Using display technologies to identify macrocyclic peptide antibiotics

Abstract

Antibiotic resistant bacterial infections are now a leading cause of global mortality. While drug resistance continues to spread, the clinical antibiotic pipeline has become bare. This discord has focused attention on developing new strategies for antimicrobial discovery. Natural macrocyclic peptide-based products have provided novel antibiotics and antibiotic scaffolds targeting several essential bacterial cell envelope processes, but discovery of such natural products remains a slow and inefficient process. Synthetic strategies employing peptide display technologies can quickly screen large libraries of macrocyclic sequences for specific target binding and general antibacterial potential providing alternative approaches for new antibiotic discovery. Here we review cell envelope processes that can be targeted with macrocyclic peptide therapeutics, outline important macrocyclic peptide display technologies, and discuss future strategies for both library design and screening.

Introduction

The continued development and spread of antibiotic resistance have left medical professionals with few options to treat multidrug resistant bacterial pathogens [1]. A recent study reports antibiotic resistant bacterial infections are now the leading cause of death globally with the highest burdens placed on locations with few available resources [2]. These problems are exacerbated by the continued stagnation of novel antibiotic discovery. Since its initial boom in the mid to late 20th century, discovery of natural small molecule antibiotics has rapidly declined, and rediscovery of known classes has become the norm. This has left the field scrambling in search of new strategies and methods to discover and develop novel antibiotics to help revitalize the pipeline and maintain pace with the development and spread of resistance.

Macrocyclic peptides are a promising class of therapeutic. They can access and bind targets deemed “undruggable” by small molecules and their cyclized structure can improve issues of peptide instability and toxicity (for review see [3,4]). Interest in macrocyclic peptide therapeutics continues to grow year over year and many macrocyclic peptide-based drugs have been FDA approved and brought to market over the last two decades [3,4].

Specifically, natural macrocyclic peptides have provided novel antibiotic drugs and have also served as scaffolds for the development of synthetic antibiotics in clinical development (for review see [5,6]). Many of these peptides target essential proteins or processes in the bacterial cell envelope to inhibit bacterial growth. Though natural macrocyclic peptide antibiotics show therapeutic promise, they also suffer from issues of limited discovery like those encountered by traditional small molecule antibiotics.

Macrocyclic peptide display technologies offer the ability to rapidly screen both small and large synthetic libraries of cyclic peptides to identify and or optimize binding to a specific target of interest or to directly screen peptide libraries for general antibacterial activity. Here we discuss some potential gram-negative cell envelope targets, list examples of known peptide inhibitors, and describe techniques for using macrocyclic peptide display technologies to identify synthetic antibiotic leads which could target these processes. We also discuss some of the challenges to using these approaches and how future library design and screening strategies may help overcome these hurdles.

1. Gram-negative cell envelope targets for peptide therapeutics

The bacterial cell envelope is the first level of defense for the bacterium against environmental threats. In gram-negative bacteria, the cell envelope is a multi-layer protective barrier that consists of an outer membrane, a thin peptidoglycan layer, and an inner membrane (for review see [7]) (Figure 1). As the forward-facing structure of gram-negative bacteria, cell envelope components are more accessible than cytoplasmic factors making them ideal targets for peptide antibiotics. Below, we list several cell envelope targets and highlight some of the known peptide-based antimicrobials which inhibit these processes (Table 1).

Figure 1. Gram-negative cell envelope drug targets.

Diagram of important gram-negative bacterial cell envelope processes with the outer and inner membranes (OM; IM) and peptidoglycan cell wall labelled. The Beta barrel assembly machinery (Bam; blue), lipopolysaccharide transport complex (Lpt; green), lipoprotein transport system (Lol; purple), Signal peptidases I and II (orange), and Disulfide bonding system (Dsb, yellow) are highlighted. Portions of lipid II and lipoprotein trafficking are also represented.

