A review on the screening methods for the discovery of natural antimicrobial peptides

Abstract

Natural antimicrobial peptides (AMPs) are promising candidates for the development of a new generation of antimicrobials to combat antibiotic-resistant pathogens. They have found extensive applications in the fields of medicine, food, and agriculture. However, efficiently screening AMPs from natural sources poses several challenges, including low efficiency and high antibiotic resistance. This review focuses on the action mechanisms of AMPs, both through membrane and non-membrane routes. We thoroughly examine various highly efficient AMP screening methods, including whole-bacterial adsorption binding, cell membrane chromatography (CMC), phospholipid membrane chromatography binding, membrane-mediated capillary electrophoresis (CE), colorimetric assays, thin layer chromatography (TLC), fluorescence-based screening, genetic sequencing-based analysis, computational mining of AMP databases, and virtual screening methods. Additionally, we discuss potential developmental applications for enhancing the efficiency of AMP discovery. This review provides a comprehensive framework for identifying AMPs within complex natural product systems.

Graphical abstract

Highlights

• •Antimicrobial peptides derived from natural products are screened in diverse methods.
• •Summary of the action mechanism involved in the antimicrobial peptide.
• •Antimicrobial peptide shows significant potential application in multiple fields.

1. Introduction

The overuse of antibiotics and the rise of multidrug resistance (MDR) in bacterial pathogens have accelerated the emergence of superbugs, generated substantial public health costs, and posed an increasingly severe threat [1,2]. According to the World Health Organization (WHO), diseases caused by multiple resistant bacteria (MDA) account for up to 700,000 deaths annually [3]. Projections suggest that by 2050, bacterial infections could result in 10 million deaths per year, surpassing cancer fatalities and causing economic losses exceeding 1 trillion USD [4,5]. Since their discovery in 1928 and development for clinical use in the 1940s, antibiotics have saved millions of lives and been integral to various critical medical procedures. However, in recent decades, there has been a notable lack of newly approved antibiotics with innovative therapeutic approaches [6]. Only a few novel classes have been identified and authorized for clinical testing [7].

Antimicrobial peptides (AMPs), often found in natural products, are crucial components of the innate immune systems of both invertebrates and vertebrates. These small-molecule peptides, typically composed of fewer than 100 amino acids, exhibit a broad spectrum of antibacterial activities, superior water solubility, excellent heat stability, and low immunogenicity [8]. Notably, AMPs can damage cell membranes through different modes, and can also centralize their activity within cells, making them promising alternatives to traditional antibiotics [9,10]. Numerous AMPs, including daptomycin [11], omigan [12], and pexiganan [13], have undergone preclinical and clinical trials [14]. Due to their potent antibacterial activity, AMPs are also referred to as “antibacterial peptides (ABPs)” or “peptide antibiotics”, and they can also effectively target fungi, viruses, pathogens, and certain tumor cells [15,16]. For clarity, all these peptides will be referred to as “AMPs” in this review.

The Antimicrobial Peptide Database (APD) is a pioneering resource that includes a tool for calculating the amino acid composition of AMPs from various families. It archives data on 3,569 peptides from six distinct life domains. These data sources encompass 380 experimentally isolated or predicted bacteriocins and peptide antibiotics from bacterial origins, 5 from archaea, 8 from protists, 25 from fungi, 371 from plants, and 2,600 from animals, including both genomically predicted and synthetic peptides. Synthetic AMPs are less abundant than those originating from natural products due to the limitations of chemical preparation techniques, like solid phase peptide synthesis and solution phase synthesis, which can lead to incomplete coupling, deprotection issues, byproduct accumulation, peptide aggregation, and prolonged reaction times [17]. In contrast to synthetic AMPs, natural AMPs, as secondary metabolites from certain organisms, offer outstanding advantages such as distinct structures and multifunctional groups [18]. Since the discovery of cecropin A in moth haemolymph by Boman and co-workers 19 in 1980, the variety of AMPs from natural sources, including animals, microorganisms, plants, and marine organisms, has grown steadily [20,21].

Extracting pure natural peptides from biological sources often results in poor yields and high costs, limiting their commercial viability [22]. Synthetic peptides are used to generate standard concentration curves for quantifying native peptides in crude extracts or fractions. However, isolating AMPs from complex natural resources is challenging, hindering the examination of their main structures, biological activities, mechanisms of action, conformational behaviors, and commercial viability. Thus, rapid screening of AMPs in natural sources has become an important research focus.

A search of relevant publications over the past 20 years (2003–2023) on PubMed, yielded 434 results for “Antimicrobial peptide screening” (Article type: Review papers). Various structures, bioactivities, and potential applications of AMPs have been summarized in several reviews. For instance, Blondelle et al. [23] compiled diverse libraries leading to de novo AMP sequences and recent structural information and optimization procedures to enhance peptide selectivity. Zou et al. [24] outlined the strategies for screening, identifying, purifying, and characterizing AMPs, and documented the novel AMPs. Wu et al. [25] highlighted the recent advances in peptide prediction using machine learning to identify potential active AMPs, anti-cancer peptides, and anti-inflammatory peptides.

Although numerous methods for screening AMPs from natural products have been reported, a comprehensive overview of their discovery and screening methods has rarely been mentioned before. Despite some AMPs advancing to clinical trials and other applications, inherent defects in natural AMPs have impeded their development. Therefore, to overcome these drawbacks, further research on the mechanisms of action of AMPs is necessary to enhance their utility in various fields. This review describes the progress made in discovering and screening methods of natural AMPs, aiming to provide a scientific reference for their developments and applications. The mechanisms of action of biofilms and reviews of the methods of screening AMPs from complex natural products will be introduced. In addition, this review will also address the limitations and prospects of natural AMPs.

2. Mechanism of action of AMPs

The fundamental structural and physicochemical attributes of AMPs significantly influence their interactions with bacterial entities, thereby shaping their antimicrobial properties and stability [26]. AMPs activity can be categorized into membrane-based and non-membrane-based mechanisms. Among them, the binding of AMPs to the cell membrane is generally considered to be the key and initial step in exerting their activity. Regardless of its operational principle, persistent adhesion between AMPs and cellular membranes remains consistent across different operational principles [27]. Specific receptors on bacterial membranes are not always necessary for these interactions. Furthermore, understanding the mechanisms of AMPs activity provides vital information and a substantial basis for effective, fast, and easy separation [28].

2.1. Mechanism of action at membrane

AMPs exert antimicrobial effects primarily by interacting with bacterial cell walls or membranes. Typically, AMPs possess a positive charge that allows them to bind electrostatically to the negatively charged cell membrane, resulting in the formation of transmembrane pores. This leads to metabolite outflow, cell destruction, and ultimately bacterial cell death [29]. Various membrane action mechanisms have been identified, including barrel-stave model, carpet model, toroidal pore model, aggregate model, electroporation model, ion carrier model, and irregular toroidal model. In the following part, four common models will be discussed. These mechanisms, including the barrow-stave model, toroidal-pore model, carpet model, and aggregation model, are illustrated in Fig. 1.

Fig. 1.

Membrane and non-membrane mechanisms of antimicrobial peptides (AMPs): (A) barrow-stave model, (B) toroidal-pore model, (C) carpet model, (D) aggregation model, (E) RNA transcription, (F) DNA replication, (G) protein folding, (H) mitochondria, and (I) cell division.

In the barrel-stave model, as the concentration of AMPs bound to the cell membrane increases, aggregation and conformational changes occur, causing the arrangement of local phospholipid head groups to bundle. The hydrophobic region of the AMP binds to the phospholipid bilayer, forming a hollow luminal structure (Fig. 1A). Both α-helical peptides and β-folded peptides utilize this mechanism to align their hydrophobic regions with those of membrane phospholipids [30,31]. For example, alamethicin employs this paradigm to perform pore-forming action [[32], [33], [34]].

In the toroidal-pore model, AMPs are vertically adsorbed to the cell membrane via electrostatic interactions. The hydrophobic region displaces phospholipids, causing the cell membrane to continuously bend inward until a ring pore is formed (Fig. 1B). Such pore formation leads to the death of the cell. Typical examples include magainin 2, lacticin Q, melittin, and arenicin [35,36]. The hydrophilic portions of the AMPs and the phospholipids work together in this scenario to generate the inner surface of the pore, whereas the barrel-stave model consists of the hydrophilic regions of the AMPs accumulating on the inner surface of the pore [37,38].

