Abstract: For high cell density cultivation of Saccharomyces boulardii, we attempted to optimize medium composition and culture conditions. Plackett-Burman design was used to recognize the significant medium components. Subsequently, response surface methodology and artificial neural network (ANN) based on central composite design (CCD) were applied to model the relationship between dry biomass production and medium composition. The optimization of medium composition was carried out using genetic algorithm (GA). The results showed that the ANN model, more suitable for complex and nonlinear modeling, had better goodness of fit and prediction performance. The optimal medium composition was obtained as follows (g/L): glucose 40.52, peptone 36.8, corn steep liquor 17.32, KNO3 14, yeast nutrients 1.5, KH2PO4 0.6 and MgSO4 0.8 shake flask cultivation using the optimized medium gave a dry biomass yield of 8.21 g/L, which was 2.39 folds higher than that obtained from the original medium. Based on these results, we determined the optimal high cell density culture conditions for S. boulardii cultivated in a 1-L fermentor as follows: temperature 30 ℃, 10% inoculum, pH 5.0 and 40% dissolved oxygen. Furthermore, we scaled up the culture process in a 50-L fermentor with the addition of glucose and peptone to maintain the glucose concentration at 3 g/L and the ammonia nitrogen concentration at 0.06 g/L. The dry biomass yield of S. boulardii reached 51.21 g/L in the large scale experiment.

Key words: Saccharomyces boulardii, neural network, genetic algorithm, enrichment medium, high cell density fermentation