Computational Design of Pore-Forming Peptides with Potent Antimicrobial and Anticancer Activities

Abstract

Peptides that form transmembrane barrel-stave pores are potential alternative therapeutics for bacterial infections and cancer. However, their optimization for clinical translation is hampered by a lack of sequence-function understanding. Recently, we have de novo designed the first synthetic barrel-stave pore-forming antimicrobial peptide with an identified function of all residues. Here, we systematically mutate the peptide to improve pore-forming ability in anticipation of enhanced activity. Using computer simulations, supported by liposome leakage and atomic force microscopy experiments, we find that pore-forming ability, while critical, is not the limiting factor for improving activity in the submicromolar range. Affinity for bacterial and cancer cell membranes needs to be optimized simultaneously. Optimized peptides more effectively killed antibiotic-resistant ESKAPEE bacteria at submicromolar concentrations, showing low cytotoxicity to human cells and skin model. Peptides showed systemic anti-infective activity in a preclinical mouse model of Acinetobacter baumannii infection. We also demonstrate peptide optimization for pH-dependent antimicrobial and anticancer activity.

Figure 1

Computational design of TBP-forming peptides. Mutations favoring TBP stabilization in the “scaled” Martini simulations. (a) Helical wheel diagram of LP1, simulation snapshots of hexameric LP1 TBP, and the effect of “N-shift” motions (black arrows) on the stability of intermolecular salt bridges and aromatic stacking interactions in LP1 TBP. LP1 interaction strengths are used as a reference (orange lines). (b) Stronger aromatic stacking on the second peptide–peptide interface of hexameric LP6 TBP using I26F substitution. (c) Stronger salt bridges resulted in octameric LP14 and heptameric LP15 TBPs. (d) Shorter N-terminal K-cluster decreased the N-shift and stabilized heptameric LP17 TBP. (e) Carboxy-terminus and complementary stacking with I26F resulted in an octameric LP26 TBP. (f) T/S substitutions caused tight packing of polar faces, resulting in narrower but octameric LP34 and LP36 TBPs. Neutral H-containing peptide ends (H ends) and carboxy-terminus resulted in octameric LP40 TBP. Snapshots were taken after 51 μs simulation using the “scaled” Martini force field, showing the side and top views of TBPs in the POPC lipid membrane. Schematic illustrations are shown for three antiparallel neighboring transmembrane peptides representing two peptide–peptide interfaces of a TBP (side and top views). Stability of stacking and salt bridge interactions was calculated as the percentage of designed interaction contacts averaged over 51 μs simulation using the standard and “scaled” Martini force fields (Table 1). Color coding: peptide hydrophilic and hydrophobic residues in green and white, respectively; basic and acidic in residues blue and red, respectively; aromatic residues in gray; membrane lipid phosphates in yellow and tails as gray panel; and yellow horizontal lines in the schematic illustrations indicate the position of lipid phosphates.

Figure 2

Computational design of peptides with switched charge distribution stabilizing TBPs. Mutations favoring TBP stabilization in the “scaled” Martini simulation. Helical wheel diagram (a), simulation snapshots of TBPs (a, b), schematic illustrations of “C-shift” motions (black arrows) and the intermolecular interactions, and the strength of these interactions in TBPs (a–c; orange lines indicate LP1 reference). Snapshots were captured after 51 μs simulation using “scaled” Martini force field, showing the side and top views of TBPs in POPC lipid membrane. Schematic illustrations represent three antiparallel neighboring transmembrane peptides with two peptide–peptide interfaces from a TBP (side and top views). Stability of aromatic stacking and salt bridge interactions was calculated as the percentage of designed interaction contacts averaged over 51 μs simulation using the standard and “scaled” Martini force fields (Table 1). Color coding: peptide hydrophilic and hydrophobic residues in green and white, respectively; basic and acidic residues in blue and red, respectively; aromatic residues in gray; membrane lipid phosphates in yellow and tails as gray panel; and yellow horizontal lines in the schematic illustrations indicate the position of lipid phosphates.

