Escherichia coli has been the most widely used host to produce large amounts of heterologous proteins. However， given an input plasmid DNA， E. coli may produce soluble protein， produce only inclusion bodies， or yield little or no protein at all. Many efforts have been made to surmount these problems， but most of them have involved time-consuming and labor-intensive trial-and-error. We hypothesized that different metabolomic fingerprints might be associated with different protein production outcomes. If so， then it might be possible to change the expression pattern by manipulating the metabolite environment. As a first step in testing this hypothesis， we probed a subset of the intracellular metabolites by partially labeling it with 13C-glucose. We tested 71 genes and identified 17 metabolites by employing the two-dimensional NMR spectroscopy. The statistical analysis showed that there existed the metabolite compositions favoring protein production. We hope that this work would help devise a systematic and predictive approach to the recombinant protein production.