Table 1:

Examples of peptide-based gram-negative cell envelope antimicrobials.

Target	Peptide	Structure	Mode of action	References
Outer/Inner Membrane	Colistin	Cyclic lipopeptide	Membrane disruption	[9]
	Protegrin-1	Cyclic β-hairpin peptide	Membrane disruption	[82]
Bam Complex	Darobactin	Cyclic modified heptapeptide	BamA inhibition	[19,20]
	Peptide 2	Linear peptide	BamD inhibition	[21]
Lpt Complex	Murepavadin	Synthetic cyclic β-hairpin peptide	LptD inhibition	[26–28]
	Thanatin	Cyclic β-hairpin peptide	LptA inhibition	[24,25]
Signal Peptidases	Arylomycin	Cyclic lipohexapeptide	Signal Peptidase I inhibition	[30]
	Globomycin	Cyclic depsipeptide	Signal Peptidase II inhibition	[31,32]
Dsb System	PFATCDS	Linear heptapeptide	DsbA-DsbB interaction	[34]

1.1. Cell membranes

The gram-negative bacterial outer membrane is composed of an outer leaflet of predominantly anionic lipopolysaccharides (LPS) and an inner leaflet of phospholipids (for review see [7,8]) (Figure 1). The outer membrane is targeted by different classes of macrocyclic cationic antimicrobial peptides (CAMPs) [9,10]. CAMPs are thought to initially bind the outer membrane through electrostatic interactions between positively charged side chains of residues like lysine, arginine, and diamino butyric acid (Dab) and the negatively charged LPS [11–13]. These interactions can lead to outer membrane destabilization allowing access to and disruption of the cytoplasmic inner membrane resulting in bacterial death. Macrocyclic CAMPs that target bacterial membranes include lipopeptides like the polymyxins and β-hairpin peptides like Protegrin-1 [9,10]. While these peptides all target bacterial membranes, their exact mechanisms appear to vary. For example, bacteria that are resistant to polymyxins are not necessarily more resistant to other types of CAMPs [14,15].

The outer membrane appears an ideal target for peptide antibiotic drug development due to ease of accessibility; however, selectivity between bacterial membranes and mammalian membranes remains a problem. Many membrane-active peptides also cause unacceptable levels of hemolysis and cytotoxicity, reducing their therapeutic utility (for review see [5,8]). A better understanding of membrane selectivity could reduce toxicity in the future. Even so, Colistin is currently used as a last-resort antibiotic against multidrug resistant gram-negative bacteria after being deemed too toxic early after its initial clinical introduction [16].

1.2. The Bam complex

The β-barrel assembly machinery (Bam) complex inserts newly synthesized β-barrel membrane proteins into the outer membrane (for review see [17]). The Bam complex consists of five subunits, BamA-E, of which BamA and BamD are essential for viability (Figure 1) [18]. Macrocyclic peptide inhibitors of BamA have been reported, including Darobactin, a cyclic heptapeptide which mimics a β-strand (Table 1) [19,20]. This structure allows Darobactin to bind BamA and lock it into a confirmation preventing insertion of other outer membrane proteins [19,20]. Additionally, a fifteen amino acid linear fragment of BamA, named Peptide 2 was found to interact with BamD and inhibit β-barrel protein folding in vitro (Table 1) [21]. The location of the Bam complex in and around the outer membrane makes it an intriguing target for macrocyclic peptide inhibitors. While parts of BamA are exposed to the extracellular environment, accessing BamD and other regions of BamA would require traversing at least part of the outer membrane.

