Introduction

Antimicrobial resistance (AMR) among bacteria causing infections in humans is increasing. Fewer treatment options are available for infections caused by resistant bacteria, leading to increased morbidity and mortality. An estimated 1.27 million deaths were directly attributable to bacterial antimicrobial resistance (AMR) in 2019 [1]. A review on AMR published in 2016 estimated that an additional 10 million deaths will be caused by AMR in 2050 if current trends continue [2]. Furthermore, the rise of AMR is hastened by viral outbreaks such as the COVID-19 pandemic; from 2019 to 2020, an increase in infections caused by resistant bacteria such as carbapenem-resistant Acinetobacter (78% increase), vancomycin-resistant Enterococcus (14%), and methicillin-resistant Staphylococcus aureus (13%) was reported [3]. The antibiotic resistance crisis is believed to be exacerbated by excessive and inappropriate use of antibiotics in human medicine and agriculture [4]. For this reason, alternative antimicrobials are sorely needed. Peptides and proteins with antimicrobial activity are produced by virtually all organisms as part of their innate immune system. Bacteria also produce antimicrobial peptides and proteins, known as bacteriocins, to inhibit each other during competition for common nutrients or niches [5,6,7,8,9]. Bacteriocins are characterized by a narrow spectrum of activity, often only inhibiting strains closely related to the producer. Furthermore, they often exhibit very potent activity in the pico- to nanomolar range towards their target strains and are thus in some cases considerably more potent than antibiotics [10]. Indeed, bacteriocins have received increasing attention as an alternative or supplement to antibiotics [11, 12].

Bacteriocins are very diverse, differing in sizes, structures, modes of action, molecular targets, and spectrum of activity. Currently, small bacteriocins (< 10 kDa) are typically classified as class I if they are post-translationally modified and class II if they are unmodified [13]. The peptides are further subdivided in both classes based on similarities in biosynthesis, structure, or sequence. Class I is the most diverse group and contains at least 12 subclasses, each based on their characteristic post-translational modification. Examples include the lantibiotics which contain lanthionine; glycocins, which are glycosylated; and circular bacteriocins, where the N- and C-termini are linked. The class II bacteriocins are generally only categorized into four subclasses (IIa–d): IIa, pediocin-like; IIb, two-peptide; IIc, leaderless; and IId, linear non-pediocin-like peptides [13]. Most bacteriocins in both class I and class II are synthesized as precursor peptides with a leader sequence that is removed during or following export to yield the active bacteriocin [13, 14]. A notable exception is the class IIc leaderless bacteriocins, which are unmodified peptides synthesized in the cell in their active form [15].

To date, class IIc includes over 20 bacteriocins which are either single-, two-, or multi-peptide bacteriocins. Among the single-peptide leaderless bacteriocins are the AurA53-like, LsbB-like, and EntL50-like groups of peptides [15, 16]. The LsbB-like bacteriocins depends on the presence of a zinc metalloprotease (RseP/Eep/YvjB) for its antimicrobial activity and kills target cells via a specific interaction with the metalloprotease [17, 18]. In contrast, all non-LsbB leaderless bacteriocins is generally believed to act directly on the bacterial membrane leading to perturbation and permeation without requiring any specific protein on the cell surface as a receptor [15]. A common feature of the non-LsbB leaderless bacteriocins is that they all seem to have a saposin-like fold [19]. In fact, it has been suggested to group the leaderless bacteriocins in two classes based on this difference in structure, namely the saposin-like and the LsbB-like [20]. In addition to the leaderless saposin-like bacteriocins, many circular bacteriocins from class I have the similar saposin-like fold, such as enterocin AS-48 (AS-48) [19].

Saposins are four (A–D) small (~9 kDa) heat-stable proteins with an important function in sphingolipid metabolism in vertebrates [21, 22]. The four saposins are produced in the lysosome or late endosome by proteolytic processing of a single 70-kDa precursor protein prosaposin [20]. Although the four saposins share relatively little sequence identity (20–40%), all four share a very similar compact globular fold consisting of four amphipathic α-helices [23,24,25]. Acidification of saposins promotes a change in conformation to an open lipid-binding form, either as a dimer or higher oligomer [23, 26,27,28,29,30,31]. The open conformation exposes a highly hydrophobic core capable of binding the alkyl chain of certain lipids, thus facilitating lipid extraction and solubilization from the membrane [32, 33]. In saposin C, the protonation of several acidic residues is thought to trigger the conformational change to a lipid-binding state [34, 35]. Despite the structural and functional similarities between the four saposins, they differ in their lipid specificities [36]. It is not known how the “bi-functional” property of the saposin-fold in both lipid recognition and conformational switching is achieved. The bacteriocin enterocin AS-48 binds to lipids and is similarly thought to dimerize in a pH-dependent manner [37].

