Bacteriocin diversity, function, discovery and application as antimicrobials

1Antimicrobial peptides are produced ubiquitously across the kingdoms of life as an ancient form of host defence. Bacteriocins are a subset of bacterially produced, ribosomally synthesized antimicrobial peptides that have been studied for almost a century. In that time, bacteriocins have been the subjects of thousands of publications, describing diverse peptide families, many of which have reported the ability of the peptides and/or their producers to control the outgrowth of food spoilage organisms and to preserve foods. Many bacteriocins are the products of safe and food grade fermenting species such as lactic acid bacteria, and some, exemplified by nisin, have transitioned in the past century from unknown antimicrobial by-products to characterized, regulated and approved natural food additives in widely consumed commercial products. Bacteriocins have found success as food preservatives in the present but are not limited to it for the future, as they have also been investigated for various applications such as growth promotion in production animals, plant growth promotion and clinical therapeutics for the treatment of fungal, protozoal, viral and, foremostly, bacterial infections.

2The increasing rates of antimicrobial resistance threatens the use of current antibiotics, and it is predicted that antibiotic-resistant microorganisms will cause millions of deaths worldwide every year by 2050, severely affecting what are now considered routine medical treatments.

3The WHO has generated a priority pathogen list highlighting an urgent need for novel treatments against antibiotic-resistant pathogens, with special mention of Mycobacterium tuberculosis and carbapenem-resistant Gram-negative bacteria for which only a few treatment options are available, followed by the Gram-positive vancomycin-resistant Enterococcus faecium and other high-priority pathogens. Bacteriocins represent a potential solution as current antibiotic alternatives or adjuncts–. They offer several pharmacological advantages such as activity against pathogens in the nanomolar and picomolar ranges, narrow spectra, general non-toxicity and distinct mechanisms of action. The diversity in bacteriocin structures and mechanisms also represents a range of ribosomally produced antimicrobials that are amenable to genetic engineering strategies enabling the production of bioengineered variants with superior efficacy against target microorganisms.

4As natural agents of microbial competition, bacteriocins offer an advantage to bacteria over sensitive neighbouring species, enabling invasion, colonization and protection of a niche in complex environments such as the gut microbiome. By exploiting these natural mechanisms, they can be used as a tool for humans to deliver antimicrobial activity as pure peptides and composite nanoparticles or can be codified within live biotherapeutics for in situ expression in microbiomes. Advances in microbiome metagenomics and other ‘omics’ technologies over the past two decades have revealed catalogues of genome-mined and metabolome-mined bacteriocins, and new families are continually being discovered, but currently discovery is outpacing characterization and application. Narrow-target spectrum bacteriocins are also increasingly desirable as we advance our understanding of host– microbiota interactions (response and regulation of host processes to microbiota composition and products) and are realizing that the impact of antibiotics on microbiome diversity should be considered. Despite lack of investment, bacteriocins are currently undergoing clinical studies to treat infections caused by antimicrobial-resistant pathogens; the outlook is bright for diverse bacteriocin applications in the next century.

5In this Review, we summarize recent advances in bacteriocin research, including the identification of novel peptides and mechanisms of action. We discuss bioengineering as a means to optimize bacteriocin activity and expression in addition to novel methods of bacteriocin discovery in the omics era. Finally, we outline the potential applications of bacteriocins producers, including live biotherapeutics for microbiome modulation, and highlight those peptides that are currently under investigation for clinical applications.

6Bacteriocins are diverse and widespread across the phyla that produce them, with enormous heterogeneity in gene cluster organization, bio-synthetic machinery and peptide structures (Table 1 and Fig. 1). Despite such diversity, bacteriocins are usually classified into two major groups: those with post-translational modifications (class I) and unmodified peptides (class II) (Box 1). In recent years, there have been increased reports of novel ribosomally produced and post-translationally modified peptides (RiPPs) and characterization of their biosynthetic machinery (extensively reviewed elsewhere). However, not all RiPPs are bacteriocins and nor are all bacteriocins RiPPs as some do not undergo substantial post-translational modification. Bacteriocins are generally considered large molecule antimicrobials relative to small-molecule antibiotics but can range from bicyclic peptides, such as the recently described darobactin (965 Da) that is smaller than the clinically used colistin (~1,200 Da), to larger circular peptides such as pumilarin (7,083 Da) (Fig. 1).

7Class I bacteriocins undergo extensive modification with enzymes that may dehydrate residues and form inter-amino acid bonds, leading to their intricate and stable structures. Such modifications can result in peptides that are extensively cyclized and dehydrated, such as polyheterocyclic klebsazolicin and corynaridin, the latter of which contains eight dehydrated residues. Lantibiotics (lanthionine-containing antibiotics) are well-established class I bacteriocins, but novel structurally related peptide families continue to be identified; such as the type V lanthipeptides (lanthidins), which are glycosylated and N-terminally dimethylated, as evident in the founding member cacaoidin. Class II bacteriocin structures include simple linear α-helical peptides such as plantaricin EF, peptides with conserved complex secondary structures such as the four-helix peptide BacSp222 (ref. ) or linear peptides with specific conserved disulfides such as pediocin-like maltaricin CPN and actifensin,. Class II bacteriocins do not feature the substantial modifications, such as dehydration and formation of unusual amino acids or linkages of class I, and previously included circular bacteriocins that are subject to N-terminal to C-terminal covalent linkage resulting in their simple circular shape and the microcins that are post-translationally conjugated to the siderophore (iron binding) moiety. For consistency within the two major classification schemes (Box 1), we suggest that circular and siderophore-modified peptides are henceforth referred to as class I bacteriocins, as they undergo modification despite relatively simple mature structures.

