Killing of Staphylococcus aureus and Salmonella enteritidis and neutralization of lipopolysaccharide by 17-residue bovine lactoferricins: improved activity of Trp/Ala-containing molecules

Bovine lactoferricin (LfcinB) has potent antibacterial, antifungal and antiparasitic activities but is also hemolytic. Our objective was to identify LfcinB17-31 derivatives with reduced hemolysis and improved antimicrobial activity via substituting Cys3, Arg4, Gln7, Met10, and Gly14 with more hydrophobic residues. Two peptides, Lfcin4 and Lfcin5, showed higher activity against Staphylococcus aureus and Salmonella enteritidis and lower hemolytic activity than the parent peptide LfcinB17-31. These peptides permeabilized the outer and inner membranes of S. enteritidis; however, Lfcin5 did not permeabilize the inner membrane of S. aureus. Gel retardation and circular dichroism spectra showed that Lfcin4 and Lfcin5 bound to bacterial genomic DNA. Lfcin4 inhibited DNA, RNA and protein synthesis. Both peptides induced the peeling of membranes and the lysis of S. enteritidis. At doses of 10 and 15 mg/kg, Lfcin4 and Lfcin5 reduced the bacterial counts in infected thigh muscles by 0.03‒0.10 and 0.05‒0.63 log10 CFU/g of tissue, respectively, within 10 h. Lfcin4 and Lfcin5 enhanced the survival rate of endotoxemic mice; reduced serum IL-6, IL-1β and TNF-α levels; and protected mice from lipopolysaccharide-induced lung injury. These data suggest that Lfcin4 and Lfcin5 may be antimicrobial and anti-endotoxin peptides that could serve as the basis for the development of dual-function agents.


Results
Physicochemical parameters, structures and electrostatic potentials on the surface of LfcinB17-31 and its derivatives. The amino acid sequences and key physicochemical parameters of the LfcinB17-31 derivatives are summarized in Table 1. The physicochemical properties of Lfcin2 were almost equal to those of the parent peptide, Lfcin1. The net charge of peptides Lfcin3-Lfcin6 was lower than that of Lfcin1. The Lfcin3, Lfcin4 and Lfcin5 peptides had the highest α -helical content at 64.7% and Lfcin6 had the lowest α -helix content at 35.2%. Lfcin2 contained a lower helical content (52.9%) compared to Lfcin1. The hydrophobicity of Lfcin3-Lfcin6 increased from − 1.207 to − 0.376, indicating that they are more hydrophobic than Lfcin1 and Lfcin2; this is likely due to the addition of a methyl group from the Gly/Cys/Gln to Ala residue substitution 28 . The instability index of Lfcin2-Lfcin6 ranged from 28.82 to 46.67. The Boman index, which is a measure of peptide affinity to proteins and the ability to establish bio-interactions, of LfcinB17-31 derivatives was in the range of 1.45-3.02 kcal/mol.
its α -helical structure and formed a turn (70.6%) and an irregular coil (29.4%) (Supplementary Table S1), perhaps due to the replacement of Arg4 and Met10 by Trp. The electrostatic potentials of Lfcin3, Lfcin4 and Lfcin5 (from ± 84.244 to ± 85.553 kT/e) were lower than those of Lfcin1 (± 98.263 kT/e), Lfcin2 (± 93.267 kT/e) and Lfcin6 (± 86.594 kT/e) ( Supplementary Fig. S1C), revealing their distinct cationic and amphipathic characters. This difference likely results in different permeabilization abilities of these peptides through electrostatic and hydrophobic interactions with bacterial membranes 29 . Lfcin4 and Lfcin5 display antibacterial activity with low hemolytic activity. As shown in Table 2, Lfcin4 and Lfcin5 showed higher activity against E. coli, S. enteritidis and S. aureus than against Lfcin1. The MIC values for Lfcin4 and Lfcin5 against S. enteritidis CVCC3377, S. aureus ATCC25923 and S. aureus ATCC43300 were in the range of 2 to 64 μ g/ml. Lfcin6 exhibited the lowest antimicrobial activity against the test strains. Similar to Lfcin1, higher MICs (> 128 μ g/ml) for Lfcin2 to Lfcin6 against S. pullorum CVCC1789 and CVCC1802, and P. aeruginosa CICC10419 and CICC21630 were observed. Moreover, Lfcin4 and Lfcin5 displayed very low hemolytic activity against mouse erythrocytes (Fig. 1A). Even at a high concentration (512 μ g/ml), the hemolytic activity of Lfcin4 and Lfcin5 was 4.87% and 2.2%, respectively, which is lower than that of the parent peptide, Lfcin1 (approximately 7.5% hemolysis at 256 μ g/ml) 14 .
