The dynamic perfusion process is a less complex alternative to the more established steady-state perfusion for process intensification in biomanufacturing. This study presents data on a dynamic perfusion process in 50 L scale using the single-use Xcellerex™ automated perfusion system (APS) for antibody production. Dynamic perfusion reached a viable cell density of 119 million viable cells (MVC)/mL with 95% viability over a period of 15 days. Its volumetric productivity was 1 g/L/d, which was equal to a steady-state perfusion and 2.5-fold higher than the reference fed-batch process. We showed that the media volume required for dynamic perfusion was ~40% lower compared to the steady-state perfusion, while maintaining productivity. In addition, dynamic perfusion required fewer interactions, demonstrating its operational simplicity. We conclude that the dynamic perfusion process can be used for more advanced automation operations, such as Process Analytical Technologies (PAT), to establish full automation capabilities.
Introduction
The current industrial standard for producing monoclonal antibodies (mAbs) in various cell lines is fed-batch cultivation. However, continuous processing through perfusion shows significant benefits for biomanufacturing in terms of product quality and yield, which makes it a good candidate for upstream process intensification. On one hand, the steady-state conditions promoted by perfusion contribute to improved product quality, and on the other hand, the increased overall space-time yield surpasses the one for fed-batch processes significantly.
Despite these benefits, the perfusion process has not emerged as the new standard in biopharmaceutical upstream processing. This is in part due to its high level of complexity and the potential difficulties linked to running processes aseptically over long periods of time.
To overcome the complexity of perfusion processes, a dynamic approach can be adopted. In a dynamic approach, a fixed volumetric media feed rate determines the growth rate and metabolic profile of the cell line over time. In other words, to promote a high cellular growth rate at the beginning of the culture, the feed rate per cell is higher, and towards the end of the culture, the feed rate per cell is lower resulting in a lower cellular growth rate (Fig 1).
This approach reduces the overall complexity of the process by removing the need for cell bleed and the daily perfusion rate adjustment. Also, the dynamic perfusion process has, due to its nature, a predictable timeframe and a shorter operational window that can be more easily fit into a manufacturing schedule.
In this study, we compared three upstream processes: fed-batch, steady-state perfusion, and dynamic perfusion. For each process we calculated:
- the volumetric productivity, which is the amount of product produced in the bioreactor per liter, per day.
- the space time yield (STY), a normalized metric to compare the overall productivity of upstream cell culture processes, regardless of whether they are batch-based or continuous.
- the total yield for each process in its entirety.
Fig 1. Representation of cell-specific perfusion rate (CSPR) vs vessel volume per day (VVD) over time in a simulated cell culture process.
MATERIALS AND METHODS
Cell culture media and feed preparations
We prepared the cell culture media by supplementing HyClone™ ActiPro™ cell culture media with a 10% (w/v) solution of HyClone™ Cell Boost™ 1 supplement, a 5% (w/v) solution of HyClone™ Cell Boost™ 3 supplement, poloxamer 188, and sodium bicarbonate in the proportions listed in Table 1. The media was then sterile filtered into an XDM storage bin bag. We filtered the base using the bottle top filter and transferred it to addition bottles.
Table 1. Cell culture media and base solutions used in this study
| Solution | Components | Concentration of individual solution | Concentration in final mixture |
| ActiPro™ media (seed train and bioreactor inoculation) |
N/A | 22.36 g/L | N/A |
| Perfusion media | 77.1% (v/v) ActiPro™ 10.8% (v/v) Cell Boost™ 1 12.1% (v/v) Cell Boost™ 3 Poloxamer NaHCO3 |
22.36 g/L 100 g/L 50 g/L N/A N/A |
17.23 g/L
10.75 g/L 6.09 g/L 2 g/L 1.39 g/L |
| Base (NaHCO3) | N/A | 75 g/L | N/A |
Seed train propagation in shake incubators
We recovered the cells from cryopreservation according to our internal standard protocol, and sub-cultured the cells every two to four days, according to the settings in Table 2.