As the arsenal of known bacteriocins is growing, discovering new bacteriocin peptides becomes increasingly difficult. The process of sampling, screening, purifying, and identifying bacteriocins is laborious and time-consuming, and often results with the identification of already described bacteriocins. The discovery of new suitable bacteriocins is arguably one of the bottlenecks in developing these peptides for biomedical applications. Recent advances in synthetic DNA combined with in vitro protein expression enable the direct synthesis of active bacteriocins [38]. In this work, we use a hybrid approach to show that existing bacteriocin families constitute a rich source of new synthetic antimicrobial peptides.

Results

Design of Hybrid Bacteriocins

The saposin-like bacteriocins are remarkably similar in structure, but differ in their activity and spectrum, indicating differences in their mechanism despite the structural similarity [39]. A mechanism has been suggested for the circular bacteriocin AS-48 which has a similar saposin-like fold, where protonation of acidic residues is thought to occur at the membrane interface that facilitates membrane insertion [37]. Interestingly, many non-LsbB leaderless bacteriocins also contain one or more acidic residues, primarily located at the C-terminal half of the peptides. As such, we hypothesized that these bacteriocins have a bi-functional property where the C-terminal half is involved in binding to specific lipids unique to sensitive cells while the N-terminal half inserts into the membrane. Based on this assumption, we constructed a library of genes encoding peptide sequences designated ISP1 (ISP, in vitro synthesized peptide) through ISP49 containing all combinations of the N- and C-terminal halves of seven leaderless bacteriocins; lactolisterin BU (LliBU; ISP1), mutacin BHT-B (BHT-B; ISP8), aureocin A53 (AurA53; ISP15), K411 (ISP22), lacticin Q (LacQ; ISP29), epidermicin NI01 (EpiNI01; ISP36), and salivaricin C (SalC; ISP43) (Fig. 1; Online Resource Supplementary Material Table S1).

Fig. 1
figure 1

Structures of the leaderless bacteriocins (from left to right): LliBU, BHT-B, AurA53 (PDB ID: 2N8O), K411, LacQ (PDB ID: 2N8P), EpiNI01 (PDB ID: 6SIG), and SalC. Structures with a PDB identifier (ID) have been solved experimentally while the remaining structures were predicted by AlphaFold2. Basic amino acids (K, L, H) are colored blue, acidic residues (D, E) are shown in red, and hydrophobic residues (F, I, L, M, V, W, A, P) are colored in gray. The N-terminal methionine is indicated in yellow

Full size image

The resulting library (Table 1) contains 49 peptide sequences, of which 42 are novel hybrid peptides. Despite exchanging the N- and C-terminal halves between the bacteriocins, structure predictions of the hybrids suggest that the saposin-like fold is not considerably affected (Online Resource Supplementary Fig. S1). The genes encoding the seven peptide sequences from which the hybrids were derived were included as a comparison and control.

Table 1 Overview of the peptide sequences encoded by the library (ISP; in vitro synthesized peptide). The isoelectric point (pI) and hydrophobicity were calculated using the Peptides (v2.4.4) library for R v4.1.1 [40]
Full size table

In Vitro Synthesis and Antimicrobial Activity Screening of Hybrid Bacteriocins

Genes encoding all the peptides in the library, including the native peptides, were used as a template for in vitro protein expression using PURExpress® In Vitro Protein Synthesis Kit (New England Biolabs). The resulting products from the in vitro synthesis were directly assayed for antimicrobial activity. Using a spot-on-lawn assay, each reaction was tested for antimicrobial activity against a panel of eight indicator bacteria: Enterococcus faecium, Enterococcus faecalis, Listeria monocytogenes, Staphylococcus aureus, Streptococcus dysgalactiae, Staphylococcus haemolyticus, Lactococcus lactis subsp. lactis, and Escherichia coli (Fig. 2).