Circular dichroism (CD) spectra of Lfcin4 and Lfcin5. Anionic surfactants such as sodium dodecyl sulfate (SDS) and fluorinated alcohols such as trifluoroethanol (TFE) are often used to mimic the cell membrane environment 30 . To analyze the structural features of Lfcin4 and Lfcin5, the CD spectra of peptides were measured in ddH 2 O, 5-40 mM SDS and 50% TFE buffer, respectively. The secondary structures of Lfcin4 and Lfcin5 in ddH 2 O were characterized predominantly by antiparallel and random coils with a characteristic positive maximum at 230 nm and a negative minimum at 200 nm ( Fig. 1B and C). The CD spectra of Lfcin4 and Lfcin5 showed an increase in α -helical content in SDS and TFE buffer (> 11.8% α -helicity) (Supplementary Tables S2 and S3) relative to the parent peptide (< 10% α -helicity) 20 . This result indicates that both SDS and TFE buffer are favorable for the formation of an α -helical structure.
Lfcin4 and Lfcin5 affect the cell surface hydrophobicity of bacterial cells. As shown in Fig. 2A and B, the cell surface hydrophobicity of S. aureus ATCC25923 and S. enteritidis CVCC3377 treated with peptides was positively correlated with the concentration of Lfcin4 and Lfcin5. For S. aureus ATCC25923, treatment with Lfcin4 resulted in an increased cell surface hydrophobicity from 5.27% to 18.87% ( Fig. 2A), whereas Lfcin5 caused a relatively small change in cell surface hydrophobicity (from 5.27% to 8.95%). The cell surface hydrophobicity of S. enteritidis CVCC3377 increased from 9.43% to 16.45% and from 9.43% to 16.74% following treatment with Lfcin4 and Lfcin5, respectively, which is markedly higher than the untreated controls. Lfcin5 exhibited higher cell surface hydrophobicity than Lfcin4 (Fig. 2B). Changes in cell surface hydrophobicity are one of the noticeable effects of peptide interactions with bacterial cells, which may be correlated with structural alterations of bacterial cell surfaces 31 .  Permeabilization of the outer membrane in S. aureus and S. enteritidis. The hydrophobic fluorophoric probe N-phenyl-1-naphthylamine (NPN) is normally prevented from entering cells because it cannot penetrate the outer membrane. Once the outer membrane is damaged, NPN is partitioned into the interior of the outer membrane, thereby exhibiting strong fluorescence in this hydrophobic environment. As shown in Fig. 3B and C, both peptides instantly permeabilized the outer membrane of S. aureus ATCC25923 and S. enteritidis CVCC337 within 1 min, and higher concentrations of Lfcin4 and Lfcin5 resulted in a stronger NPN signal, indicating that the peptides made the outer membrane more permeable. These results suggest that both Lfcin4 and Lfcin5 induce a timeand concentration-dependent NPN fluorescence increase in intact bacterial cells and that the permeabilization capacity of Lfcin4 is higher than that of Lfcin5. Lfcin4 and Lfcin5 bind to bacterial genomic DNA. An electrophoretic mobility shift assay (EMSA) was used to evaluate the binding affinity of each peptide to bacterial genomic DNA 33 . Lfcin4 and Lfcin5 interacted with the genomic DNA ( Fig. 4A, Supplementary Fig. S2). Lfcin4 impaired the migration of the genomic DNA from E. coli O157 and S. aureus ATCC25923 at mass ratios over 1.0 (peptide:DNA) and from S. enteritidis CVCC3377 at a mass ratio greater than 2.5, demonstrating the intrinsic DNA-binding ability of Lfcin4. For Lfcin5, no DNA bands were observed on the gel for E. coli or S. aureus at mass ratios of 10; however, a fraction of the genomic DNA of S. enteritidis migrated into the gel. This finding suggests that DNA-binding activity of Lfcin4 is higher than that of Lfcin5. The DNA-binding properties of peptides were further supported by the following CD assay. As shown in Fig. 4B,C and D, the DNA CD spectra were dramatically changed when peptides bound to DNA from E. coli, S. aureus and S. enteritidis, which contained a negative band at approximately 250 nm and a positive band at approximately 275 nm. There was a greater decrease in the CD amplitude but no obvious shift, suggesting that Lfcin4 and Lfcin5 interact with genomic DNA and alter the DNA conformation 34 .
