Development and Challenges of Antimicrobial Peptides for Therapeutic Applications

More than 3000 antimicrobial peptides (AMPs) have been discovered, seven of which have been approved by the U.S. Food and Drug Administration (FDA). Now commercialized, these seven peptides have mostly been utilized for topical medications, though some have been injected into the body to treat severe bacterial infections. To understand the translational potential for AMPs, we analyzed FDA-approved drugs in the FDA drug database. We examined their physicochemical properties, secondary structures, and mechanisms of action, and compared them with the peptides in the AMP database. All FDA-approved AMPs were discovered in Gram-positive soil bacteria, and 98% of known AMPs also come from natural sources (skin secretions of frogs and toxins from different species). However, AMPs can have undesirable properties as drugs, including instability and toxicity. Thus, the design and construction of effective AMPs require an understanding of the mechanisms of known peptides and their effects on the human body. This review provides an overview to guide the development of AMPs that can potentially be used as antimicrobial drugs.


Introduction
In the past several decades, multidrug-resistant bacteria have rapidly spread, causing increases in nosocomial infections and in-hospital mortality, and posing a threat to global health [1][2][3][4]. Moreover, the discovery of new classes of antibiotics has slowed down since 1987 [5,6]. The lack of new discoveries might be prompted by the conservative way we have searched for antibiotics, or this field may be saturated [5,6]; in other words, we may have already discovered many of the large natural structures that have antimicrobial activity. With the rise of antibiotic resistance, our last lines of effective antibiotics are failing [7][8][9]. Antimicrobial peptides (AMPs), a ubiquitous part of the innate immune defense in all classes of life, have been widely studied and show potential as small molecule antibiotics [10][11][12].

Antimicrobial Peptide Database
To date, only seven small AMPs have been approved by the FDA, so we extended our study to other AMPs that are under development and listed in the Antimicrobial Peptide Database (APD). A total of 3156 AMPs is listed in the APD, most of which were discovered in nature [13]. An analysis of 2700 of the 3156 AMPs in the APD showed that these peptides all have different structures and sequence motifs, and because they have a broad spectrum, they can kill a range of pathogens [42,43]. Interestingly, one-third of the AMPs are derived from frogs [44]. The average length of peptides in the APD is 33 amino acids, the median length is 28 amino acids, and more than 90% of the peptides, known as small peptides, have no more than 50 amino acids ( Figure 3A). The average hydrophobic content of the peptides is 54% (Figure 3B), and the mean peptide net charge is +3 ( Figure 3C). About 45% of the peptides do not contain cysteine; 21% and 17% of them have two cysteines and six cysteines, respectively ( Figure 3D), which reveals the potential of a disulfide bond formation between two cysteines. Antibiotics 2020, 9, x FOR PEER REVIEW 4 of 22

Antimicrobial Peptide Database
To date, only seven small AMPs have been approved by the FDA, so we extended our study to other AMPs that are under development and listed in the Antimicrobial Peptide Database (APD). A total of 3156 AMPs is listed in the APD, most of which were discovered in nature [13]. An analysis of 2700 of the 3156 AMPs in the APD showed that these peptides all have different structures and sequence motifs, and because they have a broad spectrum, they can kill a range of pathogens [42,43]. Interestingly, one-third of the AMPs are derived from frogs [44]. The average length of peptides in the APD is 33 amino acids, the median length is 28 amino acids, and more than 90% of the peptides, known as small peptides, have no more than 50 amino acids ( Figure 3A). The average hydrophobic content of the peptides is 54% (Figure 3B), and the mean peptide net charge is +3 ( Figure 3C). About 45% of the peptides do not contain cysteine; 21% and 17% of them have two cysteines and six cysteines, respectively ( Figure 3D), which reveals the potential of a disulfide bond formation between two cysteines.  In the APD, 1869 of the 2700 peptides (~70% of the database) are small cationic amphipathic peptides. However, of the FDA-approved AMPs, only one, colistin, is in this category of small cationic amphipathic peptides. Gramicidin is small (10 amino acids) and has a net charge of +2, but it contains eight hydrophobic residues and two positively charged lysine residues, which makes it a small cationic hydrophobic peptide. Many studies have proposed that membrane-active AMPs selectively target and disrupt anionic bacterial cell membranes using electrostatic interactions [45][46][47][48][49]. However, daptomycin, a small amphipathic peptide with a neutral net charge, deviates from this pattern. Because vancomycin, oritavancin, dalbavancin, and telavancin are lipoglycopeptides, they are not included in the APD, which comprises only peptides and lipopeptides.

