The grapevine flagellin receptor VvFLS2 differentially recognizes flagellin-derived epitopes from the endophytic growth-promoting bacterium Burkholderia phytofirmans and plant pathogenic bacteria



  • The role of flagellin perception in the context of plant beneficial bacteria still remains unclear. Here, we characterized the flagellin sensing system flg22–FLAGELLIN SENSING 2 (FLS2) in grapevine, and analyzed the flagellin perception in the interaction with the endophytic plant growth-promoting rhizobacterium (PGPR) Burkholderia phytofirmans.
  • The functionality of the grapevine FLS2 receptor, VvFLS2, was demonstrated by complementation assays in the Arabidopsis thaliana fls2 mutant, which restored flg22-induced H2O2 production and growth inhibition. Using synthetic flg22 peptides from different bacterial origins, we compared recognition specificities between VvFLS2 and AtFLS2.
  • In grapevine, flg22-triggered immune responses are conserved and led to partial resistance against Botrytis cinerea. Unlike flg22 peptides derived from Pseudomonas aeruginosa or Xanthomonas campestris, flg22 peptide derived from B. phytofirmans triggered only a small oxidative burst, weak and transient defense gene induction and no growth inhibition in grapevine. Although, in Arabidopsis, all the flg22 epitopes exhibited similar biological activities, the expression of VvFLS2 into the fls2 background conferred differential flg22 responses characteristic for grapevine.
  • These results demonstrate that VvFLS2 differentially recognizes flg22 from different bacteria, and suggest that flagellin from the beneficial PGPR B. phytofirmans has evolved to evade this grapevine immune recognition system.


In order to counter-attack infections and microbial colonization, plants need to detect potentially pathogenic microorganisms and to launch an innate immune response. The first layer of this recognition is ensured by pattern recognition receptors (PRRs) which detect pathogen- or microbe-associated molecular patterns (PAMPs or MAMPs) or plant endogenous molecules released during pathogen invasion, called damage-associated molecular patterns (DAMPs; Boller & Felix, 2009; Dodds & Rathjen, 2010; Monaghan & Zipfel, 2012). This recognition leads to the so-called MAMP-triggered immunity (MTI). Successful pathogens can secrete effectors that suppress or interfere with MTI, resulting in effector-triggered susceptibility (ETS). A second level of perception involves the direct or indirect recognition of pathogen effectors by intracellular immune receptors, leading to effector-triggered immunity (ETI; Jones & Dangl, 2006; Hann et al., 2010).

Flagellin, the main building protein of eubacterial flagella, is a potent defense elicitor in different plant species and is one of the best studied MAMPs (Boller & Felix, 2009). Flagellin is required for bacterium motility and chemotaxis in the environment, and therefore represents a key virulence factor for many pathogenic bacteria (Finlay & Falkow, 1997). The plant PRR responsible for flagellin perception is the leucine-rich repeat receptor-like kinase (LRR-RLK) FLAGELLIN SENSING 2 (FLS2; Gomez-Gomez & Boller, 2000; Chinchilla et al., 2006). FLS2 was first identified in Arabidopsis thaliana (AtFLS2; Gomez-Gomez & Boller, 2000), but functional FLS2 orthologs have since been identified in tomato (LeFLS2; Robatzek et al., 2007), Nicotiana benthamiana (NbFLS2; Hann & Rathjen, 2007) and rice (OsFLS2; Takai et al., 2008).

flg22, a 22-amino-acid peptide present in the N-terminal part of flagellin, is sufficient to trigger FLS2-dependent responses in several plant species (Boller & Felix, 2009). The synthetic peptide flg22, based on the flagellin sequence of Pseudomonas aeruginosa, has become a commonly used epitope that substitutes the effect of flagellin. flg22 induces rapid extracellular alkalinization, reactive oxygen species (ROS) production, activation of a mitogen-activated protein kinase (MAPK) cascade, massive transcriptional reprogramming including the up-regulation of pathogenesis-related (PR) genes, callose deposition, increased ethylene production and seedling growth inhibition in Arabidopsis (Felix et al., 1999; Gomez-Gomez et al., 1999; Asai et al., 2002; Zipfel et al., 2004). The role of flagellin perception in resistance towards pathogenic bacteria has been demonstrated in Arabidopsis against adapted and non-adapted bacteria (Zipfel et al., 2004; Hann & Rathjen, 2007; Forsyth et al., 2010; Zeng & He, 2010).

However, although MTI is well documented in the context of plant–pathogenic bacteria interactions, knowledge about the perception of endophytic plant growth-promoting rhizobacteria (PGPRs) is very scarce (Van Wees et al., 2008; Zamioudis & Pieterse, 2012). PGPRs colonize the rhizosphere of different plant species and provide beneficial effects, such as enhanced plant growth and induced systemic resistance (ISR) to biotic and abiotic stresses (Ait Barka et al., 2000; Compant et al., 2005a; Lugtenberg & Kamilova, 2009). Burkholderia phytofirmans strain PsJN (B. phytofirmans) is notably an endophytic PGPR of potato, tomato and grapevine (Sessitsch et al., 2005; Lo Piccolo et al., 2010; Mitter et al., 2013). Recently, B. phytofirmans has also been shown to colonize Arabidopsis and to promote its growth (Poupin et al., 2013; Zuniga et al., 2013). In grapevine (Vitis vinifera), B. phytofirmans closely attaches to the rhizodermal cell walls, extensively colonizes the root surface and penetrates the root internal tissues to further spread into aerial parts of the plant, such as stems and leaves, via xylem vessels (Compant et al., 2005b). By altering grapevine metabolism, B. phytofirmans confers better tolerance to Botrytis cinerea infection and to cold stress (Ait Barka et al., 2000, 2006; Fernandez et al., 2012; Theocharis et al., 2012). In grapevine cells, B. phytofirmans triggers a transient extracellular alkalinization, the production of salicylic acid (SA) and defense-related transcripts, suggesting that it is perceived by V. vinifera potentially via MAMP detection (Bordiec et al., 2011).

Several MAMPs from PGPRs have already been characterized. For instance, lipopolysaccharides (LPSs), pyoverdine and lipopeptides trigger immunity in several plant species (Zeidler et al., 2004; Tran et al., 2007; van Loon et al., 2008; Jourdan et al., 2009). Flagellin from Pseudomonas putida or Pseudomonas fluorescens is also recognized by tobacco and Arabidopsis cells (van Loon et al., 2008; Millet et al., 2010). Interestingly, the flg22 epitopes derived from Agrobacterium tumefaciens, Rhizobium meliloti and Ralstonia solanacearum flagellins are highly divergent and are not recognized by Arabidopsis and tomato (Felix et al., 1999; Bauer et al., 2001; Pfund et al., 2004). However, the potential differences between endophytic and pathogenic bacterial perception through their flagellins are not known.

