Introduction

Most plants rely for their nutrition on inorganic nitrogen (ammonium and nitrate), a macronutrient found to be limiting in most terrestrial ecosystems (Chapin et al., 2002). Legumes and actinorhizal plants have acquired the ability to establish a root nodule symbiosis with nitrogen-fixing bacteria known as rhizobia and frankiae, respectively. N2-fixing root nodule symbioses by actinorhizal plants are important contributors to ecosystems worldwide. Phylogenetically, this symbiosis occurs among eight plant families and three orders that have diverse global distributions (Benson and Silvester, 1993). Conversely, most important crop plants such as cereals are unable to use atmospheric N2 for their nutrition. Modern agriculture is therefore dependent on synthetic nitrogen fertilizers. The increasingly expensive fossil energy and the pervasive effects of atmospheric carbon dioxide associated with the industrial production of nitrogen fertilizer, together with the negative environmental impact of leached nitrogen fertilizers, have led to a renewed interest for biological nitrogen fixation.

How nitrogen fixation occurs has been studied in both free-living organisms and in symbioses for more than a century (Beijerinck, 1901; Wyss et al., 1941; Berry et al., 1990). However, key steps in the symbiotic process, including how fixed nitrogen is transferred from symbionts to host plants, are still only partially understood despite efforts made in this area (Benson and Silvester, 1993; Benson et al., 2011; Oldroyd et al., 2011; Pawlowski et al., 2011). High-throughput sequencing of expressed sequence tags (ESTs) and DNA arrays are powerful tools to understand biological processes and have been widely used recently to study nitrogen-fixing symbiotic interactions (Ampe et al., 2003; Alloisio et al., 2010; Hocher et al., 2011; Heath et al., 2012). Nevertheless, there is presently no obvious candidate gene identified in either rhizobia or Frankia to account for release of nitrogen compounds to plants. Although both symbiotic systems have some common features, both the bacteria and the plants involved are very different, and hence it is expected that different mechanisms have evolved for similar purposes in legumes and actinorhizal plants.

Frankia bacteria are able to fix nitrogen both as free-living organisms and in symbiosis. In both conditions, Frankia develops special cell types called vesicles for nitrogen fixation. These structures are formed at the tip of hyphae, are spherical with a diameter from 2 to 5 μm and are covered by several layers of hopanoid lipids to protect the nitrogenase enzyme from oxygen (Benson and Silvester, 1993; Berry et al., 1993). In vitro, growth in nitrogen-free medium (BAP−) has been found to induce the formation of these structures (Murry et al., 1984).

Defensin peptides, generally described as antimicrobial peptides, are effectors of innate immunity in plants and animals and thus permit to control growth of pathogens. In plants, members of the A family of defensins (A1, A2, A3 and A4) are known to be able to inhibit growth of a broad range of fungi without adverse effects on plant or mammalian cells. The A3 subfamily is known to cause membrane permeabilization and also hyphal branching (Fant et al., 1999; Thevissen et al., 1999). More recent studies have shown that antimicrobial peptides are also important in symbiotic relationships. An antimicrobial peptide expressed by insects has been found to control symbiotic bacteria (Login et al., 2011), and an extremely large gene family from Medicago truncatula that encodes polypeptides related to defensins was shown to have a nodule-specific expression and to be implicated in the terminal differentiation of bacterial symbionts (Mergaert et al., 2003; Van de Velde et al., 2010).

In actinorhizal symbiosis, a collection of Alnus glutinosa and Casuarina glauca ESTs has been sequenced from roots and 3-week-old nodules (Hocher et al., 2011). These analyses showed that several genes related to carbon and nitrogen metabolism, defense against pathogens or stress resistance were strongly upregulated in nodules. More recently, similar results were found for another actinorhizal plant, Datisca (Demina et al., 2013). Remarkably, all three actinorhizal plants induce a small family of defensin genes in their nodules (Hocher et al., 2011; Demina et al., 2013). In the present study, we sought to confirm the nodule-enhanced expression of defensin peptides produced by actinorhizal plants and to understand their role in the symbiotic process.

