High levels of cyclic‐di‐GMP in plant‐associated P seudomonas correlate with evasion of plant immunity

Summary The plant innate immune system employs plasma membrane‐localized receptors that specifically perceive pathogen/microbe‐associated molecular patterns (PAMPs/MAMPs). This induces a defence response called pattern‐triggered immunity (PTI) to fend off pathogen attack. Commensal bacteria are also exposed to potential immune recognition and must employ strategies to evade and/or suppress PTI to successfully colonize the plant. During plant infection, the flagellum has an ambiguous role, acting as both a virulence factor and also as a potent immunogen as a result of the recognition of its main building block, flagellin, by the plant pattern recognition receptors (PRRs), including FLAGELLIN SENSING2 (FLS2). Therefore, strict control of flagella synthesis is especially important for plant‐associated bacteria. Here, we show that cyclic‐di‐GMP [bis‐(3′‐5′)‐cyclic di‐guanosine monophosphate], a central regulator of bacterial lifestyle, is involved in the evasion of PTI. Elevated cyclic‐di‐GMP levels in the pathogen P seudomonas syringae pv. tomato (P to) DC3000, the opportunist P . aeruginosa  PAO1 and the commensal P . protegens  Pf‐5 inhibit flagellin synthesis and help the bacteria to evade FLS2‐mediated signalling in N icotiana benthamiana and A rabidopsis thaliana. Despite this, high cellular cyclic‐di‐GMP concentrations were shown to drastically reduce the virulence of P to  DC3000 during plant infection. We propose that this is a result of reduced flagellar motility and/or additional pleiotropic effects of cyclic‐di‐GMP signalling on bacterial behaviour.