1.3. The Lpt complex

The Lipopolysaccharide transport (Lpt) system carries LPS from the inner membrane across the periplasm and inserts it into the outer leaflet of the outer membrane (Figure 1). It consists of seven essential subunits denoted as LptA-G (for review see [22]). Macrocyclic β-hairpin peptide inhibitors have been identified for LtpA and LptD. LptA subunits span the periplasmic space and connect the inner membrane portion of the Lpt complex to LptD which resides in the outer membrane [23]. The macrocyclic β-hairpin peptide Thanatin [24] binds to a site of assembly between LptA subunits and prevents their association (Table 1). This disruption breaks the bridge between the cytoplasm and outer membrane and leads to cell death [25]. Murepavadin is a synthetic derivative of Protegrin-1 that inhibits LptD activity in Pseudomonas aeruginosa (Table 1) [26,27]. It binds the beta jelly roll domain of LptD and blocks LPS from reaching the outer membrane, resulting in cell death [26,28].

Both Thanatin and Murepavadin exhibit macrocyclic β-hairpin structure and inhibit different subunits of the lipopolysaccharide transport system (Lpt). Notably, both Thanatin and Murepavadin interact with sites of assembly within the Lpt complex by replacing normal subunit interactions, suggesting that this strategy could be deployed when designing synthetic macrocyclic peptide libraries to screen for antibiotic potential.

1.4. Signal Peptidase I and II

Bacterial signal peptidase (SPase) types I and II are essential enzymes found tethered to the cytoplasmic membrane and are responsible for cleaving signal peptides from secretory proteins (Figure 1) (for review see [29]). Macrocyclic peptides inhibiting both types of SPase have been identified. The arylomycin class of antimicrobial lipohexapeptides disrupt SPase I activity by binding the active site and inhibiting its activity (Table 1) [30]. Globomycin, an amphiphilic macrocyclic pentapeptide inhibits SPase II processes (Table 1) [31]. Disruption of SPase II function inhibits export of prolipoproteins to the outer membrane, thereby resulting in an accumulation of protein within the cell that results in cell death [32].

The conserved nature of SPase I and II and their relatively accessible active sites (periplasmic face of the inner membrane) make these enzymes worthwhile targets for antibiotic development. Accessibility is made more difficult in gram-negative bacteria because the outer membrane needs to be bypassed for target access

1.6. Additional targets

The disulfide bond (Dsb) forming system of enzymes catalyzes the formation of structural disulfide bonds in the periplasm of gram-negative bacteria (Figure 1) (for a review see [33]). The Dsb system is a promising candidate for antimicrobial targeting because DsbA and DsbB are essential for many pathogens to establish infections. This is because the Dsb system is required for the assembly of multiple virulence factors and a large portion of beta-lactamases. The linear peptide inhibitor PFATCDS inhibits the Dsb system by mimicking a DsbB periplasmic loop stopping DsbA-DsbB redox complex formation [34]. This suggests macrocyclic peptides could also be identified to target this system.

There are also essential gram-negative cell envelope targets that have not yet been inhibited by peptide-based antimicrobials. For example, the localization of lipoproteins (Lol) complex consists of subunits A-E and is responsible for transporting lipoproteins across the periplasm (Figure 1) [35]. Though there are no reported peptide inhibitors of the Lol complex, several small molecule inhibitors have been identified and are described elsewhere [36].

Lastly, several classes of macrocyclic peptides inhibit peptidoglycan synthesis in gram-positive bacteria primarily by targeting lipid II trafficking, incorporation, or crosslinking (for review see [37]). These include several lantibiotics and glycopeptides. Many, such as Vancomycin and Nissin, are active against gram-negative bacteria if the outer membrane is weakened [38]. This suggests that inhibiting peptidoglycan synthesis in gram-negative bacteria is possible with macrocyclic peptides if the outer membrane can be bypassed.

