Segment-based peptide design reveals the importance of N-terminal high cationicity for antimicrobial activity against Gram-negative pathogens

Abstract

Host defense antimicrobial peptides (AMPs) are recognized candidates to develop a new generation of peptide antibiotics. While high hydrophobicity can be deployed in peptides for eliminating Gram-positive bacteria, high cationicity is usually observed in AMPs against Gram-negative pathogen. This study investigates how the sequence distribution of basic amino acids affects peptide activity. For this purpose, we utilized human cathelicidin LL-37 as a template and designed four highly selective ultrashort peptides with similar length, net charge and hydrophobic content. LL-10+, RK-9+, KR-8+, and RIK-10+ showed similar activity against methicillin-resistant Staphylococcus aureus in vitro and comparable antibiofilm efficacy in a murine wound model. However, these peptides showed clear activity differences against Gram-negative pathogens with RIK-10+ (i.e., LL-37mini2) being strongest and LL-10+ weakest. To understand this activity difference, we characterized peptide toxicity, the effects of salts, pH, and serum on peptide activity, mechanism of action, and determined the membrane-bound helical structure for RIK-10+ by two-dimensional NMR spectroscopy. By writing an R program, we generated charge density plots for these peptides and uncovered the importance of the N-terminal high density basic charges for antimicrobial potency. To validate this finding, we reversed the sequences of two peptides. Interestingly, sequence reversal weakened the activity of RIK-10+ but increased the activity of LL-10+ especially against Escherichia coli, Pseudomonas aeruginosa, and Acinetobacter baumannii. Those more active peptides with high cationicity at the N-terminus are also more hydrophobic based on HPLC retention times. A database search found numerous natural sequences that arrange basic amino acids primarily at the N-terminus. Combined, this study not only obtained novel peptide leads but also discovered one useful strategy for designing novel antimicrobials to control drug-resistant Gram-negative pathogens.

Graphical Abstarct

Introduction

Antibiotic resistance constitutes a serious concern as it may put us back to the pre-antibiotic age. In particular, Gram-negative pathogens are difficult to eliminate due to the double membrane structure. It is estimated that 10 million people could die by 2050 [1]. While a comprehensive strategy, ranging from controlled use of existing antibiotics to education of the public, is required, the search of potent antimicrobials is always an option.

Most of the current drugs are derived from nature. Antimicrobial peptides (AMPs) are natural compounds that play an essential role in a variety of organisms, ranging from plants to animals [2–4]. As of September 23, 2024, the antimicrobial peptide database (APD) documented 3279 natural AMPs with 55 reported in 2024 [5–7]. Among them, 406 originated from bacteria, 265 from plants, and 2559 from animals. Of the 153 human AMPs, the major classes are defensins, cathelicidin, ribonucleases, dermcidin, histatins, and antimicrobial cytokines [4]. Most of these peptides are cationic, but dermcidin has a net charge of -2 [3]. While defensins are known to form a β-sheet structure stabilized by three pairs of disulfide bonds, such a bond is absent in human cathelicidin, which forms an α-helical structure [8] with the C-terminal tail disordered (Figure 1A) [9]. The most common mature human cathelicidin peptide, LL-37, is able to eliminate viruses, bacteria, fungi, and parasites [10,11]. Significantly, it can inhibit SARS-CoV-2, Zika, and Ebola viruses [12–14]. While conventional antibiotics are not effective against multidrug resistant bacteria and biofilms, LL-37 remains potent [15,16]. In addition, human LL-37 plays a role in chemotaxis, lipopolysaccharide (endotoxin) neutralization, and wound healing [17,18]. The wide-spectrum activity of LL-37 has stimulated a high interest in developing it into novel antimicrobials. To reduce the cost for chemical synthesis, both structural biology and high-throughput screening peptide library have been utilized to map the active regions of LL-37 [19–23]. Based on NMR structural studies of LL-37 fragments, we previously identified the major antimicrobial region corresponding to residues 17-32 (GF-17/FK-16) as well as the antimicrobial core peptide LL-37(17-29) (i.e., FK-13) [22]. GF-17 was successfully converted to 17BIPHE2 by substituting I20, I24, and L28 with D-leucine, and F17 and F27 with biphenylalanine (Table 1). 17BIPHE2 is a selective, potent, and stable peptide to eliminate the ESKAPE pathogens [24]. Subsequent synthesis of additional short peptides led to the discovery of the first minimal antibacterial peptide KR-12 [9]. This peptide template has been utilized to engineer both linear, stapled and macrocyclic peptides, including lipopeptides [25–28]. Our library screening of truncated KR-12 peptides (4 to 12 amino acids) conjugated with fatty acids with varying lengths at the peptide N-terminus led to the identification of potent and selective peptide C10-KR8 (Table 1) [28]. Our recent studies of ultrashort peptides (≤ 10 amino acids) led to the discovery of another small antibacterial peptide (RIK-10) that inhibits E. coli but not MRSA [29]. The major LL-37 peptides derived from our structural studies are depicted in Figure 1B (amino acid sequences and properties in Table 1). LL-31 corresponds exactly to the helical region of LL-37, while SK-24 corresponds to the helical region without the N-terminal hydrophobic segment (residues 1-8) separated by the hydrophilic S9 on the hydrophobic surface of LL-37 [9].

Figure 1.

LL-37 structure, active region discovery and peptide design strategies: (A) In the membrane-bound state, human cathelicidin LL-37 adopts a helical structure spanning residues 1-31, while the C-terminal residues 32-37 are not structured (PDB ID: 2K6O) [9]. (B) Identification of the antibacterial regions of LL-37 via structural studies and library approach and (C) direct use of short segments of LL-37 for peptide design.

Table 1.

Amino acid sequences and properties of the LL-37 segments and their designed peptides ¹

Peptide	Sequence	Length	Pho	Net charge	GRAVY	Boman index
LL-37	LLGDFFRKSKEKIGKEFKRIVQRIKDFLRNLVPRTES	37	35	+6	−0.72	2.99
LL-31	LLGDFFRKSKEKIGKEFKRIVQRIKDFLRNL-NH2	31	39	+7	−0.64	2.81
SK-24	SKEKIGKEFKRIVQRIKDFLRNLV-NH2	24	38	+6	−0.69	2.95
GF-17	GFKRIVQRIKDFLRNLV-NH2	17	47	+6	−0.094	2.47
17BIPHE2	GBKRLVQRLKDBLRNLV-NH2	17	47	+6	NA	NA
KR-12	KRIVQRIKDFLR-NH2	12	42	+5	−0.71	4.02
C10-KR8	C10-KRIWQRIK-NH2	8	38	+5	−1.53	4.28
LL-10	LLGDFFRKSK-NH2	10	40	+3	−0.38	2.13
LL-10+	LWGRFFRKWK-NH2	10	50	+5	−0.96	2.44
Bip-LL-10+	LWGRFFRKBK-NH2	10	50	+5	NA	NA
RetroLL-10+	KWKRFFRGWL-NH2	10	50	+5	−0.96	2.44
RK-9	RKSKEKIGK-NH2	9	11	+5	−2.26	4.6
RK-9+	RWWKKWWGK-NH2	9	44	+5	−2.24	2.36
KR-8	KRIVQRIK-NH2	8	38	+5	−0.89	4.07
KR-8+	KRWWQWWK-NH2	8	50	+4	−2.43	2.78
Bip-KR-8+	KRWBQWBK-NH2	8	50	+4	NA	NA
RIK-10	RIKDFLRNLV-NH2	10	50	+3	−0.08	2.89
RIK-10+	RWKRFLRNWV-NH2	10	50	+5	−1.19	4.03
Bip-RIK-10+	RBKRFLRNBV-NH2	10	50	+5	NA	NA
RetroRIK-10+	VWNRLFRKWR-NH2	10	50	+5	−1.19	4.03

¹Changed amino acids in ultrashort peptides are in bold. Calculated using the APD website: https://aps.unmc.edu/prediction. Pho is the hydrophobic ratio obtained by summing the eight hydrophobic amino acids (A, V, I, L, M, F, W, and C) divided by peptide length. NA: not available; −NH₂: peptide C-terminal amidation; B: biphenylalanine; C10: a fatty acid with a 10 carbon chain. This chain is not included in the property calculations.

