Transcriptional reprogramming regulated by WRKY18 and WRKY40 facilitates powdery mildew infection of Arabidopsis


(fax +49 221 5062353; e-mail


The two closely related Arabidopsis transcription factors, WRKY18 and WRKY40, play a major and partly redundant role in PAMP-triggered basal defense. We monitored the transcriptional reprogramming induced by the powdery mildew fungus, Golovinomyces orontii, during early stages of infection with respect to the role of WRKY18/40. Expression of >1300 Arabidopsis genes was differentially altered already 8 hours post infection (hpi), indicating rapid pre-penetration signaling between the pathogen and the host. We found that WRKY18/40 negatively affects pre-invasion host defenses and deduced a subset of genes that appear to be under WRKY18/40 control. A mutant lacking the WRKY18/40 repressors executes pathogen-dependent but exaggerated expression of some defense genes leading, for example, to strongly elevated levels of camalexin. This implies that WRKY18/40 act in a feedback repression system controlling basal defense. Moreover, using chromatin immunoprecipitation (ChIP), direct in vivo interactions of WRKY40 to promoter regions containing W box elements of the regulatory gene EDS1, the AP2-type transcription factor gene RRTF1 and to JAZ8, a member of the JA-signaling repressor gene family were demonstrated. Our data support a model in which WRKY18/40 negatively modulate the expression of positive regulators of defense such as CYP71A13, EDS1 and PAD4, but positively modulate the expression of some key JA-signaling genes by partly suppressing the expression of JAZ repressors.


Plants constantly encounter a plethora of phytopathogens with different infection strategies and life styles. Still, plants are able to survive because they are immune to most of these potential pathogens by a process known as non-host resistance (Lipka et al., 2008), and are susceptible to only a few specialized ones. Adapted pathogens evade and/or suppress plants’ basal defense mechanisms and manipulate host cell responses to their benefit. Of special interest are obligate biotrophic pathogenic fungi as they exclusively depend on their hosts for feeding and to complete their life cycles.

Powdery mildew fungi (Ascomycete, Erysiphales) are such a group of obligate parasites, being one of the most important groups of phytopathogens that can infect hundreds of plant species including major crops (Glawe, 2008). They exclusively feed on the epidermal cells by developing appressoria and penetration pegs within hours after spore germination. These invasion structures help to bridge the epidermal cuticle allowing haustoria formation within about 24–48 h. Haustorial complexes act as conduits for the transfer of nutrients and possibly also for fungal effector molecules (Eichmann and Hückelhoven, 2008). The fungus completes its life cycle on the leaf surface through asexual reproduction forming conidiophores within 6–7 days (Figure S1) with successive re-colonization and sporulation marking a successful infection.

How plants respond to powdery mildew invasion and how these pathogens are able to avoid plant detection and/or host defenses to establish a compatible interaction still remain poorly understood. Mutant analyses have identified a limited set of host genes required to establish compatible interactions with powdery mildew pathogens. Loss-of-function mutant alleles of mildew resistance locus O (MLO) encoding seven-transmembrane, calmodulin-binding proteins confer broad-spectrum resistance towards adapted powdery mildews both in barley and in Arabidopsis (Consonni et al., 2006). Six PMR designated loci required for susceptibility towards Golovinomyces cichoracearum and Golovinomyces orontii were isolated, and three identified as encoding a callose synthase (PMR4), a novel protein (PMR5) and a pectate lyase-like protein (PMR6) (Nishimura et al., 2003; Vogel et al., 2002, 2004). Microarray studies provided additional although partly conflicting clues on host genes and signaling pathways that may be important in this type of interaction (Chandran et al., 2009; Fabro et al., 2008; Ramonell et al., 2005; Stein et al., 2006; Zimmerli et al., 2004). Fabro et al. (2008) implicated the jasmonic acid (JA) and salicylic acid (SA) signaling pathways in partly modulating the compatible G. cichoracearum–Arabidopsis interaction: the SA-associated npr1 mutant had reduced number of fungal conidiophores compared with wild-type (WT) plants, whereas a JA response mutant, jar1, had more conidiophores. Expression comparisons between infected WT, npr1 and jar1 plants revealed complex changes in the transcriptome that were partially independent of phytohormone signaling (Fabro et al., 2008). In Arabidopsis pen3 mutants lacking a functional ABC-transporter, SA-associated genes are hyperinduced by adapted powdery mildews and resistance towards these pathogens was shown to be SA dependent (Stein et al., 2006). Finally, a recent detailed transcriptional profiling of host responses towards G. orontii infection defined a set of genes that require SA for induced expression during later stages of fungal growth and reproduction, but whose activation at early stages appears to be SA independent (Chandran et al., 2009). In nearly all cases, knowledge of early events of pathogenesis important for the establishment of fungal infection and on the specific host regulators involved in modulating these complex transcriptional changes remain unknown.