2. Cell-free macrocyclic peptide display

Cell free display of ribosomally synthesized peptides is commonly used to screen for direct peptide-protein interactions. Several in vitro display systems have been developed linking peptides to the mRNA encoding them. Ribosome display was among the first systems developed and used an mRNA encoded library lacking a stop codon, causing the ribosome to stall and form a polysome connecting the peptide to the coding mRNA [39,40]. This polysome mixture could be used for affinity selection without further purification; however, the non-covalent connections between the peptide and mRNA limited the type of screens and post-translational modifications that could be performed. To overcome this problem, an mRNA display system was developed where the peptide is covalently linked to its coding mRNA by puromycin, an antibiotic that mimics the aminoacyl end of a tRNA (Figure 2A) [41]. The resulting mRNA-peptide fusion can then be purified with poly-dT resin and reverse transcribed into a cDNA/mRNA duplex to prevent unintended binding by RNA aptamers. Variations of mRNA display have been developed, mostly focused on changing the puromycin-DNA linker [42,43]. Cell free peptide display also offers methods for incorporation of unnatural amino acids, some post-translational modifications, and conjugation to monosaccharides or antibiotics (for review see [44])

Figure 2. Peptide cyclization strategies for mRNA display.

A) An mRNA library is attached to puromycin via a DNA linker on the 3’ end. The mRNA molecules are translated in vitro and covalently linked to the C-terminus of its coding peptide through puromycin at the end of translation. Following purification, reverse transcription generates the complementary DNA to prevent binding from the mRNA aptamer in the downstream screening. B) Crosslinking of cysteines with dibromo-xylene (DBX). C) Thioether bond formed by a cysteine and 2,3-didehydroalanine (Dha). D) Crosslinking of amine groups from the peptide N-terminus and a lysine sidechain by disuccinimidyl glutarate (DSG). E) Cyclization of an N-terminal Nα-chloroacyl-Phenylalanine and a cysteine side chain. The starting amino acid was recoded to Nα-chloroacyl-Phenylalanine which enabled the cyclization to happen spontaneously upon translation. F) mRNA display of a head-to-tail cyclized peptide. After translation, a series of reactions were taken out, resulting in: i) the N-terminal portion of the peptide (up to the Cysteine residue) being attached to the C-terminal portion of the peptide through side chains (light green and pink residues); ii) Native chemical ligation (NCL) for head to tail cyclization of the N-terminal portion of the peptide.

Peptides identified through traditional mRNA display systems are linear and often suffer from issues of instability due to protease degradation in vivo. For this reason, technologies were developed to display macrocyclic peptides to increase stability and therapeutic usefulness. Macrocyclic mRNA display libraries can be produced through several different reactions all observed in natural antimicrobial peptides [45–47]: side chain to side chain, head to side chain, and head to tail cyclization. Each requires the reaction to be highly efficient and chemo selective for compatibility with downstream screening techniques. Below we detail these three types of reactions and the various ways they can be carried out.

2.1. Side Chain to side Chain

The simplest and most straightforward method of peptide cyclization is the formation of disulfide bonds between two cysteine side chains. Disulfide bonds allow for a high level of flexibility in ring size, but the spontaneous formation of disulfide bonds has low efficiency, especially if the cysteines are not near one another in the natural tertiary structure. Disulfide bond formation efficiency can be increased by adding protein disulfide isomerase (PDI) [48] and other disulfide-forming enzymes to the translation system, but even successful disulfide bonds can be reduced depending on buffer conditions, making them transient rather than permanent.

One variation of disulfide bond cyclization is cysteine alkylation involving two cysteines connected by a crosslinker such as dibromoxylene (DBX) (Figure 2B) making the linkage resistant to reduction and therefore permanent [49]. The reaction with DBX is performed in an aqueous solvent at pH 7.8 at room temperature with a mild reducing agent to prevent disulfide bond formation. The cyclization is fast, clean, and selective with an efficiency greater than 80%. This method has been used to cyclize both beta-hairpin and alpha-helix peptides and is suitable for mRNA display due to the mild conditions used in the reaction in contrast to other methods requiring harsher conditions [50,51]. This method has also been adapted to use crosslinkers with higher valency to create bicyclic and even tricyclic peptide mRNA display libraries [52].

