The synthetic elicitor 3,5-dichloroanthranilic acid induces NPR1-dependent and NPR1-independent mechanisms of disease resistance in Arabidopsis.

Immune responses of Arabidopsis (Arabidopsis thaliana) are at least partially mediated by coordinated transcriptional up-regulation of plant defense genes, such as the Late/sustained Up-regulation in Response to Hyaloperonospora parasitica (LURP) cluster. We found a defined region in the promoter of the LURP member CaBP22 to be important for this response. Using a CaBP22 promoter-reporter fusion, we have established a robust and specific high-throughput screening system for synthetic defense elicitors that can be used to trigger defined subsets of plant immune responses. Screening a collection of 42,000 diversity-oriented molecules, we identified 114 candidate LURP inducers. One representative, 3,5-dichloroanthranilic acid (DCA), efficiently induced defense reactions to the phytopathogens H. parasitica and Pseudomonas syringae. In contrast to known salicylic acid analogs, such as 2,6-dichloroisonicotinic acid (INA), which exhibit a long-lasting defense-inducing activity and are fully dependent on the transcriptional cofactor NPR1 (for Nonexpresser of Pathogenesis-Related genes1), DCA acts transiently and is only partially dependent on NPR1. Microarray analyses revealed a cluster of 142 DCA- and INA-responsive genes that show a pattern of differential expression coinciding with the kinetics of DCA-mediated disease resistance. These ACID genes (for Associated with Chemically Induced Defense) constitute a core gene set associated with chemically induced disease resistance, many of which appear to encode components of the natural immune system of Arabidopsis.

Plants utilize an abundance of mechanisms to defend themselves against invading pathogens (Chisholm et al., 2006). These defenses can be induced or constitutive (Somssich and Halbrock, 1998;Dangl and Jones, 2001). Induced defenses are often triggered by the recognition of conserved pathogen-associated molecular patterns (PAMPs), resulting in PAMP-triggered immunity (PTI; Gomez-Gomez and Boller, 2002). To counteract PTI, pathogens evolved effector molecules that can attenuate PTI, making the plant susceptible and the pathogen virulent (compatible interaction; Chisholm et al., 2006). An additional form of induced defense is triggered in response to recognition of pathogen effectors by Leu-rich repeat-containing plant resistance (R) proteins, making the pathogen avirulent and the plant resistant (incompatible interaction). This type of plant immunity is called effector-triggered immunity (ETI) or R-mediated resistance (Jones and Dangl, 2006). PAMP and effector recognition rapidly induce several well-characterized biochemical changes in the plant. These early defense features involve the production of signaling molecules, including reactive oxygen intermediates (ROIs; oxidative burst), nitric oxide, and salicylic acid (SA; Malamy and Klessig, 1992;Jabs et al., 1997;Dangl, 1998;Delledonne et al., 2002). Specifically associated with ETI is the hypersensitive response (HR), a programmed plant cell death localized to infection sites that in many cases effectively restricts pathogen spread and growth (Goodman and Novacky, 1994;Dangl et al., 1996). SA has been shown to be a key player in ETI and PTI (Malamy and Klessig, 1992;Klessig et al., 2000;Tsuda et al., 2008). Accumulation of SA is preceded by the oxidative burst, leading to downstream defense responses and potentiation of further ROI production (Shirasu et al., 1997). In many cases, plants deficient in their ability to accumulate or produce SA are unable to mount successful defense responses (Gaffney et al., 1993;Delaney et al., 1994). SA signaling is partially dependent on NPR1 (for Nonexpresser of Pathogenesis-Related genes1), a transcriptional cofactor that is required for the activation of multiple defense genes (Dong, 2004).
Plant defense responses can be induced by both biotic and abiotic stimuli, such as chemical elicitors. A chemical can be considered a defense activator if it triggers resistance to pathogens while inducing the same or similar molecular markers as biotic defense stimuli. In addition, such a compound should not be directly toxic to the pathogen (Kessmann et al., 1994). Exogenous application of chemicals such as SA, 2,6dichloroisonicotinic acid (INA), and acibenzolar-S-methyl benzo(1,2,3)thiadiazole-7-carbothioic acid S-methyl ester (BTH) has been shown to activate the plant's natural immune responses (Métraux et al., 1991;Ward et al., 1991;Uknes et al., 1992;Schob et al., 1997). Public concern over the dangers of pesticide use has spawned considerable interest in alternative methods for pest control. Many chemical pesticides currently in use rely on direct antibiotic activity, which often leads to undesirable toxic environmental side effects (Kessmann et al., 1994). Compounds that elicit a plant's innate immune response offer an attractive alternative to the application of toxic pesticides for disease-control regimes Uknes et al., 1992).
INA and BTH were discovered in screens for chemical inducers of long-lasting broad-spectrum disease resistance in cucumber (Cucumis sativus; Métraux et al., 1991;Gö rlach et al., 1996). They have the qualities of efficient defense activators as they induce defense responses in a wide variety of plant species and do not exhibit any direct antimicrobial activity. Both compounds are considered functional analogs of SA because they induce the expression of known SAresponsive genes. Notably, INA and BTH are active in nahG plants, which are unable to accumulate SA, showing that they act independently of SA perception and biosynthesis (Kessmann et al., 1993;Friedrich et al., 1996) and suggesting that they interfere with biological targets operating downstream from these steps. While INA has never been used commercially, BTH, under the names Actigard and Bion, is commercially available from Syngenta. Besides their potential use for pest control, such synthetic defense elicitors can serve as versatile tools for chemical genetic analyses of plant defense mechanisms.