In this study, we report the identification of VvFLS2, the V. vinifera receptor of bacterial flagellin, and the comparison of the recognition specificities of VvFLS2 and AtFLS2 in relation to their capability to perceive flagellin-derived immunogenic epitopes from endophytic and pathogenic bacteria. The commonly used flg22 peptide from P. aeruginosa triggered early signaling events, the expression of a set of defense genes and plant growth inhibition in grapevine. We further showed that flg22-elicited responses induced grapevine immunity against the necrotrophic fungus B. cinerea. We demonstrated that VvFLS2 is the functional ortholog of AtFLS2, as it complements the lack of responsiveness in an Arabidopsis null fls2 mutant. Our work provides the first description of an active grapevine PRR–MAMP pair. We also provide evidence that the grapevine immune response triggered by flg22 from the endophytic bacterium B. phytofirmans is lower than those triggered by the pathogen-derived flg22 peptides from P. aeruginosa or X. campestris. Interestingly, these differences were not observed in wild-type (WT) Arabidopsis, but were gained on expression of VvFLS2 in the Arabidopsis fls2 mutant, suggesting that FLS2 itself underlies the differences observed in the responses triggered by these different peptides in Arabidopsis and grapevine. To our knowledge, VvFLS2 is the first characterized receptor that differentially recognizes flg22 epitopes from a PGPR and plant-pathogenic bacteria, suggesting an evolutionary mechanism of grapevine innate immunity evasion by its beneficial endophytic bacterium B. phytofirmans.

Materials and Methods

Plant and cell culture materials

Grapevine plants (V. vinifera cv Marselan) were grown in glasshouse conditions, as described in Steimetz et al. (2012). Grapevine in vitro plantlets (V. vinifera cv Chardonnay) were micropropagated by nodal explants grown on Murashige and Skoog (MS) agar medium, as described earlier (Theocharis et al., 2012). Grapevine wild-type and apoaequorin-transformed cells (V. vinifera cv Gamay) were cultivated as described previously (Vandelle et al., 2006).

Arabidopsis (Arabidopsis thaliana) plants from WT Columbia (Col-0) and fls2 (SAIL_691_C4; Zipfel et al., 2004) and fls2/p35S::VvFLS2-GFP mutants were grown in Jiffy-7 peat pellets (Jiffy, Lorain, OH, USA) in a controlled growth chamber, as described in Dubreuil-Maurizi et al. (2011). Arabidopsis Col-0 cells were cultivated as described in Dubreuil-Maurizi et al. (2011), except that cell suspension cultures were subcultured every 7 d by transferring 5 ml of cells into 100 ml of fresh liquid MS medium including Nitsch vitamins (M0256; Duchefa, Haarlem, the Netherlands) supplied with 30 g l−1 sucrose, 0.5 mg l−1 1-naphthalene acetic acid and 50 μg l−1 kinetin.


Sequences of flg22 peptides from P. aeruginosa strain PAK (QRLSTGSRINSAKDDAAGLQIA), X. campestris pv campestris strain 305 (QRLSSGLRINSAKDDAAGLAIS), B. phytofirmans strain PsJN (TRLSSGKRINSAADDAAGLAIS) and A. tumefaciens strain C58C1 (ARVSSGLRVGDASDNAAYWSIA) were retrieved from the UniProt database and purchased from Proteogenix (Schiltigheim, France) (purity superior to 95%). Laminarin was provided by Goëmar Laboratories (St Malo, France).

Protection assays against B. cinerea

Leaf disks from the second and third adult top leaves from at least 12 different grapevine plants were floated on elicitor solutions (10 μM flg22 or 2.5 g l−1 laminarin) for 24 h, washed and placed on humid filter paper for the next 24 h. Forty-eight hours post-treatment (hpt), at least 30 disks per condition were inoculated with B. cinerea strain B05.10 (5000 conidia in a 6-μl droplet of potato dextrose broth 1 : 4) and the infection intensity was assessed 3 d post-inoculation (dpi) by measuring the average lesion diameter (Aziz et al., 2003).

Bioassays in cell cultures

ROS production, cytosolic Ca2+ variations ([Ca2+]cyt) and MAPK phosphorylation analysis in grapevine cells were performed according to Dubreuil-Maurizi et al. (2010). Briefly, cells were washed with M10 buffer (175 mM mannitol, 0.5 mM CaCl2, 0.5 mM K2SO4, 10 mM Mes, pH 5.3) and resuspended at 0.1 g fresh weight of cells (FWC) ml−1. ROS production was measured by chemiluminescence of luminol. Variations in free cytosolic calcium [Ca2+]cyt were followed using apoaequorin-expressing cells. MAPK phosphorylation was detected by Western blot using antibody raised against a phospho-Thr202/Tyr204 peptide of human phosphorylated extracellular regulated protein kinase 1/2 (α-pERK1/2, Cell Signalling, Danvers, MA, USA). Transfer quality and homogeneous loading were checked by Ponceau Red staining (0.5% Ponceau Red, 1% acetic acid). The oxidative burst in Arabidopsis cells was monitored using the same protocol as for grapevine, according to Dubreuil-Maurizi et al. (2011).

Gene expression by real-time quantitative polymerase chain reaction (qPCR)

After treatment, c. 100 mg of grapevine tissue were briefly ground with a micro-mortar before the addition of Extract All Reagent (Invitrogen). RNA extraction was then carried out following the manufacturer's instructions and the final pellet was resuspended in 30 μl of RNAse-free water. RNA purity and concentration were determined by spectrophotometry and checked on a 1% agarose gel. Reverse transcription of RNA was performed with 150 ng of total RNA, using M-MLV reverse transcriptase (Invitrogen), following the manufacturer's protocol. Real-time qPCRs were carried out as described previously (Theocharis et al., 2012), except that a 1 : 30 dilution of reverse transcribed RNA was used. The transcript level was calculated using the comparative ΔΔCt method (Livak & Schmittgen, 2001) with the EF1α and 60SRP genes as internal controls for normalization (Dubreuil-Maurizi et al., 2010; Gamm et al., 2011). Grapevine primer sequences used for qPCR have been described previously in Bordiec et al. (2011), except for VvPR1.2 (National Center for Biotechnology Information (NCBI)# XM_002274239.1), with the forward (5′-GCGTGGGTGGGGAATGCCGA-3′) and reverse (5′-GATGTTGTCCCTGATAGTTGCC-3′) primers.

