Negative Regulation of Defense Signaling Pathways by the EDR1 Protein Kinase

SUMMARY

The enhanced disease resistance 1 mutant of Arabidopsis confers enhanced resistance to bacterial and fungal pathogens. To better understand how edr1-mediated resistance occurs, we performed transcriptome analyses on wild-type and edr1 plants inoculated with the fungal pathogen, Golovinomyces cichoracearum (powdery mildew). The expression of many known and putative defense-associated genes were more rapidly induced, and to higher levels, in edr1 plants compared to wild type. Many of the genes with elevated expression encode WRKY transcription factors and there is enrichment for their binding sites in promoters of the genes upregulated in edr1. Confocal microscopy of transiently expressed EDR1 protein showed that a significant fraction of EDR1 is localized to the nucleus, suggesting that EDR1 could potentially interact with transcription factors in the nucleus. Analysis of Gene Ontology annotations revealed that genes associated with the endomembrane system, defense, reactive oxygen species production (ROS), and protein kinases are induced early in the edr1 mutant, and that elevated expression of the endomembrane system, defense and ROS-related genes is maintained for at least four days after infection.

INTRODUCTION

Plants have evolved complex mechanisms to defend themselves against pathogens (Bent & Mackey, 2007). To identify genes regulating plant defense responses, we have previously screened for Arabidopsis (Arabidopsis thaliana) mutants with enhanced resistance to virulent pathogens (Frye & Innes, 1998). The enhanced disease resistance 1 (edr1) mutant displays enhanced resistance to the fungus Golovinomyces cichoracearum, an obligate biotroph and causal agent of powdery mildew on Arabidopsis and many cucurbit species (Adam et al., 1999, Adam & Somerville, 1996). While G. cichoracearum forms conidiophores (stalks of asexual spores) on the surface of a susceptible leaf, resistance in edr1 is manifested as necrotic lesions at the site of infection and a reduction in conidiophores (Frye & Innes, 1998). Additionally, edr1 mutants have greater callose deposition and form more papillae and at an earlier time than wild-type Col-0 plants. EDR1 encodes a protein with a C-terminal kinase domain and a putative N-terminal regulatory domain (Frye et al., 2001). A recombinant protein containing just the EDR1 kinase domain is able to autophosphorylate and can phosphorylate the common kinase substrate myelin basic protein in vivo, demonstrating that EDR1 does indeed have kinase activity (Tang & Innes, 2002).

The enhanced resistance of the edr1 mutant is suppressed by mutations that reduce salicylic acid (SA) production (sid2, eds1 and pad4), or block SA perception (npr1/nim1) (Frye et al., 2001, Tang et al., 2005). Transgenic expression of NahG, which lowers endogenous SA levels, also eliminates the edr1-mediated enhanced disease resistance phenotype (Frye et al., 2001). In contrast to the requirement for SA signaling in edr1-mediated resistance, neither ethylene nor jasmonic acid appear to be necessary, as mutations in the ETHYLENE INSENSITIVE 2 (EIN2) gene or CORONATINE INSENSITIVE 1 (COI1) gene do not alter edr1-mediated disease resistance (Frye et al., 2001).

In addition to regulating responses to pathogens, EDR1 also regulates responses to abiotic stresses, such as drought. When grown under drought conditions, the edr1 mutant is dwarfed and forms lesions, while growth is normal under optimal conditions (Tang et al., 2005). These phenotypes are suppressed by mutations in the SA signaling pathway (eds1, pad4, or npr1), indicating that the drought response is also SA-dependent and may share similarity to the pathogen response in edr1. Additionally, the F-box protein mutant, ore9, which has delayed senescence in response to ethylene, restores wild-type growth under drought conditions to the edr1 mutant, but does not abate the drought-induced lesion phenotype (Tang et al., 2005).

EDR1 is most similar to the ethylene response regulator, CTR1, and four other proteins of unknown function in Arabidopsis. Despite this similarity, edr1 mutants have a normal triple response, unlike ctr1 mutants. However, when edr1 mutants are treated with ethylene, they senesce more rapidly than wild-type Col-0 (Frye et al., 2001). This response can be abolished by the presence of ein2, an ethylene signaling mutation, but it does not require SA responses (Tang et al., 2005). Taken together, the responses to pathogen, drought, and ethylene in edr1 imply that EDR1 is negatively regulating cell death in response to various stimuli.