Figure 3

In vitro antimicrobial activity and toxicity. (a–d) Antimicrobial activity is reported as the MIC values against Gram-positive (colored violet–blue–green) and Gram-negative (colored yellow–pink–orange–red) ESKAPEE pathogens. (e) Cytotoxicity against human keratinocyte cells HaCaT, human alveolar epithelial cells A549 (derived from adenocarcinoma), and murine macrophage cells RAW 264.7 is reported as the IC50 values after 24 h of peptide treatment. (f) Toxicity against the reconstructed human epidermis model is reported as the average tissue viability after 1 h treatment with peptides at 50 μM concentration and 42 h of post-treatment incubation. Ampicillin, polymyxin B, levofloxacin, melittin, and sodium dodecyl sulfate (SDS) were used as controls. Peptide mutations tested are as follows—LP3: F→W for aromatic W-stacking; LP8: D→E for K–E salt bridges; LP18: +2 e increase in net charge; LP22: negatively charged C-terminus; LP28: +3 e increase in net charge; LP29–LP37: A-, M-, V-, L-, I-, T-, Q-, S-, N-variants; LP40: H-variant with +2 e net charge; sequences are given in Table 1.

Figure 4

Anti-infective activity in the deep thigh infection mouse model. (a) Schematic of the deep thigh infection model in which bacteria are injected intramuscularly at day 4 and peptides are administered intraperitoneally also at day 4 to assess their anti-infective activity. Mice were euthanized 4 days postinfection (day 8). Each group consisted of six mice (n = 6), and the bacterial loads used to infect the mice derived from three different inocula. (b) Intraperitoneal treatment with the peptides at 10-fold MIC (i.e., LP1: 0.16 mg/kg, LP18: 0.17 mg/kg, LP28: 0.18 mg/kg, LP40: 1.33 mg/kg) reduced the bacterial load of A. baumannii (ATCC 19606) compared to the untreated control group. Polymyxin B (0.006 mg/kg) and levofloxacin (0.014 mg/kg) were used as the reference antibiotics. Statistical significance was done using one-way ANOVA followed by Dunnett’s test. Violin plots represent the median, upper quartile, and lower quartile. (c) Mouse weight was monitored to exclude possible toxic effects of the peptides. Peptide sequences are shown in Table 1.

Figure 5

Pore-forming mechanism of action of designed AMPs. (a) Peptide-induced leakage of the fluorescent dye calcein from the LUVs composed of POPC:POPG (1:1 mol/mol) lipids. Surfactant triton was used as the control, causing 100% leakage in the end. (b) Peptide-induced depolarization of the bacterial cytoplasmic membranes is detected as an increase in the fluorescence intensity of the membrane potential-sensitive dye DiSC₃-(5). (c) Peptide-induced permeabilization of Gram-negative bacterial outer membranes is indicated as an increase in fluorescence intensity of the lipophilic dye NPN. Peptides were tested at their MIC values against the respective bacteria. Polymyxin B and levofloxacin were used as reference antibiotics. (d) AFM topographical images of (i) the initial untreated, homogeneous, defect-free, and vesicle-free SLB (1 × 1 μm²) and (ii–iv) LP18 peptide-treated SLBs captured successively at increasing resolution (1 × 1 μm² > 400 × 400 nm² > 200 × 200 nm²). Color scale (height): 2 nm. SLBs were composed of POPC:POPG lipids (1:1 mol/mol). The studied P:L molecular ratio is 1:500, which corresponds to a final peptide concentration of 1 μM. Peptide sequences are given in Table 1.

Figure 6

The pH-dependent in vitro antimicrobial and anticancer activity. (a) Antimicrobial activity against E. coli (TOP10) cultured at two different pH is reported as the MIC values. (b) Activity against a panel of 22 different cancerous and noncancerous human cell lines is reported as IC50 values after 72 h of peptide treatment at the physiological pH. (c) Anticancer activity against four different cancerous human cell lines cultured in three different media, RPMI, DMEM, and PBS, which are adjusted to three different pH values 8, 7, and 6, and additionally one media without pH adjustment (∼7.6) is reported as the cell viability after 1 h treatment with peptides at 10 μM concentration. Peptide melittin, antibiotic ampicillin, and PBS buffer were used as controls. The graphs represent two independent repetitions. Peptide sequences are listed in Table 1.