Materials and methods

Bacterial strains and growth conditions

Frankia alni ACN14a (Normand and Lalonde, 1982) was grown according to Schwencke (1991) at 28 °C in defined (BAP) medium containing 5 mM ammonium chloride (N-replete condition, BAP+) or not (N-fixing condition, BAP−) (Murry et al., 1984) with some minor modifications (Alloisio et al., 2010). The 5-day-old cultures were used for in vitro analyses.

Plant growth conditions and mRNA extraction

A. glutinosa seeds were harvested from an A. glutinosa specimen growing on the Rhône River banks in Lyon, France, as described previously (Alloisio et al., 2010; Hocher et al., 2011). Seeds were grown and inoculated with the compatible F. alni strain ACN14a (Normand and Lalonde, 1982) as described previously (Alloisio et al., 2010). For RNA extraction, non-inoculated roots at 0 and 7 days post inoculation (d.p.i.), inoculated roots (7 d.p.i.) and nodules (4 weeks post inoculation) were collected and immediately frozen in liquid nitrogen as done by Hocher et al. (2011). Five biological replicates were used for each condition. Total RNA was purified from roots and nodules using the RNeasy plant mini kit (Qiagen, Courtaboeuf, France) (Alloisio et al., 2010). Residual DNA was removed from RNA samples using the Turbo DNA free kit (Ambion, Aix-en-Provence, France). RNA was quantified using a Nano-Drop spectrophotometer (Thermo Fisher Scientific, Villebon sur Yvette, France) and qualitatively assessed using agarose gel electrophoresis and/or a Bioanalyzer 2100 according to the manufacturer’s instructions (Agilent, Les Ulis, France).

EST database and microarray expression analysis

The previously reported database of A. glutinosa ESTs and microarray analysis of these transcripts (Hocher et al., 2011), which are available in the Gene Expression Omnibus database (www.ncbi.nlm.nih.gov/geo; accession number GSE24153), were analyzed. An extensive iterative BLAST search was made in order to find ESTs encoding defensin-homologous peptides. Then, differential expression of all recovered genes was determined using the microarray data set, according to Hocher et al. (2011). Briefly, Student’s t-test was applied to compare nodules versus non-inoculated roots and inoculated roots (7 d.p.i.) versus non-inoculated roots and average fold changes were calculated.

RT-qPCR analysis

Reverse transcription-PCR (RT-PCR) and quantitative real-time reverse transcription PCR (RT-qPCR) were performed with the five biological replicates of nodules, inoculated roots and non-inoculated roots (as reference). RT was performed with 10 μg of total mRNA using Transcriptor Reverse Transcriptase and oligo (dT)15 primer (Roche, Boulogne-Billancourt, France). RT-qPCR was run on a LightCycler 480 (Roche) using LightCycler 480 SYBR Green I Master (Roche) under the following conditions: 95 °C for 5 min; and 45 cycles of 95 °C for 20 s, 60 °C for 20 s and 72 °C for 15 s. The primer set for the Ag5 gene (07J18c 5′-AAACGCATCGAAAGGTGTAA-3′ and 07J18d 5′-CCCCGATATACGGAGAGT-3′) was designed using Primer 3 (Rozen and Skaletsky, 2000) and AmplifX version 1.5.4 (Nicolas Jullien; CNRS, Aix-Marseille Université, http://crn2m.univ-mrs.fr/pub/amplifx-dist) programs. Two RT-qPCRs were run for each biological replicate. Expression values were normalized using the expression level of the ubi gene (AgUbi 5′-AGGAGTCCACCCTCCATCTT-3′ and 5′-TCAGAACTCTCGACCTCCAA-3′) that encodes ubiquitin in A. glutinosa (Hocher et al., 2011).

