Plants have a sensitive perception system for the most conserved domain of bacterial flagellin


  • The Plant Journal (1999) 18(3), 265–276

*For correspondence (fax +41 61697 4527; e-mail


The flagellum is an important virulence factor for bacteria pathogenic to animals and plants. Here we demonstrate that plants have a highly sensitive chemoperception system for eubacterial flagellins, specifically targeted to the most highly conserved domain within its N terminus. Synthetic peptides comprising 15–22 amino acids of this domain acted as elicitors of defence responses at sub-nanomolar concentrations in cells of tomato and several other plant species. Peptides comprising only the central 8 to 11 amino acids of the active domain had no elicitor activity but acted as specific, competitive inhibitors in tomato cells. These antagonists suppressed the plant’s response to flagellin, crude bacterial extracts and living bacterial cells. Thus, plants have a highly sensitive and selective perception system for the flagellin of motile eubacteria.


Plants do not have an immune system analogous to vertebrates, but they nevertheless have the capacity to detect invasion by microbial pathogens and to respond with a broad set of defence responses ( Somssich & Hahlbrock 1998). Molecules released or generated during microbial entry, so-called elicitors, are thought to act as the chemical cues that are perceived by the plant and activate defence, including early responses such as the oxidative burst ( Lamb & Dixon 1997). Support for this simple model comes mainly from work with fungi, which have proven to be a rich source of various elicitors ( Boller 1995; Ebel & Cosio 1994). Among these elicitors are surface molecules characteristic for fungi in general such as glucans, chitin and chitosan oligosaccharides, as well as various glycopeptides and glycoproteins, and ergosterol ( Boller 1995).

The classic paradigm of these so-called ‘general elicitors’ from fungi is the branched ‘heptaglucoside’, a β-1,3 branched β-1,6-glucan initially purified from the cell walls of Phytophthora megasperma (reviewed by Ebel & Cosio 1994). This elicitor acts as an inducer of defence responses at the nanomolar level in soybean, and a high-affinity binding site with all the hallmarks of a receptor for this glycan structure has been cloned ( Umemoto et al. 1997 ). Elicitors of the heptaglucoside type are present in both virulent and avirulent races of Phytophthora megasperma, and these elicitors exhibit activity in both resistant and susceptible cultivars of soybean. Despite many years of work, it has not been possible in this classic case to establish a link between elicitor perception, induction of defence responses, and susceptibility or resistance ( Ebel & Cosio 1994). Nevertheless, perception of heptaglucoside and a variety of other general elicitors occur with high selectivity and sensitivity, thus suggesting that these sensory systems for non-self structures could signal the presence of potential pathogens. Sensing the presence of pathogens is also important for activation of defence in non-host resistance, i.e. in the resistance of a given plant species against microbes that do not normally colonise it ( Boller 1995).

Host resistance, i.e. the resistance of a given plant species against certain races of pathogens that can normally colonise it, is based on a set of highly specialised recognition systems for molecules produced only by certain strains of one given pathogen. Recent advances in the cloning of fungal avirulence genes ( De Wit 1997) and the corresponding resistance genes ( Hammond-Kosack & Jones 1997; Parniske et al. 1997 ) support this classic concept of gene-for-gene interaction. Avirulence gene products appear to be strain-specific polypeptides produced and secreted by the pathogen. They act as specific elicitors that are recognized, directly or indirectly, by the products of the resistance genes. These proteins, encoded by multigene families, show characteristics of plasma membrane-associated receptors with leucine-rich repeats exposed to the extracellular space ( Parniske et al. 1997 ).

Host resistance with characteristic gene-for-gene interaction also occurs with bacterial pathogens. However, rather than acting as extracellular elicitors, avirulence gene products appear to be injected into the plant cytoplasm through the action of a type-III secretion system of the pathogens ( Lindgren 1997). Consequently, the products of the plant resistance genes are intracellular proteins interacting with the bacterial avirulence gene products ( Bonas & Vandenackerveken 1997). The only well-known extracellular elicitors from bacteria are the harpins, heat-stable proteins isolated from the fire blight pathogen Erwinia amylovora ( Wei et al. 1992 ) and from the bean pathogen Pseudomonas syringae pv syringae ( He et al. 1993 ). These two harpins are both members of the hrp gene cluster encoding elements of the type-III secretion system. Except for a short stretch of 22 amino acids that appears not to be involved in its elicitor activity, they share no sequence homology ( He et al. 1993 ; Wei et al. 1992 ). Plants also respond with defence responses to hrp mutants and to bacteria that lack a type-III secretion system, suggesting that they possess additional sensory systems to perceive potential bacterial pathogens ( Lindgren 1997). Plant mutants with a generally enhanced susceptibility for infection by bacteria ( Glazebrook et al. 1997 ) indicate the existence of such additional perception systems. So far, no general bacterial elicitors that may serve as chemical cues for these perception systems are known.

In an attempt to characterise a harpin-like activity from Pseudomonas syringae pv tabaci, we instead discovered the first general elicitor produced by eubacteria, i.e. flagellin. Here we report that plant cells have a highly specific and sensitive perception system for a domain in the N-terminal part of flagellin that is highly conserved in a wide range of eubacteria.


