Self-derived structure-disrupting peptides targeting methionine aminopeptidase in pathogenic bacteria: a new strategy to generate antimicrobial peptides

Abstract

Bacterial infection is one of the leading causes of death in young, elderly, and immune-compromised patients. The rapid spread of multi-drug–resistant (MDR) bacteria is a global health emergency and there is a lack of new drugs to control MDR pathogens. We describe a heretofore-unexplored discovery pathway for novel antibiotics that is based on self-targeting, structure-disrupting peptides. We show that a helical peptide, KFF-EcH3, derived from the Escherichia coli methionine aminopeptidase can disrupt secondary and tertiary structure of this essential enzyme, thereby killing the bacterium (including MDR strains). Significantly, no detectable resistance developed against this peptide. Based on a computational analysis, our study predicted that peptide KFF-EcH3 has the strongest interaction with the structural core of the methionine aminopeptidase. We further used our approach to identify peptide KFF-NgH1 to target the same enzyme from Neisseria gonorrhoeae. This peptide inhibited bacterial growth and was able to treat a gonococcal infection in a human cervical epithelial cell model. These findings present an exciting new paradigm in antibiotic discovery using self-derived peptides that can be developed to target the structures of any essential bacterial proteins.—Zhan, J., Jia, H., Semchenko, E. A., Bian, Y., Zhou, A. M., Li, Z., Yang, Y., Wang, J., Sarkar, S., Totsika, M., Blanchard, H., Jen, F. E.-C., Ye, Q., Haselhorst, T., Jennings, M. P., Seib, K. L., Zhou, Y. Self-derived structure-disrupting peptides targeting methionine aminopeptidase in pathogenic bacteria: a new strategy to generate antimicrobial peptides.

If new classes of antibiotics are not developed, the growing global threat of multidrug antibiotic resistance could make many common infections untreatable in this century (1). Widespread antibiotic resistance occurs largely because most effective antibiotics act on only 3 classes of targets or pathways: ribosomes, cell wall synthesis, and DNA gyrase and topoisomerase (2). This restricted set of targets or pathways is related to the fact that most existing antibiotics are based on natural products discovered from extensive soil screening during the 1940s to 1960s (3). Attempts during the last 2 decades, using high-throughput screening and rational drug design, have met with limited success in delivering novel, synthetic, small-molecule antibiotics because of a lack of specificity and an inability to penetrate bacterial cells (2). A recent breakthrough in screening uncultured soil bacteria led to the discovery of teixobactin (4); it is ineffective, however, against most Gram-negative bacteria, such as Escherichia coli and Neisseria gonorrhoeae. E. coli causes diverse and serious diseases, ranging from urinary and gastrointestinal tract infections to bacteriemia and meningitis (5). N. gonorrhoeae is the causative agent of the sexually transmitted infection gonorrhea, which causes localized disease of the genital tract but is also associated with infertility and increased transmission of HIV [reviewed in Edwards et al. (6)]. Because of increasing drug resistance (E. coli forms a key part of carbapenem-resistant Enterobacteriaceae), these 2 bacteria have been ranked at the highest threat level, an “urgent threat,” by the U.S. Centers for Disease Control and Prevention (Atlanta, GA, USA) (7).

In recent years, peptide-based drugs have gained increased attention, largely due to their high success rates in clinical trials as a result of their high specificity and low toxicity (8), as well as advancements in technology for peptide delivery (9), synthesis, (10) and stabilization (11). Self-targeting, inhibitory peptides are those with a sequence derived from a polypeptide segment of the protein that it is designed to inhibit. These peptides have been used to inhibit protein-protein interactions (12, 13) or to disrupt folded structures (14, 15), but to date, they have not been used as antibiotics to specifically target the structures of essential bacterial proteins. Bacteria have hundreds of proteins that are essential for survival and growth (16, 17). Such a wide variety of essential proteins provides an opportunity to design and screen self-targeting peptides that unfold essential protein targets and kill the pathogen. In this study, we tested this strategy with 2 MDR pathogens and the essential enzyme methionine aminopeptidase (MetAP).

MATERIALS AND METHODS

Peptides

The peptides shown in Supplemental Table S1 were synthesized by Genscript (Piscataway, NJ, USA) at >95% purity with N-terminal acetylation and C-terminal amidation.

Enzyme activity assay

Methionine aminopeptidase of E. coli (EcMetAP) was expressed in E. coli BL21 (DE3) with a polyhistidine tag, purified by immobilized metal ion affinity chromatography followed by ion exchange chromatography and size-exclusion chromatography. The protein was treated with 20 mM EDTA to remove bound metal ions before size exclusion chromatography. EcMetAP was incubated with peptides for 2 h at 4°C, and then enzyme activity was measured by the catalyzed hydrolysis rate of the fluorogenic substrate methionyl aminomethylcoumarin (Met-AMC) (18), at 37°C, using an Infinite 200 Pro Multimode Microplate Reader (Tecan, Männedorf, Switzerland). The reaction solution was composed of 0.5 μM EcMetAP, 50 mM Tris (pH7.4), 100 μM Met-AMC, 100 μM MnCl₂, and 5 mM DTT.