INTRODUCTION
Plants rely on an innate immune system consisting of passive and active defences to resist potential attack by most microbial patho-gens (Dangl et al., 2013). During infection, cell surface-localized pattern recognition receptors (PRRs) perceive pathogen/microbeassociated molecular patterns (PAMPs/MAMPs), molecules released from microbial pathogens, and trigger local and systemic immune responses (Boller and Felix, 2009). The PRRs FLAGELLIN SENSING2 (FLS2) and EF-TU RECEPTOR (EFR) specifically recognize the peptidic MAMPs flg22, derived from bacterial flagellin, and elf18, derived from elongation factor thermo-unstable (EF-Tu), respectively, and activate an intracellular signal transduction cascade that results in pattern-triggered immunity (PTI) (Zipfel, 2014). Although flg22 recognition seems to be conserved in most plant species and functional FLS2 orthologues have been identified in many plants, including Arabidopsis thaliana (Gomez-Gomez and Boller, 2000) and Nicotiana benthamiana (Hann and Rathjen, 2007), a functional EFR protein has only been found in Brassicaceae (Boller and Felix, 2009;Kunze et al., 2004;Zipfel et al., 2006).As PTI is activated by features conserved across entire groups of microbes, it is sufficient to ward off a broad range of microbial threats (Boller and Felix, 2009;Zipfel, 2014).
Plants are continuously exposed to a highly complex microbial community that contains not only pathogens, but also many commensal and beneficial species. Indeed, the microbiomes of the phyllosphere and rhizosphere play critical roles in the adaptation of a plant to its environment (Berendsen et al., 2012;Guttman et al., 2014). Flagellin and EF-Tu (with the immunogenic epitopes flg22 and elf18, respectively) are typical MAMPs. They are widespread among many bacterial species, abundant and highly conserved, fulfilling important bacterial functions (Pel and Pieterse, 2013). Although the PRRs for flg22 and elf18 ensure the reliable recognition of diverse bacterial pathogens, the universal nature of these MAMPs means that beneficial and commensal microbes are also potentially exposed to immune recognition (van Loon, 2007;Van Wees et al., 2008;Zamioudis and Pieterse, 2012). Indeed, some symbiotic bacteria, such as Rhizobium, have evolved to evade flg22 recognition (Felix et al., 1999).
Pseudomonas is an important genus of plant-associated bacteria that contains both phytopathogenic and commensal species. Pseudomonas syringae pv. tomato (Pto) DC3000 is a foliar, hemibiotrophic plant pathogen that causes bacterial speck disease on tomato and A. thaliana (Xin and He, 2013). Primary infection sites are natural openings and wounds, through which bacteria migrate into the apoplast before multiplying rapidly, leading to chlorosis and necrosis of plant tissue (Xin and He, 2013). Pseudomonas aeruginosa is an opportunistic pathogen that infects immune compromised humans and can also colonize and infect plants, although this ability is limited to specific hosts (Starkey and Rahme, 2009). Attachment, colonization and proliferation of P. aeruginosa in A. thaliana have been described as having similarities to the Pto DC3000 infection process (Plotnikova et al., 2000), although no specific plant virulence factor has been identified. Pseudomonas aeruginosa growth in the intercellular space leads to systemic infection and, ultimately, to severe soft-rot symptoms (Plotnikova et al., 2000). Conversely, Pseudomonas protegens Pf-5 is a soil bacterium that colonizes the rhizosphere and promotes plant growth by suppression of a wide variety of plant diseases over a broad host range (Loper et al., 2007). Pseudomonas protegens Pf-5 produces multiple secondary metabolites, including pyoluteorin and 2,4diacetylphloroglucinol, that underpin its biocontrol capacities. In addition to the production of antibiotics, siderophore secretion and genetic features, such as broad catabolic pathways, an expanded array of efflux systems and numerous genes conferring tolerance to oxidative stress enable P. protegens Pf-5 to cope with environmental stress and microbial competition in the rhizosphere. Consistent with its commensal lifestyle, certain pathogenicity factors, such as the type-III secretion system (T3SS), are not present in the Pf-5 genome (Loper et al., 2007(Loper et al., , 2012. Like many Gram-negative bacteria, these three plantassociated Pseudomonas spp. have polar flagella that confer directed mobility and enable both the spatial colonization of plant surfaces and migration into the apoplast (Jackson, 2009).
The bacterial second messenger cyclic-di-GMP (cdG) is a key regulator of flagella expression and function in Pseudomonas (Hickman and Harwood, 2008). In general, cdG controls processes involved in the switch between single-celled motile and communal sessile lifestyles in many bacterial species (Hengge, 2009). Pseudomonas and other bacteria integrate environmental cues and intracellular signals in cdG signalling pathways, which regulate a diverse range of behaviours, including motility (Dasgupta et al., 2003), adhesion to surfaces (Newell et al., 2011), biofilm formation (Hickman et al., 2005 and virulence (Kulasakara et al., 2006). The level of intracellular cdG is coordinated by the opposing enzymatic activities of multiple diguanylate cyclases (DGCs) and phosphodiesterases (PDEs) (Hengge, 2009). In P. aeruginosa, flagella-driven motility and exopolysaccharide (EPS) production are reciprocally controlled by several cdG-dependent systems (Dasgupta et al., 2003;Lee et al., 2007), with the cdG-binding transcription factor FleQ playing a central role. FleQ is a σ 54dependent master regulator that controls the expression of genes, including the fleSR two-component system, flagella export apparatus loci and genes involved in the initiation of flagella basal body assembly (Dasgupta et al., 2003;Robleto et al., 2003). When cdG levels are low, FleQ inhibits EPS production by the repression of pel and psl EPS-operon transcription, but is required for the transcription of multiple flagellar genes. Correspondingly, on binding of cdG, FleQ both releases EPS biosynthetic gene repression and abolishes flagella gene transcription, enabling the switch from motility to EPS production and biofilm formation (Baraquet and Harwood, 2013;Hickman and Harwood, 2008). Homologues of FleQ are present in most Pseudomonas spp., including P. syringae and P. protegens (Winsor et al., 2011).
Here, we investigate the impact of increased intracellular cdG levels in different Pseudomonas spp. on plant immune responses. Bacterial extracts from the opportunistic P. aeruginosa PAO1, the commensal P. protegens Pf-5 and the pathogenic Pto DC3000 were assessed for their ability to elicit plant defence responses by measuring ROS production in A. thaliana and N. benthamiana leaves. Overexpression of the constitutively active DGC wspR19 (Goymer et al., 2006) increased cdG levels in the extracts of all three Pseudomonas spp., leading to a strongly suppressed ROS response in each case. This effect was almost entirely dependent on reduced signalling through the FLS2 receptor, a finding consistent with greatly reduced flagellin (FliC) levels seen on cdG overproduction in each case. However, although increased cdG levels led to reduced fliC expression and contributed to the evasion of FLS2-mediated immunity, A. thaliana infection by Pto DC3000 was severely compromised on DGC overexpression. We show that high cdG levels in plant-associated Pseudomonas spp. suppress flagellin expression, which correlates with evasion of the FLS2-mediated immune response. However, any advantage this might confer during plant infection is overwhelmed by the pleiotropic effects of cdG on other virulence-associated phenotypes.

