Combating virulence of Gram-negative bacilli by OmpA inhibition

Preventing the adhesion of pathogens to host cells provides an innovative approach to tackling multidrug-resistant bacteria. In this regard, the identification of outer membrane protein A (OmpA) as a key bacterial virulence factor has been a major breakthrough. The use of virtual screening helped us to identify a cyclic hexapeptide AOA-2 that inhibits the adhesion of Acinetobacter baumannii, Pseudomonas aeruginosa and Escherichia coli to host cells and the formation of biofilm, thereby preventing the development of infection in vitro and in a murine sepsis peritoneal model. Inhibition of OmpA offers a strategy as monotherapy to address the urgent need for treatments for infections caused by Gram-negative bacilli.

We performed an initial computational analysis with the SiteMap module of Schrodinger Suite 24 and identified a potentially druggable area for ligand binding in a cavity formed by the extracellular loops of the N-terminal transmembrane (TM) domain of OmpA. We hypothesized that the size and shape of this predicted druggable area of OmpA may allow for accommodation of a hexapeptide ligand. Considering that cyclization increases the proteolytic resistance and restraints the conformational flexibility of peptide sequences, we designed a virtual library of 26 C2-symmetric cyclic hexapeptides as potential OmpA-binders (Table 1). All cyclic hexapeptides contained two fixed proline residues to facilitate cyclization. This peptide library was subsequently computationally screened against the TM domain of OmpA of both E. coli and A. baumannii and ranked on the basis of the scoring function (see "Methods" section). As no three-dimensional structure of the TM domain of A. baumannii OmpA was found in the Protein Data Bank (PDB), a homology model using the i-TASSER server was built using the structures of the homologous domain of E. coli (PDB entries 1G90, 2GE4, 1QJP, and 1BXW) as template. Similar docking results were obtained for the OmpA proteins of both pathogens. Notably, all peptides containing tryptophan (Trp) and arginine (Arg) residues showed better in silico activity compared to other hexapeptides ( Table 1). The peptide-protein complex models generated by docking suggested that the best-scoring peptides bind in the same protein region and adopt similar orientations in the interaction with both proteins (Fig. S1). One of the Trp residues of the peptide acts as an anchor and inserts in the border of the β-barrel, while the remaining peptide residues contact with the extracellular flexible loops. The possible contact of the peptide with extracellular loops could affect the function of OmpA by reducing their interaction with host cells and consequently bacterial pathogenesis. It is well known that inhibition and mutation of the extracellular loops of OmpA affect the pathogenesis and the interaction of E. coli with host cells 25,26 .
To experimentally validate the virtual screening, we synthesized a set of seven peptides ( Table 2, Fig. S2). Four of them (AOA-1 to AOA-4) were Arg-Trp-containing peptides in which the position and/or stereochemistry of the amino acid residues were modified. These included the best-scoring docking peptides (AOA-1 and AOA-2). Additionally, two peptides were synthesized in which the positively charged arginine residue was replaced by a neutral hydrophilic serine (AOA-5) or by a negatively charged glutamic acid (AOA-6). As a control, the linear version (SXV4) of the in silico most active peptide was also synthesized.
In vitro effect of OmpA inhibitor peptides on A. baumannii growth and contact with host cells. In the initial screening, the bactericidal activity of the synthetic peptides and their toxicity on human lung epithelial cells (A549) were evaluated. We confirmed that these peptides did not show bactericidal or cytotoxic activities, compared with colistin, as the positive control peptide (Tables S1 and S2), and therefore fulfill all the features to be only potential blockers of OmpA, without killing either bacteria or cells. &Gly-Pro-Ala-Gly-Pro-Ala& −2.9 Table 1. Docking score of the virtual library of C2-symmetric cyclic hexapeptides. & symbol means cyclic peptides The ability to reduce A. baumannii adherence to A549 cells was tested in all the compounds synthesized. Among them, peptide AOA-2 (&Trp-D-Pro-Arg-Trp-D-Pro-Arg&) presented the greatest ability to reduce the A. baumannii adherence to A549 cells (Fig. S3).