Another side chain to side chain cyclization technique uses a thioether bond (Figure 2C) commonly found in lantibiotics, a thioether bond is formed between a cysteine side chain and either 2,3-didehydroalanine (Dha) or 2,3-didehydrobutyrine (Dhb). As it is difficult to directly incorporate these non-canonical amino acids into peptides during translation, another amino acid, selenalysine, needs to be added first using lysine codons as an intermediate [53]. The selenalysine residue is then oxidized by H₂O₂ into Dhha which will spontaneously undergo an intramolecular 1, 4-addition of the cysteine mercapto to form a thioether-glutathione bond. When adapted to mRNA display, the cysteine residues must be protected by glutathione prior to the oxidation step, later the cysteines are reduced to initiate cyclization [54].

2.2. Head to side chain

Another common cyclization method of mRNA display is head to side chain cyclization by a covalent bond between two amine groups, one from the N-terminus and the other a lysine residue using the crosslinker disuccinimidyl glutarate (DSG) [55] (Figure 2D). This cyclization protects the terminal amine group from protease degradation; however, the efficiency of intramolecular cyclization was roughly 55% in liquid for small ring sizes and decreased to ~30% when 12 amino acids were included in the ring [56]. The efficiency was moderately improved when mRNA-peptide fusion was purified and bound to solid phase during cyclization, as it helped to prevent intermolecular cyclization [57].

Limited by natural amino acid chemical groups, most cyclization methods (except for disulfide bond formation) require one or more post translational reactions. Non-canonical amino acids with additional functional groups can be incorporated in the elongation event through stop codon and sense-codon suppression [42]. However, reprogramming the initiation event is difficult due to the stringent requirement of the N-formyl group on the initiating methionine in prokaryotic translation. Suga group used two tools to reprogram translation initiation for mRNA display [58]. The first is a reconstituted E. coli translation system, referred to as the PURE system, where certain amino acids could be withdrawn to create vacant codons in the genetic table [59]. The second tool, termed Flexizyme, is an artificially evolved ribozyme that can acylate a tRNA with any amino acid [60]. With these two systems, they built a peptide library with Nα-chloroacyl-Phenylalanine as the N-terminal amino acid and a cysteine as the C-terminal residue. Cyclization of such peptides happens spontaneously after translation because the α-carbon of the Nα-ClAc group is attacked by sulfhydryl group of cysteine to close the ring [58] (Figure 2E). This method has been applied to mRNA screens against a dozen different targets and produced hits with low affinity and inhibition of target function. Further optimization of this method used stereo exclusion to select for the cysteine at the carboxyl terminus over upstream cysteines introduced by random codons [61].

2.3. Head to tail

Peptides with the N-terminal amine group cyclized with the C-terminal acyl group are exceptionally stable and commonly found in many organisms as defense tools against viruses, bacteria, fungi, and insects [62,63]. One example is θ-defensin-1, an antimicrobial peptide containing eighteen amino acids forming a beta-hairpin conformation [45,64]. Although there is great interest in screening and testing head-to-tail macrocyclic peptides for therapeutics, backbone cyclization is not easily compatible with mRNA display because the C-terminal acyl group of the peptide is covalently linked to puromycin. To overcome this problem, one option is to attach the peptide to the mRNA using a peptide side chain, this frees the C-terminal acyl group for cyclization (Figure 2F). Suga group developed an elegant series of reactions to achieve this [65]. Following peptide translation, two multireaction steps were performed: 1) connect a side chain from the region to be cyclized to the downstream peptide region. 2) intramolecular native chemical ligation (NCL) between the N-terminal amine group and a downstream thioester group introduced directly by ribosomal translation, leading to backbone cyclization. The technique has been used to generate a library of circular peptides containing thirty-four amino acids and identified inhibitors of human beta-factor XIIa [66].