Cloning of green fluorescent protein (GFP)-tagged VvFLS2 and Arabidopsis transformation

The full-length coding sequence of VvFLS2 (NCBI# XM_002272283.2) was amplified by PCR from elicited grapevine cell cDNAs using a proof-reading Pfu DNA polymerase (Finnzymes/Thermo Scientific, Pittsburgh, PA, USA) and gene-specific primers (5′-CACCATGGTGTCTGAAAGAGTCAGTTTAATCC-3′ and 5′-GGCTGATGATGATGGTAATGGAGG-3′). A gel-purified (Wizard® SV Gel and PCR Clean-Up System, Promega) single PCR product of the expected size was first directionally subcloned into pENTR™/D-TOPO® vector (Invitrogen), and then inserted into Gateway expression vector pK7FWG2 (kanamycin resistance; Karimi et al. (2007)) using Gateway LR Clonase™ II enzyme mix (Invitrogen), to give the construct p35S::VvFLS2-GFP. The correct sequence and orientation of the insert in the expression vector were verified by sequencing. The transformation of the Arabidopsis fls2 mutant (BASTA resistance) with p35S::VvFLS2-GFP was performed by floral dip, according to Clough & Bent (1998). Kanamycin-resistant transgenic plants were screened for the presence of the oxidative burst in leaf disks after flg22 treatment, as described previously (Zipfel et al., 2006). ROS production in leaf disks (two disks per plant) was analyzed by luminol reaction (100 μM luminol, 10 μg ml−1 horseradish peroxidase (HRP)) using a microplate luminescence reader (Mithras LB 940, Berthold Technologies, GmbH & Co. KG, Bad Wildbad, Germany). Total proteins were extracted by grinding two 7.5-mm-diameter leaf disks in 100 μl of 3 × loading buffer containing 50 mM dithiothreitol (DTT), and the presence of VvFLS2-GFP was detected by Western blot using rabbit anti-GFP antibodies (AMS Biotechnology, Abingdon, UK) and goat anti-rabbit-HRP secondary antibodies (Sigma-Aldrich).

Phylogenetic analysis

Protein sequences were aligned with the MUSCLE program implemented in (Dereeper et al., 2008). The maximum-likelihood phylogenetic tree was generated with SeaView version 4 software (Gouy et al., 2010) using the LG substitution model and bootstrapping with 1000 replications. The sequences of other Arabidopsis RLKs, such as EF-TU RECEPTOR (AtEFR) and Wall-associated kinase 1 (WAK1), were used as outgroups.

Confocal microscopy

Confocal microscopy was performed using a Leica TCS SP2-AOBS confocal laser scanning microscope with a 40× oil-immersion objective (numerical aperture 1.25; Leica, Nanterre, France). Fluorescent markers were visualized by excitation with an argon laser at 488 nm. GFP and FM4-64 emissions were bandpass filtered at 500–525 nm and 616–694 nm, respectively.

Growth inhibition assays on grapevine and Arabidopsis plantlets

For V. vinifera (cv Chardonnay), 2-wk old in vitro plantlets were transferred from 25-mm glass tubes to Magenta boxes containing 20 ml of liquid modified MS medium (10 g l−1 sucrose) supplemented with the different flg22 peptides at 1 μM. For Arabidopsis, seeds of the different lines (Col-0, fls2 and fls2/p35S::VvFLS2-GFP) were first germinated on solid half-strength MS medium (Duchefa M0222, 5 g l−1 sucrose), and then transferred individually in 24-well culture plates containing half-strength MS liquid medium supplemented with the different flg22 peptides at 1 μM. The fresh weight of 10 individual plants was scored after 1 and 2 wk for Arabidopsis and grapevine, respectively.


flg22 induces immune responses and resistance against B. cinerea in grapevine

To determine whether flagellin perception by grapevine triggers the responses commonly observed in Arabidopsis, tomato or tobacco (Felix et al., 1999; Gomez-Gomez & Boller, 2000; Hann & Rathjen, 2007), we first characterized the early signaling events and defense gene expression induced by flg22 (from P. aeruginosa) in V. vinifera cell suspensions. flg22 treatment induced a dose-dependent oxidative burst (Supporting Information Fig. S1). The saturating flg22 concentration of 1 μM was then used to study the defense-related events in grapevine cells. Treatment with 1 μM flg22 induced a transient increase in free [Ca2+]cyt that peaked after 4 min (Fig. 1a) and an oxidative burst with maximal H2O2 production detected at 15 min (Fig. 1b). From 5 to 30 min, flg22 induced rapid and transient phosphorylation of two MAPKs with relative molecular masses of 45 and 49 kDa, which was not observed in control cells (Fig. 1c). The expression of defense marker genes activated by different elicitors (Aziz et al., 2003, 2007; Bordiec et al., 2011) was then followed by qPCR. Among them, flg22 induced the expression of two defense genes encoding an acidic chitinase (Chit4c, Fig. 1d) and a protease inhibitor (PR6, Fig. 1e) as early as 1 hpt, with the strongest induction detected at 6 hpt.

Figure 1.

flg22-induced responses and immunity against Botrytis cinerea in grapevine. (a) Free cytosolic calcium variations measured with apoaequorin-expressing grapevine cells. (b) Time course of H2O2 production detected by luminol chemiluminescence. Values are means ± SD from three independent experiments. FWC, fresh weight of cells. (c) Activation kinetics of two mitogen-activated protein kinases (MAPKs) detected by Western blot with a human phosphorylated extracellular regulated protein kinase 1/2 (α-pERK1/2). Homogeneous loading was checked by Ponceau Red staining. (d, e) Expression of defense genes encoding an acidic chitinase (Chit4c, d) and a proteinase inhibitor (PR6, e) measured by quantitative polymerase chain reaction (qPCR). Means of triplicate data from independent biological experiments were normalized by housekeeping genes EF1a and 60SRP and compared with time 0, set as 1. (f) Development of B. cinerea at 3 d post-inoculation (dpi) on grapevine leaf disks pretreated 48 h before with 10 μM flg22 or 2.5 g l−1 laminarin (Lam), and compared with control. Values represent the mean lesion diameter ± SD ( 30 lesions per modality). Asterisks indicate statistically significant difference between control and elicitor treatment (t-test; *, < 0.05). (a–e) illustrate kinetics of responses in Vitis vinifera cell suspensions treated with 1 μM flg22 (from Pseudomonas aeruginosa), whereas (f) shows 10 μM flg22-triggered immunity in plant leaf disks. For (a), (c) and (f), similar results were obtained in at least three independent experiments.