In the carpet model, AMPs cover the phospholipid surface like a carpet, neither forming channels nor inserting peptides into the membrane hydrophobic center in a surfactant-like manner, causing membrane disruption (Fig. 1C). As the number of AMPs present on the cell membrane increases to a certain level, the membrane turns unstable, resulting in the loss of integrity. Human cathelicidin LL-37 and AMPs with β-sheet structures exhibit their activity through this mechanism [39]. Consequently, other AMP molecules penetrate the membrane, which causes the bending of the lipid bilayer, membrane rupture, intracellular contents leakage, and ultimately bacterial cell death [40].

In the aggregate model, the cationic AMPs bind to the polysaccharides or peptidoglycans located outside the anionic cytoplasmic membrane, forcing peptides and lipids to form peptide-lipid complex micelles (Fig. 1D). The aggregation enhances the susceptibility to phagocytosis, thus increasing the occurrence of a cell-killing mechanism [41,42]. Thanatin, a 21-amino acid peptide isolated from the gut of Podisus maculiventris, is a representative of this model [43].

Until now, the most efficient screening methods for AMPs components from natural products have evolved from the interaction of AMPs with membranes. All those mechanisms utilize the AMPs to break the stability of membranes and then lead to cell death. The investigation of the action mechanism of AMPs will offer insightful informations for the creation of innovative antimicrobial agents.

2.2. Mechanism of non-membrane action

Besides membrane penetration and pore creation, AMPs exhibit other mechanisms of action. Recent studies suggest that some AMPs prevent biofilm formation and disrupt existing biofilms through non-membrane actions, categorized as intracellular and extracellular activities. These include the interactions with DNA, inhibition of DNA transcription and replication, suppression of protein synthesis and folding, inhibition of cell wall and membrane formation, inhibition of enzyme activity, and interference with cell division [[44], [45], [46]]. These interactions with nucleic acids and proteins regulate gene transcription, translation, and expression or inhibit specific intracellular protein functions, thereby restraining and killing bacteria, as shown in Figs. 1E−I. For instance, S-thanatin achieves bacterial eradication by inhibiting bacterial respiration and interacting with lipopolysaccharides in vitro and a mouse model of pyogenic shock induced by multidrug-resistant bacteria [47]. Microcin J25, a ribosomally produced and post-translationally modified peptide, binds to the secondary channel of RNA polymerase, inhibiting trigger-loop folding, which is necessary for the RNA polymerase to catalyze processes efficiently. As a result, it can stop substrates from entering this channel, thus suppressing RNA polymerase activity [48].

Whether AMPs act via membrane or non-membrane mechanisms, the dual modes contribute to the reduced susceptibility of bacteria to resistance. These mechanisms often work in tandem. However, it is important to note that different mechanisms lead to the realization of the purpose of AMP through different courses of action. The realization of the antimicrobial function of certain AMPs requires interaction with the bacterial cell membrane. Some AMPs can form transient channels on the cell membrane, allowing entry into the intracellular space to exert their effects. Additionally, some AMPs cause cell death by adhering to the cell membrane, without damaging its structure, leading to cell death.

3. Efficient screening methods for the discovery of novel AMPs

In recent years, besides bacteria and their metabolic products, medicinal, herbal, and higher plants have been explored for the discovery of novel AMPs. These plants, which contain unique pharmacologically active polypeptides, have been screened and isolated through drug discovery programs. Unfortunately, most antibiotic screening methods have not progressed significantly since the invention of the traditional Waksman's methodology in the 1930s [49,50]. This method involves the step-by-step preparation of a peptide, followed by column chromatography separation and purification, combined with antibacterial experiment testing of the purified single AMP.

More specifically, the most popular method for identifying and purifying active ingredients in natural products is still the bioassay-guided fractionation (BGF) strategy. This approach determines bioactivity before identifying and purifying the active components. However, BGF has inherent drawbacks, such as being excessively time-consuming, costly, and operationally cumbersome [51]. In addition, its repeated extraction steps can result in the loss of activity and degradation of trace amounts of active peptides, finally limiting the development of AMPs from natural sources. Therefore, there is a pressing need to develop simple, rapid, and efficient methods for the discovery and screening of AMPs from natural products and extracts.

Various novel screening methods have emerged to overcome those limitations, which include whole-bacterial adsorption binding, cell membrane chromatography (CMC), liposome chromatography, capillary electrophoresis (CE), and colorimetric assays. They all focus on membrane-mediated mechanisms. Additionally, other methods, such as thin layer chromatography (TLC), colorimetric assays, fluorescence screening, genetic sequencing-based analysis, computational mining of AMP databases, and virtual screening methods have been developed. These methods utilize various physicochemical and structural parameters related to antimicrobial activity to screen and grade peptides. Subsequently, the resulting peptide sequences are then assessed and analyzed to identify the desired AMPs.

3.1. Whole-bacterial adsorption binding

Most AMPs initiate steps to their antimicrobial action by binding to bacterial cells. Bacterial adsorption binding depends on the leverage of the electrostatic interactions between AMPs and bacterial cell membranes [52]. This method involves incubating AMPs with whole bacteria and then comparing the results of liquid chromatography (LC) (specifically gel filtration chromatography) before and after incubation. Peaks that decrease or disappear after incubation are identified as potential AMPs. Subsequently, their structures are determined using mass spectrometry (MS). A schematic of the mechanism is shown in Fig. 2. Fei [53] used LC-MS to compare the spectra before and after incubation with Escherichia coli (E. coli) extract, and then collected the disappeared peaks. Samples were diluted to 1 mg/mL before and after incubation with E. coli. The antimicrobial assay showed that the disappearance of AMPs binding successfully suppressed the development of the five bacterial infections. The sequences of these AMPs were identified as KPPLNGPAL, QPVYHWYWH, VVTSSSLP, RMRLLLRRKGGQ, and RVLLLPAFAEK. Through database comparison, all of these were discovered for the first time. Tang et al. [54] discovered MDpep9, a novel peptide (sequence KSSSPPMNH), from housefly larvae. This peptide exhibited preferential binding to bacterial cell extract and facilitated in vitro membrane permeation. In a similar approach, Hao et al. [55] utilized E. coli extracts to screen AMPs derived from trypsin hydrolyzate of yak blood. Two novel AMPs (YakB-1 and YakB-2) were effectively isolated by repeated purification using reverse-phase high performance liquid chromatography (RP-HPLC), followed by an evaluation of their antibacterial activity. The whole-bacterial adsorption binding method required only several microliters of peptide solution at concentrations of several milligrams per milliliter, reducing the screening process time to less than 30 min [53]. This method, based on the membrane-action mechanism of AMPs, can be applied to peptide mixtures, enabling simultaneous co-incubation experiments with different peptide compositions.

Fig. 2.

The schematic scheme of screening antimicrobial peptides (AMPs) by whole-bacterial adsorption binding method with liquid chromatography-mass spectrometry (LC-MS).

Furthermore, in this method, the location of AMP peaks can be identified by comparing differences in liquid phase peaks before and after crude peptide incubation with whole bacteria. Compared to traditional methods, this approach offers several advantages, including scalability, shorter processing times, and lower costs, making it highly suitable for straightforward applications. However, these methods have limitations in sensitivity and capacity. Complex bacterial cells may adsorb other protein components, leading to increased nonspecific adsorption and difficulties in subsequent separation [56,57]. These challenges may explain why this method has been less frequently reported in recent years.