Current Development of Peptide Drugs
We further analyzed FDA-approved small peptide (less than 50 amino acids) therapeutics from the past 20 years (total 57 drugs; Table 1) using Drugs@FDA (http://www.fda.gov/drugsatfda) and DrugBank [50]. Details regarding newly approved compounds (submission classification: Type 1-New Molecular Entity) were extracted from Drugs@FDA, and the data were further confirmed by DrugBank. A total of 555 new molecules were approved and commercialized between January 1999 and December 2019 ( Figure 4A). Many peptide therapeutics are not included in the THPdb, and the listed molecules are not limited to AMPs. Fifty-seven of the molecules are small peptide therapeutics, and most of these (37 drugs) are receptor-binding peptides that either activate or block the specific receptors to which they bind, causing a biological response. The rest of them are inhibitors of biological pathways (15 drugs), membrane-active peptides (MAPs; 4 drugs), or have other functions (1 drug) ( Figure 4B). Below, we will discuss the various ways in which peptides can interact with cells to perform their therapeutic functions.
Antibiotics 2020, 9, x FOR PEER REVIEW 5 of 22 In the APD, 1869 of the 2700 peptides (~70% of the database) are small cationic amphipathic peptides. However, of the FDA-approved AMPs, only one, colistin, is in this category of small cationic amphipathic peptides. Gramicidin is small (10 amino acids) and has a net charge of +2, but it contains eight hydrophobic residues and two positively charged lysine residues, which makes it a small cationic hydrophobic peptide. Many studies have proposed that membrane-active AMPs selectively target and disrupt anionic bacterial cell membranes using electrostatic interactions [45][46][47][48][49]. However, daptomycin, a small amphipathic peptide with a neutral net charge, deviates from this pattern. Because vancomycin, oritavancin, dalbavancin, and telavancin are lipoglycopeptides, they are not included in the APD, which comprises only peptides and lipopeptides.

Current Development of Peptide Drugs
We further analyzed FDA-approved small peptide (less than 50 amino acids) therapeutics from the past 20 years (total 57 drugs; Table 1) using Drugs@FDA (http://www.fda.gov/drugsatfda) and DrugBank [50]. Details regarding newly approved compounds (submission classification: Type 1-New Molecular Entity) were extracted from Drugs@FDA, and the data were further confirmed by DrugBank. A total of 555 new molecules were approved and commercialized between January 1999 and December 2019 ( Figure 4A). Many peptide therapeutics are not included in the THPdb, and the listed molecules are not limited to AMPs. Fifty-seven of the molecules are small peptide therapeutics, and most of these (37 drugs) are receptor-binding peptides that either activate or block the specific receptors to which they bind, causing a biological response. The rest of them are inhibitors of biological pathways (15 drugs), membrane-active peptides (MAPs; 4 drugs), or have other functions (1 drug) ( Figure 4B). Below, we will discuss the various ways in which peptides can interact with cells to perform their therapeutic functions.   Table 1. Summary of the small peptide (less than 50 amino acids) therapeutics approved by the FDA between January 1999 and December 2019. Raw data (submission classification: Type 1-New Molecular Entity) were collected from Drugs@FDA (http://www.fda.gov/drugsatfda) and the data were further confirmed by DrugBank [50]. "MAP" is defined as "membrane-active peptide".