NMR detection of AOA-2 binding to OmpA. Considering that AOA-2 provided analogous docking results with the E. coli and A. baumannii OmpA homologs, we hypothesized that E. coli OmpA represents an optimal model system for in vitro characterization of the intermolecular interaction of AOA-2 with OmpA proteins. E. coli OmpA has been extensively characterized and the protocol to produce pure and functional E. coli OmpA has been previously reported 21 .
We assessed the binding of AOA-2 to E. coli OmpA by solution NMR. As an indication of binding, the presence of OmpA solubilized in 1 mM Triton X-100 causes signal broadening and chemical shift changes in the proton resonances of AOA-2. The 1D 1 H NMR spectrum of AOA-2 was unaltered by the presence of 1 mM Triton X-100, demonstrating that the perturbations were induced by OmpA (Fig. 1A).
The peptide-protein interaction was also clearly observable by Saturation Transfer Difference (STD) NMR experiments of AOA-2 (750 μM) in the presence of OmpA (8 μM). The resulting STD spectrum was obtained by subtracting a spectrum with irradiation at −0.75 ppm, which selectively saturated the protein, from an analogous spectrum with irradiation at 50 ppm, which did not perturb any AOA-2 or OmpA resonances. The STD spectrum reveals signals of the binding peptide that have been perturbed by saturation transfer from the protein (Fig. 1B).
Binding was further confirmed by pulsed field gradient (PFG) NMR-based diffusion experiments. In these experiments, PFG causes a decrease in signal intensity that depends on the gradient strength, and on the translational diffusion of the observed molecule, and is more pronounced for rapidly diffusing, small molecules. As shown in Fig. 1C-E, the attenuation of AOA-2 signals was smaller in the presence of the protein indicating that AOA-2 is interacting with OmpA and then has a slower translational diffusion.
In vitro effect of AOA-2 against GNB contact with host cells. AOA-2 ( Fig. 2A,B) was selected for the following studies on the basis of its greater ability to reduce the bacterial adherence (Fig. S3). This peptide showed a reduction in the adherence of A. baumannii to A549 cells by more than 60% in vitro using 0.25 mg/ mL of the compound (Fig. 2C). Due to highly conservation of OmpA among most of the GNB, the same assay using P. aeruginosa and E. coli was performed also observing a significant reduction in adherence (Fig. 2C), and without presenting bactericidal activities even at the highest AOA-2 concentration tested (Fig. S4). It is important to note that AOA-2 at concentrations (0.25-1 mg/mL) presents different bacteriostatic activities on different bacteria. Thus, it appears to have high anti-adherence/poor bacteriostatic activity on A. baumannii, and low anti-adherence/relevant bacteriostatic activity on P. aeruginosa, with an intermediate behaviour in the case of E. coli.
We previously reported that ompA-deficient A. baumannii is less adherent to abiotic surfaces and disseminate less between murine organs 14,27 and that recombinant OmpA shows high affinity for fibronectin, an extracellular matrix protein 28 . Due to this fact, assays to test the activity of this peptide blocking the interaction between fibronectin and bacteria (A. baumannii, P. aeruginosa and E. coli) were performed, and the results showed a significant decrease in the bacterial adherence to fibronectin after treatment with AOA-2 ( Fig. 2D,E). In the same line, other approach using D-amino acids has been performed recently to assess their antivirulence activity against A. baumannii and P. aeruginosa 29 , and showed that some of them can inhibit both pathogens adherence to eukaryotic cells and protect against infection of these eukaryotic cells with P. aeruginosa 29 .