The elicitor from boiled preparations of Pseudomonas syringae pv tabaci that induces medium alkalinization in tomato cells is flagellin

Phytopathogenic bacteria are known to elicit a rapid K+ efflux and a concomitant medium alkalinization as well as an oxidative burst when added to suspension-cultured tobacco cells ( Atkinson et al. 1993 ; Keppler et al. 1989 ). The same responses have been observed after treatment with cell-free preparations of harpin from Erwinia amylovora ( Baker et al. 1993 ). We used medium alkalinization as a readily measurable symptom to test the effects of Pseudomonas syringae pv tabaci, the cause of fire blight in tobacco, on cells of the non-host species tomato. The response to freshly harvested living bacteria was variable and occurred only at relatively high doses ( Fig. 1a). In contrast, bacterial preparations after heat denaturation by boiling reproducibly caused a strong alkalinization response. Alkalinization-inducing activity in the boiled preparation was associated with the soluble fraction, and the response was saturated at doses ≥ 0.3 μl ml–1. The alkalinization of the medium in response to the bacterial extracts started after a short lag phase of 2–3 min ( Fig. 1a), was paralleled by an efflux of K+, and could be completely blocked by the protein kinase inhibitor K-252a (data not shown).

Figure 1.

Alkalinization of the culture medium in response to treatment of tomato cells with preparations of Pseudomonas syringae pv tabaci.

(a) Extracellular pH of tomato cells treated with a suspension of living bacteria (living) or with different doses of this bacterial suspension after heat treatment for 5 min at 95°C (boiled). The addition of 10 μl bacterial supension per ml corresponds to the addition of approximately 107 cfu per ml of suspension containing approximately 2 106 plant cells per ml.

(b) Separation of the factor inducing medium alkalinization by anion exchange chromatography and SDS-PAGE. Coomassie-blue stained SDS gel with bacterial extract (1), and elicitor-active fraction obtained after separation on anion exchange chromatography (2). Lanes (1) and (2) were cut in slices of 2 mm, eluted and eluates assayed for induction of alkalinization in tomato cells.

The activity of heat-treated crude preparations was sensitive to treatment with trypsin, chymotrypsin and proteinase K (data not shown), indicating a proteinaceous nature of the elicitor activity. Because elicitor activity was not affected by boiling in 1% SDS, we separated the crude extract by SDS-PAGE in order to estimate the molecular weight of the elicitor ( Fig. 1b). The dried Coomassie-blue stained gel was cut into 2 mm segments and tested for induction of the alkalinization response. There was a single peak of activity that co-migrated with a major polypeptide band with an apparent molecular mass of 32 kDa. On anion-exchange chromatography this polypeptide co-purified with the alkalinization-inducing activity ( Fig. 1b). This band was excised from the gel and subjected to N-terminal sequencing. The resulting sequence of the first 20 amino acids had no similarity to harpins but shared strong similarity with the N-terminal sequences known for flagellin from Pseudomonas species like P. aeruginosa and P. putida ( Fig. 2).

Figure 2.

Alignment of N-terminal sequences of eubacterial flagellin sequences.

Schematic representation of flagellin gene structure with conserved N- and C-terminal sequences and a variable middle part (top). Alignment of N-terminal sequences with consensus sequence (bottom) and the N-terminal sequence derived from the 33 kDa protein purified from Pseudomonas syringae pv tabaci.

This sequence information demonstrated the presence of flagellin but did not rule out the presence of another polypeptide that could be responsible for the elicitor activity in the band at 32 kDa. To test whether flagellin was indeed the alkalinization-inducing component, we followed an established, independent protocol for the purification of flagellin ( Totten & Lory 1990). Flagella were sheared off the bacteria by treatment in a blender and purified by differential centrifugation under de-polymerising (pH 2.0) and re-polymerising (pH > 7.0) conditions. As shown by SDS-PAGE in Fig. 3(a), the preparation obtained by this procedure was essentially pure flagellin. The purified polypeptide contained the expected N terminus and, most importantly, was a potent elicitor of the alkalinization response ( Fig. 3b). Half-maximal alkalinization was induced at a concentration of approximately 0.5 n m (EC50). Heat-denaturation by boiling of this flagellin preparation resulted in further increase of activity and an EC50 of approximately 0.1 n m (data not shown). These results demonstrate that the plant cells have a highly sensitive perception system for bacterial flagellin.

Figure 3.

Purification and alkalinization inducing activity of flagellin.

(a) SDS-PAGE of crude flagella preparation from P. syringae pv tabaci (1), flagella sedimented by high-speed centrifugation (2), and flagellin after purification (3).

(b) Alkalinization inducing activity of purified flagellin in tomato cell suspension.

