High level secretion of recombinant human serum albumin by fed-batch fermentation of the methylotrophic yeast, Pichia pastoris, based on optimal methanol feeding strategy.

The dynamic programming method was applied to obtain the optimal specific growth rate, mu, in the fed-batch fermentation using the recombinant human serum albumin (rHSA)-producing yeast, Pichia pastoris. Based on the relationship between the specific production rate, varrho, and the specific growth rate, mu, a simple mathematical model describing the growth and rHSA production was constructed and used for calculations. Two constraints, final volume and maximum methanol feed rate, were adopted for calculations and the optimal mu resulted as follows. That is, mu was initially at the maximum value, mu(max), then decreased gradually. Finally, mu decreased to the mu(min) that gave a maximum varrho. The decline of mu was revealed to be caused by the constraint for maximum methanol feeding rate, F(max), and F(max) was constant until mu decreased to mu(min). We tried to realize the optimal mu in the fed-batch fermentation by manipulating the methanol feeding rate and obtained it. However, the observed varrho was differed from the expected one. The discrepancy between the expected varrho and observed varrho after the change of mu suggests the inapplicability of the relationship between mu and varrho to dynamic situations where mu changes. To confirm this, simulation and fed-batch fermentation runs were carried out at a methanol feeding rate that would cause a continuous change in mu. The rHSA production was simulated well, suggesting the applicability of the relationship between mu and varrho in such situations. Discontinuity in the change in methanol feeding rate of the optimal feed pattern at the time mu changed is considered to be the cause for the discrepancy between the expected and observed varrho. Therefore, a new methanol feeding strategy that could mimic the changes in mu and varrho of the optimal strategy without a discontinuity in the feeding rate was sought using a mathematical model of fermentation by trial and error. This modification in the methanol feeding rate resulted in a considerably improved varrho and 18% increase in total rHSA production compared with those obtained by the optimal strategy.