Application of constraint-based modeling for in silico analysis and prediction of CHO cell glycosylation
Date
2018
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
University of Delaware
Abstract
Advancements in biopharmaceutical research, development, and
manufacturing have led to improved treatments for ailments including leukemia,
multiple sclerosis, and diabetes. Over 70% of current therapeutics, composed of
biopharmaceuticals such as recombinant proteins, are produced by Chinese Hamster
Ovary (CHO) cells. One of the most important protein modifications enabled by CHO
cells is glycosylation, or the sequential linkage of oligosaccharide chains of varying
saccharidic composition. As glycosylation is a primary factor for determining
biopharmaceutical safety and efficacy, there is a need for increased understanding and
control of CHO cell glycosylation. In silico kinetic modeling has been shown to
provide valuable insights that have furthered our understanding of glycosylation.
However, due to the complexity of glycosylation, previous models have been built
using a large number of parameters – many of which have unknown or assumed
values. The large parametric demands of current glycosylation models make them
difficult to use for a priori analysis and prediction of glycosylation. ☐ The goal of this work is to develop a model that can be used to predict CHO
cell glycosylation without the need for a large number of parameters. Successful
development of this system will give users a way to study and predict glycosylation
without requiring parameters for culturing conditions, therapeutic protein type, and
CHO cell line modifications. This thesis establishes the framework for a novel method
of modeling glycosylation in CHO cells through application of constraint-based
analysis. Constraint-based analysis has been used as a low-parameter alternative to
kinetic models for complex biological systems. This work creates a constraint-based
model for glycosylation and introduces a series of discrete parameters that control
metabolite flux through the network. The system is used to predict changes in
glycosylation due to cell line engineering and media optimization. Our model was
found to successfully predict the effects of an Mgat1 knockout on glycosylation
patterns and validate the quantitative effects of ST6Gal1 overexpression. Additionally,
our model was able to validate the presence of media supplementations such as
glucosamine, N-acetylmannosamine, galactose, uridine, and glutamine using only the
initial and final glycosylation pattern (glycoform) as an input. ☐ Unlike kinetic models, which are computationally demanding, complex, and
require the a priori knowledge of >100 parameters, our model can make successful
predictions and generate testable hypotheses for glycosylation using <20 parameters.
We have designed a simple graphical user interface (GUI) which employs custom-
built algorithms that can calibrate these parameters to any glycoform or simulate a
glycoform given a set of parameters entered by the user. This work shows how
development of a tunable constraint-based model can be used to predict and
understand complex biological phenomena such as glycosylation. Finally, our model
can generate testable hypotheses to explain the appearance of unexpected
glycosylation patterns, resulting in an efficient way to control CHO cell glycosylation.
Description
Keywords
Applied sciences, Chinese hamster ovary, CHO, Glycosylation, Modeling, Monoclonal antibody, Therapeutic protein