We further investigated the efficiency of flg22-induced immunity in V. vinifera leaf disks challenged with the necrotrophic fungus B. cinerea. flg22 treatment applied 48 h before pathogen inoculation significantly reduced the B. cinerea lesion diameter, compared with control leaf disks (Fig. 1f). Results were comparable with those obtained by pretreatment with the β-1,3-glucan laminarin, described to trigger protection against B. cinerea in grapevine (Aziz et al., 2003).

In silico characterization of the predicted grapevine FLS2 receptor: VvFLS2

As grapevine responds to flg22 treatment, we aimed to identify the corresponding flagellin receptor in V. vinifera. The grapevine genome carries one clear gene encoding the ortholog of AtFLS2, hereby designated as VvFLS2 (CAN78669.1/XP_002272319.2), which is clearly distinct from other grapevine LRR-RLKs (Fig. 2a). Alignment with AtFLS2 permitted the identification of an upstream sequence encoding the 26 amino acids of the VvFLS2 signal peptide that was unpredicted by NCBI. The full-length VvFLS2 gene (KF562727) consists of an open reading frame of 3516 bp and contains a small 105-bp intron at position 1050, a location conserved amongst Arabidopsis AtFLS2, tomato LeFLS2 and rice OsFLS2 (Fig. 2b). Therefore, FLS2 homologs exhibit a highly conserved gene structure.

Figure 2.

In silico characterization of the putative grapevine VvFLS2 ortholog. (a) Maximum-likelihood phylogenetic tree showing the relationship between the protein sequences (GenBank) of the Arabidopsis FLAGELLIN SENSING2 (AtFLS2), its identified orthologs in tomato (LeFLS2), rice (OsFLS2) and the most similar protein sequences of Vitis vinifera, including the predicted VvFLS2. Only bootstraps higher than 500 (from 1000) are presented. (b) Exon–intron architecture of coding regions of VvFLS2 and FLS2 orthologs. Black boxes represent exons, numbers represent codons. Bar, 200 bp. (c) Deduced amino acid sequence of the cloned VvFLS2 cDNA with indication of the predicted signal peptide, N-terminal domain (LRRNT), leucine-rich repeat (LRR) domain, extracellular juxtamembrane region, single transmembrane domain, intracellular juxtamembrane region, serine/threonine (S/T) kinase domain and C-terminal tail. Amino acids matching LRR consensus according to Mueller et al. (2012) are shaded in gray. Highlighted in bold and underlined are amino acids affecting FLS2 signaling in Arabidopsis when mutated (Cao et al., 2013; Robatzek & Wirthmueller, 2013).

The predicted encoded protein of 1171 amino acids, called VvFLS2, contains a signal peptide, an LRR ectodomain, a single transmembrane domain and a non-RD-type intracellular kinase domain, also found in other FLS2 proteins (Fig. 2c; Boller & Felix, 2009). The LRR domain of VvFLS2 consists of 28 tandem repeats, similar to AtFLS2 and LeFLS2, which is one repeat more than OsFLS2. The VvFLS2 protein sequence exhibits 72% similarity with AtFLS2, 77% with LeFLS2 and 66% with OsFLS2 (Table S1). LRR domains of VvFLS2 and LeFLS2 share 64% amino acid identity, compared with 56% sequence identity found between VvFLS2 and AtFLS2 LRR domains (Table S1). As LeFLS2 and OsFLS2 showed the highest homology to AtFLS2 and were identified as functional flagellin receptors in their respective species (Robatzek et al., 2007; Takai et al., 2008), VvFLS2 was a promising candidate to function as a flagellin receptor in grapevine.

VvFLS2 functionally complements the Arabidopsis fls2 mutant and is localized at the plasma membrane

To investigate whether VvFLS2 is the true ortholog of AtFLS2, the functional complementation of the Arabidopsis fls2 mutant (Zipfel et al., 2004) was undertaken. The full-length VvFLS2 cDNA was cloned into the binary expression vector pK7FWG2, which was used to obtain stable Arabidopsis transgenic lines expressing p35S::VvFLS2-GFP. Expression of VvFLS2 in the fls2 mutant restored ROS production after flg22 treatment in 17 of the 24 independent kanamycin-resistant transgenic T1 lines tested (Fig. 3a). This ROS production was correlated with FLS2-GFP accumulation detected by Western blotting using an anti-GFP antibody (Fig. 3a). For further characterization, the stable homozygous T3 lines #3 and #15 carrying a single VvFLS2 transgene were selected. These lines were responsive to flg22 as assayed by measurement of ROS production and seedling growth inhibition triggered by flg22, whereas the fls2 mutant was unresponsive (Fig. 3b,c).

Figure 3.

VvFLS2 complements the Arabidopsis fls2 mutant and is localized at the plasma membrane. (a) Correlation between H2O2 production after flg22 treatment (100 nM) and VvFLS2-GFP protein amount detected by α-green fluorescent protein (α-GFP) immunoblot in different T1 lines fls2/p35S::VvFLS2-GFP. Equal loading was checked by Coomassie Brilliant Blue (CBB) staining. (b) H2O2 production after flg22 treatment (1 μM) in leaf disks of homozygous single copy T3 complemented lines #3 and #15. Data represent means ± SD in at least 16 leaf disks from eight plants of two independent experiments. For (a) and (b), reactive oxygen species (ROS) production was measured using the chemiluminescence of luminol and photon counts were expressed as relative luminescence units (RLU). (c) flg22-induced growth inhibition is restored in the complemented mutant fls2/p35S::VvFLS2-GFP (lines #3 and #15), when compared with Col-0 (wild-type, WT) and the fls2 mutant. Data represent means ± SD (= 15). Asterisks indicate statistically significant difference between control and flg22 treatment (t-test; *, < 0.05). Similar results were obtained in three independent experiments. (d) Subcellular localization of VvFLS2-GFP visualized by confocal microscopy in leaves of Arabidopsis fls2 mutant transformed with p35S::VvFLS2-GFP (line #3). DIC, differential interference contrast. Bars, 20 μm.