Previously, microarray analyses identified the LURP cluster, a set of Arabidopsis (Arabidopsis thaliana) genes that exhibit coordinated late/sustained up-regulation in response to the pathogenic oomycete Hyaloperonospora parasitica (Hp; Eulgem et al., 2004). A defining feature of these genes is a strong accumulation of their transcripts between 12 and 48 h after infection, while transcript levels of a second Hp-responsive gene cluster we identified predominantly accumulated within the first 12 h after infection (Evrard et al., 2009). The Hp inducibility of LURP expression was found to be independent from NPR1 (Eulgem et al., 2004). Genetic studies have functionally implicated several genes from this set in plant defense responses (Zhou et al., 1998;Knoth et al., 2007;Knoth and Eulgem, 2008). We found one member of this cluster, which encodes the WRKY70 transcription factor, to participate in the regulation of other LURPs (Knoth et al., 2007). WRKY70 appears to act downstream from ROI and SA signaling.
Another member of this cluster, CaBP22, closely matches the average Hp-induced LURP expression profile (Eulgem et al., 2004). CaBP22 (At2g41090; also known as CML10) encodes a putative calmodulin-like calcium-binding protein (McCormack et al., 2005). Biological roles of this protein have not been described. Here, we report on the analysis of the CaBP22 promoter and its use in screens for new synthetic defense elicitors. We have identified two Hp-responsive regions in the CaBP22 promoter that contain LURP A , a protein-binding motif. Using an Hpresponsive CaBP22 promoter fragment fused to GUS (uidA), we screened chemical libraries for compounds that induce the expression of this reporter gene. We identified 114 candidate elicitors and designed a set of assays to characterize their modes of action. One of these synthetic elicitors, 3,5-dichloroanthranilic acid (DCA), was shown to rapidly and transiently induce resistance to two phytopathogens by simultaneously activating two distinct branches of the plant defense signaling network. Although structurally related, we found DCA and the SA analog INA to be functionally distinct with regard to their dependency on NPR1 and the kinetics of their activities. Microarray analyses revealed a cluster of genes showing transcriptional changes strictly associated with disease resistance mediated by DCA and INA.

Characterization of the Hp-Responsive CaBP22 Promoter by 5# Deletion Analysis
To locate Hp-responsive regions in the CaBP22 promoter, we performed a 5# deletion analysis using transgenic Arabidopsis lines containing promoter fragments ranging from 65 to 1,075 bp upstream of the CaBP22 transcriptional start site translationally fused to the Escherichia coli GUS reporter gene (lines CaBP22 21075 , CaBP22 2590 , CaBP22 2333 , and CaBP22 265 ; Fig. 1A). Figure 1 (B and C) shows GUS expression responses driven by each promoter deletion visualized by histochemical staining. In lines containing CaBP22 21075 to CaBP22 2333 , GUS expression is clearly visible 7 d postinfection (dpi), with avirulent HpEmoy2 and virulent HpNoco2, and is greatly reduced (or absent) in response to mock treatment ( Fig.  1B; data not shown). There was no observable induction by HpEmoy2 or mock treatment in the CaBP22 265 lines (Fig. 1, B and C).
GUS expression in the CaBP22 21075 to CaBP22 2333 lines was significantly inducible by HpEmoy2 in quantitative 4-methyl-umbelliferyl-b-D-glucuronide fluorescence assays, while no induction was observed for the CaBP22 265 lines (Fig. 1, D and E). Notably CaBP22 2333 lines exhibited an average induction of 3-fold, which is about half the induction produced in CaBP22 21075 and CaBP22 2590 lines (6-to 7-fold). These data show that the promoter region contained in the CaBP22 2333 construct is sufficient to mediate Hp-induced GUS expression. Furthermore, there appear to be two Hp-responsive regions located in the CaBP22 promoter operating in an additive fashion. One of these Hp-response regions must be located between positions 2590 and 2333 and one between positions 2333 and 265.
Nuclear Proteins Interact with a Novel Motif in the CaBP22 Promoter Our reporter assays defined a minimal Hp-responsive region of 268 bp within the CaBP22 promoter (between positions 2333 and 265). While this region does not contain any known defense-associated ciselements, it has a 25-bp stretch consisting of two inversely repeated sequences (5#-ATTGTTTTCTTCT-GTAGAAGACCAT-3#) that is strictly conserved in the second Hp-responsive region between positions 2590 and 2333. We termed this conserved region LURP A and the inversely repeated half sites consisting of the core motif AGAAGA LURP A -CM (underlined in the 25-bp sequence shown above). This hexamer is statistically moderately enriched among promoters of the LURP cluster (as defined as cluster II by Eulgem et al. [2004]; P = 3.3 3 10 23 ).