8For most bacteriocins, the structural peptide-encoding gene also encodes a leader sequence, which comprises N-terminal residues removed from the mature peptide, and can fulfil various functions in peptide export, cleavage and modification. These roles are linked to the specific type of bacteriocin they accompany, shaping peptide production and post-translational alterations. For example, in the case of lantibiotics, conserved motifs such as the FNLD motif are observed within leader sequences, whereas circular bacteriocins exhibit considerable diversity in their leader sequences, ranging from minimal lengths as seen in lactocyclin Q (two residues) to more extensive sequences such as in amylocyclicin (47 residues). The flexibility of circular bacteriocin leader sequences was highlighted by modification of circularin A, which demonstrated functional expression with a leader of only one residue (Met), suggesting a limited role for the leader in active circularin A biosynthesis. Additionally, specific motifs within the nisin leader sequence have been shown to guide peptide export, modification and cleavage, whereas in lanthipeptides, such as mersacidin, the leader undergoes a two-step cleavage process involving initial cleavage upon exiting the cell by the transporter-peptidase MrsT and subsequent cleavage outside the cell by specific extracellular proteases such as AprE. Although many bacteriocins with leader sequences possess a dedicated leader sequence essential for their biosynthesis and activity, with exceptions such as circularin A described earlier, leaderless bacteriocins lack this and instead are directly translated as active peptides and the mechanisms of their production and export are poorly understood. Two pleckstrin homology domain-containing proteins DdE and DdF have been linked to the export of enterocin DD14 peptides in Enterococcus faecalis, but peptide export mechanisms are yet to be characterized.

9Bacteriocin-encoding gene clusters encode essential factors involved in their production and protection, typically incorporating genes encoding a core peptide, modifying enzymes for modified peptides, export machinery, self-immunity proteins that protect the producer and many clusters also contain genes encoding regulatory proteins. In the case of the unmodified bacteriocin bactofencin, a ~4 kb gene cluster encodes the peptide and an ABC transporter with a protease domain that exports and removes the leader sequence. These two genes alone (bfnA and DLSL_0052) have been demonstrated to produce active peptide when expressed heterologously in Escherichia coli. By contrast, the native producer of bactofencin A, Ligilactobacillus salivarius DPC6502, produces an immunity protein that provides self-protection and an additional protein containing a thioredoxin family domain that is hypothesized to assist in disulfide bond formation. Other small gene clusters result in more complicated peptide structures such as a cluster of 6.2 kb that contains genes for darobactin A production: a core peptide (darA), an ABC transporter (darBCD) and a single-radical S-adenosylmethionine enzyme (encoded by darE) that catalyse the formation of a Trp–Lys C–C bond and a C–O–C Trp–Trp ether bond via an unknown mechanism.

10Radical S-adenosylmethionine enzymes also catalyse the formation of α-carbon to thioether (sactionine) bonds in sactibiotics (sactionine-containing antibiotics), such as the ruminococcins, in which the gene cluster in Ruminococcus gnavus E1 consists of 17 open reading frames over 15 kb that can result in five distinct peptides. In the case of the lanthidin cacaoidin, the gene cluster spans 30 kb and contains 27 genes that encode the necessary enzymes for multiple modifications, including rhamnose synthesis and glycosyl transferases, dehydration and C-terminal S-[(Z)-2-aminovinyl-3-methyl]-d-cysteine (AviMeCys) formation. Many bacteriocin gene clusters are not constrained to specific phyla, such as glycocins and lasso peptides, which are encoded within the genomes of both Gram-positive and Gram-negative bacteria,. Bacteriocin gene clusters may be present on plasmids capable of conjugation or transposable elements, thus facilitating transfer of the bacteriocin production phenotype in a competitive environment. Recently, it was found that up to 3% of prophage encode biosynthetic gene clusters, the majority of which are for bacteriocin production and can demonstrably transfer production phenotype to nascent strains.

11Although this Review mainly focuses on the antibacterial nature of bacteriocins, they are known to have a diverse set of other biological functions. A wide range of peptides possess antiviral activities such as subtilosin from Bacillus amyloliquefaciens that displays antiviral activity against herpes simplex viruses. Peptides have been demonstrated to have cytotoxic effects towards cancer cell lines such as Laterosporulin10 (ref. ) or possess in vivo antitumour activity as was shown for microcin E492 against human colorectal cancer cells using a xenografted zebrafish model. Bacteriocins can modulate immune systems such as sublancin that can enhance phagocytic activity of macrophages and increase T cell frequency in mesenteric lymph nodes in mouse models of methicillin-resistant Staphylococcus aureus infection,. Thuricin-17, which is produced by Bacillus thuringiensis, has been shown to promote the growth of canola plants (Brassica napus L.) at high temperatures through an unknown mechanism, displaying a potential role for bacteriocins in supporting plant growth in changing climates. Finally, human streptococcal pathogens co-regulate complex bacteriocin production and competence pathways linking fratricide and natural transformation, implicating a role for bacteriocins in gene acquisition.