Table 2. Culture conditions for shake flask cultures
| Parameter | Setting |
| Incubator agitation rate | 105 rpm |
| Orbital shaker diameter | 50 mm |
| Target inoculation cell concentration | 4 days: 0.3 × 106 viable cells/mL
3 days: 0.5 × 106 viable cells/mL 2 days: 1.5 × 106 viable cells/mL |
| Culture duration per split | 2-4 d |
| Temperature | 37°C |
| CO2 incubator concentration | 7.5% |
| Target viable cell concentration before split | 4-6 × 106 viable cells/mL |
| Target viability | > 95% |
| Volume of shake flask | ≤ 1000 mL |
| Working volume of shake flask | ~1:4 liquid to flask volume ratio |
| Culture medium | ActiPro™ with 37.5 µM MSX |
N-1 seed train cell culture in ReadyToProcess WAVE™ 25 bioreactor for Xcellerex™ XDR-50 bioreactor perfusion culture
We expanded the cells from shake flask cultures into a ReadyToProcess WAVE™ 25 bioreactor as a final N-1 step, and cultured the cells according to the settings in Table 3.
Table 3. Culture conditions for N-1 seed culture in ReadyToProcess WAVE™ 25 bioreactor
| Parameter | Setting |
| Culture working volume | 10 L |
| Rocking settings | 22 rpm, 6° angle |
| Temperature setpoint | 37°C |
| pH | Measured but not controlled |
| DO | Measured but not controlled |
| CO2 flow | 7.5% |
| Gas flow rate at start | 0.25 LPM |
| Target cell concentration at inoculation | >0.5 × 106 viable cells/mL |
| Target cell concentration at harvest | 6 × 106 viable cells/mL |
| Target viability | > 95% |
| Culture duration | 4 days |
| Culture medium | ActiPro™ |
Perfusion cell culture–50 L culture in Xcellerex™ XDR-50 bioreactor and Xcellerex™ APS
Before inoculating the XDR-50 bioreactor, we conditioned the cell culture media in the bioreactor for three days (Table 4). The air and CO2 gas mixture represents 7.5% of the total gas flow. We seeded the cells from the N-1 bioreactor into the XDR-50 bioreactor to equal 1 × 106 cells/mL in 50 L working volume. After inoculation, we used ActiPro™ cell culture media to top up to the final working volume. The running conditions for the perfusion cell culture process are listed in Table 5. The operating parameters for aeration and agitation were determined using the Cytiva Bioreactor Scaler.
Table 4. Conditioning parameters in Xcellerex™ XDR-50 bioreactor prior to inoculation
| Parameter | Setting |
| Media fill volume | 36 L |
| Temperature setpoint | 37°C |
| Vessel temp PID parameters | P=4; I=30; D=0; DB=0 |
| Stirrer speed | 100 rpm |
| Fixed air flow (1 mm sparger) | 1.85 lpm |
| Fixed CO2 flow (1 mm sparger) | 0.15 lpm |
| Condition duration | 3 d |
Table 5. Culture conditions for perfusion process in Xcellerex™ XDR-50 bioreactor
| Parameter | Setting |
| Culture working volume | 50 L |
| Temperature setpoint | 37°C |
| Vessel temp PID parameters | P=4; I=30; D=0; DB=0 |
| Stirrer speed | P/V 40 (50 L=101 rpm) |
| pH setpoint | Day 0: 7.1 ± 0.05
Day 1 and onward: 6.95 ± 0.15 |
| pH control | Split range: Lower MFC 3/ Up Pump 1 |
| pH PID parameters | P=0.5; I=1.5; D=0 |
| DO setpoint | 40% |
| DO control | Lookup table 1: Air (MFC1) Lookup table 2: O2 (MFC2) |
