FIGURE 3 Monitoring protein production in CFPS by the split GFP assay. (a) Schematic representation of the workflow for monitoring protein production in CFPS by the split GFP assay. (b) Producing six NMN pathway enzymes by CFPS and monitoring their expression levels by the split GFP assay. BC: Blank CFPS reaction; NW: Nanopure water. Statistical significance was examined using a two-tailed T-test analysis. ***:P < 0.001. Error bars represent the standard deviation of three biological replicates. (c) Sensitivity of the split GFP assay. The correlation coefficients between complementation fluorescence value and the amount of HsRbk-LG were 0.9986, 0.9991, 0.9992, 0.9987, and 0.9977 for incubating for a period of 8 h, 10 h, 12 h, 14 h, and 16 h, respectively. Error bars represent the standard deviation of three biological replicates.
3.3 Prototyping enzyme homologs for NMN biosynthesis by using a normalized screening procedure
After designing a feasible biosynthetic pathway for NMN production and demonstrating the ability to monitor CFPS via split GFP assay, a normalized screening procedure, which incorporated these two technologies, was employed to rapidly prototype enzyme homologs (Figure 1b). Briefly, enzyme homologs were parallelly expressed by CFPS. For each homolog, its CFPS reaction mixture was divided into two parts: one part was used to catalyze the conversion of the desired substrate to product, and the other part was mixed with GFP 1-10 detector to regenerate the fluorescent GFP signal for monitoring its expression level. The ratio (RT/F) between the final titer of the product and the fluorescence value was calculated for each individual homolog. Then the normalized screening was performed by comparing the RT/F values based on a criterion, in which the homolog with the highest RT/F value was regarded as the most productive homolog.
To identify the putative enzyme homologs in NMN biosynthetic pathway, the amino acid sequences of HsNampt, MrNampt, HsPrs, PcPrs, HsRbk and EcRbk were used as query sequences to perform BLAST analysis against the Universal Protein Resource (UniProt) or National Center for Biotechnology Information database (NCBI) database. For a particular query sequence, some resulting fallacious sequences were discarded, and then the left sequences were used to generate a phylogenetic tree. Thus, six phylogenetic trees were obtained in total (Figure 4). Notably, since the prokaryotic NMN pathway enzymes were found more likely to be productive (Figure 3c), we had a bias to select more sequences from the phylogenetic trees that were constructed based on the enzymes from prokaryotes (i.e., EcRbk, PcPrs and MrNampt). As a result, two sequences from each phylogenetic tree constructed based on HsRbk (Figure 4a), HsPrs (Figure 4c), or HsNampt (Figure 4e), and six sequences from each phylogenetic tree generated from EcRbk (Figure 4b), PcPrs (Figure 4d), or MrNampt (Figure 4f) were selected. In summary, there were 10 homologs for each enzyme of the three-enzyme NMN pathway, which included two query homologs and eight homologs chosen from phylogenetic trees. The 30 enzyme sequences were all cloned into the pET28a-derived expression plasmid (Figure S1) for adding the linker and GFP11 tag.