3.2. CMC

CMC is a biomimetic method that mimics cell membrane interactions by binding active tissue cell membranes to specific carrier surfaces, creating a stationary phase [58]. This technique utilizes chromatographic methods to investigate the interaction patterns between drugs and receptors [59]. When AMP interacts with cell membranes, the AMP components are quickly identified and separated by combining CMC with a HPLC-MS [60]. Benesch et al. [61] explored a covalent method attaching PC amide to silica surfaces, evaluating its effectiveness for immobilizing artificial membranes in AMP screening by reversed-phase HPLC. Four AMPs, including the GS-14 (peptide sequence: Cyclo-(VOLdFPVOLdFP))] derived from natural products, were evaluated by comparing this covalent method with noncovalent columns. They found stronger AMP binding on covalent columns, particularly benefiting from cationic residues. Xiao et al. [62] loaded extracts of Jatropha curcas polypeptide (20 μL, 5 mg/mL) onto a cell membrane affinity chromatographic column. The cell membrane of E. coli was isolated and an off-line tandem technique for CMC and LC-MS was developed. The cationic JCpep7 (peptide sequence: KVFLGLK) was identified from the seeds of Jatropha curcas, demonstrating its broad-spectrum antimicrobial activity through membrane disruption. Similarly, Xiao et al. [63] developed a cell membrane affinity extraction-off-line lLC time-of-flight MS (LC-TOF-MS) method to screen AMP from the Jatropha curcas proteolytic hydrolyzate. Finally, a cationic AMP (JCpep8, peptide sequence: CAILTHKR) was isolated and evaluated by comparing HPLC spectra and finally identified by TOF-MS. JCpep8 presented antibacterial activity against E. coli, Shigella dysenteriae (S. dysenteriae), Pseudomonas aeruginosa (P. aeruginosa), Staphylococcus aureus (S. aureus), Bacillus subtilis (B. subtilis), and Streptococcus pneumoniae (S. pneumoniae), as well as presented relatively strong thermal stability and minimal hemolytic properties. Tang et al. [64] applied CMC to extract Cpep11, from bovine casein. Initially, to produce an affinity binding medium, Saccharomyces cerevisiae cell membranes were fixed onto the silica gel surface. The fraction of membrane binding could be identified by contrasting the hydrolyzate's pre- and post-adsorption fingerprint chromatograms from RP-HPLC with the adsorption medium. Ultraperformance liquid chromatography-matrix-assisted laser desorption/ionization quadrupole TOF tandem MS (UPLC-MALDI-Q--TOF-MS) was used to determine the amino acid sequence of the peptide Cpep11, which came out to be LRLKKYKVPQL. The sequence matched amino acid residues 99–109 of bovine α(s1)-casein. Wang et al. [65] combined CMC with live bacterial adsorption methods to screen AMPs from Moringa oleifera seeds. Five AMPs were also identified by LC-MS/MS, of which MCNDCGA peptide (called MOp3) was identified and showed the most significant inhibitory activity against S. aureus, causing irreversible damage to the cell membrane. CMC screening typically involves processing 10 μL peptide volumes at concentrations of hundreds of micrograms per milliliter, with a screening time of approximately 1 h (plus up to 2 h for affinity adsorption) [64]. This method leverages the specific affinity of AMPs for bacterial membranes, making it suitable for screening complex peptide mixtures even when their detailed compositions and concentrations are unknown.

Despite the advantages such as efficiency, sensitivity, and reproducibility, CMC screening methods require AMPs to be collected on a specific carrier surface, and the preparation process is complex. In addition, during the continued screening process, cell membranes could become inactive and lose their function, thus resulting in a shorter lifespan and comparatively higher costs. Moreover, the high demand for cell membranes and the relatively low capture of AMPs limits its large-scale application. The drug discoveries of natural products based on CMC are summarized in Table 1 [[61], [62], [63], [64], [65]].

Table 1.

Summary of screening antimicrobial peptides (AMPs) from natural products based on cell membrane chromatography (CMC).

Year	Source of natural products	Screened natural productsa	Antimicrobial active strains	Refs.
2010	Gram-positive bacteria Bacillus brevis	GS-14	Phospholipid bilayers of cell membranes (no antibacterial activity evaluation test)	[61]
2011	Jatropha curcas	JCpep7	S. typhimurium, P. aeruginosa, S. dysenteriae, S. aureus, B. subtilis, and S. pneumoniae	[62]
2012	Jatropha curcas meal protein	JCpep8	E. coli, P. aeruginosa, S. dysenteriae, S. aureus, B. subtilis, and S. pneumoniae	[63]
2015	Casein hydrolyzate	Cpep11	S. dysenteriae, E. coli, S. typhimurium, B. subtilis, S. aureus, and S. pneumoniae	[64]
2022	Moringa oleifera seeds	MOp3	S. aureus	[65]

^aGS-14, JCpep7, JCpep8, Cpep11, and MOp3 are the name of different peptides.

S. typhimurium: Salmonella typhimurium; P. aeruginosa: Pseudomonas aeruginosa; S. dysenteriae: Shigella dysenteriae; S. aureus: Staphylococcus aureus; B. subtilis: Bacillus subtilis; S. pneumoniae: Streptococcus pneumoniae; E. coli: Escherichia coli.

3.3. Phospholipid membrane binding

Bacterial cell membranes are complex structures composed of lipid bilayers and embedded proteins (Fig. 3). The lipid bilayer serves as the foundational framework, primarily composed of phospholipids such as phosphatidylcholine (PC), phosphatidylethanolamine (PE), phosphatidylglycerol (PG), cardiolipin (CL), and others [[66], [67], [68]]. AMPs exert their antimicrobial effects through specific interactions with these phospholipid components, particularly PG and CL. Phospholipid membrane chromatography mimics artificial cell membranes to study the AMPs-phospholipid membrane interactions and screen AMPs by different methods. Tang et al. [69] used egg-yolk PC (EYPC) and dimyristoyl-sn-glycero-PG (DMPG) as monomers to create an artificial membrane mimicking the stationary phase and demonstrating AMP interactions. They employed solid-phase extraction and HPLC to effectively screen a novel Opep12 (peptide sequence: RVASMASEKMKI) from egg albumin hydrolyzates, showing broad-spectrum antibacterial activity against both Gram-positive and Gram-negative bacteria. Joo [70] employed immobilized lipid affinity capture (ILAC) with LC-MS to replicate the phospholipid content of Gram-positive bacterial membranes immobilized on magnetic particles. ILAC effectively collected a novel AMP, hominicin, from the peptide pool of Staphylococcus hominis MBBL 2–9. Tang et al. [71] employed liposomes S. aureus in an equilibrium dialysis to purify a protein hydrolysate from anchovy (Engraulis japonicus) cooking wastewater by liposome equilibrium dialysis combined with HPLC. After enzymatic hydrolysis with antimicrobial activity, the purified AMP (ACWWP1) was identified as GLSRLFTALK. For rapid assessment of antimicrobial activity effects of AMPs, microbial-bound bilayer lipid membranes (tBLMs) combined with electrical impedance spectroscopy was developed as an efficient method. Alghalayini et al. [72] developed solid matrix-supported phospholipid bilayers to model AMP-lipid interactions, highlighting tBLMs technology for studying the effects of various antimicrobial agents on lipid bilayers. These methodologies illustrate how AMP-phospholipid interactions can be studied comprehensively, enhancing the understanding of antimicrobial mechanisms and facilitating the discovery of novel AMPs. Tang et al. [73] introduced an innovative method for screening AMPs from boiled-dried anchovies using an immobilized bacterial membrane liposome chromatography method. A new cationic AMP called Apep10 with cell membrane contact activity was successfully extracted. This method extends the concept of CMC and employs biomimetic phospholipids to simulate biological cell membranes. It overcomes challenges associated with obtaining large quantities of biological membranes and reduces issues related to decreased membrane activity. By mimicking biological membranes with biomimetic phospholipids, this method enhances sensitivity, allowing screening peptide mixtures at concentrations in the milligram per milliliter range with volumes as small as tens of microliters. The screening process itself takes less than 1 h, making it rapid and efficient [73].

Fig. 3.

Schematic diagram of antimicrobial peptides (AMPs) screening based on the phospholipid membrane method, including a fixation on the chromatographic column and magnetic particle carrier. LL-37, Opep2, MDpep9, and YakB-1/YakB-2 are the name of different peptides.

The principle of this screening method is based on the specific affinity of AMPs for phospholipid membranes, which are crucial components of cell membranes. This affinity drives the screening based on the membrane action mechanism of AMPs. Importantly, this method allows simultaneous screening of multiple components from peptide mixtures even when the concentration of individual components is unknown. Therefore, the method has plenty of advantages such as simplicity, high sensitivity, specificity, and improved stability compared to traditional methods. However, there are still important differences between phospholipid membrane chromatography and real active cell membranes. The availability of commercial phospholipids limits the biomimetic and specific properties of these artificial cell membrane materials.