In silico analysis

All Alnus symbiotic upregulated peptide (ASUP) genes were analyzed for the presence of translation start and stop sites (http://insilico.ehu.es/translate/), and for signal peptides (SignalP, http://www.cbs.dtu.dk/services/SignalP/). All sequences were aligned using CLUSTAL_X (Thompson et al., 1997).

Ag5 peptide production

The mature Ag5 peptide (the Ag5 peptide without the signal peptide as predicted by SignalP; Petersen et al., 2011) was synthesized by Proteogenix (Schiltigheim, France). This chemically synthesized peptide was purified by high-performance liquid chromatography to a 96% level.

PI/SYTO9 analysis

Microscopic observations of bacteria were made after 24 h of incubation at 28 °C with and without Ag5 peptide (1 μM). Membrane porosity was assessed by addition of 30 μM propidium iodide (PI)/20 μM SYTO9 to cultures and 75 μM PI/50 μM SYTO9 in fresh nodule sections obtained by hand cutting. After 15-min incubation in the presence of the dyes, the cells were visualized with a Zeiss Axioskop fluorescence microscope (Marly-le-Roi, France).

In vitro treatments with Ag5 peptide

Treatments were done on cultures diluted to an OD600 of 0.03 or 0.10. These diluted cells (200 μl) were treated with Ag5 peptide at different concentrations (0.1, 0.3, 1, 3, 10 and 100 μM) and incubated from 1 to 7 days at 28 °C. For determination of cations and amino-acid analysis, 500 μl dilutions of OD600 of 0.03 were made in phosphate-buffered saline and were left for 7 days in contact with Ag5 before recovering supernatants and/or cells. Data handling (normalization and statistics) was performed using Excel. Each measurement was done in quintuplicate (biological repeats).

Colony-forming unit determination

A total of 50 μl of F. alni ACN14a cells treated or not with Ag5 1 μM for 24 h were deposited onto 6 cm Ø Petri dishes with BAP− medium and spread out using glass beads. After 1 month of incubation at 28° C, the number of colonies on each dish was determined.

Nitrogenase activity and respiration analyses

Determination of nitrogenase activity of cells treated or not with Ag5 ASUP, heated or not to 100 °C for 1 h, NCR035 peptide (from Medicago) and bovine serum albumin was made with the acetylene reduction assay using a 10% vol/vol acetylene concentration. Measures were made with a gas chromatograph (Girdel series 30, Suresne, France) after 4 days of growth in the presence of acetylene (five biological replicates). To estimate the metabolically active biomass, respiration was evaluated as the amount of CO2 (in p.p.m.) liberated by the same cells using a gas chromatograph (P200 MicroGC, Agilent Technology, Massy, France) and by the tetrazolium/formazan assay (Prin et al., 1990).

Determination of cation release

Ag5 peptide was added at different concentrations to F. alni ACN14a cells that were cultured for 7 days in phosphate-buffered saline. Supernatants were centrifuged at 12 000 g for 10 min to remove cells. Then, 250 μl of supernatants were diluted fourfold with sterile water and analyzed by ion chromatography.

Amino-acid analyses

Amino-acid analysis was performed by HPLC (Agilent 1100; Agilent Technologies) with a guard cartridge and a C18 reverse phase column (Zorbax Eclipse-AAA 3.5 μm, 150 × 4.6 mm, Agilent Technologies), according to the procedure specifically developed for this system (Henderson et al., 2000). Before injection, samples were buffered with borate at pH 10.2, and primary or secondary amino acids were derivatized with ortho-phthalaldehyde or 9-fluorenylmethyl chloroformate, respectively. The derivatization process, at room temperature, was automated using the Agilent 1313A autosampler. Separation was carried out at 40 °C, with a flow rate of 2 ml min−1, using 40 mM NaH2PO4 (eluent A, pH 7.8, adjusted with NaOH) as the polar phase and an acetonitrile/methanol/water mixture (45:45:10, vol/vol/vol) as the non-polar phase (eluent B). A gradient was applied during chromatography, starting with 20% of B and increasing to 80% at the end. Detection was performed by a fluorescence detector set at 340 and 450 nm wavelengths for excitation and emission, respectively (266/305 nm for proline). These conditions do not allow detection and quantification of cysteine and tryptophan, and hence only 18 amino acids were quantified. For this quantification, norvaline was used as the internal standard and the response factor of each amino acid was determined using a 250 pmol μl−1 standard mix of amino acids. The software used was the ChemStation for LC 3D Systems (Agilent Technologies).