To determine if LURP A can interact with nuclear proteins, we performed electrophoretic mobility shift assays (EMSAs) with nuclear protein extracts from Arabidopsis seedlings that were left untreated or harvested 48 h after defense induction, a time point that coincides with high LURP transcript levels (Eulgem et al., 2004). A probe containing the full 25-bp LURP A sequence (LURP A -WT) led to a distinct constitutive shift, the intensity of which was clearly enhanced after infection with HpEmoy2 or treatment with 1 mM SA (Fig. 2, A-C). This interaction was successfully competed by 100-fold excess of unlabeled LURP A -WT probe. Additional EMSAs were conducted with several mutated probes to further delineate LURP A 's key protein-interacting regions. LURP A -M2 contains block mutations in both LURP A half sites (LURP A -HS); while LURP A -M1 and LURP A -M3 are mutated in only the Figure 1. Analysis of CaBP22-promoter::GUS 5# deletion constructs in Arabidopsis seedlings. A, Diagram of CaBP22-promoter::GUS 5# deletion series. Numbers indicate 5# promoter end points relative to the CaBP22 transcriptional start site. B, Histochemical GUS staining of 3-week-old Arabidopsis seedlings transformed with each CaBP22promoter::GUS 5# deletion construct after spray inoculation with avirulent HpEmoy2 (7 dpi with 5 3 10 4 spores mL 21 ) or water (7 dpi). Each well contains five to 10 seedlings from independent transformation events. Shown are representative examples from eight independent transformation events per construct. C, Close-up views of GUS-stained CaBP22-promoter::GUS constructs at 7 dpi with HpEmoy2. Ten inde-pendent lines per construct were analyzed. Shown are examples representing the typical behavior for each construct for at least five independent experiments. D, Fluorometric analysis of 10-d-old T2 transgenic CaBP22-promoter::GUS Arabidopsis seedlings. Light gray bars represent mock treatment (4 dpi), and dark gray bars represent HpEmoy2 treatment (4 dpi with 5 3 10 4 spores mL 21 ). Measurements were taken for eight independent lines for each construct. Shown are combined averages from all eight measurements from three independent experiments. 4-MU, 4-Methyl-umbelliferyl-b-D-glucuronide. E, Graphic representation of the fold-induction for each deletion construct. Mann-Whitney U test with a cutoff of P , 0.05 was used to test for significant differences, marked a, b, or c. The error bars in D and E represent SE. first (TCTTCT; LURP A -5#HS) or second (AGAAGA; LURP A -3#HS) LURP A half site, respectively ( Fig. 2A). The Hp-induced LURP A shift was successfully competed by 100-fold excess of unlabeled LURP A -WT and LURP A -M1 probes, while unlabeled LURP A -M2 and LURP A -M3 probes competed the LURP A -mediated shift less efficiently (Fig. 2D). EMSAs performed with labeled LURP A -M1 produced a shift similar to that of the unmutated LURP A -WT probe, while no shift was detected using the labeled LURP A -M2 probe (Fig. 2E). Labeled LURP A -M3 produced a shift of similar size but severely reduced intensity compared with the wildtype probe.
In summary, these data show that nuclear DNAbinding factors interact with LURP A in a sequence-specific manner that is mainly facilitated by the downstream half site LURP A -3#HS. We observed a significant decrease in reporter activity between CaBP22 2590 (which contains two copies of LURP A ) and CaBP22 2333 lines (which contains a single copy of LURP A ) and complete loss of activity in CaBP22 265 lines (which lacks LURP A ). This suggests that a single copy of this inverted repeat is sufficient for Hp-responsive CaBP22 promoter activity, but the additional copy of it may enhance this promoter's responsiveness to defense stimuli. As CaBP22 represents the average expression pattern of the LURP cluster and LURP A -CM is statistically moderately enriched in promoters of this gene set, AGAAGA-containing protein-binding sites or related motifs may contribute to the defense-associated coregulation of LURP members. Consistent with a possible role of AGAAGA-containing promoter sites in controlling LURP expression, derivatives of this motif from two other LURP promoters, ZAT7 and WAK1, successfully competed with LURP A for nuclear DNA-binding factors (Fig. 2F).

A Chemical Screen Reveals a Small Molecule Elicitor of LURP Expression
As shown above, the CaBP22 2333 promoter mediates GUS expression in response to avirulent HpEmoy2. To determine if this promoter is defense specific, we tested several known biotic and abiotic stimuli (Supplemental Table S1) in an Arabidopsis line homozygous for a single insertion site of the CaBP22 2333 construct. This line clearly expressed GUS in response to treatment with the bacterial pathogen Pseudomonas syringae, SA, wounding, and virulent HpNoco2. However, reporter expression was not induced by any of the other stimuli tested (jasmonic acid, ethephon, kinetin, abscisic acid, CaCl 2 , MgCl 2 , NaCl, indole-3-acetic acid, GA 3 , 2,4-dichlorophenoxyacetic acid, or submersion). Due to the apparent specificity of the response pattern of CaBP22 2333 for defense-inducing stimuli, we considered this line an excellent choice for a high-throughput screen for synthetic elicitors. Oneweek-old seedlings of the homozygous CaBP22 2333 line grown in liquid growth medium on 96-well plates were incubated for 24 h with library compounds at final concentrations of 4 to 20 mM followed by histochemical staining to visualize GUS expression (Fig.  3A). Screening a total of 42,000 diverse organic compounds (see "Materials and Methods") identified 114 candidates that reproducibly induced GUS expression in the CaBP22 2333 line. Many of these elicitors are structurally related to SA, while several others do not have any obvious similarity to known defense elicitors.
We found one of these 114 compounds, DCA ( Fig.  3B), which has not been reported as a plant defense inducer, to be particularly active in detailed follow-up analyses. Histochemical staining of the CaBP22 2333 line submerged in varying DCA concentrations (saturation Top, Example of a 96-well screening plate containing 7-d-old liquid-grown CaBP22 2333 seedlings after a 24-h incubation with library compounds at concentrations ranging from 4 to 20 mM. Bottom, Screening plate after GUS histochemical staining. The blue circle indicates the "hit well." B, Structure of DCA. C, GUS histochemical staining of CaBP22 2333 seedlings comparing reporter responses to SA and DCA after a 24-h incubation treatment at the indicated concentrations. D, Doseresponse curve plotted from GUS fluorometric assays measuring DCAinduced GUS activities at the indicated concentrations with proteins extracted from 7-d-old liquid-grown CaBP22 2333 seedlings after a 24-h incubation with DCA-treated medium. Mean and SE values calculated from three independent experiments are shown. 4-MU, 4-Methyl-umbelliferyl-b-D-glucuronide. E, Trypan Blue staining of Col-0 wildtype seedlings incubated for 24 h in DCA-containing medium at the indicated concentrations. Dark blue color indicates cell death (toxicity). All staining experiments were performed at least three times with similar results. treatment) revealed that this elicitor is active at concentrations as low as 0.1 mM, making it significantly more potent than SA, which did not trigger GUS expression at concentrations under 5 mM (Fig. 3C). Quantitative dose-response curves after saturation treatments with DCA revealed a maximal effective concentration of approximately 15 mM. Application of higher concentrations resulted in a sharp decline in GUS activity (Fig. 3D). Its median effective concentration (EC50) was determined to be 6 mM. To examine DCA-induced phytotoxicity, we used trypan blue staining of seedlings after saturation treatment. Dark blue staining, indicating cell death, was prevalent in approximately 50% of the seedlings treated for 24 h with 40 mM DCA and in nearly 100% of the seedlings treated with DCA concentrations of 50 mM or higher (Fig. 3E). However, no cell death was observed at concentrations showing effective reporter activation (5-20 mM), suggesting that cell death was not responsible for GUS reporter activation but may be the cause of the observed decline of GUS activity at higher concentrations. Based on these assays, DCA is clearly a potent inducer of GUS expression in the CaBP22 2333 lines. Consistent with this effect, DCA triggered increased binding of nuclear Arabidopsis proteins to LURP A (Fig. 2C).