Fig. 2
figure 2

Spot-on-lawn assay assessing antimicrobial activity of the seven leaderless bacteriocins (first column) and hybrid peptides synthesized in vitro. Reaction mixtures of each peptide (ISP1-ISP49, corresponding to Table 1) were spotted inside the corresponding squares (5 µl). Indicators used were A Lac. lactis subsp. lactis IL1403, B Ent. faecium LMGT 3104, C L. monocytogenes LMGT 2653, D Ent. faecalis LMGT 2333, E Staph. aureus ATCC 14458, F Strep. dysgalactiae LMGT 3890, G Staph. haemolyticus LMGT 4071, and H E. coli DH5α

Full size image

Of the seven native bacteriocins tested, six of them displayed activity against at least one indicator. From the hybrid peptides tested, 11 (11/42; 26%) of them showed activity against at least one indicator (Table 2). The hybrids ISP4, ISP5, ISP6, ISP20, and ISP26 produced large zones of inhibition against several of the pathogenic strains. Particularly active was ISP26 which inhibited all strains except for S. haemolyticus and E. coli. ISP4 displayed activity against Lac. lactis, E. faecium, L. monocytogenes, and S. aureus, but was inactive against E. faecalis. Similarly, ISP5 was inactive against E. faecium but displayed good activity against the other strains, except for S. haemolyticus and E. coli.

Table 2 Overview of bacteriocins and hybrid peptides exhibiting inhibition of at least one of the indicators. Clear zone of inhibition, +; diffuse zone of inhibition, (+); no inhibition, -
Full size table

It was also interesting to note that some of the hybrid peptides appeared to display altered inhibition spectra compared to their native counterparts. AurA53 and EpiNI01 showed only weak activity against L. monocytogenes, while AurA53-EpiNI01 (ISP20) showed good activity (clear zone of inhibition). Similarly, LliBU and K411 had no or negligible effect against Strep. dysgalactiae; an inhibition zone was observed for the LliBU-K411. In fact, four of the hybrids with the N-terminal half from LliBU (4/7; ~ 60%) showed good activity towards several indicators, while only one hybrid with the C-terminal half from LliBU (ISP37) showed weak activity.

To further assess the antimicrobial activity of these hybrid bacteriocins, we sought to establish a bacterial production scheme that would allow us to obtain larger quantities of the hybrid peptides. However, obtaining active peptides by heterologous expression was not successful (see Online Resource Supplementary Material). The most promising hybrid, ISP26 (K411-LacQ), was therefore obtained synthetically which allowed us to confirm its antimicrobial activity and to assay its activity against a larger panel of bacterial species (Table 3).

Table 3 Antimicrobial activity of synthetic ISP26 against a larger panel of bacterial species. Using a spot-on-lawn assay, 5 µl of ISP26 was spotted on plates overlaid with the indicator. Pure DMSO was also spotted in equal volume as a negative control and showed no clear zones of inhibition against any of the indicators. A clear zone of inhibition at 0.01 mg/ml, +++; 0.1 mg/ml, ++; 1 mg/ml, +; no inhibition, -
Full size table

Using a spot-on-lawn antimicrobial assay, the synthetically obtained ISP26 was shown to have broad-spectrum activity against most Gram-positive species tested, including Bacillus, Lactobacillus, Lactococcus, Listeria, Pediococcus, and Streptococcus. The exception is staphylococci, where sensitivity towards ISP26 varied considerably between species and strains. For example, while activity is observed against Staph. haemolyticus and Staph. hominis, Staph. epidermidis appears insensitive to ISP26. Furthermore, among the S. aureus strains tested initially, S. aureus LMGT 3022 showed a clear zone of inhibition at 0.1 mg/ml while S. aureus LMGT 3242 was not inhibited even at 1 mg/ml of the bacteriocin. To investigate if the variation in sensitivity was peculiar to the tested strains, ISP26 was further assayed for activity against a panel of 134 S. aureus strains from our laboratory collection (Online Resource Table S3).

Of all tested strains, 77 (58%) were insensitive to ISP26. Interestingly, while most of the Gram-negative species tested were insensitive to ISP26, Acinetobacter baumannii was shown to be sensitive to ISP26. To further explore the activity of ISP26 against Acinetobacter sp., another panel of indicators belonging to this genus was assayed for sensitivity (Table 4; Online Resource Supplemental Material Fig. S1).

Table 4 Antimicrobial activity of ISP26 against some Acinetobacter species. Clear zone of inhibition at 0.01 mg/ml, +++; 0.1 mg/ml, ++; 1 mg/ml, +; no inhibition, -
Full size table

All species of Acinetobacter that was tested showed sensitivity towards bacteriocin ISP26 at 1 mg/ml and four species were also inhibited at 0.1 mg/ml (Ac. ursingii, Ac. towneri, Ac. lwoffii, Ac. gen sp. 9), clearly demonstrating the antimicrobial effect of ISP26 against Acinetobacter.