Triggering of plant immune responses during plant–pathogen interactions are generally thought to involve either pattern recognition receptors detecting PAMPs (pathogen-associated molecular patterns) or major resistance (R) proteins, recognizing isolate-specific pathogen effectors (Chisholm et al., 2006). Major host R genes encode predominately two classes of nucleotide-binding site–leucine-rich repeat (NBS–LRR) proteins. One class contains a coiled coil domain at the N-terminus (CC–NBS–LRR), whereas the other class, Toll/interleukin1 receptor–nucleotide binding site–leucine-rich repeat (TIR–NBS–LRR) proteins have an N-terminal domain with similarity to the cytoplasmic domain of Drosophila and human Toll-like receptors (Tan et al., 2007). In barley, the CC–NBS–LRR intracellular R protein mildew A (MLA) interacts with the effector molecule AVR10 and, in conjunction with WRKY transcription factors Hv-WRKY1 and Hv-WRKY2, functions in the nucleus to provide resistance to the barley powdery mildew Blumeria graminis. Virus-induced gene silencing (VIGS) of Hv-WRKY1/2 suggested a repressor function both in basal defense as well as in effector-triggered immunity (Shen et al., 2007). WRKY transcription factors have been implicated in regulating various host defense responses towards phytopathogens (Eulgem and Somssich, 2007; Pandey and Somssich, 2009), and the Arabidopsis members, WRKY18, WRKY40, and WRKY60, showed the highest sequence relatedness to Hv-WRKY1/2. Otherwise susceptible Arabidopsis Col-0 plants, when simultaneously mutated in WRKY18 and WRKY40 were resistant to G. orontii infection, indicating the existence of a similar feedback repression system (Shen et al., 2007). Currently however, the Arabidopsis functional homologue of the barley MLA (R) protein remains unknown. Thus, the level at which wrky18 wrky40 mutant plants gain resistance, and the molecular basis of the feedback repression system remains elusive. Interestingly, wrky18 wrky40 mutant plants do not constitutively express known defense genes (Shen et al., 2007).

In this study we attempted to identify genes/pathways that are directly and indirectly regulated by the Arabidopsis WRKY40 transcription factor in an infection-dependent and infection-independent manner. Based on previous studies we speculated that the powdery mildew pathogen G. orontii can manipulate WRKY18 and WRKY40 transcription factor activities to modulate plant defense responses to its advantage (Shen et al., 2007). We analyzed the temporal dynamics of host gene expression during Arabidopsis-powdery mildew infection in WT and wrky18 wrky40 genotypes. Using whole genome microarrays, we attempt to decipher some of the early transcriptional reprogramming occurring during powdery mildew infection, and further elucidate the dynamics of JA, SA/NPR1, camalexin and EDS1 signaling pathways during the first 48 h of infection. Moreover, we evaluate the contribution of WRKY40 alone in altering the enhanced resistance phenotype of wrky18 wrky40 mutants towards G. orontii and identify direct in vivo targets of this transcription factor.


Loss-of-WRKY18/40 positively affects pre-invasion resistance to G. orontii

Single Arabidopsis wrky18 and wrky40 mutants subjected to G. orontii infection showed no differences in pathogen growth and susceptibility compared with WT Col-0 plants (Figure S3), whereas the wrky18 wrky40 double mutant displayed enhanced resistance indicating functional redundancy between these two genes (Shen et al., 2007). To test whether resistance of the double mutant is a consequence of altered host cell entry at early stages of infection, fungal penetration counts were performed. When wrky18 wrky40 plants were infected with G. orontii spores, only 35% penetration was observed compared with >83% fungal penetration of WT plants (Figure 1a). To determine if one gene was functionally sufficient to restore susceptibility, and to define the contribution of WRKY40 in this process, we over-expressed WRKY40 in the wrky18 wrky40 mutant (Figure 1). Two independently transformed lines (8/6 and 12/2) differing in their WRKY40 protein levels (Figure S8a) were analyzed following G. orontii infection. Both transgenic lines were clearly susceptible showing strong fungal secondary hyphae formation and conidiophore production (Figure 1b). However, although fungal penetration efficiency was clearly enhanced compared with the double mutant (59–65%; Figure 1a), host cell entry rates did not reach WT levels. This indicates that strong pre-invasion resistance requires loss of both WRKY18 and WRKY40 functions but that WRKY40 over-expression alone is sufficient to re-establish substantial susceptibility towards this pathogen.

Figure 1.

 Arabidopsis wrky18 wrky40 mutants show enhanced resistance towards G. orontii due to reduced fungal penetration.
(a) Penetration levels of G. orontii after 48 hpi. Three genotypes, WT (Col-0), wrky18 wrky40, and WRKY40-HA complemented wrky18 wrky40 lines, 8/6 and 12/2, were assayed. WRKY40-HA complementation of the double mutant partially restores susceptibility. Values represent means ± SE from four to five biological replicates, each comprising of three fields.
(b) Phenotypes of Col-0, wrky18 wrky40 and lines 8/6 and 12/2 shown after 10 dpi. **Significantly different at P < 0.01; *Significantly different at P < 0.05.

Temporal expression of WRKY18 and 40 during G. orontii infection

WRKY18 and WRKY40 appear to act as negative regulators of G. orontii resistance in a partially redundant manner. We used quantitative real-time PCR (qPCR) to analyze the inducibility and the temporal kinetics of their expression during the compatible interaction of Arabidopsis with G. orontii (Figure 2). Compared with basal levels detected in untreated control leaves, a > 10-fold increase in WRKY40 transcripts was detected within 8 hour post inoculation (hpi) with elevated levels already observed 4 hpi. Basal WRKY18 transcript levels in untreated leaves were below detection limits, but were detectable within 4 hpi demonstrating induced expression of the gene at this early time point. Compared with transcript levels at 4 hpi, a strong increase of >7-fold was observed for WRKY18 within 8 hpi (Figure 2). The rapid WRKY18 and WRKY40 transcript accumulations upon G. orontii infection corroborate their involvement in regulating some early plant responses. The elevated levels of both transcripts are already detected in host cells at early stages of fungal appressorium formation but prior to the fungal penetration event. WRKY18 and WRKY40 transcription rates were also induced within 4 h post treatments when WT plants were subjected to salicylic acid (SA) and to methyl jasmonate (MeJA), two endogenous phytohormone signaling molecules, and to the PAMP flg22 (data not shown). In wrky18 wrky40 plants no significant level of either transcript was detected during the course of infection confirming the loss-of-function nature of the double mutant (Figure 2).

Figure 2.