This study took a different avenue to peptide discovery. Since small peptides are cost effective to make, we hypothesized that potent and selective peptides could be designed by directly cutting LL-37 (Figure 1A) into overlapping ultrashort segments (≤ 10 amino acids). In the previous study, six peptides were investigated. Here we selected four peptides from the helical region of LL-37 for peptide design (Figure 1C). Three of these natural segments, LL-10, RK-9, and KR-8, were not antibacterial and only RIK-10 showed a moderate activity in standard MHB [29]. We then enhanced the activity of all these segments via knowledge-based design. Our knowledge-based design was conducted based on the statistical differences between natural and synthetic peptides. Due to optimization, synthetic peptides tend to have higher basic amino acids and hydrophobic amino acids [7]. Arginine (Arg) appears to be more potent than lysine in conferring activity to peptides, especially against viruses [14]. Despite a low abundance in natural AMPs [5], the bulky aromatic tryptophan (Trp) is widely utilized in peptide design since it is especially powerful in membrane binding and conferring activity to short synthetic peptides [29–32]. In addition, arginine tends to have a preferred association with tryptophan [32]. Likewise, one can further enhance peptide activity by substituting tryptophan with biphenylalanine (Bip), which is more hydrophobic [24]. We hypothesized that appropriate substitutions with Trp, Bip, and Arg within ultrashort peptides derived from LL-37 could improve their antimicrobial activity. We then tested antibacterial and antibiofilm activity of the designed peptides both in vitro and in vivo. We also investigated the mechanism of action of these peptides and determined the three-dimensional structure of the most potent peptide. The similar activity against Gram-positive bacteria but clearly different activity against Gram-negative pathogens allowed us to further hypothesize that the distribution of cationic amino acids along the sequence plays a role because these peptides possess similar net charge and hydrophobic ratios. To view the charge distribution, we programmed and generated charge density plots for these designed peptides. The cationic amino acids of three peptides with good activity against Gram-negative bacteria were found mostly at the N-termini. Next, we validated this finding by making sequence reversed peptides. Finally, we searched the antimicrobial peptide database and identified numerous anti-Gram-negative peptides with such a charge distribution in the sequences, thereby uncovering one useful strategy for designing effective AMPs to combat Gram-negative pathogens. Here we report these results.

Materials and Methods

Chemicals and Peptides.

All the chemicals were purchased from established vendors such as Thermo Fisher Scientific Inc and MilliporeSigma. Peptides were made by Genemed Synthesis, Inc. (San Antonio, TX). All the peptides were > 95% pure by HPLC. The correct mass of each peptide was validated by Mass Spectrometry. Peptides were quantitated by measuring UV absorbances at 280 nm (with W) or absorbance differences between 215 and 225 nm (without W) based on the Waddell’s method [33].

Antibacterial Assay.

Bacteria used in this study include S. aureus USA300 LAC, S. epidermidis 1457, E. coli E423-17, E. coli E416-17, P. aeruginosa E411-17, K. pneumoniae E406-17, and A. baumannii B28-16. Antibacterial activity of peptides was tested using the CLSI standard procedure [34] with minor modifications as described [35]. In brief, a peptide concentration gradient with twofold dilution was made in the 96-well polystyrene microplate at 10 μL per well. Bacteria were grown to the exponential phase (i.e., optical density at 600 nm ≈ 0.5), diluted to ~10⁵ CFU/mL in MHB, and partitioned into the 96-well microplate at 90 μL per well. The microplates were incubated at 37°C overnight and read on a ChroMate 4300 Microplate Reader at 600 nm (GMI, Ramsey, MN). The peptide concentration in the wells without bacterial growth is the minimum inhibitory concentration (MIC).

For pH effects on peptide activity, the pH of the medium was adjusted to a targeted pH for autoclave and then remeasured at room temperature. For salt and serum effects on peptide activity, stock NaCl solution or human serum were added to the media.

Killing kinetics.

For bacterial killing kinetics study, we growed MRSA USA300 and P. aeruginosa E411-17 to the exponential phase (OD₆₀₀ ~ 0.3) in MHB. The culture was diluted to OD₆₀₀ 0.001 (~10⁵ CFU/mL). One milliliter of bacterial suspension was then mixed with the peptide, colistin or daptomycin at 4× MIC and incubated at 37°C. After 15, 30, 60, 90, and 120 min incubation, 50 μL was taken and serially diluted with 1×PBS. Then, 50 μL of the diluted suspension was plated on Mannitol Salt Agar (MRSA USA300) or Cetrimide Agar (P. aeruginosa E411-17) plates (NEOGEN, MI). The plates were incubated overnight at 37°C for bacterial CFU count.

Hemolytic Assay.

The hemolytic effects of peptides were evaluated using a previous method [36]. In brief, human red blood cells (UNMC Blood Bank) were washed (3×) and diluted to 2% using the saline solution. Aliquots of 90 μL were mixed with 10 μL of serially diluted peptide solutions. After incubation at 37°C for 1 h, plates were centrifuged at 500g for 5 min and aliquots of the supernatant were transferred to a new 96-well microplate to measure the amount of hemoglobin released at 545 nm (relative to 1% Triton X-100) using a ChroMate Microplate Reader.

Cytotoxicity Assay.

Human keratinocytes (HaCaT) were grown in Dulbecco’s Modification of Eagles’s Medium/High Glucose (DMEM, Hyclone, UT) containing 10% fetal bovine serum (FBS) (Mediatech, Corning, Manassas, VA). Human liver HepG2 cells were cultured in ATCC-formulated Eagle’s Minimum Essential Medium with 10% FBS. Human lung A549 cells were grown in ATCC-formulated F-12K Medium containing 10% FBS. Cells were grown at 37°C in 5% CO₂, and the medium was changed every other day. Peptides were diluted in a 96-well polystyrene microplate (10 μL each well) and mixed with 90 μL of cell suspensions (10⁵ cells/mL) in their respective media. The mixture was incubated at 37°C in a 5% CO₂ for 24 h. Then, 20 μL of MTS reagent (MTS cell proliferation assay kit, Promega, WI) was added to each well and incubated at 37°C for 2 h. The absorbance was determined at 492 nm using a microplate reader. The percentage viability was determined using the following formula: Percentage viability = (OD value of treated cells/OD value of untreated cells) × 100, where OD is the optical density.

Bacterial Membrane Permeabilization.

Bacterial membrane permeation was conducted as described [37]. In brief, the peptide ladder was prepared as described for the antibacterial assay with 10 μL in a black COSTAR 96-well plate. Propidium iodide (PI) (MP Biomedicals, Solon, OH) was dissolved in DMSO (Thermo Fisher Scientific, NY) to 20 mM. This PI stock solution was further diluted to 1 mM with water and 2 μL of 1 mM PI was added to each well. Exponential phase MRSA USA300 or P. aeruginosa E411-17 was diluted to OD₆₀₀ 0.11 with tryptic soy broth (TSB) and 88 μL was added to each well. The plate was incubated at 37°C with continuous shaking in a FLUOstar Omega (BMG LABTECH, NC) microplate reader. The sample was read every 5 min for 24 cycles with excitation and emission wavelengths at 584 nm and 620 nm, respectively.