| DO PID parameters | P=0.5; I=5; D=0; DB=0 |
| Gas flow control strategy | MFC1 Air (20 µm)
MFC2 O2 (20 µm) MFC3 CO2 (1 mm) MFC4 Air (1 mm) |
| Look-up table for DO control | Look Up Table1 Mapped to MFC 1 0 0.1 SLPM 30 0.3 SLPM 100 0 SLPM Look Up Table2 Mapped to MFC 2 0 0 SLPM 30 0 SLPM 100 2 SLPM |
| Fixed air flow for CO2 stripping (1 mm sparger) | Start at 0.5 lpm and increase incrementally with 0.5 lpm up to 3.5 lpm (0.07 VVM) when pCO2 > 15 kPa. |
| Target inoculation cell concentration | 1 MVC/mL |
| Target viable cell concentration | 80 MVC/mL |
| Target viability | > 85% |
| Culture duration | 14 days (minimum 10 days of perfusion) |
| Target VCD for starting perfusion | 4-5 MVC/mL |
| Perfusion rate (VVD) | Day 2-4: VVD corresponding to minimum and approximately 40 pL/cell/d until reaching 0.75 VVD
Day 4 and onward: 0.75 VVD |
| Recirculation flow rate | 3.06 L/min; Equals 1000 s-1 shear rate over filter membrane |
The samples were taken daily both from the bioreactor and from the permeate, and treated and analyzed according to Table 6.
Table 6. Sample plan for the Xcellerex™ XDR-50 bioreactor perfusion run
| Measured parameter | Assay type | Type of sample | Bioreactor volume (mL) | Permeate volume (mL) |
| Carbon dioxide | ABL9 | Untreated | 1 | N/A |
| pH (off-line) | ABL9 | Untreated | ||
| Viable cell concentration | Vi-CELL XR and XM40 | Untreated | 1 | |
| Viability | Vi-CELL XR and XM40 | Untreated | ||
| Glucose | Cedex Bio | Supernatant after centrifuged at 2500 RCF for 5 min | 1 | 1 |
| Lactate | Cedex Bio | |||
| Glutamine | Cedex Bio | |||
| Glutamate | Cedex Bio | |||
| Ammonium | Cedex Bio | |||
| IgG | Cedex Bio | |||
| LDH | Cedex Bio | |||
| Osmolality | Osmometer | 0.02 | N/A | |
| Product quality (MW, charge, N-glycan | HPLC-IEX, SEC, LC-MS | Supernatant from above also filtered through 0.2-0.45 µm. | 1 | 8 |
RESULTS
N-1 seed train in ReadyToProcess WAVE™ 25 bioreactor
We seeded the ReadyToProcess WAVE™ 25 bioreactor at 0.72 MVC/mL and 10 L working volume. Viable cell density reached 4.6 MVC/mL and a cell viability of 99.4% after three days of batch culture (Fig 2A). Population doubling time (PDT) remained steady between 22 and 30 hours (Fig 2B). After three days, we used the entire cell culture to seed the XDR-50 bioreactor. We connected one of the tubing on the Cellbag to one of the tubing on the XDR-50 bioreactor bag through ReadyMate™ disposable aseptic connectors (DAC) and transferred the cell culture using a peristaltic pump.
Fig 2. (A) Cell concentration and viability in the N-1 WAVE™ 25 bioreactor. (B) Population doubling time (PDT) in the N-1 WAVE™ 25 bioreactor.
Cell growth and IgG concentration during perfusion in XDR-50
We seeded the XDR-50 bioreactor to 1.1 MVC/mL and 50 L working volume and grew in batch mode for two days. On day 2, we started perfusion, at a viable cell concentration of 4.3 MVC/mL. The cells reached a plateau on day 10 and maintained a viability above 95% throughout the culture, reaching a final VCD of 119 MVC/mL on day 16 (Fig 3) when the culture was harvested.
Fig 3. Cell concentration and viability in XDR-50 perfusion culture.