Bacterial growth assays

E. coli strains MG1655 (nonpathogenic) obtained from American Type Culture Collection (ATCC; Manassas, VA, USA), CFT073 (a uropathogenic), B2C (an enterotoxigenic), EDL933 (enterohemorrhagic; from Harry Sakellaris, Griffith University), and EC958 (an MDR uropathogenic) (19) were cultured from an initial optical density at 600 nm (OD₆₀₀) of 0.001 in EZ Rich Defined Medium (RDM; Teknova, Hollister, CA, USA) supplemented with 100 mM Tris and 225 μM resazurin (pH 7.4 in 37°C) in 96-well plate wells with frequent shaking. Growth of bacteria was monitored by the fluorescent product converted from resazurin by viable cells. Fluorescent signals excited at 380 nm and emitted at 460 nm in the course of cell culture were monitored by the Infinite 200 Pro Multimode Microplate Reader (Tecan) or a Microplate Absorbance Spectrophotometer (Bio-Rad, Hercules, CA, USA) for strain EC958.

Minimum inhibitory concentration

Peptide minimum inhibitory concentration (MIC) was determined according to the Clinical and Laboratory Standards Institute guidelines (20) (i.e., the lowest concentration of an antimicrobial agent that prevents visible growth of a microorganism in an agar or broth dilution assay over 20-h growth). The growth of microorganisms was also quantified by measuring cell density at OD₆₀₀ absorbance after 20-h incubation at 37°C.

Minimum bactericidal concentration

After MIC assays, undiluted and 10-fold serial dilutions (to decrease the peptide concentration remaining from the MIC assay) of samples were plated onto antibiotic-free lysogeny broth agar plates. CFUs were counted after overnight incubation at 37°C. The minimum bactericidal concentration (MBC) is determined as the lowest concentration of peptide that reduces the viability of the initial bacterial inoculum by ≥99.9%. The peptide is regarded as bactericidal if the MBC is no more than 4 times the MIC, or as bacteriostatic, if otherwise (21).

Intracellular MetAP-inhibition assay

E. coli. MG1655 was cultured in RDM to log phase (OD₆₀₀ = 0.3–0.7). The cell density was then adjusted to OD₆₀₀ = 0.1 with RDM and further incubated in the absence or presence of the KFF-EcH3 peptide at various concentrations for 1 h at 4°C. The cells were then quickly washed and resuspended with chilled Dulbecco’s PBS buffer. The collected cells were aliquoted for assays of EcMetAP activity and genomic DNA quantification. For intracellular MetAP assay, cells with 1 mM DTT were lysed by ultrasonication, and the hydrolysis rate of Met-AMC in the supernatant was measured as described above. The genomic DNA in samples were quantified after cells were stained with Hoechst H33342 for 10 min at 4°C and then washed again. The fluorescence signals emitted at 465 nm from the H33342-DNA complex were measured. The effect of peptide KFF-EcH3 on intracellular EcMetAP activity was indicated by the ratio of EcMetAP activity to the cellular DNA that are normalized to 1 in the cell sample in the absence of peptide.

Seminative and denaturing SDS-PAGE and Western blot analysis

Cells cultured in RDM were incubated with or without 4 µM KFF-EcH3, P-EcH3 linked with a cell-permeating peptide (KFF)₃K at the N-terminal, and P-EcH3 synthetic peptide with the same sequence as Helix 3 of EcMetAP, at 37°C for 1 h. The cells were washed in PBS 3 times to remove extracellular peptide, then lysed by ultrasonication. Cell debris was removed by centrifugation. The cell extracts were untreated for seminative SDS-PAGE, or treated with 5 mM DTT and boiled for 15 min before loading for denaturing SDS-PAGE. Proteins were transferred from the gel onto a PVDF membrane, which was washed in Tris-buffered saline with Tween20 (TBST) 3 times, blocked in TBST+1% BSA, and incubated with 1:1000 diluted biotinylated EcMetAP antibody (30283-05121; AssayPro, St. Charles, MO, USA) at 4°C overnight. After primary antibody incubation, the membrane was washed 3 times and incubated with a 1:10,000 dilution of horseradish peroxidase–conjugated streptavidin. Blots were washed with TBST 3 times and digitally imaged with enhanced chemiluminescence substrate.

Overexpression of MetAP in E. coli

The EcMetAP gene was cloned into the pFLAG vector (Addgene, Cambridge, MA, USA) by using the Gibson Assembly Kit (New England Biolabs, Ipswich, MA, USA). The pFLAG-EcMetAP vector or an empty pFLAG control vector was introduced into the EcMetAP-deficient E. coli cells CLA-MAP5 (22) by heat shock, and the cells were cultured in RDM with 0.2% arabinose. EcMetAP expression was induced with 0.1 mM IPTG in RDM with 0.2% arabinose. The growth of CLA-MAP5 is limited in the absence of arabinose (Supplemental Fig. S5A), but equivalent growth was seen for cells containing the pFLAG-EcMetAP or the empty pFLAG vector that were grown in the presence of 0.2% arabinose.

Drug-resistance assay

To monitor the development of drug resistance, E. coli MG1655 cells cultured in RDM were passaged daily with 0.5, 1, and 2× MIC of KFF-EcH3 or the control antibiotic enrofloxacin from MilliporeSigma (Burlington, MA, USA). After each passage, the MIC was updated as the lowest concentration, in which there was no significant 20-h growth at 37°C. The culture with 0.5× updated MIC was then diluted to ∼10⁶ cfu/ml and passaged once more with 0.5, 1, and 2× updated MIC of the peptide or the control antibiotic.