3. Phage and antibacterial macrocyclic peptide display

Methods of peptide display are available using several biological organisms, including bacteriophage, bacteria, and yeast. These display libraries can be quickly screened using multiple rounds of panning, fluorescent sorting, or using next-generation sequencing to identify peptides from large libraries with an activity of interest. Here we focus on methods using macrocyclic peptide phage display to identify potential target interactors and an antibacterial display method for identification of macrocyclic peptides with general antibiotic activity targeting the gram-negative cell envelope called surface localized antimicrobial display (SLAY).

3.1. Phage peptide display

Phage display is a well-established method to identify peptides that can bind to a specific target. In peptide phage display, sequences are cloned for expression as fusions to specific phage coat proteins. Large pools of phage are generated to produce diverse peptide libraries. A phage population can then be screened for interactions between displayed peptides and a desired target. These interactions can be further enriched over multiple selection cycles [67]. Phage display of peptides also offers methods for post-translational modification including phosphorylation and conjugation (for review see [44]).

Display of macrocyclic peptide libraries can also be achieved using phage (for review see [68,69]). One method involves encoding cysteines at the ends of the variable region to potentiate a disulfide bond. Though this method has proven successful in identifying macrocyclic peptide ligands, there is no guarantee that a disulfide bond will form between the two cysteines of every peptide in the library. Also, if a disulfide bond does form it can be reversibly reduced in certain environments. A recent modification of this method has been developed using macrocyclic organo-peptide hybrids (MOrPHs) [70]. This method involves the replacement of one cysteine with the noncanonical cysteine-reactive amino acid O-(2-bromoethyl)-tyrosine (O2beY), incorporated using amber stop codon suppression and an engineered aminoacyl-tRNA synthetase. O2beY spontaneously reacts with cysteine to create a nonreducible thioether bridge when displayed. This method is reported to have successfully isolated macrocyclic peptides with nanomolar binding affinity to their target. Chemical crosslinkers can also be used to produce macrocyclic peptides in phage display by creating bicyclic peptides similarly to cell-free systems [71]. This technology involves the inclusion of three cysteines in the peptide variable region displayed on the phage coat protein which are subsequently connected with the trivalent thiol-reactive compound tris-(bromomethyl) benzene (TBMB) in situ, resulting in the production of bicyclic peptides on the surface of the phage library.

Macrocyclic peptide phage display is beneficial for screening exceptionally large libraries in a fast, straightforward manner and is ideal for finding lead peptides binding and potentially inhibiting the activity of soluble cell envelope targets like some of those discussed above. However, interaction with a potential target does not guarantee inhibition of its activity. Thus, further investigation and optimization of lead macrocyclic peptide hits would be required to ensure inhibition of the desired target’s activity.

3.2. Surface localized antimicrobial display

A couple major drawbacks of the display technologies mentioned above is that there is no selection for reaching the desired target location and selection only occurs for target binding, not inhibition of the target’s activity. This makes any lead peptide identified in these screens unlikely to be effective as a therapeutic drug without additional evaluation and further optimization. During treatment of a bacterial infection, an antibiotic must interact with a bacterium at its cell surface and then migrate to the target location. For gram-negative bacteria this often means bypassing the outer membrane.

Recently a new strategy was developed using bacterial peptide display to select for peptides inhibiting bacterial growth from the extracellular space entitled surface localized antimicrobial display or SLAY for short. SLAY involves the display of a large peptide library on the bacterial cell surface encoded on an inducible plasmid system. The peptide library is tethered to the outer membrane via a transmembrane domain and the linker is long enough to allow peptides to reach potential periplasmic targets [72,73] (Figure 2). The library is transformed into the desired bacteria and screened by monitoring the plasmid populations in induced and uninduced cultures via next-generation sequencing to quantify the number of reads present for each plasmid encoded peptide. A drop in reads in the induced samples, relative to the uninduced samples, implies growth inhibition and antibacterial potential. This technique was used to initially screen 800,000 random peptide sequences leading to the identification of hundreds of hits. Further examination of a subset of hits confirmed the antibiotic activity of ~20 novel antibiotic peptides [72].