Lfcin4 inhibits the synthesis of DNA, RNA and protein precursors. Despite the focus on bacterial membranes as targets for AMPs, several AMPs have additional intracellular targets 25 . The incorporation of radioactive precursors into DNA, RNA, peptidoglycans, and proteins was used to evaluate the effects of Lfcin4 and  effective at reducing the bacterial load in treated mice. At the start of therapy, in treated mice, 8.97 (the Lfcin4 group) and 9.07 (the Lfcin5 group) log 10 CFU/g of S. aureus ATCC25923 were detected in the thigh. Five hours after treatment, the bacterial counts in mice treated with Lfcin4 and Lfcin5 were all reduced by − 0.01-0.10 and − 0.01-0.57 log 10 CFU/g, respectively, compared to untreated mice (Fig. 6). Ten hours after treatment, an increase of 0.19-0.21 log 10 CFU/g was detected in untreated control mice. Treatment with 10 or 15 mg/kg Lfcin4 resulted in a reduced bacterial burden by 0.03 and 0.10 log 10 CFU/g, respectively. This reduction was less than that observed following vancomycin treatment (− 0.06 log 10 CFU/g) (Fig. 6A). At 5 h, treatment with 15 mg/kg Lfcin5 reduced the bacterial count by 0.57 log 10 CFU/g. Treatment with 10 mg/kg vancomycin resulted in a decrease of 0.1 log 10 CFU/g (Fig. 6B). At 10 h post-treatment with 15 mg/kg Lfcin5, approximately 76.6% of S. aureus cells had been killed. This indicates that treatment with 15 mg/kg Lfcin5 is more effective at reducing the bacterial burden than treatment with 10 mg/kg vancomycin (− 1.2%) and Lfcin4 (20.5%) (Fig. 6B).   (Fig. 7A). The mice injected with LPS began to die 12 h after inoculation; all mice had succumbed within 24 h. Treatment with 10, 15 or 20 mg/kg Lfcin4 improved the survival rates to 75%, 100% and 100%, respectively. Mice treated with 10, 15 or 20 mg/kg Lfcin5 responded in a dose-dependent fashion, with survival rates of 50%, 62.5% and 75%, respectively. The survival rate of mice treated with 15 mg/kg   colistin was 75% (Fig. 7A). These results indicate that both Lfcin4 and Lfcin5 protect mice from lethal LPS challenge in vivo.
To determine whether the protective activities of Lfcin4 and Lfcin5 are associated with inflammatory cytokines, the serum levels of proinflammatory cytokines interleukin-6 (IL-6), interleukin-1β (IL-1β ) and tumor necrosis factor-α (TNF-α ) were determined in endotoxemic mice at different time points using an enzyme-linked immunosorbent assay (ELISA). As shown in Fig. 7B, the concentrations of serum IL-6, IL-1β and TNF-α were 779.5-115.8, 100.1-6.3 and 158.9-67.8 pg/ml, respectively, for endotoxemic mice treated with Lfcin4 and were 761.6-240.2, 155.2-47.7 and 173.8-78.3 pg/ml, respectively, for mice treated with Lfcin5. The levels of IL-1β and TNF-α were significantly lower than those from the corresponding LPS control groups (717.2-294.9, 258.0-193.2 and 330.3-120.5 pg/ml, respectively) and the colistin groups. These data suggest that treatment with either Lfcin4 or Lfcin5 inhibits the production of proinflammatory cytokines.
To determine whether Lfcin4 or Lfcin5 protects mice from LPS-induced lung injury, the lungs of mice were examined histologically at 8 h-7 d after treatment. As shown in Fig. 7C, no pathological change was found in the lungs of mice not injected with LPS, whereas LPS-challenged mice treated with saline developed acute lung injury, characterized by alveolar septum distention, inflammatory cells, an infiltration of red blood cells, and a certain degree of lung tissue distortion. In contrast, after treatment with Lfcin4 or Lfcin5, there was apparently less lung damage at 8 h-4 d and no obvious pathological differences in the lung tissues were observed at 7 d. The efficacy of both Lfcin4 and Lfcin5 was higher than that of colistin. These data demonstrate that, similar to colistin, both Lfcin4 and Lfcin5 protect mice from lethal LPS challenge in vivo.