Bacterial Membrane Depolarization.

Bacterial membrane depolarization was conducted as described previously [38]. Briefly, an overnight culture of MRSA USA300 or P. aeruginosa E411-17 was inoculated into TSB and grown to exponential phase. Bacteria were washed with 1×PBS, re-suspended in twice the volume of 1×PBS containing 25 mM glucose, and incubated at 37°C for 15 min. Then, 500 nM (final concentration) of the DiBAC4 (3) bis-(1, 3-dibutylbarbituric acid) trimethine oxonol (ANASPEC, CA) was added and vortexed gently. Aliquots of 90 μL of the energized bacteria solution were loaded to the 96 well plates (Corning COSTAR, AZ) and placed in a FLUOstar Omega microplate reader (BMG LABTECH, NC). Fluorescence was read for 20 min at excitation and emission wavelengths of 485 nm and 520 nm, respectively. Then, 10 μL of peptide solutions was added and fluorescence readings were recorded for 40 min. Triton X-100 (0.1%) was used as a positive control.

In vitro Antibiofilm Activity.

Inhibition of Bacterial Attachment.

Bacterial attachment was conducted as described [16,17]. In brief, MRSA USA300 or P. aeruginosa E411-17 was grown overnight in TSB media to an optical density (OD₆₀₀) of ~1.0. Then, 180 μL of this culture were added to each well of the microtiter plates containing 20 μL of various MIC folds (1×, 2×, 4×, and 8× MIC) of peptides or antibiotics. The plates were then incubated at 37°C for 1 h. Next, media was aspirated, washed with 1× PBS, and 200 μL of TSB containing XTT 10% [2, 3-bis (2-methyloxy-4-nitro-5-sulfophenyl)-2H-tertazolium-5-carboxanilide] (ATCC, VA, USA) solution was added. After incubation at 37°C for 2 h, absorbance was read at 450 nm using a ChromateTM microtiter plate reader. TSB containing 10% XTT served as blank, while bacterial culture treated with water served as a positive control. Percentages of viable cells in biofilms were plotted by assuming 100% growth in water treated control.

Effects of Peptides on Established Biofilms in Vitro.

Bacteria (S. aureus USA300 LAC or P. aeruginosa E411-17) were grown up in TSB media overnight. A second inoculation was made the next day to reach an exponential phase (OD₆₀₀ ~0.4). Microtiter plates (96 wells, Corning Costar Cat No. 3595), after aliquoting with 180 μL of the culture to each well, were incubated at 37°C for 24 h to form biofilms in TSB. Media were aspirated post incubation and the attached biofilms were washed with 1× PBS to remove the planktonic bacteria. Each well was aliquoted with 20 μL of 10× peptide solution and 180 μL of 10% MHB, and plates were further incubated at 37°C for 24 h. Biofilms treated with water served as the positive control while media without bacterial inoculation served as the negative control. Live cells in the biofilms were quantitated using the XTT as described above. Absorbance was read at 450 nm (only media with XTT containing wells served as the blank) using a ChromateTM microtiter plate reader. Percentage biofilm growth (The plot legend is antibiofilm activity) for the peptides was plotted by assuming 100% biofilm growth in bacterial control alone. The data were represented as mean ± SD, plots were generated using GraphPad prism 7, where * indicates p<0.05, ** p<0.01, and ***P<0.001, and ****p<0.0001 (one-way analyses of variance).

In Vivo Efficacy of the Designed Peptides.

Preparation of Inoculum.

An overnight culture of S. aureus USA300 LAC was inoculated in MHB and incubated at 37°C for 3 h. The suspension was centrifuged at 5,000 g for 5 min and the supernatant was discarded. The bacteria were re-suspended in 1×PBS and the OD₆₀₀ was adjusted to 0.6 (~3 × 10⁷ CFU/mL).

Experimental Animals.

Female BALB/c mice (6-8 weeks, ~20 g) were fed with standardized food (Teklad Laboratory diet for rodents) and water (Hydropac^® Alternative Watering System) ad libitum. Mice were kept in ventilated cages (IVCs) at a temperature of 20-24°C, humidity of 50-60%, 60 air exchanges per hour, and a 12/12-hour light/dark cycle. All materials, including IVCs, lids, feeders, bottles, bedding, and water, were autoclaved before use. All animal manipulations were performed in a class II laminar flow biological safety cabinet. The study was approved by the Institutional Animal Care and Use Committee (IACUC) of UNMC (Protocol no. 22-015-08-FC).

In Vivo Biofilm Assay.

Mice were anesthetized by intraperitoneal injection of ketamine-xylazine (100 mg/kg + 10 mg/kg). The dorsal back hair was removed with a clipper and depilatory cream, followed by cleaning the shaved area with isopropyl alcohol (70%) and povidone iodine swabs. Two full-thickness skin wounds with a diameter of 6 mm were then created using a disposable biopsy punch (Integra Miltex, MA). The wounds were immediately inoculated with MRSA USA300 (10 μL of ~3 × 10⁷ CFU/mL) and covered with a transparent film (3M Tegaderm^™, Deutschland GmbH, Neuss, Germany). Buprenorphine (0.05 mg/kg, SC) was injected and mice were kept in individual cages. After 24 h of inoculation and biofilm formation, treatment groups received 3×10 μL each peptide or daptomycin dissolved in sterile water. Treatment was applied topically on the wound. The infected control group received an equal volume of sterile water. After 24 h treatment, mice were euthanized under CO₂ and specimens were collected using sterile 8 mm biopsy punch (Integra Miltex, MA, USA) into sterilized tubes containing 2 mL of PBS. Tissues were blended using a homogenizer, diluted further in 1×PBS and plated on mannitol salt agar plates. Finally, the plates were incubated at 37°C for 20 h and the CFUs were counted next day.

Structural determination of RIK-10+ by NMR spectroscopy.

NMR data were collected for RIK-10+ (>95% pure) in complex with deuterated DPC_d38 (1:60 molar ratio) in 300 μL 90% H₂O/10% D₂O on the 600 MHz Bruker spectrometer at 25^oC and pH 4.5 as described [36]. The NMR data were processed using NMRpipe [39] and signals were assigned via Pipp [40]. Structures were then determined as described elsewhere by using Xplor-NIH [41]. Structural quality was checked using Procheck [42]. Structural images were produced using MOLMOL [43].

Charge and hydrophobic density plots for antimicrobial peptides.

Charge and hydrophobic densities for each amino acid in helical peptides were calculated by writing an in-house R program. The window size was four based on the 3.6 amino acids per turn in an α-helix. The calculations started from the N-terminus and ended at a residue whose fourth count is the C-terminal residue. Multiple sequences could be handled in one shot. Figures were made based on the calculated values using GraphPad Prism 7.

Determination of peptide hydrophobicity by RP-HPLC.

Peptide hydrophobicity was measured on a reversed phase HPLC system equipped with a C18 Gemini-NX column (4.6×250 mm, 5 μ particle size, 110 Å pore size). The peptide was eluted with a solvent gradient of 20% to 80% and detected by UV at 220 nm. The mobile phase flow rate was 1.0 ml/min. Both eluents contains 0.1% (v/v) of trifluoroacetic acid (TFA). Peptide retention time was recorded for comparison.

Statistical Analysis.