Some of the peptides identified in the initial SLAY screen contained cysteines potentiating macrocyclic structure through disulfide bond formation. Further characterization of one of these peptides, symbah-1, showed it contained an intermolecular disulfide bond and a beta-hairpin secondary structure. This confirmed that SLAY can be used to display and identify structured macrocyclic peptides with antibacterial activity [14]. The SLAY method has since been modified for identification of macrocyclic peptide antibiotics specifically through the inclusion of cysteines on the terminal ends of the variable region [73]. A subsequent SLAY screen of a peptide library with terminal cysteines and beta-hairpin secondary structure identified dozens of new macrocyclic antimicrobial sequences [74] Innovative screening strategies like SLAY which can screen directly for inhibition of bacterial growth can significantly improve the efficiency with which lead synthetic macrocyclic peptides with clinical antibiotic potential are identified.

4. Future strategies for macrocyclic peptide library design and screening

For an antibiotic to be clinically effective it must do more than simply inhibit bacterial growth. It must also be able to reach the site of infection, remain stable and soluble upon accessing its target, and be non-toxic to human cells. This requires overcoming several problems commonly encountered by peptide therapeutics including, protease lability, rapid clearance, and a lack of membrane specificity.

Antibiotics with the above attributes remain quite difficult to identify. Current estimations predict thousands, if not millions, of bacterial species need to be investigated to find one novel and viable natural product lead [75,76]. This suggests traditional screening approaches for antibiotics simply cannot keep pace with the current rate of resistance development. This means any new methods of synthetic discovery, including macrocyclic peptide display strategies, must efficiently identify many leads in a high-throughput manner to be a viable path forward. Each peptide lead will likely require further modification like incorporation of non-canonical amino acids since all current clinically used macrocyclic peptide antibiotics contain such residues. Here we discuss some macrocyclic peptide library design and display screening strategies which may help increase the efficiency of clinical lead identification in the future.

4.1. Strategies for macrocyclic peptide library design

Display screens of peptides with completely random sequence potentiate the identification of peptides with truly unique sequence and structure but may not identify a high percentage of viable therapeutic hits, especially for screens only focusing on target binding. A more reliable way to identify leads with potent antibiotic potential is to use known peptide antibiotics as library scaffolds. Using natural peptide antibiotics as synthetic scaffolds has already proven effective in optimizing lead peptides like Thanatin [77], and Protegrin-1 [78], even leading to clinically relevant drugs like Murepavadin [27]. The large-scale library and high-throughput screening offered by display techniques improves on these smaller scale analog screening methods and could significantly improve antimicrobial potency.

Alternatively, macrocyclic library scaffolds may be biased toward the structural characteristics and properties of known peptide antibiotics to enhance attributes like stability and target binding. Most known peptide antibiotics are highly charged, helping them to interact with negatively charged membranes. They also often adopt alpha helical or beta hairpin secondary structures which can help stability and binding. This knowledge can be used to bias libraries toward one or more of these characteristics to improve the efficiency with which viable leads are identified.

Another potential strategy is to have macrocyclic library scaffolds mimic portions of essential proteins involved in essential complex interactions. As mentioned in section one, Thanatin [25], Peptide 2 [21], Murepavadin [26,28], and PFATCDC [34] mimic fragments of larger proteins which engage in essential complex interactions. Theoretically, macrocyclic peptide library scaffolds could be designed to mimic portions of essential cell envelope proteins involved in critical complex interactions and then screened for competitive binding or antimicrobial activity using macrocyclic peptide display technologies. Though this method is still unproven, the examples mentioned above demonstrate the potential such a strategy might have on improving efficient identification of lead peptides.

