Bacterial and yeast strains used in this study are listed in Table 1. Bacteria were grown for 24 h either on nutrient agar plates or broth under shaking (200 rpm), at 28 or 37°C depending on the tested strain. Yeasts were grown for 48 h on Sabouraud agar medium at 37°C. Fungal strains are listed in Table 2. Aspergillus fumigatus ATCC 16424, A. terreus MHR1, and Scedosporium apiospermum EC13 were also grown on Sabouraud agar medium, but at 28°C during 5 days. Conidia were then harvested in sterile water, filtered to remove hyphae and adjusted to a working concentration of 10⁵ conidia/ml. Botrytis cinerea B05.10 and Alternaria brassicicola Abra43 were cultured in 9.5 cm Petri dishes containing Potato Dextrose Agar (PDA, Becton Dickinson) under a 16 h photoperiod at 22°C for 2 weeks to induce sporulation. Then, conidia were harvested in 3 ml of sterile water, filtered through miracloth (EMD Chemicals) and adjusted to a working concentration of 10⁵ conidia/ml.

Synthetic native peptides, armadillidin H (molecular weight of 5259.69 g/mol) and armadillidin Q (molecular weight of 5250.68 g/mol), with the respective amino acid sequences (without C-terminal amidation) GHLGRPYIGGGGGFNRG GGFHRGGGFHRGGGFHSGGGFHRGGGFHSGGSFGYR and GHLGRPYIGGGGGFNRGGGFHRGGGFHRGGGFQSGGGFHRGGGFHSGGSFGYR were purchased from ProteoGenix Corporation (Schiltigheim, France). Stock solutions were prepared in 8% acetonitrile at a final concentration of 2.7 mM and stored at -80°C. Working solutions (380 μM) were prepared by dilution in sterile water. All other reagents were purchased from Sigma–Aldrich (Saint-Louis, MO, USA) unless stated otherwise.

The molecular masses of haemocyte crude extracts [prepared as described in (Herbiniere et al., 2005)] and synthetic peptides were determined by electrospray ionization mass spectrometry (ESI-MS) with a Xevo Q-TOF (Waters, Milford, MA, USA) mass spectrometer. Samples were suspended in 50% acetonitrile/0.2% formic acid (v/v). LC-MS mass spectra were performed in positive mode with a cone voltage ramping from 20 to 40 V. The spray voltage was set to 3.0 kV, the source temperature to 120°C and the desolvation temperature to 450°C. The LC separation was conducted on a ProSwift^® RP-1S (Dionex) analytical reverse-phase HPLC column (4.6 mm × 50 mm). Separation was carried out using a water/acetonitrile/formic acid 0.2% (v/v) solvent system. After an initial 5 min wash without acetonitrile, elution was achieved at a flow rate of 0.5 ml/min with a 10 min linear gradient from 0 to 40% acetonitrile, followed by a 5 min wash with 80% acetonitrile.

Minimum inhibitory concentrations (MIC) of both armadillidins toward various bacterial and yeast strains were measured according to the method detailed elsewhere (Verdon et al., 2008). Yeast suspensions [0.5 × 10³ to 2.5 × 10³ blastospores/ml in RPMI 1640 medium buffered to pH 7.0 with 0.165 M MOPS (morpholinepropanesulfonic acid)] were incubated with twofold dilutions of a 380 μM armadillidin stock solution. Suspensions were incubated for 48 h at 28 or 37°C depending on the tested strain.

Hyphal growth inhibition tests were performed in 96-wells plate (Nunc) containing Malt extract (15 g.l^-1; Duchefa) medium for B. cinerea B05.10 and A. brassicicola Abra43 or RPMI MOPS medium for A. fumigatus ATCC 16424, A. terreus MHR1 and S. apiospermum EC13 strains. Different concentrations of the working peptide solution were prepared in sterile water. A volume of 5 μl of conidia suspension (10² conidia/μl) was added to the incubation medium containing 50 μl of the peptide solution and 50 μl of culture medium. In parallel, 50 μl of sterile water was added instead of the peptide solution. The microtiter plates were placed on an orbital shaker (30 rpm) and incubated for 36 h at 22°C. Each well was observed under an inverted microscope (LEICA, DMI 6000B) and the length of three hyphae (hl) was measured on three independent replicates. Length averages were calculated and the percentage of inhibition was determined according to the following formula: % inhibition = (100-(100/hl_control) × hl_peptide).

