Model-based characterization of operational stability of multimeric enzymes with complex deactivation behavior: An in-silico investigation

摘要

A comprehensive characterization of the enzyme kinetics and stability is required in order to design biotransformation processes efficiently. In particular, the characterization of the operational stability (stability under process conditions) of biocatalysts lacks established procedures and the few reported procedures have hardly been evaluated by transfer to reactor scale. In the scope of this modeling study three potentially applicable characterization methods were tested: (i) progress curves recorded at different temperatures; (ii) isothermal continuous reactor operation at different (elevated) temperatures; and (iii) continuous reactor operation with a temperature gradient. Input data from those procedures were generated in-silico by introducing two virtual dimeric enzymes with defined, complex deactivation mechanisms. The applied degradation pathways account for experimentally proven phenomena, which were neglected in operational stability descriptions so far, such as multiple intermediate unfolding state and subunit dissociation. By introducing a random normal error data sets with typical experimental variance were generated. On the basis of these data a model-based experimental analysis for parameter estimation was applied which included the testing of different simplified models (with increasing complexity until over-parameterization became evident). The obtained fits were evaluated by the lack of fit variance and the significance level of the F-statistics. Both criteria indicated a dimeric model with two intermediate unfolding states as best description, which was the model with the highest similarity to the models that had been applied to produce the data in the first place. When subjecting the obtained parameterization to a virtual process design task the quality of the prediction could be easily tested by comparison with the prediction from the original mechanistic model that was used to produce the input data. The obtained results clearly indicate that the often used Lumry-Eyring type models should not be used as empiric description for potentially complex multimeric enzyme deactivation mechanisms as it is done in current operational stability investigations. Next, for the continuous reactor procedures a nice correlation between the obtained significance level of the F-statistics and the accuracy of process prediction was found demonstrating the usefulness of these experimental procedures for parameter estimation. Further, it was possible to estimate a specific threshold significance level that can be used as an indicator for reliable parameterizations with respect to process design.