AOA-2 effect on biofilm formation by GNB. Many reports have demonstrated the involvement of
OmpA in biofilm formation by GNB 10,27,30,31 . These reports showed that ompA-deficient A. baumannii and E. coli did not produce biofilm 10,27,31 . Here, we showed that treatment with AOA-2 diminishes the biofilm formation of all the reference strains and clinical isolates of A. baumannii, P. aeruginosa, and E. coli (Table 3). Bacterial biofilms have significant impacts in industrial and clinical settings and there is therefore an urgent need to develop new compounds to prevent biofilm formation. Our data with AOA-2 showed remarkable results and are consistent with those previously reports for other inhibitors 29,32,33 . In vitro and in vivo effect of AOA-2 against GNB virulence. We have proved using different assays that AOA-2 is able to block the interaction between different bacteria and host cells, however the real aim of this compound is to prevent the cell death. The presence of 0.25 or 0.5 mg/mL AOA-2, prevented cell death dependent on bacterial adherence 10,34 , with the exception of E. coli strain ATCC 25922 (Fig. 3A). Interestingly, AOA-2 has  Table 2. List of synthesized hexapeptides based on the computationally screened library.
no effect on cell death caused by ompA-deficient A. baumannii which confirm that AOA-2 specifically acts by inhibiting OmpA function (Fig. S5). The failure of AOA-2 against E. coli may be attributable to the presence of other highly virulent factors circumventing the loss of OmpA 35 , acting differently to those of P. aeruginosa and A. baumannii. The difference in activity between A. baumannii and the other two bacterial strains is due to the fact that the compound was designed specifically to be active against A. baumannii OmpA; however, it is also active against other GNBs due to the high sequence homology of the OmpA protein, therefore we may think that better results could be achieved for both E. coli and P. aeruginosa, by using its own OmpA or homologous protein.
We have demonstrated that AOA-2 reduced the interaction of GNB with host cells and prevented their death in vitro, and further studies in mice did not reveal any evidence of toxicity at a dose ≤40 mg/kg (Table S3). To prove the bioavailability of AOA-2 when administered i.p. to mice pharmacokinetic studies indicated that AOA-2 showed good activity (Fig. S6).
In a peritoneal sepsis murine model, administration of 10 mg/kg AOA-2 to mice 2 h after intraperitoneal (i.p.) challenge with minimal lethal doses of A. baumannii ATCC 17978, P. aeruginosa PaO1, and E. coli ATCC 25922 strains reduced the bacterial load in the spleen (4.54, 4.04, and 3.36 log CFU/g, respectively) and lungs (4.34, 4.29, and 3.47 log CFU/g, respectively) (Fig. 3B). In addition to this significant decrease of the bacterial load positive blood cultures and mouse mortality were also decreased (Fig. 3C). The possible explanation of the good therapeutic effect of AOA-2 could be the fact that AOA-2 would affect the interaction of bacteria with host cells which prevents their evasion from the immune host defence. These results are consistent with previous observations that absence of OmpA in A. baumannii and E. coli is associated with the loss of their virulence in vivo by reducing their dissemination between organs and animals mortality 14,15 . In both panels, diffusion gradients were 2% (bottom) and 98% (top) of their maximum value (≈55 G/cm). (E) Ratio of the intensity (100*I G /I 2% ) of the AOA-2 signal resonating at ca. 7.50 ppm for a given diffusion gradient strength (I G ) with respect to that measured (I 2% ) with the weaker diffusion gradient (2%). Peak intensity ratios were obtained in the absence (black squares) and in the presence of 8 μM OmpA (grey circles). As an indication of binding, AOA-2 appeared to diffuse slower (smaller signal attenuation) in the presence of OmpA than in isolation.
SCIEnTIFIC REPORtS | 7: 14683 | DOI:10.1038/s41598-017-14972-y So far, we have reported a cyclic hexapeptide (AOA-2) able to inhibit the GNB adherence to host cells and, consequently, an increase in the cell viability. At the same time, this peptide shows a potent inhibitor activity on the biofilm formation by these GNB. AOA-2 in vivo administration shows a large decrease in the bacterial loading in both spleen and lungs together with a significant decrease in mouse mortality, especially in A. baumannii.