Pseudomonas syringae DC3000 or B. megaterium F04 cells were grown to an OD₆₀₀ between 0.15 and 0.35. Bacteria were then appropriately diluted in 10 mM sodium phosphate buffer (pH 6.8) to 10⁶ CFU/ml. Serial twofold dilutions of armadillidin H were achieved in sterile 10 mM sodium phosphate buffer (pH 6.8) and added (15 μl) to bacterial suspension (285 μl) at a starting concentration of 4.75 μM. Suspensions were incubated for 1 h at 28 or 37°C depending on the tested strain. Controls were run without peptide (only with the peptide solvent containing acetonitrile). The number of colony-forming units (CFU) was determined by plating 10-fold serial dilutions of bacterial suspensions on NB agar plates after 24 h of incubation at 37°C for B. megaterium or 36 h at 28 °C for P. syringae.

Exponentially growing bacteria in NB (B. megaterium F04 and P. syringae DC3000) were appropriately diluted in 10 mM sodium phosphate buffer (pH 6.8) to obtain a concentration of 10⁶ CFU/ml. Aliquots of 950 μl were incubated (37°C for B. megaterium and 28°C for P. syringae) with 50 μl of 10 mM sodium phosphate buffer (pH 6.8) containing armadillidin H at a final concentration of 1.2 μM for P. syringae and 4.75 μM for B. megaterium. Controls were run without peptide. At different times, suspensions were diluted in 10 mM sodium phosphate buffer (pH 6.8) and spread onto NB agar plates. After an overnight incubation at 37°C for B. megaterium or 36 h at 28°C for P. syringae, CFU were counted.

Exponentially growing bacteria in NB (B. megaterium F04 and P. syringae DC3000) were appropriately diluted in 10 mM sodium phosphate buffer (pH 6.8) to obtain a concentration of 5.10⁷ CFU/ml. Bacteria were then treated for 15 min at 28°C (P. syringae DC3000) or 37°C (B. megaterium F04) with armadillidin H concentrations that induce the highest loss of cultivability (determined by counting CFU as described above for bactericidal activity assays) and that allow the presence of a pellet to work on: 9.5 μM for P. syringae and 2.37 μM for B. megaterium. A. brassicicola Abra43 mycelium was collected after 24 h of culture at 22°C in absence or in presence of 7.6 μM armadillidin H, centrifuged (5000 × g, 10 min) and fixed according to the following procedure. Cells were fixed for 1 h with 2.5% glutaraldehyde in 1 M phosphate buffer, pH 7.1. After PBS washes, bacterial cells were post-fixed for 45 min in 1% osmium tetraoxyde in phosphate buffer. Dehydration was carried out using successive incubations of increasing ethanol concentrations (from 70 to 100%). Cells were then resuspended in 100% ethanol and separated for SEM or TEM. Each part was centrifuged at 10,000 rpm for 10 min. For TEM, the P. syringae DC3000 or B. megaterium F04 pellet was included in epon resin and after 24 h of polymerisation,70 nm sections were cut using an ultramicrotome UC6 (Leica). Uranyl acetate (2% in 70% ethanol) and lead citrate were used as contrasting agents for electron microscopy (JEOL 1010 à 80 KV). TEM on cells was recorded using Olympus digital camera Quemesa with Item software. For SEM, bacterial cells were deposited on a 12 mm cover glass and dried by HMDS treatment (hexamethyldisilazane). The surface of the coverglass was sputter-coated in a vacuum with an electrically conductive 25 nm thick layer of gold–palladium alloy coating system (BALTEC Scd 005 sputter coater). SEM images were then recorded with a scanning electron microscope (JEOL 840) at 15 kV.