WRKY18/40 expression is rapidly and transiently induced upon powdery mildew infection.
Temporal expression of WRKY40 (upper panel) and WRKY18 (lower panel) in WT (solid lines) and in wrky18 wrky40 plants (broken lines) upon pathogen challenge determined by qPCR. At time 0, plants were infected with G. orontii spores, and induced levels were compared with constitutive levels at the time of induction. Increase in transcript accumulation was detected within 4 hpi, and maximum accumulation was recorded 8 hpi. nd, not detectable.

Reprogramming the Arabidopsis transcriptome during early G. orontii infection

We used Arabidopsis whole genome microarrays (ATH1 array) to characterize the dynamic early transcriptional reprogramming events in leaves of WT and wrky18 wrky40 plants during the infection with G. orontii. The first 24 hpi are critical for the establishment of pathogenicity comprising both fungal penetration and haustoria formation (Figure S1). Furthermore, based on our qPCR studies, 8 hpi appears to be a decisive time point in determining WRKY18/40-dependent changes in gene expression. Therefore, we profiled the transcriptomes of both genotypes before (0 hpi-constitutive state) as well as at 8 hpi (penetration stage) of the infection process.

A total of 1351 genes were differentially regulated between WT-uninduced and -8 hpi with G. orontii whereas 1561 genes were differentially regulated between similarly treated wrky18 wrky40 mutant plants (Figure 3a,b; Tables S1 and S2). Nine hundred and twenty genes were common between both genotypes.

Figure 3.

 Early transcriptional reprogramming in WT and wrky18 wrky40 plants upon G. orontii infection.
(a) Venn diagram illustrating total number and overlap of genes affected in WT and wrky18 wrky40 8 h post G. orontii infection.
(b) Changes in the transcript profiles of wrky18 wrky40 (upper panel) and WT (lower panel) plants 8 hpi compared with their respective profiles in the non-infected (0 hpi) states.
(c) Differentially regulated genes between wrky18 wrky40 and WT plants at 0 h (lower panel) and 8 hpi (upper panel). Numbers along solid lines (x-axis) indicate genes up-regulated (dark gray) and down-regulated (light gray).

In WT 8 hpi plants, members of several transcription factor family genes such as MYB, AP-2 type, and bZIP were either up- or down-regulated (Table S1). This differential regulation of transcription factor coding genes also included 10 WRKY members all of which in this case were up-regulated. We checked whether expression of any WRKY members was affected in wrky18 wrky40. Indeed, enhanced transcript accumulations of WRKY11 and WRKY33 were not observed in wrky18 wrky40 plants upon G. orontii infection (data not shown), indicating possible regulation of WRKY11 and WRKY33 by WRKY18 and WRKY40.

To uncover additional WRKY18/40 targets, we compared the expression profiles of WT and wrky18 wrky40 plants (Figure 3c). A total of 167 genes were differentially regulated between the two genotypes under non-infected conditions, whereas 333 genes were differentially regulated 8 hpi with G. orontii, respectively (Tables S3 and S4). Most of the genes affected in the mutant showed up-regulation supporting the negative regulatory function of these WRKY factors (Figure 3c). In particular, the genes encoding an AP2/ERF-type transcription factor designated RRTF1 (At4g34410; Khandelwal et al., 2008), and an oxidoreductase (DIN11; At3g49620), showed highest elevated transcript levels in untreated mutant plants compared with WT, suggesting that they may be under direct negative control by WRKY18/40. Similarly, expression of five additional AP2/ERF-type transcription factors (ERF6, DDF1, ERF13, ATERF-1, ERF11) and five genes encoding members of the JAZ family of JA-signaling repressors (JAZ1, JAZ5, JAZ7, JAZ8, and JAZ10; Katsir et al., 2008) showed clearly elevated levels in the mutant (Tables S1 and S2). The ethylene responsive pathways were also affected during G. orontii infection. Two genes, the calmodulin-like 38 (CML38) and the ethylene response factor11 (ERF11) that are linked to the WRKY40–WRKY33 co-regulatory networks (ATTED-II; were up-regulated four- and three-fold in wrky18 wrky40 constitutively and following pathogen challenge, respectively. Similarly, ERF6 and ERF13 transcript levels were already five- and three-fold elevated in uninfected wrky18 wrky40 plants, whereas ERF1 was three-fold up-regulated in the mutant following infection.

Gene Ontology (GO) terms for 118 and 219 genes of the 167 (0 hpi) and 333 (8 hpi) differentially up- and down-regulated genes (more than two-fold) were obtained and grouped into 20 functional classes (Figures S4). Of particular interest was the observation that transcriptional regulators comprised the largest group of genes (>30%; 32 of 104) showing elevated expression levels in uninfected mutants compared with WT plants indicating that WRKY18/40 function negatively controls these regulators. Other genes having higher transcript levels in 8 h infected mutant than in infected WT plants were CYP71A12 (six-fold), CYP79B2 (2.3-fold), CYP71A13 (2.8-fold) and CYP71B15/PAD3 (4.5-fold) encoding enzymes of camalexin biosynthesis (Rauhut and Glawischnig, 2009), and flavin-dependent monooxygenase 1 (FMO1; 5.1-fold), a positive regulator of the ENHANCED DISEASE SUSCEPTIBILITY1 (EDS1)/PHYTOALEXIN-DEFICIENT4 (PAD4) defense signaling pathway (Bartsch et al., 2006). Genes encoding cell wall related and auxin response functions were among the largest GO groups that were differentially down-regulated in wrky18 wrky40 (Figure S4).