In addition, in agreement with the presence of a predicted signal peptide and a transmembrane domain (Fig. 2c), VvFLS2-GFP was localized to the cell periphery on confocal microscopy analysis of leaves from fls2/p35S::VvFLS2-GFP #3 plants (Fig. 3d). The green fluorescence of VvFLS2-GFP followed the plasma membrane, shrinking during plasmolysis triggered by 1 M NaCl (Fig. S2a), and co-localized with the red fluorescence of the plasma membrane probe FM4-64 (Fig. S2b), demonstrating that VvFLS2 is localized at the plasma membrane. Together, these results show that VvFLS2 is a functional flg22 receptor capable of complementing the loss of FLS2 in Arabidopsis.

flg22 from B. phytofirmans is a weak elicitor in grapevine

Burkholderia phytofirmans is a PGPR well adapted to grapevine, and promotes a very marked plant growth (Ait Barka et al., 2000; Compant et al., 2005b; Lo Piccolo et al., 2010). Compared with P. syringae pv pisi, the perception of B. phytofirmans triggers weak defense responses in grapevine (Bordiec et al., 2011), whereas a marked PR1 gene expression was observed in Arabidopsis pPR1::GUS seedlings (Methods S1; Fig. S3a). In grapevine, the elicitation of two defense genes by a boiled crude extract from B. phytofirmans was greatly affected by proteinase K treatment (Fig. S3b). Moreover, purified flagellin from B. phytofirmans was sufficient to elicit Arabidopsis PR1 gene expression (Fig. S3c,d). All of these results suggest that flagellin might be an active MAMP of B. phytofirmans.

To investigate whether and how grapevine perceives flagellin from its associated PGPR, the eliciting activity of the flg22 peptide, based on the flagellin sequence from B. phytofirmans strain PsJN (Bp flg22), was tested in grapevine cells and compared with flg22 from P. aeruginosa strain PAK (Pa flg22), X. campestris pv campestris strain 305 (Xc flg22) and A. tumefaciens strain C58C1 (At flg22). Pseudomonas aeruginosa PAK and X. campestris 305 have been described previously as plant pathogenic bacteria in lettuce and Arabidopsis, respectively (Rahme et al., 1997; Sun et al., 2006). Compared with the classical Pa flg22 sequence, Xc flg22 and At flg22 have four and 12 amino acid substitutions, respectively (Fig. 4a). Interestingly, the Bp flg22 epitope possesses six amino acid substitutions compared with the Pa flg22 sequence, but only three compared with Xc flg22 (underlined T1Q, K7L and A13K; Fig. 4a).

Figure 4.

flg22 from the plant growth-promoting rhizobacterium (PGPR) Burkholderia phytofirmans triggers weaker defenses in grapevine than flg22 from Pseudomonas aeruginosa or Xanthomonas campestris. (a) Alignment of flg22 from bacteria P. aeruginosa (Pa), X. campestris (Xc), B. phytofirmans (Bp) and Agrobacterium tumefaciens (At). Arrows indicate key amino acids for flg22-eliciting activity in tomato cells (Felix et al., 1999). Underlined amino acids differ between Bp flg22 and Xc flg22. (b) Dose–response of flg22-induced H2O2 production in grapevine cells 15 min post treatment, measured by luminol chemiluminescence. Data are expressed as a percentage of the response induced by 1 μM Pa flg22 and graphics show means ± SD of three independent experiments. (c) flg22-induced defense gene expression in grapevine cells at 6 h post-treatment (hpt). The expression of defense genes encoding a β-1,3-glucanase (Gluc), a PR1 protein (PR1.2), a proteinase inhibitor (PR6) and an acidic chitinase (Chit4c) was evaluated by quantitative polymerase chain reaction (qPCR). (d) Kinetics of Chit4c gene induction triggered by the different flg22 epitopes. For (c) and (d), means of triplicate data were normalized by housekeeping genes EF1a and 60SRP and compared with water-treated control, set as 1. Defense gene induction was quantified in grapevine cells after treatment by 1 μM of Bp flg22, Pa flg22, Xc flg22 or At flg22.

Dose–response oxidative burst assays revealed that Bp flg22 triggered the production of H2O2 in grapevine, but to a lesser extent than Pa flg22 or Xc flg22 (Fig. 4b). The determination of the half-maximal response (EC50) revealed that Xc flg22 was the most active epitope with EC50 ~ 80 nM, compared with EC50 ~ 300 nM for Pa flg22. The low activity of Bp flg22 is illustrated by an EC50 estimated at c. 8 μM if higher concentrations reached the same plateau. Indeed, at the maximal concentration tested (10 μM), Bp flg22 was still less active than either Pa flg22 or Xc flg22 at 500 nM. Finally, At flg22 seems to be inactive in grapevine as no H2O2 production could be detected even at a concentration of 10 μM.

The expression of typical grapevine defense marker genes (Aziz et al., 2003; Bordiec et al., 2011) was followed at 1, 6, 9 and 24 hpt using 1 μM of each flg22 peptide. On the whole, maximal inductions were observed at 6 hpt. At this time point, Pa flg22 and Xc flg22 induced a strong accumulation of the four defense gene transcripts encoding a β-1,3-glucanase (Gluc), PR1.2, a proteinase inhibitor (PR6) and an acidic chitinase (Chit4c; Fig. 4c). By contrast, Bp flg22 induced only a weak expression of Gluc, PR1.2 and PR6 (Figs 4c, S4a), whereas an intermediate up-regulation of Chit4c was detected (Fig. 4c). However, the Bp flg22-triggered Chit4c gene expression was very transient compared with the long-lasting effect of Xc flg22 and Pa flg22 treatments (Fig. 4d). In addition, the 17.3 gene, which is an SA marker in grapevine (Bordiec et al., 2011), was strongly induced at 1 hpt by Xc flg22 and Pa flg22, but not by Bp flg22 (Fig. S4b). Treatment with At flg22 was totally inactive in defense gene expression (Figs 4c,d, S4a,b). Our results show that Bp flg22 elicits only weak defense responses in grapevine.