Tomato cells have a sensitive and selective chemoperception system for a highly conserved peptide domain of flagellin

The purified flagellin was cleaved using cyanogen bromide, and the resulting fragments were assayed for induction of the alkalinization response. Activity was found only in the N-terminal fragment with a molecular mass of < 10 kDa (data not shown). Amino acid sequences deduced from flagellin genes of various eubacteria show homology in the N-terminal and C-terminal domains of flagellin while the middle part is highly variable in size and sequence ( Wilson & Beveridge 1993; scheme on top of Fig. 2). The most conserved domain is a stretch of amino acids close to the N terminus of the flagellins. We speculated that the plant might have a perception system for this conserved domain as a common determinant of bacteria. Based on the sequence of P. aeruginosa, a 22 amino acid peptide spanning the conserved domain was synthesised ( Fig. 2, underlined sequence). This peptide, named flg22, proved to be an extremely potent elicitor of the alkalinization response ( Fig. 4a). It was active at a threshold of about 1 p m, and the concentration required for halfmaximal activity (EC50) was 30 p m.

Figure 4.

Activity of peptides spanning the conserved N-terminal domain of flagellin.

(a,b) Dose-response curves for alkalinization induced by peptides covering 22 amino acid residues of the conserved region (flg22), or peptides spanning only part of this domain as indicated by the legend in the middle.

To further delineate the peptide determinant perceived by the plant cells, peptides lacking varying numbers of amino acid residues from the N-terminal or C-terminal end of the domain spanned by flg22 were synthesised and assayed for activity ( Fig. 4). The removal of the N-terminal glutamyl residue (flg21) did not affect the activity ( Fig. 4a). Omitting the arginyl residue in position two or up to five more residues resulted in peptides (flg20–flg15) with slightly reduced activity and EC50 values three to five times higher than the EC50 of flg22. Further trimming at the N-terminal end (flg14–flg12) strongly reduced activity, and EC50 values increased by more than one order of magnitude for every residue removed. However, even flg13 still induced a full alkalinization response when given at high enough concentrations (EC50 of 300 n m, Fig. 4a).

Using flg15 as the smallest, highly active peptide, the effect of shortening at the C terminus was studied ( Fig. 4b). Removal of two or three amino acid residues (flg15-Δ2, flg15-Δ3) caused a three- and 10-fold loss in activity, respectively, and the maximal pH increase induced with saturating doses was somewhat smaller. Peptides lacking four or more residues at the C terminus (from flg15-Δ4 to flg15-Δ9) were completely inactive, indicating a strong drop in activity by removal of the Leu residue present at position 12 of flg15.

Several structural analogues of flg15 with replacements of single amino acid residues with alanine were synthesised. None of these replacements abolished activity completely but replacements of the two Asp residues or the Gly residue in the center of flg15 all caused 10- to 100-fold decreases in activity compared to flg15 ( Table 1). In contrast, the replacement of several other amino acids, notably the Leu residue at position 12 of flg15, had no apparent negative effect on activity. Permutation of the Asp-Asp-Ala sequence in the middle of flg15 also led to a more than 100-fold reduction in biological activity ( Table 1).

Table 1.  Alkalinization-inducing activity of flg15 variants and peptides corresponding to the homologous domain of flagellin in different bacterial species
Biological activity
PeptideAmino acid sequenceEC50 (nM)% b
  1. a No significant alkalinization at 10 μm (EC50 >> 10 μm).

  2. b Calculated as ([EC 50 of flg15] × 100/EC50).

flg15A.tum. RVGDASDNAAYWSIAinactive a0.0
flg15R.mel. RVGQAADNAAYWSIAinactive a0.0
flg15E. coliRINSAKDDAAGQAIA0.2030.0

Peptides corresponding to the homologues of flg15 from several other bacteria were synthesised. The flg15 homologue from Escherichia coli, also conserved exactly in the flagellin sequences of Bordetella bronchiseptica and Proteus mirabilis, was also highly active ( Table 1). Two plant-associated bacteria, Agrobacterium and Rhizobium, have flagellins that are exceptionally divergent in the flg15 domain ( Fig. 2). The A. tumefaciens and the R. meliloti flg15 homologues, termed flg15 A.tum. and flg15R.mel., respectively, were completely inactive in the bioassay, even at micromolar concentrations ( Table 1).

The active peptides of flagellin induce an oxidative burst and other elements of the defence response in tomato cells

We devised a plate assay to visualize elements of the defence response in tomato cells. Cells were spread as flat lawns on Petri dishes, and test solutions containing different amounts of flg22 or the fungal elicitor xylanase were applied locally (scheme on top of Fig. 5). In the presence of the pH-indicator chlorophenol red, a colour change towards pink-red indicated alkalinization of the medium. This response became visible within 10 min of treatment as a ring around the site of application that slowly increased with the radial diffusion of the elicitor. The threshold for flg22 to induce a visible alkalinization after 10 min was 2 pg, corresponding to less than 1 fmol of the peptide ( Fig. 5). In the presence of the peroxidase substrate 5-aminosalicylate, dark-brown staining indicated production of active oxygen species. This response became visible after 30 min and continued to develop over time ( Fig. 5). In the absence of an indicator, a brown coloration around the sites of flg22 application became visible after 20 h, indicating that flg22 can elicit a type of necrotic or hypersensitive response in the cultured cells. Xylanase, applied as a positive control for elicitor activity ( Bailey et al. 1992 ; Felix et al. 1993 ), similarly induced all three types of responses in tomato cells.

Figure 5.

Plate assays for induction of alkalinization, oxidative burst and necrotic response.