15N/14N measurement

To assess whether the Ag5 peptide was catabolized, F. alni ACN14a cultures were grown for three subsequent subcultures in BAP+ medium with 15N-labeled NH4+ to convert all cellular nitrogen to 15N. A final subculture was made with and without 14N Ag5 peptide. Amounts of 15N (labeled cellular) and 14N (peptide) were measured using 4 mg per sample. Each sample was placed in a tin capsule, pyrolyzed and then passed through a reduction column to convert all nitrogen to N2. The N2 was then analyzed by isotope ratio mass spectrometry (SCI, CNRS, Villeurbanne, France).

Antibodies

The chemically synthesized Ag5 was used to develop a polyclonal anti-Ag5 antibody in rabbits by standard intravenous injection using a commercial service (Covalab, Villeurbanne, France; http://www.covalab.com).

Microscopy

For light microscopy, samples were observed using a × 10 objective on a Zeiss Axioskop microscope. For fluorescent microscopy, nodules and roots were dipped into a paraformaldehyde 4% (vol/vol) fixative solution in phosphate-buffered saline. After 1 week at 4 °C, the fixative was removed by rinsing several times with phosphate-buffered saline before embedding the tissue in 1.3% (wt/vol) agar. Subsequently, samples were dehydrated through a graded ethanol/H2O series and transferred to 1-butanol, at 4 °C, for 1 week. Samples in agar were then impregnated and embedded in melted Paraplast resin (Hexcel Corp., Stanford, CT, USA). Wax blocks were kept dust free until sectioning. Tissue sections (5 μm thick) were obtained using a Thermo Scientific Microm HM340E microtome (Thermo Fisher Scientific Inc.,Waltham, MA, USA). Sections were placed on poly-lysine-coated slides, dried overnight in a 37 °C oven, and stored at 4 °C before immunostaining. Confocal microscopy was made at the Centre Technologique des Microstructures–Plateforme de l’Université Claude Bernard Lyon 1.

Immunolocalization

To determine the localization of Ag5 in Alnus tissues, immunolocalization was performed. The 4-week-old nodules were used for this analysis. The immunostaining and microscopy were made according to Login et al. (2011).

Results

Defensin genes are induced in symbiosis

In a previous expression microarray analysis (Hocher et al., 2011), a gene encoding a defensin peptide was found to be present among the most highly upregulated genes in A. glutinosa nodules (3 weeks post inoculation). In the present study, an extensive screening of the A. glutinosa database (Hocher et al., 2011) found 15 different ESTs coding for defensin-like peptides that were all upregulated in nodules (Table 1). Moreover some of these peptides were also upregulated as early as 7 d.p.i. during the early infection stage (Table 1).

Table 1 ASUPs are highly upregulated during symbiosis
Full size table

To unravel the defensin peptide functions in the A. glutinosa–Frankia symbiosis, the Ag5 ASUP was selected for an in-depth study because the Ag5 gene displayed the highest differential expression at 7 d.p.i. and was ranked as the eleventh most upregulated gene in 3-week-old nodules (Hocher et al., 2011) (Table 1). Early induction of Ag5 was further confirmed by RT-qPCR, revealing a 6.15-fold upregulation (P=0.00783).

