Once an upstream process has been successfully established, it must be scaled for clinical manufacturing. The scale-up process typically requires a series of intermediate volumes and pilot runs to increase a 3L process to 2000L. This stepwise approach is time consuming, requiring up to three months and slowing progress towards important milestones.

This article describes the direct scale-up of a 3L monoclonal antibody (mAb) production process to 2000L without requiring intermediate-scale steps at 200L and 1000L or pilot runs, ensuring safety, efficiency, and robustness. Using this approach, scaling for production runs becomes faster, more predictable, and consistent. Elimination of intermediate steps and pilot runs can accelerate time to market and deliver a competitive advantage.

Three key elements of a streamlined scale-up process will be highlighted: optimizing cell culture process at 3L scale, adjusting sparging and spinning at the 2000L level, and finally, design of an optimized clarification step. While this study was performed using a CHO cell line, the methodology applies to other cell lines and recombinant proteins.

Optimizing Cell Culture
Cell expansion is a critical step in process development in which a wide variety of culture conditions must be adjusted to optimize cell growth. The whole cell expansion process takes about two to three weeks at 2kL and starts with the thawing of viable cells from a cell bank vial transferred into a series of increasing volume shake flasks. The cells are then moved from the flask into a wave mixed bioreactor and then into a 200L bioreactor. This bioreactor is the final seed step, also known as the N-1 bioreactor. Eventually, the cell expansion is pursued in the 2000L bioreactor, where the production of the therapeutic molecule is finally occurring.

The first steps in establishing a robust cell expansion are to determine shake flask, wave bag, and N-1 bioreactor volumes and then set minimum and maximum conditions for each container, including temperature, CO₂, and stirring.  Then, a target for viable cell density (VCD) must be set. The goal is a sufficient quality of cells reaching the seeding specification and volume while ensuring exponential growth; two key parameters affecting VCD are passage and amplification. It is important to define how VCD differs depending on the day of passage and initial VCD of the seeding to move the cells from the initial container to the next larger container, thus defining your cell expansion train. Having too many or too few cells in the shake flask or wave bag bioreactor can affect the growth and expansion timing.

Once these foundational culture conditions are determined, a process development study at the 3L bioreactor scale is conducted to determine the best production parameters. This step ensures that the highest therapeutic protein production will be achieved while keeping the impurities within an acceptable range prior to the purification stages. Figure 1 summarizes results obtained in 3L and 2000L Mobius Bioreactors using the scale-up approach presented in this article, in which key process parameters were plotted against the number of days in culture (Note: all Figures appear in the image slider at the top of this article). Parameters include VCD (A) and viability (B). A decrease in viability could indicate the release of impurities into the medium, including host cell proteins (HCP), DNA, and cell debris which would impact the clarification step and further downstream purification. Other parameters to assess include lactate (C) and titer (D).

Sparging and Stirring at 2000L Scale
In addition to optimizing the cell culture process, important considerations when scaling from 3L to 2000L without intermediate steps are the system characteristics of the bioreactor, as these can affect the mixing and sparging conditions. Mixing efficiency can vary based on the impeller type, position, size and geometry of the tank. Sparging efficiency can vary based on the sparger size, position, and bubble size. Differences in fluid movement and the gas/liquid interaction can result from baffle type and position. In addition, regulation systems and equipment capacities do not vary in a linear manner and therefore will not affect growth and production in a linear manner. 

To address these challenges, use oxygen mass transfer modeling prior to production and use in-line proportional-integral-derivative (PID) regulation optimization during production.

Oxygen Mass Transfer Modeling
Comparable dissolved oxygen (DO) levels must be reached at 3L and 2000L bioreactors to ensure similar process performances as this critical substrate directly impacts cell metabolism. The oxygen mass transfer model was designed to maintain a stable K_La in a different bioreactor. K_La is the capacity of the bioreactor system to deliver oxygen to the cells in given operating conditions and is essential for accelerating the scale-up process; it is defined by the equation K_La = f (P/V; Vs).  The main parameters that can cause changes in K_La are the velocity of the gas used for sparging (Vs) and the power input of the stirring (P/V).

To create the model, a maximum and minimum for air flow (Vs) and power input (P/V) are first defined for different bioreactor geometries. The resulting KLa data are then used to create spatial models for different bioreactor scales (Figure 2).

A proof-of-concept study compared the use of K_La with the use of the volume of air under standard conditions per volume of liquid per minute (VVM) to ensure scalability and reproducibility of processes at 3L and 200L. VVM represents the quantity of gas purged into the bioreactor.