Data were analyzed using GraphPad Prism 7 (GraphPad Software, Inc., San Diego, CA, USA) and measured values were expressed as mean ± standard deviation. One-way analyses of variance (ANOVA) were used to compare the mean values among treatment groups. P < 0.05 was considered statistically significant.

Results

Peptide design.

Four peptide segments, including LL-10, RK-9, KR-8, and RIK-10 (Figure 1C), were previously produced based on the long helical region (Figure 1A) of human LL-37 determined by 3D triple-resonance NMR spectroscopy [9]. While LL-10, RK-9, and KR-8 are not antibacterial in the standard Mueller Hinton Broth (MHB), RIK-10 is active against E. coli but not S. aureus USA300 [29]. To enhance the activity of these peptides, we made three- to five amino acid substitutions in each segment based on the preferred amino acids in amphipathic peptides [7]: (1) In LL-10, L2 was changed to W, and D4 was altered to R, and S9 was replaced with tryptophan (W); (2) four W residues were deployed in both RK-9+ and KR-8+ to increase peptide activity; and (3) In RIK-10, we substituted D4 with R and changed both I2 and L9 to W. The peptide sequences before and after changes, including their physiochemical properties, are shown in Table 1 where activity-enhanced sequences are indicated with “+”. As these amino acid changes retain the amphipathic patterns, the new peptides have the potential to retain the helical structure as observed for the parent peptide. After these changes the four designer peptides are comparable in length (8-10 amino acids), hydrophobic ratios (Pho: 44-50%), and net charges (+3 to +4). The hydrophobic ratios and net charges in these small peptides are similar to those in the majority of natural AMPs in the APD (https://aps.unmc.edu) [6,44]. However, there are some variations in GRAVY [45] and Boman index [46] in Table 1. The GRAVY values of the two Trp-rich peptides, RK-9+ and KR-8+, are twice as negative as those of LL-10+ and RIK-10+ with comparable values (~−1). While three of the designed peptides have a Boman index in the range of 2.36 to 2.78, RIK-10+ has the highest value at 4.03 kcal/mol (Table 1). We also generated helical wheels for these peptides by assuming an α-helical structure. Except for RIK-10, the original LL-37 segments (Table 1) are not perfectly amphipathic (see the helical wheel plots in Figure 2A) [47,48]. However, they are entirely amphipathic after amino acid substitutions (Figure 2B), leading to increased hydrophobicity <H> and hydrophobic moment <μH> for each peptide. Although amino acids may change positions on helical wheels, sequence reversal did not alter the amphipathic pattern of the designed peptides (Figure 2C). Hence, peptide hydrophobicity and hydrophobic moment remained the same before and after sequence reversal (cf.: Figures 2B and 2C).

Figure 2.

Helical wheel plots of short segments of LL-37: (A) natural peptide segments LL-10, RK-9, and RIK-10, (B) activity enhanced peptides LL-10+, RK-9+, and RIK-10+, and (C) reversed sequences of LL-10+, RK-9+, and RIK-10+. KR-8+ is not provided since the HeliQuest program did not produce anything. Indicated underneath each panel for each peptide are hydrophobicity <H> and hydrophobic moment <μH>. The helical wheel was generated using the Heliquest program (https://heliquest.ipmc.cnrs.fr/cgi-bin/ComputParams.py).

Antimicrobial activity of the designed peptides.

We then measured the antimicrobial activity of the four designed peptides (Table 2) using the standard microdilution assay in MHB. All these peptides are highly purified with HPLC purity greater than 95% (Figure S1). Two Gram-positive (S. aureus and S. epidermidis) and three Gram-negative bacterial strains (P. aeruginosa, E. coli, and A. baumannii) were tested. Most of the Gram-negative bacteria contain extended-spectrum beta-lactamase (ESBL) and are resistant to beta-lactam antibiotics. LL-10+ and RK-9+ showed moderate activity (MIC 8-32 μM) against four of these bacteria except for A. baumannii (MIC > 32 μM). While KR-8+ and RIK-10+ were moderately active against Gram-positive Staphylococcal strains, including methicillin-resistant S. aureus (MRSA), they displayed excellent activity against Gram-negative pathogens, especially E. coli and P. aeruginosa (MIC 4-8 μM). In addition, RIK-10+ was more active against A. baumannii than the other three peptides (Table 2). It appeared that the activity of these peptides against Gram-negative pathogens is in the following order: RIK-10+ > KR-8+ > RK-9+ > LL-10+ (see the Graphic). Since these peptides are more active against Gram-negative bacteria than Gram-positive bacteria, they are highly desired considering a general lack of new drugs against these pathogens. Hence, we have designed unique peptides (e.g., KR-8+ and RIK-10+ in Table 1), which differ from our previously designed LL-37 peptides such as C10-KR8 and 17BIPHE2 with broad antibacterial activity against both Gram-positive and Gram-negative pathogens (Table 2). As positive controls, daptomycin and colistin displayed anticipated activity spectra against Gram-positive and Gram-negative bacteria only, respectively.

Table 2.

Antimicrobial activity of human LL-37 derived antimicrobial peptides ¹

Peptide	Gram-positive bacteria		Gram-negative bacteria
	S. aureus USA300	S. epidermidis 1457	P. aeruginosa E411-17	E. coli E423-17	A. baumannii B28-16
LL-37	>32	>32	16	8	8
LL-10+	16	32	16	32	>32
RK-9+	8	16	16	8	>32
KR-8+	8	16	8	4	16
RIK-10+	16	>32	4	4	8
C10-KR8	4	4	8	8	4
17BIPHE2	4	8	4	4	4
Colistin	ND 1	ND	2	2	2
Daptomycin	0.5	0.5	ND	ND	ND

¹Peptide activity is represented as minimal inhibitory concentration (MIC) in μM. ND: activity not detected at the highest concentration tested.

We also attempted to further increase peptide activity by substituting Trp with Bip. Bip-KR-8+ was obtained by altering both W4 and W7 of KR-8+ to Bip, while Bip-LL-10+ was designed by substituting W9 of LL-10+ to Bip. In Bip-RIK-10+, both W2 and W9 of RIK-10+ were replaced by Bip (Sequences in Table 1). Such a Trp to Bip change increased peptide activity against Staphylococcal strains (MIC 1-4 μM). In particular, Bip-RIK-10+ was potent against both Gram-positive and negative pathogens (MIC 1-2 μM), including K. pneumoniae E406-17, which was killed by RIK-10+ at 32 μM (not shown).

Bacterial growth inhibition and killing kinetics.

We then followed bacterial growth inhibition kinetics at varying peptide concentrations. In the case of S. aureus USA300, all the four peptides did not inhibit its growth completely until 16 μM. At 8 μM, KR-8+ (Figure 3A) and RK-9+ (Figure 3B) were more effective than LL-10+ (Figure 3C) and RIK-10+ (Figure 3D). These inhibition curves agree with the MIC values for these peptides against MRSA (Table 2). The growth curves of P. aeruginosa, however, were more distinct for the four peptides. The peptide concentration for complete Pseudomonal inhibition was in the following order: RIK-10+ (4 μM) > KR-8+ (8 μM) > LL-10+ (16 μM) > RK-9+ (32 μM) (Figure 3, E–H). These results are consistent with the MIC values in Table 2 and provided additional insight into bacterial growth inhibition kinetics. As RIK-10+ is more effective against Gram-negative bacteria, we also compared its bacterial killing kinetics. At 4 × MIC, RIK-10+ was able to eliminate P. aeruginosa in 60 min, whereas colistin killed the same bacterium in 30 min (Figure S2A). At the same MIC fold, it took 90 min to kill MRSA USA300 similar to daptomycin (Figure S2B). These results suggested stronger antimicrobial power of RIK-10+ to attack Gram-negative pathogen than Gram-positive bacteria, consistent with MIC values (Table 2).

