The key to a successful scaleup on any biological process is to maintain the exact conditions that a cell experiences with increased volume, according to Byron Rees, scientific and laboratory services manager at Pall Biotech. “By creating a clear strategy and then complementing that by working with well-characterized bioreactors, you can more effectively match quality through the scale-up process,” he said.
The current surge of interest in the “Bioprocessing 4.0” concept has been adapted to Pall’s scaling up process; in particular, the utilization of digital technologies, artificial intelligence and machine learning to allow process understanding and performance prediction through smart, data-driven approaches.
“Pall is doing a lot of work to incorporate these concepts as a standard of operation,” Rees continued. “It is not about fitting an industry trend. It is about working to create automated data that we can use to continually refine our processes for consistency and quality.”
Some of the important parameters that need to be carefully monitored include oxygen transfer rates and liquid-liquid mixing times, as these can be affected by reactor volume. Adding base solution to help lower pH in a culture will mix and equilibrate readily in a low volume bench top bioreactor. However, increasing the scale to 2000 L runs the risk of generating harmful pH gradients; similarly, the amount of oxygen available to cells should be maintained as the process is scaled up, so it does not become rate limiting. Agitator speeds and air flow rates directly impact both mixing time and oxygen transfer rates, thus understanding these relationships will allow definition of the operating conditions and their correlation to each scale.
Using the data generated on the Pall Allegro Stirred Tank Reactor (STR) systems, Rees and his team were able to demonstrate that by maintaining power input and superficial gas velocity one can deliver consistently similar mixing times and oxygen transfer rates. For process development, the adoption of these parameters allows one to test operating conditions to determine the best cell growth and product yields, that can be effectively operated at the final manufacturing scale.
Other factors to consider are the types of sparger used as a gas flushing device, vessel geometry, materials of construction, and how effectively carbon dioxide is removed from the bioreactor.
Rees considered the related problem of cellular aggregation when moving to large-scale production vessels. “In my experience, when you see a lot of cellular aggregation in process development stages, you need to try to mitigate it immediately. The reason being that as you scale up, it’s unlikely to resolve itself. Cell aggregates will have negative impact on the culture as they are more likely to be inviable and this will have an impact on the healthy cells in the culture.”
Typical causes of cell aggregation are incomplete or partial adaptation to a different growth medium or platform, e.g., adapting and adherent cells to suspension. Sub-optimal conditions in the bioreactor include poor mixing and pH gradients, or shear stress caused by excessive agitation or gassing. Again, Rees endorses a good scale-down model which will help avoid these conditions at the production scale.
Rees also addressed the choice of cell lines for biologics production. “If we can grow cells at the process development scale, we can normally take them forward by replicating the conditions that gave the desired results. However, if we see poor growth at the start of the development, we will generally take it back to the growth media and agitation strategy to learn what conditions produce the most favorable cell growth and production conditions. Some cell lines can be sensitive to shear stress, but this can often be remedied by adding more shear protectants to the growth medium as opposed to lowering agitation speed, where you may compromise mixing time or oxygen transfer.”
During process development the more data that may be extracted from a single run, the better it is for the life of the process. This is the point at which one can look closely at the platform and fully characterize the bioreactor run in terms of pH profile, dissolved gasses, and nutrients and metabolites to correlate results to events in the process. From such information you can determine optimum feed strategies, bioreactor controller output or, for gene therapy processes, the optimal time for transfection.
Rees summed up his current views of Pall’s honing of its technological platform, “As we transfer processes from labware into the benchtop reactors, and then into larger pilot scales, we are generating and learning from the data in real time with consistent overlaying of data to identify any differences which could indicate issues with scalability.”
