(133g) Towards the Implementation of QbD to the Production of Therapeutic Monoclonal Antibodies with Desired Glycosylation Patterns

The US FDA and the European Medicines Agency have recently proposed the implementation of the Quality by Design (QbD) paradigm to the manufacture of biopharmaceuticals. Its implementation requires the use of all available knowledge of a given product, including the parameters that affect its quality, for the design, optimization and control of the manufacturing process. The goal is to ensure that quality is built into the product at every stage of the manufacturing process. Most licensed monoclonal antibodies (mAbs) are based on the immunoglobulin G isotype and contain a consensus N-linked glycosylation site on the Cg2 domains of their heavy chains. Studies have found that the oligosaccharides attached to this site dramatically influence the efficacy of mAbs as therapeutics either by reducing their serum half-life or by directly affecting the mechanisms they trigger in vivo, thus defining glycosylation as a critical quality attribute of mAbs under the QbD scope. It has been recently proposed that detailed mathematical models will play a critical role in the design, control and optimization of biopharmaceutical manufacturing processes under the QbD scope.To our knowledge, there are currently no mathematical models that relate mAb glycosylation with cell culture conditions.

Several reports have shown that glycosylation is directly affected by the intracellular availability of nucleotide sugar donors (NSDs), which are the co-substrates for the glycosylation reactions that occur in the Golgi apparatus. During culture, cells synthesize all the relevant NSDs from glucose through the nucleotide sugar metabolic pathway. In an effort to relate process conditions with mAb glycosylation, we have generated a dynamic mathematical model for this metabolic pathway. The kinetic model was reduced based on the methodology described by Nolan and Lee (Methods in Bioengineering: Systems Analysis of Biological Networks, 2009). In order to estimate the unknown parameters of the reduced model, the intracellular concentration of glucose and eight NSDs was monitored through the course of a batch culture of CHO cells. The cells were harvested and their metabolism quenched. The cells were subsequently lysed and their contents lyophilized for analysis. The intracellular glucose concentration was determined using a fluorescence-based assay kit and the intracellular NSD concentrations were determined with an ion-pairing reverse-phase liquid chromatographic technique. Finally, a series of fed-batch cell culture experiments were performed in order to validate the reduced model and assess confidence in the estimated parameters.

Our mathematical model for NSD metabolism was then coupled to a model for mAb N-linked glycosylation in the Golgi apparatus to generate a direct link between extracellular glucose concentration, which is a readily measurable process variable, and protein glycosylation. The model for mAb glycosylation approximates the Golgi apparatus to a plug flow reactor so that cisternal maturation is represented. Recycling of Golgi-resident proteins (glycosylation enzymes and transport proteins) was also considered. The rate expressions for glycosylation reactions were derived based on their reported kinetic mechanisms and transport of nucleotide sugar donors from the cytosol to the Golgi lumen was modelled to serve as a link between glycosylation and cellular metabolism. Optimization-based methodologies were developed for estimating enzyme and transport protein concentration profiles. The resulting model is capable of reproducing glycosylation profiles of commercial mAbs. It can further reproduce the effect gene silencing of the FucT glycosylation enzyme and cytosolic nucleotide sugar donor depletion have on the mAb oligosaccharide profile.

Such a combined model has great potential for the design, control and optimization of manufacturing processes that produce mAbs with built in glycosylation-associated quality as proposed under the QbD paradigm.