Bioinformatics sequence analysis showed that Ag5 ASUP is 94 amino acids (aa) long and contains a predicted N-terminal signal peptide of 25 aa (Supplementary Figure S1), thus leaving a 69 aa sequence after removal of the peptide leader. It is a medium length cationic peptide with a net charge of 11, an isoelectric point of 8.99 and a molecular weight of 8145.26 Da. The defensin structure present in most ASUP, including Ag5, is: Signal peptide-XnCX3SXTWX3CX5CX3CX3EXAX2GXCXnCXCXnCXn, (n=1−30), and this is similar to the pattern found in class A3 defensins (Fant et al., 1999) but differs from the one described for the NCR peptides of Medicago (Mergaert et al., 2003). Other important differences from these NCR peptides are their size, 69 aa in Ag5 and 63 aa on the average for the other ASUPs, whereas NCR peptides only have a mean length of 40 aa. Similarly, the number of cysteine residues is 4 or 6 in NCR peptides and 8 for ASUPs.

Ag5 colocalizes with Frankia in nodules

Ag5 was strictly colocalized with Frankia-infected nodule cells (Figure 1) and no signal was observed in other nodule tissues or in roots. Guan et al. (1997) proposed four zones from nodule apex to base: the Frankia-free meristematic zone, the infection zone with hyphae, a nitrogen-fixation zone with mature vesicles and the senescence zone. In longitudinal nodule sections, a fluorescence intensity gradient was found, with no detectable signal at the apex, a moderate intensity in the infection zone, an intense fluorescence in the central section and a lowered intensity in the senescing zone (Figures 1b and d). Within the infected cells of the nitrogen-fixation zone, the fluorescence signal is strongly associated with the vesicles (Figure 1h), demonstrating that the Ag5 peptide is targeted to the Frankia endosymbionts.

Figure 1
figure 1

Ag5 ASUP colocalizes with Frankia in nodules of Alnus glutinosa. Negative control with rabbit serum before intramuscular injection of Ag5 (a, c, e) and immunofluorescence localization of Ag5 with anti-Ag5 antibody (b, d, f, h). Longitudinal (a, b, c, d) and cross-sections (e, f) of the nodule are shown. (g) Alnus roots carrying nodules infected with F. alni ACN14a. Blue: DAPI; red: autofluorescence; green: Alexa Fluor 488 dye anti-rabbit antibody (Life Technologies, Saint Aubin, France). c, infected cells; h, hyphae; nc, noninfected cells; v, vesicles; x, xylem. Zones in nodule: I, meristem zone (free of Frankia), II, infection zone (mostly hyphae of Frankia), III, nitrogen-fixation zone (mostly vesicles of Frankia) and IV, senescence zone.

Full size image

Ag5 binding in N-fixing cultures of ACN14a, treated or not with Ag5, was also analyzed in order to determine the localization of the peptide in the bacterial cells (Supplementary Figure S2). Ag5 was detected in both hyphae and vesicles but only at the periphery of the Frankia cells, suggesting binding of the peptides to the bacterial cell walls.

Effects of Ag5 on in vitro grown Frankia

To address the physiological function of Ag5, cultures of F. alni were treated with the Ag5 peptide in vitro. Evaluation of the response of F. alni ACN14a cells to increasing concentrations of peptide was made by measuring nitrogen-fixation activity and cellular respiration. Both analyses gave similar results with the maintenance of strong activities when the Ag5 peptide concentration was equal or inferior to 1 μM but with a sharp drop when the concentration was increased (Figure 2). However, a notable difference was observed, with cellular respiration gently decreasing at low concentrations of peptide, whereas nitrogen-fixation activity remained stable or indeed increased until after a 1 μM concentration was reached. Other peptides used as controls showed no significant differences, whereas heat-treated Ag5 showed similar effects to the nontreated one (Supplementary Figure S3). The ability of bacteria to grow on BAP+ plates was also tested after treatment with 1 μM Ag5. Despite the high metabolic activity of the bacteria at this peptide concentration, we observed a decrease of two orders of magnitude in colony-forming ability (from 400 to 4 colony-forming units). The symbiotic function of Frankia inside the host is nitrogen fixation. Therefore, we used nitrogen-fixation activity as an indicator to determine the maximum tolerance to the Ag5 peptide and a 1 μM concentration of the peptide was used in further studies.