AtFLS2 and VvFLS2 have different recognition specificities

Burkholderia phytofirmans is able to colonize Arabidopsis and to stimulate its growth in laboratory conditions (Poupin et al., 2013; Zuniga et al., 2013). However, both effects are less pronounced than in grapevine (Compant et al., 2005b; Zuniga et al., 2013), suggesting potential differences in the perception of this bacterium between grapevine and Arabidopsis. Previous studies have reported different perception specificities between FLS2 from tomato and Arabidopsis (Felix et al., 1999; Bauer et al., 2001; Sun et al., 2006; Robatzek et al., 2007; Mueller et al., 2012). We therefore characterized the eliciting activity of Bp flg22 in Arabidopsis. In Arabidopsis cells, Bp flg22 triggered an oxidative burst comparable with that triggered by Pa flg22 and Xc flg22 (Fig. 5a, left). As a significant correlation has been observed between flg22-eliciting activity and seedling growth inhibition (Vetter et al., 2012), we carried out seedling growth inhibition assays in Arabidopsis. The seedling growth inhibition induced by the three active flg22 epitopes was not observed in the fls2 mutant, indicating that their perception was strictly FLS2 dependent (Fig. 3c and data not shown). On WT Col-0, we showed that the level of reduction in seedling weight after treatment with Bp flg22 was comparable with the growth inhibition caused by Pa flg22 and Xc flg22 (Fig. 5b, left). The growth inhibition activity of flg22 peptides correlated with their ability to induce comparable PR1 expression, as revealed using pPR1::GUS-expressing plants (Fig. S5a). In addition, similar levels of expression of the immune genes CYP71A12, MYB51, WRKY11 and AT5G25260 (Millet et al., 2010) were monitored in Arabidopsis roots (Fig. S5b). As published previously (Felix et al., 1999; Bauer et al., 2001), At flg22 did not elicit any biological response (Figs 5a,b, S5a).

Figure 5.

Species-specific differences in flg22 perception between Arabidopsis and grapevine revealed by reactive oxygen species (ROS) production and plant growth inhibition. (a) ROS production was measured at 15 min by chemiluminescence of luminol on Arabidopsis thaliana or Vitis vinifera cell suspension cultures after treatment with Bp flg22, Pa flg22, Xc flg22 and At flg22 peptides or water control. Data are means ± SD of three independent experiments and are expressed as a percentage of the response induced by 1 μM Pa flg22. (b) Arabidopsis and grapevine growth inhibition triggered by 1 μM Bp flg22, Pa flg22, Xc flg22 and At flg22 and compared with water control (= 30). Different letters indicate significant differences (one-way ANOVA test followed by a Tukey–Kramer test; < 0.05). Similar results were obtained in at least three independent experiments.

In contrast with its strong eliciting activity in Arabidopsis, we found that Bp flg22 is a weak elicitor in grapevine. In addition to activating only a weak oxidative burst (Fig. 5a, right) and defense gene induction (Fig. 4c,d), Bp flg22 did not significantly inhibit grapevine plantlet growth, in contrast with observations on Xc flg22 treatment (Figs 5b, right, S4c). Indeed, Xc flg22 was highly active in grapevine, inducing strongly both defense gene expression and growth inhibition (Figs 4c,d, 5b, right). In addition, grapevine plants challenged with Xc flg22 displayed a root darkening phenotype which was not observed on treatments with other flg22 peptides (Fig. S4c,d) or in Arabidopsis (data not shown).

Given the polymorphism existing between AtFLS2 and VvFLS2 (Table S1), we tested whether FLS2 was responsible for the observed species-specific differences in flg22 perception using growth inhibition assays on fls2/p35S::VvFLS2 Arabidopsis seedlings. Although, in WT Col-0, all three flg22 epitopes exhibited similar biological activity (Fig. 6a; Table S2), Xc flg22-challenged fls2/p35S::VvFLS2 plants were consistently and significantly smaller than Bp flg22-challenged plants (Fig. 6b; Table S2). Therefore, the expression of VvFLS2 into the fls2 background conferred differential flg22 responses characteristic for grapevine (compare Fig. 6b and Fig. 5b, right). These results suggest that VvFLS2 has evolved to distinguish flagellin originating from the grapevine-associated PGPR B. phytofirmans.

Figure 6.

FLAGELLIN SENSING2 (FLS2) is responsible for recognition specificities between grapevine and Arabidopsis. Arabidopsis growth on 1 μM Bp flg22, Pa flg22 and Xc flg22 in wild-type (WT) Col-0 plants (a) or in the complemented mutant fls2/p35S::VvFLS2 line #3 (b). Data represent means ± SD (= 30). Different letters indicate significant differences and ns shows non-significant differences (one-way ANOVA test followed by a Tukey–Kramer test; < 0.05). Similar results were obtained in at least three independent experiments.

Burkholderia phytofirmans overcomes Xc flg22-induced MTI to colonize grapevine plants

We hypothesized that B. phytofirmans flagellin partially evades strong recognition to enable successful plant colonization. As a corollary, we investigated whether the activation of MTI by fully active flagellin-derived peptide would reduce PGPR colonization. Roots of grapevine plantlets grown in vitro were exposed to 1 μM Xc flg22, which displayed the strongest eliciting activity in grapevine (Figs 4c,d, 5a,b), for either 1 min (co-treatment) or 24 h before the inoculation with living B. phytofirmans bacteria (Methods S2). Surprisingly, we observed that treatment with Xc flg22 did not affect the colonization of grapevine leaves or roots (Fig. S6). These results suggest that B. phytofirmans may overcome Xc flg22-induced MTI to colonize grapevine plants.


The FLS2–flg22 perception system is conserved in grapevine and triggers a typical MTI

The flagellin-derived flg22 peptide spans the highly conserved N-terminal region and is an active elicitor in many plant species. Immune responses elicited by flg22 of P. aeruginosa have been characterized in detail in multiple studies, including Arabidopsis and tomato (Boller & Felix, 2009). flg22 triggers Ca2+-associated membrane depolarization, pH alkalinization, oxidative burst, MAPK cascade activation, transcriptional reprogramming, ethylene production, callose deposition and seedling growth inhibition (Gomez-Gomez et al., 1999; Asai et al., 2002; Zipfel et al., 2004; Jeworutzki et al., 2010). Not much is known about flagellin perception in grapevine (V. vinifera). Here, we show that flg22 elicits grapevine immune responses, such as [Ca2+]cyt variations, an oxidative burst, phosphorylation of two MAPKs and defense gene expression (Fig. 1), and growth inhibition (Fig. 5b). Our results are in accordance and complement the recent study of Chang & Nick (2012), who reported flg22-triggered medium alkalinization, oxidative burst, induction of several defense genes, but not cell death, in V. vinifera and the North American species V. rupestris. In cell cultures from V. vinifera cv Pinot Noir, the EC50 of apoplastic alkalinization was estimated at 877 nM (Chang & Nick, 2012), whereas our data indicated that V. vinifera cv Gamay cells are around three times more sensitive to flg22 (EC50 ~ 300 nM; Fig. 4b). Moreover, V. rupestris, a species considered as naturally more resistant to different diseases, is more sensitive to flg22 perception with an estimated EC50 of c. 5 nM (Chang & Nick, 2012). These differences may be caused by different species or cultivar responsiveness in the Vitaceae family, as observed previously in Brassicaceae, where variation in flg22 perception mostly results from changes in FLS2 protein abundance (Vetter et al., 2012).