Camalexin biosynthesis and accumulation are affected in wrky18 wrky40

Camalexin accumulation enforces defenses to penetrating pathogens and is the main Arabidopsis phytoalexin induced at infection sites (Rauhut and Glawischnig, 2009). Genes encoding all three enzymes of camalexin biosynthesis were up-regulated in our microarrays. Conversion of indole-3-acetaldoxime (IAOx) to indole-3-acetonitrile (IAN), catalyzed by CYP71A13, marks a committed step (Nafisi et al., 2007), whereas PAD3 catalyzes the final step in camalexin biosynthesis (Schuhegger et al., 2006). We examined expression changes for both of these genes over a time course in WT and wrky18 wrky40 plants by qPCR. Compared with uninfected plants, G. orontii-infected wrky18 wrky40 plants had >15-fold elevated CYP71A13 transcript levels, whereas infected WT plants showed only a four-fold increase. Similarly, the wrky18 wrky40 mutant accumulated overall more PAD3 transcripts than WT plants (Figure 4b).

Figure 4.

 Loss-of-WRKY18/40 function up-regulates the accumulation and biosynthesis of camalexin and the EDS1 signaling pathway upon G. orontii infection.
(a) Camalexin levels were determined in WT (open bars) and wrky18 wrky40 (solid bars) plants before (0 hpi) and at 24 hpi with G. orontii.
(b) Temporal expression of G. orontii-induced host genes CYP71A13 and PAD3 essential for camalexin biosynthesis and (c) of EDS1, PAD4 and FMO1 in WT (solid lines) and wrky18 wrky40 (broken lines) plants as determined by qPCR at the indicated time-points. Samples were collected and gene expression levels were calculated with respect to time 0. **Student’s t-test, n = 10, P < 0.05.

We next examined camalexin levels prior to and 24 hpi with G. orontii. Compared with WT plants, the wrky18 wrky40 mutant already had elevated levels of camalexin in uninfected tissue (Figure 4a). Nevertheless, these levels increased significantly in both genotypes upon infection, with the mutant accumulating 18-fold higher concentrations of camalexin than WT. These findings substantiated our microarray studies and revealed that loss-of-WRKY18 and WRKY40 functions resulted in increased biosynthesis and accumulation of camalexin, which is further strongly enhanced upon G. orontii infection. The pre-existing higher camalexin levels found in uninfected wrky18 wrky40 plants may in part be due to the nearly two-fold elevated transcript levels observed for CYP79B2, CYP71A13 and CYP71B15/PAD3.

EDS1 signaling is up-regulated in wrky18 wrky40 during G. orontii infection

The five-fold elevated FMO1 transcript levels detected in 8 hpi wrky18 wrky40 versus WT plants suggested that the EDS1/PAD4 defense signaling pathway may be positively affected in this mutant. We therefore performed time course studies to monitor altered EDS1, PAD4 and FMO1 expression following G. orontii infection. Within 4 hpi, transcription of EDS1 was >22-fold enhanced in the wrky18 wrky40 mutant, compared with an eight-fold increase in WT plants (Figure 4c). Similarly, transcription of PAD4 encoding the EDS1 interacting partner was also five-fold up-regulated in wrky18 wrky40 plants within 4 hpi, whereas in WT plants maximum PAD4 expression was observed later (8 hpi) and only reached levels that were three-fold higher over uninoculated controls (Figure 4c). Whereas FMO1 acts as a positive regulator of EDS1/PAD4 signaling pathway (Bartsch et al., 2006; Mishina and Zeier, 2006), Nudix hydrolase7 (NUDT7) acts as negative regulator of this pathway (Bartsch et al., 2006). Consistent with our microarray and qPCR findings for EDS1 and PAD4, FMO1 was also clearly up-regulated, with higher transcript levels detectable in wrky18 wrky40 than in WT plants (Figure 4c). NUDT6 and -7 transcription appeared not to be affected upon infection although basal transcript levels of NUDT7 were elevated in the mutant (Figure S5). Thus, wrky18 wrky40 plants accumulate higher levels of transcripts encoding important components of EDS1/PAD4 defense signaling upon pathogen challenge.

WRKY18 and 40 positively regulate JA signaling in response to G. orontii

Phytohormone signaling networks enable plants to coordinate and fine-tune different responses. JA and SA are the two major phytohormone signaling systems that regulate host defense responses (Glazebrook, 2005). Our expression analyses revealed significant up-regulation of five genes encoding repressors of JA signaling in uninfected wrky18 wrky40 mutants. Thus, we examined the expression of two genes, LOX2 and AOS, encoding JA biosynthesis enzymes, along with the expression of JAZ7, JAZ8 and JAZ10 in time course experiments (Figure 5). WT plants showed 10-fold elevated levels of LOX2 and AOS transcripts 4 hpi with G. orontii. In contrast, pathogen-induced expression of both genes was completely absent in wrky18 wrky40 (Figure 5a). As anticipated, transcript levels of all three JAZ genes were clearly elevated in unchallenged wrky18 wrky40 plants (Figure 5b). Four hours post pathogen challenge, expression of all three JAZ genes declined significantly. In the case of JAZ8 and JAZ10, transcript levels continued to decline over the infection period reaching levels similar to those in WT plants at 24 hpi (Figure 5b). In contrast, enhanced pathogen-dependent transient expression of JAZ7 was observed in both genotypes, although more pronounced in WT (Figure 5b). These findings suggest that WRKY18/40 negatively control expression of certain JAZ genes and that failure to induce the JA pathway in wrky18 wrky40 plants may be due to elevated levels of these JAZ repressors. The eight-fold differential increase in RRTF1 transcript detected in untreated mutant plants by microarray analysis was also validated by qPCR (Figure 5a). As with expression of JAZ7, expression levels of RRTF1 declined drastically upon pathogen challenge.

Figure 5.

WRKY18/40 positively regulate JA-signaling but negatively regulate RRTF1.
Temporal expression of the JA-biosynthesis genes LOX2 and AOS and of RRTF1 (a) along with three genes, JAZ7, JAZ8 and JAZ10, encoding negative regulators of JA signaling (b) in WT (solid lines) and wrky18 wrky40 (broken lines) plants upon infection with G. orontii as determined by qPCR at the times indicated. Inducibility of LOX2 and AOS are negated in the wrky18 wrky40 mutant.

