The limits of conventional antibiotics in treating pathogenic microbes and the increasing prevalence of antibiotic-resistant pathogens have led to the exploration of viable alternatives, including antimicrobial peptides. Only a limited number of antibiotics are available for clinical use, and they have similar activity spectrum and action mode [44, 45]. AMPs have distinct advantages over antibiotics such as remarkable structural and functional diversity, and immunomodulatory activity. Some AMPs show a broad range of action, which can be effectively used to treat multi-microbial infections including both Gram-positive and Gram-negative bacteria [45, 46].
Currently, no antibiotic peptides are available for clinical use. However, a number of AMPs are under clinical trials and development, and their applications are not limited to directly killing pathogens: e.g. pexiganan (for the treatment of bacterial infections, to diabetic foot ulcers), omiganan (catheter infections and rosacea), hLF-11 (bacterial and fungal infections in immunocompromised stem cell transplantation), novexatin (fungal infections), CZEN-002 (vaginal candidiasis), LL-37 (wound healing), PXL01 (prevention of post-surgical adhesion formulation), Iseganan (oral mucositis), and PAC-113 (oral candidiasis) .
Despite their potential, the development and availability of AMPs for clinical use is met with several challenges. Primarily, the high production cost and low scalability of chemical peptide synthesis prevent the widespread development and adoption of AMPs as a viable clinical treatment . Biological production using recombinant prokaryotic systems is a viable alternative to chemical synthesis, but issues such as toxicity to host cells, degradation of the product by protease, and low yield must be addressed .
In this study, a SUMO tagging system was used to prevent toxicity of the expressed antimicrobial peptide to host cells. The commonly used fusion carriers such as thioredoxin (12 kDa) and glutathione-S-transferase (GST, 26 kDa) have several advantages associated with increased solubility, promotion of proper folding and prevention of toxicity of AMPs [9, 46, 47]. However, GST increases the relative molecular weight ratio of carrier proteins to the peptides, which leads to low AMP yields. Also, several GST fused-AMPs expressed in E.coli showed proteolytic activity, resulting in inefficient or failed AMP productions. Thioredoxin is more favorably used for peptide production than the GST due to its small size, allowing high peptide yield attributed to the high peptide-to-carrier ratio . However, the proteases used for the release of attached peptides from the carriers are more expensive and more sensitive to pH and chaotropes when compared to SUMO protease . Moreover, AMPs tagged to aggregation-promoting carriers such as PurF fragment, PaP3:30 and ketosteroid isomerase have shown toxicity to host bacterial cells [9, 48].
The SUMO tag has many similar advantages to the thioredoxin or/and GST systems including high yield resulting from the high ratio of peptide-to-tagging protein, enhanced solubility, and no toxicity to host cells, but it also has unique advantages over other tags. Sumoase can recognize the tertiary structure of SUMO and cleave it from substrates with no attachment of unwanted amino acids to peptides [24,25,26]. The precise cleavage of target proteins from SUMO fusions by sumoase is confirmed by comparing the molecular mass and N-terminal amino acid sequence of the released peptides to the corresponding synthetic peptides using matrix-assisted laser desorption/ionization time-of-flight mass spectrometry (MALDI-TOF) and Edman degradation. More detailed information about the current tagging and expression systems for the production of antimicrobial peptides in E.coli is included in Additional file 2: Table S1.
In this study, we confirmed the proper cleavage of abaecin from SUMO using MALDI-TOF, but there was 5-aa-long C-terminal deletion from the abaecin. In contrast to our expectation, the attached SUMO tag (overall charge − 5) didn’t protect the abaecin (overall charge + 4, Fig. 3b) via electrostatic interaction from endogenous proteases in an E.coli system. Small size peptides with high cationic content are highly susceptible to proteolytic attack in E.coli as shown in Piers et al.’s study . In the study, direct expression of small size cationic peptide such as human neutrophil peptide 1 (HNP-1, a kind of defensin) wasn’t successful, while its transcript was detected. Even when the HNP-1 and other cationic peptide, synthetic cecropin/melittin hybrid (CEME), were fused to GST, the fusion proteins were proteolyzed. However, the proteolysis wasn’t observed when anionic pre-pro defensin sequence was inserted between the GST and the cationic peptides. Based on our analysis, a positively charged patch created by two Arg (R) at 12th and 13th positions (Fig. 3b) seemed to not be recognized by endogenous proteases possibly due to charge neutralization effect by the electrostatic interaction with SUMO, while the second patch created by two Lys (K) at 28th and 30th positions (Fig. 3b) seemed subjected to proteolytic attack. Another possibility of the C-terminal deletion could be a consequence of excessive sonication, resulting in denaturation of the fusion protein and breakdown at the C-terminus. Also, it can’t be ruled out that the sumoase could aberrantly cleave the abaecin. It is thought that sumoase is highly specific and active in a broad range of conditions; it cleaves effectively in a wide range of pH (5.5–10.5) and temperatures (4–37 °C), and even in detergents such as 2 M urea or 20 mM DTT or β-mercaptoethanol . However, the cleavage study at the junction between SUMO and a partner protein showed that the cleavage didn’t work when the first amino acid of the partner protein was Pro (P) . Also, the introduction of a stretch of several Try (W) residues close to the cleavage site caused random cleavage within the fusion protein due to an inaccessibility of sumoase by steric hindrance, which, however, was relieved by addition of 1 M urea and led to the release of correct size peptide . Considering the high content of proline in abaecin, the repeated kinks generated by 10 prolines in the abaecin could create a steric hindrance when attached to SUMO, which would lead an unexpected random cleavage. To investigate these issues mentioned above, SUMO could be further modified in a way which introduces additional glycine resides at the cleavage junction to release steric hindrance. In our unpublished study, cleavage of SUMO-cecropin B at the junction by sumoase happened only after introduction of additional glycine residues. Also, genetically modified SUMO which has more negative charge could provide a better protection to cationic peptides.
