New fusion protein systems statistically designed to avoid inclusion body formation in Escherichia coli.

Abstract

Using human interleukin-3 (hIL-3) as a model heterologous insoluble protein, three gene fusions were constructed that code for a native E. coli protein at the N-terminus and hIL-3 at the C-terminus. The three native E. coli proteins, NusA, GrpE, and bacterioferritin (BFR), were chosen based on their favorable cytoplasmic solubility characteristics as predicted by a statistical solubility model for recombinant proteins in E. coli. Modeling predicted the probability of soluble fusion protein expression in the following order: NusA (most soluble), GrpE, BFR, and thioredoxin (least soluble). Expression experiments showed that NusA/hIL-3 fusion protein was expressed almost completely in the soluble fraction while GrpE/hIL-3 and BFR/hIL-3 exhibited partial solubility at $\rm 37\sp\circ C.$ Thioredoxin/hIL-3 was expressed almost completely in the insoluble fraction. hIL-3 was purified to homogeneity from the NusA/hIL-3 fusion protein using an N-terminal histidine tag, factor Xa protease cleavage, and anion exchange chromatography. These results have a broad significance to the biotechnology field since insoluble heterologous protein expression in E. coli continues to be a pervasive problem in both industrial and academic research. Also, these results represent a successful example of how statistical modeling can be used to design a novel protein expression system for a specific cell type.