In summary, this drug discovery program must be considered as an initial stage of the development of a novel class of antimicrobial agents. The efficacy of AOA-2 can be further tested in experiments administering this new peptide in combination with antimicrobial agents used in the clinical setting, with the aim to improve the therapeutic efficacy in severe infections, avoiding their mortalities despite the currently available antimicrobial therapies.

Methods
Bacterial strains. The list of bacterial strains was regrouped in Table S4. Docking and molecular modelling. All molecular docking calculations, and preparation of protein and ligand systems were done with SCHRODINGER software package 24 . Protein structures of the transmembrane domain of A. baumannii OmpA were obtained by homology modeling. Homology modeling constructs a model of the "target" protein from its amino acid sequence and an experimental three-dimensional structure (X-ray or NMR) of a related homologous protein ("template"). I-TASSER online server (http://zhanglab.ccmb.med.umich. edu/I-TASSER/) was used for homology modeling 36 . The amino acid sequence was taken from Genbank, and template structures used in the modeling were taken from the E. coli, OmpA transmembrane domain structures presented in the Protein Data Bank (PDB) depository. Two PDB entries obtained from NMR (1G90 and 2GE4), and two X-Ray crystals (1QJP, 1BXW) were used in all homology models. Best-ranked models were prepared with Protein Preparation Wizard, minimized 24 , and further relaxed including short (5 ns) explicit solvent molecular dynamics (MD) simulations with Desmond software 37 . The SiteMap module of the Schrodinger Suite 24 identified a potentially druggable area for ligand binding in the border of opening of the transmembrane domain of OmpA and flexible extracellular loops. Same area was used for constructing docking grids for Glide. A library of cyclic hexapeptides was used for virtual screening. All peptide structures were built, optimized by LigPrep 24 and additional MacroModel conformational search techniques were applied to all molecules. Thus, for each molecule of the library were obtained 10 to 20 different molecular conformations. All received conformations were docked using XP Glide protocol and best-scored results for each hexapeptide were stored.

Protein. A recombinant 37 kDa polypeptide fragment containing the transmembrane and periplasmatic
domains of the E. coli OmpA protein was purchased from NovoPro (China). OmpA samples contained Triton X-100 to mimic the membrane environment.

Nuclear Magnetic Resonance (NMR).
All NMR experiments were recorded at 25 °C on a Bruker Avance 800 MHz spectrometer equipped with a cryoprobe. Unless otherwise is stated, AOA-2 samples (750 μM) were prepared on 10 mM sodium phosphate buffer (pH 7.3), 100% D 2 O, 0.01% NaN 3 , 1 mM Triton X-100 in the absence or in the presence of OmpA (8 μM). Saturation transfer difference 38 (STD) experiments were obtained using a pseudo-2D pulse sequence for the interleaved acquisition of on-resonance and off-resonance spectra. Total saturation time was 3 seconds. Suppression of the residual water signal was achieved by excitation sculpting 39 . The final STD spectrum was obtained by subtracting the on-resonance spectrum, which saturates the protein (irradiation at −0.75 ppm), from the off-resonance one (irradiation at 50 ppm), which does not affect any ligand or OmpA resonances.
Diffusion NMR experiments were performed using the LED pulse sequence 40 with gradient lengths of 3 ms and a diffusion delay of 110 ms. The total inter-scan delay was 4.5 s. A series of experiments applying diffusion gradients of 2%, 50%, 75%, and 98% of its maximum value (55 G cm −1 ) were acquired for AOA-2 (750 μM) in 10 mM sodium phosphate buffer (pH 7.3), 100% D 2 O, 0.01% NaN 3 , 1 mM Triton X-100 in the absence and in the presence of OmpA (8 μM).
