E. coli, the classical workhorse organism of biotechnology, is a mature and versatile host for the production of various proteins. Due to the high obtainable expression levels and the ease of genetic manipulation, E. coli was chosen as the first secretion vector.

For E. coli, we chose pET28a with its intrinsic T7 promoter and RBS as the vector backbone.

Basic Construction

Applying PCR and the Golden Gate Assembly method, we first cloned the proteins of interest into pET28a. Subsequently, we used the resulting expression vectors to transform E. coli TOP10, which is an ideal cloning strain. Following sequence confirmation and vector expansion in TOP10, the vectors were transferred into E. coli BL21 for expression, which promised satisfactory secretion levels.

Improved Construction

Although E. coli is desirable for achieving maximal protein production within the cytoplasm, targeting the proteins to extracellular compartments may be difficult in this host, since E. coli has an outer membrane.

To avoid this obstacle and increase the secretion efficiency, we adopted a gene named kil. Kil is a bacteriocin release protein (BRP) which leads to a permeabilization of the outer membrane and could thus be used for the secretion of proteins of interest into the culture medium¹.

However, high-level induction of kil results in cell lysis and death. To solve this problem, we selected a medium-strength promoter, J23117, from the iGEM PARTS library and fused it to the 5’ end of the kil coding sequence. We also included a lac operator between J23117 and kil, which resulted in a simple inducible synthetic kil expression gene. Consequently, kil was only expressed in the presence of IPTG, which ensured the survival of the bacteria under normal conditions.

B. subtilis is a Gram-positive bacterium and as such does not have an outer membrane, which enables it to secrete proteins directly into the medium. As a result, B. subtilis may offer higher product yields for secreted proteins, which makes it an attractive alternative to E. coli.

Basic Construction

We chose the shuttle plasmid pBES as vector backbone. PBES could be manipulated easily and multiplies quickly in E. coli, so we conducted all molecular cloning procedures in E. coli. Once successfully constructed, the plasmids were extracted and used to transform B. subtilis for subsequent expression.

Improved Construction

To increase the expression and secretion levels, we examined two promoters - PaprE and P43. These two promoters are widely used for heterologous protein production in B. subtilis². By fusing either PaprE or P43 to the coding sequence, we hoped that B. subtilis would achieve a high secretion efficiency.

The target proteins included 3A-SUP (triple SpyTag-SUP), 3A-mSA (triple SpyTag-mSA), 3A-RFP (triple SpyTag-SUP), 3B (triple SpyCatcher) and 2B (double SpyCatcher). Since a number of important attributes of these fusion proteins remained uncharacterized, and especially their exact folding state in the cytoplasm, we were unable to rationally choose or design a signal peptide in each individual case. We were, however, able to build a signal peptide library to simply screen for the best candidates. The ultimate goal was to select the most appropriate ones from a large number of available signal peptides. However, due to time constraints, we only chose a comparatively limited number of signal peptides to test for their potential to mediate the secretion of the target proteins.

Based on previous studies on their secretion performance, 4 different signal peptides, derived from LTIIb, PhoA, PelB and OmpA, were chosen for E. coli and 7 signal peptides, derived from Epr, Bpr, LipA, YjfA, SacB, NprE and ImdA, were chosen for B. subtilis³.

To test the overall system, we first chose OmpA and ImdA, from E. coli and B. subtilis respectively, to determine whether well-established signal peptides could guide the target proteins out of the cytoplasm. This was done in order to exclude that emergent features of the complex protein folding of our multi-module monomers somehow generally inhibited secretion. Applying the Golden Gate Assembly method, we cloned these two signal peptides into the pSB1C3 vector backbone, amplified the resulting construct by PCR, and constructed the corresponding expression vectors by ligating the signal peptides along with the other expression cassette elements into either pET28a or pBES.

After we had succeeded in detecting the target proteins in culture supernatants, we kept on constructing vectors with other signal peptides. To accomplish this task, we first amplified the backbones with the signal peptides we had already constructed, and subsequently combined them with other signal peptides (Fig. 3.).

Each of the proteins of interest has its specific biochemical properties, such as SUP’s affinity to uranyl ions or SpyTag’s irreversible binding to SpyCatcher. Unfortunately, none of these intrinsic properties could be used to directly assay for protein secretion. Although SDS-PAGE could be used as a universal method to separate proteins with different molecular weights, it is not specific enough to verify the results of secretion experiments. Thus, in order to accurately and efficiently assay for protein secretion, we modified the proteins of interest with two effective molecular labels — a His-tag and a FlAsH-tag.

His-tag

The His-tag is an artificial amino acid motif that contains at least six histidine residues and is widely used for recombinant protein purification because of its affinity for resin-immobilized bivalent nickel or cobalt ions³.