We have also shown that flg22 induces resistance against the fungal pathogen B. cinerea. In grapevine, flg22 perception triggers the expression of some PR genes (Figs 1, 4, S4a,b) and the activation of the phytoalexin pathway leading to stilbene biosynthesis (Chang & Nick, 2012), two mechanisms known to delay B. cinerea spreading (Coutos-Thevenot et al., 2001; Aziz et al., 2007). Our data are in agreement with previous work showing that flg22 induces resistance to B. cinerea also in Arabidopsis (Ferrari et al., 2007), and support the concept that MTI confers broad-spectrum disease resistance regardless of the origin of the MAMP perceived. Thus, our results indicate that flg22-triggered immune responses are shared between V. vinifera and A. thaliana.

Broad varieties of plant species are highly sensitive to flg22 and carry a functional FLS2 receptor in their genomes (Boller & Felix, 2009). The successful complementation of the Arabidopsis mutant fls2 with the closest grapevine FLS2 ortholog, VvFLS2, demonstrates its function as a grapevine flagellin receptor. The signaling pathways downstream of the flg22–FLS2 perception system are highly conserved between species, as demonstrated by heterologous expression of VvFLS2, OsFLS2 or LeFLS2 in Arabidopsis (our results; Takai et al., 2008; Mueller et al., 2012). In concordance, the overall organization of the FLS2 gene is conserved between the four functionally characterized flagellin receptors, with a unique intron in the 3′ end (Fig. 2b). The FLS2 protein structure is also conserved in grapevine, as VvFLS2 contains an ectodomain comprising 28 tandemly arranged LRRs and a typical non-RD kinase intracellular domain (Fig. 2c). The predicted signal peptide and the transmembrane domain of VvFLS2 target the protein to the plasma membrane (Figs 3d, S2), as demonstrated previously in Arabidopsis (Robatzek et al., 2006; Beck et al., 2012). Lastly, the N-terminal cysteine pair (C61/C68) required for normal processing, stability and function of AtFLS2 (Sun et al., 2012), as well as the residues G318, G493, T867, S938, D997, T1040, G1064, T1072 and P1076, identified to affect AtFLS2 function when mutated (Cao et al., 2013; Robatzek & Wirthmueller, 2013), are strictly conserved in VvFLS2. To sum up, our study indicates that the flg22–FLS2 perception system is conserved in V. vinifera, as in most higher plants, thus supporting a concept of an ancient origin of flagellin perception in plants (Boller & Felix, 2009).

Perception of B. phytofirmans-derived flg22 induces weaker defense responses in grapevine than do X. campestris- or P. aeruginosa-derived flg22

We observed that different flg22 peptides derived from different bacteria had distinct eliciting activities in grapevine. Bp flg22 derived from the non-pathogenic endophytic bacterium B. phytofirmans exhibited a weak oxidative burst and defense gene expression compared with the same epitope derived from the plant pathogenic bacteria Paeruginosa and X. campestris. The expression of several defense genes, such as Chit4c, was only transiently induced by Bp flg22, whereas others were not significantly activated (e.g. PR6 and 17.3; Figs 4, S4a,b). Accordingly, Bp flg22 did not trigger a significant growth inhibition of grapevine plantlets (Figs 5, S4c,d). Thus, Bp flg22 is a weak elicitor in grapevine and triggers only partly flg22-responsive events. Indeed, the gene 17.3, which is exclusively regulated by SA in grapevine (Bordiec et al., 2011), was activated by Xc flg22 and Pa flg22, but not by Bp flg22 (Fig. S4b). These results suggest that the SA signaling pathway might not be activated by Bp flg22, but only by the two other epitopes. Moreover, the kinetics of gene induction were very distinct. Although the three epitopes induced Chit4c expression at a similar level at early time points (1 h), the induction of expression of this gene was very transient after Bp flg22 treatment (Fig. 4d). By contrast, X. campestris-derived flg22 displayed a strong eliciting activity in grapevine, as demonstrated by the low EC50 value in oxidative burst assays, strong induction of defense genes and marked growth inhibition.

Key amino acids described as crucial for flg22-eliciting activity (Felix et al., 1999; Bauer et al., 2001; Sun et al., 2006) are unchanged in Bp flg22. However, the three amino acid substitutions between the most active peptide in grapevine Xc flg22 and Bp flg22 are sufficient to strongly increase its EC50 value from c. 80 nM to c. 8 μM, a 100-fold difference in sensitivity (Fig. 4). In tomato, deletion of the first seven N-terminal amino acid residues of flg22 did not strongly affect the biological activity, as the flg15 sequence remained fully active (Felix et al., 1999). However, fls2 protoplasts expressing AtFLS2 are 1000-fold more sensitive to flg22 relative to flg15 (Mueller et al., 2012). Together, these results indicate the importance of the N-terminal part of flg22 for its perception in Arabidopsis and grapevine, but not in tomato. Interestingly, mutation K13A of flg22 has been reported previously to decrease its biological activity to 60% in tomato (Felix et al., 1999), whereas the mutation K13S has a minimal effect on flg22-eliciting activity in Arabidopsis (Sun et al., 2006). It would be interesting to perform substitutions of these distinct amino acids in Bp flg22 in order to identify their role in FLS2 perception, as performed previously for pathovar variants of Xc flg22 (Sun et al., 2006).

Certain pathogenic or symbiotic bacteria, such as Ralstonia solanacearum, A. tumefaciens, Azoarcus sp. and Mesorhizobium loti, have specific flg22 sequences that are not recognized by FLS2 (Felix et al., 1999; Pfund et al., 2004; Buschart et al., 2012; Lopez-Gomez et al., 2012). Other bacteria are able to reduce or increase their flagellum content depending on the stage of colonization (Achouak et al., 2004; Bardoel et al., 2011, 2012). Another evasion strategy is flagellin glycosylation that masks its perception (Taguchi et al., 2009; Hirai et al., 2011). Weak recognition of their MAMPs, such as flagellin, or even their loss can facilitate host tissue colonization by plant-associated bacteria. Our data suggest that alterations in the Bp flg22 sequence might be a successful adaptation of B. phytofirmans to avoid recognition by the host VvFLS2.