Cytotoxicity assay

Human cell ME180 (cervical epidermoid carcinoma cells), A549 (lung epithelial carcinoma cells), and Caco-2 (colorectal adenocarcinoma cells) were obtained from ATCC and maintained according to the supplier’s instructions. The assay was performed in a black 96-well plate, with cells grown to near complete confluence (>95%). Before the experiment, all cells were washed once with HBSS (Thermo Fisher Scientific). Cell medium supplemented with 1% (v/v) peptides dissolved in DMSO at various concentrations in 100 µl was then added. The plates were incubated in 37°C and 5% CO₂ for 3 h, after which 1 µM Sytox Green Nucleic Stain (Thermo Fisher Scientific) was added to each well. The plates were further incubated for 10 min in 37°C and 5% CO₂, and total-well fluorescence was measured (504/523 nm) with an Infinite 200 Pro Multimode Microplate Reader (Tecan). Negative control wells contained medium supplemented with 1% (v/v) DMSO, and positive control wells contained medium supplemented with 0.1% (v/v) Triton X-100. Data are shown as mean fluorescence units from 3 replicate wells ± sd.

Thermofluor assay

The fluorescence generated by bound Sypro Orange with exposed hydrophobic residues allows the determination of the stability curve and melting temperature. The Thermofluor assay is composed of 4 µM EcMetAP, 20 µM peptide and 1× Sypro Orange (Thermo Fisher Scientific) in 10 mM Tris, 20 mM MgCl₂, and 1 mM DTT (pH 7.4). The assay mixture was first held at 20°C for 5 min and then heated to 70°C in stepwise increments of 0.2°C for 30 s. The fluorescence was monitored by the CFX96 Real-Time PCR System (Bio-Rad). Fluorescence data were normalized by subtraction of blank controls that did not contain EcMetAP protein and were aligned according to baselines. The melting temperature was calculated by averaging the inflection points of the fluorescence slopes of 4 independent measurements.

Tertiary structural changes by NMR heteronuclear single quantum coherence spectra

This method requires the introduction of a ¹⁵N-label to the protein by expressing the EcMetAP in M9 minimal medium with ¹⁵NH₄Cl as the sole nitrogen source. The protein was purified by the same process as previously described. ¹H-¹⁵N heteronuclear single quantum coherence (HSQC) spectra were collected using a 600-MHz NMR spectrometer (Bruker, Preston, VIC, Australia) at a temperature of 300 K. The sample contained 20 μM EcMetAP dissoved in 50 mM Tris (pH 7.4), 100 mM MgCl₂, 5 mM DTT, 90% H₂O, and 10% D₂O, in the absence or presence of 500 μM P-EcH3 or P-EcS12 peptide. The MetAP protein concentration used in NMR [and circular dichroism (CD) analysis] is higher than that used in enzyme activity inhibition experiments, to enable us to reach the detection threshold of these methods.

Secondary structural changes by circular dichroism

CD spectra were collected with a spectropolarimeter (J-1500; Jasco, Easton MD, USA) with a 1-mm optical path length cuvette. The CD spectra were acquired at 25°C with 4 μM EcMetAP in 10 mM Tris, 20 mM MgCl₂, and 1 mM DTT (pH 7.4), with or without 5, 10, or 20 μM P-EcH3 (1:1.25, 1:2.5, or 1:5 MetAP:P-EcH3 peptide molar ratio, respectively) or 20 μM P-EcS12 peptide. The samples were scanned 5 times from 195 to 260 nm with a 0.5-nm interval. CD spectra were corrected for background. That is, the CD spectra of the peptide at the corresponding concentration was subtracted from the protein-peptide CD spectra. The protein secondary structure content was analyzed with K2D3 (23).

Metadynamics molecular dynamics simulations

Molecular dynamics (MD) simulations were performed to investigate interactive mechanisms between EcMetAP and the self-derived peptides P-EcH1, -H2, and -H3. Each peptide was randomly placed manually near EcMetAP, and the systems were then solvated within periodic boxes of transferable intermolecular potential-3P water molecules. The systems were neutralized and maintained at a salt concentration of 100 mM by adding Na⁺ and Cl⁻ ions. The particle mesh Ewald method (24, 25) was used to treat the electrostatic interaction, and the cutoff was set to 1.0 nm. The cutoff of the van der Waals interactions was also set at 1.0 nm. All the bonds were constrained with the linear constraint solver algorithm, and thus the MD time step was set to 2 fs. The Berendsen algorithm (26) was used for both temperature and pressure coupling. Each of the systems was first subjected to a minimization of 1000 steps, followed by an equilibrium run with an NPT ensemble at 1 atm and 300 K for 20 ns, where N is the number of molecules, V is the volume of the system, and T is the temperature. Then, the last conformation was used as the initial structure for further simulations. All the simulations were performed with Gromacs 4.6.2 (27) using the Amber99SB-ILDN force field (http://www.gromacs.org/), where ILDN is the 1-letter code for the side chains whose potentials are modified (28). In addition, we used an advanced sampling method (MD) (29) to explore the conformational space. It accelerates barrier-crossing events by periodically adding repulsive gaussian potentials on the collective variables (CVs). We used 2 CVs to describe the binding behavior of the peptides: the distance between the center of mass of the peptide and that of the protein and the residue-residue contact numbers between a peptide and EcMetAP based on the Cα-Cα distance (<5 Å). The height of the gaussian potential was set to 0.5 kJ/mol, and the width was set at 0.35 and 1.5 for the 2 CVs, respectively. The time interval for depositing the gaussian potentials was 1 ps. The simulation time of each system was 300 ns. A Gromacs plugin (PLUMED; v.1.3) (30) was used for the MD simulation. We calculated the free-energy landscapes by summing the added gaussian potentials along the CVs defined above. We also clustered the structures of the MetAP within the deepest basin to examine dominant conformations of EcMetAP. Each simulation took 2 mo on a 12-core single node machine on a 4000L cluster (Sugon, Beijing, China).