The NPR1/SA pathway has been implicated in Arabidopsis susceptibility towards G. cichoracearum (Fabro et al., 2008). Furthermore, NPR1 gene expression is partly under the regulatory control of SA inducible WRKY transcription factors (Yu et al., 2001). Therefore we examined NPR1 expression during Arabidopsis–G. orontii infection in wrky18 wrky40 plants. No differences in NPR1 transcription were observed between the WT and wrky18 wrky40 genotypes (Figure S6). These results are in agreement with the findings of Chandran et al. (2009), showing that the activation of WRKY40 expression during early stages of G. orontii infection is SA independent whereas NPR1 expression requires SA.

In vivo interaction of WRKY40 with target promoters

Our data imply that powdery mildew challenge elicits rapid accumulation of WRKY18 and WRKY40, which negatively modulate the expression of positive regulators of defenses involved in the camalexin and EDS1/PAD4 signaling pathways and negatively modulate the expression of repressors of the JA-signaling pathway. The promoters of the genes involved in these pathways have several W box elements (Figure S7) indicating that such modulation could be partly due to direct interaction with WRKY factors including WRKY18 and WRKY40. To examine if some of the genes are direct targets of WRKY40, in vivo chromatin immunoprecipitation (ChIP) assays were performed using the functional complementation line 8/6 expressing a WRKY40-HA fusion gene driven by the CaMV 35S promoter in the wrky18 wrky40 genotype (Figure 1). Formaldehyde cross-linking was performed on control (0 hpi) and on 4 hpi plant material. Candidate genes were selected based on their clear differential expression profiles (mutant versus WT) in microarray and/or qPCR analyses, and on the presence of W box elements within their respective promoters. These included EDS1, FMO1, PAD3, PAD4, CYP71A13, LOX2, AOS, RRTF1, JAZ8, JAZ10, and At4g16260 (glycosyl hydrolase). ChIP experiments followed by semi-quantitative PCR were performed to narrow down the most promising candidate genes. We observed a clear enrichment for EDS1 and RRTF1 on 4 hpi and for JAZ8 on 0 hpi material, respectively. These genes were selected for more detailed qPCR analyses. Bar charts of the qPCR analyses of the ChIP experiment are illustrated in Figure 6. The normalized amounts of EDS1, RRTF1, and JAZ8 immunoprecipitated promoter fragments, from the 35S:WRKY40-HA 8/6 line, were plotted against normalized amounts of corresponding immunoprecipitated promoter fragments from the wrky18 wrky40 mutant (normalization details see Experimental Procedures). A 7.6- to 18.3-fold increase was observed for the EDS1 promoter fragment (Figure 6a) using two independent biological replicates (ChIP1 + ChIP2), whereas the RRTF1 promoter fragment showed the highest, 9- to 22-fold, enrichment associated with WRKY40-HA (Figure 6b). Specificity of binding at these two target promoters was verified by performing similar ChIP experiments with a transgenic Arabidopsis line ectopically expressing a HA epitope-tagged version of WRKY70. WRKY70 has previously been shown to act as an integrator at the convergence point of SA- and JA-mediated signaling events in plant defense (Li et al., 2004). Little or no enrichment was observed for WRKY70 at these sites (Figure S8). In addition, approximately five-fold enrichment was observed for the JAZ8 promoter fragment containing two proximal W boxes (Figure 6c, Amplicon 7–8). Importantly, no enrichment over background was observed for one additional JAZ8 promoter region harboring distal W box elements (Figure 6c, Amplicon 5–6) demonstrating a degree of WRKY40 binding specificity for distinct W boxes. These data combined with the fact that EDS1, RRTF1 and JAZ8 transcript levels are increased in wrky18 wrky40 compared with WT plants provide evidence that WRKY40 acts as a direct transcriptional repressor of these three genes.

Figure 6.

EDS1, RRTF1 and JAZ8 are direct targets of WRKY40.
The upper panels in (a–c) show schematic representations of the EDS1, RRTF1 and JAZ8 promoter regions containing W box clusters. The two variants of the W box (T/C TGAC T/C, black bar; TGAC T/C, gray bar) are depicted. The numbers indicate W box positions in the respective promoters relative to the ATG start codon (bent arrows). The numbered half-arrows indicate the position of the primer sets used for qPCR. Results for primer pairs 1 and 2 (EDS1, a), 3 and 4 (RRTF1; b) and amplicons 5–6, and 7–8 (JAZ8; c) are shown. Bottom panels in (a–c) represent ChIP assays analyzed by qPCR showing binding of WRKY40-HA to the EDS1, RRTF1 and JAZ8 promoter regions, respectively. Two biological independent ChIP enrichments were performed (ChIP1 and ChIP2) on two sets of plant material at 4 hpi (a + b) and 0 hpi (c). Amplicon primer pair 5–6 for JAZ8 resulted in no enrichments. The amplicon of each promoter region and of each genotype (mutant = wrky18 wrky40, complementation line = 35S:WRKY40-HA) was first normalized to the corresponding Actin (At3g18780) amplicon, followed by a second normalization to the wrky18 wrky40 mutant control line.


We monitored the transcriptional reprogramming induced by G. orontii during early events of fungal establishment and explored the role of two closely related WRKY transcription factors, 18 and 40, in this interaction. We found that WRKY18/40 negatively affect pre-invasion defenses in Arabidopsis and deduced a subset of genes that appear to be under WRKY18/40 control. In particular, we could demonstrate the direct, in vivo physical interaction of WRKY40 to W box containing promoter regions of the key regulatory genes EDS1 and JAZ8, and of the AP2/ERF transcription factor gene, RRTF1.

