Optimization and Modeling of Bacterial Processes

Bifidobacterium longum and Streptomyces peucetius are two totally different bacterial species with respect to environmental conditions and process objective. These were used as model organisms in this study. The objective of the B. longum process was to produce viable biomass suitable for use as food additive. The objective of the S. peucetius process was to produce a medically significant compound toxic to the organism. This study presents the process optimization for both model organisms by the use of experiment designs. The critical down-stream processing stage concerning B. longum, freeze-drying, was also optimized using response surface methods. Yeast extract and glucose concentrations together with L-cysteine-HCl·H2O concentration were found most effective concerning B. longum ATCC 15707 growth. The first two were optimized with temperature using a CCC experiment design. Optimal growth and glucose consumption was achieved with temperatures as high as 40C, a glucose concentration of 20 g l and yeast extract concentration of 35 g l. With a mixture design, soy peptone, beef extract, bacto peptone and tryptone were identified as growth favoring medium components for S. peucetius var. caesius N47. A CCF experiment design was constructed for the optimization of environmental conditions concerning S. peucetius growth, glucose consumption and εrhodomycinone production. Raising the cultivation temperature to 35C favored growth and glucose consumption, but 30C was found best for ε-rhodomycinone production. At the optimal temperature, a high aeration control setpoint and a high pH value yielded the best results for all responses. The survival of B. longum ATCC 15707 during freeze-drying was found highly temperature dependant. With a proper temperature control strategy during the freeze-drying process over 160% better product activity was achieved with a 50% shorter drying time compared to constant temperature freeze-drying. Kinetic cultivation parameters concerning S. peucetius were investigated using both batch and continuous cultivation data. Coefficients for substrate consumption (YXS 0.536 g g and mS 0.54 mg g h) and product formation (YPX 12.99 mg g and mP 1.20 mg g h) were calculated from chemostat results, and a μmax value of slightly over 0.10 h was observed. These parameters were used in the kinetic modeling of the cultivation. A best overall fit from the kinetic modeling was obtained when the logistic equation was used for the modeling of growth. Metabolic flux analysis (MFA) was applied to the chemostat data. This implied that ε-rhodomycinone production is almost linearly dependant on the citric acid cycle (TCA) rate. The kinetic modeling approach gave relatively good simulation results, but could not be used for prediction. This was, however, successfully done using neural networks.