Modeling of recombinant enzyme inactivation and prediction of N-linked glycosylation site-occupancy and microheterogeneity.

摘要

The inactivation of the tissue-type plasminogen activator protein (tPA) by a glycation reaction with glucose was identified as another possible mechanism between hyperglycemia and cardiovascular disease. Kinetic modeling revealed identical rates of glycation for glycosylation variants of r-tPA. Glycation at a single residue was found necessary to result in enzymatic activity loss. Computational techniques identified possible residues in the protease domain at which inactivating glycation may occur. A glucose-independent inactivation mechanism, as a result of protein-protein interactions, was also observed and found dependent on r-tPA glycosylation at N184. Simulations were performed for the optimization of fed batch feeding parameters for the production of r-tPA in a stirred-tank reactor in the presence of these inactivation mechanisms. The optimal harvest period was identified as the total r-tPA activity of the culture approached a maximum value, which served as the objective function of the optimization. Feeding profiles in the presence and absence of specified metabolite control were examined.; Novel neural network-based models were developed for the prediction of N-linked glycosylation characteristics. Variable site-occupancy and microheterogeneity classification were found to be predictable quantities of polypeptide glycosylation. Intracellular oligosaccharide transfer to a polypeptide is known to be either robust or dependent on cell culture conditions during pharmaceutical production. Model predictions were optimized when based on an input of a portion of the polypeptide primary sequence. Further intracellular enzymatic processing of the oligosaccharide results in complex-type, high mannose or hybrid branching of the glycan structure. A neural network model was created for the prediction of the major fraction of a heterogeneous mixture of glycoforms. Predicted values of secondary structure elements and residue solvent accessibility were found to best predict neural network testing data sets. The primary structure was effectively eliminated from the neural network input vector space. These results further emphasized the notion that site-occupancy remains dependent upon the primary sequence of the polypeptide and glycosylation microheterogeneity remains governed by secondary structure elements and three-dimensional properties of the folded glycoprotein.