Discussion

The construction of the hybrid bacteriocins presented in this work revealed active new-to-nature bacteriocins with increased inhibition spectrum and an apparent increase in potency compared to their native counterparts. The constructed hybrid bacteriocins were shown to inhibit the growth of WHO priority pathogens, including Ac. baumannii, Staph. aureus and Ent. faecium, and other important pathogens such as Ent. faecalis and L. monocytogenes. As such, these hybrid peptides may serve as an important addition or supplement to antibiotics for the treatment of infections from these pathogens in the future.

Notably, the hybrid peptide ISP26 displayed broad spectrum antimicrobial activity, including also Gram-negative Ac. baumannii and other Acinetobacter species. Antimicrobial activities of bacteriocins derived from Gram-positive bacteria active against Gram-negative species are quite rare, and only a few examples of bacteriocins with activity against Ac. baumannii have been reported [41, 42]. Multidrug- or even pandrug-resistant Ac. baumannii are emerging pathogens associated with a range of infections, and there is a pressing need to find new therapeutics [43, 44]. It should therefore be further explored whether saposin-like bacteriocins could be used alone or in combination with other antimicrobials to fight Ac. baumannii infections.

To design the library of genes encoding hybrid bacteriocins in this work, we hypothesized that the saposin-like bacteriocins are “bifunctional,” where the N- and C-terminal halves of the peptides serve different functions. This idea is derived from a proposed mechanism of action for the circular bacteriocin AS-48, and structure and sequence similarities shared between this bacteriocin and other saposin-like bacteriocin [19, 37, 39]. In the proposed mechanism for AS-48, protonation of four glutamic acid side chains is believed to occur in the acidic environment of the membrane interface, resulting in the transition to a membrane-interacting dimeric form. The protonated glutamic acids of AS-48 recognize and associate with the phosphate moiety of a phospholipid, which was demonstrated by crystallography [37]. It is not known whether the non-LsbB leaderless bacteriocins act as single peptides or form dimers or multimers, however, many saposin-like peptides (SAPLIPs) are known to form dimers in their active form [31, 45, 46]. In fact, the transition to the dimeric form for many SAPLIPs is believed to occur upon interacting with the membrane [24, 25, 31, 47]. If a similar mechanism is also employed by the saposin-like bacteriocins, their activity and inhibition spectrum may depend on the presence of, or abundance of, certain phospholipids that will vary between species. However, the strain-to-strain differences in sensitivity of S. aureus against ISP26 may suggest the involvement of specific cell surface features (Online Resource Table S3). Although our hybrid approach successfully resulted in saposin-like hybrid peptides with improved activities, we could not assign any obvious or distinct role of the N- and C-terminal halves of the peptides. Understanding the mechanism and differences in activity between the hybrids will thus need more investigation.

Our data shows that in vitro protein synthesis is a well-suited tool to screen for active, new-to-nature antimicrobials. Nevertheless, there are numerous challenges with in vitro synthesis of bacteriocin-like peptides, which practically limit this method to small-scale screening. For example, we have experienced that antimicrobial activity from reactions is lost upon storage/incubation past 24 h, presumably due to aggregation and/or precipitation [48]. We have also not been successful at purifying the peptides from the reaction mixture. Furthermore, in vitro synthesis of bacteriocins sometimes fails, which was also evident in our study. For example, the native bacteriocin BHT-B was included as a control in the screen and known to be active towards Lac. lactis [49]. However, no inhibition zone was produced by the in vitro expression of BHT-B in our screen, indicating low- or no yield from the reaction. It is therefore likely that more of the hybrid peptides with no reported activity in the screen (Fig. 1) may be false negatives due to the failed synthesis of some peptides. The variation in synthesis efficiency also precludes the direct comparison of potency between the peptides. Our results suggest that several of the hybrid peptides have higher potency than the peptides from which they were derived (e.g., ISP4-6 has larger zone than ISP1 towards Lac. lactis, see Fig. 1); however, we cannot exclude that this is just a result of different synthesis efficiency in the in vitro reactions.