Receptor-Binding Peptides
Receptor-binding peptides, which include both agonists and antagonists, constitute the major category of therapeutic peptides that have been approved by the FDA. These peptides have been used as therapeutics and diagnostics for applications other than infectious diseases (see Table 1). Seventeen of the 37 FDA-approved receptor-binding peptides are insulin and its analogs, which are used for treating diabetes. Four of the 37 have been utilized as anti-cancer drugs: two for prostate cancer and two for neuroendocrine tumors. Some of the FDA-approved peptides have immune-modulating effects.
However, no receptor-binding peptide has yet received FDA-approval as an antimicrobial therapeutic [51]. Small-molecule drugs have been widely studied to modulate the immune system, e.g., drugs that interact with the toll-like receptor [52][53][54]. Given this capacity, it is possible that receptor-binding peptides could be used to treat infections by stimulating the immune system. Future directions of research may investigate the use of peptides to modulate the immune system instead of, or in addition to, killing bacteria directly.
In fact, several multifunctional AMPs have been used experimentally to modulate the immune response and kill pathogens [55][56][57]. For example, human cathelicidin LL-37 and human β defensins activate the toll-like receptor signal in the innate immune system [58,59]. Nevertheless, these peptides may be a double-edged sword in that a higher dose of AMPs (e.g., cathelicidin LL-37) or their proteolytic peptide fragments could result in off-target effects and trigger additional chronic inflammatory diseases, e.g., atopic dermatitis, rosacea, psoriasis, and hidradenitis suppurativa [60]. Clinical studies of LL-37 as a topical treatment for chronic leg ulcers has demonstrated safety [61]. LL-37 has entered phase II clinical studies for further investigation of its antimicrobial activity and its ability to modulate inflammation and the healing rate of diabetic foot ulcers (see more information on ClinicalTrials.gov website: https://clinicaltrials.gov; ClinicalTrials.gov Identifier: NCT04098562). More clinical studies of LL-37 are needed to explore the efficacy and potential side effects of this molecule [61]. In future work, synthetic peptides that have precise immunomodulatory effects together with direct antimicrobial activity may be designed as a promising route.

Membrane-Active Peptides (MAPs)
Five out of the seven FDA-approved AMPs are MAPs (Table 1): gramicidin, daptomycin, oritavancin, telavancin, and colistin. More specifically, gramicidin is a pore-forming peptide that forms ion channels as a transmembrane dimer. Daptomycin is a membrane lytic peptide that does not form pores but co-clusters with anionic lipids and lyses the membrane. These peptides aggregate and assemble in bacterial membranes, promoting membrane depolarization via different pathways [16,23,24,[62][63][64]. Oritavancin and telavancin are dual-mechanism AMPs: they (i) inhibit bacterial cell wall synthesis and (ii) disrupt bacterial membranes. Although a few studies have suggested that these peptides have features similar to those of another pore-forming AMP, nisin [28][29][30]65], their actual mechanisms of membrane disruption remain unclear. Oritavancin and telavancin may either form membrane pores or channels, or lyse the membrane. Colistin forms pore-like aggregates in the bacterial cell membrane and disrupts the membrane; thus, it results in lytic cell death [66,67].
Membrane pore-forming AMPs constitute a large subgroup of MAPs. These peptides bind to cell membranes and spontaneously assemble in the lipid bilayer as a channel or pore-like structure, though not all are cytolytic ( Figure 5). Well-known natural examples are gramicidin [68], colistin [67], melittin [69,70], maculatin [71], and alamethicin [72]. The two common models for the channel structures are barrel-stave and toroidal, depending on how the peptide interacts with the lipid headgroups [73]. These oligomeric structures can be a homogeneous population of oligomers or have diverse multimeric sizes that can conduct water and ions across the membrane. The pore size, which varies among different peptides, can be measured by several biophysical techniques and molecular dynamics simulations [44,71,[74][75][76][77]. Membrane-lytic peptides, e.g., daptomycin [23,24,26,78,79], colistin [67], LL-37 [80], aurein 1.2 [81,82], and piscidin 1 [83,84], disrupt cell membranes, like detergents. These peptides accumulate on the membrane surface, carpet the membrane at a critical threshold concentration, and destabilize and permeabilize the membrane structure. Some membrane-lytic peptides, e.g., melittin and colistin, form pores at low peptide concentration and lyse the cell membrane above a threshold concentration or interact with specific membrane types [85].
Unlike receptor-binding peptides or peptide inhibitors that have specific binding targets, membrane-active peptides, whose activity is limited to specific cell membranes, are not well-defined. Their specificity is usually caused by their hydrophobic moment and electrostatic interactions, but exact mechanisms have not yet been determined [86][87][88]. These properties limit the ability to precisely tune the selectivity of membrane-active peptides toward a specific bacterial species [44,70,71,74,89,90]. Bacterial membranes are generally composed of more negative charges [91], whereas mammalian cell membranes contain abundant cholesterol and sphingomyelin, which make the membranes more rigid [92,93]. The advantage of utilizing membrane-active peptides for antibiotics is that bacteria may have less of a chance to develop drug resistance [94]. A deeper understanding of the molecular mechanisms underlying bacterial membrane disruption may enable further fine-tuning of the hydrophobic moment and charge distribution, and improvement of specificity.