The mAbs IgG titer in the bioreactor increased, in accordance with the increased viable cell concentration, up to 2.8 g/L on day 15 (Fig 4). The mAbs titer in the permeate and harvest was proportional to the VCD. The titer levelled out on day 11 and remained steady around 1.3–1.4 g/L until day 15. Hence, the product transmission steadily decreased over time to 45% on day 15.
Fig 4. IgG titer in bioreactor, harvest line, and sieving coefficient in XDR-50 perfusion culture.
Nutrient and metabolite concentration in the perfusion culture
In the initial batch phase, glucose was consumed from 5 g/L to 2.5 g/L (Fig 5A). Once perfusion started, glucose accumulated and peaked at 6.6 g/L on day 5 before decreasing steadily until day 12 when it was totally depleted. Lactate production followed the rapid glucose consumption and the fast-growing phase during the first six days (Fig 5B). On day 7, the culture switched to lactate consumption and was totally depleted on day 15.
Fig 5. Metabolite and nutrient concentrations in XDR-50 perfusion culture.
We observed glutamate consumption from the beginning, which was fully depleted by day 7 (Fig 5C). Lactate dehydrogenase (LDH) increased during the culture as cell concentration increased. However, the slope increased towards the end of the culture when the viability decreased (Fig 5D). Glutamine was stable around 0.9 mM between day 2 and day 7(Fig 5E). Afterwards, we observed a shift to production and later consumption. Ammonia concentration increased from day 8 and peaked at 5.1 mM on day 15 (Fig 5F).
Environmental conditions of the perfusion culture
We observed an increase in osmolality up to day 5, where it peaked at 350 mOsmol/kg (Fig 6A). This was related to base additions needed to control the pH during the first 7 days because of the high lactate production and increasing respiratory CO2 levels (Fig 6B).
Fig 6. Osmolality (A) and pCO2 (B) in the XDR-50 perfusion culture.
At the time of inoculation, the pCO2 concentration was 7.2 kPa (Fig 6B), which was in accordance with the CO2 gas mixture of 7.5% during media conditioning. Between day 1 and 3, we observed a slight decrease to 4.5–5 kPa, which was maintained during those three days. From day 4, the pCO2 concentration increased above 15 kPa, during the fast-growing phase, which we controlled by adjusting the stripping air gas flow through the 1 mm sparger. To control the foam generated by the increasing gas flows, which we applied to control DO and pCO2, an antifoam agent was added manually every day.
Perfusion rate and cell culture productivity
During the first 3 days, we increased the volumetric perfusion rate from 0.25 to 0.5 VVD (Fig 7). By the fourth day, we further increased it to 0.75 VVD, which was kept constant for the remaining culture time. The stepwise increase was to limit an overflow of nutrients in the early stages of the cell culture, which could eventually limit cell growth and productivity. In addition, the increase was set to mirror the volumetric perfusion rate to a cell-specific perfusion rate (CSPR) of around 40 pL/cell/d. Once the volumetric perfusion rate was set to 0.75 VVD, the CSPR decreased with increasing viable cell concentration, stabilizing at 7 pL/cell/d. At this point, the cells stopped growing due to nutrient limitations, manifested in the increase in LDH and ammonia. Overall, the perfusion rate was controlled accurately throughout the process.
Fig 7. Perfusion flows during the XDR-50 perfusion culture.
The volumetric productivity during the dynamic perfusion process increased until day 11 but levels off when entering the growth plateau at around 1 g/L/d (Fig 8). This is consistent with the steady-state perfusion culture. Two fed-batch cultures, 50 L and 500 L scales as references, display lower volumetric productivity compared to the perfusion cultures.
Fig 8. Volumetric productivity for XDR-50 dynamic perfusion culture (green), XDR-50 steady-state perfusion (orange), 50 L (light blue) and 500 L (dark blue) fed-batch cultures.