Previous transcriptomic studies mainly focused on later stages of powdery mildew infection i.e. haustoria formation (18–24 hpi; (Fabro et al., 2008), or pathogen growth and reproduction (1–7 days post inoculation (Chandran et al., 2009). We found that expression of hundreds of host genes are altered within 8 hpi indicating that signaling between the pathogen and individual host cells already occurs a few hours post contact of the spores with the leaf surface at about the stage of penetration, but certainly prior to establishment of the extra-haustorial matrix. It is important to note that G. orontii invades only epidermal cells (Lipka et al., 2008). Hence, local gene expression differences within such cells may actually be higher than the values observed using whole leaf samples. Interestingly, nine out of the 24 genes most highly and differentially expressed (>3-fold) in uninfected wrky18 wrky40 plants encode transcriptional modulators. This implies that WRKY18/40 are central negative regulators controlling distinct downstream transcriptional programs.

The role of phytohormones and distinct signaling pathways during compatible interactions of Arabidopsis and powdery mildews is not firmly established (Chandran et al., 2009; Consonni et al., 2006; Fabro et al., 2008). Although WRKY18/40 expression was observed to be dependent on the SA biosynthetic gene isochorismate synthases1 (ICS1) at later stages of G. orontii infection (1–7 dpi), induction of these genes at the early infection stage (6 hpi) was ICS1 independent (Chandran et al., 2009). Compared with WT, wrky18 wrky40 plants constitutively express 2.4–8-fold higher levels of five co-regulated JAZ family members, namely JAZ1, JAZ5, JAZ7, JAZ8 and JAZ10, and their promoters contain several W box elements (data not shown). JAZ proteins are key repressors of JA signaling (Chico et al., 2008). Although we could not detect in vivo associations of WRKY40 at JAZ7 and JAZ10 promoter sites nor to LOX2 and AOS (data not shown), we did observe specific interaction with a region of the JAZ8 promoter containing two perfect W boxes (Figure 6). Currently only one other transcription factor, MYC2, has been identified as a direct regulator of JAZ promoter activity. However, expression of JAZ7 and -8 does not require MYC2 (Chung et al., 2009). Our studies imply that WRKY40 affects JA signaling by directly controlling the expression of a subset of these negative regulators. Lack of mutants for JAZ7, -8 and -10 currently hamper functional testing of their role in our plant–pathogen interaction. Also, the contribution of WRKY18 in this process remains to be elucidated, although it exerts a positive regulatory function, along with WRKY40, in JA signaling (Wang et al., 2008).

The role of RRTF1 in plant defense remains to be rigorously tested, but recent reports indicate its involvement in regulating redox homeostasis during stress, and expression of RRTF1 is dependent on COI1, a central regulator of JA signaling (Khandelwal et al., 2008; Wang et al., 2008). Moreover, RRTF1 is part of a core redox signaling sub-network that also includes EDS1 and WRKY33 (Khandelwal et al., 2008). Possibly, all three of these genes are under WRKY18/40 control.

WRKY factors form a functionally interconnected transcriptional network (Eulgem and Somssich, 2007), and WRKY18/40 may thus indirectly modulate gene expression by directly controlling other WRKY activities. Indeed, loss-of-WRKY18/40 function appears to limit the pathogen-induced accumulation of WRKY11 and WRKY33 transcripts. PAMP-induced expression of WRKY33 was shown to depend on W box promoter elements that are bound by WRKY factors in planta (Lippok et al., 2007). Compared with WT, wrky18 wrky40 mutants displayed rapid transient induction of CYP71A13 and an overall increased accumulation of PAD3 transcripts. WRKY33 binding to the PAD3 promoter in vivo has been demonstrated, but no binding was detected within the CYP71A13 promoter (Qiu et al., 2008). Our results suggest that CYP71A13 may represent a direct target of WRKY18/40 while PAD3 is an indirect target that is directly controlled by WRKY33.

In general, expression of stress response genes is transient and their transcription relies on a combination of distinct sets of co-activators and negative regulators (Lopez-Maury et al., 2008). It remains to be elucidated how WRKY18/40 negatively regulate transcriptional outputs. In the case of RRTF1 and JAZ8 one can envision a direct repressor function on basal gene expression. However, CYP71A13, EDS1 and PAD4 regulation may be different. Expression of these genes in wrky18 wrky40 is not significantly elevated in the absence of pathogen infection. Rather, loss-of-WRKY18/40 function results in a pathogen-dependent but exaggerated increase in their transcript levels (Figure 4). A simplistic model would be that other WRKY factors (i.e. WRKY60) could substitute to form weaker but sufficient repressor complexes at these promoters. Replacement of WRKY factors at specific W box sites has been demonstrated for two parsley promoters (Turck et al., 2004). Activation of these genes would require an activator complex that is dependent of pathogen-triggered signaling. Elevated transcript levels observed in the mutant may be the consequence of a higher association of the activator(s) to the transcription complex due to failure of the weaker corepressors to counter this effect. This could imply that binding sites for default repressors and sites for signal-induced activators overlap, a situation often found in inducible eukaryotic promoters (Affolter et al., 2008). Naturally, alternative scenarios are conceivable that may also involve altered regulation at the post-transcriptional level.