Figure 2
figure 2

Strong activity of Frankia cells is detected when concentration of Ag5 ASUP is ≤1 μM. Nitrogen-fixation activity is measured by the acetylene reduction activity per μl of culture. Triangles, respiration analyses; diamonds, CO2 p.p.m.; squares, tetrazolium/formazan OD490 nm.

Full size image

Ag5 increases membrane porosity

Defensin-type antimicrobial peptides are known to permeabilize bacterial membranes and this is considered a major mode of antimicrobial activity for these peptides (Brogden, 2005). To assess the effect of the Ag5 peptide on bacterial membrane structures in vitro, treated and nontreated Frankia cultures were stained with PI and SYTO9. SYTO9 is a freely diffusing DNA fluorescent stain, whereas the PI fluorescent DNA stain can only penetrate cells following lethal or nonlethal perturbations in membrane permeability (Boulos et al., 1999). Untreated Frankia cells showed mostly SYTO9-stained (green) hyphae (Figures 3e, e′ and e″), whereas the proportion of PI-stained (red) hyphae was higher in cells treated with 1 μM Ag5 ASUP (Figures 3f, f′ and f″). However, the most important differences were observed in the nitrogen-fixing Frankia vesicles. The vesicles were mostly stained red with PI after addition of Ag5 ASUP to the medium (Figures 3f, f′ and f″), whereas they were mostly stained green with SYTO9 in untreated control cultures (Figures 3e, e′ and e″). Thus, the Ag5 peptide at 1 μM induces membrane permeabilization without affecting nitrogenase activity.

Figure 3
figure 3

Vesicle permeability in nodules and in vitro. SYTO9–PI analysis shows porosity differences between Frankia hyphae and vesicles in fresh nodule sections (scale 5 μm) observed by confocal microscopy (a, b, c, d) and in Frankia cultures without Ag5 (e, e′, e″) and with 1 μM Ag5 (f, f′, f″) (scale 10 μm) observed by fluorescence microscopy. Phase-contrast (a, e, f); green fluorescence with SYTO9 (b, e″, f″); red fluorescence with PI (c, e′, f′); superimposed image (d). h, hyphae; v, vesicles. Arrows indicate vesicles.

Full size image

Next, we evaluated Frankia membrane porosity in nodules using the same SYTO9–PI staining procedure (Figures 3a–d). Confocal microscopy of nodule sections showed that hyphae were mostly stained green with SYTO9 (Figures 3b and d), suggesting membranes were not affected. In contrast, the vesicles were strongly stained red with PI, suggesting an increased porosity (Figures 3c and d). These results are in agreement with the in vitro activity of the Ag5 peptide and the in vivo localization of the peptide in the nodule cells, both suggesting that the peptide is affecting vesicles more than hyphae in live cultures.