Figure 3.

Growth curves of S. aureus USA 300 (A-D) and P. aeruginosa E411-17 (E-H) treated with LL-10+, RK-9+, KR-8+, and RIK-10+ in MHB.

Effects of pH, salts, and serum on peptide activity.

It is known that media conditions could compromise activity of human LL-37. In human lung, LL-37 becomes less active at pH 6.8 than at pH 7.4. Due to binding, LL-37 can be inactivated by human apolipoprotein A-I [49]. To further understand antimicrobial potential of these peptides, we also evaluated the effects of pH, physiological salts, and human serum on peptide activity. In the case of S. aureus, all the peptides retained activity or even became more active at a basic pH of 8.1, but lost activity at an acidic pH of 6.2 (Table S2). In the presence of 150 mM NaCl, both LL-10+ and RIK-10+ lost activity (MIC > 32 μM). In contrast, RK-9+ and KR-8+ retained some activity (MIC 16-32 μM). These peptides retained some activity in the presence of 5% serum, but lost this activity at a higher serum concentration. Under these conditions, however, all the designer peptides appeared to behave better than the parent peptide LL-37 (Table S2). As positive controls, two lipopeptides, C10-KR8 [28] and daptomycin, remained potent against S. aureus USA300 under all these conditions. C10-KR8 is a robust peptide we designed previously by conjugating KR-8 (a short peptide of LL-37) with caprylic acid at the N-terminus. As a negative control, colistin was inactive under these conditions against MRSA.

We also evaluated antibacterial activity of these four designer peptides under different pH, salt and serum conditions using P. aeruginosa E411-17. As a negative control, daptomycin did not display any activity against P. aeruginosa under all the conditions tested here (Table S3). Interestingly, all the peptides remained pseudomonacidal even at the acidic pH of 6.2 (Table S3). Moreover, the activity of KR-8+ was not compromised in the presence of 150 mM NaCl. As in the case of MRSA, human serum (10%) could reduce the activity of these peptides. Of note, the parent peptide LL-37 retained some activity in 10% serum. Like colistin (positive control), our previously LL-37 designed C10-KR8 peptide [28] also retained pseudomonacidal activity under these conditions. The version of C10-KR8 we used was made of D-amino acids, substantially reducing its binding to human serum proteins as we proved via mass spectrometry proteomic studies [28].

Cytotoxicity of the designed peptides.

Next, we asked whether the newly designed peptides are toxic. We first used the well-established hemolytic assays. Like daptomycin and colistin, LL-10+ and RK-9+ did not lyse human red blood cells even at 400 μM. KR-8+ started to show some toxic effect only at 400 uM, whereas RIK-10+ showed a dose-dependent low-level hemolysis only at high peptide concentrations (200 μM). At 400 μM, KR-8+ and RIK-10+ caused 18% hemolysis (Figure 4A). Since the 50% hemolytic concentrations (HC₅₀) were far greater than 400 μM, all the four designed peptides are highly selective. These short peptides are more selective than its parent peptide LL-37, the major antimicrobial peptide GF-17, and its engineered peptide 17BIPHE2 [50].

Figure 4.

Peptide dose-dependent cytotoxicity to (A) human red blood cells (hRBC), (B) human skin HaCaT cells, (C) human HepG2 liver cells, and (D) human A549 lung cells. Colistin and daptomycin were included as controls. The results were expressed as mean ± standard deviation.

We also evaluated hemolysis of the above peptides after incorporation of Bip (Table S1). Unfortunately, Bip-RIK-10+ became more toxic than RIK-10+ with a HC₅₀ value of ~ 30 μM. Although Bip-KR-8+ showed a high HC₅₀ of >200 μM, its HC₁₀ (10% hemolytic concentration) was only 15 μM. The toxicity of these two peptides resulted from the introduction of two Bip amino acids. Bip-LL-10+, with only one Bip, was of some interest with HC₅₀ >200 μM and HC₁₀ at 100 μM (Table S1). However, the HC₁₀ values for LL-10+, RK-9+, KR-8+, and RIK-10+ were >400, >400, 300, and 220 μM, respectively, all much less toxic than those Bip-containing peptides. Hence, we did not study these Bip-containing peptides further.

To provide additional evidence for peptide safety, we also evaluated peptide toxicity using other mammalian cells. In the case of human keratinocytes (HaCaT), LL-10+, RK-9+, and KR-8+ showed little toxicity even at 100 μM similar to daptomycin and colistin. Only RIK-10+ displayed some toxicity at 100 μM with a 50% lethal concentration (LC₅₀) of 100 μM (Figure 4B). We also tested peptide toxicity to human liver cancer cells (HepG2) commonly used for hepatocytoxicity assays. Like colistin and daptomycin, LL-10+ and RK-9+ showed subtle effects. KR-8+ appeared to be most toxic with LC₅₀ at 12.5 μM. Intriguingly, this effect was not dose dependent as cell viability did not change with increase in peptide concentration. Similar to the effect on human skin cells, RIK-10+ showed a low toxicity below 50 μM with an LC₅₀ at 100 μM. Another two peptides, LL-10+ and RK-9+, were less toxic with LC₅₀ > 100 μM, comparable to daptomycin and colistin (Figure 4C). The high cell selectivity of these peptides is also evident in the case of human A549 lung cells with LC₅₀ >100 μM (Figure 4D). These results indicate that three out of the four peptides, LL-10+, RK-9+, and RIK-10+, are rather safe at the MIC values required to kill these pathogens in Table 2. It appeared that these LL-37 derived peptides might have stimulated the growth of human skin HaCaT cells, but not human liver HepG2 or lung cells at low concentrations.

Mechanism of action of the designed peptides.

Human LL-37 is known to act on bacterial membranes via the carpet model or by forming the barrel-stave pore or toroidal pore [51–53]. The fragments of LL-37, including GF-17 and GI-20, share the same mechanism as indicated by membrane permeabilization [50]. Since LL-10+, RK-9+, KR8+, and RIK-10+ are all designed based on LL-37, it is likely that these short peptides also target bacterial membranes. To confirm this, we conducted membrane permeabilization experiments with and without the peptide. In this experiment without the peptide, the indicator dye could not enter bacteria, associate with DNA, and emit fluorescence (Figure 5). In the presence of the designed peptides, an increase in fluorescence implied membrane permeabilization of S. aureus USA300. In contrast, LL-37 failed to do so since it is known to be inactive in TSB (Figure 5A) due to media inactivation [29]. Together with panels B and C, a peptide dose-dependent membrane permeabilization (1-32 μM) was evident. With an increase in peptide concentration till 8 μM, fluorescence increase started to be evident with RK-9+ at the top. This remained the same until 32 μM where RIK-10+ became slightly more powerful than RK-9+. To study the effect of bacterium Gram type, we also used P. aeruginosa. At 8 μM, RIK-10+ started to permeate bacterial membranes, while the effects of other peptides were marginal. Despite some delay, LL-37 reached a much higher level of membrane permeation than RIK-10+. The same trend recurred at 32 μM. To verify this, we also studied E. coli and A. baumannii. Remarkably, we observed similar membrane permeabilization of E. coli and A. baumannii by LL-37, which shows similar activity against these Gram-negative bacteria (Figure S3). Consistently, RIK-10+ was strongest, whereas LL-10+ was weakest in the case of E. coli. In all the cases, RIK-10+ was able to permeate bacterial membranes faster than LL-37 although it did not reach the magnitude achieved by its parent peptide. RIK-10+ is superior to other designed peptides in permeating the membranes of all the three Gram-negative bacteria. This is interesting considering only RIK-10 was active against E. coli before antibacterial activity enhancement. These results underscore the significance of the RIK-10 sequence derived from the core antimicrobial region of LL-37 in elimination of Gram-negative pathogens (Figure 1).