N. gonorrhoeae bacterial strains and growth conditions

N. gonorrhoeae 1291 (a male gonococcal urethritis isolate, provided by M. A. Apicella, University of Iowa, Iowa City, IA, USA) (31) and MDR reference strains WHO-K, -L, and -P (provided by Prof. M. Lahra, University of New South Wales, Sydney, NSW, Australia) were grown at 37°C with 5% CO₂ on GC agar (Oxoid) plates supplemented with IsoVitaleX (BD Biosciences, San Jose, CA, USA) for 16-h, and MIC experiments were performed as previously described for E. coli, with an initial OD₆₀₀ of 0.05.

Infection assays

Human transformed cervical epithelial (TCX) cells were provided by M. A. Apicella and maintained in keratinocyte-serum free medium (Thermo Fisher Scientific) supplemented with bovine pituitary extract (25 mg/L) and epidermal growth factor 1-53 (16 µg/L). The cells were grown to confluence in tissue culture–coated 96-well plates and infected with a multiplicity of infection of 10 by N. gonorrhoeae 1291 for 1 h. The cells were then washed 3 times with PBS and treated with medium supplemented with gentamicin (100 µg/ml) for 15 min to kill the extracellular bacteria. The TCX cells were washed again 3 times with PBS before media supplemented with various concentrations of peptides were added. The plates were incubated for 6 and 12 h, respectively, after which the cells were lysed by adding saponin (1% v/v). Cell lysates were plated on GC agar and colony-forming unit (CFU) counts were performed the next day. Data are shown as mean CFU counts from 3 replicate wells ± sd. Control wells included no treatment (medium only) and 20 µM KFF peptide.

RESULTS

Identification of self-targeting peptides of MetAP

To investigate whether self-targeting, structure-disrupting peptides can be used as antibiotics, we studied the essential E. coli protein, MetAP. MetAP facilitates the maturation of proteins by catalyzing the excision of N-terminal initiator methionine residues from nascent proteins in organisms. We focused our studies on the helical regions in MetAP, because helices are one of the most common secondary-structure motifs used for peptide and peptidomimetic drug discovery (32, 33). MetAP has 4 long helical segments (H1, H2, H3, and H4) (34) (Fig. 1A). We tested the enzymatic activity of recombinant EcMetAP in the presence of 10 µM of each of 4 synthetic peptides—P-EcH1, P-EcH2, P-EcH3, and P-EcH4—that have sequences identical to H1, H2, H3, and H4 of EcMetAP, respectively (Supplemental Table S1). P-EcH3 had the highest inhibition of EcMetAP activity [half maximum inhibitory concentration (IC₅₀) = 3.9 μM], and P-EcH1 had low inhibition (IC₅₀ = 28.6 µM) (Fig. 1B, C). However, P-EcH2, P-EcH4, or the β-hairpin–enclosing strands 1 and 2 (P-EcS12, used as a control) did not show inhibition of EcMetAP activity (Fig. 1B).

Figure 1 .

Identification and characterization of self-inhibitory peptides. A) The structure of the EcMetAP in complex with a small-molecule inhibitor [5-(2-chlorophenyl)furan-2-carboxylic acid in yellow] binding to the functional site (PDB ID: 1XNZ). The structure has 4 helices, labeled H1–H4, and a β-hairpin region that encloses sheets 1 and 2 (S12 in blue). As an example, P-EcH3 is shown as a peptide derived from H3 of EcMetAP. B) Enzyme activity of recombinant EcMetAP was inhibited strongly by 10 µM P-EcH3, weakly by P-EcH1, but not inhibited by P-EcH2, P-EcH4, or P-EcS12. Error bars are based on the standard deviation of triplicate experiments. C) Enzyme activity of recombinant EcMetAP was inhibited in a concentration-dependent manner by P-EcH3 and P-EcH1, with IC₅₀ shown. Activity was not inhibited by the shuffled P-EcH3 (P-sEcH3). D) E. coli was inhibited in growth by KFF-EcH3 at concentrations ≥2 µM (labeled by concentrations only), but has normal growth in the presence of the DMSO solvent only, P-EcH3 only (32 µM), or the cell-penetrating (KFF)₃K-GSG (KFF) peptide only (32 µM). E) Intracellular EcMetAP in E. coli was inhibited in a concentration dependent manner by P-EcH3 linked with the (KFF)₃K-GSG cell-penetrating peptide (KFF-EcH3). Both (C, E) were normalized to the maximum of 100% according to the control cell sample in the absence of peptide. F) Seminative and denaturing SDS-PAGE and Western blot analysis in the presence and absence of KFF-EcH3, along with the corresponding Coomassie Blue staining results demonstrated the ability of KFF-EcH3 to reduce 1 intracellular state of EcMetAP. G) No development of resistance to the KFF-EcH3 peptide (circles) was seen in E. coli over 30 d of consecutive passage in the presence of 0.5× MIC KFF-EcH3. In contrast, a 512-fold increase in MIC was seen in response to the control antibiotic enrofloxacin (open squares).