4.2. Using computational analysis and machine learning to inform future library design

Advancements in computational analysis and machine learning continue to grow at impressive rates. These advancements have created a renaissance of synthetic antibiotic discovery by using datasets accumulated from previous screens. Such analyses are helping to describe small molecule characteristics affecting antimicrobial activity and the ability to bypass bacterial membranes [79,80].

Such computational advancements have also taken place with peptide antibiotics, where computer aided approaches are beginning to offer insights into the optimization of lead peptides both de novo and in silico (for review see [81]). These principles could also be applied to the design of macrocyclic peptide libraries to enhance the likelihood of identifying peptides with antimicrobial activity and other therapeutic qualities through display screens. Current datasets for macrocyclic peptide antibiotics are quite small in comparison to small molecules, so further expansion of macrocyclic peptide antibiotic datasets using the screening techniques described here could also help expand these datasets and enhance our future predictive power.

4.3. Screening in bacterial pathogens under more therapeutically relevant conditions

All drug screening strategies should strive to select for the precise therapeutic characteristics desired from a large library of candidate molecules in a high throughput manner. Traditional synthetic antibiotic screening strategies are high throughput, but often do not screen for many of the characteristics necessary for therapeutic success. If the goal is to isolate synthetic antibiotics that are clinically effective against specific bacterial pathogens, then the conditions of the screen should attempt to accurately reflect the conditions during treatment. Advancements such as SLAY consider attributes like target accessibility and growth inhibition beyond simple target binding, but still do not replicate the environment in which the drug must remain active and do not screen for activity against specific multidrug resistant pathogens of interest.

To date, SLAY screens have only been performed in lab strains of E. coli grown in traditional laboratory growth medium. The surface display system used in SLAY has also shown functionality in other gram-negative bacteria, including strains closely related to the known pathogens Acinetobacter baumannii and Pseudomonas aeruginosa [72]. To screen for therapeutic lead peptides more accurately, SLAY could be adapted to function in more clinically relevant or even multidrug-resistant strains. Using SLAY in such strains potentiates screening under more therapeutically relevant conditions like growth in blood serum or performing screens during actual tissue infection using a murine or insect model of infection. Such advancements would allow us to select for macrocyclic peptides that kill a specific clinical pathogen and are active in therapeutically relevant conditions, improving the likelihood with which we isolate clinically relevant lead peptides. Such screening advancements, or the development of other novel display methodologies, could help us more efficiently identify new drugs to combat the development and spread of multidrug resistant pathogens in the future.

5. Conclusions

New methods of antimicrobial discovery are critically needed to help revitalize the clinical antibiotic pipeline. Many natural macrocyclic peptide-based products show an ability to bind and inhibit essential processes within the bacterial cell envelope, but traditional methods of natural discovery remain inefficient. Macrocyclic peptide display technologies offer a high throughput synthetic alternative to traditional natural discovery. Advancements in such technologies are allowing screens for antimicrobial function rather than simple target binding, increasing the efficiency with which antimicrobial products are identified. Further advancements such as incorporation of machine learning into library design and use of more therapeutically relevant screening conditions could further increase the efficiency with which clinically relevant products are identified.

Figure 3. Surface localized antimicrobial display of macrocyclic peptides.

Diagram of the surface localized antibacterial macrocyclic peptide display system. The display machinery is transported to the periplasm by a signal peptide which is cleaved and then is embedded into the outer membrane (OM) by encoding a fragment of E. coli OmpA. A long tether then leads to the variable region of the peptide library. Macrocyclic structure is potentiated through two cysteines (yellow) encoded on the terminal ends of the variable region potentiating a disulfide bond.

Highlights.

• More efficient strategies for antibiotic discovery are critically needed

• Macrocyclic peptide-based products can target bacterial cell envelope processes

• Cell-free and cell-based display technologies function with peptide macrocycles

• This permits high-throughput screening for novel macrocyclic peptide antibiotics