For the study of proteins that exist as complexes and act together for certain biological functions, it is often crucial to co-express different protein species. It is also a common practice to co-express the target proteins together with molecular chaperones or their interacting proteins, in order to increase the functional fractions of the recombinant proteins which are otherwise prone to form aggregates. For the co-expression of different proteins in vivo, the target genes should be cloned in a single multi-cistronic construct or in separate plasmids. Both of these cloning procedures pose significant technical challenges. First, the construction of a polycistronic expression vector is complicated, since multiple cycles each consisting of numerous cloning steps are required. The design of a polycistronic vector becomes more complicated as the number of ORFs increases. In addition to the difficulties involved in cloning, the translation of a downstream ORF could be affected by that of the upstream ones, which makes it difficult to control the expression level of the individual proteins. On the other hand, for the co-expression of multiple genes cloned in separate vectors, the mutual compatibility of the replication origins and selection markers should be considered, in order to prevent the segregation of the plasmids. Moreover, even in the circumstances where all of these requirements are fulfilled, it is not unusual to observe the biased production of one species when heterologous proteins are expressed together. As a result, attempts to develop an efficient strategy for the co-expression of multiple genes have met with limited success, basically because the expression of proteins relies upon living cells. In contrast, with the goal of making protein synthesis independent of cell growth, cell-free protein synthesis could be a straightforward alternative for the co-expression of different proteins. Furthermore, compared to the conventional in vivo methods, the cell-free synthesis system offers excellent throughput of protein expression. Initially, we expected that the co-expression of recombinant proteins would be able to be readily achieved by programming the reaction mixture with the corresponding plasmids. Contrary to our expectation, however, in certain cases of our experiments to co-express different pairs of DNA, only one species of the added DNA pairs effectively directed the protein expression. On the other hand, certain pairs of DNAs equivalently directed the expression of the corresponding proteins. In the course of reviewing the experimental set-ups, we recognized that the “compatible” pairs of DNAs commonly carried identical nucleotide sequences at their 5'-ends.

It is well known that the overall efficiency of protein synthesis is strongly affected by the nucleotide sequences downstream of the start codon, which are frequently referred to as the downstream boxes. Many of the recent reports on this subject indicate that the identity of the bases in the downstream boxes affects the initiation and/or early elongation phase of translation. Therefore, based on this observation, we assumed that the efficiency of co-expression during cell-free protein synthesis is determined by the relative efficiency of the downstream boxes. We therefore examined whether the introduction of common nucleotide sequences onto the 5'-ends of non-compatible ORFs would affect the bias in the protein expression in our cell-free protein synthesis system. Indeed, it was found that the bias in the protein synthesis was abolished, and both of the encoded proteins were successfully expressed when the template DNAs were designed to begin with identical nucleotide sequences. We assumed that the biased co-expression level within some pairs and not in others arose from either transcriptional or translational differences among the templates. Each template of the co-expression pairs which showed both biased and non-biased expression was transcribed in vitro and the protein expression levels of these templates were compared. There seemed to be no discrimination of the templates at the level of transcription and this result did not correlate with the protein expression level. Consequently, it was deduced that the ‘bias' of protein expression takes place at the level of translation. Moreover, as long as the template DNAs had the same initial sequences, the compatibility of the protein expression was maintained when the number of DNA species was increased, and we were able to express at least four different proteins. The relative amounts of the expressed protein were kept unchanged compared to those when they were expressed individually, indicating that the templates shared equivalent amounts of the translation machinery and that this machinery was evenly allocated to each of the templates. By introducing an N-terminal fusion partner encoded by an identical downstream box onto the templates of the co-expression pairs, it is possible to ensure that each template is translated without being interfered with. We expect that the presented results will contribute to providing a general platform for studying the structure and function of interacting proteins. We anticipate that the proposed approach will offer a powerful tool for the preparation of unbiased protein libraries, as well as for studying the structure and functions of interacting proteins.

Extended Abstract Status: Not Uploaded