Due to the length (50–53 aa) and high hydrophobicity of the peptides, commercial peptide synthesis can be difficult and costly. For this reason, we sought to produce the hybrid bacteriocins heterologously in E. coli, the only species insensitive to all peptides tested. However, all our attempts at expressing and purifying these hybrid peptides in E. coli were unsuccessful (see Online Resource Supplementary Material). The expression of ISP26 and ISP29 in E. coli as C-terminal fusions to maltose-binding protein (MBP) also failed. We could not determine why expression failed in E. coli, although clones expressing MBP-ISP26/29 exhibited severely attenuated growth. It could be speculated that these peptides are active when delivered in the cytoplasm of E. coli (e.g., inner membrane perturbation/permeabilization), and that the lack of activity in our screen is due to the outer membrane barrier. A failure to express a similar protein fusion in E. coli was recently reported by Malesevic et al.; in this work, the authors tried to express a fusion of LliBU (ISP1) to MBP [50]. LliBU is a peptide of similar physicochemical properties as the hybrid peptides. It has also been documented that E. coli has a great ability to degrade peptides, such as bacteriocins, when fused to proteins [51,52,53].

Furthermore, since the native producers of all bacteriocins in the library are Gram-positive species, nisin-inducible expression of the hybrid peptide in the Gram-positive Lac. lactis was attempted (see Online Resource Supplementary Material). While we were able to heterologously express and purify MBP-ISP26 in Lac. lactis NZ9000 (Online Resource Supplemental Material Fig. S2), the release of functional (antimicrobial) ISP26 from the fusion protein using the incorporated TEV cleavage tag was unsuccessful with the set of conditions tested here. Thus, alternative methods and design of the constructs to express and purify the peptide may be needed to overcome this problem in the future. This could, for example, include so-called sandwich approaches, where both termini of the peptide are fused to innocuous proteins that can be excised without scars (e.g., SUMO and intein proteins) [51, 52].

In this work, we show that previously characterized leaderless bacteriocins can serve as scaffolds for the construction of new-to-nature antimicrobials with improved properties. Additionally, screening new sequences for antimicrobial activity can provide invaluable insight into antimicrobial determinants. Very little is known about the mechanism of most bacteriocins and the factors that determine their potency and spectrum. A better understanding of the features shared among active peptides versus inactive ones, and those active towards certain species but not others, can allow us to rationally design new peptides targeting high priority pathogens. However, more research is needed to fully characterize these peptides and to assess their therapeutic potential as well as to find cost-effective strategies for their production.

Materials and Methods

Strains and Growth Conditions

All strains used in this study are listed in Table 5. Lac. lactis was grown at 30 °C in M17 broth (Oxoid) supplemented with 0.4% glucose (GM17), and E. coli was grown in LB at 37 °C (180 rpm). All remaining strains were grown in BHI (brain heart infusion; Oxoid) at 37 °C without shaking.

Table 5 Indicator strains used for assessing activity and inhibition spectrum of in vitro synthesized peptides
Full size table

In Vitro Protein Expression and Antimicrobial Assay

Bacteriocin peptide sequences were reverse translated and codon was optimized for E. coli K12 using GENEius (Eurofins Genomics, Germany). All genes were synthesized by Pepmic Co. Ltd (Suzhou, China) and supplied in pET-3a. Plasmids were solubilized to 250 ng/µl in Milli-Q water and used directly as templates for in vitro protein synthesis using PURExpress® In Vitro Protein Synthesis Kit (New England Biolabs). Reactions of 50 µl using 500 ng of template per reaction were assembled according to the protocol provided by the manufacturer in a 96-well plate. The 96-well plate was sealed using heat-sealing film and incubated at 30 °C for 4 h with vigorous shaking at 1200 rpm using a microplate shaker (PMS-1000i, Grant-Bio, Grant Instruments Ltd., Shepreth, UK). All reactions were immediately assayed for antimicrobial activity using a spot-on-lawn assay. Briefly, an overnight culture of the indicator strain was diluted 50-fold in growth medium (see above) containing 0.8% agar and poured over an agar plate (10 × 10 cm, square). After solidification, 5 µl of each reaction mixture was spotted onto the plate and allowed to dry. All plates were incubated at 30 °C overnight for the appearance of inhibition zones.

The peptide ISP26 was obtained from Pepmic Co. Ltd (Suzhou, China) at > 95% purity and solubilized in pure molecular biology grade dimethyl sulfoxide (DMSO; Sigma-Aldrich, D8418) to a stock concentration of 1 mg/ml (170 µM); all dilutions were prepared in DMSO. Sensitivity towards ISP26 was determined using a spot-on-lawn assay as described above, with the exception of Acinetobacter which was assayed according to the EUCAST disk diffusion methodology (Online Resource Supplemental Material Fig. S2). For Acinetobacter species, colonies were suspended in saline (0.9% NaCl) until McFarland 0.5 and spread on a Mueller–Hinton agar plate using a sterile cotton swab. ISP26 was spotted (5 µl) on the agar plates as indicated. An equal volume of pure DMSO was spotted in all assays as a negative control.