Figure 5.

Designer peptides permeabilized the membranes of S. aureus USA300 (A-C) and P. aeruginosa E411-17 (D-F) as indicated by a dose-dependent fluorescence increase. RFU: Relative fluorescence intensity. Results were expressed as mean ± standard deviation.

Cationic AMPs may also depolarize bacterial membranes. This effect was measured in PBS buffer primed with glucose. Under such a condition, LL-37 was observed to be most powerful in membrane depolarization, even more effective than the positive control Triton-X100 at 8 μM and above for both S. aureus (Figure S4, A–C) and P. aeruginosa (Figure S4, D–F). Compared to LL-37 or Triton-X100, the membrane depolarization effects of the designed ultrashort peptides were generally weak. At 32 μM, KR-8+ was slightly better than other short peptides in the case of S. aureus. All the designed peptides only weakly depolarized bacterial membranes of P. aeruginosa even at 32 μM. KR-8+ was at the top of the four designed peptides in this experiment. Taken together, these short designed peptides also depolarize bacterial membranes although weakly compared to its parent peptide LL-37 [8,10].

Antibiofilm capability of the four designed peptides.

As most of bacteria are in the biofilm state, we then compared antibiofilm activity of the four peptides designed based on ultrashort LL-37 segments. Both S. aureus USA300 and P. aeruginosa E411-17 were included in bacterial attachment (Figure 6 A & B) and preformed biofilm disruption experiments (Figure 6 C & D). In the case of MRSA, RIK-10+ at 8 fold of its MIC was found to be more potent than the other three peptides against bacterial attachment, the first step in biofilm formation. Importantly, RIK-10+ was also more potent than daptomycin treated at 8 × MIC (Figure 6A). A dose-dependent antibiofilm effect was also observed for 24-h preformed biofilms of MRSA. At 4 × MIC, peptides LL-10+ and KR-8+ were less effective than RK-9+ and RIK-10+. The latter two appeared to be more potent than daptomycin at 4 × MIC and 8 × MIC (Figure 6C).

Figure 6.

Inhibition of bacterial biofilm formation and disruption of preformed biofilms in vitro. Surface attachment of S. aureus USA300 (A) and P. aeruginosa E411-17 (B). Metabolic activity of the 24-h preformed biofilms of S. aureus USA300 (C) and P. aeruginosa E411-17 (D) after peptide treatment for 24 h. The metabolic activity of the biofilm-associated bacteria was assessed using the formazan-based MTT assay on biofilms formed for 24 h in TSB media.

In the case of P. aeruginosa, these peptides were comparable in inhibiting bacterial attachment at 2 × MIC or below (Figure 6B). At 8 × MIC, RIK-10+ appeared to be slightly more effective than other peptides. However, they were all less effective than colistin at 4× or 8× MIC. We also tested the biofilm-disruption ability of these peptides. The dose-dependent effect on P. aeruginosa (Figure 6D) was less pronounced than that on MRSA (Figure 6C). As we demonstrated for other LL-37 peptides [16], these designer small peptides are also anticipated to inhibit biofilm formation probably via direct bacterial killing at and above MIC.

In vivo antibiofilm activity of the four peptides in a murine wound model.

One major hurdle for wound healing is bacterial infection and biofilm formation. The current APD has annotated 33 AMPs to have wound healing effect [7]. Human LL-37 is a typical example that can eliminate bacteria and promote wound healing [17,54,55]. To further compare the antibiofilm ability of the four designer peptides, we also evaluated their in vivo efficacy in treating the 24-h preformed biofilms of S. aureus USA300 in murine wounds (see methods for technical details). These peptides, when directly loaded to the wound, were able to reduce the bacterial burden by 2-3 logs after one treatment for 24 h (Figure 7). They showed similar effects, consistent with similar MIC values against MRSA (Table 2). The antibiofilm effects of these peptides were also comparable to daptomycin in the same animal model. After formulation, an eight-residue LL-37 mimicking peptide can further eliminate bacterial pathogens and shows a synergistic effect with nanomaterial in promoting wound healing [55]. These results are encouraging as the small size of these designer peptides could substantially reduce the production cost compared to LL-37.

Figure 7.

Anti-biofilm activity of the four designer peptides against S. aureus USA300 in vivo. S. aureus USA300 biofilms were formed in the wounds created on the back of BALB/c mice for 24 h followed by treatment with water (untreated control), KR-8+, RK-9+, LL-10+, RIK-10+ or daptomycin (all at 10 mg/kg per wound) for 24 h. Results were expressed as mean ± standard deviation.

Membrane-bound structure of RIK-10+ determined by 2D NMR spectroscopy.

LL-37 has a long helix covering residues 1-31 in complex with membrane-mimicking sodium dedecylsulphate (SDS) or dioctanoylphosphatidylglycerol (D8PG) (Figure 1A) [9]. Also, the C-terminal tail of LL-37 remains disordered in complex with E. coli lipopolysaccharides (LPS) [9]. The helical structure is retained for its long fragments in the membrane bound state or even in PBS [9,22,50]. It can be projected that the small peptides designed here are potentially helical as well (Figure 2). To substantiate this, we determined the 3D structure of RIK-10+, which is a more potent anti-Gram-negative peptide (Table 2 and Figure 3). Since the peptide acted on bacterial membrane (Figure 5), we utilized lipid micelles to mimic bacterial membranes. This study used dodecylphosphocholine (DPC) [56] for two reasons: First, this deuterated lipid could give a signal dispersion in NMR spectra comparable to that of D8PG where a deuterated version is not yet available [57,58]. Second, we obtained similar structures for LL-37 peptides bound to DPC and D8PG [58,59]. At a peptide:lipid molar ratio of 1:60, one peptide was bound to one micelle. This sample enabled us to determine the membrane-bound structure of RIK-10+ by using the established solution NMR method [60]. Peptide proton signals were assigned using the Wüthrich sequential assignment method. We obtained 101 NOE restraints for this 10-residue peptide. In addition, we generated 16 dihedral angle restraints for residues 3-10 based on chemical shifts, which are known to improve structural quality [61,62]. The NMR-determined 3D structure is presented in Figure 8. The structure was determined to high quality with 90% of the backbone angles in the most favored Ramachandran region and 10% in the additionally allowed region. Residues 3-10 were found to be helical with a superimposed backbone RMSD of 0.32 Å (Figure 8, A and B). When viewed from the end of the structure, the amphipathic nature of RIK-10+ was evident with W2, F5, L6, and W9 forming the hydrophobic surface (Figure 8C). The amphipathic pattern of the structure could also be seen in the potential surface with basic charge in blue and hydrophobic amino acids in light purple (Figure 8D). A further analysis of the 2D NOESY spectra revealed that the peptide sidechains directly interacted with residual protons from DPC. These were intermolecular NOE cross peaks from well-resolved DPC protons (0.6 ppm) to sidechains of R1, W2, K3, R4, and W9, confirming interfacial location of these amino acids in the peptide-lipid complex.

Figure 8.

Three-dimensional structure of RIK-10+, a peptide designed based on RIK-10 derived from the major antimicrobial region of human cathelicidin LL-37. The NMR structure was determined in a membrane-mimetic environment (peptide:DPC molar ratio: 1:60) at 25^oC and pH 4.5.