Pseudomonas syringae DC3000 or B. megaterium F04 cells were grown to an OD₆₀₀ between 0.15 and 0.35. Bacteria were centrifuged (10000 × g, 30 s) and pellets were suspended in 10 mM sodium phosphate buffer (pH 6.8) to approximately 5.10⁷ CFU/ml. Bacterial suspensions were treated with armadillidin H at a final concentration of 9.5 μM for P. syringae and 2.37 μM for B. megaterium. Suspensions were incubated for 15 min at 28 or 37°C depending on the tested strain. Controls were run without peptide and with 0.1% Triton X-100 for maximal lysis activity. One half of the cell suspensions was analyzed using epifluorescence microscopy (Olympus BX41) after 15 min of bacterial staining with 5 μM SYTO9 (S9) and 7 μM Propidium Iodide (PI) from the Live/Dead BacLight kit L-7005 (Invitrogen, Cergy Pontoise, France), as recommended by the supplier. Damaged bacteria (PI positive) and total bacteria (S9 positive) were counted in each observed field for a total of 10 fields/slide and three slides/condition. Results were expressed as permeabilization: % permeabilization = (number of PI⁺ bacteria/total number of bacteria) × 100. The other half of the bacterial suspensions was appropriately diluted in 10 mM sodium phosphate buffer (pH 6.8) and spread on NB agar plates. After an overnight incubation at 37°C for B. megaterium or 36 h at 28°C for P. syringae, CFU were counted.

For A. brassicicola Abra43, the membrane permeabilization assay was adapted from Joubert et al. (2010). Briefly, 24 h-old germinating conidia grown on malt extract were exposed to 7.6 μM armadillidin H or 0.1% Triton X-100 as positive control for 4 h. After treatment, SYTOX Green was added to the incubation mixture (final concentration of 2 μM) as indicated by the supplier and samples were observed using epifluorescence microscopy as described above.

Hemolytic activity of armadillidin H was measured by detection of released hemoglobin from fresh human erythrocytes. Prior to the assay, the blood was centrifuged (2000 × g, 3 min, 4°C) to collect erythrocytes. Erythrocytes were washed three times with PBS and adjusted to a final concentration of 10% (∼1.10⁸ cells/ml as determined by blood cells counting with a hemocytometer). Reactions were performed in 1 ml mixtures containing 1% erythrocytes (v/v) and variable amount of peptides or an equivalent volume of PBS buffer. Serial twofold dilutions of peptide were performed in PBS buffer and added to 1% human erythrocytes at a starting concentration of 19 μM. Samples were incubated at 37°C for 30 min. Erythrocytes were removed by centrifugation (2000 × g, 3 min, 4°C) and the absorbance of the supernatant was measured at 576 nm. The level of hemolysis when suspending erythrocytes in 0.1% Triton X-100 was considered 100%. Three experiments were carried out in duplicates.

Human blood was obtained from the French national blood transfusion organization, “Etablissement Français du Sang - Centre Atlantique,” under the agreement number CA-PLER-2014 089. Donors gave their written and informed consent for the use of blood samples in research.

Circular dichroism spectra were recorded on a JASCO V-670 spectrometer in the range 190–260 nm, every 0.5 nm (cell length 1 cm), with an armadillidin H sample of 19 μM in 100 mM phosphate buffer, pH 6.0. Then, 50% TFE were added. For spectral deconvolution, the three computer programs Selcon3, CDsstr, and CONTIN/LL have been used, available in the CDPro software package (Sreerama and Woody, 2000) or used at the internet-based CD analysis site DICHROWEB (Whitmore and Wallace, 2008), with selected bases corresponding to soluble proteins (either structured or denaturated).