Aliquots of cell suspensions were placed in Petri dishes. For assaying the alkalinization response, the pH indicator chlorophenol red (10 μm) was added. For assaying the oxidative burst, cells were supplied with the peroxidase substrate 5-aminosalicylic acid (10 μm). Using a narrow tipped pipette most of the medium was removed, leaving a lawn of (wet) cells in the dish. Droplets of 1 μl H2O as a negative control, or H2O containing 10 μg xylanase as a positive control, or different amounts of flg22 were pipetted onto the lawn of cells as indicated. Petri dishes were analyzed optically by a flat-bed scanner at intervals.

Perception of flagellin by cells of other plant species

We examined cell cultures derived from other plants for their response to flagellin and peptides derived from the conserved N-terminal domain. Purified, boiled flagellin induced an alkalinization response in suspension-cultured cells of Lycopersicon peruvianum (a wild relative of tomato), tobacco, potato and Arabidopsis ( Table 2). The peptides flg22 and flg15 also induced the alkalinization response in these cultures. Suspension cultures of rice did not show a detectable response to flagellin, flg22 or flg15, although they reacted with an alkalinization response to chitin oligomers ( Table 2). Flg15-Δ5, flg15 A.tum. or flg15R.mel. did not cause alkalinization in any of these cultures when added at concentrations up to 10 μm (data not shown).

Table 2.  Alkalinization response in cell cultures of different plant species
EC50 for induction of the alkalinization response (nM)
Cell cultureFlagellinflg22flg15Chitin fragments a
  • n.r. No response when treated with concentrations up to 1000 n m of flg22 and flg15, and up to 100 n m of flagellin, respectively.

  • a

    Alkalinization in response to the chitin fragment penta-N-acetyl chitopentaose. A. thaliana cells and rice cells show much higher sensitivity for chitin fragments with a higher degree of polymerization (data not shown).

Lycopersicon peruvianum0.50.050.052
Arabidopsis thaliana30.3301000

Peptides truncated at the C-terminal end act as antagonists of flagellin

Peptides comprising less than 12 amino acids of the N-terminal part of flg15 were inactive as inducers in our bioassays. We tested these peptides for inhibitory effects and observed that the peptides flg15-Δ4 to flg15-Δ7 did suppress responses induced by the active peptides flg15 and flg22. Suppression could be overcome by increasing concentrations of active peptides, indicating that the truncated peptides act as competitive inhibitors. In Fig. 6(a), this suppressive effect on the induction of alkalinization by flg15 is shown for flg15-Δ4 and flg15-Δ7. The EC50 for the flg15-dependent induction of alkalinization increased from 0.08 n m in the absence of truncated peptides to 7 n m in the presence of 10 μm flg15-Δ4 and 28 n m in the presence of 10 μm flg15-Δ7, respectively. Peptides flg15-Δ6 and flg15-Δ5 also showed antagonistic effects with efficiencies similar to the ones of flg15-Δ7 and flg15-Δ4. In contrast, peptides that were further shortened at the C-terminal end, such as flg15-Δ8 and flg15-Δ9, exhibited only a marginal (flg15-Δ8) or no inhibitory effect at all (flg15-Δ9). The antagonistic peptides flg15-Δ4 to flg15-Δ7 did not affect the induction of the alkalinization response by unrelated elicitors such as chitin fragments ( Fig. 6b), ergosterol or xylanase (data not shown). No inhibitory or antagonistic activity was observed for flg15 A.tum. or flg15R.mel. when added at concentrations of up to 10 μm (data not shown). In summary, peptides comprising the N-terminal 8–11 amino acid residues of flg15 show no activity as agonists but act as specific, competitive antagonists.

Figure 6.

The peptides representing the N-terminal part of flg15 act as competitive suppressors of flg15 in tomato cells.

(a) Effects of 10 μm flg15-Δ4 and 10 μm flg15-Δ7 on the alkalinization response induced by different concentrations of flg15.

(b) Effect of 10 μm flg15-Δ7 on alkalinization response induced by different concentrations of the chitin fragment penta-N-acetylchitopentaose (CH5).

Although the truncated peptides flg15-Δ4 to flg15-Δ7 did not exhibit agonist activity in cells of tobacco and A. thaliana either, they had no clear antagonistic effect when applied together with flg22 or flg15 (data not shown), indicating a recognition process slightly different from the one in tomato cells.

Flagellin is a major determinant for the perception of motile bacteria in tomato cells