Peptides Having Other Inhibitory Mechanisms
Other potential targets for inhibiting bacterial growth include DNA, RNA, and ribosomes (protein synthesis). Peptides with these mechanisms include edeine [96,97], tuberactinomycins [98], and dityromycin [99]. Edeine is an antimicrobial pentapeptide that binds to the binding site (P-site) of both 30S subunits and 70S ribosomes, thus inhibiting the binding of aminoacyl-tRNA and blocking translation initiation [96,97]. Tuberactinomycins, a group of cyclic peptides, inhibit prokaryotic protein synthesis and group I self-splicing via binding to the G-binding site and backbone of the intron RNA [98]. The antimicrobial cyclic decapeptide dityromycin has been shown to block Membrane-lytic peptides, e.g., daptomycin [23,24,26,78,79], colistin [67], LL-37 [80], aurein 1.2 [81,82], and piscidin 1 [83,84], disrupt cell membranes, like detergents. These peptides accumulate on the membrane surface, carpet the membrane at a critical threshold concentration, and destabilize and permeabilize the membrane structure. Some membrane-lytic peptides, e.g., melittin and colistin, form pores at low peptide concentration and lyse the cell membrane above a threshold concentration or interact with specific membrane types [85].
Unlike receptor-binding peptides or peptide inhibitors that have specific binding targets, membrane-active peptides, whose activity is limited to specific cell membranes, are not well-defined. Their specificity is usually caused by their hydrophobic moment and electrostatic interactions, but exact mechanisms have not yet been determined [86][87][88]. These properties limit the ability to precisely tune the selectivity of membrane-active peptides toward a specific bacterial species [44,70,71,74,89,90]. Bacterial membranes are generally composed of more negative charges [91], whereas mammalian cell membranes contain abundant cholesterol and sphingomyelin, which make the membranes more rigid [92,93]. The advantage of utilizing membrane-active peptides for antibiotics is that bacteria may have less of a chance to develop drug resistance [94]. A deeper understanding of the molecular mechanisms underlying bacterial membrane disruption may enable further fine-tuning of the hydrophobic moment and charge distribution, and improvement of specificity.

Peptides Having Other Inhibitory Mechanisms
Other potential targets for inhibiting bacterial growth include DNA, RNA, and ribosomes (protein synthesis). Peptides with these mechanisms include edeine [96,97], tuberactinomycins [98], and dityromycin [99]. Edeine is an antimicrobial pentapeptide that binds to the binding site (P-site) of both 30S subunits and 70S ribosomes, thus inhibiting the binding of aminoacyl-tRNA and blocking translation initiation [96,97]. Tuberactinomycins, a group of cyclic peptides, inhibit prokaryotic protein synthesis and group I self-splicing via binding to the G-binding site and backbone of the intron RNA [98]. The antimicrobial cyclic decapeptide dityromycin has been shown to block elongation factor G (EF-G)-catalyzed translocation by disrupting the contact between EF-G and ribosomal protein S12, so that it deactivates the ribosome-EF-G complex and prevents translocation [99].

