A comparison of statistical approaches used for the optimization of soluble protein expression in Escherichia coli.

During a discovery project of potential inhibitors for three proteins, TNF-α, RANKL and HO-1, implicated in the pathogenesis of rheumatoid arthritis, significant amounts of purified proteins were required. The application of statistically designed experiments for screening and optimization of induction conditions allows rapid identification of the important factors and interactions between them. We have previously used response surface methodology (RSM) for the optimization of soluble expression of TNF-α and RANKL. In this work, we initially applied RSM for the optimization of recombinant HO-1 and a 91% increase of protein production was achieved. Subsequently, we slightly modified a published incomplete factorial approach (called IF1) in order to evaluate the effect of three expression variables (bacterial strains, induction temperatures and culture media) on soluble expression levels of the three tested proteins. However, soluble expression yields of TNF-α and RANKL obtained by the IF1 method were significantly lower (<50%) than those obtained by RSM. We further modified the IF1 approach by replacing the culture media with induction times and the resulted method called IF-STT (Incomplete Factorial-Stain/Temperature/Time) was validated using the three proteins. Interestingly, soluble expression levels of the three proteins obtained by IF-STT were only 1.2-fold lower than those obtained by RSM. Although RSM is probably the best approach for optimization of biological processes, the IF-STT is faster, it examines the most important factors (bacterial strain, temperature and time) influencing protein soluble expression in a single experiment, and can be used in any recombinant protein expression project as a starting point.