In the tomato cells, the antagonistic peptides flg15-Δ4 to flg15-Δ7 could be conveniently used to examine to what extent the elicitor activity of intact flagellin or crude bacterial preparations was due to the conserved domain spanned by the flg15 peptides. In the presence of 10 μm flg15-Δ5, the alkalinization induced by 1 n m flagellin purified from P. syringae pv tabaci was blocked ( Fig. 7a), indicating that the only determinant of flagellin that is perceived by the cells is the domain spanned by the flg15 peptide. Similarly, a strong inhibition of the response induced by suspension of living P. syringae pv tabaci or crude, boiled extract derived from this preparation was observed in the presence of flg15-Δ5 ( Fig. 7a). Alkalinization-inducing activity was observed in crude extracts prepared from various bacteria including Erwinia carotovora, Erwinia chrysanthemi, different pathovars of Pseudomonas syringae (pv glycineae, pv tomato, pv syringae, pv phaseolicola), P. aeruginosa, P. fluorescens and E. coli. In contrast, no activity was observed in extracts prepared analogously from Xanthomonas campestris (pv vesicatoria, pv juglandis, pv brassica rapa), A. tumefaciens and R. meliloti. Figure 7(b) shows representative examples for extracts obtained from P. aeruginosa, P. syringae pv tomato and P. syringae pv syringae. The flagellin antagonists suppressed induction by the crude extracts of all the bacteria that contained alkalinization-inducing activity, as exemplified for flg15-Δ5 and extracts of P. aeruginosa, P. syringae pv tomato and P. s. pv syringae ( Fig. 7b). These results show that the elicitor-active epitope is common to many different bacterial species and they suggest that flagellin is the major, if not the only, determinant capable of stimulating rapid responses in our bioassays with tomato cells.

Figure 7.

The peptide flg15-Δ5 antagonizes the induction of rapid medium alkalinization by living bacteria and crude bacterial preparations.

(a) Extracellular pH of tomato cells treated with 10 μl ml–1 of a suspension of living P.s. pv tabaci, with 0.3 μl ml–1 of this bacterial suspension after heat treatment, or with 3 n m of flagellin purified from P.s. pv tabaci. Where indicated, cells were pre-treated for approximately 3 min with 10 μm flg15-Δ5.

(b) Effect of pre-treatment with 10 μm flg15-Δ5 on alkalinization induced by 0.3 μl ml–1 of heat-treated extracts obtained from P.s. pv tomato, P.s. pv syringae or P. aeruginosa, respectively.

Induction of responses in tomato leaf tissues

We also tested flagellin peptides for the induction of rapid responses in tissues of tomato plants. Leaf or stem slices suspended in H2O showed an alkalinization response when treated with flg22 and flg15. However, this pH increase was usually small and variable (< 0.4 pH units), probably due to a small number of accessible cells at the surface of the tissue, and due to a relatively high initial extracellular pH (data not shown). In contrast, these tissues consistently showed rapid, easily measurable production of active oxygen species when treated with active flagellin peptides. Using a luminol-based assay, a rapid oxidative burst was observed after treatment with 10 n m flg15 but not when treated with 10 n m flg15 in combination with 30 μm flg15-Δ5 or with 30 μm flg15-Δ5 alone ( Fig. 8a). No oxidative burst was observed in leaf tissues treated with 10 μm flg15 A.tum. or flg15R.mel. ( Fig. 8b). These results indicate that tomato cells in leaves and cell suspension perceive flagellin with the same specificity and respond to this stimulus with the same responses. Induction of the oxidative burst in plant cells has been observed to be blocked by protein kinase inhibitors ( Schwacke & Hager 1992) and inhibitors of NADPH oxidase ( Levine et al. 1994 ). Similarly, induction of oxidative burst by flagellin peptides was blocked by pre-treatment of tomato leaf tissue with 1 μm of the protein kinase inhibitor K-252a or 10 μm of the NADPH oxidase inhibitor diphenyleneiodonium ( Fig. 8c).

Figure 8.

The flagellin peptide flg15 induces an oxidative burst in tomato leaf tissue.

Luminescence of tomato leaf slices in a solution with luminol and peroxidase after treatment with flagellin-derived peptides. Light emission at the very beginning of the experiments is caused by phosphorescence of the green tissue.

(a) Luminescence after treatment with flg15 and/or flg15-Δ5, as indicated.

(b) Luminescence after treatment with flg15 A.tum. or flg15R.mel..

(c) Luminescence in leaf tissue pre-treated for 5 min with 1 μm of the protein kinase inhibitor K-252a or 10 μm of the NADPH oxidase inhibitor diphenyleneiodonium.

Induction of ethylene biosynthesis is a further response characteristic for plants attacked by pathogens or treated by elicitor preparations ( Boller 1995). Leaf tissues treated with flg15 at concentrations of 10 n m or 10 μm, respectively, showed strongly enhanced production of ethylene compared to mock treated controls ( Table 3). No induction was observed after treatment with 10 μm flg15R.mel. or 10 μm flg15-Δ4 ( Table 3).

Table 3.  Induction of ethylene biosynthesis in tomato leaves
TreatmentEthylene production (pmol g–1) a
  • a

    Ethylene production was measured as the amount of ethylene accumulating during the first 2 h of treatment. Values shown represent mean and standard deviation of five replicates.