3.4. Membrane-mediated CE

As the mechanism of the antimicrobial action of AMPs is its interaction with bacterial cytoplasmic membranes, the evaluation of in-solution interactions between membranes has also become a possible way to develop AMP screening methods. CE based on phospholipid membranes has the potential applications as a screening method. CE is a micro-separation method for polypeptides based on the high correlation between their effective charge and the relative molecular mass ratio to ionic mobility. The critical principle of CE analysis is still based on the AMP and membrane interactions. To establish a CE analytical model that simulates the binding of cell membranes to bacteria, screening was based on the different interaction effects (mainly electrostatic interactions) between the two. Xiao et al. [74] examined the binding process of JCpep8 to bacteria using live S. aureus cells as a pseudo-stationary phase in capillary electrochromatography. The thermodynamic parameters and binding constants were investigated, and electrostatic and hydrophobic effects were proved to have a major effect on the binding interaction. In addition, antimicrobial activity kinetics and outer and inner membrane destruction were studied by time-dependent experiments. It was found that the JCpep8 killed microorganisms mainly through a wall/membrane-targeted pore-forming mechanism. This antimicrobial process provides a perspective for the screening of AMPs. In addition, to obtain a relatively long lifetime for efficient and repeatable research, Tang et al. [75] developed liposomes constructed from E. coli membrane lipids as a pseudo-stationary phase in capillary electrochromatography, which facilitated the valuation of a relatively long lifetime with Apep 10. Liposomes were used as model membranes because of their structural similarities between liposomes and natural cell membranes (both of them are phospholipid bilayers). For comparison, mixed phospholipid liposomes were also used as a pseudo-stationary phase, and the results of combining the constant and capacity factors showed that the former exhibited more biomimetic properties. An in-depth knowledge of the interactions between AMPs and native bacterial membrane lipids may be helpful for the selection and designation of AMPs. Dufort-Lefrancois [76] explored the mechanism and found a binding constant between indolicidin and the bacterial receptor-lipopolysaccharide. The interaction of sphingomyelin with two AMPs, namely indolicidin (peptide sequence: ILPWKWPWWPWRR) and indolicidin 45 (peptide sequence: ILPWKWPWAPARR), were evaluated in their work using affinity CE. Two AMPs were injected into the capillary with sphingomyelin [77]. It was found that the indolicidin and indolicidin 45 exhibited different migration times, which indicated the interaction of sphingomyelin with the AMPs. Then, Parihar [78] explored the hydrophobicity of the AMPs indolicidin and indolicidin 45 via the above CE methods. Moreover, bacterial membrane structures have been established for their convenience and low cost. Mills and Holland [79] evaluated the lipophilicity of polypeptides by bilayer micellar electrokinetic capillary chromatography to screen membrane affinity AMPs. A high concentration of surfactants above the critical micelle concentration was added to the background electrolyte, and a discrete and modifiable bilayer was formed to provide a plasma bacterial membrane, which is similar to the surface profile, thus the degree of membrane partitioning led to different AMP retention behaviors. The secondary structure of AMPs drives membrane insertion. The capacity factor indicates membrane affinity, and the long retention time correlates with better lipophilicity. Three cationic AMPs, i.e., indolicidin, melittin, and magainin 2, were evaluated by the membrane-mediated CE, and the rapid estimation of their lipophilicity and affinity membrane was achieved. The influence of high surfactant concentrations on the MS analysis could be further addressed to make it more accessible for detailed structural analysis functions.

Another potential screening method using CE relies on the principle of the AMP mechanism related to the effective charge of the AMPs. Tůmová et al. [80] evaluated 12 cationic AMPs with known sequences, containing over three variable amino acids. Physical-chemical parameters including effective charge by capillary isokinetic electrophoresis and the ion mobility of AMPs were examined by capillary zone electrophoresis. The charge value is an important parameter for estimating the strength of the electrostatic interactions between AMPs and the bacterial membrane, revealing the mechanism of action of drugs against pathogens. CE offers the advantages of high efficiency, good resolution, excellent reproducibility, and high speed for the identification of the interactions between AMPs and bacterial cell membranes. While the technology is in its infancy, it may not meet the mature requirements for screening AMPs from natural products and may also lead to false negative and false positive results. Therefore, further development is required for AMP screening. Šolínová et al. [81] employed the converted CE method, namely the free-flow zone electrophoresis, for the AMP dipeptide β-alanyl-tyrosine (β-Ala-Tyr) isolation from the extract of butterfly larvae after the RP-HPLC pre-separation. This dipeptide exhibited antimicrobial activity against both Gram-negative and Gram-positive bacteria, as well as antifungal properties. Meanwhile, by transferring the CE methods into the free-flow zone electrophoresis method, the separation concentration of AMPs has been greatly improved from 1 mg per milliliter [80] up to tens of milligrams per milliliter [81], and separation capacity increased from 0.2 μL per run (less than 30 min) [80] up to tens of milligrams per hour of continuous separation [81].

The membrane-mediated CE method is mainly based on the mechanism of AMPs on bacterial membranes. This method offers several advantages, including high screening efficiency, good reproducibility, short analysis time, high separation efficiency, rapid operation, and high sensitivity. However, there are drawbacks, such as missed or erroneous detections. The theoretical foundations of this technique require further exploration. Consequently, these strengths, and notable limitations, have primarily been utilized for assessing binding to bacterial cell plasma membranes. Furthermore, each AMP must undergo individual analysis for screening purposes. However, further development is needed to effectively screen complex natural product samples for active AMPs.

3.5. TLC

TLC is an operable and visual method for AMP screening [82], which combines TLC for the separation of antimicrobial substances such as AMPs, polysaccharides, phospholipids, etc.. As a separation method, TLC can be used in combination with MS or nuclear magnetic resonance (NMR) for in-situ or off-line structure identification, although it seems difficult to connect TLC with antimicrobial activity evaluation and drug screening modules. Bioautography, developed in 1946 [83], has been widely applied for biological detection. TLC combined with bioautography has been applied for active substances in complex mixtures since 1961 [84,85]. TLC-bioautography can be performed in an agar culture-medium, such as agar contact diffusion bioautography and agar overlay bioautography. It could also be used in nutrient broth, like TLC-direct bioautography, which means the mixture was separated by TLC, the microbials were sprayed or pressed close to it on a thin layer, and after cultivation and treatment with an appropriate color reagent, the antimicrobial components were screened. Shetty et al. [86] separated the metabolite AMPs compounds produced by Streptomyces parvulus (S. parvulus) using TLC with dichloromethane-ethyl acetate. The inoculated nutrient agar was placed under TLC plates for diffusion, followed by agar contact bioautography. The antimicrobial activity was detected by the cup plate method, which was performed in sterile Petri dishes, and purified peptides were added for incubation and evaluation. Three TLC bioautography methods were evaluated by Grzelak et al. [87] to screen natural anti-tuberculosis peptides from actinomycete extracts, contacts, agar-overlay, and direct bioautography. In the contacts method, a TLC plate with separate compounds is dipped into the bacterial suspension before incubation. However, this method requires a high degree of precision in the contact operation. In the agar-overlay method, a TLC plate with separated compounds is placed on a Petri dish, and overlaid with agar inoculated with bacteria before incubation. The compounds were then directly screened by bioautography and their structures were identified by TLC-MS and TLC-NMR. The agar-diffusion bioautography method is suitable for polar and pure substances, whereas the direct method matched almost all types of substances and presented the highest sensitivity. Ecumicin, a new anti-tuberculosis macrocyclic tridecapeptide, appeared to have selective anti-tuberculosis activity. Besides, Jaskiewicz et al. [88] described a rapid AMP screening method using TLC. The crude peptides were separated on a TLC plate, and the TLC direct bioautography was then employed before incubation. After treatment of the TLC plates with resazurin for cell staining, the potential AMPs showed yellow spots for autography. CAMEL (KWKLFKKIGAVLKVL-NH₂) and two lipopeptides, Pal-KK-NH₂, and Pal-KKK-NH₂, were acquired utilizing this method, and showed high antimicrobial activity. Ramya et al. [89] discovered the antimicrobial compounds from sea slugs. The compounds extracted from the sea slugs were uniformly applied to the plate. After spraying with the visualizing reagent (ninhydrin reagent), purple to pink spots on the plate indicated the positions of the amino acids and peptide compounds.

The TLC-bioautography method screens peptide mixtures by first separating them via chromatography on plates, followed by in-situ co-culturing with bacteria. During this process, peptides exhibiting antimicrobial activity interact with microbes through various mechanisms. Dyes are subsequently used to indicate changes (or lack thereof) in microbial metabolites. This method operates on the principle that AMPs exert a bacteriostatic effect, resulting in a distinct dye color compared to the control. Hence, it is suitable for screening both membrane and non-membrane-active AMPs based on their respective mechanisms of action. This method has the advantages of simplicity, rapidity, operability, and visibility for the discovery of antimicrobial active peptides from natural product mixtures. TLC combined with MS or NMR allows the structural identification of AMP components. Natural product peptide extracts are typically found at concentrations of milligrams per milliliter, with capacities ranging from tens of microliters to several milliliters. However, the process of separating natural product mixtures and evaluating their antimicrobial properties required several hours [90]. Despite its utility, the TLC method faces challenge due to limitation in the concentration range, making it difficult to meet industrial requirements, particularly in terms of establishing calibration with standard curves. This method was designed to analyze crude extract samples, which could be further developed for the screening of components in complex natural systems.