KFF-EcH3 mediates inhibition of intracellular EcMetAP and the growth of E. coli

To facilitate entry of the P-EcH3 peptide into the bacterial cell, we added a cell-permeating peptide (KFF)₃K [known to penetrate E. coli (35) and mammalian cells (36) without cytotoxicity] and a flexible linker Gly-Ser-Gly (GSG) to the N-terminus of P-EcH3 (named KFF-EcH3 for simplicity). The growth of E. coli K-12 MG1655 was inhibited by ≥4 µM KFF-EcH3 and partially inhibited by 2 µM during a 10-h culture (Fig. 1D). In contrast, P-EcH3 or (KFF)₃K-GSG alone at 32 µM has no effect on E. coli growth. KFF-EcH3 has a minimum inhibition concentration (MIC) of 4 µM for K-12 MG1655 (Supplemental Fig. S1). Plating of samples from the MIC assay onto peptide-free medium indicated that KFF-EcH3 is bactericidal, with a MBC of 4 µM [Supplemental Table S2; MBC was based on a concentration that reduces initial bacterial inoculum by ≥99.9% (21)]. The same KFF-EcH3 MIC of 4 µM was observed for clinical isolates of uropathogenic E. coli (UPEC) CFT073, enterotoxigenic E. coli B2C, enterohemorrhagic E. coli EDL933, and the MDR strain, UPEC isolate EC958 (Supplemental Fig. S2A–D). All of these strains have identical sequences of EcMetAP. These experiments show that KFF-EcH3 inhibits bacterial growth and survival.

The effect of KFF-EcH3 on intracellular EcMetAP activity was measured by incubating whole-cell bacteria with the peptide, followed by a wash to remove extracellular peptide, cell lysis, and enzymatic inhibition assays, as above. Figure 1E shows the concentration-dependent inhibition of intracellular EcMetAP by KFF-EcH3 in E. coli K-12 MG1655 cells. The IC₅₀ for cellular level EcMetAP inhibition is ∼10 µM. To further demonstrate the direct intracellular interaction of KFF-EcH3 with EcMetAP, whole-cell bacteria were once again incubated with the peptide, followed by a wash to remove extracellular peptide, and then whole-cell lysates were subjected to PAGE under denaturing or seminative conditions, followed by Western blot with an anti-EcMetAP antibody. In denaturing conditions, 1 fully-denatured EcMetAP band was observed when cells were incubated in the presence or absence of peptides (Fig. 1F). However, under seminative conditions, several EcMetAP bands were present (representing various denaturation states), and the intensity of one of the bands with high apparent MW was significantly reduced by peptide treatment. Together, these data confirm that KFF-EcH3 directly interacts with EcMetAP inside the cell and results in loss of activity.

KFF-EcH3 specifically targets the EcMetAP and does not display off-target toxicity

To demonstrate the specificity of the KFF-EcH3 peptide for the EcMetAP, we tested inhibition of other Gram-negative bacteria, and of E. coli–overexpressing MetAP. KFF-EcH3 was tested against N. gonorrhoeae strain 1291 and showed no effect on cell growth up to 20 µM (Supplemental Figs. S3 and S4) even though EcMetAP and NgMetAP share as high as 57% identity and 73% similarity in their sequences (but only 20% identity in Helix 3, Supplemental Table S1). Similarly, KFF-EcH3 did not inhibit the growth of Neisseria meningitidis strain MC58 (58% sequence identity to EcMetAP), Haemophilus influenzae strain 2019 (65% sequence identity), or Moraxella catarrhalis strain 25239 (58% sequence identity) (tested up to 16 µM; Supplemental Fig. S4). On the other hand, by overexpressing EcMetAP in E. coli, an increased concentration of peptide was needed to inhibit growth, with an increase in the MIC of KFF-EcH3 from 4 µM (E. coli with empty pFLAG vector) to 16 µM (E. coli with pFLAG-MetAP) (Supplemental Fig. S5).

To further demonstrate specificity and display that MetAP inhibition by KFF-EcH3 is not related to nonspecific toxicity mediated by the KFF cell-permeating portion of the peptide, we generated a shuffled control peptide KFF-sEcH3. We divided P-EcH3 into 4–6 aa blocks and shuffled the blocks randomly (P-sEcH3, see Supplemental Table S1 for the sequence) so that the control peptide maintained the amino-acid composition and local sequence features of EcH3. The shuffled P-sEcH3 peptide is unable to inhibit the enzymatic activity of EcMetAP (Fig. 1C), and KFF-sEcH3 does not inhibit growth of E. coli K-12 MG1655 (up to 32 µM) (Supplemental Fig. S6).

The KFF-EcH3 peptide did not display any toxicity (tested up to 20 µM) to 3 human cell lines (ME180, A549, and Caco-2 via Sytox Live/Dead staining; Thermo Fisher Scientific) (Supplemental Fig. S7, and via visual scoring of cell morphology). EcMetAP had 47% identity to human MetAP11, 39% to human MetAP12, and 21% to human MetAP2.

No resistance to the inhibitory peptide KFF-EcH3 is detected

We monitored the resistance of E. coli to KFF-EcH3 by serially passaging E. coli in the presence of 0.5× to 2× MIC (2–8 µM initially) of KFF-EcH3 or that of the control antibiotic enrofloxacin (0.25–1 µM, initially) over a period of 30 d. The MIC of KFF-EcH3 did not change, whereas the MIC of enrofloxacin increased 512-fold (from 0.5 to 256 µM) over 30 d (Fig. 1G). This result shows that E. coli was unable to develop resistance against KFF-EcH3 during the 30-d experimental time period.