A mechanism for CTR1-mediated ethylene regulation proposes that two F-box proteins, EBF1 and EBF2, target ethylene-inducible transcription factors for proteasome-mediated degradation (Guo & Ecker, 2003, Potuschak et al., 2003). This degradation is dependent upon an active CTR1 protein and in the absence of CTR1, the transcription factor EIN3 can accumulate. Ethylene represses CTR1 activity, preventing the activity of EBF1/2 and this allows EIN3 to accumulate and activate ethylene responses. It is possible that EDR1 may be negatively regulating cell death responses in a similar manner. Mutations in the F-box protein ORE9 can block ethylene-induced cell death in the edr1 mutant, as well as the drought-induced growth inhibition, suggesting that a repressor of these phenotypes accumulates in the ore9 mutant. However, not all edr1-mediated responses can be blocked by ore9, indicating that ORE9 regulates just a subset of EDR1-mediated responses.

All known edr1 mutant phenotypes can be suppressed by a specific missense mutation in the KEEP ON GOING (KEG) gene, which encodes an E3 ubiquitin ligase that is responsible for the degradation of the ABA-inducible transcription factor ABI5. This result suggests that EDR1 may be mediating cell death via a mechanism similar to the regulation of ethylene responses by CTR1, namely the targeting of transcription factors to the proteasome. Consistent with this model, qRT-PCR analyses revealed that some ABA-inducible genes are expressed more highly in edr1 mutant plants and this enhanced expression is abolished by the keg-4 mutation (Wawrzynska et al., 2008).

Despite the extensive work performed on the edr1 mutant, there is still little information about how EDR1 negatively regulates cell death, particularly in response to G. cichoracearum. To investigate the control of cell death in the edr1 mutant, we performed microarray experiments to identify genes whose regulation was affected by the edr1 mutation in the presence of powdery mildew. As expected, many of the genes upregulated in the edr1 mutant were defense response genes, indicating that EDR1 negatively regulates defense signaling pathways and that removal of such repression in the edr1 mutant results in enhanced resistance. Significantly, the EDR1 protein was found to localize at least part of the time to the nucleus, suggesting that EDR1 may be regulating the stability and/or activity of defense-related transcription factors directly.

RESULTS

Identification of genes regulated by EDR1

Wild-type Col-0 and edr1 mutant plants were inoculated with G. cichoracearum and tissue was collected at 18, 36, and 96 hours post inoculation (hpi). By 18 hrs, the fungus has germinated, penetrated the epidermal cells and begun to form haustoria (Fabroet al., 2008). By 36 hrs, infected cells have begun to form papillae and deposit callose. By 96 hrs stalks of asexual spores (conidia) begin to form on wild-type leaves, but very few form on edr1 leaves; however, no cell death is observable in wild-type or edr1 plants even at 96 hpi and visible powder has not begun to form (Frye & Innes, 1998). Tissue was also collected from plants immediately prior to inoculation as an uninfected control (0 h). High quality RNA was prepared from the collected tissue, including four biological replicates per genotype per time point, and analyzed using Affymetrix ATH1 Gene chips.

To identify genes that are negatively regulated by EDR1, we first selected genes that were upregulated greater than two-fold in edr1 compared to wild-type Col-0 at any time point and that were determined to be significantly different (p≤0.05) using the Benjamini-Hochberg correction (Benjamini & Hochberg, 1995). This correction should reduce the false discovery rate to less than 5%. Additionally, genes that were upregulated greater than two-fold in edr1 or wild-type Col-0 after inoculation compared to uninoculated plants were selected. These datasets were then compared to identify genes that were upregulated in an edr1- and pathogen-dependent manner. Genes whose expression was higher in edr1 than in Col-0 at any time and was also higher in either Col-0 or edr1 or both after pathogen inoculation were selected (areas bounded by yellow oval in Fig. 1; Table S1). This subset of genes contained 553 probe sets corresponding to 545 annotated genes. We refer to this subset as the edr1&pm-upregulated gene set. Note that because of cost issues, we did not include an uninoculated control at each time point, thus it is a formal possibility that some of the genes included in the powdery mildew-induced gene set are upregulated due to circadian changes in gene expression instead of, or in addition to, powdery mildew infection. Never-the-less, all genes included in the edr1&pm-upregulated gene set are more highly expressed in the edr1 mutant than in WT Col-0 during at least one time point.

