In producing large amounts of heterologous proteins, researchers most often use Escherichia coli as a host thanks to its extensively studied genetics, simple growth procedure, and low cost. However, the desired protein is often produced only in a form of inclusion bodies. Researchers have tried to devise a way to circumvent such a problem, and the ones using fusion partners seem to be the most successful. Based on our previous observation that the host metabolome was related to the outcomes of protein expression patterns, we proceeded to perturb the metabolome by applying a salt stress to see if we could shake up metabolite compositions to make them better suited for soluble expression of the target protein. We examined a subset of the metabolites which had been partially labeled with input 13C‐glucose. We tested 11 genes under 4 different NaCl concentrations, and identified 18 metabolites using the heteronuclear single quantum coherence NMR experiment. Most of the proteins kept their expression profiles unchanged, but two proteins were converted from inclusion bodies to a soluble form with increased NaCl concentration. Through the statistical analysis, we could identify a region where the soluble protein production was favored in the metabolite space. We hope that this work would provide an alternative strategy to produce the recombinant proteins in their soluble or native forms, not only in E. coli but also in other hosts.

Protein-ligand interaction is one of the highlights of molecular recognition. The most popular application of this type of interaction is drug development which requires a high throughput screening of a ligand that binds to the target protein. Our goal was to find a binding ligand with a simple detection, and once this type of ligand was found, other methods could then be used to measure the detailed kinetic or thermodynamic parameters. We started with the idea that the ligand NMR signal would disappear if it was bound to the non-tumbling mass. In order to create the non-tumbling mass, we tried the aggregates of a target protein, which was fused to the elastin-like polypeptide. We chose the maltose binding proteinas a test case, and we tried it with several sugars, which included maltose, glucose, sucrose, lactose, galactose, maltotriose, and β-cyclodextrin. The maltose signal in the H-1 NMR spectrum disappeared completely as hoped around the protein to ligand ratio of 1:3 at 298 K where the proteins aggregated. The protein signals also disappeared upon aggregation except for the fast-moving part, which resulted in a cleaner background than the monomeric form. Since we only needed to look for a disappearing signal amongst those from the mixture, it should be useful in high throughput screening. Other types of sugars except for the maltotriose and β-cyclodextrin, which are siblings of the maltose, did not seem to bind at all. We believe that our system would be especially more effective when dealing with a smaller target protein, so both the protein and the bound ligand would lose their signals only when the aggregates formed. We hope that our proposed method would contribute to accelerating the development of the potent drug candidates by simultaneously identifying several binders directly from a mixture.