DCA Induces Rapid and Transient Resistance to Hp
We further examined if the induction of CaBP22 2333 :: GUS expression by DCA translates to defense activation in soil-grown plants. Figure 4A shows that application of DCA via foliar spray induced reporter activity in 2-week-old soil-grown CaBP22 2333 seedlings at concentrations ranging from 10 to 500 mM. We also pretreated 2-week-old soil-grown ecotype Columbia (Col-0) seedlings by foliar spray with varying concentrations of DCA 24 h prior to challenge with virulent HpNoco2. The extent of Hp spore formation was assayed 7 d after pathogen challenge. Plants . DCA induces disease resistance. A to D, All experiments were conducted with 14-d-old soil-grown CaBP22 2333 or Col-0 seedlings. Mean and SE values were calculated from three independent trials. A, GUS histochemical staining of CaBP22 2333 seedlings 24 h after chemical spray treatment with the indicated concentrations of DCA and SA. B, Col-0 seedlings were sprayed with DCA and SA at the indicated concentrations 24 h prior to spray infection with virulent HpNoco2 (3 3 10 4 spores mL 21 ). Spores were counted at 7 dpi. C, Kinetic analysis of chemically induced disease resistance; Col-0 seedlings were sprayed with 100 mM of each chemical at the indicated times prior to HpNoco2 (3 3 10 4 spores mL 21 ) spray infection. Spores were counted at 7 dpi. D, Quantification of Pst growth by colony-forming units (cfu). Col-0 seedlings were pretreated with 100 mM of the indicated chemicals 24 h prior to dip inoculation with virulent Pst. Bacteria were extracted at day 0 (gray bars) or day 3 (black bars). Significant differences were tested using the Mann-Whitney U statistical test (P , 0.05). fw, Fresh weight. E, Pst grown in liquid culture with 100 mM of the indicated chemicals or 100 mg mL 21 hygromycin (HYG). OD600, which represents the optical density of bacteria at 600 nm, was measured at the indicated times after inoculation. SE values were all less than 0.05 and so are not visible on the graph.
pretreated with DCA at concentrations as low as 10 mM displayed a significant reduction of Hp spore numbers compared with mock-pretreated plants (Fig. 4B). Maximal effects with regard to CaBP22 2333 ::GUS induction and suppression of HpNoco2 spore formation were observed after spray application of 100 mM DCA (Fig.  4, A and B). Spray application of DCA to soil-grown plants at this concentration did not cause any detectable amount of cell death within 8 d (Supplemental Fig. S1). Therefore, for all further experiments, DCA was applied by foliar spray to 2-week-old soil-grown seedlings at a concentration of 100 mM.
To examine the kinetics of DCA-induced HpNoco2 resistance compared with other defense elicitors (SA and INA), plants were pretreated with 100 mM of each compound at specific times ranging from 1 h to 6 d prior to pathogen challenge (Fig. 4C). As noted in the reporter assays, DCA is unmistakably more potent than SA, which was unable to induce full resistance under these conditions at any of the tested time points. Chemical pretreatment with DCA or INA 24 h prior to Hp infection induced full resistance, while mock-or 100 mM SA-pretreated plants showed prolific development of Hp sporangiophores, which contain mature asexual spores. Interestingly, DCA and INA both induced strong resistance as early as 1 h after treatment. However, despite the structural relatedness of INA and DCA (Fig. 5A), INA-induced resistance was long lasting, whereas DCA-induced resistance began to decline between 3 and 6 d after chemical treatment (Fig. 4C). The surprisingly early defense induction triggered already 1 h after DCA treatment coincided with a similarly fast induction of WRKY70 and CaBP22 expression (Supplemental Fig. S2). These data clearly show that DCA is a potent elicitor of Hp resistance and that its activity is both rapid and reversible.

DCA Reduces Growth of the Bacterial Pathogen Pseudomonas syringae in Planta Only
We also tested the ability of DCA to induce resistance to the bacterial pathogen Pseudomonas syringae pv tomato DC3000 (Pst). Plants pretreated with 100 mM DCA and INA 24 h prior to dip inoculation with Pst showed no visible disease symptoms at 5 dpi, while 100 mM SA-and mock-treated plants appeared highly diseased, exhibiting extensive chlorosis and necrosis (data not shown). Quantification of in planta bacterial growth revealed that plants pretreated with DCA showed the greatest reduction in bacterial growth, followed by INA and SA pretreatment, respectively (Fig. 4D). To determine if DCA possesses direct antibacterial properties, we monitored the growth of Pst in liquid medium containing 100 mM DCA, SA, INA, or the antibiotic hygromycin. None of the tested defense inducers reduced bacterial growth at their bioactive concentrations, while hygromycin completely eliminated growth of Pst (Fig. 4E). These data show that DCA, like INA or SA, induces resistance to Pst without exhibiting direct antibiotic activity.