AtFLS2 and VvFLS2 have different recognition specificities

Our results also show that V. vinifera and A. thaliana did not perceive similarly the different flg22 epitopes. Thus, differences in MAMP recognition exist between AtFLS2 and VvFLS2. As flg22 binding is mediated by the FLS2 LRR ectodomain (Dunning et al., 2007; Robatzek & Wirthmueller, 2013), it is interesting to note that the LRR domains of AtFLS2 and VvFLS2 share only 56% of amino acid identity (Table S1). Similarly, the LRR ectodomain of AtFLS2 shares only 55% identity with the LRR ectodomain of LeFLS2, which possesses species-specific traits for flg22 recognition (Robatzek et al., 2007; Mueller et al., 2012; Robatzek & Wirthmueller, 2013). Comparing the LRRs of eight FLS2 orthologs, Boller & Felix (2009) identified conserved amino acids of β-strands only in LRR 1 and LRR 22–28. Interestingly, domain swap experiments between AtFLS2 and LeFLS2 narrowed down the potential Pa flg22 binding domain to LRRs 7–10 for the RINSAKDD core sequence (Mueller et al., 2012; Robatzek & Wirthmueller, 2013). Mutational scanning of LRR domains has also indicated that LRRs 9–15 play an important role for FLS2 function (Dunning et al., 2007). However, no clear conservation has been found between different plant species in these crucial LRRs 7–15 (Boller & Felix, 2009), potentially explaining the different sensitivities of A. thaliana and V. vinifera towards flg22 treatment. Although, in wild-type Arabidopsis, all three tested peptides induced immune responses of similar intensity (Figs 5, S5), the fls2 mutant expressing VvFLS2 gained a flagellin responsiveness profile specific to grapevine (Fig. 6). These results clearly suggest that the differences observed between Arabidopsis and grapevine are caused, at least in part, by the different FLS2 proteins. Future work should reveal which polymorphisms underlie the different perception specificities of AtFLS2 and VvFLS2.

Burkholderia phytofirmans overcomes MTI in Arabidopsis and grapevine to colonize plants

The eliciting activity of B. phytofirmans was mainly conserved in the boiled extract (Fig. S3) and proteinase K treatment greatly affected the eliciting activity, indicating that it is mostly proteinaceous compounds that are responsible for the elicitation. Moreover, the purified flagellins from B. phytofirmans and Bp flg22 are strongly active in Arabidopsis. Thus, flagellin seems to be a main MAMP of Bphytofirmans, even if we cannot exclude the possibility that other elicitors, such as LPS, might participate in the elicitation process (Erbs & Newman, 2012). The eliciting properties of flagellins from endophytic bacteria have been studied in a few plant systems. For instance, flagellins from P. putida and P. fluorescens induced different early defense responses in tobacco cells depending on their origin (van Loon et al., 2008). Similarly, boiled extracts from different strains of endophytic PGPRs P. fluorescens and P. putida differentially stimulated H2O2 and phytoalexin production in grapevine cell suspensions (Verhagen et al., 2010). Unfortunately, flagellin sequences from these bacteria are not known, and it is therefore difficult to make a structure–activity correlation.

Burkholderia phytofirmans is a PGPR naturally associated with grapevine (Ait Barka et al., 2000; Lo Piccolo et al., 2010). Although this bacterium is not known to be associated with Arabidopsis in nature, it is able to colonize this plant under laboratory conditions (Poupin et al., 2013; Zuniga et al., 2013). Our data suggest that, although Bp flg22 has weak elicitor activity in grapevine, it is strongly active in Arabidopsis. These results suggest that VvFLS2 and/or flagellin from B. phytofirmans may have undergone evolutionary changes allowing the adapted endophytic bacterium to colonize its natural host plants without inducing a strong MTI. However, the addition of the strongly eliciting Xc flg22 during the first stages of bacterialization did not interfere with the colonization process in grapevine (Fig. S6). Moreover, in Arabidopsis, Bp flg22 triggers a strong growth inhibition which contrasts with the described PGPR effect (Poupin et al., 2013; Zuniga et al., 2013). On the basis of these data, it seems that B. phytofirmans may ultimately neutralize plant immunity induced by flg22 or other MAMPs using a strong evasion process that could be related to ETS (Jones & Dangl, 2006) to successfully colonize plants. Therefore, the bacterium may inhibit MTI by injecting effectors. Interestingly, no potential secreted effectors have been identified in its sequenced genome (Sessitsch et al., 2005; Weilharter et al., 2011; Mitter et al., 2013). Moreover, although B. phytofirmans possesses all relevant type 3 secretion system (T3SS) genes, the gene encoding the needle-forming protein is absent, suggesting that this T3SS apparatus is not functional (Mitter et al., 2013). Furthermore, a cell culture filtrate of B. phytofirmans did not suppress flg22-induced Arabidopsis defense responses (S. Dorey, unpublished data), in contrast with that shown previously for P. fluorescens and Bacillus subtilis (Millet et al., 2010; Lakshmanan et al., 2012). Another hypothesis is that bacteria might regulate MAMP responses by lowering ethylene production. Indeed, B. phytofirmans is known to reduce the level of ethylene in plants via its 1-aminocyclopropane-1-carboxylate deaminase activity (Onofre-Lemus et al., 2009; Sun et al., 2009), and endogenous ethylene is known to control FLS2 expression (Boutrot et al., 2010; Tintor et al., 2013). Further experiments will be needed to investigate the mechanisms underlying the multilayered evasion of plant immunity by the PGPR B. phytofirmans.


We thank Jean-Pierre Métraux for providing us with the Arabidopsis Col-0 cell suspension and Fabienne Baillieul for fruitful discussions. We also acknowledge Agnès Klinguer, Annick Chiltz and Sandra Villaume for excellent technical assistance. This work has benefited from the facilities of the Centre de Microscopie INRA Dijon/Université de Bourgogne, Plateforme DImaCell and the expertise of Christine Arnould and Elodie Noirot (INRA, UMR1347 Agroécologie, Plateforme DImaCell, Centre de Microscopie INRA/Université de Bourgogne, BP 86510, F-21000 Dijon, France). This work was supported by Agence Nationale de la Recherche and Comité National des Interprofessions des Vins d'appelation d'origine (Génoplante ‘Safegrape’ project 08-GENO-148G), by the Conseil Régional de Bourgogne and Bureau Interprofessionnel des Vins de Bourgogne (PARI AGRALE12, Lucie Trdá and Jani Kelloniemi), by funds from the Region Champagne-Ardenne and the INTERREG IV program France-Wallonie-Vlaanderen (Phytobio project, F.O.), by The Gatsby Charitable Foundation and by the Biotechnology and Biological Sciences Research Council (BBSRC) grant BB/G024936/1 (‘ERA-PG PRR CROP’) to C.Z.