Cell culture and infection. Human type II pneumocyte cell line A549 derived from a human lung carcinoma were obtained from American Type Culture Collection (LGC, UK) and were grown in DMEM medium (Invitrogen, Spain) supplemented with 10% heat-inactivated fetal bovine serum, vancomycin (50 μg/mL), gentamicin (20 μg/mL), amphotericin B (0.25 μg/mL) (Invitrogen, Spain) and 1% HEPES (Invitrogen, Spain) in a humidified incubator, 5% CO 2 at 37 °C. A549 cells were routinely passaged every 3-4 days. The cells were seeded 24 h in 96-well plates for cellular viability assay, and in 24 well plates for adhesion and immnunofluorescence assays. Cellular viability. Peptide toxicity. A549 cells were incubated with AOA-1, AOA-2, AOA-3, AOA-4, AOA-5, AOA-6, or negative control peptide SXV4 (0, 0.25, 0.5 and 1 mg/mL) 48 h with 5% CO 2 at 37 °C. Prior the evaluation of the peptides cytotoxicity, A549 cells were washed three times with prewarmed PBS. Then, peptides cytotoxicity was initially assessed quantitatively by monitoring the mitochondrial reduction activity using the MTT assay as described previously 27 . The percentage of cytotoxicity was calculated from the optical density (OD) as follow: [(OD of treated cells / mean OD of non-treated cell) × 100].
Bacterial cytotoxicity. A549 cells were infected with 10 8 CFU/mL of A. baumannii ATCC 17978, P. aeruginosa PaO1, and E. coli ATCC 25922 strains pretreated with AOA-2 or SXV4 (0, 0.25 and 0.5 mg/mL, 30 min) for 24 h with 5% CO 2 at 37 °C. Prior the evaluation of bacterial cytotoxicity, we firstly removed viable bacteria from A549 cells cultures and washed A549 cells five times with pre-warmed PBS. Then, cellular viability was assessed as indicated above.
Fibronectin-binding assay. Fibronectin-binding assay was performed as described previously 28 . Briefly, A. baumannii ATCC 17978, P. aeruginosa PaO1, and E. coli ATCC 25922 grown overnight at 37 °C in LB were resuspended in PBS and collected by centrifugation at 5,000 × g for 10 min. Bacteria were washed twice in sterile PBS, resuspended in the same sterile buffer, and incubated with AOA-2 or SXV4 (0, 0.25 and 0.5 mg/mL, 30 min). Then, 50 µL of bacterial suspension were mixed with 50 µL of PBS, added to fibronectin coated wells and incubated 2 h at room temperature for bacterial adsorption. Non-adhered bacteria were discarded and wells were washed six times with sterile PBS to remove unbound bacteria. Adherent bacteria were then collected with 125 µL of sterile PBS containing 0.5% Triton X-100. Diluted lysates were plated onto sheep blood agar and incubated at 37 °C for 24 h for enumeration of developed colonies and then the determination of the number of bacteria that attached to fibronectin.
Biofilm. An abiotic solid surface biofilm formation assay was performed as described previously 27 . In brief, we used an overnight culture of 2 reference strains and 8 clinical isolates of A. baumannii, 1 reference strain and 8 clinical isolates of P. aeruginosa, and 1 reference strain and 6 clinical isolates of E. coli diluted 1:1 in fresh Luria Bertani (LB) in 96 well plates that were incubated in presence or not of 0.25 mg/mL AOA-2 without shaking at 37 °C during 24 h. Biofilm was stained with crystal violet 0.4% (v/v) and quantified at 580 nm after solubilization with ethanol 95%.
Animals. Immunocompetent C57BL/6 female mice (16-18 g) were obtained from University of Seville facility; they had a sanitary status of murine pathogen free and were assessed for genetic authenticity. Animals were housed in regulation cages with food and water ad libitum. This study was carried out in strict accordance with the recommendations in the Guide for the Care and Use of Laboratory Animals 41 . The protocol was approved by the Committee on the Ethics of Animal Experiments of the University Hospital of Virgen del Rocío, Seville (2012PI/238). All surgery was performed under sodium thiopental anaesthesia, and all efforts were made to minimize suffering.