Structure-Activity Analysis
In order to determine the features of DCA that are important for its defense-inducing activity, we analyzed structural analogs of this molecule. Initially, 18 DCA analogs were tested for their ability to induce reporter activity in CaBP22 2333 seedlings using the liquid growth assay established for high-throughput screening (Supplemental Table S2). From these, we selected six compounds that represent a range of activities while maintaining obvious structural similarities to DCA plus INA and SA for detailed analyses (Fig. 5A). Dose-response curves were generated illus- Figure 5. Structure-activity analysis. A, Chemical structures of DCA analogs analyzed. B, Fluorometric analysis of GUS activities induced by DCA analogs using proteins extracted from 2-week-old soil-grown CaBP22 2333 seedlings 48 h after spray treatment with compounds at the indicated concentrations (mM). The mean and SE values were calculated from three independent replicates. 4-mu, 4-Methyl-umbelliferyl-b-Dglucuronide. C, HpNoco2 growth inhibition assay. Two-week-old soilgrown Col-0 seedlings were spray infected with HpNoco2 at 48 h after treatment with varying concentrations (mM) of each DCA analog and then assayed at 7 dpi for disease symptoms (spores). A value of 100% inhibition = 0 spores. The assay was repeated three times with similar results.
trating the activities of each compound in two different types of assays: (1) GUS activity triggered in CaBP22 2333 seedlings using fluorometric 4-methylumbelliferyl-b-D-glucuronide assays (Fig. 5B); and (2) inhibition of Hp spore development during a normally compatible plant-Hp interaction (Fig. 5C). Each analog displayed similar behavior in both assays, and three major activity trends were revealed: strong, moderate, and weak. DCA, like INA, showed strong GUS induction (greater than 7-fold) and nearly 100% inhibition of spore development at low micromolar concentrations. At the opposite end of this spectrum, benzoic acid, anthranilic acid, 3-chlorobenzoic acid, and 5-chloroanthranilic acid were very weak inducers of GUS activity (less than 2-fold) and were unable to mediate full Hp defense at any of the tested concentrations. Both 3,5dichlorbenzoic acid and 3-chloroanthranilic acid mediated a moderate inhibition of spore development (50%-75% at 100 mM) and triggered a medium level of GUS activity similar to that induced by SA (approximately 4-fold). Generally, dichlorinated molecules showed the strongest defense-inducing activity. Anthranilic acid and benzoic acid analogs with single chlorines exhibited strongly reduced or abolished bioactivity in both assays, while their completely dechlorinated derivatives were inactive. Removal of the amino group from DCA significantly reduced its biological activity. Additionally, DCA, 3,5-dicholorbenzoic acid, and 3-chloranthranilic acid were all more potent than SA. In summary the structure-activity analysis showed that no substitutions for DCA were well tolerated, as activity was lowered in all tested DCA analogs. Furthermore, chlorination, particularly in the 3 position, is required for biological activity, and anthranilic acid derivatives were consistently more active than their comparable benzoic acid analogs.

DCA Acts Downstream or Independently of SA Perception and Is Partially Dependent on WRKY70 and NPR1
To determine at what hierarchical level DCA interferes with defense signaling, we used reverse transcription (RT)-PCR to examine transcript levels of two LURPs (CaBP22 and WRKY70) after treatment with 1 mM SA and 100 mM DCA in Col-0 and several wellcharacterized defense signaling mutant backgrounds (Fig. 6A). As anticipated, SA and DCA treatments transcriptionally induced the endogenous LURP genes CaBP22 and WRKY70 in Col-0 seedlings. DCA-and SA-induced LURP expression remained unaltered in the eds1, ndr1, and pad4 mutants known to be blocked upstream from SA (Aarts et al., 1998;Jirage et al., 1999) and in the npr1 mutant, which is compromised in some signaling processes downstream of SA perception (Dong, 2004). The LURP-inducing activity of DCA, unlike that of SA, was also not blocked in nahG plants. Only in the wrky70 mutant was the DCA and SA inducibility of LURP transcript accumulation blocked.
This outcome was not surprising, as WRKY70 was shown previously to affect the expression of CaBP22 and LURP1 (Knoth et al., 2007). These data show that DCA targets a WRKY70-dependent branch of the defense signaling network.
To confirm the effects of some of the tested mutations on DCA activity, we analyzed DCA-mediated resistance to HpNoco2 (Fig. 6B). Col-0 and mutant plants were pretreated with DCA prior to spray infection with HpNoco2. DCA induced strong Hp resistance in Col-0 and nahG plants, nearly fully suppressing the formation of HpNoco2 spores. DCAinduced resistance was significantly compromised in the wrky70 mutant but was not fully abolished. In contrast to the results of the RT-PCR analysis, DCAmediated resistance was weakly reduced in npr1 plants. The npr1-3 and wrky70-3 mutants used appear to be null alleles (Cao et al., 1997;Knoth et al., 2007). Thus, in the absence of NPR1 or WRKY70, DCA is still able to activate defense reactions to a certain extent. This may indicate that NPR1-and WRKY70-dependent mechanisms are partially redundant. Alternatively or additionally, DCA may also induce defense mechanisms that are completely independent from NPR1 and WRKY70.
The fact that the impact of npr1 on DCA-mediated HpNoco2 resistance is only weak was surprising, as the defense-inducing activity of the structurally related INA is known to be fully dependent on NPR1 (Cao et al., 1994;Delaney et al., 1995). In fact, a side-byside comparison confirmed that HpNoco2 resistance mediated by INA is fully blocked in npr1 plants, while resistance mediated by DCA is somewhat reduced in this mutant (Fig. 6C). In the npr1 background, DCA suppressed Hp spore formation to 43% of the level observed in untreated plants, which is approximately 4-fold less than the DCA-mediated suppression of spore levels to 10% in Col-0.
Taken together, these data show that DCA triggers both NPR1-dependent and NPR1-independent defense responses (Fig. 6D). Interaction of DCA with defense signaling pathways is likely to occur either downstream or independently of SA perception/ accumulation and is partially dependent on WRKY70. The partial independence from NPR1 and the transient nature of its defense-inducing activity functionally discriminates DCA from INA.