Discussion
Lfcin is a multifunctional peptide that targets antibiotic-resistant bacterial strains, fungi, viruses, parasites, and tumors, as well as directly binds to the bacterial endotoxin LPS and is thereby already being used in different preclinical and clinical trials 11 . In this work, we designed five 17-residue LfcinB17-31 derivatives (Trp-rich and Ala-rich) and first studied their antibacterial activities; Lfcin4 and Lfcin5 were found to be the most active peptides among those tested against both S. aureus and S. enteritidis, displaying more than 2-to 8-fold increases in antibacterial activity over the parent peptide, Lfcin1 (LfcinB17-31) ( Table 2).
The α -helical conformation is necessary for the activity of some AMPs such as magainin and cecropin 35,36 . In this study, there was also a positive correlation between the α -helical content and antibacterial activity of LfcinB17-31 derivatives. Both Lfcin4 and Lfcin5 had the highest α -helix content at 64.7% and the most potent antimicrobial activities (Tables 1 and 2), which may be related to Ala at positions 3, 7 and 14 since Ala is favorable for α -helix formation 37 . Conversely, Lfcin6, with the lowest α -helical content, was inactive, wherein Arg and Met modifications in positions 4 and 10 by Trp resulted in a decrease in α -helix content from 60% to 35.2% (Tables 1  and 2). Furthermore, the CD spectra verified that both Lfcin4 and Lfcin5 contain an amphipathic α -helical structure ( Fig. 1B and C, Supplementary Table S2). Therefore, it appears that the α -helical structure of the LfcinB17-31 derivatives is the most important structural parameter affecting antibacterial activity, while the antimicrobial activity of these peptides was less dependent on their positive charge.
Rekdal et al. reported that an increase in hydrophobicity leads to an increase in antibacterial activity 18 . In our study, the enhanced activity of certain peptides may be due to the increased hydrophobicity by the replacement of Cys, Arg, Gln, Met, and Gly residues at positions 3, 4, 7, 10, and 14 in Lfcin1 by an aliphatic Ala or an aromatic Trp or Phe with larger hydrophobic side chains, which contribute largely to the depth of penetration into the hydrophobic core of a bilayer and the strength of interaction with negatively charged membranes, respectively 38 . Lfcin4 and Lfcin5 contain the largest number of hydrophobic residues. These results are consistent with previous conclusions that the increase in activity of Lfcin peptides (LFB A3, LFB Q7 and LFB G14) may be due to increased hydrophobicity that comes with the replacement of Cys, Gln or Gly residues by Ala 20 . Additionally, Trp is essential for the pore-forming properties of the peptide 27 . LfcinB uses Trp solely as a hydrophobic anchoring force to help thread the peptide through the membrane by its hydrophobic character 21 . However, it is also evident that there is an upper limit to the number of Trp residues that is tolerated, as peptide Lfcin6 (containing four Trp residues at positions 3, 6, 8, and 10) in this study exhibited much less antimicrobial activity than Lfcin4 (containing three Trp residues at positions 4, 6 and 8) and Lfcin5 (containing three Trp residues at positions 6, 8 and 10). This finding is in agreement with a previous study that reported that the LFB derivatives containing three additional Trp residues had lower antimicrobial activity than the LFB peptide with two Trp residues 21 . Cation-π interactions and hydrogen bonds occur mainly between aromatic residues (e.g., Trp and Phe) and residues with positively charged side chains (e.g., Arg and Lys), which may allow the peptides to penetrate deeper into the membrane 22 . All of these factors combined (i.e., α -helical content, hydrophobicity and intermolecular interactions) may have contributed to a more efficient activity of Lfcin4 and Lfcin5 against bacteria than their parent peptide.