Although the C-terminus of abaecin sustained a 5-aa deletion after sumoase cleavage, the 29-aa-long abaecin showed antimicrobial activity against B. subtilis in the functionality study, while the SUMO-abaecin fusion protein had no bacteriolytic activity as shown in Fig. 4. Furthermore, the bacteriolytic activities were consistent with previous reports in which the antimicrobial activity of abaecin alone was further enhanced with combinatory treatment with other pore-forming AMPs [31, 43]. Although abaecin has high binding efficacy to DnaK like other proline-rich DnaK-binding AMPs including metalnikowins, metchnikowins, onocin Onc72, apidaecin Api88, drosocin and pyrrhocoricin, the abaecin hasn’t a conserved motif, YL/IPRP . So we analyzed both abaecin amino acid sequences such as 34-aa and 29-aa-long abaecins for the functional binding sequence to DnaK using limbo server and found that the full length abaecin has two binding sequence sites: one is at N-terminus (1YVPLPNV7, score 1.7) and the other one is at C-terminus (25NPKIKWP31, score 2.8). So we assume that the reason why the 29-aa long abaecin still has antimicrobial activity (Fig. 4) is due to the presence of N-terminal binding sequence to DnaK (Fig. 3c).
The main purpose of this study was to introduce a new expression platform for the production of antimicrobial peptides in E.coli system in a way that does not harm the host cell. Abaecin was chosen as a reference AMP because it targets the intracellular molecule DnaK, prokaryotic heat shock protein 70 (Hsp70), and the binding of abaecin to DnaK can cause protein metabolism to be compromised in E. coli, resulting in host cell death . The 6xHisSUMO-abaecin fusion protein, however, did not exhibit toxicity to E. coli host cells. The transformed E. coli cells grew in liquid culture at the similar growth rate to untransformed ones (Fig. 1c). Furthermore, the purified fusion protein showed no anti-bactericidal activity against B. subtilis (Fig. 4a, b). The normal growth of the transformed E. coli and the lack of antimicrobial activity of the fusion protein to B. subtilis proved that the lethal activity of the AMP to the host cells can be properly shielded by SUMO but its activity can be restored by the release from the tag. The antimicrobial activity of AMPs is generally determined by the hydrophobicity and the net positive charge. As reported in the previous study, the electrostatic interactions between the positive charges (between + 4 and + 6) of AMPs and negatively charge residues within SUMO (overall charge − 5) seem to play a key role in neutralizing the bacteriolytic activity and protecting AMPs from degradation by endogenous proteases . In this regard, although the protection of abaecin from proteolytic degradation by SUMO failed, the C-terminal deleted abaecin still had antimicrobial activity. Furthermore, the toxicity of abaecin was, in some way, successfully prevented by the SUMO tag via the presumed electrostatic interaction.
The SUMO tagging system is also advantageous because it has no disulfide bond. Due to the reducing environment of the E. coli cytoplasm, proteins that require disulfide bonds fail to achieve their active forms and ultimately form inclusion bodies or are degraded by proteases. To achieve and maintain their active forms, such proteins need to be redirected to periplasmic space or secreted, which are energy-consuming and could potentially lower target protein yield. However, the SUMO tagging system does not require the additional steps because it only contains one cysteine in its entire 96 amino acid sequence.