Microarray Analyses Reveal Transcriptional Changes Associated with Chemically Induced Disease Resistance
We reasoned that DCA-triggered transcriptome changes that follow the temporal pattern of DCAmediated resistance are very likely to be of key importance for a successful pathogen defense. To this end, we hybridized to Affymetrix ATH1 GeneChips RNA from Col-0 seedlings that were treated for 48 h or 6 d with mock solution, DCA, or INA. We examined responses at these two time points, because both DCA and INA efficiently suppressed Hp spore formation between 24 and 72 h after treatment but substantially differed in their efficiency at 6 d after treatment. To further examine the role of NPR1 in DCA-mediated disease resistance, we also analyzed responses at 48 h after DCA or mock treatment in the npr1 mutant. Differentially expressed genes (DEGs) in response to the chemical treatment were identified statistically among three biological replicates using as a cutoff a false discovery rate of less than 0.05. The up-and down-regulated DEGs identified by this analysis are summarized in Figure 7A  Most importantly, our Venn analyses shown in Figure 7A revealed a cluster of 142 DEGs whose patterns of transcriptional changes and disease resistance triggered by DCA or INA match. Hence, they represent transcriptional changes strictly associated with DCA-and INA-induced defense and are likely to be functionally important for these immune responses. These genes showed significantly altered transcript levels 48 h after DCA or INA treatments as well as 6 d after INA treatment but not 6 d after DCA treatment (Fig. 7C). This ACID (for Associated with Chemically Induced Defense) cluster contains 137 genes that were coordinately up-regulated and five genes that were coordinately down-regulated by all three defenseinducing treatments (48 h after DCA or INA treatments and 6 d after INA treatment). Consistent with the partial suppression of DCA-mediated defense responses by NPR1, 20% of the 137 up-regulated ACID genes exhibit NPR1-independent expression, while the remaining genes are expressed in an NPR1-dependent manner. All of the five coordinately down-regulated genes are NPR1 dependent (Fig. 7B). Figure 6. Analysis of DCA activity in known defense mutants. A, RT-PCR analysis of CaBP22 and WRKY70 expression in Col-0 (wild type) and mutant or nahG backgrounds 24 h after spraying 2-week-old seedlings with water (mock), 1 mM SA, or 100 mM DCA. Actin8 (ACT8) is shown as a loading control. At least three biological replicates showed consistent results. B and C, Two-week-old seedlings were sprayed with virulent HpNoco2 at 24 h after pretreatment with water, 100 mM DCA, or 100 mM INA as indicated. Spores were counted 7 d after spray infection. Mean and SE values were calculated from three independent experiments. Mann-Whitney U test (P , 0.05) was used to determine significant differences among the different plant lines. Dark gray bars indicate mock pretreatment, light gray bars indicate 100 mM DCA, and checked bars indicate 100 mM INA treatment. D, Model illustrating how DCA may interfere with defense signaling. DCA targets regulators operating downstream or independently of SA, triggering both NPR1-dependent and NPR1-independent defense responses. The latter branch targets a WRKY70-dependent node of the defense signaling network. Additional DCA-triggered pathways may involve NPR1 and WRKY70 paralogs or defense regulators unrelated to them.
In order to infer possible molecular processes contributing to DCA/INA-mediated disease resistance, we used the FuncAssociate program to identify enriched Gene Ontology (GO) terms (Berriz et al., 2003). This program identified several statistically overrepresented GO terms in the set of 137 up-regulated ACID members (Supplemental Table S9). GO terms representing kinase activity, transferase activities, and calmodulin binding were ranked the highest in this set. Both calmodulin-binding proteins, which sense Ca 2+ fluxes, and protein kinases and are known to be important for plant defense signaling (Zhou et al., 1995;Zhang and Klessig, 2001;Kim et al., 2002). The overrepresentation of these terms within the ACID cluster supports the conclusion that these genes are important for disease resistance, as suggested by their strict correlation with successful pathogen defense.

DISCUSSION
Plants have an intricate immune system that responds to pathogen infections via a complex regulatory network. Synthetic bioactive molecules that interfere or interact with defined signaling mechanisms can serve as powerful tools for the dissection of regulatory networks and complement the use of mutants, natural messenger molecules, or ligands (Kawasumi and Nghiem, 2007). In order to expand the existing repertoire of elicitors that can be used for analyses of the plant defense network, we initiated screens of several diversity-oriented chemical libraries for new inducers of pathogen-responsive reporter genes.
We performed elicitor screens targeting regulatory mechanisms controlling the expression of LURP genes. As CaBP22 represents the average response pattern of the LURP cluster, we first dissected its promoter and Displayed are ratios of transcript levels triggered by the individual chemical treatments compared with the respective mock treatments. Magenta represents up-regulated relative to control, and blue represents down-regulated relative to control; the brightest color indicates a greater than 8-fold differential expression.
identified a Hp-responsive region containing LURP A . This motif appears to be related to the defense-associated TL1 element (CTGAAGAAGAA; Wang et al., 2005), which contains the LURP A core motif (LURP A -CM). Thus, TL1/LURP A -type promoter elements may play a wider role in defense gene regulation.