RESULTS
Increased Pseudomonas cdG levels lead to a suppressed ROS response in N. benthamiana and A. thaliana In order to examine the effect of increased cdG levels on the induction of plant immune responses by Pseudomonas spp., we first produced a broad-host-range expression vector for a constitutively active allele of the DGC wspR (pBBR2/5-wspR19), and transformed the three species included in this study (Pto DC3000, P. protegens Pf-5 and P. aeruginosa PAO1). In agreement with previous findings (Goymer et al., 2006), wspR19 expression resulted in aggregative, wrinkled colony morphologies, enhanced Congo Red dye binding, which indicates the production of polysaccharides, such as cellulose, and reduced motility (Fig. S1, see Supporting Information). Overexpression of DGC genes in trans leads to the intracellular accumulation of cdG in various bacterial species (Hengge, 2009). Accordingly, the expression of wspR19 led to increased cdG levels in all three Pseudomonas spp., as determined by liquid chromatography-tandem mass spectrometry (LC-MS/MS) analysis (Table 1). The cdG levels measured for the wild-type (WT) containing the empty vector and for the DGCoverexpressing strains (wspR19) of P. protegens Pf-5 and P. aeruginosa PAO1 were comparable with the levels described previously for PAO1 (Malone et al., 2010). Interestingly, wspR19 expression increased cdG concentration to a much lesser extent in Pto DC3000 compared with the increases seen in P. protegens Pf-5 and P. aeruginosa PAO1.
Next, we investigated the effect of high cdG levels in Pseudomonas on the induction of PTI in different plants by measuring the ROS burst on the application of bacterial extracts. The extracts of Pto DC3000 and P. protegens Pf-5 overexpressing the DGC wspR19 triggered a reduced ROS burst in both A. thaliana and N. benthamiana compared with WT extracts (Fig. 1A,B,D,E), whereas extracts of P. aeruginosa PAO1 express-ing wspR19 only caused a reduced ROS in N. benthamiana (Fig. 1E), but not A. thaliana (Fig. 1C). To test whether the cdG molecule itself is responsible for the suppression of ROS responses, we added chemically synthesized cdG to the Pto DC3000 bacterial extract. No difference in the A. thaliana ROS response was seen compared with the Pto DC3000 WT extract alone. Likewise, we saw no ROS response on addition of 1 μM cdG alone (Fig. 1D). Bacterial extract from a previously characterized P. aeruginosa PAO1 mutant with increased intracellular cdG levels, ΔyfiR (Malone et al., 2010), triggered a similarly reduced ROS response to the PAO1 wspR19-overexpressing strain (Fig. 1E). Thus, we conclude that the intracellular activity of cdG in the bacterium leads to loss of ROS production triggered by the tested bacterial extracts in N. benthamiana, and in A. thaliana for Pto DC3000 and Pf-5 extracts.

