Assessment of the interactions between bioprocess conditions and protein glycosylation in antibody-producing mammalian cell cultures

• Main Author: Jimenez Del Val, Ioscani
• Other Authors: Kontoravdi, Cleo ; Nagy, Judit
• Published: Imperial College London 2013
• Subjects: 660
• Online Access: http://ethos.bl.uk/OrderDetails.do?uin=uk.bl.ethos.616862

The pharmaceutical industry is going through a rather turbulent period. Many blockbuster drugs have fallen off patent over the past two years and many more are expected to do so in the near future. In response, pharmaceutical companies have continued searching for products that will replace those that have lost patent protection. However, drug development and approval is extremely time-consuming and costly. So that this critical issue is addressed, industry experts and regulatory agencies have jointly proposed the implementation of Quality by Design (QbD) principles in the development and manufacture of all new drugs. Adoption of QbD is expected to reduce drug development cost and approval time. It is also expected to encourage innovation by developing drugs, and the processes used to manufacture them, around the mechanisms that relate process inputs with end product quality. Within this context, monoclonal antibodies (mAbs) are currently the highest-selling products of the biopharmaceutical industry and are projected to account for nearly half of the world’s top-selling drugs by 2018. All currently commercialized mAbs contain N-linked glycans (complex carbohydrates) bound to their protein backbone. These carbohydrates, in turn, have been widely reported to impact the safety and efficacy of mAbs. Furthermore, it has widely been reported that bioprocess conditions heavily impact the composition and distribution of these glycans. For these reasons, mAb glycosylation is considered a critical quality attribute (CQA) of these therapeutic proteins under the QbD scope. Based on QbD principles, the objective of this thesis was to generate a mathematical model that mechanistically relates the effect of nutrient availability throughout cell culture with the glycan profile of a mAb. The model was constructed from three individual ones. The first model describes the N-linked glycosylation process which occurs in the Golgi apparatus. The second model is unstructured and describes cell culture dynamics. The third and final model describes the biosynthetic pathway for nucleotide sugars. All three models were developed independently, but were adapted with features so that they could be interconnected. The glycosylation model approximates the Golgi apparatus to a single plug flow reactor where resident proteins (glycosylation enzymes and transport proteins) are recycled from distal portions of the Golgi space to proximal ones. Optimisation-based methods were developed to estimate unknown parameters of the model. The cell culture dynamics model was developed to represent cell growth, nutrient consumption and mAb synthesis. It was originally based on Monod kinetics, but was adapted to include experimentally-encountered complexity. The model for nucleotide metabolism was heuristically reduced from 35 constituting reactions to 7. Additional mechanistic features were adapted or included to ensure model fidelity. Experimentally, batch cultures were performed with hybridoma (CRL-1606 from ATCC). Data for viable cell density, glucose, glutamine, lactate, ammonia and mAb titre were collected. Intracellular samples were produced by perchloric acid extraction. These samples were then analysed for nucleotide sugar content using a high performance anion exchange chromatographic method which was optimized to quantify eight nucleotide sugars and four nucleotides in 30min. mAb bound glycans were analysed by MALDI mass spectrometry. The experimental data was used to estimate the unknown parameters of the models. The models – along with their associated parameters – were then combined to produce a coupled model that mechanistically relates nutrient availability with mAb glycosylation-associated quality. With further validation, such a model could be used for bioprocess design, control and optimization.