The LfcinB peptide exerts a minor permeabilizing effect on the cytoplasmic membrane of susceptible bacteria; however, it does not lyse the bacteria 25,39 . In our study, both Lfcin4 and Lfcin5 induced a concentration-dependent leakage of entrapped calcein, with the former having more potent interactions with negatively charged phospholipids than the latter. Additionally, the effects of Lfcin4 and Lfcin5 on Gram-negative bacteria differ from those on Gram-positive bacteria. Compared to Lfcin5, Lfcin4 demonstrates relatively stronger outer membrane permeabilization capacity against S. aureus and S. enteritidis (Fig. 3B and C). Moreover, Lfcin4 permeabilized the inner membrane of S. aureus and induced the leakage of cytoplasmic content, although Lfcin5 did not cause any significant damage to the integrity of the inner membrane; only shrinkage of cells was observed (Figs 2C and  3E). It is very interesting that the effect of Lfcin5 on S. enteritidis was more rapid and more profound than that of Lfcin4 (Fig. 3F and G), as observed by SEM (Fig. 2C). These results suggest a possible intracellular target for Lfcin4 and Lfcin5. The interactions between the peptides and bacterial genomic DNA examined in our study imply that both Lfcin4 and Lfcin5 bind to genomic DNA from S. aureus, S. enteritidis and E. coli (Fig. 4). This result is contrary to the previous finding that LfcinB did not bind to pGEM-β GAL plasmid DNA 25 . Moreover, we found that Lfcin4 inhibited DNA, RNA and protein synthesis in S. aureus whereas Lfcin5 did not (Fig. 5), which is similar to LfcinB17-41 and other Trp-rich peptides derived from human lysozymes that have been shown to translocate across lipid bilayers and inhibit macromolecular synthesis 25,39 . Therefore, it appears that Lfcin4 and Lfcin5 use multiple mechanisms to kill bacteria.
Conversely, in this study, Lfcin4 and Lfcin5 demonstrated excellent in vivo activity against S. aureus in mice (Fig. 6), as does human Lfcin1-11 40 . A dose-dependent effect was observed, where increasing doses of Lfcin4 and Lfcin5 from 10 to 15 mg/kg of body weight resulted in a greater reduction of viable bacteria in infected thighs of 0.85 log 10 CFU/g of tissue compared to control mice (Fig. 6). We also showed that both Lfcin4 and Lfcin5 increased the survival of endotoxemic mice and protected the lungs from acute injury induced by LPS and that they were more effective than colistin ( Fig. 7A and C). Moreover, these peptides significantly inhibited the release of cytokines (IL-6, IL-1β and TNF-α ) in mice challenged with LPS (Fig. 7B). Together, these results suggest that both Lfcin4 and Lfcin5 are potential endotoxemia therapeutics.
In conclusion, we designed a series of LfcinB17-31 derivatives based on net charge, α -helical content and hydrophobicity. Both Lfcin4 and Lfcin5 showed stronger antimicrobial activity than the parent peptide and had very low hemolysis. Lfcin4 and Lfcin5 disrupt bacterial cell membranes and interact with bacterial DNA. DNA, RNA and protein synthesis were inhibited in S. aureus by Lfcin4 but not by Lfcin5. Both Lfcin4 and Lfcin5 protected mice from S. aureus infection and from LPS-induced lethality. Our findings strongly suggest that Lfcin4 and Lfcin5 (Trp-rich and Ala-rich) could serve as promising antimicrobial and anti-endotoxin agents for further clinical applications.
Antibacterial and hemolytic activity. The MIC values of LfcinB17-33 derivatives against bacteria were determined by the microtiter broth dilution method 41 . Briefly, 10 μ l of serially diluted aliquots of peptides were added to each well of 96-well microtiter plates, followed by the addition of 90 μ l exponential phase bacteria (10 5 CFU/ml) and incubated for 16-18 h at 37 °C. The MIC value was determined as the lowest peptide concentration at which the peptide completely inhibited the growth of bacteria.
The hemolytic activities of Lfcin4 and Lfcin5 were evaluated based on the amount of released hemoglobin from erythrocyte suspensions of healthy mouse blood according to a previous method 42 . Blood cells were centrifuged at 2000 rpm for 5 min at 4 °C and washed three times with a saline solution (0.9% NaCl). Erythrocytes were resuspended in 0.9% NaCl, and 100 μ l aliquots were mixed with 100 μ l of peptides at different concentrations ranging from 2 to 1024 μ g/ml. The mixtures were then added to 96-well plates. After incubation at 37 °C for 1 h, cells were centrifuged at 5000 rpm for 5 min, and the absorbance of the supernatant was measured at 540 nm on a microtiter plate reader. Controls for 0% and 100% hemolysis release were determined with 0.9% NaCl (A NC ) and 0.1% Triton X-100 (A PC ), respectively. Hemolytic percentages are expressed as follows: hemolysis Cell surface hydrophobicity assay. The effects of AMPs on the cell surface hydrophobicity of S. aureus ATCC25923 and S. enteritidis CVCC3377 were measured using the hexadecane partitioning method 41 . Cells were grown to stationary phase in Mueller-Hinton broth at 37 °C, harvested by centrifugation at 5000 rpm for 4 min, washed twice with sterile normal saline, and suspended in 0.1 M KNO 3 (pH 6.2) to an OD 600 nm of 0.4. Lfcin4 or Lfcin5 was added to a final concentration of 1× , 2× or 4 × MIC and incubated for 10 min at 37 °C. Sterile normal saline was used as a negative control, and the optical density at 600 nm was measured as OD 0 . Subsequently, a 1.2-ml cell suspension was mixed with 0.2 ml of hexadecane by vortexing for 1 min, and then, the mixture was incubated for 15 min. The OD 600 nm value of the aqueous phase was determined as OD 1 . The hydrophobicity percentages were calculated by the following equation: hydrophobicity (%) = (1 − OD 1 /OD 0 ) × 100%. All experiments were repeated three times.