3.6. Colorimetric assays

Spectroscopic analysis of microbial activity for AMP screening is a widely employed indirect method for screening AMPs, focusing on spectral changes in target components. This approach aids in assessing the bacterial state in the presence of AMPs, analyzing microbial virulence, comparing proliferation rates, and enumerating microbes. As a result, for high-throughput screening of antimicrobial agents from natural products, colorimetric assays and fluorescence methods are commonly employed in spectroscopy due to the accessibility and ease of operation of these instruments.

The colorimetric method relies on the power reduction of living microbes' dehydrogenases, particularly those in mitochondria, to quantitatively assess microbial viability. This is achieved by detecting changes in spectral features (brightness, saturation, etc.) following substrate reduction. However, due to the short half-life of AMPs in complex environments, overnight incubation can potentially underestimate biological activity. Traditional long-term liquid medium cultures may also struggle to provide accurate results. Therefore, the colorimetric method, employing reducing substrate dyes to monitor microbial activity via metabolic pathways, is widely favored for AMP screening. It allows for the rapid determination of antimicrobial activity. There are two main types of representative dyes as reducing substrates, including tetrazolium-based dyes, such as 2,3-bis[2-methoxy-4-nitro-5-sulfophenyl]-5-[(phenylamino)carbonyl]-2H-tetrazolium hydroxide (XTT), 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT), p-iodonitrotetrazolium violet (INT), and 3-(4,5-dimethylthiazol-2-yl)-5-(3-carboxymethoxyphenyl)-2-(4-sulfophenyl)-2H-tetrazolium (MTS), etc.. Tetrazolium-based dyes can be reduced by living bacteria to form formazan, and different tetrazolium salt derivatives as substrates will produce formazan with different colors. Other cost-effective and universal dyes are resazurin and resazurin-based reagents, including PrestoBlue and AlamarBlue (Fig. 4). Blue resazurin can be irreversibly reduced by living bacteria to pink resorufin with red fluorescence. Therefore, it can be used to evaluate the toxicity, antimicrobial activity, and screening of antimicrobial components.

Fig. 4.

Representative reducing substrate dyes of colorimetric screening method: (A) tetrazolium-based dyes and (B) resazurin-based dyes. XTT: 2,3-bis[2-methoxy-4-nitro-5-sulfophenyl]-5-[(phenylamino)carbonyl]-2H-tetrazolium hydroxide; MTT: 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide; MTS: 3-(4,5-dimethylthiazol-2-yl)-5-(3-carboxymethoxyphenyl)-2-(4-sulfophenyl)-2H-tetrazolium; INT: p-iodonitrotetrazolium violet.

MTT, a traditional positive-charged tetrazolium salt with a yellow color in an aqueous solution, is reduced by dehydrogenases in living bacteria to form purple insoluble formazan. It offers the advantage of low background interference [91]. MTT was described for evaluating antimicrobial inhibitors in the 1980s [92,93], and MTT has since been widely used for AMP screening [94]. Pannecouque et al. [95] elaborated the protocol and applied MTT colorimetry in the screening of anti-viral compounds, which remains of interest due to its ability to provide detailed information, inactivate the virus thereby protecting the experimenter from infection risk, and low cost (even though the organic solvent is especially required in the methods). XTT, an improved tetrazolium salt of the first generation, produces orange-yellow formazan with a maximum absorbance at 450 nm [96]. While XTT has better solubility compared to MTT and can be measured without dimethyl sulfoxide (DMSO) dissolution, it is unstable and challenging to store. In 1988, Vince et al. [97] used XTT to evaluate drug activity, demonstrating its reduction to formazan dye in the presence of hydrogenases and its application in screening nucleoside analogs with potential antiviral activity. The XTT method, following the Vince's protocol, is employed for screening AMPs from natural products. Gustafson et al. [98] discovered the macrocyclic peptides circulins A and B from Chassalia parvifolia extracts as novel human immunodeficiency virus (HIV) inhibitors, utilizing XTT for cytotoxicity detection and virus-induced in-vitro sterilization. Recently, the XTT colorimetric assay method has been employed widely to evaluate microbial growth-related properties during AMP screening. Mechesso et al. [99] screened the active ultrashort peptides KR-8 and RIK-10 among engineered LL-37mini peptides by developing a diluted Mueller-Hinton broth medium to observe the antimicrobial and antibiofilm activity. XTT was employed to estimate the inhibition of microbe attachment or biofilm formation by detecting the attachment on plates or in biofilm viable cells after washing the medium. Besides, Boaro et al. [100] applied XTT in cytotoxicity assays to evaluate the activity of wasp venom-converted AMPs. In addition, INT, as a tetrazolium-based reagent, is commonly used to detect lactate dehydrogenase leaking from cells, and can form a purple formazan dye upon reduction. Ogbole et al. [101] screened peptides, including cyclic peptides derived from Nauclea diderichii Merrill and Ixora brachypoda DC, showing inhibitory activity against bacteria, fungi, and B and C type enteroviruses. Both antimicrobial and antifungal detection was performed by INT-based colorimetric redox detection later. INT colorimetric assay was commonly employed in AMP screening. However, INT may not always detect all bacterial inhibitions by control drugs. The MTS reagent has also been developed for AMP screening as it offers higher sensitivity and requires a one-step reaction [102]. Mohammadi et al. [103] evaluated the anti-leishmanial activities of three AMPs (piscidin originating from tilapia, dicentracin-like peptide from the hippocampus, and moronecidine-like peptide from Asian sea bass fish) by measuring their action on the major Leishmania promastigotes and their cytotoxicity against cells using the MTS-based method. Piscidin displayed significant activity against major Leishmania promastigotes and dicentracin-like and moronecidine-like peptides, killing major Leishmania and producing anti-Leishmania factors.

Resazurin has also been commonly used in AMP screening apart from the tetrazolium-based colorimetric reagents [104]. It is water-soluble and can be reduced by oxidoreductases and dehydrogenases to form the colored fluorescent resorufin. Colorimetric methods respond rapidly to natural peptides within seconds, offering high-throughput capabilities for evaluating cell viability and determining minimum inhibitory concentration (MIC). These methods exhibited the advantages of simplicity and high throughput, which present potential and compatibility with automated operations using 96-well plates. Despite their advantages, these methods provide limited information on the interaction between AMPs and microbes and require time-consuming pre-testing of microbial cultures. However, they remain widely used for determining antimicrobial activity and cytotoxicity, often serving as a common screening method for AMP discovery. Coban et al. [105] modified the traditional broth microdilution method into a resazurin-based colorimetric antibiotic sensitivity test for the early rapid detection of vancomycin (glycopeptide)-resistant enterococci infection. Resazurin, therefore, was employed for the fast evaluation of microbial growth. This research inspired the development of efficacy evaluation approaches for other AMPs. Karyne et al. [106] investigated the in-vitro activity of the bee venom peptide melittin to different strains of Acinetobacter baumannii using resazurin to evaluate MIC. The extreme drug-resistant strains were found to be susceptible to melittin. While these methods need to be further improved to accommodate screening requirements, detecting multiple pathogens simultaneously is an effective solution. Besides, PrestoBlue and AlamarBlue, as resazurin-based reagents, have also been utilized in the screening of AMPs. Rodriguez et al. [107] used nano LC-MS analysis to sequence peptide segments of the mollusc Nerita versicolor, and screened three potential antimicrobial peptides (Nv-p1, Nv-p2, and Nv-p3) through bioinformatics prediction. Toxicity evaluation of Nv-p1, Nv-p2, and Nv-p3 peptides was carried out by the PrestoBlue-based method. These AMPs exhibited anti-activity against P. aeruginosa and three fungi. Wang et al. [108] designed and expressed a phage-derived recombinant PK34 AMP, in which the microplate-based AlamarBlue assay was employed for MIC estimation. These resazurin-based reagents have been frequently used for MIC evaluation.

In addition, new colorimetric methods for screening AMPs have been developed. Zhao and Sugihara [109] explored the functions of polydiacetylene for the detection of six AMPs including melittin, magainin 2, α-hemolysin, PGLa, LL-37, and human defensin human neutrophil peptide-1 (HNP1). The 10,12-tricosadiynoic acid and 1,2-dioleoylsn-glycero-3-phosphocholine formed polydiacetylene vesicles showed different colors due to the lipid-AMP peptide interactions. The intensity ratio between polydiacetylene peaks at 645 nm (blue) and 545 nm (red) corresponded to the AMP concentrations. The ligand concentration produced a median effect of 50% (EC₅₀) and the Hill coefficient was correlated with the net charges and mechanisms of antimicrobial activity, respectively. In addition, the synergistic effect of the two types of peptides was predicted in this study, which was combined with a 96-well plate, thus presenting the potential application of high throughput screening.

