The bacterial flagellin (FliC) epitopes flg22 and flgII-28 are microbe-associated molecular patterns (MAMPs). Although flg22 is recognized by many plant species via the pattern recognition receptor FLS2, neither the flgII-28 receptor nor the extent of flgII-28 recognition by different plant families is known.
Here, we tested the significance of flgII-28 as a MAMP and the importance of allelic diversity in flg22 and flgII-28 in plant–pathogen interactions using purified peptides and a Pseudomonas syringae ∆fliC mutant complemented with different fliC alleles.
The plant genotype and allelic diversity in flg22 and flgII-28 were found to significantly affect the plant immune response, but not bacterial motility. The recognition of flgII-28 is restricted to a number of solanaceous species. Although the flgII-28 peptide does not trigger any immune response in Arabidopsis, mutations in both flg22 and flgII-28 have FLS2-dependent effects on virulence. However, the expression of a tomato allele of FLS2 does not confer to Nicotiana benthamiana the ability to detect flgII-28, and tomato plants silenced for FLS2 are not altered in flgII-28 recognition.
Therefore, MAMP diversification is an effective pathogen virulence strategy, and flgII-28 appears to be perceived by an as yet unidentified receptor in the Solanaceae, although it has an FLS2-dependent virulence effect in Arabidopsis.
The perception of conserved molecular patterns of microbes, called PAMPs (pathogen-associated molecular patterns) or MAMPs (microbe-associated molecular patterns), is an important first line of plant defense, known as pattern-triggered immunity (PTI) (Jones & Dangl, 2006; Segonzac & Zipfel, 2011). The perception of the 22-amino-acid flagellin epitope flg22 is one of the most studied examples of PTI. The flg22 epitope binds directly to the pattern recognition receptor (PRR) FLAGELLIN SENSING2 (FLS2) (Chinchilla et al., 2006), after which FLS2 interacts with the adaptor protein BAK1 (Chinchilla et al., 2007b; Heese et al., 2007) and several other receptor-like kinases (RLKs) (Roux & Zipfel, 2012). These interactions lead to the rapid efflux of Ca2+ and the activation of calcium-dependent protein kinases (CDPKs) (Boudsocq et al., 2010), the generation of reactive oxygen species (ROS) (Felix et al., 1999) and the activation of mitogen-activated protein (MAP) kinase cascades (Rasmussen et al., 2012), triggering a complex defense response which includes FLS2-dependent stomatal closure to interfere with pathogen invasion (Melotto et al., 2006; Zeng & He, 2010), callose deposition to strengthen plant cell walls (Gómez-Gómez & Boller, 2000) and the induction of pathogenesis-related genes to restrict pathogen growth (Gómez-Gómez et al., 1999; Chinchilla et al., 2007a). Pretreatment of Arabidopsis plants with the flg22 epitope before pathogen inoculation decreases pathogen growth, and fls2 mutant plants are more susceptible to Pseudomonas syringae pv. tomato (Pto) infection following spray inoculation (Zipfel et al., 2004), demonstrating that flg22 recognition by FLS2 has biological relevance. Interestingly, flagellin recognition evolved in parallel in plants and animals (Ausubel, 2005; Zipfel & Felix, 2005). In mammals, the PRR TLR5 recognizes extracellular flagellin (Hayashi et al., 2001) and the intracellular receptor NLRC4 recognizes flagellin inside macrophages (Kofoed & Vance, 2011; Zhao et al., 2011).
An important strategy of plant pathogens to avoid PTI is the injection of immunity-suppressing effector proteins directly into host cells (Chisholm et al., 2006; Jones & Dangl, 2006; Cunnac et al., 2011). This strategy is best exemplified by the molecular mechanisms of the type III-secreted effector proteins AvrPto and AvrPtoB of Pto, both of which suppress FLS2-mediated immunity (Shan et al., 2008; Xiang et al., 2008; Cheng et al., 2011; Martin, 2012). As MAMP-containing proteins and other molecules are, by definition, essential for a pathogen's life cycle and/or pathogenicity (Jones & Dangl, 2006; Segonzac & Zipfel, 2011), pathogens cannot avoid PTI by losing essential proteins containing MAMPs, and allelic variation of MAMPs is expected to be limited by evolutionary constraints on their structure (Bittel & Robatzek, 2007; Boller & Felix, 2009; McCann et al., 2012). Nonetheless, a few studies have suggested that some pathogens are able to alter MAMPs to avoid PTI. For example, a single amino acid change in flg22 of the plant pathogen Xanthomonas campestris pv. campestris severely attenuated or eliminated the perception of flg22 by FLS2 (Sun et al., 2006); however, the mutation had no effect on pathogen growth during infection. Furthermore, no known alleles of flagellin from Ralstonia solanacearum elicit PTI (Pfund et al., 2004), but the effect of the evasion of flagellin recognition on pathogen growth during infection is not known. In addition, post-translational modifications of flagellin, including glycosylation, can have a major impact on the elicitation activity of flagellin (Taguchi et al., 2003; Takeuchi et al., 2003). Flagellin proteins from the human pathogens Bartonella bacilliformi, Campylobacter jejuni and Helicobacter pylori escape detection by TLR5 through mutations of amino acids in the known interaction surface (Andersen-Nissen et al., 2005). The diversity of flagellin perception as a result of allelic variation in FLS2 has also been demonstrated in that the flg15 epitope of Escherichia coli flagellin is recognized by tomato (Solanum lycopersicum), but not by Arabidopsis or Nicotiana benthamiana (Meindl et al., 2000; Bauer et al., 2001).