The consequences of transcriptional reprogramming should be reflected in qualitative/quantitative altered host metabolite levels resulting in resistance of wrky18 wrky40 plants towards G. orontii. Although correlative, our findings that wrky18 wrky40 mutants accumulated high camalexin levels and were resistant to this powdery mildew lends support to the importance of this compound in determining the outcome in this interaction. Proteases and endo-membrane components involved in protein quality control and secretion have been associated with hormone responsive signal transduction events and disease resistance (Beers et al., 2004; Wang et al., 2005). Several protease genes including At3g49340 were more strongly up-regulated in challenged wrky18 wrky40 mutants than in WT plants. Genes encoding components of the endo-membrane system and endoplasmic reticulum (ER) response (e.g. CYP71A12, CYP81F2, wall-associated kinase (At2g34500), BiP3 (At1g09080), and type I phosphodiesterase (At4g29700)) were also differentially regulated. CYP81F2 encodes a P450 monooxygenase essential for pathogen-induced accumulation of antifungal compounds (Bednarek et al., 2009). BiP3 has been associated with the ER response, and its up-regulation in wrky18 wrky40 resistant plants intuitively suggests that in susceptible WT plants, G. orontii may manipulate part of the host ER response system by impinging on WRKY18/40 functions. Defense activation is often preceded by mitogen activated protein kinase (MAPK) signaling (Colcombet and Hirt, 2008). Uninfected wrky18 wrky40 plants constitutively expressed five-fold elevated transcript levels of a mitogen-activated triple kinase gene MAPKKK19 (At5g67080).

Although wrky18 wrky40 lines over-expressing WRKY40 showed a high degree of susceptibility, penetration efficiency did not reach WT levels (Figure 1). Because WRKY18/40/60 can form homo- and heterodimers (Xu et al., 2006), appropriate WRKY18/40 heterodimerization may be partly required for maximal transcriptional reprogramming. Alternatively or additionally, WRKY18 and 40 may target different subsets of genes. Due to extreme difficulties in generating plants expressing a WRKY40 transgene under its native promoter, interpretation of our current results rely on ectopic expression of this gene. Despite this caveat, these lines show no obvious phenotypic differences to WT plants and do not constitutively express defense genes. Moreover, as illustrated by ChIP, overexpressed WRKY40 shows binding selectivity towards some promoters and even at such sites only to a subset of the overall number of W box elements present (Figure 6).

We hypothesize that during the early course of G. orontii infection, the JA signaling pathway is required for plant susceptibility, whereas the EDS1 signaling pathway is needed to maintain resistance. WRKY18/40 negatively modulate the expression of positive regulators of defense such as EDS1 and the camalexin biosynthetic genes, but positively enhance JA-signaling by partly suppressing the expression of genes encoding negative regulators of this pathway (Figure S9). This enables the pathogen to overcome early host-defenses and ultimately leads to susceptibility. Alternatively, WRKY18/40 regulate induced expression of genes whose products are required for pathogen growth and reproduction. However, very few such host susceptibility factors enhancing powdery mildew pathogenesis have been identified (Eichmann and Hückelhoven, 2008). Moreover, despite the fact that constitutive activation of JA and SA signaling can result in enhanced resistance towards powdery mildew infection, comprehensive studies employing a range of SA, JA and ET signaling mutants failed to reveal a clear-cut role of these pathways in G. orontii resistance. Thus, we favor a model in which G. orontii, in part via the regulatory action of WRKY18/40, temporally alters the balance between SA and JA signaling at early stages of infections, keeping host defenses suppressed and thereby allowing the pathogen to gain access to host nutrients. Future investigations will aim at functionally dissecting the WRKY40/18-dependent pathways and the role of key WRKY40 downstream targets in modulating host plant defenses. Such studies will include extensive evaluation of phytohormone dynamics (especially of JA and SA), analysis of selected mutants, and the generation of transgenic lines modified in the expression of multiple key components identified in the current study.

In conclusion, our study of G. orontii-Arabidopsis interaction allowed us to monitor some of the early events of pathogenesis, and to home in on the role of WRKY18/40 transcription factors as negative modulators of pathogen-triggered endogenous signaling and plant defense responses. These studies allowed us to identify key early-response in vivo target genes of WRKY40, but more importantly, provide us with the needed tools to identify direct WRKY40 targets on a global scale thereby helping to elucidate the role of this transcription factor in modulating host responses during powdery mildew infection.

Experimental Procedures

Plant material

All experiments were performed with Arabidopsis Columbia-0 wild type (WT) plants, or mutants in Col-0 background. Single wrky18 and wrky40 mutants were from GABI-Kat (line 328G03; wrky18-1) and SLAT collection of dSpm insertion lines, respectively, as described earlier (Shen et al., 2007; Tissier et al., 1999). T-DNA insertions were within introns 2 and 1 of the WRKY18 and WRKY40 genes, respectively (Figure S2). Double mutants (wrky18 wrky40) were obtained by crossing homozygous single mutant lines (Shen et al., 2007; Xu et al., 2006). Seeds were germinated on Jiffy pots (Alwaysgrows, to prevent patho-infection from garden soil, and plants were maintained under short-day conditions in sterilized closed cabinets (Sneijder-chambers, All the experiments were performed on approximately 5-week-old plants.

Plant pathogen treatment

Virulent Arabidopsis powdery mildew pathogen, Golovinomyces orontii, was maintained on Col-0 WT Arabidopsis plants at 20°C, 16 h light/8 h darkness, 80% relative humidity. Conidiospores were inoculated on WT or indicated wrky mutants and penetration counts were scored 48 hpi as described earlier (Consonni et al., 2006).

Over-expression of WRKY40

The Arabidopsis WRKY40 full-length cDNA was cloned into the Gateway® ( compatible binary vector pAM-Kan-2×35S-intronWK33:GW-HA (GenBank accession no. AY027531). The construct was transformed into A. tumefaciens strain GV3101-pMP90 (Koncz and Schell, 1986). Agro-mediated transformation into wrky18 wrky40 plants was performed as described (Logemann et al., 2006). Transformants were selected on kanamycin and expression of WRKY40 was confirmed by qPCR and by western-blot analyses using an anti-HA antibody. T2, T3 and T4 plants were used for further analysis.