GUS reporter activity mediated by the CaBP22 2333 region was only observed after a variety of defense stimuli activating the SA signaling cascade or wounding, but not after other treatments. Hence, the CaBP22 2333 line seemed to be a specific and reliable reporter system for our elicitor screens. We observed a low hit rate of 0.3%, and as anticipated, many of the 114 LURP inducers we identified are structurally related to SA. However, we also identified several structurally novel compounds in this screen. One candidate, DCA, was particularly active in detailed follow-up analyses. The EC50 value for GUS expression in CaBP22 2333 seedlings after 24 h of DCA saturation treatment was found to be 6 mM, which is 10-fold lower than the published value for PR1 induction by SA (65 mM; Pillonel, 2001). Low active concentrations are often correlated with high target specificity and a decrease in unwanted side effects (Burdine and Kodadek, 2004). DCA did not exhibit any herbicidal activity at the bioactive concentrations; however, it proved to be phytotoxic at higher concentrations. To our knowledge, there are no reports of defense induction by exogenous application of DCA, and despite some structural similarities, it is chemically distinct from known defense inducers. In addition, DCA represents a substructure repeatedly observed among the other 114 elicitors we identified. It is also readily commercially available. Combined, these results made DCA an interesting candidate for further analysis of its role in defense gene activation.
DCA induced resistance to two phylogenetically distinct pathogens at concentrations much lower than SA without exhibiting direct antibiotic activity. DCA did not induce HR-type cell death or an oxidative burst (data not shown) at biologically active concentrations, suggesting that it acts downstream or independently of these defense responses. The wrky70 mutant was the only tested Arabidopsis line deficient in defense regulation, with suppressed DCA inducibility of CaBP22 expression. Consistent with this, the wrky70 mutant exhibited significantly reduced DCAmediated resistance. In contrast to the results of the HpNoco2 defense assays, LURP expression responses were not affected in the npr1 mutant. To clarify the role of NPR1 in DCA-induced resistance, we performed microarray analyses in the npr1 mutant background 48 h after DCA treatment. Of the 137 DCA-inducible ACID genes, 20% exhibited NPR1-independent transcriptional up-regulation. As anticipated, the NPR1independent ACID subset contains several LURPs, including WRKY70, LURP1, NPR4, and WAK1. Both the NPR1-dependent and NPR1-independent subsets of the ACID cluster contain a multitude of additional genes implicated in defense responses. Based on these observations, we propose that DCA operates downstream or independently of SA, activating a WRKY70/ LURP-dependent branch of the defense signaling network as well as, weakly, a separate NPR1-dependent branch (Fig. 6D). We cannot exclude that additional NPR1-independent and WRKY70-independent signaling routes are activated by DCA. Such alternative pathways may involve paralogs of NPR1 and/or WRKY70.
The defining features of DCA-type elicitors are the presence of the 3-and 5-position chlorines and an amino group at position 2. All of the tested analogs conform to Lipinski's rule of five (Lipinski et al., 1997) and have low polar surface area values, which suggests that they should all be readily absorbed by cells. However, compared with DCA, all of tested analogs showed a reduced defense-inducing activity, suggesting that each tested region of this compound is important for its activity. Although functionally distinct, DCA, SA, and INA have several common structural features. SA and DCA share the backbone structure of a benzoic acid substituted at the 2 position. However, this core structure (anthranilic acid in the case of DCA) is inactive with regard to defense induction. Comparative analyses of structure-activity relationships of DCA and INA turned out to be more complex, because their efficiencies were nearly identical regarding both CaBP22 2333 ::GUS induction and defense activation. DCA and INA share a common structure of a dichlorinated six-member ring with a carboxy group. Their closest common structure, 3,5-dichlorobenzoic acid, also induces CaBP22 2333 ::GUS expression and Hp resistance, albeit at a significantly reduced level. Exchange of a carbon atom by a nitrogen atom at position 4 of the ring converts this molecule to INA, while addition of an amino group to position 2 of the ring results in DCA. Hence, both INA and DCA can formally be considered as two representatives of a continuum of related defense-inducing molecules. However, some key differences between DCA and INA were observed in our defense assays. As discussed above, npr1 only mildly affected DCA-induced resistance. This functionally discriminates DCA from SA, INA, and BTH, as their defense activation is fully blocked in npr1 Lipinski et al., 1997;Knoth et al., 2007). Taken together, these data suggest that despite some structural similarities and a qualitatively related response, DCA seems to shift the balance between NPR1-dependent and NPR1-independent responses toward the NPR1-independent ones.
Besides the ACID set, our microarray analyses defined a second interesting cluster, which comprises genes that are specifically up-regulated by 48 h DCA and but not by INA treatment (173 genes). GO analysis of this set show that it is highly enriched for genes annotated for involvement in defense responses. Strongly overrepresented in this set are genes involved in phosphorylation, phosphate metabolism, phosphotransferases, and phosphokinase activity. Whereas in the set of genes specifically up-regulated by 48 h INA (235 genes), GO analysis does not show any enrichment for phosphorylation-related mechanisms (Supplemental Table S10). This supports the conclusion that DCA and INA may be acting on different (but possibly related) targets, leading, on the one hand, to the activation of defense responses specific for each of these elicitors and, on the other hand, to a set of common defense responses.
In addition, DCA and INA differ substantially in their kinetic behavior. Consistent with previous reports (Uknes et al., 1992), we found INA to induce long-term disease resistance, whereas DCA-induced resistance is transient. This feature of DCA should allow rapid and reversible induction of immune responses at any developmental stage with limited side effects. Permanent defense activation often results in fitness costs, due to the toxicity of defensive products and resource allocation away from growth or reproduction. For example, possibly due to its long-term activity, INA was insufficiently tolerated by some crop plants to warrant practical use as a plant protection compound . Furthermore, the mutants cpr1-1 (for constitutive expresser of PR1-1) and ssi1 (for suppressor of SA-insensitivity1) exhibit constitutive defense responses causing severe dwarf phenotypes (Shah et al., 1999;Jirage et al., 2001). Thus, chemicals that transiently activate plant immunity may be beneficial in combating virulent pathogens that threaten crops only during a limited period of time. A transiently active compound like DCA may allow finetuned control of defense induction coordinated with the plant's needs, thereby decreasing any unwanted side effects caused by long-term defense activation. The distinct kinetic characteristics of DCA and INA may be due to differences of their lifetimes in planta. Alternatively, the two compounds may differ regarding their modes of target interaction. Future studies will have to address their fate in planta, the identification of their biological targets, and details of their interference with these targets.