We have determined previously that a second epitope of flagellin, termed flgII-28, is sufficient to trigger immunity in tomato (Cai et al., 2011). The flgII-28 epitope was identified on the basis of two nonsynonymous mutations in almost identical strains of the bacterial speck disease pathogen Pto, which suggested selection for evasion of flgII-28 perception by tomato. In fact, the two derived alleles of flgII-28 (flgII-28K40 and flgII-28Col338), which are present in Pto strains typical of recent bacterial speck disease outbreaks, trigger a weaker immune response in tomato cultivar cv ‘Chico III’ than does the ancestral flgII-28T1 allele, which was the predominant allele in Pto strains that caused disease outbreaks before 1980 (Cai et al., 2011). Some Pto strains also have a mutation in flg22, but this mutation does not affect significantly the strength of the tomato immune response (Cai et al., 2011).
Here, we show that there is significant variation in both the strength of PTI elicited by different flagellin proteins in the same plant and in the perception of the same flagellin among different plants. Importantly, we show a significant effect of allelic variation in flg22 and flgII-28 on the outcome of plant–pathogen interaction in the absence of any effects on bacterial motility, thus revealing that allelic variation in MAMPs is an effective PTI avoidance mechanism in the evolutionary arms race between P. syringae pathogens and plants. Intriguingly, the effect of allelic variation in flgII-28 on bacterial growth in Arabidopsis is dependent on FLS2, although recognition of the flgII-28 peptide appears to be independent of FLS2 based on multiple lines of evidence. This suggests that flgII-28 may modulate indirectly flg22 perception by FLS2 and that solanaceous plants are probably equipped with a second flagellin receptor that recognizes flgII-28.
Materials and Methods
Peptide synthesis and storage
Peptides were synthesized to 70–90% purity by EZ Biolab (Carmel, IN, USA), with the exception of the flg22 consensus peptide (QRLSTGSRINSAKDDAAGLQIA) used in Figs 4(d) and Supporting Information S6, which was synthesized by Biomatik (Cambridge, ON, Canada). Water was added to the peptides to bring them to a concentration of 5 μM, except for the flgII-28 peptides which were insoluble in water (for some alleles) and were, instead, initially dissolved in a solution of 50% dimethylsulfoxide (DMSO) and diluted further in water. Dissolution in DMSO is unlikely to affect the ROS elicitation potential of the peptides because flgII-28T1 is an equivalent elicitor on S. lycopersicum whether dissolved in water or 50% DMSO (not shown). To maintain activity, peptides were stored in small aliquots to avoid multiple freezing and thawing. All dilutions to obtain working concentrations were performed in ultrapure H2O. Peptide solutions were stored at −20°C for short-term usage or −80°C for long-term storage.
Measurement of ROS generation
Previously described protocols were used to quantify the production of ROS following elicitation with peptides (Chakravarthy et al., 2010). Briefly, leaf disks of 4-wk-old plants (except eggplant and bean for which 6- and 3-wk-old plants, respectively, were used) were punched out with a #1 cork borer and floated adaxial side up overnight at room temperature in 200 μl double-distilled H2O (ddH2O) in individual wells of a clear-bottomed, 96-well microassay plate (Greiner Bio-one, Kremsmünster, Austria). Sixteen hours later, the water was replaced with a solution containing 1 μM peptide (or a concentration otherwise noted in the figure legend), 34 μg ml−1 luminol and 20 μg ml−1 horseradish peroxidase (Sigma, St Louis, MO, USA), all in ddH2O. The luminescence of each well was then measured immediately using a Synergy HT plate reader (Biotek, Winooski, VT, USA) employing the bottom camera at an optical sensitivity of 225 every 3 min for 60 min. For comparison among different peptides, multiple leaf punches were taken from the same leaf to allow for the peptides to act on the same leaf (but different sections) to control for variability among leaves. Six to eight leaves were used for each experiment.
The following plants were screened via the ROS assay in this work: Arabidopsis thaliana ecos. ‘Columbia-0’, ‘Cvi-0’ and ‘WS’ (The Arabidopsis Information Resource, Stanford, CA, USA); radish (Raphanus sativus) cv ‘Champion’ (Ferry Morse, Futon, KY, USA); turnip (Brassica rapa) cv ‘Purple Top White Globe’ (Ferry Morse); cauliflower (Brassica oleracea) cv ‘Toscana 2’ (Ferry Morse); snapdragon (Antirrhinum majus) cv ‘Cook's Tall’ (Cooks Garden, Warminster, PA, USA); morning glory (Convolvulaceae) (Ferry Morse); celery (Apium graveolens) cv ‘S. V. Pascal’ (Seeds of Italy, Harrow, UK); bean (Phaseolus vulgaris) cv ‘Red Mexican’ (Vermont Bean Seed Company, Randolph, WI, USA); tomato (Solanum lycopersicum) cvs ‘Rio Grande’, ‘Chico III’, ‘Sunpride’ (Bavicchi, Ponte San Giovani, Italy) and ‘Roter Gnom’ (Canadian Plant Genetic Resource Center, Ottowa, ON, Canada); pepper (Capsicum annuum) cvs ‘California Wonder’ (Wyatt-Quarles Seed, Garner, NC, USA) and ‘Jalapeno Early’ (Burpee, Warminster, PA, USA); eggplant (Solanum melongena) cv ‘MM643’ (INRA, Avignon, France); Nicotiana benthamiana, N. tabacum cv ‘Burly’ (personal laboratory stocks); potato (Solanum tuberosum) cv ‘Red Maria’ (Maine Potato Lady, Guilford, ME, USA).