The colorimetric method operates on the bacteriostatic effect of the AMPs, where peptides prevent bacterial reduction of the dye, resulting in no color change. This method is compatible with the AMPs acting through various mechanisms, including those affecting membranes and non-membrane targets. For evaluating biofilm inhibitory concentrations of AMP candidates, concentrations as low as tens of milligrams per milliliter and volumes up to tens of milliliters are used, typically requiring a 48 h incubation period [110]. This method facilitates accurate measurement of peptides at known concentrations and enables simultaneous testing of multiple peptides without the requirement for extensive chromatographic instrumentation, accommodating up to 384 samples with multi-well plates. While the method offers advantages such as simplicity, speed, cost-effectiveness, and efficiency, it also has drawbacks including low sensitivity, limited selectivity, poor repeatability, and limited applicability across concentration ranges.

3.7. Fluorescence-based screening

The fluorescence screening method was developed because of its high sensitivity. The discovery of AMPs relied on the measurement of cell absorption or the release of fluorescent dyes [41,[111], [112], [113]]. Kodedová et al. [114] developed a fast fluorescence 3,3′-dipropylthiacarbocyanine iodide (DiS-C3(3)) assay that enabled the detection of hyperpolarization and permeabilization of the Candida plasma membrane. DiS-C3(3) passed effortlessly through the plasma membrane and bound to cellular components, which was accompanied by a shift in its maximum absorption wavelength and fluorescence intensity changes in permeabilized cells. In combination with a 96-well microplate, this method was employed for the high-throughput screening of new AMPs that can directly target the surface of Candida in less than 2 h. The Lasioglossin peptide LL-III was also identified in the venom of Lasioglossum laticeps bees via this method and the results showed broad-spectrum antifungal activity against several species of Saccharomyces, Osmotolerant yeast, and Candida. Nuti et al. [115] developed a microfluidic system for the production and screening of AMPs, which was employed to evaluate the specificity of AMP membranes. In this system, AMPs were synthesized via a cell-free protein method using double emulsion droplets. The AMPs are then bound to two different artificial monolayer lipid vesicles in different ratios through positive and nonpolar features. The membrane destruction potential of AMPs, along with their antimicrobial activity and cytotoxicity to mammalian cells was determined by measuring the amounts of two self-quenched concentrations of fluorescent dyes 5(6)-carboxyfluorescein and sulforhodamine B, which were released by the rupture of two vesicles types. The alpha-hemolysin from S. aureus, Pneumolysin from S. pneumoniae, meucin-25 from the venom glands of the scorpion Mesobuthus eupeus, and cathelicidin-BF from the venom of the snake Bungarus fasciatus were evaluated in this system and the expected results were obtained. The peptide concentration in fluorescence screening methods can be reduced to several micrograms per milliliter, with volumes typically in the range of tens of milliliters range. This screening process is completed in less than 2 h [114]. Fluorescence methods primarily rely on the interaction between AMPs and cell membranes, making them ideal for screening AMPs that target membrane mechanisms. They are effective for analyzing individual peptides at known concentrations. Fluorescence screening methods offer advantages such as high sensitivity, low sample consumption, and no need for additional transporters. However, they may encounter challenges related to interference factors, limited separation efficiency, and the need for improved stability of fluorescent signals.

3.8. Genetic sequencing-based analysis

Genomics, a transcriptomics-based sequencing method for AMP screening, has been investigated as a cutting-edge method that utilizes the similarity of gene and peptide sequences in the AMP family. It exhibits the advantages of high efficiency and short time consumption [116,117]. In addition, it is obtained through the hydrolysis of proteins (i.e., cytokines and lectins) expressed by other immune genes. Further studies are required to determine the antimicrobial activity and phylogenetic relationships among the different factors.

Yi et al. [118] employed the Fish-T1K Database with transcriptomics for high throughput screening of AMPs and immune peptides. Yi et al. [119] also utilized the transcriptomes of gills from 87 Actinopterygii species for the screening of short nucleotide sequences, and the innate immune defense factors of fish were described via homologous searches and AMP screening. They also analyzed the genomes and transcripts of Boleophthalmus pectinirostris and Periophthalmus magnuspinnatus. A total of 507 AMP transcripts and 449 new sequences were identified. This method relies on the known AMPs genes as templates to predict gene sequences. Finally, hemoglobin β1 and amylin were found to have a high inhibiting activity on Micrococcus luteus. This method efficiently screened AMPs. However, it is exclusively effective for homologous species, and processing such a wide range of datasets requires advanced data processing. Pandi et al. [120] employed deep learning of the genome, metagenomic data, and a cell-free protein synthesis method to discover and rank 500 AMP candidates. The results showed that up to six of the de novo AMPs demonstrated antimicrobial activity against multi-drug resistant bacteria. Deep generative variational autoencoders (VAEs) were trained using peptide sequences from the UniProt database. The pre-trained VAE was then subjected to transfer learning using another dataset with a known AMP to generate an AMP from scratch. Cell-free protein synthesis is a DNA-based biosynthesis method that utilizes the in-vitro transcription translation systems, which have the advantages of high efficiency and high throughput. It took only 24 h for the cell-free protein synthesis and bioactivity tests of the AMPs, with comparisons of 28 days by synthesis and wet lab experiments (including MIC measurement via the broth microdilution method, hemolytic activity, acute in-vivo toxicity analysis, and circular dichroism spectroscopy) according to the report of Das et al. [121]. Pandi et al.'s work [120] also decreased the cost of producing AMPs for screening two strains in parallel to less than 10 USD, excluding the cost of DNA synthesis or use of polymerase chain reaction (PCR) primers. This achievement was attributed to a much higher throughput (a 15-fold increase compared to Das et al.'s work [121]) at similar success rates (12.6% vs. 10%), as well as the fact that this sort of one-pot cell-free protein synthesis pipeline could reduce the price of producing peptides to approximately 1 USD (a 14-fold increase compared to the non-one-pot cell-free protein synthesis system [120]). The next step will focus on designing products with different properties (low drug resistance, systemic usage, etc.), in combination with other machine learning techniques, to develop various anti-microbial characteristics such as selectivity, good stability, and iterative optimization. Consequently, swift and precise analytical methods to develop the potential and practical applications of high-throughput genetic sequencing-based analysis technology are blooming to become extremely imperative. In addition, genetic sequencing-based analyses are extremely efficient and rapid and can yield large amounts of AMP information simultaneously. The high-throughput genetic sequencing-based screening method took less than 4 h in silico post-generation screening to obtain each AMP gene sequence from 90,000 CLaSS-generated AMP sequence candidates [121] based on the database-based approach, which is suitable for both the mechanism of non-membrane and membrane action. This method is based on the principle of homologous AMP gene sequences, which relies on existing AMP genes as templates to predict the AMP gene sequences; therefore, it is mainly applicable to homologous species. In addition, owing to the high-throughput performance of computers, they are suitable for the simultaneous screening of multiple peptides with known structures, prediction of their structures, and validation of their antimicrobial properties. However, it is important to note that after screening using this method, each AMP must be analyzed and validated using other wet-lab screening methods, as described above. However, there are some limitations, such as the reliance on known AMP genes as templates to predict gene sequences, which is only effective for homologous species. Conventional analysis software can no longer satisfy the requirements of numerous datasets. Fast and accurate analytical software and methods must be developed to demonstrate the advantages and application value of high-throughput sequencing technologies.

3.9. Computational mining of AMP databases

The AMP database was established over 20 years ago to store peptide bioinformatics resources and simple predictions for AMPs [122,123]. The databases collect AMP information, such as length, structure, cysteine, lysine, and other vital amino acids. All biological sources, such as bacteria, insects, amphibians, and mammals, can significantly affect the positive charge number and hydrophobicity of AMPs, which in turn would affect their antimicrobial activity, cytotoxicity, and mode of action towards different targets [124]. However, because of the huge amount of manpower and resource costs for in-vivo or in-vitro detection of AMP and gradual verification, the yields of various AMPs are limited by isolation methods or cherished raw materials, which still need to be further improved to support complete in-vivo and in-vitro experiments as well as other voluminous experiments, such as those for the evaluation of draggability, activity, and stability. Moreover, antimicrobial peptides are limited by their toxicity and unpredictability. This is reflected in the fact that sometimes the final toxicity evaluation is carried out after an elaborate screening phase for antimicrobial activity, at which point an unfortunate “unacceptable toxicity” result is obtained, one of the possible reasons for this being that the relationship between structure and toxicity has yet to be researched and summarized. Therefore, it is necessary to develop prediction tools based on databases and computer methods for large-scale computational mining of AMP database screening methods (Fig. 5).