Using the constant VVM approach, stirring was set to different values at 3L and 200L scales to achieve the same volume metric for power input for the two bioreactor sizes at approximately 10 W/m³ to go from 75 mL/min at the 3L scale to 7.5 L/min at the 200L scale. As shown in Figure 3, there was much greater oxygen transfer to cells in the 200L bioreactor (K_La of 31.8 h^-1 against 6.1 h^-1 at 3L) is much higher. In contrast, keeping the K_La as similar as possible between the two scales (6.2 h^-1) resulted in better dissolved oxygen control (Figure 4).

The scale-up strategy using a constant K_La (and power input) is a successful approach to improve control of dissolved oxygen and reduce the maximum O₂ injection rate, which was set to 1.6 L, to a level five times less than with the standard VVM approach. A direct result of this approach is the scalability and reproducibility of the process, as measured by titer and specific productivity, at both 3L and 2000L scales (Figure 5). With the constant VVM approach, a performance gap was observed. 

PID Optimization
A proportional–integral–derivative controller (PID controller) is a control loop mechanism employing feedback for applications such as a bioreactor that requires continuously modulated control. A PID algorithm consists of three basic coefficients: proportional, integral and derivative, which are varied to get an optimal response.

Optimization of PID for oxygen regulation in the bioreactor must be performed in-culture with high oxygen consumption levels, which occurs during the expansion phase before reaching maximum VCD. For this purpose, use the Ziegler & Nichols tuning method. This method starts by zeroing the integral and differential gains and then raising the proportional gain until the system is unstable.

Fine tuning the PIDs, in this case, the 1/P, can further improve dissolved oxygen level regulation. This cannot be transferred from one system (software/hardware) to another. The step must be performed with the different bioreactors, just as for the K_La /Power input modelization.

To use the PID method for optimizing sparging and stirring, set the oxygen flow rate at a high value. The deadband, which creates a window in which the PID controller maintains the system output, allows a continuous action of the three parameters. To determine the ultimate gain (Ku), I and D values are fixed at zero and P increased until it reaches Ku. The goal is a quick, stable, and consistent oscillation (Figure 6).  The Ku and period of oscillation (Tu) are then used to calculate the associated I and D values. This information facilitates the adjustment of regulation needed for bioreactor scale-up.

Clarification Design
Before downstream purification, the bioreactor harvested cell culture fluid must be clarified to remove cell debris and this step must be scalable. Selection and sizing of the appropriate depth filtration cassette at the smaller scale is a critical first step. During development studies, pressure and volume monitoring will generate a resistance model (pressure/flux) needed to enable sizing for the 2000L process based on small-scale data with a safety margin. Indeed, manufacturing limitation such as equipment footprint or pump maximal flow rate does not enable a direct linear scale-up from bench scale to manufacturing scale.

Manufacturing constraints such as bioreactor volume (L), filtering area (m²) and process time (h) can be entered into the model to determine the resulting pressure set below the maximum operating pressure (Figure 7). Use this tool to advance directly to the 2000L scale from a 3L bioreactor.

Conclusion 
This whitepaper describes a direct, efficient and robust scale up of a mAb production process from a 3L bioreactor to 2000L without any intermediate volumes. The knowledge generated during process development to define process tolerances for volume, gas level, pH, glucose, as well as the use of a model designed to keep the oxygen mass transfer coefficient (K_La) stable in any bioreactor size, were key success factors. Mixing efficiency, sparging efficiency, fluid movement and gas/liquid interaction can all vary among bioreactor sizes and were modeled and evaluated to maintain the same mass transfer coefficient (K_La).

This approach allowed the application of similar conditions for production at the 2000L scale as developed for the 3L bioreactor. By combining a deep understanding of process dynamics with the use of Mobius bioreactors, process scale-up from 3L bench-scale bioreactors to 2000L becomes faster, more predictable, and consistent. Pilot runs and intermediate-scale steps at 200L and 1000L can be eliminated, accelerating development workflows, time to clinic, and reducing costs. 

---

Steven Strubbe is a technical leader in the End-to-End Solutions team at MilliporeSigma. He is involved in coordinating activities related to upstream and downstream process development, scale up, technology transfer, and process validation for manufacturing in a GMP-certified environment. He received his Master in Engineering in Biotechnology at ENSTBB a biotechnology training center in France specialized in the production, purification and analysis of therapeutic recombinant proteins and monoclonal antibodies. BioReliance End-to-End Solutions is an established, full-service biologic CDMO with facilities in France, China and the U.S.