(701g) Controllability Analysis of Protein Glycosylation In CHO Cells

Biopharmaceuticals represent the largest growing class of therapeutics with a US market of approximately $50 billion in 2009 and an expected steady increase in future sales (1).  Many biopharmaceuticals, such as the therapeutic monoclonal antibodies Herceptin and Avastin, are produced as recombinant proteins from Chinese hamster ovary (CHO) cells that are cultivated in bioreactors.  As with other manufactured products, these therapeutic proteins are effective only when their product quality attributes (bioactivity, potency, purity, etc.) lie within a specific range of values.  Of the many factors that affect the quality and bioactivity of these proteins, arguably the most important is glycosylation, a post-translational modification in which a carbohydrate chain, termed a glycan, is added to a protein.  To function as intended in vivo, most therapeutic protein treatments validated for human use must have a precise distribution of glycans (i.e., specific percentages of glycans with specific sugar monomers such as galactose, sialic acid, or fucose).  Unlike other cellular processes such as DNA replication and protein production, however, glycosylation has no master template, and, as a result, glycan formation and attachment are subject to variability and are often non-uniform.  Consequently, regulatory agencies, such as the Food and Drug Administration (FDA) and European Medicines Agency (EMEA), are now encouraging biopharmaceutical manufacturers to develop strategies to control glycosylation online during production.  However, online glycosylation control, which will involve active monitoring and effective control of key process variables that are critical to product quality, is yet to be implemented in the biopharmaceutical industry for a variety of reasons, mostly attributable to the complexity of these bioprocesses, the non-availability of on-line measurements, and the lack of comprehensive control paradigms tailor-made for such processes.

Our goal is to develop—and validate experimentally—such a comprehensive strategy for effective online, real-time control of glycosylation.  To achieve this goal, however, it is imperative first to assess the controllability of the process of glycosylation. Specifically, with x defined as the vector of the percentages of each glycan form present in the overall glycan pool from a batch of therapeutic protein, we must determine if protein glycosylation can be directed from any initial state x(0)=x₀ to any arbitrarily specified desired final state x_f, in finite time, via admissible manipulations of available process variables and operating conditions. Unfortunately, even though there is sufficient mechanistic knowledge to enable the development of a first-principles process model of glycosylation, by the very nature of the glycosylation process, with its convoluted reaction schemes and complex network architecture of approximately 23,000 reactions and almost 8,000 glycoforms, it is entirely impossible to carry out standard theoretical closed-form, state-space controllability analysis with the resulting mathematical model.  While numerical analysis is always possible in principle, the intrinsic nonlinearity and sheer complexity of the mathematical model demands that even model simulations must be carried out systematically and judiciously if the required system controllability information is to be extracted efficiently from the simulation results. 

In this presentation, we discuss a simulation-based procedure for assessing the controllability of glycosylation predicated upon employing statistical design of experiments to vary glycosylation enzyme and sugar nucleotide concentrations (model inputs) strategically.  The resulting glycosylation responses (model outputs) are analyzed via ANOVA to identify the enzymes and sugar nucleotide donors that exert the most significant effects on the predominant glycans found in recombinant proteins produced with CHO cells.  The statistical analysis results are subsequently used to assess the degree to which glycosylation is controllable.  In addition to providing useful insight into the extent to which online glycosylation control is achievable in practice in an actual experimental process, we also discuss how the procedure provides a framework for developing directed approaches for the genetic engineering of cell lines to produce protein with a specific desired glycan structure.

[1] Aggarwal, S., What's fueling the biotech engine-2009-2010. Nat Biotechnol, 2010. 28(11): p. 1165-71.