Control 24 ± 12
Flg15 (10 n m) 460 ± 80
Flg15 (10 μm) 1140 ± 280
Flg15-Δ4 (10 μm) 12 ± 2
Flg15R.mel. (10 μm) 36 ± 0


Plant cells possess highly sensitive and selective perception systems for a variety of molecules typical for fungal surface structures. Examples for such elicitors originating from lower fungi of the class Oomycetes comprise the classic heptaglucoside ( Ebel & Cosio 1994; Umemoto et al. 1997 ) and the different elicitins and glycoproteins produced by various Phytophthora species ( Baillieul et al. 1995 ; Nürnberger et al. 1994 ; Ricci et al. 1989 ). In our own work, using rapid responses of suspension cultured tomato cells as bioassays, we have previously identified sensitive chemoperception systems for molecules characteristic of fungi in general. In particular, these cells specifically perceive the fungal sterol ergosterol ( Granado et al. 1995 ), the fungal cell wall constituent chitin ( Felix et al. 1993 ), as well as N-linked glycopeptides with fungal-type mannosyl linkages ( Basse et al. 1992 ). There are indications that plants also have perception systems for characteristic bacterial structures: various types of living bacteria as well as preparations of heat-killed bacteria can trigger rapid responses in plant cell cultures and defence responses in intact plant tissues ( Atkinson et al. 1993 ; Jakobek & Lindgren 1993; Keppler et al. 1989 ). In the present work we identified flagellin as a first general bacterial elicitor and demonstrated a highly sensitive and specific chemoperception system for this highly characteristic surface structure of motile bacteria.

The bacterial flagellum consists of a long helical filament and a rotary motor which is anchored in the cell surface ( Schuster & Khan 1994). The filament consists of flagellin subunits, 10–40 000 molecules per flagellum, which self-assemble at the distal tip after export through the hollow core of the flagellum ( Schuster & Khan 1994; Wilson & Beveridge 1993). Export and self-assembly are determined by the conserved N-terminal and C-terminal ends of flagellin while the central part is highly variable in size and sequence and may determine polymorphism of flagella from different bacteria ( Wilson & Beveridge 1993).

It is the most constant and unique domain of flagellin, residing close to the N terminus of flagellins, which acts as a very potent elicitor of responses in our bioassays. The minimum peptide with full activity spans 15 amino acid residues of this domain. Variants of this peptide shortened at N-terminal or C-terminal ends show a strong drop in activity for every amino acid removed. Interestingly, peptides truncated at the C terminus, flg15-Δ4 to flg15-Δ7, show no activity as inducers but act as competitive and specific antagonists of flagellin. This is best explained by competition between the antagonistic peptides and the elicitor-active peptides for the same binding site on the hypothetical flagellin receptor of the tomato cells. Experiments with radiolabeled flagellin peptides indicate the presence of such a binding site with high-affinity for flagellin agonists and somewhat lower affinity for the antagonists in membrane preparations of tomato cells (T. Meindl, G. Felix and T. Boller, in preparation). With respect to receptor activation by flg15 this indicates a mechanism involving binding of the N-terminal part and activation of signalling by the C-terminal part as two functionally separate steps. A similar bipartite mechanism for binding and signalling has also been found for other stimuli in plant cells, in particular for the yeast-derived glycopeptide elicitors ( Basse et al. 1992 ) and the wound hormone systemin ( Meindl et al. 1998 ).

Specific, competitive antagonists are useful tools for distinguishing different qualities of stimuli present in crude preparations. For example, the flagellin peptide flg15-Δ5 antagonised induction by living cells of P. syringae pv tabaci or by boiled preparations of these bacteria. These antagonists also completely blocked activity present in all other elicitor-active bacterial preparations tested. This includes preparations from phytopathogenic bacteria like E. carotovora, E. chrysanthemi, and different pathovars of P. syringae (pv glycineae, pv tomato, pv syringae, pv phaseolicola) but also bacterial preparation from E. coli, P. aeruginosa and P. fluorescence. Thus, the elicitor-active epitope occurs in a wide variety of bacteria and, in all of these preparations, it is the predominant, if not the only, component triggering rapid responses, such as medium alkalinization in the tomato cells.

Interestingly, several preparations from plant-associated genera of bacteria such as Agrobacterium, Rhizobium and Xanthomonas did not induce rapid responses in the tomato cells. While there is no sequence information available for flagellins of Xanthomonas species, sequences of Agrobacterium and Rhizobium exhibit exceptional divergence in the N-terminal conserved domain of flagellin. Peptides synthesised according to these divergent sequences proved completely inactive in our bioassays and are apparently not perceived by the plant cells. This is reminiscent of our previous studies on the perception of ergosterol ( Granado et al. 1995 ). Most higher fungi have membranes rich in ergosterol, and spores from these fungi activate responses in our bioassays with the tomato cells because they release some ergosterol. However, some of the most typical plant-associated fungi, the biotrophic rust and mildew pathogens, contain only modified ergosterol that is inactive in our bioassay. It is tempting to speculate that the ergosterol receptor as well as the flagellin receptor represent elements of a ‘non-self’ perception system of plants ( Boller 1995), and that microbes adapted to grow in or on plants may have been under selection pressure to modify or lose these determinants detected by the chemical sense of plant cells. The structure activity requirements for the elicitor activity of flagellin described in this paper opens up the possibility of testing this hypothesis by introducing elicitor-active flagellin into Agrobacterium or Rhizobium or, vice versa, by replacing elicitor active flagellin in virulent and non-virulent bacteria.

