Fast development of Pichia pastoris GS115 Mut(+) cultures employing batch-to-batch control and hybrid semi-parametric modeling.

In this paper, we implemented a model-based optimization platform for fast development of Pichia pastoris cultures employing batch-to-batch control and hybrid semi-parametric modeling. We illustrate the methodology with a P. pastoris GS115 strain expressing a single-chain antibody fragment (scFv) by determining the optimal time profiles of temperature, pH, glycerol feeding and methanol feeding that maximize the endpoint scFv titer. The first hybrid model was identified from data of six exploratory experiments carried out in a pilot 50-L reactor. This model was subsequently used to maximize the final scFv titer of the proceeding batch employing a dynamic optimization program. Thereupon, the optimized time profiles of control variables were implemented in the pilot reactor and the resulting new data set was used to re-identify the hybrid model and to re-optimize the next batch. The iterative batch-to-batch optimization was stopped after 4 complete optimized batches with the final scFv titer stabilizing at 49.5 mg/L. In relation to the baseline batch (executed according to the Pichia fermentation guidelines by Invitrogen) a more than fourfold increase in scFv titer was achieved. The biomass concentration at induction and the methanol feeding rate profile were found to be the most critical control degrees of freedom to maximize scFv titer.