12Bacteriocins, such as antimicrobial peptides across all kingdoms of life, are typically cationic hydrophobic peptides, many of which are active at negatively charged bacterial cell surfaces–. The activities of bacteriocins, once they reach the cell surface, are diverse, including specific interactions with cell-wall or cell-membrane components, general membrane interaction and pore formation, or receptor-mediated cell penetration and interference with intracellular processes (for extensive mechanistic reviews, see refs. –) (Fig. 2).

13Nisin is well established to have a dual mechanism of action, binding the cell-wall precursor molecule lipid II, thereby inhibiting peptidoglycan biosynthesis and subsequent pore formation leading to cell death. Initially, it was believed to form a pore complex of eight nisin and four lipid II molecules, although more recent evidence has shown that the pore complexes continue to grow causing massive cell membrane damage (Fig. 2). Numerous post-translationally modified bacteriocins share a conserved nisin-like N-terminal lipid II binding domain (rings A and B), but do not possess the same C-terminal domain, and do not form pores as a result. This has been recently been demonstrated for the nisin-like peptides rombocin A and cesin. Many bacteriocins do not target lipid II and instead use a different receptor or lipid component as a binding site,.

14Searching the literature with the term bacteriocin returns more than 11,000 publications, describing a heterogenous group of compounds that mostly share two general features: bacterial ribosomally synthesized peptides/proteins that are antagonistic towards other bacteria. These compounds can range from multiprotein contractile nanomachines to antimicrobial secreted peptides. The lack of a precise definition after a century of bacteriocin research results in a field that can often be contradictory and confusing.

15Jacob et al. first coined ‘proteins of the colicin type’ as bacteriocins following the initial work of Gratia in 1925 describing colicins as the source of strain-specific antagonistic activity between Escherichia coli (see the figure). Subsequent publications used bacteriocin to describe lytic proteins from Gram-positive bacteria, such as megacin from Bacillus megaterium. The discovery of nisin was published in the same decade as colicin but Gram-positive bacteria-derived antimicrobials were not as widely described as bacteriocins for decades. In 1976, Tagg et al. acknowledged that no universally accepted definition existed and criticized the use of the term bacteriocin for poorly characterized compounds. It was proposed that bacteriocin should describe bactericidal compounds containing a biologically active protein moiety. The 1993 Klaenhammer scheme expanded this definition, which is the basis for modern bacteriocin classification. Peptides were grouped into class I (post-translationally modified) and class II (unmodified), with larger proteins grouped as class III (ref. ). Research on bacteriocins from Gram-positive bacteria grew considerably, and in 2005 Cotter et al. suggested that bacteriocin be reserved for peptides rather than heat-labile proteins. The term has not ceased to be used for these larger proteins. Other schemes have been proposed including up to four classes, and many bacteriocins are now popularly referred to as RiPPs (ribosomally produced and post-translationally modified peptides),,. We suggest consistent usage of the two major class schemes for bacteriocins to provide clarity for future researchers navigating the literature.

16Pediocin-like bacteriocins (class II) interact with the mannose phosphotransferase system (Man-PTS) forcing the channel into an open state, which was recently confirmed by cryo-electron microscopy. The Man-PTS transporter is a receptor for several class II bacteriocins, such as lactococcins A and B and garvicins A, B and C, as was recently confirmed by deletion mutants, protein modelling and cryo-electron microscopy,, and is the target for microcin E492 after ‘Trojan horse’ translocation across the outer membrane. The circular bacteriocin garvicin ML relies on the maltose ABC transporter to target Lactococcus lactis, whereas AS-48 is believed to interact based solely on electrostatic and hydrophobic interactions to insert directly into the lipid bilayer, leading to destabilization,. Direct membrane interaction is also the mechanism for the four helix-bundle bacteriocins epidermicin NI01 and aureocin A53, which have been found to have distinct multimodal ‘flowering’ pore-forming activity.

17Bacteriocin activities are typically restricted within the Gram status of the producer; that is, peptides derived from Gram-positive bacteria are typically far less active against Gram-negative bacteria without assistance of an outer-membrane permeabilizing agent,. Contradictory to this, a semipurified supernatant containing bacteriocins from Lacticaseibacillus paracasei CNCM I-5369 was found to be active against a panel of Gram-negative bacteria, which was confirmed by heterologous expression. Bioengineering and rational design have been utilized in the absence of an outer-membrane permeabilizing compound (which would typically be required for Gram-positive-derived bacteriocins to inhibit Gram-negative bacteria) to produce variants of nisin with up to 12-fold greater activity including bactericidal activity against E. coli and Acinetobacter baumanii. A hybrid peptide consisting of enterocin CRL35 fused with microcin V (Ent35–MccV) was shown to be active against the targets of both individual peptides, including Listeria monocytogenes and E. coli.