Measurement of ethylene production following peptide treatment
Fully expanded leaves of 4–6-wk-old Solanum lycopersicum cv ‘MoneyMaker’ (Thompson & Morgan, Ipswich, Suffolk, UK) and N. benthamiana, both grown in long-day conditions (14 h photoperiod, 25°C), were cut into 2-mm slices and floated on water overnight. Leaf slices were transferred to 6-ml glass tubes containing 0.5 ml of an aqueous solution of the peptide being tested. Vials were closed with rubber septa and ethylene accumulating in the free air space was measured by gas chromatography after 3.5–4 h of incubation.
Bayesian analysis of ROS assays
For the analysis of ROS production, we modeled the data for each plant–peptide combination using a shifted Gompertz decay curve (Eqn 4 in Methods S1). Estimation of the unknown modeling parameters was performed using a Markov Chain Monte Carlo algorithm. To initially address whether ROS is elicited by a plant–peptide combination, we calculated the posterior probability of the null hypothesis (no ROS response) given the data. That is, we fitted the data to the Gompertz decay with the parameters Θ1 = Θ2 = 0 – a function of a flat line with no slope – and determined the posterior probability of this fit. To compare the ROS elicited by different peptides on the same plant, we determined the Bayesian credible intervals of differences between peptides on each plant for five parameters describing the kinetics of ROS production: peak intensity, offset, increase rate, change point and decrease rate (Fig. S1). (See Methods S1 for more details.)
Agrobacterium-mediated transient expression
For transient expression in N. benthamiana, Agrobacterium tumefaciens harboring FLS2p::FLS2-3xmyc-GFP (Robatzek et al., 2006) and FLS2p::SlFLS2-GFP (Robatzek et al., 2007) in pCAMBIA2300 and the empty vector were grown overnight in yeast extract broth (YEB) medium. Bacteria diluted in infiltration medium (10 mM MES, pH 5.6, 10 mM MgCl2) were pressure infiltrated into the leaves of 3–4-wk-old N. benthamiana plants. Leaves were assayed for ethylene biosynthesis 2 d after infiltration.
pH shift in Solanum peruvianum cell cultures
Cell cultures of S. peruvianum (Nover et al., 1982), formerly Lycopersicon peruvianum, were used as described previously (Meindl et al., 1998). Briefly, peptides were introduced into aliquots of 5–8-d-old cell suspensions on a rotary shaker at 120 rpm. Immediately after, the pH was measured with a glass electrode pH meter and recorded using a pen recorder. The plant wound hormone systemin (Meindl et al., 1998) was used as a positive control for cell culture alkalinization.
Virus-induced gene silencing of FLS2
Virus-induced gene silencing (VIGS) was performed using the Tobacco rattle virus (TRV) system described in Liu et al. (2002). Mixtures of A. tumefaciens containing pTRV1 and pTRV2 were prepared as described previously (Meng et al., 2013) and syringe infiltrated into cotyledons of 9-d-old ‘Rio Grande’ tomato seedlings. Plants were kept for c. 3 wk post-infiltration in a growth chamber with 20°C day and 18°C night temperatures in 50% relative humidity with a 16-h photoperiod. Plants were transferred to 24°C day and 20°C night temperatures 1 wk before performing the ROS assays.
Swim plate assay for the quantification of bacterial motility
Swim plates were prepared by making King's B medium (King et al., 1954) plates with 0.3% agar. Swim plates were always prepared on the same day as used. Plates were inoculated via ‘toothpick inoculation’ using a 2-μl pipette tip (Rainin, Columbus, OH, USA) by placing the tip in 1-d-old bacterial colonies on a plate and then gently placing the pipette tip on the swim plate. The pipette tip was always placed aligned with the radial lines of the plate. The plates were then placed in a 28°C incubator for 2 d. The diameter of swimming of each strain was measured using a ruler held perpendicular to the radial lines of the plate, and therefore perpendicular to the orientation of the pipette tip during inoculation, to nullify the slight variation resulting from imperfectly consistent pipette tip inoculations. Larger 150 × 15-mm2 Petri dishes (VWR, Radnor, PA, USA) were used to accommodate seven bacterial strains on the same swim plate to control for variability among plates.
Deletion of fliC from Pto DC3000
A flagellin-deficient mutant of Pto DC3000 was generated according to the method described previously (Shimizu et al., 2003). Approximately 0.8-kb fragments located on each side of fliC were amplified by PCR with primers (Table S1). Each amplified DNA fragment for upstream and downstream regions was ligated at the BamHI site and inserted into the EcoRI site of the mobilizable cloning vector pK18mobsacB. The resulting plasmid containing the DNA fragment lacking fliC was transformed into E. coli S17-1. The fliC deletion mutant was obtained by conjugation and homologous recombination. Specific deletions were confirmed by PCR using the primers PC1 and PC4.
Cloning of fliC and complementation of Pto DC3000∆fliC
Genomic DNA was extracted from the Pto strains DC3000, K40, T1 and ES4326, and used as a template for PCR amplification of fliC. Primers (Table S1) were designed to amplify 326 bp upstream (except for ES4326 with 327 bp upstream) and 52 bp downstream of the fliC open reading frame (ORF) with Sac1 and Xho1 restriction enzyme sites on the forward and reverse primers, respectively. This region contains both the fliC ORF and nearly the entire flanking regions up to the flanking genes in the flagellum gene cluster. PCR products were cleaned using an AccuPrep PCR purification kit (Bioneer, Alameda, CA, USA) and digested by SacI and XhoI (NEB, Ipswich, MA, USA) at 37°C for 3 h. The broad expression vector pME6010 was similarly digested with SacI and XhoI. All digested products were separated using gel electrophoresis and cleaned using an AccuPrep gel extraction kit (Bioneer), and each PCR product was individually ligated into the digested pME6010 vector using DNA ligase (Takara Bio, Shiga, Japan). Because of the orientation of restriction enzyme sites in the multiple cloning site of pME6010, each PCR product was ligated 3′ to 5′ with respect to the Pk promoter of 6010; therefore, fliC is under the control of its native promoter in these vectors. The resulting vectors were transformed via conjugation into Pto DC3000∆fliC.