FlAsH-tag

A FlAsH-tag is a tetracysteine-motif-tag (-FLNCCPGCCMEP-) which binds with high affinity and specificity to the biarsenical dye FlAsH-EDT2 and thereby forms a fluorescent complex⁴.

We added the His-tag sequence between the signal peptide and the coding sequence of the protein of interest, so that it would be exposed at the N-terminus of the secreted protein after the signal peptide was removed during the Sec-pathway secretion process. A His-tag-reactive antibody was commercially available, enabling us to conduct further assays using western-blot analysis.

We additionally fused the FlAsH-tag to the C-terminus of the protein of interest. Because the outer membrane of E. coli is naturally impermeable to FlAsH-EDT2⁴, undesired off-target labeling of intracellular proteins is avoided, rendering the assay highly selective for secreted proteins.

A schematic illustrating all the components used in the library design is shown in Fig. 6.

We expressed the recombinant proteins in E. coli strain BL21. The bacteria were cultured in M9 minimum medium for 14 hours. We subsequently separated the cytoplasmic and periplasmic components from the culture supernatants and examined them for the presence of recombinant proteins using western blot analysis. An overview of the results is shown in Table. 1.

The results show that all four signal peptides possess the ability to mediate secretion, but the efficiency varies. For some combinations, such as LTIIb-mSA, the secreted portion was larger than the portion remaining in the cytoplasm, while for others, such as PelB-3B, the opposite was true. This shows that the interaction between signal peptides and the proteins of interest significantly affects the secretion efficiency, as expected. Moreover, the total expression levels of the recombinant proteins are also different from one signal peptide to another. For example, proteins with the signal derived from OmpA on their N-terminus consistently showed higher expression levels than their respective combinations with any of the other three signal peptides. The reason for this may be due to the interaction between the RBS from the vector and the coding sequences of the signal peptides. It’s worth mentioning that cytoplasmic precursors of the proteins always have bigger molecular weights than their respective secretion products in the culture supernatants. This is because the signal peptides were removed during the secretion process.

It should be noted that overall efficiency decreased considerably when the proteins of interest were co-expressed with the Kil protein, most likely because Kil directly reduced the viability of the host cells. Therefore, the strength of the promoter controlling kil expression must be finely tuned to optimize the production and release of secretory proteins. Additional experiments are needed to supplement these findings.

We expressed the combined fusion proteins in B. subtilis strain RiK1285, which lacks the proteases AprE and NprE, that might otherwise degrade secreted proteins of interest. We first examined the secretion efficiency of constructs containing either 3A-SUP or 2B, together with one of the signal peptides derived from YjfA, NprE, LipA or SacB. The bacteria were cultured in M9 medium at 37°C for 14 hours before examination.

Unfortunately, none of our target proteins were detectable in the culture supernatants from M9 medium. We subsequently changed the culture medium to 2xYT and prolonged the culture time to 48 hours. After SDS-PAGE analysis, the protein band patterns of strains carrying 3A-SUP and 2B constructs were similar, so we concluded that the recombinant proteins were not expressed at all. Further western blot analysis confirmed this assumption. We concluded that there might be a cryptic problem related to the general B. subtilis plasmid design, and we consequently decided to stop the examination of other constructs for B. subtilis.

We instead attempted to at least express the recombinant proteins in the cytoplasm, and exchanged the PaprE promoter for P43, which is more efficient [2]. Following the same procedures as described above, we still could not detect protein expression through either SDS-PAGE or western blot analysis (data not shown).

There are two possible reasons accounting for this failure. Firstly, the proteins might have failed to fold properly, rendering them insoluble. Secondly, as mentioned above, the proteins may also have not been expressed at all. Since the two promoters we used are known to be strong promoters in this host according to earlier studies, the problem may be due to RBS design or an as of yet unknown problem with our B. subtilis strains and protocols.

We intended to quantitatively measure the secretion of recombinant proteins using the FlAsH-tetracysteine assay. The FlAsH-EDT2 molecule only shows strong fluorescence when bound to a tetracysteine motif, so that the intensity of fluorescence in the samples could be expected to be directly proportional to the amount of recombinant protein bearing a tetracysteine-tag.

Therefore, we purified 3A-SUP tagged with the tetracysteine motif, and directly used it to calibrate the fluorescence intensity of the FlAsH-EDT2 probe based on target protein concentration. The fluorescence intensity did change with the magnitude of protein concentration, as expected. However, we did not find a good linear relationship between the target protein concentration and probe fluorescence (Table. 2.). Hence, only qualitative results could be obtained using this assay.