18Some bacteriocins interfere with essential intracellular processes, in particular those that target Gram-negative bacteria, which lack exposed peptidoglycan and inner cell membrane components. The TOMM klebsazolicin inhibits protein synthesis by blocking the ribosome exit tunnel, but its extracellular uptake mechanism is unknown. The crystal structures of the lasso peptides microcin J25 and capistruin have revealed their specific sites of transcription inhibition within the RNA polymerase secondary channel following uptake mediated by a host of transport proteins. Sactibiotics are a class known for their narrow spectra of inhibition. Recently, it was suggested that the R. gnavus- produced and Clostridioides- targeting sactibiotic ruminicoccin C could act through nucleic acid synthesis inhibition in a manner similar to metronidazole, but this has yet to be confirmed.

19Despite ongoing characterization of the mechanisms of action of bacteriocins, many peptides of interest remain uncharacterized. Similarly, the recently described bacteroidetocins, a family of anti-Bacteroidales pediocin-like bacteriocins, have an unknown mechanism of action as Bacteroides species lack the mannose-PTS. Although most research to date has focused on mechanisms of bacteriocins as free agents in liquid culture, a two-component Gram-negative bacteria-derived bacteriocin CdzCD was recently identified on the basis of cell-to-cell contact-dependent killing, for which the outer-membrane protein PerA was required for sensitivity, and has been suggested as the target receptor for CdzCD binding. With vast recent improvements in protein structural modelling and prediction, we may begin to see more in-depth bacteriocin–target complex structures and gain insights into yet unknown mechanisms.

20The risk of target strains developing resistance against bacteriocins because of continuous use is a key concern, given the current situation with antimicrobial resistance and the antimicrobial arms race. Bacteriocin producers are self-protected by the action of immunity proteins, which may bind or export the peptides, or block the action of peptides on protein complexes. Resistance to bacteriocin activity can occur through general mechanisms, such as cell-wall thickening causing resistance to the lipid II binding lasso peptide, siamycin I, or by truncation of a protein target similar to the transporter of the mannose-PTS but high-level resistance occurs through acquisition of specific resistance genes. Bacteriocins can be inactivated owing to the activity of specific enzymes such as the nisin resistance protein, which degrades nisin. Worryingly, some pathogenic species express generalized transporter proteins such as Streptococcus agalactiae bacitracin type efflux protein (SaNsrFP) that protects against numerous cell-wall targeting compounds including lantibiotics. As bacteriocins have not been extensively used in the clinical setting, the extent or risk for resistance in clinical use is unknown and to our knowledge for many classes of bacteriocins, no resistance mechanisms have been characterized yet.

21One of the most appealing attributes of bacteriocins is their gene-encoded nature, which enables the alteration of peptides using simple techniques and the generation of new-to-nature structures with unknown characteristics (Fig. 3). Wild-type bacteriocin producers often generate low bacteriocin yields because production is a high-energy process for the cell and is thus tightly controlled. This can result in excessive production costs that render bacteriocins non-viable commercially. Production optimization has therefore been examined by altering growth parameters in addition to exploiting their gene-encoded nature by manipulating the native producers and/or peptide expression in more efficient or peptide-insensitive heterologous hosts,.

22Bacteriocins of all classes have been subject to heterologous expression to express novel silent genes, understand their mechanism of biosynthesis, or to improve their production for commercial viability. Although post-translationally modified (class I) bacteriocins require essential modification enzymes to produce active mature peptides, class II bacteriocins are particularly suited to heterologous expression as many are ‘simple’ linear peptides. A range of Gram-positive bacteria, Gram-negative bacteria, eukaryotic hosts and diverse plasmids have been exploited for this such as E. coli for expression of pediocin PA-1 (syn. pediocin AcH), bactofencin A and lactocin AB,, L. lactis for expression of plantaricin JK and enterocin NKR-5-3B, and the yeasts Pichia pastoris (syn. Komagataella phaffii) and Kluyveromyces lactis for the expression of enterocins,.

23Beyond heterologous expression, simple optimization of culture conditions can increase bacteriocin yield, as demonstrated by researchers by improving garvicin KS production from Lactococcus garvieae as high as 60-fold when cultured in pasteurized milk and tryptone, compared with GM17 media. Combining the improved media with increased gene dosage of the core peptide of the native producer further increased production to the high concentration of 1.2 g of peptide per litre. Increasing gene copy number has also improved production of the sactibiotic thurincin H, in which an additional three copies of the structural gene, thnA, resulted in double the yield from broth culture. Synthesis of plantaricin E/F in Lactobacillus plantarum strains is induced by both an auto-inducing peptide PlnA1 and acetate via a distinct interaction with the PlnB1 response regulator. Exploiting this mechanistic understanding enabled an increase in the yield of plantaricin in culture by addition of sodium acetate at a concentration of 1 g l−1 (ref. ). Limited studies report extensive optimization or comparative purification yields, as peptide purification is typically a means of characterizing peptides at small scale in vitro rather than for potential commercial application. However, to scale up to in vivo experiments with pure peptide and for bacteriocin production to become more commercially viable, optimization must be explored in greater detail.