Inhibition is caused by disruption or unfolding of target structure

Inhibition of enzyme activity and lack of bacterial resistance to the self-derived peptide may be caused by the disruption of the target’s structure, as suggested by Western blot analysis (Fig. 1F). One expects that resistance is less likely to develop in structure-disrupting antibiotics, because protein cores are highly conserved, and mutations in the structural core are likely to destabilize, rather than stabilize, the whole structure and result in loss of activity, unless complementary mutations evolve at the same time. To confirm structure disruption, we monitored the exposure of hydrophobic residues in EcMetAP in the presence and absence of P-EcH3, using the Thermofluor temperature-denaturation assay (37). EcMetAP alone, or EcMetAP with 20 µM control peptide P-EcS12, have a low level of hydrophobic exposure between 20 and 40°C and a sharp unfolding transition at ∼48°C (Fig. 2A). In contrast, EcMetAP with 20 µM P-EcH3 peptide has a much high level of hydrophobic exposure even at 20°C, with a gradual unfolding transition up to ∼48°C. This result suggests that P-EcH3 has already caused structural unfolding of EcMetAP at room temperature. Indeed, the CD spectrum of EcMetAP at 25°C was mostly unchanged in the presence of P-EcS12 but changed significantly when EcMetAP is mixed with P-EcH3 (Fig. 2B). According to the program K2D3 (23), which estimates secondary structure contents from CD spectra, the helical content of EcMetAP is reduced in a concentration-dependent manner from 18% to 16, 11, or 2 in the presence of 5, 10, or 20 µM P-EcH3, respectively, but is unchanged in the presence of 20 µM P-EcS12 (18% vs. 18%).

Figure 2 .

Self-targeting by structure disruption. A) Temperature denaturation indicates that EcMetAP no longer had a cooperative folding/unfolding transition in the presence of P-EcH3 (no effect is seen with 20 μM control peptide P-EcS12). B) CD spectrum of EcMetAP only (black), in the presence of 5, 10, and 20 μM P-EcH3 and 20 μM P-EcS12 (blue), respectively. A significant change in the CD spectrum accompanied with loss in helical structures according to program K2D3 is observed when various concentrations of P-EcH3 as labeled were added. C) ¹H-¹⁵N HSQC NMR spectrum of 20 μM ¹⁵N-labeled EcMetAP (blue) and 20 μM ¹⁵N-labeled EcMetAP in complex with 0.5 mM P-EcH3 (red). D) Same as in C, but P-EcH3 is replaced by P-EcS12. Disappearance and broadening in ¹⁵N-chemical shifts of EcMetAP indicated denaturation in the presence of P-EcH3 but not P-EcS12. E) The binding free-energy landscape between P-EcH3 and EcMetAP projected onto the number of contacts and the distance between them by MD simulations (expressed in kilojoules per mole). F) The representative structure of the largest cluster within the deepest basin with P-EcH3 colored in cyan and H3 colored in green as in Fig. 1A.

We further investigated possible tertiary structural changes of EcMetAP using ¹H-¹⁵N HSQC NMR spectroscopy. Folded proteins have an intrinsic ¹⁵N and ¹H chemical shift dispersion, whereas denatured spectra show less dispersed and largely overlapped ¹H-¹⁵N HSQC signals (38). The ¹H-¹⁵N HSQC signals of EcMetAP had a significant reduction in the number of peaks in the presence of P-EcH3 (Fig. 2C), but not in the presence of negative control peptide P-EcS12 (Fig. 2D). These data confirm that P-EcH3 inhibition of EcMetAP is caused by structural disruption or denaturation of the target enzyme.

A strong interaction between the self-targeting peptide and target protein structural core is key for structure disruption

To examine the mechanism of structure-disrupting peptides at the atomic level, we performed all-atom MD simulations. Structure disruption requires large conformational changes of proteins that are too time consuming to simulate by using regular MD simulation techniques, because free-energy barriers separate different protein conformations. We used an advanced sampling method named metadynamics (29), which accelerates barrier-crossing events by periodically adding repulsive gaussian potentials to smooth the normally rugged free-energy landscape along the predefined coordinates (collective variables). The free-energy landscape of the simulated system with the smoothing potentials can be calculated afterward by removing their effects computationally. We obtained the free-energy landscape of EcMetAP mixed with its self-derived inhibitory peptide (P-EcH3 or P-EcH1) or noninhibitory peptide (P-EcH2) along with 2 collective variables: the distance between the center of mass of the peptide and that of EcMetAP and the number of residue-residue contacts between them (defined by the Cα-Cα distance <5 Å).

Figure 2E shows that the free energy for the mixture of P-EcH3 and EcMetAP has the lowest free-energy minimum (dark blue) centered at ∼65 contacts, with a distance of ∼3 nm between them. Figure 2F displays the structure of the complex between EcMetAP and P-EcH3 in the cluster center of the free-energy minimum. It shows that helices H1, H3, and H4 are partially denatured [a 44% reduction of helical content from the native structure based on the secondary structure assignment by using the Dictionary of Protein Secondary Structure, (39)]. The partial denaturation of these helices allowed the interaction of P-EcH3 with the structural core of EcMetAP that is buried in the native conformation. In particular, we found that the native contacts between EcH3 and the C-terminal region of EcMetAP (EcS12 in blue) were partly replaced by the same contacts between P-EcH3 and the C-terminal.