Conclusions and Future Perspectives
Peptide therapeutics have only made up a minority of all new molecular entities approved by the FDA (Figure 4A). Peptides have mostly been utilized to treat bacterial skin infections, pink eye, or wounds [100,101], e.g., Neosporin ® (gramicidin; manufactured by Monarch Pharmaceuticals, Inc., Bristol, TN, USA), Cubicin ® (daptomycin; manufactured by Merck & Co., Inc., Kenilworth, NJ, USA), Vancocin ® HCl (vancomycin; manufactured by ANI Pharmaceuticals, Inc., Baudette, MN, USA), Orbactiv ® (oritavancin; manufactured by Melinta Therapeutics, Inc., New Haven, CT, USA), Dalvance TM (dalbavancin; manufactured by Allergan Sales, LLC, Irvine, CA, USA), and Vibativ ® (telavancin; manufactured by Theravance Biopharma, Inc., San Francisco, CA, USA). Several AMPs have been approved for direct injection into the human body, e.g., Cubicin, Vancocin, Orbactiv, Dalvance, Vibativ, and Coly-Mycins, because of their longer elimination half-life (ranging from hours to days) and better pharmacokinetics [31][32][33][34] compared with gramicidin or other AMPs. However, most of these lipopeptide antibiotics (except colistin) are used for treating Gram-positive bacterial infections, and only a few of them have been administered as oral solutions or tablets because of their poor penetration of the intestinal mucosa [102]. Oral vancomycin (Vancocin) is limited to the treatment of Gram-positive bacterial infections, such as Clostridium difficile diarrhea and staphylococcal enteritis, because of its poor absorption and ingestion in the body and the severity of these infections. AMPs to treat infections caused by Gram-negative bacteria are clearly needed.
Although vancomycin has been approved by the FDA, several clinical studies have shown that it may cause kidney damage in some patients or at high doses. Oritavancin and dalbavancin were, in fact, developed to improve the antibacterial activity of vancomycin, so that the dose could be reduced and toxicity lowered or prevented. Although the side effects of these compounds are mild, their effectiveness against drug-resistant Gram-positive organisms and for long-term treatment remains ambiguous [103][104][105]. Telavancin, another derivative of vancomycin, is more effective for treating a range of drug-resistant Gram-positive bacteria, but it has been reported that it may induce acute kidney injury and have a higher death rate than vancomycin [106,107]. Colistin may cause damage to the kidneys and the central nervous system in adult patients, and heavy use of colistin can result in the occurrence of colistin-resistant bacteria, making it problematic for regular use [108][109][110]. Other extensively studied pore-forming AMPs, such as alamethicin and melittin, are hemolytic and cytotoxic [69,111,112]; therefore, no clinical study has been conducted (see more information on ClinicalTrials.gov website: https://clinicaltrials.gov). This suggests that controlling the selectivity, reducing the toxicity, and lowering unexpected side effects are essential to the design of AMPs as human medicines [38,49].
The seven FDA-approved AMPs are small, with a molecular weight between 1145 and 1882. They are composed of several noncanonical amino acids and have chemical modifications or cyclic structures. These features optimize their pharmacokinetics and extend their elimination half-life so that they resist enzymatic degradation. The APD contains 3156 AMPs, 98% of which were discovered in nature [13]: many, in fact, were extracted from the skin secretions of frogs [44,[113][114][115] or are toxins from other species, e.g., bees, snakes, and wasps [86,116,117]. In contrast, all the FDA-approved AMPs were discovered in or derived from Gram-positive bacteria commonly found in the soil: Brevibacillus brevis (gramicidin), Streptomyces roseosporus (daptomycin), Amycolatopsis orientalis (vancomycin, which is the prototype of oritavancin, dalbavancin, and telavancin), and Paenibacillus polymyxa (colistin) [118]. This coincidence is not surprising, as soil bacteria are also the source of many conventional antibiotics.