Charge and hydrophobic density plots of the designed peptides.

Because basic charge and hydrophobicity are the two most important physiochemical parameters of amphipathic helical peptides, we calculated both charge and hydrophobic densities for the four peptides. As α-helix has 3.6 residues per turn, we rounded it up to a window size of 4. For charge density calculations, we summed the number of arginine (R) and lysine (K) divided by the window size (there is no histidine in these sequences). Likewise, we calculated the hydrophobic density per residue by summing all the eight hydrophobic amino acids (A, V, L, I, F, M, C, and W) defined in the APD [6,7] and divided by the same window size of 4. A computing R program was written to facilitate these calculations. Both charge and hydrophobic density calculations started from the N-terminal amino acid of each sequence and ended when it reached the last four residues at the C-terminus. We focused on charge plots (Figure S5) since the hydrophobic plots (Figure S6) were rather homogeneous except for KR-8+. For LL-10+, charge density increased from the N-terminus to the C-terminus with the highest at 0.75. In contrast, the plots for another three peptides were higher at the N-termini. Interestingly, RK-9+, KR-8+, and RIK-10+ (MIC 4-8 μM) were all more potent against E. coli than LL-10+ (MIC 32 μM). It appeared that the location of the high-density charge at the N-terminus was important in E. coli killing. To confirm this, we decided to reverse peptide sequences. LL-10+ and RIK-10+ were selected as they presented an exactly opposite charge plots (Figure S5, A & G). These two retro peptides were chemically synthesized and subjected to antibacterial testing. Remarkably, LL-10+ became much more potent against all the bacteria after sequence reversal (named retroLL-10+). In contrast, retroRIK-10+ became much weaker (Table 3). However, the effects are larger in the cases of Gram-negative bacteria than Gram-positive bacteria. As anticipated, their charge density plots were also reversed (Figure S5, B&H). Such a difference is not clear in the helical wheel plots after sequence reversal (Figure 2C). To provide additional insight, we also measured HPLC retention times for LL-10+ and RIK-10+ before and after sequence reversal. Remarkably, RIK-10+ and retroLL-10+ showed longer retention times (i.e., more hydrophobic) than retroRIK-10+ and LL-10+ (Table 3). These results confirmed the importance of the N-terminal location of the highly charged regions for antibacterial activity especially against Gram-negative bacteria and peptide hydrophobicity also played a role.

Table 3.

Effects of sequence reversal on antibacterial activity and hydrophobicity of designed peptides ¹

Name	S. aureus USA 300	S. epidermidis 1457	P. aeruginosa E411-17	E. coli E423-17	A. baumannii B28-16	HPLC retention time (min)
LL-10+	16	32	8	32	64	9.87
RetroLL-10+	8	32	4	4	32	11.21
RIK-10+	16	64	4	4	8	11.81
RetroRIK-10+	32	128	16	32	64	8.91

¹Peptide activity is represented as minimal inhibitory concentration (MIC) in μM.

Charge and hydrophobic plots shine new light on nature’s design of human cathelicidin.

Since RIK-10 is the shortest antibacterial unit of LL-37, we were curious how the charge and hydrophobic plots of LL-37 look like. In the charge-density plot (Figure S7A), the magnitude of charge gradually decreased from the N-terminal to the C-terminal region, providing one interpretation for its antibacterial activity against Gram-negative bacteria. The hydrophobic density plot of LL-37 (Figure S7B) is also interesting. There were three 50% hydrophobic zones: residues 1-4, 17-21, and 24-31. A clear hydrophobic valley existed between the first and second zones due to the presence of hydrophilic S9 on the hydrophobic surface as observed in the 3D structure of LL-37 [9]. Another two 50% hydrophobic zones with a dip at residues 22-23, when combined, correspond exactly to the major antimicrobial region we previously discovered from NMR structural studies [22]. Overall, the charge and hydrophobic plots of LL-37 are rather complementary to each other since a highly charged region is achieved at the expense of hydrophobic amino acids (Figure S7).

Charge density plots for antimicrobial peptides against Gram-negative bacteria obtained from the APD.

To identify additional examples, we searched the APD for peptides with activity against Gram-negative only. The following search criteria were applied: (1) active against Gram-negative bacteria only; (2) peptide length less than 50 amino acids; (3) peptides containing both arginine and lysine; (4) peptides with known helical structure; and (5) natural AMPs. These database filters led to a total of 11 peptides in Table S4. Of note, numerous peptides deployed more basic amino acids at the N-termini (K: red and R: blue). This can be viewed more clearly from the charge density plots for seven peptides (Figure S8, F–L). For comparison, LL-37 is provided in Figure S8A. Biologically, these peptides are active against Gram-negative pathogens such as E. coli (Table S4). While the charge density plot of amphibian ocellatin-PT8 (Figure S8B) showed a relatively even distribution of positively charged R and K along the sequences, the basic amino acids of amphibian thaulin-1 (Figure S8E) are mostly located at the C-terminus. Such different charge distribution patterns may explain in part why they showed weak activity. However, two peptides with 4-5 charge clusters (Figure S8, C & D) showed good activity against E. coli. Hence, other charge distribution patterns such as an even distribution are also useful, enriching charge distribution patterns in natural AMPs.

Discussion

Antimicrobial peptides are host defense molecules of innate immunity that rapidly stop invading microbes. Despite structural diversity, AMPs have been unified into four classes: linear (e.g., human LL-37), sidechain-linked (e.g., defensins), sidechain-backbone-linked (e.g., daptomycin), and backbone-linked (i.e., cyclotides) peptides based on chain connection patterns [44]. Among them, linear peptides are extensively studied and some important discoveries have been made. For over 1000 amphibian peptides, which are mostly helical after binding to bacterial membranes, net charge (mainly due to lysine) increases whereas hydrophobic ratio (mainly due to leucine) decreases with the increase in peptide length [63]. In addition, peptides with low cationicity and high hydrophobicity tend to kill primarily Gram-positive bacteria [35,36]. However, a high net charge is frequently observed for peptides against Gram-negative bacteria [5,63]. Consistent with this database finding, we succeeded in converting GF-17, a wide-spectrum human LL-37 major antimicrobial peptide [64], into narrow-spectrum peptides that are active against either only Gram-positive or Gram-negative bacteria [9,22,65]. This study has advanced our knowledge by identifying sequence features that determine peptide activity against Gram-negative bacteria. We identified this feature by designing multiple short and active peptides because longer peptides such as human cathelicidin LL-37 can contain different functional regions, which could complicate the analysis. After activity enhancement of the four small segments of LL-37 (Figure 1), all the peptides achieved a net charge and hydrophobic content comparable to the majority of natural AMPs in the APD [7]. Interestingly, these short peptides demonstrated similar activity against MRSA both in vitro (Table 2) and in vivo (Figure 7). Nevertheless, they vary in activity against Gram-negative bacteria (see the Graphic). We noticed a gradient increase of peptide activity against Gram-negative bacteria from LL-10+, RK-9+, KR-8+, to RIK-10+. It appears that the antibacterial activity difference is related to Boman index with higher values for KR-8+ and RIK-10+ than LL-10+ and RK-9+ (Table 2). Structurally, these peptides are anticipated to be helical like the parent peptide LL-37 (Figure 1A) as the changes we made keep the amphipathic pattern (Figure 2). This is indeed the case after we completed the 3D structure of RIK-10+ by the improved 2D NMR spectroscopy, which includes the use of natural abundance heteronuclear chemical shift information in structural refinement [62]. However, it is challenging to correlate helical structures with the differences in activity against Gram-negative bacteria [62].