Fig. 5.

The overall workflow of the antimicrobial peptides (AMPs) database-based screening process, including database-based computational tool screening and experimental evaluation of antimicrobial activity. APD: Antimicrobial Peptide Database; CAMP: Collection of Antimicrobial Peptides; dbAMP: Data bank Antimicrobial Peptides.

To date, more than ten databases have been used to collect AMPs physicochemical properties, toxicity, and specificity data, and provide simple prediction functions. The four most widely used AMP databases are the Database of Antimicrobial Activity and Structure of Peptides (DBAASP), which contains three-dimensional (3D) structural information, chemical structures and synthesis methods of AMPs, the Collection of Antimicrobial Peptides (CAMP) database, with a more accurate AMP prediction ability, the APD database, which has the AMP timeline, and the Data bank Antimicrobial Peptides (dbAMP) database, which has the AMP finder function. Zhang et al. [125] established a long short-term memory (LSTM) model and deep learning-based protein-peptide prediction model (DeepPep) to generate de novo peptides with higher-predicted binding affinity to the major proteases of severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2), and applied it for the rapid screening of peptides. Menousek et al. [126] employed the APD database to screen 30 short peptides (less than 25 residues) predicted as potential inhibitors of methicillin-resistant S. aureus (MRSA). Six of these peptides showed anti-microbial activity against MRSA, and only one AMP, DASamP1, exhibited in-vitro antimicrobial activity against MRSA and inhibited biofilm formation in the early stage. The purpose of establishing a database of specific preferences is helpful in the promotion of AMP screening.

The rapid development of AI (including machine learning and deep learning models) has provided outstanding tools, and the creation of different algorithms has replaced manual extraction from large databases to simplify the process of identifying new AMPs [124,[127], [128], [129], [130]]. The AMP drug discovery process can be accelerated by technology advancements at every stage, including target identification, virtual screening, de novo drug design, predictive modeling, primer optimization, synthesis planning, and clinical trial analysis. Ma et al. [127] combined several neural network models for natural language processing, including LSTM, and attention and bidirectional encoder representations from transformers models, into a unified pipeline to screen for AMP from the human gut microbiome database. A total of 2,349 peptide sequences were found, 181 of which were verified to have antimicrobial activity, with a positivity rate of over 83%. This approach allowed the model to screen short peptide sequences to distinguish between similar peptides and showed increasing insights and self-learning abilities. Eleven AMPs were found to have potent antimicrobial properties against antibiotic-resistant Gram-negative pathogens. The iterative improvement of AI and enhancement of these algorithms, indicating increasing accuracy and computational efficiency, will continue to have a profound impact on the AMP drug discovery process.

In addition to the development of basic technologies, the technological foundation of AI encompasses several key elements critical to shaping efficacy, efficiency, and innovation in the field of AI discovery in AMP research [130,131]. First, high-fidelity, comprehensive datasets covering genomics, proteomics, and clinical records are essential for the training of AI models [123]. Second, advances in hardware technology, such as graphics processing units and specialized AI processors, have accelerated drug discovery and development. Synergistically integrating AI with other technologies, such as robotics and lab automation, automated platforms that are skilled in performing high-throughput screening, and other lab functions can enhance the drug discovery environment. The maturation of tailored cheminformatics and bioinformatics tools for large-scale data analysis (e.g., the various databases described above) is critical [132]. Furthermore, the interoperability of different data formats and platforms (such as big health databases, genome repositories, and clinical trial databases) enhances the ability of AI algorithms to extract comprehensive insights [133]. Computational mining of the AMP database screening method showed a high throughput of seven predicted top-performing AMPs from 512 billion nonapeptide sequences in 19 days [124]. This method can be applied to both non-membrane and membrane action. Computational mining of AMP databases as a promising AMPs screening technique requires the rapid development of computational mining of AMP database systems for AMP medicines and biologically active ingredients, which will probably continue to evolve as AI technology progresses.

In summary, this method is suitable for the prediction of multiple peptides and validation of their antimicrobial properties owing to the strong and promising performance of the computational mining of the AMP database method. However, similar to the genetic sequencing-based analysis, it is necessary to analyze each AMP via wet experiments, as described above, after virtual screening. This method exhibits some advantages such as high screening efficiency, short time, and high precision. However, the screening results need to be further verified. It is generally only suitable for primary screening, relying solely on database mining and not capable of discovering new types of AMPs. Therefore, it is best suited for the extended mining of the screening stage. Finally, as complexity increases, it is important to pioneer models that rationally explain predictive functions. The development of AI platforms as a powerful tool to propel biotech companies towards drug discovery has already opened a promising direction for future pharmaceutical companies.

3.10. Virtual screening

Virtual screening is another promising technique for screening AMPs. Computer-based methods have been used to discover new ligands based on their biological structures [134]. The rapid development of AMP drugs and bioactive component virtual screening systems may begin by considering the ongoing progress in machine learning, which means that it is feasible to find and create novel chemical AMPs that are not found in the natural world [135]. Notably, bacteria can produce AMPs in intricate human physiological environments.

This study provides a scientific foundation for the direct identification of active compounds from microbiome information based on microbial ecological connections. Li et al. [136] prepared sturgeon sperm peptides using enzymatic hydrolysis of sturgeon sperm protein extract as a raw material. The sequence of spermary peptides (SSPs) was analyzed by LC-MS/MS, the AMPs were virtually screened by computational prediction tools and molecular docking methods, and the candidate peptides were prepared by the solid phase synthesis method. The results demonstrated the superior inhibitory action of the SSPs against E. coli, with an inhibitory rate of 76.46%. The peptides NDEELNKLM and RSSKRRQ were obtained using computer prediction tools and molecular docking software. Balmeh et al. [137] employed molecular docking analyses to screen antiviral polypeptides by identifying the peptide with the best affinity for four viral proteins of the coronavirus disease 2019 (COVID-19). The AMPs lactococcin G from Lactobacillus plantarum and glycocin F from Lactococcus lactis were selected as the best antiviral drug candidates for further preclinical experiments. Heymich et al. [138] extracted AMPs from chickpeas and hydrolyzed storage protein legumes using the digestive protease chymotrypsin. Subsequently, ultra-high performance liquid chromatography-triple quadrupole TOF tandem MS (UHPLC-QqTOF-MS) was used to confirm the peptide spectra, and 21 potential AMPs in the hydrolyzates were identified. Among them, Leg1 (RIKTVTSFDLPALRFLKL) and Leg2 (RIKTVTSFDLPALRWLKL) showed antibacterial activity against 16 different bacteria, including pathogens, putrefying bacteria, and two antibiotic-resistant strains. Handley et al. [139] evaluated proline-rich AMPs (PrAMPs) using the Data Repository of Antimicrobial Peptides (DRAMP) database (http://dramp.cpu-bioinfor.org/) to discover new PrAMPs by comparing them with known bacterial protein-folding companion DnaK inhibitors. Eight novel peptides were identified as potential additions to the PrAMP family. Antibacterial naps, attacin-C, P9, and PP30 were chemically synthesized and identified. They proposed a change to the following definition of PrAMPs: peptides with more than 25% of proline content, +1 or higher net charge, regulation of Dnak and/or 70S ribosomes, response to bacterial infections, and significant antimicrobial activity. Virtual screening methods for AMPs are capable of docking one billion compounds in approximately two weeks when utilizing 10,000 CPU cores simultaneously [140]. The method is adapted to both the mechanism of non-membrane and membrane action and is capable of simultaneously screening multiple substances using computer tools.

These methods offer practical and cost-effective in silico solutions for the initial stages of complex-system compound library discovery. The molecular docking method, leveraging structural data to simulate protein-ligand interactions, provides rapid insights into extensive chemical databases, supported by numerous validated cases demonstrating their effectiveness. However, the speed of results generation relies on certain assumptions and approximations, which may compromise algorithmic precision. In contrast, physics-driven molecular simulations offer more accurate and realistic depictions of protein-ligand binding dynamics but involve a slower process and demand significant computational resources [141].