The flagellum is an important virulence factor for bacteria pathogenic to animals and plants ( Finlay & Falkow 1997). Most plant-associated bacteria are motile in the free-living state. Flagella-driven chemotaxis may be important in the early interaction with the host plant, particularly in the rhizosphere ( Vande Broek & Vanderleyden 1995). Motility is also important for foliar pathogens to reach internal sites in the leaves of the host plants ( Beattie & Lindow 1995; Hatterman & Ries 1989). Intact flagella attached to bacteria are not expected to act as the direct stimuli for the perception system described in this paper. Depolymerization or fragmentation of the flagellum is probably required to obtain elicitor-active flagellin subunits or fragments that can diffuse through the plant cell wall and reach a receptor site expected to reside on the plasma membrane. Cytological studies have shown that bacteria usually shed their flagella inside the plant tissue ( Hatterman & Ries 1989). Further mechanisms that could result in elicitor active products may consist of a ‘leaky’ export and assembly process for the flagellum, or to depolymerization and degradation processes occurring within plant tissue (e.g. by protease activities in the apoplast). In our bioassays with suspension cultured cells, approximately 1–10 living bacteria per plant cell were sufficient to trigger significant alkalinization. Release of elicitor activity appeared to be a limiting factor since depolymerization of the flagella by boiling or by treatment at low pH greatly increased the elicitor activity of the bacterial suspension. However, ‘release’ from intact bacteria occurring with slower kinetics would not be easily detectable in our assays which monitor only the first few minutes of the interaction. Nevertheless, based on the sensitivity of the flagellin perception observed in cultured cells (detection limit of approximately 10 p m), one can estimate that a single flagellum, even when disintegrated only partially into its 10–40 000 flagellin subunits, could signal the presence of a bacterium to the adjacent plant cell.

Rapid release of active oxygen species is a characteristic response of plant cells to elicitor preparations or inoculation with microbial pathogens ( Lamb & Dixon 1997; Mehdy et al. 1996 ). Active oxygen species released during this response have been implicated in the regulation of defence-related genes and the induction of the hypersensitivity response ( Levine et al. 1994 ). Furthermore, by oxidative cross-linking of proteins, active oxygen species have been found to render the plant cell wall less digestible by microbial enzymes ( Brisson et al. 1994 ). The oxidative burst in plants is reminiscent of the respiratory burst observed in leukocytes where the production of active oxygen species serves a major role as bactericidal mechanism ( Babior 1995). Leukocytes perceive invading bacteria via receptors for bacterial-derived chemotactic peptides. Common to these peptides is N-formylation of their initial methionine residue, a feature highly characteristic for bacterial protein synthesis. In preliminary experiments we did not observe reactions to these chemotactic peptides in plant cells (data not shown). However, plant cells responded to an equally characteristic bacterial epitope, bacterial flagellin, with an oxidative burst.

Perception of flagellin, like the perception of chitin fragments, appears to be well conserved in many species and was observed in cell cultures derived from A. thaliana and several solanaceous species. Although there was some variation with regard to sensitivity, all cultures responded to flagellin, flg22 and flg15, and they were not elicited by flg15-Δ5, flg15 A.tum. or flg15R.mel.. Among the cell cultures tested, only rice cells showed no response to flagellin and flagellin peptides. It remains to be seen whether this is a constant trait of rice, or even of monocots in general, or whether it is due to the loss of responsiveness that occurred in this particular cell culture line during years of in vitro growth. As described in the accompanying paper, loss of responsiveness appears to have also occurred in one of the several ecotypes of A. thaliana tested ( Gómez-Gómez et al. 1999 ). Seedlings of sensitive A. thaliana ecotypes respond to flagellin-derived peptides with defence-related responses and, surprisingly, with a marked inhibition of growth ( Gómez-Gómez et al. 1999 ). We used these findings to start a genetic approach for elucidating the molecular basis of flagellin perception and for studying the putative role of this chemosensory process for plant defence.

Genetic studies on the side of the pathogens and on the side of the host plants have provided the basis for explaining the molecular mechanism underlying resistance of plants against specific, highly specialized pathogens that follow the classic gene-for-gene relationship. Therefore, the products of plant resistance genes can be regarded as highly selective chemoperception systems directed against the products of avirulence genes, i.e. polypeptides unique for single strains of a pathogen ( Dangl et al. 1995 ; De Wit 1997). Our studies highlight an additional strategy to sense the presence of microbes involving chemoperception systems for microbial surface molecules that are highly characteristic for whole classes of microorganisms. As is the case with the perception systems for general elicitors from fungi, it remains an open question at present as to how the perception of the general bacterial elicitor flagellin is related to disease susceptibility and resistance. The fact that a variety of different plant species possess such a highly sensitive and selective perception system may be taken as an indication for selection pressure to maintain the capability to sense flagellin. Based on the molecular characterization of this perception system described in this report, our ongoing work concentrates on a genetic approach in A. thaliana to investigate the role of flagellin perception in interactions between plants and symbiotic, neutral and pathogenic bacteria.

Experimental procedures


Peptides were synthesised by F. Fischer (Friedrich Miescher-Institute, Basel, Switzerland) or by Bio-Synthesis Inc. (Lewisville, TX, USA). Peptides were dissolved in H2O (stock solutions of 1–10 m m) and diluted in a solution containing 0.1% BSA and 0.1 m NaCl. Penta-N-acetylchitopentaose was obtained from Seikagaku Corp. (Tokyo, Japan) and xylanase from Trichoderma viride was purchased from Fluka (Buchs, Switzerland).