Because basic and hydrophobic amino acids are the two most important parameters for cationic AMPs [8–11,23], we calculated charge and hydrophobic density plots to gain additional insight into peptide activity differences against Gram-negative bacteria. While the hydrophobic plots are similar for the four designed short peptides, charge density plots differ. Remarkably, three of the designed peptides deployed cationic amino acids primarily at the N-terminus, corresponding to stronger activity against E. coli (Table 2). In the case of LL-10+, a peptide derived from the N-terminus of LL-37 (Figure 1), its highly charged region was located at the C-terminus, corresponding to the poorest activity in inhibiting E. coli. We propose that the location of a highly basic region at the N-terminus of these short peptides is important for antimicrobial activity against Gram-negative pathogens. Subsequently, we validated this finding by making sequence reversed peptides. As anticipated, our sequence reversal increased the activity of LL-10+, but decreased the activity of RIK-10+ against nearly all the bacteria (Table 3). However, the activity increase or decrease is more pronounced for Gram-negative bacteria than Gram-positive bacteria for both peptides. Previously, Neubauer and colleagues [66] investigated the retro analog concept by comparing six AMPs with and without sequence reversal. Most of these AMPs do not follow the sequence pattern of LL-10+ or RIK-10+. Three amphibian peptides, aurein 1.2, citropin 1.1, and temporin A [67–69], deployed 1-2 basic amino acids in the middle, while CAMEL, a hybrid peptide of cecropin(1-7) and magainin(2-9), has a relatively even distribution of cationic residues along the sequence [70]. Pexiganan, a 22-residue magainin derivative, has two basic clusters at both the N and C-termini [71]. Interestingly, sequence reversal only clearly increased the activity of omiganan (formerly MBI 226), a synthetic analog of bovine indolicidin [72], against a panel of Gram-negative bacteria, including E. coli, K. pneumoniae, and P. aeruginosa, while its activity either did not change or changed by two fold against E. faecalis, S. aureus, and S. pneumoniae, all Gram-positive bacteria [66]. Also, sequence reversed omiganan (r-omiganan) with three basic amino acids relocated to the N-terminus became more hydrophobic with a longer HPLC retention time. This is exactly in line with the results of our sequence reversal of LL-10+ and RIK-10+. The peptides are more hydrophobic and active when basic amino acids have been shuffled to the N-terminus (Table 3).

Our finding here is in line with the two distinct amphipathic peptides: horine and verine we designed recently [36]. Horine, with the hydrophobic motif WWW at the N-terminus, kills mainly Gram-positive MRSA. In contrast, verine, with triple basic amino acids RRR at the N-terminus, kills both Gram-positive and Gram-negative bacteria. We also observed the deployment of high cationicity at the N-terminal portion of LL-37. These results underscore the importance of a proper positioning of basic amino acids in peptide sequences for potency against Gram-negative pathogens. Indeed, multiple natural sequences from invertebrates (e.g., insects) in the antimicrobial peptide database [7] share a similar charge distribution pattern (Figure S8). Most of 34 insect cecropins place multiple lysine, arginine, histidine, or both lysine and arginine at the N-terminus. Cecropins could preferentially eliminate Gram-negative bacteria over Gram-positive bacteria [46,73,74]. Likewise, 12 insect moricins, with broad bactericidal activity, take advantage of a similar design for host defense [75]. Hence, the placement of high cationicity at the peptide N-termini constitutes a natural strategy to control invading pathogens, especially Gram-negative bacteria. Driven by electrostatic interactions, such a molecular design would enable the host defense bullet to more effectively reach the targeted pathogen like an arrow followed by binding to bacterial membranes with higher affinity.

We have been searching for short active peptides within LL-37 via structural studies to reduce the production cost [9,22]. We previously identified a series of antimicrobial fragments from LL-37 with KR-12 being the smallest (Figure 1B) [9]. Recently, we reported the identification of KR-8, which only became active after MHB medium dilution. Previously, a potent peptide designed based on KR-8 was named LL-37mini [29]. Since the amino acid sequence of KR-8+ investigated here is unique and keeps more sequence features of LL-37, we gave KR-8+ an easy-to-remember name LL-37mini1. Among the four ultrashort natural segments of human LL-37 (Figure 1), however, only RIK-10 is able to inhibit the growth of E. coli without media dilution. After peptide activity enhancement, it is remarkable that RIK-10+ remains more potent than other small designer peptides. Since both KR-8 and RIK-10 with an overlapping sequence are derived from the same major antimicrobial region of LL-37, we refer to RIK-10+ as LL-37mini2, which has gained numerous merits: being short, potent, and highly selective. These LL-37mini peptides are 2-4 residues shorter than KR-12, the minimal antibacterial peptide we initially reported [9]. Taken together, this study has not only deepened our fundamental understanding of natural AMPs but also brought about effective peptides for developing novel antimicrobials to combat difficult-to-kill Gram-negative pathogens.

Conclusion

Previous LL-37 based peptide design utilized the active regions discovered from either structural biology or peptide library [22–24]. While KR-12, a previously identified smallest antibacterial peptide [9], has been widely utilized in engineering various antimicrobial constructs (for a recent review, see ref. [76]), this study took another step in peptide design by dissecting the helical region of human cathelicidin LL-37 into overlapping ultrashort fragments with 10 amino acids or less. The statistical differences between natural and synthetic AMPs [7] directly guided our design. A similar net charge, hydrophobic ratio, and length for these designer peptides (LL-10+, RK-9+, KR-8+, and RIK-10+) might be responsible for their comparable anti-MRSA activity. However, they showed clearly different MICs against Gram-negative bacteria with RIK-10+ being strongest. This is remarkable since RIK-10+ is derived from the core antimicrobial region of LL-37 we discovered based on structural studies [22]. As these peptides are very small and highly selective, they are interesting candidates for further development. While high cationicity has been recognized to be important for killing Gram-negative bacteria, this study also revealed the importance of charge distribution patterns for such a peptide activity. LL-10+ with charged residues mostly at the C-terminus was weakest in inhibiting E. coli, while other peptides with higher activity placed their basic charges mainly at the N-terminus. When the sequences for LL-10+ and RIK-10+ were reversed, their activity in combating bacteria also reversed (i.e., strong to weak or vice versa). More active peptides with high cationicity at the N-terminus also possess higher hydrophobicity, contributing to improved antibacterial activity, especially against Gram-negative pathogens (Table 3). Since numerous natural AMPs in the antimicrobial peptide database [6,7] deployed more basic amino acids at the N-terminal region, our observation uncovered one useful strategy for designing new peptide antibiotics to fight drug-resistant Gram-negative pathogens.

ABBREVIATIONS USED

AMPs

antimicrobial peptides

APD

the antimicrobial peptide database

CFU

colony forming units

DPC

dodecylphosphocholine

D8PG

dioctanoylphosphatidylglycerol

ESKAPE pathogens

Enterococcus faecium, Staphylococcus aureus, Klebsiella pneumoniae, Acinetobacter baumannii, Pseudomonas aeruginosa, and Enterobacter species

FBS

fetal bovine serum

HPLC

high performance liquid chromatography

LL-37

a linear human antimicrobial peptide with 37 amino acids starting with a pair of leucine

MHB

Mueller Hinton Broth

MIC

minimum inhibitory concentration

MRSA

methicillin-resistant Staphylococcus aureus

NMR

nuclear magnetic resonance spectroscopy

NOE

nuclear Overhauser enhancement

NOESY

nuclear Overhauser enhancement spectroscopy

PBS

phosphate buffer saline

TSB

tryptic soy broth

UNMC

University of Nebraska Medical Center

UV

ultraviolet