FLS2-mediated ROS is specifically reduced by wspR19 expression
The ROS burst can be initiated by the activation of several different PRRs (Macho and Zipfel, 2014). As the bacterial extracts used in our experiments are made from whole-cell lysates, they may contain numerous different MAMPs. Consequently, we tested whether the suppressed ROS bursts seen for the wspR19 extracts depend on one or more specific plant receptors.
To identify the major eliciting agents in our bacterial extracts, we characterized the ROS response produced by Pseudomonas WT extracts on a series of different A. thaliana PRR knock-out mutants. The ROS response of efr, cerk1-2 and lym3-1 plants (CERK1 and LYM3 are involved in bacterial peptidoglycan perception; Willmann et al., 2011) to all tested extracts was comparable with that of Col-0 ( Fig. 2). Conversely, total ROS accumulation was drastically reduced in fls2 and even more so in efr fls2 and bak1-5 bkk1 plant lines (relative to Col-0) after application of WT extracts from Pto DC3000 and P. protegens Pf-5 ( Fig. 2A,B). This suggests that the ROS signal is mainly triggered by flagellin and, to a lesser extent, by EF-Tu for these two extracts. WT P. aeruginosa PAO1 extract triggered equally high ROS production in Col-0, efr and fls2 plants, whereas the ROS response was strongly diminished in efr fls2 and bak1-5 bkk1 (Fig. 2C). Previous work has shown that Data represent the means of two biological replicates with '±' representing the standard error (n = 2). Asterisks indicate statistical significance (**P < 0.01; ***P < 0.001) between wild-type (WT) and wspR19 extracts based on an unpaired Mann-Whitney test.
flg22-or elf18-triggered ROS production is drastically reduced in bak1-5 bkk1 because of the role of BAK1 and BKK1 as co-receptors for FLS2 and EFR (Chinchilla et al., 2007;Heese et al., 2007;Roux et al., 2011;Sun et al., 2013). Furthermore, the signal transduction pathways of both receptors converge into the same signalling pathway, which means that the intensity of FLS2 and EFR activation is not additive and both elicitors can compensate for each other (Macho and Zipfel, 2014). Our results suggest that, under the culture conditions used here, the main elicitor in Pto DC3000 and P. protegens Pf-5 extracts is flagellin, whereas, in the P. aeruginosa PAO1 extract, flagellin and EF-Tu are equally strong elicitors. Next, we repeated the ROS burst experiments for the A. thaliana PRR knock-out mutants with extracts from wspR19-overexpressing strains. Overexpression of wspR19 in Pto DC3000 and P. protegens Pf-5 suppresses the ROS burst elicited by their extracts in A. thaliana Col-0 and the efr knock-out mutant (Fig. 3A,B). Because the major elicitor of ROS production in the extracts of these two species is probably flagellin ( Fig. 2A,B), this suggests that suppression of the ROS response by wspR19 expression is a result of decreased flagellin perception by FLS2. Extracts from a Pto DC3000 ΔfliC strain were tested as a control. Pto DC3000 wspR19 extracts induced slightly higher ROS production in Col-0 and efr plants compared with extracts from the ΔfliC strain, but ROS was equally low in fls2 and efr fls2 (Fig. 3A). The slightly higher ROS production for wspR19 extracts in plants containing the functional FLS2 receptor is probably the consequence of residual flagellin levels in these strains. Pseudomonas aeruginosa PAO1 wspR19-expressing extracts induced equally strong ROS production in Col-0 and fls2, whereas the ROS response was reduced in efr plants, in which ROS production mainly derives from activation of the FLS2 pathway (Fig. 3C). The perception of flagellin and EF-Tu by FLS2 and EFR, respectively, has been extensively studied in A. thaliana (Boller and Felix, 2009); there is no current evidence for other A. thaliana receptors recognizing flagellin or EF-Tu, nor for the existence of additional FLS2 or EFR ligands.As the presence of either flagellin or EF-Tu in the PAO1 extracts seems to be sufficient to trigger the full ROS response in the plant (Fig. 2C), ROS production is only reduced if elements of both recognition systems are simultaneously absent. Thus, the reduction in ROS production on wspR19 expression in P. aeruginosa PAO1 also appears to be FLS2 dependent.
To examine whether the effect of wspR19 expression on ROS accumulation also depends on FLS2 in N. benthamiana, we silenced NbFls2 using virus-induced gene silencing. NbFls2 silencing was confirmed by abolished ROS production on application of 10 nM flg22 peptide (Fig. 3D). ROS accumulation on treatment with WT extracts from all three Pseudomonas spp. was strongly reduced in NbFls2-silenced plants (TRV:NbFls2) compared with mock-silenced control plants (TRV:GFP) (Fig. 3D-F), indicating that, within the bacterial extracts, flagellin is the major elicitor of the ROS burst in N. benthamiana. Nicotiana benthamiana is unable to perceive EF-Tu, as it lacks a functional EFR homologue (Zipfel et al., 2006). The suppressive effect of wspR19 expression is dependent on FLS2, as there was no significant difference in the ROS signal between TRV:NbFls2 and TRV:GFP plants when the wspR19-expressing extracts were Nicotiana benthamiana (D, E) leaf discs after treatment with Pseudomonas extracts, expressed as relative light units (RLU) over 60 min. Leaf discs were treated with extracts from Pseudomonas syringae pv. tomato (Pto) DC3000 (A, D), P. protegens Pf-5 (B, E) or P. aeruginosa PAO1 (C, E) carrying empty pBBR2 vector (WT), pBBR-wspR19 (wspR19), PAO1 ΔyfiR or 1 μM purified cdG. Values are means ± standard error (n = 8). Asterisks indicate statistically significant difference (*P < 0.05; ***P < 0.001) between treatment of wild-type extracts and other samples based on a two-tailed Mann-Whitney test. tested ( Fig. 3D-F). Together, these results indicate that the reduced A. thaliana and N. benthamiana ROS responses induced by extracts of Pseudomonas spp. with increased cdG levels are dependent on signalling through the FLS2 receptor.