Plant protection and plant infection assays
Plant protection assays were performed in a similar manner to previously described protocols (Zipfel et al., 2004; Cai et al., 2011). Sixteen hours before infection, 1 μM of each peptide was infiltrated into the adaxial side of at least five attached leaves on two to four different 4- or 5-wk-old plants via blunt end syringe, and marked to indicate the sites of infiltration. Plants were placed in high humidity for 16 h before spray inoculation of 10 ml of bacterial suspension per pot at an optical density at 600 nm (OD600) of 0.01 using a Preval® sprayer (Preval, Coal City, IL, USA). Silwet L-77 was not added to the bacterial suspension to avoid the creation of artificial conditions in which flagella are probably less essential. The plants were then covered with clear plastic flat covers to maintain high humidity for 16 h following inoculation. After the number of days indicated in each figure legend, the regions of the leaves that had been previously infiltrated by peptide were removed using a 0.52-mm2 cork borer and placed in a tube containing 200 μl of 10 mM MgSO4 solution and three 2-mm glass beads. The tube was placed in a mini-bead beater (Biospec Products, Inc., Bartlesvill, OK, USA) and shaken for 90 s to grind the leaf and release endophytic bacteria. This suspension was diluted in 10 mM MgSO4 and plated on a King's B medium plate with appropriate antibiotic selection. After 2 d, colony-forming units were counted. Similar procedures were used to quantify the growth of the Pto DC3000∆fliC strains on A. thaliana and S. lycopersicum. Arabidopsis thaliana fls2 mutant plants (Xiang et al., 2008) were kindly provided by Jian-min Zhou (Chinese Academy of Sciences, Beijing, China).
For inoculation by infiltration, plants were infiltrated via a blunt-end syringe at an OD600 of either 1 × 10−3 (Arabidopsis) or 1 × 10−4 (tomato). Population assays were performed as described above.
Diversity exists in flg22 and flgII-28 in the P. syringae species complex
As we found unexpected diversity among Pto strains in both the flg22 and flgII-28 epitopes (Cai et al., 2011), we assembled the flg22 and flgII-28 alleles from all available genomes of the P. syringae species complex. The positions within flg22 and flgII-28 that are mutated in Pto strains are conserved in P. syringae strains belonging to other pathovars (Table S2); however, some strains show overall sequence variation in both epitopes, particularly P. syringae pv. maculicola ES4326, a strain recently reassigned to P. cannabina pv. alisalensis (Pcal) (Bull et al., 2010). The subsets of flg22 and flgII-28 alleles tested in this article are shown in Fig. 1.
Mapping of the locations of flg22 and flgII-28 to the known flagellin structure (Yonekura et al., 2003) revealed that flg22 and flgII-28 are structurally similar regions of flagellin: both are found within loop regions located between alpha helices (data not shown). Loops define the angle between flanking alpha helices and are thus important determinants of protein structure, which may, in fact, explain the relative sequence conservation of both flg22 and flgII-28.
flgII-28 recognition is limited to a subset of the Solanaceae family
We have shown previously that the flgII-28 alleles of Pto strains K40 (same genotype as LNPV17.41) and Col338 trigger less ROS production and less stomatal closure than the flgII-28 allele of Pto T1 on tomato cv ‘Chico III’ (Cai et al., 2011). To test whether plants other than tomato can recognize flgII-28, ROS production was measured in 18 different plant genotypes across six different families, with multiple Brassicaceae and Solanaceae species tested. The posterior probability (P(H0|Data)) that the observed data for each plant–peptide combination can be fitted to a function with no slope (i.e. no ROS production; see 'Materials and Methods' for more details) was then calculated. Although flg22T1 triggered ROS generation in 15 of the tested plants, flgII-28T1 triggered ROS production in only six tested plants – all of which were in the Solanaceae family (Table 1). Two of the plants that did not respond to flg22T1 were ecotypes of Arabidopsis known to be defective in the FLS2 locus: Ws-0 (Gómez-Gómez et al., 1999) and Cvi-0 (Dunning et al., 2007); the third was celery (Apium graveolens). The flgII-28T1 peptide, on the other hand, only elicited ROS in the three tested tomato cultivars, the two tested cultivars of pepper (Capsicum annuum) and the one tested cultivar of potato (Solanum tuberosum) (Table 1). Graphical representations of selected ROS curves are depicted in Fig. S2.
Table 1. Production of reactive oxygen species (ROS) in response to either flg22T1 or flgII-28T1 across the plant kingdom
ROS response of selected species of the Brassicaceae, Solanaceae and other plant families to flg22 and flgII-28 peptides: +, strong response; \, no response; +\, weak to no response. Data represent the analysis from one experiment of six leaf punches. Similar results were obtained in 2–12 independent experiments (plants tested in only two experiments labeled with *).
Posterior probability of the data being fitted to the null model (no response – the function being fitted is a flat line with no slope) explaining the observed data (i.e. the probability that there is no measurable production of ROS). See the 'Materials and Methods' section for more details.
In c. 25% of experimental repeats, flgII-28 triggered significant amounts of ROS production cautioning the variability among plants in this assay.