By comparison, the global free-energy minimum of the mixture of P-EcH1 and EcMetAP is more difficult to reach at a shorter distance of 1.5 nm (Supplemental Fig. S8A), compared to 3 nm for the mixture of P-EcH3 and EcMetAP. Moreover, the free-energy minimum is located at less interaction (only ∼20 contacts) between P-EcH1 and EcMetAP, compared to 65 contacts between P-EcH3 and EcMetAP. Supplemental Fig. S9A further shows that the free-energy minimum of the mixture of P-EcH2 and EcMetAP has little contact between the 2 molecules. The representative conformations in the largest structural cluster showed that P-EcH3 is considerably more structure disruptive (Fig. 1F) than either P-EcH1 (Supplemental Fig. S8B) or P-EcH2 (Supplemental Fig. S9B). These simulation results indicate that EcMetAP has the strongest interaction with P-EcH3, followed by P-EcH1 and P-EcH2. This trend is consistent with experimental results of P-EcH3 as a strong inhibitor, P-EcH1 as a weak inhibitor and P-EcH2 as a noninhibitor of EcMetAP (Fig. 1B).

Identification of structure-disruption peptides by a statistical energy function

The above simulation suggests that the ability to substitute and interact with the residues buried in the core of EcMetAP is key for structure disruption. In other words, the stronger the interaction between a helical segment and the remaining portion of MetAP, the higher the probability that the same-sequence peptide can break into the folding core of MetAP to substitute the original helix during the dynamic opening motion of the protein. We computed the interaction energy of the peptide segment (EcH1, -H2, -H3, and -H4) with the remaining portion of EcMetAP, according to a statistical energy function based on the distance-scaled, finite, ideal-gas reference (DFIRE) state (40, 41). Indeed, the interaction energy is the strongest for EcH3, followed by EcH1, -H4, and -H2 (Supplemental Table S1). This trend is the same as the inhibition capacity of these peptides (P-EcH3 > P-EcH1 > P-EcH4 ≈ P-EcH2, Fig. 1B).

Identification of NgMetAP peptides and their inhibition of gonococcal growth and infection of epithelial cells

DFIRE analysis of MetAP of N. gonorrhoeae (NgMetAP) revealed that NgH1 had the strongest energy score (i.e., the strongest predicted interaction with the remaining portion of NgMetAP) followed by P-NgH3 (Supplemental Table S1). This result was different from EcMetAP, where EcH3 had the strongest predicted interaction, despite the relatively high sequence identity between EcMetAP and NgMetAP (57%) and between EcH1 and NgH1 (54%). Gonococcal growth was investigated in the presence of the corresponding self-derived peptides (P-NgH1 and P-NgH3) after linkage with (KFF)₃K-GSG (KFF-NgH1 and KFF-NgH3, respectively). KFF-NgH1 inhibited bacterial cell growth (Fig. 3A) with an MIC of 2 µM for N. gonorrhoeae strain 1291 (Supplemental Fig. S10A). Slightly higher MIC values of 4 µM were observed for the MDR strains WHO-K, -L, and -P (Supplemental Fig. 10B–D), which had an up-regulated efflux pump (31). KFF-NgH3, on the other hand, did not inhibit cell growth of N. gonorrhoeae (Supplemental Fig. S11), consistent with NgH1 having the strongest interaction energy score (Supplemental Table S1). As a control experiment for the specificity of the peptide, we investigated the effects of KFF-NgH1 on the cell growth of E. coli K-12 MG1655 and found that it was unable to inhibit cell growth at the highest concentration tested (20 µM) (Supplemental Fig. S12).

Figure 3 .

KFF-NgH1 inhibits N. gonorrhoeae growth and infection. A) N. gonorrhoeae growth was completely inhibited by P-NgH1 linked to the (KFF)₃K-GSG cell-penetrating peptide at concentrations ≥5 µM, but has normal growth in the presence of the DMSO solvent only (green), or the cell-penetrating (KFF)₃K-GSG (KFF) peptide only (black). B) Complete inhibition of N. gonorrhoeae infection of human cervical epithelial cell was seen at 6 and 12 h in the presence of 10 and 20 µM KFF-NgH1, relative to no peptide control and KFF only. N.D., no detectable CFUs.

We also performed TCX infection assays with N. gonorrhoeae and observed a 100% inhibition of infection at 6- and 12-h time points in the presence of 10 and 20 µM KFF-NgH1, respectively, relative to assays of no peptide control and KFF only (Fig. 3B). Similar results were seen, whether the peptide KFF-NgH1 was added at the time of infection, or 1 h after infection.

DISCUSSION

We have described a prototype of a newly identified class of antibiotic peptide. This peptide is derived from a polypeptide segment of an essential protein involved in structural core interactions. Incubation of the peptide with the target protein leads to disruption of structure and loss of function. Computational evaluation of interaction energies provides an effective method for identification and rational design of self-targeting peptides that can partially denature their target protein. For the MetAP target, fusing the self-targeting peptide to a cell-penetrating peptide enables the peptide to enter the bacterial cell and leads to species-specific inhibition of cell growth. This result was shown for E. coli with the peptide KFF-EcH3 and for N. gonorrhoeae with the peptide KFF-NgH1.