Microarray analysis

Microarray analysis was performed using Arabidopsis ATH1 GeneChip arrays (Affymetrix, Samples were collected from WT and wrky18 wrky40 plants at 0, and 8 hpi with G. orontii. Three plant samples were pooled and hybridized to a single array (one biological replicate; one data point). For each data point, three independent biological replicates (in total nine plants) were collected and separately hybridized for both genotypes (WT and wrky18wrky40). In total 12 chips were hybridized and processed at Integrated Functional Genomics, IZKF, University of Münster (Germany). Data were analyzed using GeneSpring GX (Agilent Technologies, Differential expression of genes was calculated using a combination of anova (Benjamini–Hochberg multiple testing correction) at P ≤ 0.05 and a ≥ 2-fold cut-off. Differential regulation of selected genes was verified in independent experiments using quantitative real-time PCR (qPCR) assays. All samples have been uploaded to ArrayExpress, accession E-MEXP-2371.

Quantitative real-time PCR (qPCR)

Transcript analyses in time course experiments were performed by qPCR. 100 mg tissue, from three to four biological replicate plants, was harvested from untreated (0 hpi) or G. orontii treated (4, 8, 24 and 48 hpi) Col-0 WT and wrky18 wrky40 plants. Total RNA was extracted using TRI reagent (Ambion, and reverse transcribed to produce first-strand cDNA with ‘SuperScript first-strand synthesis system for reverse-transcription PCR’ with oligo (dT) primers, following the manufacturer’s protocol (Invitrogen, SYBR green assays were developed using ‘Brilliant SYBR Green qPCR core reagent kit’ (Stratagene, with gene-specific primers (Table S5) and following the manufacturer’s protocol. PCR was performed on ‘iQ5 multicolor real-time PCR detection system’ (Bio-Rad, All qPCR assays were performed with cDNA corresponding to 100 ng RNA before reverse transcription. To simplify data interpretation, expression levels in control plants (0 hpi) were fixed to 1 and relative values were calculated. The Actin gene (At3g18780) served as internal control. For qPCRs after chromatin immunoprecipitation (ChIP), data analysis similar to published works (Fode et al., 2008; Gonzalez-Lamothe et al., 2008) was adapted, where the amplicon levels in the wrky18 wrky40 mutant were calibrated to 1 and relative fold-enrichment in the WRKY40-HA complemented lines was calculated. Standard curves were prepared to monitor amplification efficiency; Actin served as internal reference for the first normalization; and the 2−ΔΔCT method was used for data analysis, both in the time course and in the ChIP assays. Briefly, the Ct (threshold cycle) values of target genes obtained for the immunoprecipitated samples from the WRKY40-HA complemented line were normalized to the endogenous reference gene actin (ΔCt = Cttarget − Ctreference) and compared with those values obtained from the calibrator line wrky18 wrky40 (ΔΔCT = ΔCtsample − ΔCtcalibrator).

Camalexin quantification

One hundred milligrams of tissue was harvested from untreated or 24 h G. orontii-treated Col-0 WT and wrky18 wrky40 plants. Tissue samples were homogenized in DMSO (2.5 μl mg−1 FW), centrifuged for 20 min at 21 000 g at 4°C, supernatant was collected in a fresh tube and subjected to HPLC analysis. A standard curve was generated using a dilution series of camalexin and the amounts were quantified. Five biological replicate plants were used in every treatment group.

Chromatin immunoprecipitation (ChIP)

Plants were grown for 5–6 weeks and infected with G. orontii as described (Consonni et al., 2006). Seeds of plants expressing a WRKY70-HA construct in the Gateway® compatible binary vector pAM-Kan-2×35S-intronWK33:GW were kindly provided by Dr. R. Birkenbihl (MPIZ Köln). Four hours post inoculation and 0 hpi the leaf material was harvested and fixed in 1× PBS with 0.02% Triton, 0.02% Tween 20 and 1% formaldehyde (fixing solution) for 30 min at room temperature. Leaf material (about 10 plants) was placed in a 250 ml beaker with 100 ml of fixing solution and vacuum infiltrated two to three times for 5 min within the 30 min fixation time period until the material became translucent. Glycine was added to the final concentration of 0.125 m to quench excess formaldehyde.

The fixed material was stored at −80°C until subsequent extraction. The method described by Gendrel et al. (2005) was used for soluble cross-linked chromatin extraction with following modifications. Phosphatase inhibitor cocktail-1 (Sigma, was added to the nuclei extraction and nuclei lysis buffers. The cross-linked chromatin was sheared using a Bioruptor™ sonicator ( for a total 8 min with 30-sec continuous pulses and 90-sec interruption periods, with instrumentation settings at medium power. The degree of chromatin shearing was checked on 0.7% agarose gels. Chromatin fragments were immunoprecipitated using magnetic beads as previously described (Knight et al., 2003). Dynabeads M-280 pre-coated with sheep anti-rabbit IgG (Invitrogen) were incubated overnight with rabbit polyclonal antibodies to HA (ChIP grade, ABCAM, at 4°C in phosphate-buffered saline containing 0.1% BSA fraction V (Sigma). After washing, the beads were incubated overnight at 4°C with 25 μg of soluble cross-linked chromatin in ChIP dilution buffer (Gendrel et al., 2005) containing protein inhibitor (ROCHE, and phosphatase inhibitor (Sigma) on a nutator. The beads were resuspended in 250 μl TE buffer pH 8.0 containing 0.5% SDS with 2 μg of RNase (ROCHE), 2 μg of proteinase K (Invitrogen) and incubated overnight at 65°C to reverse cross-links. DNA was extracted using phenol–chloroform and precipitated with ethanol. The ChIP fragments were resuspended in 50 μl TE pH 8.0. One microlitre was used for qPCR.


We thank Pawel Bednarek for the help with the HPLC, Rainer Birkenbihl for providing seeds of the 35S:WRKY70-HA transgenic line and for critical comments to the manuscript. This work was funded by the Max Planck Society, in part by an International Max Planck Research School fellowship to M.S.