Ag5 releases nitrogen-rich compounds from Frankia

We analyzed the possibility that the Ag5 peptide facilitates the release of nitrogen-containing metabolites from Frankia cells. First, we examined the release of amino acids from cultured cells. Significant differences were found between supernatants of Ag5-treated and untreated cells (Figure 4 and Supplementary Table S1). Remarkably, some amino acids were released whereas others were not. Glutamine (Gln) was 200 times and glutamic acid (Glu) was 30 times more abundant in supernatants of cells treated with 1 μM Ag5 grown under nitrogen-fixing condition, with a concentration increase of ∼1 μM for Gln and 0.5 μM for Glu. Other amino acids such as arginine (Arg), tyrosine (Tyr), lysine (Lys), methionine (Met), leucine (Leu) and asparagine (Asn) were also recovered in larger amounts in treated cell supernatants, but much less so than Gln and Glu. Amino acids present in Frankia cells of the same cultures were also analyzed (Figure 4 and Supplementary Table S1). The amino-acid abundance was generally not affected by the peptide treatment except for Glu whose concentration dropped strongly with peptide concentrations of 0.3 μM or higher. Thus together, these measurements indicate an important release of Gln in the spent medium of Ag5-treated cells, correlated with a drop of the cellular Glu level.

Figure 4
figure 4

Release of amino acids from N-fixing Frankia cultures following treatment with Ag5 ASUP after 7 days in phosphate-buffered saline (PBS). (a) Amino acids in supernatants and (b) amino acids in Frankia cells. Error bars are s.d.

Full size image

In addition, the release of other compounds from cells was also evaluated. Quantification of NH4+, Mg2+, Ca2+ and K+ levels in the culture supernatant was performed by ion chromatography (Supplementary Table S2). The quantity of these ions after addition of Ag5 ASUP was not affected at peptide concentrations up to 3 μM, indicating that at these peptide concentrations, there is no unspecific leakage out of the cells. Only when a high concentration of Ag5 (10 μM) was used, incompatible with physiological activities like N2 fixation and respiration, was an increase in NH4+ observed in the cell supernatant.

To exclude the possibility that the amino-acid increase in the spent medium results from Ag5 peptide degradation, the stability of Ag5 was assessed in the presence of Frankia. The 15N/14N ratios were determined in a 15N-labeled F. alni ACN14a culture with 14N Ag5 added. All the nitrogen recovered from the supernatant was 15N, indicating that the nitrogen originated only from the bacterial cells and not from degradation of the added Ag5 peptide.

Discussion

The Ag5 gene displays strongly increased expression and was ranked as the eleventh most upregulated gene in 3-week-old Alnus nodules (Hocher et al., 2011). This level of induction was comparable to that of known actinorhizal nodulins, suggesting a central role for this peptide in the host response to the infection with Frankia bacteria. Nevertheless, the upregulation of Ag5 and the other ASUPs does not necessarily mean by itself that they play a role in symbiosis as these peptides could be generated as a generic response to microbial invasions. Interestingly, ASUP-like genes that were also upregulated in nodules of Casuarina plants (Hocher et al., 2011) were not upregulated in mycorrhizae (Tromas et al., 2012), supporting the possibility of a specific role for them in the Frankia symbiosis. In addition, immunolocalization showed that the Ag5 peptide specifically targets the Frankia cells in the nitrogen-fixation zone of the nodule. Ag5 ASUP was most abundant on the vesicle surfaces, suggesting a direct relationship with the symbiotic process, beyond the first steps of the interaction. This specific binding to vesicles could be related to the presence of the thick hopanoid lipidic external layers in vesicles that are not present on the hyphae (Berry et al., 1993).

We demonstrate here that Ag5 peptide caused changes in the membrane properties of Alnus-infective Frankia strains and a release of cellular metabolites from cells. In the context of pathogenic interactions, such a leakage induced by antimicrobial peptides may help to weaken pathogens by decreasing or abolishing membrane potential, causing leakage of nutrients or releasing effectors that will boost host defenses. In symbiosis, the Ag5-induced membrane porosity may rather play a role in nutrient exchanges. The analysis of membrane porosity in nodule sections and in vitro cells suggests a specificity of the Ag5 action on nitrogen-fixing vesicles that were strongly permeable to PI whereas hyphae were not. This indicates an increased porosity of nitrogen-fixing vesicles relative to vegetative hyphae in host cells. This observation contrasts with those obtained from the Rhizobium–legume nodules, where the nitrogen-fixing Rhizobium bacteria were not stained with PI (Haag et al., 2011). However, our observations confirm a previous hypothesis proposing that the permeability of vesicles in nodules is higher than that of in vitro vesicles (Tisa and Richards, 2004). The increased membrane permeability induced by Ag5 and possibly other ASUPs may affect the capacity of the Frankia cells to reproduce as indicated by the drop in colony-forming capability of in vitro grown cells treated with 1 μM Ag5 but this does not or hardly affect nitrogen-fixation and metabolic activity of those cells.