Our hypothesis that the self-targeting peptides penetrate the bacteria, specifically target MetAP, and disrupt its structure, was supported by a series of independent and complementary findings in E. coli. The ability of the peptide to penetrate the cell is supported by the finding that the MetAP-targeting peptide inhibited the growth of E. coli only when it was linked to the cell-permeating peptide (i.e., KFF-EcH3), whereas the unlinked peptide alone or the cell-permeating peptide alone [P-EcH3 or (KFF)₃K-GSG, respectively] had no effect on growth. The shuffled control peptide (KFF-sEcH3), which has the same amino acid composition and local sequence features of P-EcH3 and is linked to the cell-permeating peptide, also had no effect on E. coli growth. In addition, the specificity of the interaction between KFF-EcH3 and MetAP of E. coli is highlighted by the finding that overexpression of EcMetAP in E. coli increased the MIC of KFF-EcH3 4-fold. However, KFF-EcH3 had no effect on the growth of 4 other Gram-negative bacteria (MetAP homologs with 57–65% sequence identity to EcMetAP), and the KFF-NgH1 peptide derived from the gonococcal NgMetAP had no effect on the growth of E. coli. Cell permeation and EcMetAP specificity of KFF-EcH3 are further supported by the inhibition of intracellular MetAP enzymatic activity, as well as reduction of one of its intracellular states in seminative Western blot analysis. These 2 experiments were performed with E. coli incubation, with the peptide then washed thoroughly before cell lysis, centrifugation, and analysis, to ensure the removal of any extracellular or membrane-bound peptide and allow the assessment of intracellular peptide activity only. Finally, the ability of P-EcH3 to interact with and disrupt the structure of EcMetAP was supported by biophysical characterization according to temperature-dependent exposure of hydrophobic residues and secondary and tertiary structure characterization, using ThermoFluor temperature denaturation, CD spectrum analysis, and NMR spectroscopy. Computational simulations and energetic calculations of the interactions between P-EcH3 and EcMetAP also supported the disruption of the structure of EcMetAP. The combined data strongly support that direct, intracellular interaction between KFF-EcH3 and EcMetAP is responsible for the bactericidal activity of KFF-EcH3, rather than the extracellular aggregation and nonspecific effects of the peptide. However, additional studies are needed to confirm the intracellular disruption of MetAP structure and the mechanism of action in vivo.

It has been suggested that such structure-disrupting peptides are less susceptible to resistance because structure-stabilizing core residues are highly conserved (i.e., not as mutable as surface residues) and mutations in structural cores are often destabilizing and lead to misfolding and aggregation (14, 15). In this work, we have demonstrated for the first time, to our knowledge, that the MIC of self-derived peptide KFF-EcH3 does not change, whereas the MIC of enrofloxacin increases 512-fold (from 0.5 to 256 µM) over 30 d. This result confirms that bacteria are more difficult to induce resistance against structure-disrupting peptides.

Moreover, targeting nonactive sites makes self-derived peptides more specific because active sites of proteins such as MetAP are often conserved between bacteria and humans (42). In other words, structure-targeting peptides will be highly specific. In this study, we developed 2 species-specific peptides by targeting a region of MetAP that is not perfectly conserved between E. coli and N. gonorrhoeae. Species specificity was confirmed by examining KFF-EcH3 against several other Gram-negative species. Such species-specific peptides will target pathogens only, and make minimum disruption to the wider microbiome (43). However, we can control the species specificity of the peptides. A highly conserved protein target across different bacterial species will provide the opportunity for a broadly effective antimicrobial peptide.

To summarize, all living organisms produce antimicrobial peptides as part of their innate immune response (44). In the field of therapeutics, peptide-based drugs have gained increased attention because of high specificity and low toxicity (8), as well as advancements in delivery (9), synthesis (10), and stabilization (11). Given the fact that each bacterium has hundreds of essential proteins, the approach described herein opens up an untapped reservoir for further investigation of many potential structure disrupting peptide antibiotics.

Supplementary Material

This article includes supplemental data. Please visit http://www.fasebj.org to obtain this information.

Glossary

CFU

colony-forming unit

CV

collective variable

DFIRE

distance-scaled, finite, ideal, reference state

EcMetAP

methionine aminopeptidase of Escherichia coli

HSQC

heteronuclear single quantum coherence

IC₅₀

half maximal inhibitory concentration

MBC

minimum bactericidal concentration

Met-AMC

methionine aminomethylcoumarin

MetAP

methionine aminopeptidase

MIC

minimum inhibitory concentration

NgMetAP

methionine aminopeptidase of Neisseria gonorrhoeae

RDM

rich defined medium

TCX

transformed cervical epithelial

TBST

Tris-buffered saline with Tween

UPEC

uropathogenic E. coli

AUTHOR CONTRIBUTIONS

J. Zhan and H. Jia designed and conducted all biophysical experiments and cell inhibitions associated with all E. coli strains (except MDR UPEC EC958), collected and analyzed the data, and wrote the manuscript; E. A. Semchenko and K. L. Seib designed and conducted all experiments and analyses associated with N. gonorrhoeae, N. meningitidis, H. influenzae, and M. catarrhalis and the human cell line and wrote the manuscript; Y. Bian and J. Wang performed MD simulations and data analysis and wrote the manuscript; A. M. Zhou performed the initial cell growth inhibition assay of E. coli; Z. Li and Y. Yang performed bioinformatics analysis; S. Sarkar and M. Totsika performed cell inhibition assay of E. coli strain MDR UPEC EC958 and data analysis; Q. Ye contributed to purification of EcMetAP and enzymatic activity assay; T. Haselhorst contributed to NMR HSQC data acquisition and analysis; H. Blanchard and F. E.-C. Jen contributed to structural analysis of EcMetAP; M. P. Jennings contributed to experimental design and composition of the manuscript; Y. Zhou conceived of the study, participated in its design, assisted in data analysis, and drafted the manuscript; and all authors read and approved the final manuscript.