24Nisin is perhaps the most genetically manipulated bacteriocin with thousands of unique variants having been generated using random and site-direct PCR mutagenesis techniques, in addition to an entire expression system developed based on the tight regulation of nisin production. Mutagenesis methods have produced nisin variants with increased activity against biofilms and variants with greater activity against Gram-negative bacteria such as Cronobacter sakazakii, Salmonella typhimurium and E. coli. Rombocin K, a bioengineered variant of nisin-like rombocin A, was found to have a twofold lower minimum inhibitory concentration than the original peptide when tested against Listeria monocytogenes. The inhibition spectrum of nisin has also been refined in the derivatives M17Q, T2L and HTK, which possess increased anti-Staphylococcus activity and reduced anti-Lactococcus activity,. Similarly, saturation mutagenesis of mutacin 1140 generated 418 single amino acid variants of which Phe1Ile ultimately emerged as the most promising in vivo candidate for Clostridioides difficile infection and associated diarrhoea, which in turn underwent a second round of mutagenesis optimization resulting in a more potent peptide,. Such methods may improve a promising bacteriocin for application or be used to overcome a limitation, such as the oxidation-resistant M31L variant of pediocin PA-1, in which oxidation-associated activity reduction challenges the wider application of wild-type peptide. Bioengineering can be used as a means to overcome proteolytic degradation, such as nisin S29P that is immune to the action of the nisin resistance serine protease and cesin R15G that has improved resistance to trypsin.

25Recently developed methods have utilized incorporation of non-canonical amino acids into the peptide backbone, which can then be conjugated to other compounds via ‘click chemistry’. Using the non-canonical and more-hydrophobic amino acid ethionine in the place of methionine, heterodimeric and fluorescently labelled nisin variants have been generated, each of which retained antimicrobial activity. A similar method lipidated nisin variants, which had improved bio-activity and specificity against pathogenic strains. Conjugation of bacteriocin peptides to molecules such as fatty acids can be exploited to produce bacteriocins that form colloidal stable assemblies, a feature which could be utilized to overcome problems with solubility of hydrophobic peptides. Novel methods such as this can be used to further expand, edit and hone already diverse peptide functionality and future downstream applications.

26In recent years, bacteriocin discovery has been catapulted into the meta-omics era with the availability of vast genomic data sets and dedicated bioinformatic software enabling in silico screening of bacterial genomes and large-scale metagenomic mining for the presence of novel bacteriocin gene clusters (Fig. 3). Despite convincing evidence that bacteriocin-encoding genes are widely distributed across many genera, confirming bacteriocin production from individual strains isolated during screening studies has proven difficult. Before the increased accessibility of next-generation sequencing technologies, the majority of novel bacteriocins were identified in vitro through laborious screening of hundreds or thousands of bacterial isolates. These culture-dependent screening methods face logistical microbiological limitations such as the choice of media on which samples are plated, the choice of indicator species and the growth conditions (pH, atmosphere composition and temperature), which can all lower the likelihood of bacteriocin detection. Failure to detect bacteriocin production does not mean lack of bacteriocin presence, but could be due to the inadvertent choice of an insensitive indicator species or a limited or lack of induction and production of an active bacteriocin. Genomic approaches that are agnostic of active expression and rely on bioinformatic predictions of the presence of biosynthetic gene clusters can overcome these issues, but are not without their own limitations. For example, in silico approaches alone do not confirm biological activity nor do they enable the purification and characterization of a bacteriocin. It is likely that a combination of in silico and culture-dependent methods will prove most effective in terms of identifying and isolating novel functional bacteriocins.

27As sequencing technology has improved and sequence databases have grown, a range of tools have been developed for mining bacteriocin genes encoded in bacterial DNA. Tools have been developed that span from ‘point-and-click’ interfaces suitable for non-bioinformaticians to command-line controlled packages to detect biosynthetic gene clusters in genomic and metagenomic data sets at a range of assembly levels. BAGEL and antiSMASH are well-tested and frequently updated tools that are designed for the identification of bacteriocin (and secondary metabolite) biosynthetic gene clusters in assembled bacterial genomes,. Both tools are useful for the characterization of an unknown antimicrobial-producing isolate but are limited by predictions based on enzyme homology and predicted core peptide characteristics, which may miss truly novel compounds. In addition, they can be low to medium throughput based on the skill level of the user. Despite this, they have enabled the discovery of bacteriocins in several hundreds to thousands of genomes, such as the identification of 101 lanthipeptide gene clusters from 223 Bacillus cereus strains, the characterization of the biotechnological potential of 213 diverse lactic acid bacteria and the genome-guided expansion of a bacteriocin family from an initial 11 members to widespread detection of homologues in hundreds of strains. More recently, a simple tool for mining specific RiPP classes based on BAGEL has been developed to interrogate metagenomic data sets for bacteriocin biosynthetic gene clusters.

28With continuous growth in computing power and bioinformatic tool development, artificial intelligence has also been introduced into bacteriocin research, with machine learning being utilized to predict biosynthetic gene clusters and predict chemical structures following post-translational modification. RODEO (Rapid ORF Description and Evaluation Online) was the first tool available to use machine learning and was able to expand the number of known lasso peptides 10-fold upon its initial publication and subsequently increased the number of known glycocins from 6 to 10 distinct characterized members using only a single enzyme as a starting point,.