Membrane permeability assay. Effects of Lfcin4 and Lfcin5 on model membranes. Large unilamellar
vesicles (LUVs) composed of POPC/POPG (1:3) were used as a model anionic membrane 43 . LUVs were dissolved in chloroform, dried under nitrogen gas and vacuum-dried for 2 h. The lipid film was rehydrated in 20 mM HEPES (150 mM NaCl, 1 mM EDTA, pH 7.4). To make calcein-encapsulated unilamellar liposomes, the lipid film was suspended in 5 mM sodium HEPES (containing 100 mM calcein, pH 7.5). The liposome suspensions were freeze-thawed in liquid nitrogen five times and extruded ten times through two stacked polycarbonate filters (100 nm pore size) in an Avanti Mini-Extruder. Untrapped calcein was removed using a Sephadex G-50 column (Amersham Pharmacia Biotech, Little Chalfont, UK). Peptides were then added to the LUVs. After a 10-min incubation at room temperature, the release of calcein form vesicles was monitored for 10 min on a Tecan Infinite M200 PRO plate reader (TECAN, Austria) with an excitation wavelength of 485 nm and an emission wavelength of 535 nm. The total calcein fluorescence (F T ) was determined by the addition of 0.1% Triton X-100. Dye leakage was calculated according to the following equation: leakage rate (%) = ((F − F 0 )/(F T − F 0 )), where F 0 indicates the fluorescence of each sample at T = 0.
Effects of Lfcin4 and Lfcin5 on the outer membrane. The effects of Lfcin4 and Lfcin5 on the outer membrane of S. aureus ATCC25923 and S. enteritidis CVCC3377 were determined by measuring the uptake of NPN 44,45 . Mid-log phase cells were washed with sterile normal saline three times and suspended in 5 mM HEPES buffer (pH 7.2, containing 5 mM glucose) to an OD at 600 nm of 0.5. Subsequently, NPN (10 mM) and peptide (1× , 2× or 4 × MIC) solutions were added to 96-well plates. The relative fluorescence intensity was measured using an Infinite M200 PRO plate reader. The excitation and emission wavelengths were set at 328 and 438 nm, respectively. Treatment with PBS severed as a negative control, and treatment with ampicillin or colistin was used as a positive control.
Effects of Lfcin4 and Lfcin5 on the inner membrane. The permeabilization of the inner membrane was performed according to a previous report by Li et al. with slight modifications 33 . S. aureus ATCC25923 and S. enteritidis CVCC3377 cells in mid-log phase were washed with 10 mM PBS buffer three times, resuspended in the same buffer (1 × 10 8 CFU/ml), and incubated with or without 1 × MIC peptides at 37 °C for 5, 30 and 120 min. Bacterial cells were then washed twice with PBS, fixed with 50 μ g/ml PI and analyzed using a BD FACS Calibur Flow Cytometer (Franklin Lakes, New Jersey, USA). Data were analyzed using CellQuest Pro software (BD, USA).
Effects of Lfcin4 and Lfcin5 on bacterial genomic DNA. Genomic DNA was extracted from S. aureus ATCC25923, E. coli O157 and S. enteritidis CVCC3377 using a TIANamp Bacteria DNA Kit (TIANGEN Biotech Co., Ltd., Beijing). The gel retardation experiment was performed by mixing the bacterial DNA with different concentrations of Lfcin4 or Lfcin5 in 20 μ l binding buffer (10 mM Tris-HCl, pH 8.0, 5% glycerol, 1 mM dithiothreitol, 1 mM EDTA, 20 mM KCl, and 50 μ g/ml bovine serum albumin) 41 . The ratios of peptide to DNA were 0, 0.5, 1, 2.5, 5.0, and 10.0 (w/w). The mixtures were incubated for 10 min at 37 °C and then subjected to electrophoresis on a 0.7% agarose gel.