The relative strength of the immune response triggered by flgII-28 and flg22 alleles is dependent on plant genotype
To test whether the derived flgII-28 alleles (flgII-28K40 and flgII-28Col338) are generally weaker elicitors of PTI relative to the ancestral flgII-28T1 allele, we performed ROS analyses for both the ancestral and derived flgII-28 peptides on all Solanaceae tested above. We indeed observed that the derived alleles of flgII-28 triggered significantly less ROS than flgII-28T1 on both tested cultivars of tomato. By contrast, the derived alleles of flgII-28 triggered only slightly less ROS on pepper (Fig. 2).
To determine whether the differences in ROS elicitation were statistically relevant, we modeled the ROS data for each plant–peptide combination using a shifted Gompertz decay curve and determined the differences in the kinetics of the ROS response between different peptides on the same plants (see the 'Materials and Methods' section, Methods S1 and Fig. S1 for more information). Using this approach to compare the ROS curves depicted in Fig. 2, flgII-28K40 and flgII-28Col338 were found to give rise to ROS elicitation with significantly lower peak intensity and longer offset than flgII-28T1 in both tomato cv ‘Rio Grande’ and tomato cv ‘Sunpride’. Contrastingly, no significant differences among peptides were detected in ROS elicitation in pepper cv ‘Jalapeno Early’ (Tables S3, S4). Extending this statistical analysis to all other plants capable of detecting flgII-28T1 (Table 1), flgII-28K40 and flgII-28Col338 were found to elicit significantly less ROS than flgII-28T1 on potato cv ‘Red Maria’ and tomato cv ‘Chico III’. However, as for pepper cv ‘Jalapeno Early’, the derived flgII-28 peptides do not trigger significantly less ROS on pepper cv ‘California Wonder’ (Tables S5, S6).
A single nonsynonymous single nucleotide polymorphism (SNP) distinguishes the flg22 allele of the Pto strain Col338 from the flg22 allele present in all other Pto strains, including Pto DC3000 and T1 (Fig. 1). Peptides of this flg22 variant triggered stronger responses in some Solanaceae, but weaker responses in others, when compared with flg22DC3000 (Fig. 2, Tables S7, S8).
Ethylene production is another quantifiable PTI response (Boller & Felix, 2009). Figure 3 shows that flgII-28K40 and flgII-28Col338 elicit significantly less ethylene biosynthesis than flgII-28T1 in tomato cv ‘Money Maker’, further confirming that flgII-28K40 and flgII-28Col338 trigger weaker immune responses in tomato relative to flgII-28T1. Moreover, flgII-28T1 is at least as active in eliciting ethylene biosynthesis in tomato as flg22.
Overall, these data indicate that allelic variability in both the epitopes of flagellin and in the plant PRRs that recognize them determine the intensity of PTI, suggesting co-evolution between MAMPs and PRRs. We conclude that the derived lineages of Pto have evolved to avoid flagellin recognition by S. lycopersicum and/or S. tuberosum (and possibly other related Solanaceae not tested here), but these mutations in flagellin do not make it universally more stealthy in all plant interactions.
Arabidopsis cannot perceive flgII-28 peptide
We further tested the ability of A. thaliana to detect flgII-28 in four independent PTI assays: the plant protection assay (Zipfel et al., 2004), callose deposition (Adam & Somerville, 1996), ethylene production (Boller & Felix, 2009) and cell culture alkalinization (Meindl et al., 1998). All four assays confirmed that Arabidopsis does not respond to flgII-28 peptide (Fig. S3A,B,E,F). We also observed that the flgII-28 allele of the Arabidopsis and tomato pathogen Pto DC3000, which differs by two amino acid residues from flgII-28T1 (Fig. 1), does not elicit ROS in A. thaliana (Fig. S3C), nor do two other alleles of Pto flgII-28 (Fig. S3D). Overall, the detection of flgII-28 seems to be limited to a subset of species in the Solanaceae family, contrasting with flg22, which is broadly recognized across the plant kingdom (Table 1, Albert et al., 2010).
FLS2 is not sufficient for the recognition of flgII-28
To test whether flgII-28 is recognized by FLS2, we transiently expressed FLS2 from either tomato cv ‘Roter Gnom’ (Robatzek et al., 2007), which is capable of detecting both flgII-28 and flg22 (Fig. S4), or A. thaliana, which is nonresponsive to flgII-28 (Table 1), in nonresponsive N. benthamiana plants. Plants were treated with flg22, flgII-28 or flg15E. coli – a truncation variant of flg22 that is recognized by tomato but not by A. thaliana or N. benthamiana (Bauer et al., 2001; Robatzek et al., 2007). Both FLS2 proteins accumulated in N. benthamiana (Fig. S5), and the expression of SlFLS2 conferred to N. benthamiana responsiveness to flg15E. coli (Fig. 4a), confirming the functionality of the receptor. However, transient expression of either allele of FLS2 did not confer responsiveness to flgII-28 (Fig. 4a).
The addition of flgII-28 peptide to an S. peruvianum cell culture saturated for the flg22 response triggers a further increase in extracellular alkalinization and, vice versa, the addition of flg22 peptide to an S. peruvianum cell culture saturated for the flgII-28 peptide triggers a further increase in extracellular alkalinization (Fig. 4b). Taken together, these results suggest that flgII-28 is recognized by a PRR distinct from FLS2.
Moreover, VIGS of FLS2 in tomato had no effect on flgII-28 recognition, but attenuated significantly ROS production elicited by flg22 treatment (Fig. 4c). Recent sequencing of the tomato genome (Consortium, 2012) revealed a close paralog located adjacent to the original FLS2 on chromosome 2, which we designated as FLS2.2. The DNA fragment in our FLS2-VIGS construct, cloned using FLS2.1 (Solyc02 g070890), has eight stretches of at least 21 nucleotides with perfect identity to FLS2.2 (Solyc02 g070910), and the expression of both genes was decreased by c. 40% (Fig. S6). These data suggest that neither copy of SlFLS2 contributes to the recognition of flgII-28.