29Bioinformatics can have an important role in identifying potential bacteriocins, but this should be supported by functional characterization by isolation of a producing strain, chemical synthesis or heterologous expression from genomic or synthetic DNA in a suitable host. When candidate bacteriocin-producing strains are identified in silico but production cannot be demonstrated, or when putative producers are not available, E. coli has been used as a heterologous bacteriocin expression ‘workhorse’. Expression in E. coli strains has enabled the characterization of structural peptides and biosynthetic machinery for glycocins and lantibiotics,,. Bioinformatically mined class II bacteriocins have also been expressed in E. coli, including pediocin PA-1 and lactobacilli- derived pediocin-like peptides, enabling the identification of genes essential for bacteriocin production,. Gram-positive hosts have also been extensively utilized for the expression of bacteriocins, including the reliable nisin-inducible expression system in L. lactis, which has recently been used for the functional characterization of the circular bacteriocin circularin A mined from a clostridial genome. Genome mining paired with heterologous expression is a powerful combination to explore novel compounds but can be low throughput and failure-prone depending on the choice of the host and vector. Researchers recently developed a powerful platform combining high-throughput bioinformatics and an integrated robotic system to express 96 RiPP gene clusters from diverse bacterial phyla involving 383 bio-synthetic genes, with a success rate of 86% (ref. ). High-throughput methods such as this could be used to explore the Integrated Microbial Genome Atlas of biosynthetic Gene Clusters (IMG-ABC) database that consists of 20,000 gene clusters, of which a small number has been experimentally validated.

30Whole metabolomic methods are of growing interest in bacteriocin and RiPPs research (for an extensive recent review, see ref. ). Peptidogenomics, which is mass spectrometry-guided genome mining for peptides, have successfully identified a novel lanthipeptide and production gene cluster from the cell extract and genome of a single strain. MetaMiner has been developed to search for novel compounds from billions of mass spectra and compare with diverse metagenomic data sets such as human microbiome samples. ‘Omic-mining’ can successfully identify novel peptides and biosynthetic gene clusters in an untargeted manner, but without ‘wet-laboratory’ demonstration of compound-specific antimicrobial activity, all such compounds should not be classed as bacteriocins. As the number of known biosynthetic gene clusters continues to grow, it is inevitable that biological characterization will lag behind and will have to be upscaled to explore applications of novel peptides.

31Fermentation as a means of food preservation, and bacteriocins unknowingly as by-products of thereof, has been utilized by humanity for millennia. Much of early bacteriocin research was performed in the context of food-fermenting organisms, such as the description of nisin in the context of delayed dairy fermentation owing to producer inhibition, and continues to be studied to this day,. Vast amounts of historical and ongoing research continue to describe the efficacy of bacteriocin producers (as fermentation starters or adjuncts) and their products to inhibit spoilage organisms in different products. However, despite decades of bacteriocin use in the food industry (nisin was designated as generally recognized as safe by the FDA in 1988 (ref. )) and the array of bacteriocins that have been identified to date, only a small number are currently commercially available as food preservatives (mainly nisin A, nisin Z and pediocin PA-1). Most bacteriocin-preserving agents are available as partially purified fermentates, some of which also contain the producing organism. Nisin is available as a powder with degrees of purity under the product names Nisaplin® (Danisco, Dupont), Chrisin® (Chr. Hansen) and vegetal or white NisinA/Z® and NisinA/Z® P (Handary). Pediocin PA-1 is used as an antilisterial additive in food production in a number of powdered fermentate forms such as ALTA 2351/2341 (Kerry Group) and Fargo 23® (formerly in production by now defunct Quest International) approved by EFSA’s Qualified Presumption of Safety status. A formulation containing Carnobacterium maltaromaticum CB1 and derived bacteriocins labelled Micocin® (Griffith Foods) has been approved for use in North America and Canada for the preservation of meats.

32Microbiome research continues to grow exponentially with considerable efforts directed towards associating particular microorganisms or microbial compositions with human diseases or conditions–. Identifying such correlations will naturally lead to attempts to develop tools for modulating the microbiome to enhance specific health-associated taxa or to reduce disease-associated taxa. Nutrition and lifestyle are obvious microbiome modulating strategies but these are essentially ‘blunt tools’, generating largely nonspecific and potentially widespread changes to gut microbiota composition and functionality. Antibiotics are also blunt tools, and despite being necessary for treatment and prevention of infectious agents, their lack of precision causes collateral damage to beneficial gut microorganisms, risking adverse health events such as C. difficile over-growth or the expansion of antibiotic-resistant strains,,. This effect is not restricted to the gut, with antibiotic treatment also restructuring microbiomes of the skin, vagina and oral cavity–. Bacteriocins may prove to be an exciting alternative as potent antimicrobials, which may have narrow target spectra to target specific genera and species, and can therefore be utilized to more precisely inhibit pathogens or pathobionts, leaving benign commensals unharmed. Several class II bacteriocins have been found not to alter the overall gut microbiota composition but to reduce specific taxa such as Enterococcaceae and Clostridium spp. when consumed by mice. Thuricin CD, produced by B. thuringiensis, has also been shown to specifically reduce clostridia in a faecal fermentation model, whereas vancomycin, a commonly used treatment in clinical settings, has a broad impact on the bacterial community. Even broad-spectrum bacteriocins can induce only subtle changes in microbiome compositions when compared with traditional antibiotics, as was demonstrated with microcin J25 relative to rifampicin in a continuous culture porcine model. Targeted microbiome modulation should also consider potential ecosystem effects as removal of a targeted species or genus could lead to unintended increases or decreases in others. Simplified model ecosystems such as the OMM12 model have shown that bacteriocin production can have broad effects on neighbouring strains, and removal of a single species, in this case, a bacteriocin-producing strain, leads to the increased abundance of numerous others.