Space-time yield (STY) increases with increased productivity and reaches 0.61 g/L/d on day 15 for the dynamic perfusion culture (Fig 9). This is similar to the reference steady-state perfusion culture. The two reference fed-batch cell cultures show a similar STY (0.3 g/L/d) between themselves. However, this is lower than the perfusion cell cultures, which is consistent with higher productivity.
Fig 9. Space-time yield for XDR-50 dynamic perfusion culture (green), steady-state perfusion culture (orange), and fed-batch cultures of 50 L (light blue) and 500 L (dark blue).
Dynamic perfusion showed a very similar total accumulated yield to the steady-state perfusion, reaching 413 g and 437 g, respectively on day 15 (Fig 10). Compared to the 50-L fed-batch cell culture, with a total yield of 231 g on day 14, dynamic perfusion shows a 1.8 times higher output.
Fig 10. Accumulated product yield in dynamic perfusion (green), steady-state perfusion (orange) and fed-batch (blue) 50 L cultures.
Media consumption during cell culture
On day 15, the media consumption normalized to the vessel volume was 9.9 L/L in the dynamic perfusion process compared to 16.1 L/L in the steady-state perfusion process (Fig 11). This resulted in 39% less media being consumed in the dynamic perfusion process. In part, this can be explained by the lower volumetric perfusion rate and lack of cell bleeding in the dynamic perfusion process.
Fig 11. Media consumption per vessel volume
Media consumption during cell culture
The pressure profile in mammalian perfusion processes usually displays very low pressures. Trans-membrane pressure (TMP) was negative for the first 9 days but shifted to positive or zero for the remaining culture time (Fig 12). This is not a concern in mammalian cell culture perfusion processes, where the flow rates are too low to maintain any significant pressure across the filter membrane, and the permeate flow is controlled by the permeate control pump rather than TMP. dP is the pressure difference between the filter inlet and outlet. During the perfusion process, dP increased as more resistance was built up along the filter due to fouling. This is particularly an issue at the inlet of the filter.
Fig 12. Pressure profile in terms of dP and TMP in the XDR-50 perfusion process.
Discussion
It was clear that the volumetric productivity was higher in the dynamic perfusion culture than in the fed-batch cultures referenced. The volumetric productivity was approximately 1g/L/d in perfusion compared to approximately 0.4 g/L/d in fed-batch, which represents a 2.5-fold improvement. Additionally, we estimate that the volumetric productivity could be further improved by approximately 20% with additional filter optimization studies. Performing a filter change after day 15 of the perfusion process could benefit and prolong the culture time significantly.
In addition, we observed that the media consumption for the steady-state perfusion process was nine times higher compared to a fed-batch process. However, the dynamic perfusion process needed only five times more cell culture media to achieve a similar daily accumulated titer as the steady-state perfusion culture.
Overall, dynamic perfusion was very easy to operate, and only required performing a daily manual sampling and monitoring of process parameters. This operation, due to its simplicity, could easily be fully automated by enabling automated sampling, PAT and remote monitoring.
Conclusion
In this study, we demonstrated a successful 50 L scale dynamic perfusion process in Xcellerex™ XDR-50 bioreactor with Xcellerex™ APS operated at 0.75 VVD over 15 days in total.
We demonstrated that the volumetric productivity increased by 2.5× compared to fed-batch processes. The simplicity in operating a dynamic perfusion application was significant in comparison to other perfusion processes (e.g., steady-state perfusion and hybrid perfusion), but also compared to fed-batch processes with less manual interactions.
Finally, we conclude that dynamic perfusion processes are an appealing alternative to both fed-batch processes and to more advanced perfusion processes. Compared to steady-state processes, there is lower media consumption and reduced complexity in a much shorter time, while still maintaining a similar volumetric productivity. In comparison to fed-batch processes, the volumetric productivity of the dynamic perfusion process is significantly higher, and with the potential of being less labor-intensive within a very similar time frame.
CY55065-20Nov25-AN