Figure 1.

Venn diagram showing overlap between powdery mildew-induced and edr1-induced gene sets. The yellow oval indicates the set of edr1&powdery mildew-upregulated genes used for the majority of analyses.

Upregulation of defense genes in the edr1 mutant

Many of the genes identified in the edr1&pm-upregulated gene set (Table S1) are known to be involved in plant defense responses. For example, PBS3 and PAD4 are both required for SA accumulation (Glazebrook et al., 1996, Nobuta et al., 2007), while PR-3, PR-4, THI2.1, PDF1.4, and ATTI1 are associated with JA-inducible defenses. Commonly, SA and JA defenses have antagonistic modes of action (Li et al., 2004), but can also have additive effects depending on hormone concentration and the type of pathogen encountered (Mur et al., 2006). PR-3 and PR-4 encode a chitinase and chitin binding protein with antifungal activity, respectively (Potter et al., 1993, Verburg & Huynh, 1991). THI2.1, PDF1.4, and ATTI1 are all induced by JA and may also have antifungal or antimicrobial activity (Epple et al., 1995, Silverstein et al., 2005). Other defense-associated genes in the edr1&pm-upregulated gene set include 12 genes with leucine-rich repeat (LRR) domains, including three encoding a Toll/Interleukin-1 Receptor (TIR) class nucleotide-binding leucine-rich repeat (NB-LRR) disease resistance protein.

The edr1&pm-upregulated gene set al.so contains five receptor-like kinases, including RLK5 and RLK6. RLK5 and RLK6 were identified in a search for genes that are regulated by pathogen-inducible transcription factors and are known to be induced by SA treatment and by pathogens (Du & Chen, 2000). In addition to the RLKs identified, there are also eight putative leucine-rich repeat kinases. LRR kinases can act as receptors to transmit signaling information, often during defense responses (Shiu & Bleecker, 2001). An additional 11 kinases are present in the edr1&pm-upregulated gene set, indicating that phosphorylation cascades are important for the defense mechanism induced by G. cichoracearum in edr1.

Significantly, the edr1&pm-upregulated gene set contains at least 28 genes encoding transcription factors. The largest family of transcription factors in this gene set is the WRKY family, of which there are eight members, or 1.5% of the annotated genes in this gene set, compared to 61 out of the 22810 genes on the ATH1 chip (0.27%). There are at least 75 WRKYs in the Arabidopsis genome and many WRKYs have been implicated in controlling aspects of plant defense responses (Bhattarai et al., 2010; Eulgem et al., 2000). WRKY transcription factors have a conserved DNA-binding domain, which contains a WRKY motif that is required for DNA binding (Ciolkowski et al., 2008). WRKYs also contain a zinc-binding region in the DNA-binding domain. WRKYs bind to the sequence (T)TGAC(C/T), known as the W-box, in the promoter sequence of target genes (Ciolkowski et al., 2008).

A second class of transcription factors overrepresented in the edr1&pm-upregulated gene set is the AP2/ERF family. There are seven AP2/ERFs present, comprising 1.28% of genes, compared to 89 on the ATH1 chip, or 0.39% of genes. AP2/ERF transcription factors were originally identified as genes that were induced in response to the hormone ethylene, but have since been shown to include genes that are induced in response to pathogen and during JA-inducible defenses (Gutterson & Reuber, 2004). In fact, one AP2/ERF, ORA59, has been found to integrate JA- and ET-mediated signaling pathways (Pre et al., 2008). AP2/ERF family transcription factors bind to the GCC-box (GCCGCC) to activate transcription.

The edr1&pm-upregulated gene set is also enriched for genes involved in reactive oxygen species (ROS) accumulation and turnover. Eleven genes annotated with putative peroxidase function were identified. Peroxidases act to oxidize other molecules through the use of H₂O₂ or O₂, either as a way of preventing toxicity or to signal (Yoshida et al., 2003). One of the peroxidase genes identified the edr1&pm-upregulated gene set, ATP2a, has also been shown to be upregulated in response to wounding and may play a role in pathogen responses (Cheong et al., 2002).