33Therapeutics for human or animal consumption must be carefully formulated to be delivered in an active form to the correct anatomical site for function. For example, a small-molecule antibiotic must survive harsh environments during gastrointestinal transit to be absorbed and function systemically to prevent or treat infection. Bacteriocins have long been thought to be proteolytically degraded during gastrointestinal transit and therefore must be protected by encapsulation for applications directed at the large intestine. Recent evidence has challenged this dogma and shows that some bacteriocins, such as nisin, may be more resistant than previously thought, with detectable intact nisin surviving in vivo gastrointestinal transit in pigs and microcin J25 resisting degradation during simulated human digestion,. Despite this, many bacteriocins can be readily degraded by proteases in gastric transit,, and peptide delivery to the gut faces challenges of producing sufficient quantities to administer to large animals and the degree of purity required for in vivo studies. Fortunately, bacteriocins can also be deployed in the form of a probiotic or live-biotherapeutic agent that arrives at the desired destination and delivers a bacteriocin dose in situ. Researchers recently showed that genetically modified E. coli Nissle 1917 expressing microcin l47 in a preclinical mouse model significantly reduced carbapenem-resistant Klebsiella pneumoniae after 1 week oral administration of 108 cells per day without alteration of the resident microbiota. E. coli expressing the unmodified enterocins A/B and hiracin JM79 have also been shown to reduce vancomycin-resistant Enterococcus when provided in drinking water ad libitum (5 × 108 cfu ml−1) in a vancomycin-resistant Enterococcus-colonized mouse model. Live bacteriocin producers can also be used to treat or modulate microbiomes outside the gut, such as the use of an L. lactis strain producing lacticin 3147 in a liquid paraffin-based emulsion that was used to treat bovine mastitis (109 cfu 5 ml−1 dose) as successfully as the standard antibiotic treatment.

34Bacteriocin producers that are closely related to the intended target species can have the added benefit of potentially being able to occupy the same niche, potentially providing future colonization resistance against pathogens. Bacteriocin-21-producing E. faecalis were found to evict their drug-resistant relatives in the mouse intestine when provided in drinking water ad libitum (5 × 108 cfu ml−1) without perturbation of the overall community. However, the plasmid-encoded nature of this bacteriocin resulted in conjugation to native enterococci providing immunity and the capacity to produce the bacteriocin. These data facilitate a scenario in which a multidrug-resistant strain could gain another tool to aid colonization and therefore must be cautiously considered when developing bacteriocin producers as targeted live biotherapeutics.

35Other methods for ensuring unscathed bacteriocin delivery to site of action include through formulation with other compounds that increase peptide stability under biologically relevant conditions. Silica mesoporous matrices have been utilized as protective delivery vehicles for bactofencin A and nisin A, resulting in active peptides that are resistant to trypsin and pepsin degradation, respectively. Hydrogels have been developed to ensure sustained release of bacteriocins over time, including an injectable polysaccharide gel containing nisin A active for up to 10 days, and a two-phase release gel of subtilosin developed for inhibition of Gardnerella vaginalis, which causes bacterial vaginosis. Formulation of bacteriocins into nanoparticles to increase the antimicrobial activity or widen the spectrum of the resulting nanoconjugate is a growing trend for food preservation and clinical application (reviewed elsewhere). Recently, a mix of five bacteriocins adsorbed on alginate nanoparticles was found to be 500-fold more active than the peptides against clinically relevant E. coli, and microcin J25-chitosan nanoparticles were active against enterotoxigenic E. coli in addition to ameliorating lipopolysaccharide-induced cytotoxicity of macrophages and downregulating lipopolysaccharide-stimulated pro-inflammatory cytokines. Emerging technologies such as these should continue to improve and develop bacteriocin candidates for application in food preservation and in clinical settings.

36In addition to few commercial bacteriocin products on the market, none is currently approved for human therapeutic use. Research and development for clinical application of bacteriocins has been chronically underfunded, with 70 novel antimicrobials under development in 2020, of which none were bacteriocins. Despite this lack of considerable investment, many preclinical and some clinical studies have been performed demonstrating that bacteriocins, and particularly lantibiotics, exert antimicrobial activity in vivo in tolerable quantities (Table 2). Several bacteriocins and their producers have been investigated for the oral treatment of C. difficile infection and associated diarrhoea. Preclinical trials have assessed OG-716 (mutacin derivative) for the treatment of C. difficile infection and associated diarrhoea when administered via oral gavage (20 mg kg−1 bodyweight) in golden hamsters, finding the peptide to be safe and effective at preventing relapse,. A bacteriocin-producing Bacillus velezensis strain as a live biotherapeutic for the treatment of C. difficile infection has been proven effective preclinically in mice (single oral dose of 5 × 108 cfu) and in miniature swine (single oral dose of 5 × 108 cfu and 28 days administration of 5 × 109 cfu in capsules) and recently concluded a phase I clinical trial. The semi-synthetic thiopeptide LFF571 (derived from GE2270A) has undergone phase I and II clinical trials (oral administration of 200 mg four times a day for 10 days) for C. difficile infections and was determined to be comparable to vancomycin treatment,.