Nuclear magnetic resonance spectra were recorded on an advance III HD BRUKER 700 MHz spectrometer equipped with a TCI cryoprobe, and were processed with the NMRPipe/NMRDraw software package (Delaglio et al., 1995). A set of 2D ¹H NOESY (160 ms), 2D ¹H TOCSY (80 ms), natural abundancy ¹³C HSQC were performed at 298 K on a 0.1 mM (300 μl in a 3 mm tube) solution of armadillidin H in 100 mM phosphate buffer, pH 5.2, and 10% of D₂O ¹H chemical shifts were referenced to the DSS signal at 0 ppm. After lyophilisation, the same sample was solubilized in H₂O/TFE (50/50, v/v) to acquire 160 ms NOESY, 80 ms TOCSY experiments, and ¹³C HSQC with the previously defined parameters.

To find out if armadillidin is the only AMP produced and stored in A. vulgare haemocytes (Herbiniere et al., 2005), a crude extract prepared from haemocytes of 100 animals was analyzed by LC-MS. A weak peak was observed at 9.3 min corresponding to the retention time found for synthetic armadillidin (not shown). The corresponding spectrum (Figure 1A) showed the same multicharged ions as those observed for the analysis of synthetic armadillidin (Figure 1B). As expected, the calculated monoisotopic molecular mass was 5256.15 Da. Surprisingly, a second set of multicharged ions was observed, close to the armadillidin peaks. The monoisotopic molecular mass calculated for this second peptide was 5247.13 Da. We hypothesized that the difference of 9 Da between these two peptides could correspond to the replacement of a His (H) residue by a Gln (Q) or a Lys (K) residue. To verify this hypothesis, an A. vulgare transcriptome generated by another research program (ANR-10-BLAN-1701 ImmunSymbArt; Sequence Read Archive SRA database accession numbers: SRX564995-SRX565004) and, available in the laboratory, was screened using the blastx alignment tool¹. Results showed the presence of two variants of armadillidin encoded in the genome of A. vulgare, the second one differing from the previously described one (Herbiniere et al., 2005) by only one amino acid substitution: a H is replaced by a Q at position 33. Therefore, we named this variant armadillidin Q and the first discovered armadillidin was renamed armadillidin H.

In order to determine the putative sensitivity of various microbial strains to both armadillidins, an antibacterial activity assay was performed and the MIC was determined for each tested strain. MIC was defined as the lowest concentration of peptide that totally inhibits the visible growth of a selected bacterial or yeast strain after a chosen incubation period (24 h for bacteria and 48 h for yeasts). Results are given in Table 1. Armadillidin H was shown to be active against all the tested bacteria, excepted for two staphylococcal strains (Table 1). Among sensitive strains, four appeared to be highly susceptible as they were inhibited at low peptide concentrations (∼2–5 μM). Similar results were obtained for armadillidin Q as determined MICs were in the same range as for the parent peptide (Table 1). However, concerning yeasts, no growth inhibition has been measured for both peptides (Table 1). All together, these results indicated that both armadillidins displays a similar broad antibacterial spectrum.

When tested against A. brassicicola Abra43, both peptides displayed a high inhibitory effect on mycelium growth. Indeed, the growth was completely stopped at 7.6 μM for armadillidin H and at 15.2 μM for armadillidin Q (Table 2) and conidial germination never occurred even after 2 weeks of culture (data not shown). The difference in the MIC of the two peptides against A. brassicicola Abra43 corresponds to only one dilution. Because we did not test intermediary dilutions, MIC for armadillidin Q could be close to 8 μM thus not significantly different. For B. cinerea B05.10, conidial germination still occurred when treated at the highest tested peptides concentration (38 μM; Table 2), indicating that B. cinerea B05.10 is less sensitive to armadillidins than A. brassicicola Abra43. However, a 50% hyphae length reduction was observed at 15.2 μM for armadillidin H and 19 μM for armadillidin Q (Table 2). Similar results were obtained for S. apiospermum EC13 with a 50% hyphae length reduction observed at 4.7 μM for armadillidin H and 9.5 μM for armadillidin Q (Table 2). No effect was observed on the mycelium length for the two others fungi, A. fumigatus ATCC 16424 and A. terreus MHR1, over the tested concentration range of both armadillidins.

As both peptides displayed the same inhibition spectrum with close MICs values, we decided to use only armadillidin H for the subsequent biological tests and the structure determination. Moreover, B. megaterium F04, P. syringae DC3000, and A. brassisicola Abra43 were chosen as models to study the mode of action of armadillidin H as they were representative of sensitive Gram-positive, Gram-negative and fungi, respectively. In addition, these microorganisms are known to be relatively abundant and ubiquitous in soil where A. vulgare is living (Fierer et al., 2012).

To verify whether armadillidin H has the capacity to kill bacteria or to only inhibit their growth, B. megaterium F04 and P. syringae DC3000 were incubated for a duration of 1 h in the presence or absence of the peptide. After treatments, serial dilutions were spread on NB agar plates so as to quantify viable bacteria. As shown in Figure 2, a decrease of CFU was obtained while increasing armadillidin H concentration, in a dose-dependent manner. No bacterial growth was observed at 4.75 μM for B. megaterium F04 and 1.18 μM for P. syringae DC3000 indicating that armadillidin H acts as a bactericidal peptide. Moreover, the time-killing curves revealed that armadillidin H induced the killing of all the B. megaterium population within 5 min (Figure 3A). By contrast, the killing of all the P. syringae population was achieved faster as no CFU was detected after a 30 s incubation period with armadillidin H (Figure 3B). These results indicated that the bactericidal effect of armadillidin H occurs very quickly on both Gram-positive and Gram-negative bacteria.

To gain insight into the direct effect of armadillidin H on the morphology of treated cells, we used SEM and TEM approaches. Representative electron micrographs of selected experiments are shown in Figures 4–6 for P. syringae, B. megaterium, and A. brassicicola, respectively. Control P. syringae cells had a smooth and normal surface morphology (Figures 4A,C). However, incubation with 9.5 μM armadillidin H showed severe surface damages like membrane wrinkling and bubbling (Figures 4B,D). Moreover, many cells had lost their intracellular organization and membrane integrity. Regarding B. megaterium, untreated samples consisted mainly of single cells or diplobacilli displaying an intact smooth surface (Figure 5A). TEM analysis also revealed intact cell walls and cytoplasmic membranes (Figure 5C). The armadillidin H treatment greatly affected the cell morphology of B. megaterium with a deep roughening of the cell surface (Figure 5B) and a considerable loss of the typical subcellular organization with different steps in peptide-mediated killing (Figure 5D). Indeed, it began with the condensation of cytoplasmic material and detachment of the cell membrane (Figure 5D). In the next step, this condensation evolved to a highly electron-dense amorphous region (Figures 5D,E). Some alterations of the cell wall were also noticed. Finally, cells adopted a ghost-like appearance highlighted by the presence of a large electron-lucent area (Figure 5F). The cells seemed to have lost a part of their cytoplasmic contents, although the overall cell shape was still recognizable. Concerning mycelium of A. brassicicola, in the control condition, it displayed a smooth surface and numerous ramifications (Figure 6A). In contrast, mycelium of A. brassicicola treated with 7.6 μM armadillidin H showed a complete disruption of branching and an uncontrolled cell proliferation (Figure 6B). At higher magnification, vesicle-like structures protruding from the mycelium cell wall were observed, leading to a granular and irregular surface (Figure 6C). TEM observations confirm these findings (Figures 6D–F). Indeed, untreated mycelium (Figure 6D) showed cells with smooth surface and cell-wall, the plasma membrane is also clearly defined. The cytoplasm is well-structured and dense with visible organelles and endoplasmic reticulum. By contrast, fungal cells treated with armadillidin H (Figures 6E,F) have lost their regular shape. We observed extracellular blebs, already visible in SEM. The cell-wall is rough and the cytoplasm structure is clearly altered. Plasmalemma and organelles appear to be degraded showing that cells are probably collapsed.

