(471b) A Holistic Modelling Framework Describing CHO Cell Metabolism and Both Antibody and Host Cell Protein Glycosylation
AIChE Annual Meeting
2019
2019 AIChE Annual Meeting
Pharmaceutical Discovery, Development and Manufacturing Forum
Modeling and Design of Biopharmaceutical Processes
Wednesday, November 13, 2019 - 8:21am to 8:42am
In the present work, a holistic model is utilized to describe changes in mAb and host cell protein (HCP) glycoform distribution in fed-batch cultures of Chinese hamster ovary (CHO) cells under different strategies of galactose and uridine feeding, accounting for enzyme regulation. Specifically, a previously developed software (GLYMMER) is used to construct the network for HCP glycosylation and subsequently reduce networkâs complexity to accurately describe the experimentally observed glycoprofile of HCPs in CHO cells. Stoichiometric tables extracted from GLYMMER are then fed to a kinetic modelling framework describing (i) CHO cell metabolism, (ii) Nucleotide Sugar Donor (NSD) synthesis and (iii) mAb N-linked glycosylation, resulting in a model that is able to simultaneously describe both HCPs and mAb N-linked glycosylation, based on cellular metabolism. The framework includes a refined glycosylation model that describes the reactions taking place in both the ER and Golgi apparatus, introducing this way the variability of initial, high-mannose structures as an input to the processing steps in the Golgi apparatus. The model uses five continuous stirred tank reactors to describe and segregate the reactions in the ER and Golgi apparatus, including four ER-localized and thirteen Golgi-localized enzymes. Substrate-driven enzyme regulation is also introduced in the model to account for the overexpression of glycosylation enzymes in response to culture supplementation with NSD precursors.
The modelling framework successfully describes the experimentally measured glycoform distribution of the secreted mAb and the intracellular HCPs of CHO cells in a fed-batch culture experiment, within a maximum 3% error range, while at the same time closely describing CHO cell metabolism and antibody production (R2 > 0.9). The proposed holistic modelling framework sets the basis for the in silico prediction and control of the product-of-interest and HCP glycoprofile, providing insights on the usage of NSDs towards mAb and HCP glycosylation. In addition, the glycosylation model can be easily adapted to broader reaction networks yielding highly complex glycan structures, using stoichiometric tables obtained by the GLYMMER software and automatically generating the reaction network while minimizing the simulation time by approximately 90% compared to the initial modelling framework.