(100h) A Stochastic Model of Glycosylation of Monoclonal Antibodies

Most proteins expressed in eukaryotic cells undergo the post-translational modification known as glycosylation in which a glycan (sugar moiety) is attached to the core of the protein. The addition of the sugar can either be to the amine group on the protein (N-glycosylation) or to the hydroxyl group (O-glycosylation). Of these, N-glycosylation, in which the glycan is added to the core of the protein in the lumen of the endoplasmic reticulum of the cell, followed by subsequent modification in the Golgi apparatus, has the greater impact on protein structure and properties. (For instance, the different N-glycans attached to the protein affect immunogenicity, in vivo half-life, and the receptor recognition of the protein.) Several commercially relevant biopharmaceuticals such as monoclonal antibodies (mAbs), erythropoietin, etc., are glycosylated proteins whose final product quality characteristics depend on the distribution of the different glycoforms. In order to ensure consistent product quality, therefore, regulatory agencies are encouraging manufacturers to monitor and control the quality of the product during the course of manufacturing. This requires the implementation of an appropriate control strategy, and developing such a strategy for effective control of glycosylation requires a mathematical model that represents the process of glycosylation adequately. Several models exist for predicting the distribution of glycoforms in different cellular systems^1-4 on the basis of overall production of protein in the culture. However, the glycan addition to a protein is a molecular phenomenon occurring within the microscopic volumes of vesicles in the Golgi apparatus, essentially invalidating the continuum assumption for glycan concentrations. Under such circumstances, a stochastic model will provide a more appropriate representation of the mechanisms of glycosylation and glycoprotein formation.

In this work, we develop what is, to our knowledge, the first stochastic model of glycosylation, and discuss the key challenges in the implementation of such a model, along with our approach to handling these challenges.  We present results of our stochastic model’s prediction of the glycoform concentration profiles and compare them to corresponding deterministic model predictions.

References

1.         Umaña, P.; Bailey, J. E. Biotechnology and Bioengineering 1997, 55, 890-908.

2.         Krambeck, F. J.; Betenbaugh, M. J. Biotechnology and Bioengineering 2005, 92, 711-728.

3.         Hossler, P.; Mulukutla, B. C.; Hu, W. S. Plos One 2007, 2.

4.         Jimenez Del Val, I.; Nagy, J. M.; Kontoravdi, C. Biotechnology progress 2011, 27, 1730-1743.