4. Comparison of current screening methods

The methods for AMP screening are evolving towards greater speed, sensitivity, high throughput, simultaneous targeting of multiple complex components, ease of operation, intelligence, and predictive automation. While each screening method has its strengths and weaknesses, a comparative analysis in Table 2 [54,64,73,80,90,110,114,119,126,140] can provide guidance on their application.

Table 2.

Summary and comparison of screening methods for antimicrobial peptides (AMPs).

Screening method	Employed stages	Performance characteristics	Challenges	Time-consuming degreea	Throughput	Refs.
Whole-bacterial adsorption binding	Antimicrobial activity evaluation	Simplicity process and short screening time	Relatively poor accuracy	+	Relatively high	[54]
CMC	Antimicrobial activity evaluation	High sensitivity, good reproducibility, and fast	Complex to operate	+++	Relatively high	[64]
Phospholipid membrane binding	Antimicrobial activity evaluation	Multiple selectivity, easy operation, good stability while maintaining its specificity, and a long lifetime of material	Different from active CMC	++	Relatively high	[73]
Membrane-mediated CE	Antimicrobial activity evaluation	Good resolution, good reproducibility, and quick identification of the interaction between AMP and bacterial membranes	The theory and technology are not mature	++	Medium	[80]
TLC	Antimicrobial activity evaluation	Rapid, operable, visible, adaptable to the structural identification instruments, and suitable for micro isolation	The theory and technology are limited by concentration range	+	Low	[90]
Colorimetric assays	Antimicrobial activity evaluation	Simple and suitable for antimicrobial activity or cytotoxicity determination	The application limited by across concentration range	+++	Medium	[110]
Fluorescence-based screening	Antimicrobial activity evaluation	Highly sensitive, low sample consumption, and no need for the extra transporter	The fluorescence signal is unstable and low anti-interference ability	+++	Medium	[114]
Genetic sequencing-based analysis	Computational tool-based screening	High efficiency, high throughput, and low average cost	Homologous gene templates are required	++	High	[119]
Computational mining of AMP databases	Computational tool-based screening	High efficiency, short time, and precision	The results need further verification and are generally only suitable for primary screening	++	High	[126]
Virtual screening	Computational tool-based screening	More efficient, small sample size demand, and large screening scale	The false positive rate is high and the accuracy needs to be improved	+	Relatively high	[140]

^aThe more +, the more time needed for the screening methods.

CMC: cell membrane chromatography; CE: capillary electrophoresis; TLC: thin layer chromatography.

AMP screening methods that rely on interactions with membranes (including whole-bacterial adsorption binding, CMC, phospholipid membrane binding, membrane-mediated CE, and fluorescence-based screening) have seen considerable advancements. These improvements have notably accelerated screening speeds, reducing the time from several days required by traditional broth incubation methods to just minutes in the new approaches. However, these methods predominantly focus on membrane-based mechanisms of AMP action and may not effectively screen for non-membrane mechanisms. Therefore, alternative methods remain advantageous for comprehensive screening of natural AMPs, ensuring the broader coverage of their diverse mechanisms of action.

Regarding screening tools, methods based on chromatographic principles (such as CMC, phospholipid membrane binding, membrane-mediated CE, and TLC) have demonstrated outstanding capabilities for simultaneous screening and separation of mixed peptide solutions. This eliminates the need to pre-identify concentrations of individual components in the mixture or undertake additional purification steps beforehand. Such methods significantly streamline preprocessing steps, particularly beneficial given the complex peptide fractions found in natural products, which often consist of unknown mixtures of polypeptides at concentrations ranging to several milligrams per milliliter. Chromatographic-based screening methods not only enhance efficiency and throughput but also enable analysis of the integrated antimicrobial actions of complex natural product components. However, the limited number of chromatographic separation channels may not always compare favorably with colorimetric multi-well plate tools in terms of screening reproducibility, especially for assays like MIC in subsequent wet experiments. Consequently, researchers often favor alternative methods for such high-reproducibility screening tests.

Moreover, spectroscopy-based screening methods, such as colorimetric assays and fluorescence-based techniques, have significantly enhanced sensitivity, capable of detecting sample concentrations as low as a few micrograms per milliliter. The use of multi-well plates and the ability to screen virtually all types of AMPs greatly expand their applicability and facilitate integration into automated screening workflows. In parallel, screening methods employing computerized tools and databases, including computational mining of AMP databases, genetic sequencing-based analyses, and virtual screening, have garnered considerable interest. These approaches can substantially increase screening throughput beyond human capabilities and offer predictive insights into unknown AMP structures. Additionally, advancements in microfluidics and methods for cell-free protein synthesis present clear advantages for automating the screening process.

Unfortunately, until now, a method that effectively combines all these advantages has not yet been developed. Screening methods relying on computerized tools and databases have shown high performance but are limited by the available data in existing databases, thus hindering significant advancements. Moreover, the validation process is constrained by the inherent limitations of wet-lab screening methods. It is noteworthy that, except for chromatographic-based methods (excluding membrane-mediated CE) and computerized tools utilizing databases, other screening approaches necessitate the concentrations of purified AMP candidates. Furthermore, regardless of the AMP screening method, a combination of other methods can be used to screen, or coalesce various AMPs into a single sample to obtain the results of macroscopic antibacterial effect.

5. Conclusion

Natural AMPs serve as the primary defense against pathogenic microorganisms in the host immune system and are widespread across various organisms. While membrane-based mechanisms are considered the most common for AMP activity, recent research has revealed diverse pathways through which AMPs exert antimicrobial effects, necessitating further exploration of their mechanisms and diversity.

Advancements in precision instrument processing technologies have spurred the development of robust screening methodologies in discovering a variety of natural materials, enzyme inhibitors, and bioactive components to overcome the limitations of conventional methods. One innovative approach is the at-line nano fractionation (ANF) method, which integrates high-content capabilities into screening. ANF involves chromatographic separation of complex samples, a partial flow fraction analyzed by high-resolution MS for structural identification, and another partial flow evaluated by biological activity assays to obtain an activity database. The bioactive molecule screening process is realized via the fitting of the chromatogram, MS, and activity spectrum. ANF-based screening platforms, which integrate MS, traditional liquid-phase, and the customized 384-well microfluidic fraction collector, have great potential for high throughput discovery of AMPs from natural products.

Despite its potential, the traditional liquid-phase analysis in AMP screening from extremely heterogeneous natural products still faces challenges, such as the requirement of effective detectors beyond ultraviolet (UV) absorption, and the limitations in achieving high throughput beyond 384-well microplates. No universal colorimetric method exists to evaluate all types of AMPs, hindering the widespread application of ANF in AMPs screening. While the ANF method requires further development, its significant potential lies in offering novel insights and guidelines for AMP screening. One-time screening of all types of AMPs from raw materials without repetitive steps, as well as scalable screening methods for industrial applications, are imperative to minimize the interference from the complex components in natural products. Methods employing various solid supports (i.e., polymer monolithic columns and magnetic beads) for AMP screening from natural products could improve the flexibility, speed, and efficiency in future research.

Moreover, focusing on the membrane interaction-based approach enhances the customization of AMP screening for membrane mechanisms, and provides effective tools for studying drug-membrane interactions and developing predictive models. These approaches combined with computerized tools or database-based methods have the potential to further improve screening throughput and prediction accuracy. In conclusion, advancing our understanding of AMP mechanisms of action will facilitate their clinical applications across healthcare, nutrition, and agriculture. This knowledge forms the foundation for future AMP production on a commercial scale, driving innovation in diverse industries.

CRediT authorship contribution statement

Bin Yang: Writing – review & editing, Writing – original draft, Project administration, Investigation, Funding acquisition, Conceptualization. Hongyan Yang: Writing – review & editing, Writing – original draft, Investigation, Conceptualization. Jianlong Liang: Writing – review & editing, Writing – original draft, Visualization, Investigation. Jiarou Chen: Project administration, Investigation. Chunhua Wang: Writing – review & editing, Resources, Investigation. Yuanyuan Wang: Writing – review & editing, Visualization, Resources, Funding acquisition. Jincai Wang: Visualization, Resources, Funding acquisition. Wenhui Luo: Writing – review & editing, Resources, Funding acquisition. Tao Deng: Writing – review & editing, Supervision, Resources, Funding acquisition. Jialiang Guo: Writing – review & editing, Supervision, Resources, Funding acquisition, Conceptualization.

Declaration of competing interest

The authors declare that there are no conflicts of interest.