We defined the ACID cluster as a set of genes strictly associated with defense activation by two separate defense-inducing chemicals, DCA and INA. In addition, 118 of the 142 ACID members also respond to a third synthetic elicitor, BTH (Wang et al., 2006). The ACID cluster contains many known defense-related genes and is highly enriched for genes associated with calmodulin binding and kinase activity. Among the upregulated ACID genes are six genes encoding WRKY transcription factors, which have been associated with plant immune responses (Eulgem and Somssich, 2007). Also, there are 19 genes encoding Leu-rich repeatcontaining proteins represented in this cluster. Conspicuously absent from the overrepresented GO attributes for the ACID set are terms annotated for processes associated with upstream/early defense responses, like the HR, cell death, peroxidases, response to ROIs, and SA biosynthesis genes. The lack of enhancement of these terms is consistent with our observation that DCA targets only a subset of the defense signaling network downstream of these early responses. This can be further illustrated by comparing GO terms from transcriptome changes triggered by other defense stimuli. For example, GO analysis of genes up-regulated by flg22 treatment (Zipfel et al., 2004) contains all of the GO terms found in the ACID cluster. However, the flg22 set also includes terms for HR response, transcriptional regulation, and defense response that were not overrepresented in the ACID cluster. This implies that DCA activates a distinct subset of pathogen-inducible defense responses.
Several LURPs (originally defined as cluster II by Eulgem et al. [2004]) were found to be up-regulated by 48 h DCA or INA but were not included in the ACID cluster. CaBP22 is found in this set, as it is not differentially expressed after the 6 d INA treatment. Consistent with this, cabp22 T-DNA mutants did not display detectable defense-associated phenotypes (data not shown). This shows that while CaBP22, as a WRKY70 target gene, is an excellent marker for defense activation, it is functionally not essential for such processes. Another LURP member that falls into the same category as CaBP22, ECS1, is up-regulated in several plantpathogen interactions (Aufsatz et al., 1998;Eulgem et al., 2004). However, ECS1, like CaBP22, is not vital for effective defense activation (Aufsatz et al., 1998). These observations highlight the usefulness of the ACID cluster for the identification of new components essential for full plant immune responses.
We have developed a specific and reliable highthroughput screening system for synthetic elicitors and identified DCA, a plant defense activator that triggers a defined aspect of the plant defense network. However, its target(s) remain(s) to be determined. Screens for proteins directly targeted by DCA or operating downstream from DCA perception may reveal new components of the plant immune response. The fact that DCA strongly triggers NPR1-independent defense makes such screens very useful and may overcome limitations of previous strategies that often identified npr1 alleles (Cao et al., 1994;Delaney et al., 1995;Shah et al., 1997). We are currently analyzing some other candidates identified in our chemical screen and continue to screen chemical libraries for inducers of CaBP22 2333 ::GUS as well as other pathogenresponsive reporter genes that are not inducible by DCA. We expect to provide a collection of compounds that interact with distinct hierarchical levels of the plant defense signaling network. These synthetic elicitors will be invaluable tools for the fine dissection of defense mechanisms and may lead to the development of novel pesticides tailored to enhance a crop's inherent defense capabilities.

Chemical Treatments
Stock solutions were prepared in DMSO. Stocks were added directly to the growth medium for treatment of liquid-grown plants. Stock solutions were diluted in water and sprayed on soil-grown plants at the indicated times and concentrations with Preval sprayers until imminent runoff. Final DMSO concentrations never exceeded 0.002%. Mock treatment was application of 0.002% DMSO in water. Chemicals were supplied from Sigma. To test for chemically induced disease resistance, the plants were sprayed with chemicals at the indicated concentrations and times prior to pathogen challenge. Disease symptoms were analyzed as described above.

Microarray Preparation and Data Analysis
Total RNA was isolated from seedlings using TRIZOL as outlined above (Invitrogen). RNA was processed and hybridized to the Affymetrix Arabidopsis ATH1 genome array GeneChip following the manufacturer's instructions (Affymetrix) by the University of California at Riverside Core Instrument Facility. Three independent biological replicates were performed for each treatment. Microarray analysis was performed in the statistical programming environment R using Bioconductor packages. Raw expression values were normalized using the robust multichip averaging algorithm. Analysis of DEGs was performed with the LIMMA package (Smyth, 2004). The Benjamini and Hochberg method was selected to adjust P values for multiple testing and to determine false discovery rates (Benjamini et al., 2001).
As a confidence threshold, an adjusted P value of #0.05 was chosen (compared with mock treatments with water). To visualize the DEG sets, hierarchical clustering was performed using the Cluster and TreeView programs (Eisen et al., 1998). Overrepresented GO terms were identified with the FuncAssociate program (Berriz et al., 2003).
The microarray data have been deposited in MIAME-compliant format in the GEO database under the accession number GSE13833.

Supplemental Data
The following materials are available in the online version of this article.
Supplemental Figure S1. Trypan blue staining of 2-week-old soil-grown Col-0 wild-type seedlings 8 d after spray application of mock solution (water), 100 mM DCA, 500 mM DCA, or 2 mM SA; dark blue color indicates cell death (toxicity).
Supplemental Figure S2. DCA induces rapid accumulation of WRKY70 and CaBP22 transcript levels.
Supplemental Table S1. Comparison of the ability of different stimuli to induce GUS reporter expression in the CaBP22 2333 line.
Supplemental Table S2. Comparison of GUS reporter induction for several DCA analogs.
Supplemental Table S7. Genes differentially expressed in npr1 48 h after DCA.
Supplemental Table S8. Members of the ACID cluster.
Supplemental Table S9. Enriched GO terms in the ACID cluster.
Supplemental Table S10. Comparison of enriched GO terms after treatment with two distinct defense activators.