An unstructured model has been developed for the batch culture of a recombinant yeast with a GAL promoter, in galactose media containing ethanol. The major goal was modelling of the catabolic repression of galactose on ethanol consumption, inhibition on cell growth by ethanol, and plasmid instability. Batch cultures were carried out for several initial concentrations of galactose and ethanol for this purpose.
The decrease in the plasmid-bearing cell population during the culture was observed to be severe with a medium containing galactose as carbon source without ethanol(galactose-only medium). Its fraction at the end of the run(at 38 hours) was no more than 32%. The major cause of such reactor instability was found to be the plasmid instability since the depression of plasmid-bearing cell growth, which is another potential cause for the reactor instability, was observed to be negligible. The plasmid loss frequency could be represented appropriately by a saturation type function of specific galactose consumption rate.
When galactose and ethanol existed concurrently in the fermentor, ethanol consumption rate was repressed by galactose when its concentration was high. As galactose was consumed and its concentration became below certain level, galactose and ethanol started to be consumed simultaneously. When ethanol concentration was above 40 g/L, galactose consumption and thus cell growth stopped.
To observe the roles of ethanol, batch culture with ethanol-only media was carried out. Ethanol was an inhibitor on cell growth as well as the carbon source. The specific growth rate with ethanol was lower than galactose, but cellular yield for ethanol was higher. The plasmid-bearing cell and the plasmid-free cell differed from each other in terms of ethanol consumption kinetics. But, the kinetics of catabolic repression by galactose on ethanol consumption was almost the same for both cell types.
A model including all the aspects mentioned above was proposed. Some model parameters were determined directly from the experimental data. The other parameters were determined initially by an eye-bowling approach and eventually by using an evolutionary strategy which was known to be easy to apply and robust. Although our model was quite complex and had many parameters, the result of parameter determination was satisfactory and the model with those estimated parameters fit the experimental data quite well.