37Beyond C. difficile infection, the lantibiotic NAI-107 (microbisporicin) has undergone a series of preclinical rodent studies for severe antimicrobial-resistant infections with promising results. In animal studies, nisin and lacticin-3147 have shown promise in the treatment of mastitis, and lantibiotic-producing Staphylococcus hominis ShA9 decreased S. aureus, improving atopic dermatitis when delivered to the arms in a lotion (106 cfu cm−2 applied twice daily for 1 week) and meeting primary and secondary endpoints in a phase I clinical trial. Mutacin 1140-derived topical treatments (5 or 10 mg kg−1 per day for 7 days) have been proven safe and more effective than vancomycin for methicillin-resistant S. aureus infection in mice and have also been effective when administered intravenously (10 mg kg−1 per day for 6 days). Fewer trials have been performed with class II bacteriocins, but recently the effect of an Abp-118-producing L. salivarius live biotherapeutic on recurrent acute otitis media (middle ear infection) in children was examined. The treatment was proven highly effective (84% reduction in episodes over 6 months of daily consumption of 109 cfu) and as one of the most common infections in children, a commercial bacteriocin therapy could greatly reduce antibiotic prescription (Table 2).

38Thiostrepton, derived from Streptomyces spp., has long been used in veterinary medicine for the topical treatment of acute otitis media, but has never been approved for human use nor have any other thiopeptides. However, it is one of several bacteriocins gaining interest beyond antimicrobial application. Thiostrepton selectively and potently represses the cancer-associated transcriptional activator and therapeutic target, FoxM1 (ref. ). Diverse bacteriocins have been studied in vitro and in vivo for chemotherapeutic treatment of different cancer types including nisin, microcin E492 (ref. ) and enterocin LNS18 (ref. ), with none having yet progressed to phase I clinical trials. Moli1901 (duramycin/lancuvotide) has undergone phase I and II clinical trials as a therapeutic for cystic fibrosis as a host chloride channel activator, although with limited efficacy,. Bacteriocins are not immune to the pitfalls of drug development and can be abandoned in early clinical trials owing to the poor performance or difficulties in production. Ultimately, the pharmacokinetic and pharmacodynamic attributes of bacteriocins under investigation are critical limiting factors in enabling them to become therapeutic agents. Vastly increased funding of discovery, characterization, preclinical and human trials is required to adequately advance suitable bacteriocin candidates to the frontline of clinics, which currently continues to be sorely lacking.

39Bacteriocins are a diverse group of antimicrobial peptides produced across a number of bacterial phyla, and ongoing discovery efforts continue to reveal biological insights. These discoveries include not only novel peptide families but also instances of structurally similar peptides and biosynthetic machinery present in taxonomically unrelated species. This phenomenon may shed light on aspects of horizontal gene transfer and the evolution of antimicrobial peptides, akin to the convergent evolution observed in Eukaryotic defensins.

40Bacteriocin discovery has been transformed in recent years from the labour-intensive process of identifying producer isolates in vitro to the in silico genome mining that can yield hundreds of related peptide sequences within seconds. Machine learning has become an invaluable tool in detecting bacteriocin sequences within vast genomic data sets, and with continuous advancements in computing power and artificial intelligence training, we can expect the identification of even more novel peptides and perhaps even novel bacteriocin classes in the future. Despite the growing number of known peptides, our understanding of their mechanisms of action lags behind. Investigating the cellular targets of bacteriocins is crucial for choosing the correct antimicrobial for specific pathogenic species and for understanding resistance mechanisms. This will also enable the rational design of bacteriocins through peptide bioengineering, potentially leading to the development of highly targeted and potent therapeutic agents.

41Bacteriocins hold great promise in complex microbial communities such as the gut microbiome with their ability to selectively target specific pathogens or pathobionts without disrupting the overall community structure and function. However, the potential unintended consequences of such interventions must be carefully considered to avoid disruptions to microbial ecosystems that could lead to undesirable health effects. One substantial challenge is translating laboratory discoveries into clinical applications. Bacteriocins offer a viable alternative or adjunct therapy in the treatment of infections, especially against multidrug-resistant pathogens. Realizing these opportunities will require substantial funding and rigorous clinical studies. In addition to their antimicrobial properties, structurally related peptides have been identified that lack antimicrobial activity but possess other valuable bioactivities, such as pain relief or anti-inflammatory effects. These findings expand the horizon of bacteriocins beyond antimicrobials, conceivably offering multipurpose therapeutic agents.

42In conclusion, bacteriocins have the potential to act as precise and versatile tools in an ongoing battle against antimicrobial-resistant microorganisms. As they continue to be explored, a path is paved whereby these remarkable molecules could have a pivotal role in safe-guarding human health and well-being. Although challenges lie ahead, the potential for bacteriocins as tools to eliminate clinically relevant microorganisms and precisely shape microbial communities is an exciting and promising avenue for research.