Scale-up studies are needed for assessing cell culture production system options and for testing nutrient supplementation techniques as well. With the many supplementation options available, choices need to be made as early in product development as possible because advantages can change with scale. One published fed-batch scale-up study testing from 3 L up to 2,500 L highlights items to be considered in addition to the nutrient supplementation process such as the impact of pH and CO₂ control (1). That thorough study showcases epratuzumab biomanufacturing, which passed two global phase 2 clinical trials and met all specifications for process robustness, product consistency, and reproducibility.

As cultures increase in cell concentration and volume during scale-up into bioreactors, not only are nutrients depleted, but waste products can accumulate to sufficient levels to limit product formation. As a metabolic end product of glucose metabolism, lactate is usually considered to be detrimental to cultures, but after it is synthesized to quite high levels, a number of cell lines will begin to use it through the more efficient citric acid cycle (also known as the tricarboxylic acid cycle or TCA cycle). That happens especially under low-glucose feeding strategies, but only as long as sufficient oxygen is present.

PRODUCT FOCUS:ALL PROTEINS

PROCESS FOCUS:PRODUCTION

WHO SHOULD READ:MANUFACTURING AND PROCESS DEVELOPMENT

KEYWORDS:CULTURE MEDIA, FED-BATCH, PERFUSION, MEDIA SUPPLEMENTS, PROCESS CONTROL, ANDDESIGN OF EXPERIMENTS

LEVEL:INTERMEDIATE

So it's important to define all physical and chemical parameters early into culture scale-up. Even the mode of oxygen delivery needs to be analyzed. Sparging with air or oxygen to maintain appropriate O₂ levels can become problematic because of excessive cell sheer, which may lower growth rates or decrease viability. Very tiny bubbles can be especially detrimental to cells. And sparging can also cause significant foaming at the surface of the bioreactor, which may also decrease the number of viable cells, although various detergent-like solutions are available to reduce the effect. Oxygenation limitations can become significant in very high-density cultures (as in hollow-fiber bioreactors) even where bubble sheer is not an issue.

Scale-Up Modeling

The number of bioreactor runs that can realistically be performed on a production system is limited. Modeling has proven to be successful in defining appropriate nutrient supplementation schemes, especially when it's validated through a limited number of scale-up cultures. Most examples given here represent relatively complicated models used with smaller-volume processes, so they may offer relevant options to be considered.

Selvarasu used available metabolic pathway maps for testing with hybridoma cultures producing monoclonal antibodies (MAbs) in fed-batch mode (2). Bioreactor sampling for major media components such as glucose, glutamine, and other amino acids (as well as ammonia and lactate waste products and MAbs) were compared with values predicted from a model. Results showed the model to be fairly accurate and useful for development of a nutrient supplementation protocol, especially when it was used in conjunction with smaller-scale bioreactor runs.

Henry et al. used logistic fit of simpler noninstrumented batch and semicontinuous cultures to compare cell growth kinetics, nutrient consumption, metabolite, and product formation data with those of more complex continuous and perfusion models (3). They found good correlation. Goudar et al. also found logistic modeling to be superior to more commonly used standard polynomial fitting or unstructured kinetic modeling in describing batch and fed-batch data (4).

In a different approach, Dhir et al. used “dynamic optimization” to arrive at a superior supplementation protocol (5). They developed a control algorithm that allowed on-line readjustment of supplementation depending on the differences in real-time parameters observed in a culture to those predicted by the model. Using their model to optimize hybridoma growth in a fed-batch bioreactor, the team obtained a 44% increase in cell density, resulting in a 31% productivity increase over a fixed, off-line optimization protocol. The model should be usable for other cell systems too.

Another modeling system — a macrokinetic model based on stoichiometric balance — was developed by Zhou et al. using a 653 myeloma (6). It describes dynamic balances among lactate, alanine, reduced nicotinamide adenine dinucleotide (NADH), and adenosine-5′-triphosphate (ATP) during metabolism of glucose, glutamine, and other amino acids. The model provides data to assist in estimation of specific substrate consumption rates and specific growth rates as well as oxygen uptake and acetyl–coenzyme-A (a-CoA) formation rates, which the team validated using batch and fed-batch culture experiments.