AOA-2 in vivo toxicity. The Reed and Munch method was used 42 . Groups of 6 mice were intraperitoneally (i.p.) inoculated with a single dose of AOA-2, starting at 10 mg/kg, in 0.5 mL NaCl 0.9%, and the solution were serially duplicated until 100% mortality was reached. The maximum tolerated dose (LD0), the LD50, and LD100 were defined as the concentrations causing 0, 50, and 100% mortality, respectively. LD50 value was determined using the Probit method.
AOA-2 pharmacokinetics. Serum AOA-2 level was determined in healthy mice after a single i.p. administration of 10 mg/kg AOA-2. After 5, 15, 30, 60,120, 240, 480, and 1440 min, blood was extracted from the periorbital plexuses of the anesthetized mice; three mice were used for each time point. The AOA-2 level was determined using a HPLC-tandem mass spectrometry (LC-MS/MS). The maximum concentration in serum (Cmax; reported in mg per liter), the area under the concentration-time curve from time zero to ∞ (AUC 0-∞ ; reported in mg-min per liter), and the terminal half-life (t 1/2 ; reported in min) were calculated using a computer-assisted method 43 . A. baumannii, P. aeruginosa and E. coli peritoneal sepsis models. Murine peritoneal sepsis models with A. baumannii ATCC 17978, P. aeruginosa PaO1 or E. coli ATCC 25922 strains were established by i.p. inoculation of the bacteria 44 . Briefly, mice were inoculated with 0.5 mL of the bacterial suspensions, which were incubated for 20-24 h in LB at 37 °C and mixed in a 1:1 ration with a saline solution containing 10% (w/v) porcine mucin (Sigma, Spain). The minimal bacterial lethal dose 100 (MLD100), LD50 and LD0 were determined by inoculating various groups of mice (6 mice per group) with decreasing amounts of A. baumannii ATCC 17978, P. aeruginosa PaO1 and E. coli ATCC 25922 strains inocula from 8.5 to 2.3, 9.11 to 3.3, 9.13 to 3.47 log CFU/mL, respectively, and monitoring the survival of the mice for 7 days. Therapeutic effect of AOA-2 in murine models of peritoneal sepsis. The murine peritoneal sepsis models by A. baumannii ATCC 17978, P. aeruginosa PaO1 or E. coli ATCC 25922 strains were established by i.p. inoculation of the bacteria. Briefly, animals were infected i.p. with 0.5 mL of the MLD100 of A. baumannii ATCC 17978, P. aeruginosa PaO1, or E. coli ATCC 25922 strains mixed 1:1 with 10% porcine mucin. AOA-2 therapy was administered 2 h after bacterial inoculation. Fifthy-six mice were randomly ascribed to the following groups: 1). controls (without treatment), and 2). AOA-2 administered at 10 mg/kg/d i.p. 2 h after bacterial inoculation with each strain. Mortality was recorded over 72 h. After death or sacrifice of the mice at the end of the experimental period, aseptic thoracotomies were performed, and blood samples were obtained by cardiac puncture for qualitative blood cultures. The spleen and lungs were aseptically removed and homogenized (Stomacher 80; Tekmar Co., USA) in 2 mL of sterile NaCl 0.9% solution. Ten-fold dilutions of the homogenized spleen and lungs were plated onto Sheep blood agar for quantitative cultures (log CFU/g of spleen or lungs).

Statistical analysis.
Group data are presented as means ± standard errors of the means (SEM). For in vitro studies, the Student t test was used to determine differences between means. Differences in bacterial spleen and lung concentrations (mean ± SEM log CFU per gram of tissue) were assessed by analysis of variance (ANOVA) and post-hoc Dunnett's and Tukey's tests. Differences in mortality (%) and blood sterility (%) between groups were compared by use of the χ2 test. P values of <0.05 were considered significant. The SPSS (version 17.0; SPSS Inc.) statistical package was used.