However, pretreatment of tomato leaf disks with flg22 reduced significantly ROS elicitation by the subsequent addition of flg22 or flgII-28 relative to mock pretreatment (Fig. 4d), suggesting that flg22 and flgII-28 recognition share at least one common downstream signaling component. Because flgII-28 typically elicits significantly more ROS than flg22 in tomato, this result is not simply caused by the exhaustion of the reagents used in the assay or exhaustion of the ROS substrates within cells following flg22 pretreatment.
Alleles of P. syringae flg22 are not universally recognized in either Brassicaceae or Solanaceae
The recently sequenced Pcal ES4326 strain (Baltrus et al., 2011) causes disease in Arabidopsis (Dong et al., 1991) and is related to strains that cause disease on various other Brassicaceae (Bull et al., 2010, 2011). Because flg22ES4326 is the most divergent flg22 allele relative to all other known P. syringae flg22 epitopes (only 12 identities between flg22ES4326 and flg22DC3000; Fig. 1; Table S2), we hypothesized that ES4326 and ES4326-like strains have evolved a flg22 sequence adapted to avoid recognition. Indeed, flg22ES4326 failed to elicit PTI based on ROS assays and did not promote resistance to P. syringae in either Arabidopsis (Fig. 5a,b; Table S9) or tomato (Fig. 5c,d; Table S9). Therefore, diversity within P. syringae flg22 does not only affect the strength of the PTI response, but can even lead to complete avoidance of PTI.
When performing plant protection assays in tomato, pretreatment with flg22ES4326 consistently led to more pathogen growth relative to mock-treated plants (Fig. 5d), a result confirmed on a second tomato cultivar (Fig. S7). We hypothesize that flg22ES4326 may interact with FLS2 in an antagonistic manner, reducing the detection of native flagellin from the pathogen during infection. Such characteristics have been described for truncated versions of flg22 (Felix et al., 1999) and flg22Rsol, the flg22 variant of R. solanacearum (Mueller et al., 2012). To test this hypothesis, we used cultured cells of wild tomato S. peruvianum, which respond to flg22DC3000 and flgII-28T1 with a rapid extracellular alkalinization response (Fig. 5e). As hypothesized, treating the cells with flg22ES4326 did not lead to extracellular alkalinization, but inhibited a subsequent response to flg22 (Fig. 5e), confirming that flg22ES4326 acts as an antagonist for flg22, possibly by engaging the FLS2 receptor without triggering an immune response. Notably, pretreatment with flg22ES4326 did not reduce the alkalinization response to flgII-28T1 (Fig. 5f), further corroborating that flgII-28 is not an active ligand for FLS2.
Allelic diversity in FliC does not impact motility in a Pto DC3000∆fliC background
MAMPs are considered to be favorite targets for recognition by the plant immune system because they represent conserved structures with essential roles in invariant microbial processes. Thus, we tested whether the observed mutations in flagellin affect the swimming motility of P. syringae, which is the primary function of the flagellum. To quantify swimming motility, soft-agar swim plates were used in which flagellar-motile bacteria will spread from an initial point of inoculation (Hazelbauer et al., 1969). In this assay, wild-type Pto strain K40 and Pcal strain ES4326 were both significantly less motile than the other tested strains (Fig. S8). However, in addition to the allelic differences in fliC, there are many more genetic differences between the analyzed strains. Therefore, to specifically quantify the effect of the fliC SNPs in an isogenic background, we deleted the endogenous fliC gene in Pto DC3000 and complemented it with four different fliC alleles under the control of their native promoters. As expected, Pto DC3000∆fliC was found to be significantly impaired in motility. By contrast, strains complemented with the four fliC alleles were indistinguishable from each other and swam out from the initial point of inoculation at a level just below that of wild-type Pto DC3000 (Fig. 6). We conclude that the mutations in flagellin do not substantially affect motility.
Allelic diversity in fliC affects disease development, but not growth, of P. syringae in tomato
To test whether the fliC mutations affect bacterial growth in planta, tomato plants were infiltrated with Pto DC3000 wild-type, Pto DC3000∆fliC and the four complemented Pto DC3000∆fliC strains described above. We hypothesized that the alleles of fliC containing the stealthier versions of flgII-28 would allow for higher growth relative to Pto DC3000. However, all strains grew to equally high population densities 3 and 4 d following inoculation by syringe infiltration (Figs 7a, S9, respectively) and spray inoculation (Fig. S10A). It is surprising that even the uncomplemented Pto DC3000∆fliC strain was able to reach a population density as high as that of the other strains, even following spray inoculation in the absence of a surfactant. We hypothesize that Pto DC3000 either uses forms of motility other than swimming during infection of tomato, or that the cost of being nonmotile is counterbalanced by lower PTI elicitation as a result of the absence of flg22 and flgII-28 in this mutant.
Intriguingly, although the various fliC alleles did not affect bacterial growth in the Pto DC3000∆fliC background, they did lead to significant differences in the development of leaf necrosis (Fig. 7b,c), suggesting that in planta expression of FliC protein can interfere with symptom development, which was recently demonstrated (Wei et al., 2012), in an allele-dependent manner.