Numerous approaches to peptide design have been introduced to make AMPs less toxic for humans while maintaining or improving their potency to eliminate bacteria [11,42], e.g., rational design [119,120], combinatorial peptide libraries [75,121], high-throughput screening [122,123], database-guided approaches [124,125], structure-function-guided design [86,126,127], and molecular dynamics simulations [44]. Three major methods to improve AMP function have been described: (i) High-throughput screening can be used to identify potential AMPs [128][129][130]. The SPOT-synthesis technique, for instance, has been applied to medium-or high-throughput screens; with this technique, peptide arrays are synthesized on a cellulose membrane, and the peptides can be easily cleaved from the support for screening. In addition, combining computer algorithms, automated synthesis, and automated screening for drug design can rapidly accelerate and reduce the cost of labor for drug discovery [131]. (ii) Conjugation of peptides to other active molecules (for example, antibodies) or incorporation of peptides into nanoparticles or dendrimers allows the advantages of both types of biomolecules to be combined and overcomes their weaknesses [132]. Synthesizing AMP polymers using dendrimer or other AMP nanoparticles to increase the local concentration of the AMP can lower the required dose and combat multidrug-resistant bacteria [133][134][135][136]. (iii) The development of in vitro tests and computational predictions, for example, testing for similarities with allergens, can help to evaluate the immunogenicity and allergenic potential of newly developed AMPs [137], e.g., the basophil activation test [138], cytokine assays [139], lymphocyte activation analysis [140], and Structural Database of Allergenic Proteins (SDAP) [141][142][143]. Although these tools cannot fully predict the allergenic effects of a new AMP in the clinic, they provide a promising preclinical tools for evaluating peptide-based drugs.
We compared the AMPs from the APD [13] with the seven FDA-approved AMPs and found that the peptide sequences and physicochemical properties (e.g., hydrophobic content and net charge) vary widely among AMPs (Figure 3). These features merit additional study to determine what contribution they make to antimicrobial activity and ultimate clinical utility. It is unclear why the development of many natural peptides has stopped during preclinical stages or why the peptides failed to show sufficient antimicrobial activity and drug-like properties in clinical studies [5,10,134,144]. Their poor performance may derive from differences between the clinical setting and their native conditions. We speculate that in nature, peptides may participate in cooperative pathways with other chemical compounds or enzymes, and these conditions may increase their potency against bacteria. Synthetic AMPs used in isolation, on the other hand, may not have equally strong antimicrobial activity.
Based on these observations, it may be important not only to design synthetic AMPs that have antimicrobial activity, but to optimize them for desirable clinical properties, such as: (i) stabilizing peptide structures and introducing non-canonical amino acids into the peptide sequences to extend their elimination half-lives; (ii) mining antimicrobial agents from Gram-positive bacteria in soil and understanding which biochemical properties make these suitable for human use; (iii) discovering AMPs that can modulate the immune system; (iv) evaluating the potential synergy of AMPs with other chemical compounds or enzymes to enhance antimicrobial activity; (v) optimizing the design of computational tools for peptide therapeutics and high-throughput screening; and (vi) developing appropriate in vitro models that mimic in vivo conditions to evaluate the allergenic effects. Overall, AMPs can be a tool against drug-resistance bacteria and a source of promising therapeutics to treat infectious diseases. More investigation in the clinical setting is suggested. Funding: This work was funded through a Defense Threat Reduction Agency grant (HDTRA11810041) to TKL. DTRA did not have any role in the design of the study, data analysis, or manuscript writing. TKL and CHC were supported by the National Institute of Allergy and Infectious Diseases of the National Institutes of Health under Award Number U19AI142780. The content is solely the responsibility of the authors and does not necessarily represent the official views of the NIH.