Expression of wspR19 leads to impaired accumulation of flagellin
Our data from the ROS assays in A. thaliana and N. benthamiana indicated that wspR19 expression in the three tested Pseudomonas spp. affected the amount of the MAMP flagellin in the extracts, thus reducing the FLS2-induced ROS burst relative to WT. We tested the accumulation of flagellin in the bacterial extracts by Western blot analysis using a commercial anti-FliC antibody. Flagellin accumulation was drastically reduced in Pto DC3000 and P. protegens Pf-5 (Fig. 4A,B) and undetectable in P. aeruginosa PAO1 (Fig. 4C) on wspR19 expression compared with the respective WT extracts. Studies in P. aeruginosa have shown that cdG negatively affects the expression of flagella genes by binding to FleQ, a master regulator of flagella gene transcription (Baraquet and Harwood, 2013;Dasgupta et al., 2003;Hickman and Harwood, 2008). This would explain the reduced accumulation of flagellin in the wspR19-expressing extracts compared with WT strains, and the correspondingly weaker ROS response as a result of reduced FLS2 activation (Fig. S2, see Supporting Information).
To examine whether the down-regulation of flagella synthesis through the elevation of cdG levels is a strategy adopted by Pto DC3000 on plant colonization, we attempted to extract and measure cdG from Pto DC3000 growing in planta. Unfortunately, despite our best efforts and the examination of large amounts of infected plant tissue, we were unable to reliably quantify in planta cdG levels. Although this remains a highly interesting question, it appears that in planta cdG measurements are currently beyond the limit of our technical capabilities.

Virulence of Pto DC3000 during infection is drastically reduced by wspR19 expression
As FLS2-mediated flg22 perception impedes the success of bacterial infection (Forsyth et al., 2010;Hann and Rathjen, 2007;Zeng and He, 2010;Zipfel et al., 2004), we tested the effect of wspR19 expression on the virulence of Pto DC3000 during the infection of A. thaliana plants. The growth of Pto DC3000 wspR19 was drastically reduced after spray infection of A. thaliana Col-0 (Fig. 5). This suggests that high intracellular cdG levels exert a strong negative effect on Pto DC3000 virulence, which, in turn, cancels out any benefit from the evasion of FLS2-triggered immunity. FLS2-induced defences against the virulent Pto DC3000 may only be effective when the flagellin-triggered immune responses are activated in the early stages of the colonization process (Zipfel  al., 2004). In order to dissect the different effects of wspR19 overexpression on plant infection, we syringe infiltrated Pto DC3000 WT and the wspR19-expressing strains directly into the apoplast, bypassing the critical initial colonization steps. Interestingly, the Pto DC3000 and wspR19-overexpressing strains were equally virulent after infection by infiltration (Fig. 5), suggesting

Fig. 4 High cyclic-di-GMP levels suppress FliC accumulation in Pseudomonas.
Extracts from Pseudomonas syringae pv. tomato (Pto) DC3000 (A), P. protegens Pf-5 (B) and P. aeruginosa PAO1 (C) carrying the empty pBBR2 vector (WT) or pBBR-wspR19 (wspR19) and Pto DC3000 ΔfliC were analysed by Western blot using α-flagellin antibody. Membranes were stained with Coomassie Brilliant Blue (CBB) as loading controls. that the infectious disadvantage arising from increased cdG levels is associated exclusively with the initial stages of plant colonization and migration into the apoplast. We also tested the effect of high cdG levels on the virulence of P. aeruginosa PAO1 during infection of A. thaliana. However, P. aeruginosa PAO1 did not proliferate or induce any disease symptoms in the A. thaliana ecotypes Col-0, Wassilewskija (Ws-0) or Llagostera (Ll-0) (data not shown). Previously, plant infections with P. aeruginosa have been described for strain PA14, which contains additional pathogenicity islands that might determine its virulence in A. thaliana (Plotnikova et al., 2000).

