scale perfusion process in Xcellerex™ XDR-50 bioreactor and APS
The large-scale perfusion culture reached a viable cell concentration of 34.6 MVC/mL and a cell viability of 99.2% on Day 7 (Fig 7).
Fig 7. Cell growth during large-scale perfusion cell culture run.
Glucose consumption decreased from 6 to 4 g/L at Day 2, when perfusion started and was then maintained steady at around 4 g/L (Fig 8).
Fig 8. Glucose concentration in the large-scale perfusion culture.
Since the CSPR was kept constant at 80 pL/cell/d during the entire process, the volume added and consequently harvested increased with increased cell concentration (Fig 9). At the start of perfusion, the VVD is at 0.23 VVD but increases steadily up to 3.3 VVD towards the end of the process.
Fig 9. Perfusion flows in terms of CSPR and VVD in the large-scale XDR-50 perfusion cell culture.
Transient transfection in Xcellerex™ XDR-10 bioreactor with perfused cells
Cells from the perfusion cell culture were inoculated into an Xcellerex™ XDR-10 bioreactor bag at 0.46 MVC/mL in a 5 L starting volume (Fig 10). On Day 4, the cells were diluted with HyClone™ peak expression medium to a working volume of 9.5 L. The cells were transfected by adding the transfection complex solution to the bioreactor with a final volume of 10 L. After an additional 3 d of batch growth, the cell culture was harvested at a VCD of 5.2 MVC/mL and a viability of 88.9%.
Fig 10. Cell growth and viability of cell culture in XDR-10 bioreactor.
The transfection efficiency for rAAV5 production was comparable to the control experiments and produced rAAV5 titers and number of full particles above the acceptance criteria (Fig 11).
Fig 11. Normalized rAAV5 titers from control and XDR-10 cultures. Titers were analyzed with qPCR (VG/mL) and ELISA (VP/mL).
DISCUSSION
In this work, we examined the growth, metabolic, and process factors that determine N-1 perfusion performance. While perfusion is becoming more common, its application in viral vector manufacturing is still relatively new. Through targeted scale-down studies, we identified the key parameters required to design effective media, optimize perfusion operation, and enable successful scale-up.
What the scale-down models enabled
Our studies on the scale-down models helped run a successful scale-up:
- The preliminary growth profile and nutrient consumption rate profiles allowed us to identify nutrient limitations in the cell culture medium.
- We determined a working CSPR and a perfusion starting point to maintain exponential growth and high cell viability.
Media and supplementation strategy
Our pre-studies helped determine an appropriate perfusion cell culture media. Using commercially available media, and supplementing with glucose and glutamine, kept the residual glucose and glutamine concentrations stable and sustained a CSPR of 80 pL/cell/d without compromising viability.
Value of small scale screening
Running the initial screening cell cultures in small-scale volumes (12 mL tube bioreactors) provided insight into nutrient consumption rates and limitations. This allowed us to determine an initial CSPR for the scale-down perfusion runs.
Scaling considerations
To translate performance from the scale-down models to the large-scale XDR-50, three elements were critical:
- Filter:working volume ratio—to keep cell retention dynamics comparable.
- Cross flow rate (shear) across the hollow fiber—to align residence time and mass transfer.
- Flow path design—matching the volumetric ratio in both scales to align residence time in the recirculation loop.
Perfusion process operational complexity
Perfusion processes require increased cell culture media volumes and additional flow path components. However, it does not add complexity of operation for the user. Xcellerex™ APS automates most operations, limiting the daily tasks to sampling and perfusion rate adjustments.
Why this matters for AAV manufacturers
Introducing N-1 perfusion intensifies the seed train of HEK293 cells and AAV5 manufacturing. The main benefits are:
- Shorter production process with higher seeding density.
- Reduced costs and footprint by using smaller seeding bioreactors.
- Increased process robustness due to decreased variability in the process.
Fig 12. Different process options for N-1 perfusion in AAV manufacturing
Conclusions
- N-1 perfusion process for HEK293 cells consistently reached ~ 35 MVC/mL with high viability, using Xcellerex™ XDR-50 bioreactor with Xcellerex™ APS.
- Transfection efficiency from the perfused seed in the Xcellerex™ XDR-10 bioreactor, exceeded the acceptance criteria for rAAV5 titers and full particles.
- Effective small-scale analysis (with tube bioreactors and scale-down models) helped define CSPR, perfusion timing, and media supplements required for large-scale production.
- Process intensified for AAV manufacturing with N-1 perfusion with a shorter seed train, smaller seed bioreactors, and improved facility flexibility.
FAQs
What CSPR should I target for HEK293 N‑1 perfusion?
Empirically, ~ 80 to 90 pL/cell/d worked well in our system. Pseudo‑perfusion sizing indicated that ≥ 90 pL/cell/d was required to avoid carbon limitation; large‑scale success was achieved at 80 pL/cell/d with media supplementation.
When should I start perfusion?
Around 2 MVC/mL (mid‑log) balanced nutrient availability, viability, and growth rate in our runs.
Does N‑1 perfusion complicate operations?
The Xcellerex™ APS automation kept day‑to‑day tasks to sampling and perfusion‑rate adjustments, despite rising VVD at constant CSPR.
What were the production results with a perfused seed?
Xcellerex™ XDR‑10 transient transfection runs showed VG and VP titers at or above controls (VG 0.23 vs 0.21, VP 1.08 vs 1.00, normalized), meeting acceptance criteria.
CY57427