Bacteria and bacterial elicitor preparations

Pseudomonas syringae pv tabaci (strain 511) was provided by M. Wybrecht (Novartis, Basel, Switzerland) and grown in King’s B broth at 26°C on a rotary shaker. Bacteria were harvested by centrifugation, washed once with H2O and resuspended in H2O (10% of original volume). Crude bacterial elicitor was prepared by boiling the suspension for 5–10 min and removing bacterial debris by centrifugation. Other bacterial strains were obtained from Deutsche Sammlung von Mikroorganismen und Zellkulturen (DSM GmbH, Braunschweig, Germany), and were grown and extracted as described above.

Partial purification of the elicitor from Pseudomonas syringae pv tabaci was carried out by anion-exchange chromatography and SDS-PAGE. A DEAE-cellulose (DE-52, Whatman, Maidstone, UK) column was equilibrated with 10 m m Tris–HCl at pH 7.5, loaded with crude elicitor preparation in the same buffer, and eluted with a linear gradient from 0 to 0.5 m NaCl. Elicitor active fractions were separated by SDS-PAGE and stained with Coomassie-blue. The dried gel was cut into 2 mm segments, the segments were eluted overnight in 0.1% SDS, and the eluates were tested for induction of the alkalinization response in tomato cells at a dose of 1 μl eluate per ml of suspension.

Purification of flagellin

Flagella were prepared as described previously ( Totten & Lory 1990). Bacteria were collected by centrifugation (7000 g,10 min), resuspended in 50 m m Na-phosphate (pH 7.0), 10 m m MgCl2, and treated in a Wareing blender for 60 sec to shear off the flagella. Cells and cell debris were removed by low speed centrifugation (60 min at 10 000 g), and flagella were collected by high-speed centrifugation (60 min at 100 000 g). The pellet was resuspended in H2O and the pH was lowered to 2.0 by glycine-HCl to dissociate the flagella. After removal of insoluble material by centrifugation (60 min at 100 000 g), the supernatant containing the dissociated flagellin molecules was adjusted to pH 7.0 with NaOH.

Plant cell cultures

The tomato cell line Msk8 was maintained as a suspension culture and used 4–10 days after subculture for experiments, as described previously ( Felix et al. 1993 ). Cell cultures of rice, line OC ( Baba et al. 1986 ), tobacco ( Felix & Meins, Jr. 1987), potato ( Schweizer et al. 1996 ), Lycopersicon peruvianum (Felix & Boller 1995) and Arabidopsis thaliana ( May & Leaver 1993) were cultured as described elsewhere.

Alkalinization response

To measure alkalinization of the growth medium (the alkalinization response), 2.5 ml aliquots of the cell suspensions were placed in open 20 ml vials on a rotary shaker at 120 cycles per min. The pH in the medium was continuously measured using a small combined glass electrode (Metrohm, Herisau, Switzerland) and registered on a pen recorder.

Plate assays for alkalinization response, oxidative burst and necrotic response

Aliquots of cell suspensions (3–6 ml, depending on the density of the suspension) were placed in Petri dishes (Ø 50 mm). For assaying the alkalinization response, the pH indicator chlorophenol red (10 μm) was added. For assaying the oxidative burst, cells were supplied with the peroxidase substrate 5-aminosalicylic acid (100 μm). Using a narrow tipped pipette most of the medium was removed, leaving a thin 1–2 mm thick layer of (wet) cells in the dish. Test substances in a volume of 1 μl were applied locally onto the lawn of cells, and the Petri dishes were analyzed optically by a flat-bed scanner at intervals.

Oxidative burst and ethylene biosynthesis in tomato leaves

Fully expanded leaves of 2- to 3-month-old tomato plants (L. esculentum cv Moneymaker) grown in the greenhouse were cut into 2 mm slices and floated on H2O overnight. For measuring the oxidative burst, active oxygen species released by tomato leaf tissue were measured by a luminol-dependent assay ( Keppler et al. 1989 ). Slices were transferred to assay tubes (two slices corresponding to approximately 20 mg fresh weight) containing 0.1 ml of H2O supplied with 20 μm luminol and 1 μg horseradish peroxidase (Fluka). Luminescence was measured in a LKB 1250 luminometer (LKB Wallac, Turku, Finland) for 20 min after the addition of the test solution.

For assaying ethylene production, leaf slices (approximately 50 mg fresh per assay) were transferred to 6 ml glass tubes containing 1 ml of an aqueous solution of the peptide being tested. After vacuum infiltration of the leaf slices (5 min, water pump), vials were closed with rubber septa and ethylene accumulating in the free air space was measured by gas chromatography after 2 h of incubation.


We thank Dr Ko Shimamoto (Plantech Research Institute, Yokohama, Japan) for providing the cell culture line of rice; Franz Fischer (Friedrich Miescher-Institute, Basel, Switzerland) for the synthesis of various flagellin peptides; Renate Matthies and Dr Jan Hofsteenge (Friedrich Miescher-Institute) for protein sequencing services; Martin Regenass for maintaining the cell cultures and for technical assistance; and Drs Lourdes Gómez-Gómez and Scott Peck (Friedrich Miescher-Institute) for critical reading of the manuscript.