DISCUSSION
Plants sense and respond to the presence of pathogenic microbes by the recognition of MAMPs, ubiquitous microbeassociated molecules, using plasma membrane-localized PRRs. MAMP perception triggers an intracellular signal transduction cascade and a subsequent defence response (Boller and Felix, 2009). Flagellin, the central structural unit of the flagellum filament, is a strong immunogen that elicits PTI in plants (Boller and Felix, 2009). Our experiments, measuring ROS production in A. thaliana and N. benthamiana leaves after application of whole-cell extracts, revealed that flagellin from Pto DC3000, P. protegens Pf-5 and P. aeruginosa PAO1 contributes substantially to the induction of this early plant immune response. The flagellum is an important bacterial organelle, enabling the bacterium to sense and explore its environment, and contributing to the initial attachment to surfaces (Rossez et al., 2015). Flagelladriven motility plays a critical role in both effective plant infection and commensal rhizosphere colonization. It facilitates colonization of plant surfaces, migration into the apoplast and movement through the soil towards the nutrient-rich environment of the plant root (Lugtenberg et al., 2001). The role of flagella during plant interactions is therefore somewhat ambiguous. Plantassociated bacteria must balance the requirement for flagelladriven motility in colonization and infection with the downside of potential immune recognition.
To successfully colonize and infect plants, both pathogenic and commensal bacteria need to evade recognition, or to suppress host immune responses. Bacteria have evolved a variety of mechanisms to evade flg22-triggered immunity (Rossez et al., 2015). One important strategy for evading plant immune recognition is the control of flagella synthesis. Bacteria can down-regulate or switch off flagella expression during infection when motility is unnecessary. Although the expression of flagella genes is controlled by a number of different signal inputs, the transcriptional regulator FleQ plays a central role in this process via cdG (Baraquet and Harwood, 2013;Hickman and Harwood, 2008), with FleQ inactivation by cdG binding leading to the downregulation of flagella gene expression in Pseudomonas spp.
( Baraquet and Harwood, 2013). Here, we suggest a direct link between cdG-regulated flagella synthesis and evasion of plant immunity. This relationship is widespread, and applies to commensal as well as specific and non-specific plant pathogens. Elevated cdG levels in Pto DC3000, P. protegens Pf-5 and P. aeruginosa PAO1 reduce extracellular flagellin levels, and thus help the bacteria to evade the FLS2-mediated immune response in N. benthamiana and A. thaliana plants. Reduced flagellin levels appear to explain most, if not all, of the ROS-suppressive effect of cdG overproduction, as little or no additional effect was observed on wspR19 expression in either fls2 or efr fls2 plants. Although flagellin is a major elicitor in the bacterial extracts from all three Pseudomonas spp., EF-Tu appears to be a stronger elicitor in P. aeruginosa PAO1 extracts relative to Pto DC3000 and P. protegens Pf-5. MAMPs from different species generally vary in their eliciting potential (Clarke et al., 2013;Lacombe et al., 2010) as a result of allelic variation in the recognized epitope. The immunogenic elf18 sequences from P. protegens Pf-5 and P. aeruginosa PAO1 are identical, which suggests that the difference in their ability to trigger a ROS burst might be a result of post-translational modifications or protein abundance in the extract.
cdG regulates various phenotypic output pathways that define the bacterial lifestyle, and is consequently likely to play an important role in controlling the association of bacteria with plants. This relationship is indirect, and relates entirely to the control of bacterial phenotypes. In contrast with the mammalian innate immune system in which a specific immune receptor for cdG has been found (Burdette et al., 2011), no direct effect of cdG on plant immunity has been reported to our knowledge, and purified cdG did not affect plant ROS production in our experiments. Recently, researchers have started to examine the role of cdG signalling in the interactions of commensal and pathogenic Pseudomonas spp. with their host plants. The DGCs WspR and Rup4959 play important roles in effective wheat rhizosphere colonization by the commensal species P. fluorescens F113 and P. putida KT2440, respectively (Barahona et al., 2011Matilla et al., 2011). In P. syringae, artificial elevation of cdG levels by overexpression of a DGC gene induces pleiotropic responses, including reduced motility, increased EPS production and enhanced biofilm formation, which together produce aberrant plant interaction phenotypes (Perez-Mendoza et al., 2014). More specifically, deletion of the putative PDE gene bifA results in both decreased motility and reduced virulence in Pto DC3000 (Aragon et al., 2015). Highlighting the relevance of cdG for bacterial virulence, the DGC Chp8 is part of the hrp regulon, a gene cluster containing the T3SS and expressed in response to plant signals in Pto DC3000 (Engl et al., 2014). Overexpression of Chp8 produces high cellular cdG levels, once again leading to EPS production and reduced flagella expression (Engl et al., 2014).
Overexpression of a DGC in trans drastically changed the global cdG concentration in all three bacterial species tested here. In turn, increased cdG levels trigger significant changes in the various signalling networks controlled by this important second messenger (Hengge, 2009). This is borne out by the major shifts in colony morphology, polysaccharide production and motility seen for the wspR19-overexpressing strains. Despite these pleiotropic effects, our research shows that the suppressive effect of cdG on the plant immune response depends almost entirely on the reduced signal transduction through the FLS2 receptor, as a consequence of reduced flagellin production in the cdG-overproducing strains.
Although cdG overproduction enables bacteria to evade FLS2mediated immunity, high levels of cdG actually lead to drastically reduced Pto DC3000 virulence during A. thaliana infection. We propose that the reduction in virulence on wspR19 expression is largely a result of the loss of flagella-driven motility. Restriction of bacterial growth as a result of flagellin recognition in A. thaliana is highly effective against non-adapted pathogens (Forsyth et al., 2010;Li et al., 2005;de Torres et al., 2006;Zeng and He, 2010). Conversely, against adapted pathogens, such as Pto DC3000, FLS2-mediated immunity is only effective when the bacteria are detected in the early stages of infection (Hann and Rathjen, 2007;Zeng and He, 2010;Zipfel et al., 2004). During these initial infection steps, bacteria must trade off the evasion of FLS2-mediated immune responses with the loss of flagellar motility. It has been reported that the virulence of a Pto DC3000 ΔfliC mutant is compromised on spray infection in A. thaliana Col-0, but that the mutant grows as well as WT when the bacteria are syringe infiltrated into the apoplast (Clarke et al., 2013;Li et al., 2005). Likewise, in our experiments, overexpression of wspR19 negatively affected bacterial growth only during spray infections, in which bacteria have to attach to the plant surface and migrate into the apoplast, but had no effect on virulence following leaf infiltration.
In addition to the loss of flagellar motility, other pleiotropic effects of cdG signalling may contribute to the compromised virulence of Pto DC3000 overexpressing wspR19. Thus, infiltration of bacteria directly into the apoplast might also bypass these virulence-associated cdG pathways. These include a reduction in pili-driven motility (Kazmierczak et al., 2006), interference with the correct deployment and function of the T3SS (Kulasakara et al., 2006), overproduction of EPS or other attachment factors in an inappropriate context (Gal et al., 2003) and as yet undefined effects on secondary metabolism and small molecule secretion (Malone et al., 2010). Clearly, the negative effects of high cdG levels are far less severe during an established infection, although whether the lifestyle transition from colonization and initial infection to apoplastic proliferation is accompanied by a significant increase in the intracellular level of cdG in Pto DC3000 remains to be determined. Here, we established a potential role for the bacterial second messenger cdG in plant immune evasion, and showed that this effect is mediated by a reduction in the levels of the MAMP flagellin, and consequently a reduced FLS2-mediated host immune response. However, bacteria cannot simply overproduce cdG whenever they encounter host plants; the complex intracellular signalling networks controlled by cdG play important roles in mediating the initial stages of plant infection, and flagelladriven motility appears to be at least as important to infection as immune system evasion, at least until the infection is established.