To examine whether peptide binding causes secondary structure changes in bacterial DNA, CD measurements were carried out on an MOS-450 spectropolarimeter using a quartz cuvette with a 1.0 mm path length. The ratios of peptide to bacterial DNA were 0, 5 and 10. The spectra were recorded from 220 to 320 nm at 25 °C. The CD data represent the average of 10 scans with a 20 s bandwidth.  3 H-leucine, and 3 H-glucosamine (40 μ Ci/ml) were added to cultures to measure macromolecular synthesis and incubated for 20 min at 37 °C. Cold 25% trichloroacetic acid (TCA) was added to the mixture and placed on ice for 30 min. After centrifugation, pellets were washed twice with 25% TCA, dried and counted with scintillation fluid on a MicroBeta 1450 scintillation counter (Perkin Elmer) 46 .

Effects of Lfcin4 and
In vivo experiments. Animal care and all experimental protocols were approved by the Laboratory Animal Ethical Committee and its Inspection of the Feed Research Institute of Chinese Academy of Agricultural Sciences (AEC-CAAS-20090609) and were performed in accordance with ARRIVE guidelines 47 .
Thigh infection model. The thigh infection protocol was performed as previously described 48 . Female BALB/c mice (6 weeks old, 22 ± 2 g) were rendered neutropenic (neutrophils, 100/mm 3 ) by intraperitoneal injection with 150 and 100 mg cyclophosphamide per kg of body weight for 4 d and 1 d before infection, respectively. Mice were inoculated with 100 μ l of S. aureus ATCC25923 cells (1.0 × 10 7 CFU/ml) and given a single dose (0.2 ml) of Lfcin4 or Lfcin5 (10 or 15 mg/kg of body weight) or vancomycin (10 mg/kg of body weight) by tail vein injection at 2 h post-infection. Mice were humanely euthanized by cervical dislocation at 5 and 10 h after treatment. Thighs were removed and homogenized in sterile PBS for a CFU assay for S. aureus ATCC25923.
LPS-induced endotoxemia model. Male C57BL/6 mice (8 weeks old, 22 ± 2 g) were intraperitoneally injected with 30 mg/kg of LPS (0.2 ml) from E. coli O111:B4 cells as previously described 6,49 . Mice were intraperitoneally injected with Lfcin4 or Lfcin5 (10, 15 or 20 mg/kg of body weight in 0.2 ml) or colistin (15 mg/kg of body weight in 0.2 ml) at 0.5, 8 and 24 h after inoculation with LPS, respectively. Mice injected with LPS or saline only served as positive or mock-treated controls, respectively. Survival was recorded every 12 h and followed for up to 7 d.
Serum was separated from mice at 2, 8 and 20 h following the administration of LPS (10 mg/kg). The levels of IL-6, IL-1β and TNF-α in endotoxemic mice were detected at Jiaxuan Biotech. Co., Ltd. (Beijing, China) using an ELISA kit according to the manufacturer's protocol.
Thirty minutes after intraperitoneal injection of LPS (10 mg/kg of body weight in 0.2 ml), mice were intraperitoneally injected with Lfcin4, Lfcin5 or colistin (15 mg/kg of body weight in 0.2 ml). Mice were sacrificed at 8 h, 1 d, 4 d, and 7 d after LPS injection. The lungs were removed, washed with PBS and fixed in 4% paraformaldehyde at 4 °C for 24 h. After they were washed with PBS and dehydrated with a graded series of ethanol (75-95%), the tissues were infiltrated with xylene and embedded in paraffin wax. Sections were cut, stained with hematoxylin and eosin and examined under a light microscope. Mice injected with LPS or only PBS served as positive or mock-treated controls, respectively. Statistical analysis. Statistical analysis was performed using software SPSS21.0 (SPSS, USA). Data are expressed as the mean ± standard error of the mean (SEM). One-way repeated analysis of variance (ANOVA) (Institute Inc., Cary, NC, USA), Duncan method and the Mann-Whitney rank test were used to determine the statistical significance. A p value of < 0.05 was considered statistically significant.