Conference Dates

May 6-11, 2018


Genome scale models (GSM) have become a useful tool to connect different omics dataset into a single computational framework, thus giving a good overview of the flux distribution and metabolites interconnections in a specific environmental condition. A community genome-scale metabolic network reconstruction of Cricetulus griseus and cell line specific models have been recently developed 1. The main objectives with the use of the published CHO DG44 model were to enhance industrial bioprocess performance by suggesting genetic or metabolic targets, as well as strategies for medium optimization, and by bringing more fundamental knowledge about CHO cell metabolism. In a first step, some corrections were required in order to improve the biological relevancy of the predicted intracellular fluxes. The optimization method chosen was Parsimonious Flux Balance Analysis, based on the assumption that the cell is using a minimum amount of enzymes to reach a maximized objective value, under steady state. As the predictions were generating a lot of infeasible cycles, silencing of amino acid transporters that do not involve protons or sodium in the model allowed to reduce the incoming flow of amino acids and led to disappearance of infeasible cycles in the flux distribution solution. Four reactions involved in central carbon metabolism were manually added in the model, and some reactions were removed from the model to improve predictions such as the cytosolic enzyme fumarase, mainly localized in mitochondria, or L-asparaginase which is not reported to be present in CHO cells. As initial predictions for lactate production rate were overestimated compared to experimental measurements, the assumption of lipid accumulation during cell culture was added in the form of a constraint for a minimal level of triglyceride synthesis in the model (Figure 1). In a second step, the accuracy of the prediction from the curated model was tested with three independent data set obtained from a fed-batch experiment with a CHO DG44 cell line producing a monoclonal antibody in 2L stirred tank glass bioreactors. For modelling with GSM, pre-calculated input values are required in order to constraint the model with the environmental conditions, and thus to give a prediction that is representative of the experimental condition. Uptake rates of essential nutrients initially present in extracellular medium and consumed as the cells grow were used as the limit for a maximum uptake rate in the model. The objective function chosen was maximization of growth rate or maximization of specific productivity. As a result, correlation coefficients between experimental value and prediction indicate a good fit for growth rate and specific productivity (Qp) (Figure 2). Predicted amino acid consumption rates were comparable to experimental values, indicating the accuracy of the predicted stoichiometric requirements for cell growth and energy, except for 19% of the fluxes studied (Figure 3). As the extracellular flux values seem to correlate with experimental data, prediction of intracellular flux rates were analyzed at different timepoints of the culture, showing the activation of multiple metabolic pathways. Based on the results obtained, optimization of cell culture medium was performed to identify the limiting metabolites during the process that could lead to an increased growth rate and Qp.

Please click Additional Files below to see the full abstract.

94-Poster 40.pdf (139 kB)

Included in

Engineering Commons