Phenotypic tests
Colony morphologies were examined for 5-μL spots of Pseudomonas overnight cultures, after overnight incubation at 28°C on 1.3% agar minimal medium (M9 salts, 0.4% pyruvate) plates containing 0.004% Congo Red dye. Photographs were taken with a Leica M165 FC microscopy system (Leica, Wetzlar, Germany). To measure swimming motility, 0.3% KB agar plates containing the appropriate antibiotics were poured and allowed to set and dry for 1 h in a sterile flow chamber. Plates were inoculated with 2-μL spots of Pseudomonas overnight cultures, and incubated for 2 days at room temperature.

Preparation of bacterial extracts
Bacteria were grown on KB or LB plates (1.3% agar) containing the appropriate antibiotics. Pto DC3000 and P. protegens Pf-5 were grown at 28°C for 21 h and 18 h, respectively, and P. aeruginosa PAO1 at 37°C for 10 h. Bacteria were scraped off the plates and resuspended in sterile demineralized water. Cells were lysed by boiling at 95°C for 15 min with intermediate vortexing and immediately placed on ice. Bacterial extract was obtained by centrifugation of the lysed cells for 10 min at 16 000 g at 4°C, and the supernatant was collected for further analysis. Bacterial extracts were normalized to 5 μg/mL total protein concentration determined using the Bradford assay (BioRad, Hercules, CA, USA).

Bacterial infections
Arabidopsis thaliana Col-0 and fls2 plants were grown for 4-5 weeks. Pto DC3000 cultures were grown overnight at 28°C in liquid KB medium containing 25 μg/mL kanamycin to an optical density at 600 nm (OD600) between 0.6 and 1.0. Bacteria were resuspended in 10 mM MgCl2 and adjusted to OD600 = 0.0002 [10 5 colony forming units (cfu)/mL] for syringe infiltration and to OD600 = 0.05 (2.5 × 10 7 cfu/mL) for spray infection. Shortly before spraying, 0.02% Silwet L-77 (Lehle Seeds, Round Rock, TX, USA) was added to the suspension. Plants were sprayed until run-off and covered during the first day of infection. Four different plants per bacterial strain were infected and two leaf discs (7 mm in diameter) per plant were collected at 2 days post-infection in 10 mM MgCl2 and homogenized using a drill-adapted pestle. Serial dilutions were plated on KB agar containing 25 μg/mL kanamycin and 25 μg/mL nystatin, and colonies were counted after 2 days of incubation at 28°C. Equal inoculations with different bacterial strains were confirmed by serial dilution and colony counting of the initial infection suspensions.