Allelic diversity in fliC affects the virulence of P. syringae in Arabidopsis
To test the effects of the identified SNPs in a second pathogen–host interaction, Arabidopsis was inoculated with the complemented Pto DC3000∆fliC strains. Interestingly, Pto DC3000∆fliC complemented with fliCK40 grew to significantly lower population densities than the other strains following inoculation by infiltration (Fig. 8a). Because flagella are known to play only a minor role in apoplastic growth (Schreiber & Desveaux, 2011), we also queried for differences in fitness among the strains following spray infection, and confirmed the fitness defect of Pto DC3000∆fliC complemented with fliCK40 (Fig. S10B). Importantly, no surfactants were used in the spray inoculations to avoid the creation of an artificial condition in which flagella are less important. This strain also triggered less severe disease symptoms (Figs S10D, S11). The difference between Pto DC3000∆fliC complemented with either fliCK40 or fliCT1 is particularly intriguing, considering that the only amino acid difference between these two strains is a single S→F mutation in the flgII-28 region of flagellin – a region that does not, as an isolated peptide, elicit PTI in Arabidopsis. To determine whether the observed differences between the two strains were dependent on FLS2, leaves of an fls2 Arabidopsis mutant line (Xiang et al., 2008) were inoculated by either spraying or infiltration. Interestingly, the differences between the two strains were abolished (Figs 8b, S10C). The differences were also abolished in Arabidopsis ecotypes that have a defect in FLS2 (Fig. 8c). We conclude that diversity in flagellin beyond flg22 can affect significantly P. syringae virulence in an FLS2-dependent manner.
To rule out the possibility that the observed differences between complemented Pto DC3000∆fliC strains were caused by mutations accumulated during the transformation of the Pto DC3000∆fliC strain with the different fliC alleles, and not by the different fliC alleles themselves, the results for growth and necrosis development in tomato and growth in Arabidopsis were confirmed with a second set of independently transformed isolates of all strains (Fig. S12).
As flg22 has been found to be sufficient to elicit PTI in multiple clades of seed plants belonging to angiosperms and even gymnosperms (Felix et al., 1999; Albert et al., 2010), studies of flagellin-triggered plant immunity have mainly focused on the flg22 epitope, although a flg22-independent flagellin perception system has been proposed for rice (Che et al., 2000). However, in performing a population genomics analysis of a large collection of Pto strains (Cai et al., 2011), we identified a second PTI-eliciting flagellin epitope, flgII-28. The study of pathogen diversity can thus lead to new insights into plant–microbe interactions (Cai et al., 2011), and can provide an efficient strategy for the identification of new MAMPs, as recently confirmed by others (McCann et al., 2012).
Continuing the characterization of flagellin diversity here, we have shown the following: flgII-28 elicits immunity in several Solanaceae species, but not in any members of other plant families tested so far; significant and biologically relevant allelic diversity exists at both MAMP loci of flagellin (flg22 and flgII-28), demonstrating the ability of adapted pathogens to overcome PTI through MAMP diversification; allele-dependent differences in PTI are also dependent on plant genotype, implying natural variability in the plant receptor(s) for these epitopes; the PTI-altering mutations in flagellin do not come at the cost of microbial fitness in the tested conditions, suggesting that MAMPs are under weaker purifying selection than previously thought; FLS2 is probably not the flgII-28 receptor, although allelic diversity in flgII-28 affects plant virulence in an FLS2-dependent manner in Arabidopsis.
flgII-28 elicits PTI responses in several Solanaceae species
Based on the flgII-28 alleles and plant species analyzed here, the recognition of flgII-28 is limited to a subset of Solanaceae species. The most obvious explanation for this result is that flgII-28 recognition evolved in a relatively recent ancestor of some Solanaceae. However, at this point, we cannot exclude the possibility that flgII-28 recognition is widespread in the plant kingdom, and that the plants that were nonresponsive to flgII-28 in this study recognize an as yet untested allele of flgII-28.
Significant diversity in MAMPs and MAMP perception reveals an evolutionary PTI arms race
Because of the relative differences in the strength of PTI triggered in closely related Solanaceae by the different alleles of the flgII-28 peptide, we infer the importance of allelic variability in both MAMPS and corresponding plant PRRs. In addition, Vetter et al. (2012) have recently described extensive variation in the recognition of flg22 across Arabidopsis ecotypes as a result of allelic differences in FLS2 sequences, differences in FLS2 protein abundance and differences in downstream signaling components. However, the observed variability in flgII-28 recognition among Solanaceae appears to be primarily linked to the sequence diversity of a putative flgII-28 receptor. Indeed, differences in the abundance of a receptor or in elements in downstream signaling between plant species would equally affect the recognition of all flgII-28 alleles and not lead to relative differences in the recognition of flgII-28 alleles as observed here.
As flgII-28K40 and flgII-28Col338 elicit less PTI in tomato and have almost entirely replaced the ancestral flgII-28T1 alleles in Pto populations in Europe and North America over the last 30 yr, we hypothesized that these alleles have recently evolved under selection pressure in tomato agricultural settings (Cai et al., 2011). However, we also noticed that the flgII-28K40 allele appeared in two separate genetic lineages of Pto (Cai et al., 2011), suggesting that these alleles might have been acquired through horizontal gene transfer. As we have now found that the same alleles also trigger less PTI in other plants, such as potato, it is plausible that these alleles pre-existed in the P. syringae population, evolved under selection pressure for PTI avoidance on hosts other than tomato and were only later acquired by Pto through horizontal gene transfer.
However, the significant reduction in bacterial growth and disease development in Arabidopsis during Pto DC3000 infection caused by the flagellin allele of Pto K40 relative to Pto DC3000 shows that adaptation for PTI avoidance on some plant species (here tomato) may have a collateral effect of increasing PTI on other plants (here Arabidopsis). This may explain why the mutations that gave rise to the flgII-28 alleles of Pto K40 and Col338 are not present in any other sequenced P. syringae strains. For the first time, effects of flg22 and flgII-28 on bacterial growth were observed by expressing different fliC alleles under their native promoters in the same genetic pathogen background during infection. These results thus reveal the biological significance of allelic diversity in MAMPs.