As the observed ultrastructure of armadillidin H-treated cells strongly suggested a membranolytic step in the mode of action of this peptide, we investigated the membrane permeabilization potency of this peptide using propidium iodide that easily penetrates cells with compromised membranes. In parallel, the viability of bacteria was assayed by CFU counts so as to potentially correlate cell permeabilization and bacterial death (Figure 7). In the case of B. megaterium, 2.37 μM armadillidin H induced 95.3% (±5.1%) permeabilization of the cell population and a 5 log-reduction of the CFU number (from 6.9 ± 0.3 to 1.8 ± 0.2; Figure 7A). Concerning P. syringae, 9.5 μM armadillidin H induced 95.4% (±1.1%) permeabilization of the cell population while no CFU growth was observed (from 7.7 ± 0.5 to 0; Figure 7B). Thus, armadillidin H induced permeabilization of the bacterial membrane, thereby provoking, directly or not, cell death. Concerning the membrane integrity of A. brassicicola, as seen in Figure 8, very little intracellular green fluorescence (SYTOX Green) was detected in untreated hyphae (Figure 8A). By contrast, many intracellular green areas were observed after 4 h of exposure with 7.6 μM armadillidin H resulting from dye uptake (Figure 8B). Similar results were observed when hyphae were incubated with 0.1% Triton X-100 as a control (Figure 8C). Taken together, these results indicate that armadillidin H permeabilizes A. brassicicola hyphae.

To investigate whether armadillidin H has any effect on mammalian membranes, its hemolytic potency was assayed. After incubating fresh human erythrocytes with the peptide up to 19 μM concentrations, no hemoglobin release was observed, indicating that armadillidin H does not cause lysis of erythrocyte membrane (Figure 9).

Preliminary to the experimental structural evaluation, the sequence of armadillidin H was submitted to MedoR (MEtaserver of DisORder), a metaserver for predicting protein disorder (Lieutaud et al., 2008). Whatever the predictor questioned by the MedoR metaserver, no secondary structure could be predicted. The FoldIndex software (Prilusky et al., 2005) augurs armadillidin H to be unstructured in the range 1–53, i.e., from the first to the last residue. The two CD spectra of armadillidin H recorded in 100 mM phosphate buffer pH 6.0, with or without the addition of 50% TFE display the same shape. When a deconvolution program, whatever the bases of soluble proteins chosen – structured or not – is applied, a large majority of residues are predicted to be unstructured or in turn, with only 25–30% of residues which could be in an extended conformation. However, the prediction of β-sheet by CD is generally trickier than the prediction of helices. Due to the dilution by a factor of two, the CD spectrum with TFE is less intense, but totally superimposable with the spectra recorded without TFE (Supplementary Figure S1), which means that the peptide does not show any propensity to form secondary structures when the media mimics membrane environment.

The NMR spectra showed a weak dispersion of 1H chemical shifts in the amide region (between 7.7 and 8.5 ppm), indicative of an unstructured peptide. Moreover, Hα and Cα chemical shifts, known to be particularly sensitive to secondary structures, are restricted. Finally, the superimposition of TOCSY and NOESY spectra (data not shown) does not highlight any additional peaks in the NOESY spectra that could be indicative of medium or long range NOE connectivities typical of 3D structuration. Then armadillidin H is not structured in our experimental conditions (0.1 mM in phosphate buffer pH 5.2). The addition of 50% of TFE (Figure 10) does not lead in significant shifts (except for the NH protons sensitive to “pH” variations induced by the addition of TFE). Hα and Cα chemical shifts, sensors of secondary structure variations, are not influenced by the addition of TFE. This means that the peptide remains unstructured, even in presence of TFE 50%, then it does not show any propensity to form secondary structures when the media mimics membrane environment.