Quantification of cdG concentrations
The extraction and quantification of cdG were performed using highperformance liquid chromatography-coupled tandem mass spectrometry (HPLC-MS/MS) analysis (Spangler et al., 2010). For cdG extraction, bacteria were grown on plates, as described for the production of bacterial cell extracts. Cells were scraped off and resuspended in 300 μL of ice-cold extraction solvent, a mixture of acetonitrile-methanol-water (40 : 40 : 20, v/v/v), and incubated for 10 min at 4°C to extract nucleotides. The cell suspension was heated to 95°C for 10 min, cooled and centrifuged at 13 000 g for 5 min. The supernatant was stored and extraction of the resulting pellet was repeated twice with 200 μL of extraction solvent at 4°C, omitting the heating step. The combined supernatants were evaporated until dryness at 40°C in a miVac vacuum concentrator (Genevac, Ipswich, UK) and the dried residue was resuspended in 200 μL of water. The protein content was determined by dissolving the cell pellet in 200 μL of 0.1 M NaOH, and then heating for 15 min at 95°C, before centrifugation and the measurement of protein concentration in the supernatant using a NanoDrop (Thermo Fisher Scientific, Waltham, MA, USA) at λ = 280 nm. Final cdG concentrations were expressed as pmoles per milligram of bacterial protein. Extractions were performed with two independent bacterial cultures as biological duplicates.

Measurement of ROS burst
The generation of ROS was measured as described previously . Eight leaf discs (diameter, 4 mm) per A. thaliana genotype or N. benthamiana plant were collected into 96-well plates and allowed to recover overnight in sterile water. The water was then replaced with a solution containing 17 mg/mL luminol (Sigma-Aldrich, St. Louis, Missouri, USA), 200 μg/mL horseradish peroxidase (HRP) (Sigma-Aldrich) and either bacterial extracts with 50 ng/mL protein concentration or 1 μM cdG (BioLog, Bremen, Germany) or 10 nM flg22 peptide (EZBiolab, Westfield, IN, USA) with the sequence described previously (Felix et al., 1999). Luminescence was recorded over a 60-min time period using a charge-coupled device camera (Photek Ltd., St Leonards on Sea, East Sussex, UK). Statistically significant differences between ROS outputs were determined by a two-tailed Mann-Whitney test, and experiments were repeated at least twice independently in each case.

Virus-induced gene silencing
Virus-induced gene silencing in N. benthamiana was performed using the Tobacco rattle virus (TRV) system, as described previously (Peart et al., 2002). The TRV-RNA1 construct was contained in a pBINTRA6 vector and the TRV-RNA2 constructs TRV:GFP and TRV:SU were located in a pTV00 vector (Peart et al., 2002). The NbFls2 silencing construct was amplified from N. benthamiana cDNA using the primers 5'-CGACGACAAGACCCTT ACCTTTTTCATACCTTTG and 5'-GAGGAGAAGAGCCCTGGTGGAATATTTCC, and subsequently cloned into the pYY13 vector (Dong et al., 2007). Agrobacterium tumefaciens GV3101 pMp90 containing the binary TRV-RNA1 and TRV-RNA2 constructs was resuspended in infiltration buffer consisting of 10 mM 2-(N-morpholino)ethanesulfonic acid (pH 5.5), 10 mM MgCl2 and 150 mM acetosyringone, and mixed in a 2 : 3 ratio (RNA1 : RNA2) with a final OD600 of 1.0. Two-week-old N. benthamiana plants were infiltrated with the Agrobacterium solution. Infection and systemic spread of the virus were monitored in TRV:SU plants, which were bleached as a result of reduced chlorophyll content on silencing. Silenced plants were used for experiments 3 weeks after infiltration. Successful silencing of NbFls2 was examined in a ROS assay applying 10 nM of flg22 peptide.

Immunoblot analysis
Bacterial extracts were separated on a 12% sodium dodecylsulfate polyacrylamide gel electrophoresis (SDS-PAGE) gel and blotted on polyvinylidene difluoride (PVDF) membrane blocked with 5% milk in TBS with 0.1% [v/v] Tween 20. The flagellin protein was detected with a monoclonal α-flagellin antibody specific to purified flagellin from P. aeruginosa (mabg-flapa, Invivogen, Toulouse, France) used in a 1 : 1000 dilution and α-mouse-HRP (Sigma-Aldrich) used in a 1 : 10 000 dilution as secondary antibody. Visualization was achieved with chemiluminescent ECL pico solution (Thermo Fisher Scientific) and imaged with a LAS 4000 ImageQuant system (GE Healthcare, Little Chalfont, Buckinghamshire, UK). Equal loading of protein was determined by Coomassie Brilliant Blue staining of the blotted membrane.

SUPPORTING INFORMATION
Additional Supporting Information may be found in the online version of this article at the publisher's website: