(283b) Model-Based Design of Experiments for Antibody Glycosylation in CHO Cell Culture

With a growing number of monoclonal antibody (mAb) drugs entering the global biopharmaceutical products market, it has become more critical to achieve desired product quality targets in a consistent manner. Glycosylation[*] is considered as a critical quality attribute in biologic manufacturing because it can affect the half-life, immunogenicity, and pharmacokinetics of therapeutic mAbs [1]. Multiple mechanistic models have been developed to predict the glycosylation distribution of mAbs for control and optimization purposes [2, 3]. However, these models include many kinetic and stoichiometric parameters to estimate, making it challenging to adapt these models across cell lines and culture conditions.

To overcome this challenge, we first developed a package in Python to solve a multiscale model consisted of a cell culture submodel to predict cell growth and mAb production, a nucleotide sugar donor (NSD) synthesis model to predict intracellular NSD concentrations, and a golgi model to predict the mAb glycoform distribution. The multiscale model includes over 100 parameters and it is well-known that biological system models usually have severe identifiability problems, making it challenging to estimate these parameters reliably without cautious design of experiments (DOE) [4]. This motivated us to apply a receding horizon method to design the optimal experiments for accurate estimation of the model parameters. The DOE algorithm was deployed to drive the experiments for producing mAb from Chinese hamster ovary (CHO) cells. By adjusting the feed flow rates and the concentrations of the medium supplements online, the most informative data are generated for parameter estimation. If the accuracy of the parameters is satisfactory, the experiment is terminated; Otherwise, the feeding strategies are adjusted based on the DOE algorithm to generate more data. The iterative DOE and parameter estimation are continued until the parameters reach the desired accuracy.

Our results show that the model-based DOE results in significant reduction in parameter confidence intervals. The validated and sufficiently predictive model is used to optimize the feeding strategies towards a more desired glycosylation profile of the mAb while maintaining cell viability and productivity. Simulation case studies are supported by experimental results. These models and methodologies support the development of a Quality-by-Design (QbD) approach for control of glycosylation profiles for biotherapeutic proteins.

[1] E. Edwards, M. Livanos, A. Krueger, A. Dell, S.M. Haslam, C. Mark Smales, D.G. Bracewell, Strategies to control therapeutic antibody glycosylation during bioprocessing: Synthesis and separation, Biotechnol Bioeng 119(6) (2022) 1343-1358. https://doi.org/10.1002/bit.28066.

[2] P. Kotidis, P. Jedrzejewski, S.N. Sou, C. Sellick, K. Polizzi, I.J. Del Val, C. Kontoravdi, Model-based optimization of antibody galactosylation in CHO cell culture, Biotechnol Bioeng 116(7) (2019) 1612-1626. https://doi.org/10.1002/bit.26960.

[3] Y. Luo, V. Kurian, L. Song, E.A. Wells, A.S. Robinson, B.A. Ogunnaike, Model‐based control of titer and glycosylation in fed‐batch mAb production: Modeling and control system development, AIChE Journal (2023) e18075. https://doi.org/10.1002/aic.18075.

[4] C. Kontoravdi, E.N. Pistikopoulos, A. Mantalaris, Systematic development of predictive mathematical models for animal cell cultures, Computers & Chemical Engineering 34(8) (2010) 1192-1198. https://doi.org/10.1016/j.compchemeng.2010.03.012.

[*] Glycosylation is the controlled enzymatic modification of an organic molecule by addition of a sugar molecule.