Overall, the SUMO-abaecin fusion proteins were detected more in soluble fractions than insoluble ones (Fig. 2a, b) with no toxicity to the host cells. There have been many reports that SUMO fusions increased the solubility of difficult-to-express proteins in E.coli, such as GFP, metalloprotease (MMP13)  and severe acute respiratory syndrome coronavirus (SARS CoV) proteins including 3CL protease, nucleocapsid protein and spike C protein [49, 50]. However, in contrast to antimicrobial peptides, those proteins are not toxic to the host cells. To conclusively distinguish the effect of SUMO on the solubility of the fusion protein, the expression of abaecin alone would need to be compared with SUMO-abaecin fusion protein, which, however, was not examined in this study due to technical issues such as possible low stability and potential toxicity to host cells. As mentioned above, the direct expression of the small size and high cationic content of AMPs are highly susceptible to endogenous proteolysis . Furthermore, the expression of abaecin alone could be lethal to host cells because the peptide inhibits intracellular DnaK , which is a central organizer of the chaperone network in E.coli. The chaperone protein interacts with ~ 700 cytosolic proteins. Among them, ~ 180 proteins are relatively prone to be aggregation and rely extensively on DnaK during and after their initial folding .
The translation efficiency of codon-optimized heterologous sequences was evaluated in this study through partial or whole optimization in order to find the most accommodating nucleotide sequence in E.coli system. Codon usage was adjusted according to the codon usage preference of a gene, psbA, which is highly expressed in prokaryotic systems, and then three different combinations of codon-optimzed sequences were created. The codon adjustment was performed in a way to increase the compatibility between 5’ UTR of the psbA promoter and the 5′ coding region of the fusion gene, with an increase of AT content of the SUMO sequence to 63.5% from 59.5%. In contrast to our expectation, the codon optimization of N-terminal SUMO of the fusion protein failed to improve the expression level over the non-optimized counterpart. As seen in the Fig. 2a, b, the native sequence of SUMO performed better than its corresponding codon-optimized sequence for the expression of the fusion proteins. The expression level of 6xHisSUMO-abaecin (native – codon-optimized) was 2.8 or 3.5 times higher than that of 6xHisSUMO-abaecin (codon-optimized - native) or 6xHisSUMO-abaecin (codon-optimized – codon-optimized), respectively (Fig. 2b). This is likely due to the stability of the transcribed mRNA of the native SUMO sequence, which may be relatively higher than the stability of other codon-optimized sequences . Another possible explanation is that the compatibility of the 5’ UTR of the promoter with the 5′ coding sequence of the codon-optimized SUMO could be compromised, resulting in an unstable or inefficient translational initiation complex [53, 54]. It is generally thought that translational efficiency is influenced by the efficiency of the formation of translational initiation complex and thus the first ~ 30–50 codons are considered more important than the rest of the sequence [55, 56]. Therefore, the marginal difference of expression level between codon-optimized SUMO fused native and codon-optimized abaecins could be a conseqeunce of the inefficient formation of the translational initiation complex.
Recent studies have found that the reciprocal functional interaction of abaecin with pore-forming peptides occurred not only with the peptides co-expressed in the same species, but with ones from other species [13, 31, 43]. Pores created in the membrane by the pore-forming peptide allow abaecin to access its intracellular target, DnaK. The inhibited heat shock proteins compromise protein metabolism, and the damaged pores remain unrepaired, thus allowing abaecin even greater access to its target. Abaecin’s ability to increase membrane permeabilization consequently reduces the minimal inhibitory concentrations of both abaecin and other pore-forming peptides in a reciprocal manner [31, 43]. Likewise, the activity of the pore-forming peptide cecropin B from Hyalophora cecropia was potentiated in a combinatorial treatment with abaecin (Fig. 4), showing that abaecin can be used with diverse pore-forming peptides to inhibit bacteria which are renitent to conventional antibiotics.
One of the challenges that the application of AMPs presents is the demand for effective and patient-friendly delivery system, particularly, for the patients with chronic diseases. Currently, most AMPs under clinical development are designed to target local infections using a topical formulation. Only a few AMPs are being developed for systemic delivery . Although an oral delivery system is most likely preferred due to straightforward administration, AMPs require special consideration because peptides are rapidly broken down in the gastrointestinal tracts due to the high concentration of proteases and high acidity. In this respect, edible plants can be used as a delivery platform, by which peptide based drugs bioencapsulated within the plant cells can be protected from the harsh environment of the gastrointestinal tract. But the drugs can be released into the intestine by the break-down of the cell walls by cellulolytic bacteria. The plant expression and delivery system can also eliminate the concern of endotoxin contamination, which causes fatal septic shock to recipients. Furthermore, the oral delivery of peptides by edible plant cells eliminates the need for expensive downstream purifications, reducing product cost and benefiting patients [18, 20, 21]. In our future study, we will evaluate the efficacy of the plant expression system as well.