Our results showed the release of amino acids from cultured Frankia cells to the media, induced by the presence of Ag5 ASUP. The major amino acid released from Ag5-treated Frankia is glutamine. Enzyme activity analyses of Frankia grown in N2-free media have shown an important activity of glutamine synthetase (GS) in vesicles and hyphae, whereas glutamine oxo-glutarate amino transferase (GOGAT), glutamate dehydrogenase (GDH) and alanine dehydrogenase (AlaDH) are restricted to hyphae (Schultz and Benson, 1990). Thus, the increased level of Gln in the supernatant agrees with this previous study and with the strong reduction in the cellular glutamic acid concentration after Ag5 treatment (Figure 4). However, the total amount of Gln in cells treated with Ag5 is higher than the amounts measured in control cells. This could be because of a change in enzymatic activities (for example, restriction of further assimilation for purine biosynthesis and so on) following Ag5 treatment. Another possibility is release of NH4+ to the supernatant, and this release could trigger the very fast formation of Gln from Glu inside the cells as shown by Berry et al. (1990). The efflux of glutamate from in vitro grown cells may be simply because of the fact that glutamate is the most abundant amino acid in F. alni cells (Berry et al., 1990) linked to its position at the crossroads of many metabolic pathways (Kissen et al., 2010). In symbiotic tissues, enzymatic activities related to ammonia assimilation are on several points different from those in pure cultures, with, in particular, a GS activity that is mostly restricted to the plant host cells (Hirel et al., 1982; Guan et al., 1996), resulting in a lowered Glu–Gln cycling in the Frankia vesicles and an expected NH4+ accumulation. Transcriptomic analyses also point out that in nodules Frankia GSII expression level is low compared with nitrogen-fixing condition in the free-living state (Alloisio et al., 2010). Thus, ammonium could be the major nitrogen compound exuded from the bacteria to the host cells helped by the permeability increase because of Ag5 and possibly other ASUPs.

An ideal confirmation for our results would be the knockdown of Ag5 expression by RNA interference mediated by Agrobacterium transformation of Alnus to determine its effect in nodules. However, the presence of several homologous peptides with probably related functions makes this approach unlikely to be successful. Moreover, previous Alnus transformation experiences have always failed or only transitorily transformed protoplasts were obtained (Mackay et al., 1988; Franche et al., 1998).

In recent years, reports about small secreted cysteine-rich peptides that control many aspects of plant signaling have emerged. Their known functions range from the control of reproduction to defense against pathogens (Marshall et al., 2011). The role of antimicrobial peptides has been well studied in the context of host–pathogens interactions (Maroti et al., 2011), and more recently in the symbiosis between the legume Medicago and Sinorhizobium where the NCR peptides block division of endosymbiotic cells, resulting in endoreduplication of the intracellular rhizobia (Van de Velde et al., 2010). Here we describe a new role for the antimicrobial peptide Ag5. Our work shows that the membrane porosity of in vitro nitrogen-fixing Frankia is modified by Ag5 to release Gln and Glu. This suggests that this peptide and possibly also the other ASUP peptides are implicated in trophic exchanges between the endosymbiont and the host. Our results suggest that these peptides act as host effectors that are plugged into the membrane of the endosymbionts and form specific channels permitting to release nutrients from the nitrogen-fixing Frankia vesicles. This work paves the way for the study of the Frankia symbiosis system, in which further studies of other ASUPs and their functions should be pursued.