The complete absence of PTI elicited by the flg22 allele of Pcal ES4326 in Arabidopsis and tomato is also remarkable. Interestingly, Pcal ES4326 does not have the effector avrPto, which interferes with FLS2 kinase activity (Baltrus et al., 2011), and the Pcal ES4326 homolog of the effector avrPtoB is missing the ubiquitin-ligase domain (Guttman et al., 2002), which has been reported to ubiquitinate FLS2, leading to its degradation (Göhre et al., 2008). This suggests that the evasion of PTI through allelic diversification at MAMP loci is similarly efficient in overcoming PTI as the delivery of PTI-suppressing effector proteins.
Diversity in flg22 and flgII-28 among closely related pathogens is particularly striking, because MAMPs are considered to be essential for important functions (Jones & Dangl, 2006; Bittel & Robatzek, 2007; Lacombe et al., 2010), such as motility in the case of flagellin, and thus under purifying selection. However, to our surprise, none of the differences in flg22 or flgII-28 led to deficiencies in motility in vitro when placed in the same genetic background (Pto DC3000∆fliC), demonstrating that there is significant room for divergence at both of these epitopes without cost to motility, thus unsettling the theory that PTI is more evolutionarily stable than its effector-recognizing corollary – effector-triggered immunity. Importantly, the differences in flgII-28 and flg22 within the Pto lineage are the only polymorphisms in FliC; therefore, the success of these mutations cannot be explained by intragenic compensatory changes, which have previously been implicated as essential for bacteria to mutate MAMPs to avoid recognition but maintain function (Andersen-Nissen et al., 2005).
What is the flgII-28 receptor?
In bacterial flagellin, flg22 and flgII-28 are physically linked by a stretch of only 33 amino acid residues, hinting at the possibility that both epitopes might act via the same receptor, FLS2. However, when applied as separate synthetic peptides, both epitopes act as distinct and independent MAMPs on tomato cells. For example, cells pretreated with highly saturating concentrations of one of the peptides still respond to the other peptide with further response (Fig. 4b). Therefore, the hypothesis of FLS2 as receptor for both epitopes would imply that flgII-28 and flg22 have distinct binding sites within FLS2, because the two peptides do not compete for binding (Fig. 4b), and that binding of flgII-28 is a peculiar feature of some of the solanaceous FLS2 orthologs. As an argument against this hypothesis, we observed that gene silencing of FLS2 in tomato attenuated the response to flg22, but not flgII-28, in the ROS assays. Importantly also, heterologous expression of the tomato ortholog Sl FLS2 is not sufficient to confer recognition of flgII-28 to N. benthamiana (Fig. 4a). This strongly suggests that tomato has at least one additional factor that specifies the perception of flgII-28. We tentatively termed this yet-to-be-identified factor FLS3 (Flagellin sensing3). At present, we cannot exclude a role of FLS2 in the perception of flgII-28. For example, the postulated FLS3 might be a co-receptor or signaling component that is specifically required for FLS2-dependent detection of flgII-28, and only present in Solanaceae species that respond to flgII-28. However, based on the cumulative evidence from our experiments, we rather conclude that it is more likely that the postulated FLS3 acts as the genuine receptor for flgII-28. Candidates for such a FLS3 receptor are among the hundreds of orphan RLKs and receptor-like proteins in higher plants, many of which show species-specific variation that could fit with the occurrence of flgII-28 perception in plants.
However, if FLS2 is not the flgII-28 receptor, how can we explain the FLS2-dependent virulence effect of mutations in flgII-28 during the infection of Arabidopsis? Multiple assays showed that flgII-28 peptide itself is not an elicitor in Arabidopsis (Fig. S2) and flgII-28 variants only had an FLS2-dependent virulence effect when in the context of the entire FliC protein. Therefore, Arabidopsis might have an FLS3 allele that cannot bind flgII-28 peptide by itself, but requires a larger region of FliC to interact with flgII-28. FLS2 may act as a necessary co-receptor for FLS3, and FLS2–flg22 binding may even be necessary to place the flgII-28 region of FliC in proximity to the flgII-28 binding site of FLS3. This is reminiscent of the mammalian macrophage flagellin receptor NLCR4, which acts as a co-receptor hub for multiple MAMPs – both flagellin and type III secretion system components – after the elicitors bind to the NOD-like receptors NAIP5 and NAIP2, respectively (Kofoed & Vance, 2011; Zhao et al., 2011). An equally likely alternative hypothesis is that mutations in flgII-28 lead to FLS2-dependent PTI modulation through an indirect mechanism. For example, mutations in flgII-28 may alter the release of flagellin monomers during flagella assembly, or may cause conformational changes in the FliC structure, altering the exposure of flg22 to FLS2.
An emerging parallel in plant and animal innate immunity
Interestingly, the proposed TLR5 recognition site of flagellin in animals (Yoon et al., 2012) partially overlaps with flgII-28, whereas flg22 does not seem to play any role in TLR5–flagellin binding. Therefore, future studies of FLS3-mediated recognition of flagellin in plants, in particular, the identification of FLS3 and the elucidation of flgII-28–FLS3 interaction, will be highly informative in unraveling a new, intriguing example of convergent evolution in plants and animals.
We thank John McDowell (Virginia Tech) for discussions and a critical review of the manuscript. Research in the Vinatzer laboratory is supported by the National Science Foundation (NSF) (Grant #0746501), in the Chinchilla laboratory by the Swiss National Foundation (Grant #31003A_138255) and in the Martin laboratory by the NSF (Grant #IOS-1025642), the National Institutes of Health (NIH) (Grant #R01-GM078021) and the TRIAD Foundation.