Other genes that are present in the edr1&pm-upregulated gene set include six genes that encode small heat shock proteins (sHSP). sHSPs can act as molecular chaperones and have been identified in regulation of responses to various stresses and developmental processes, including apoptosis (Basha et al., 2006). Genes encoding glutathione S-transferase genes are also present. GSTs are involved in regulating cellular redox state and are often induced during defense responses (Wagner et al., 2002).

Nine genes encoding FAD-binding domain containing proteins were also identified as part of the edr1&pm-upregulated gene set. These genes are closely related to a sunflower gene that encodes an antimicrobial protein with carbohydrate oxidase activity, Ha-CHOX (Custers et al., 2004). Ha-CHOX catalyzes the production of hydrogen peroxide using glucose as a substrate. When Ha-CHOX is overexpressed in tobacco, it confers greater resistance to Pectobacterium carotovorum. The nine FAD binding domain-containing genes in the edr1&pm-upregulated gene set are all members of the same family of proteins, sharing similarity across their entire length (Supplemental Fig. S1).

To determine if the FAD-binding domain genes are induced by other pathogens, data from publicly available microarrays were analyzed using the Genevestigator web-based interface (https://www.genevestigator.ethz.ch/gv/index.jsp) (Hruz et al., 2008, Zimmermann et al., 2004). Using Bimax clustering, the available high quality arrays were analyzed for conditions where the FAD genes from our dataset were expressed in similar patterns (Fig. 2). The majority of these genes were induced by multiple pathogens, including fungi (e.g. Botrytis cinerea), oomycetes (Phytophthora infestans) and bacteria (P. syringae). Additionally, several of these genes were also induced by the MAMPs (microbial associated molecular patterns) elf18, elf26, and chitin and by some abiotic stresses, including osmotic and oxidative stresses. These results point to a role for this family of genes in controlling defense and stress responses, perhaps through the production of H₂O₂.

Figure 2.

BiMax clustering of FAD binding domain-containing genes. Red indicates induction and green indicates suppression, with brighter colors indicating greater effect of the indicated treatment. The yellow box indicates the two genes with highly similar expression patterns over the indicated treatments.

Analysis of transcription factor motifs

To determine if the edr1&pm-upregulated gene set was enriched for genes that are induced by WRKY family transcription factors, a kilobase region of sequence upstream of the ATG was collected for all genes with AGI numbers in our dataset. These regions were then searched for W-boxes to determine the frequency of this element. As a control, we analyzed the equivalent upstream regions from all genes that remained unchanged (<1.155 fold up or down) in edr1 versus wild-type, and also remained unchanged at all time points after infection in both edr1 and wild-type plants (a total of 472 genes). Significantly, the upstream regions of genes from our dataset were enriched for W-boxes, with 910 elements, or a frequency of 1.59 elements per gene, compared to the unchanged dataset, where there was a frequency of 0.79 elements per gene (Table 1). These data strongly suggest that WRKY-regulated genes are upregulated in the edr1 mutant after pathogen treatment.

Table 1. Frequencies of transcription factor binding sites in promoter regions of edr1&pm-upregulated genes.

A 1 kb region 5’ of the start codon for each gene was analyzed. This information was also calculated for genes that were unchanged in edr1 relative to Col-0 and unchanged after pathogen infection in either genotype, and for all available upstream sequences in TAIR (33,518 sequences).

	#elements (edr1&pm- upregulated)	#elements (unchanged)	#elements (genome)	frequency (edr1)	frequency (unchanged)	freq uen cy (gen ome)
Wbox (TTGACC/T)	910	514	40,948	1.59	0.79	1.22
GCC box (GCCGCC)	52	60	3,528	0.091	0.127	0.11
MYC2 (CACATG)	267	204	16,556	0.47	0.43	0.49
ARF (TGTCTC)	270	284	17,690	0.47	0.60	0.53
MYB2 (C/TAACG/TG)	591	535	36,727	1.03	1.11	1.1
GATA (T/AGATAG/A)	1968	1404	107,927	3.43	2.98	3.2
DREB (CCGAC)	544	570	37,212	0.998	1.21	1.11

Using the same promoter scanning analysis as for the WRKY transcription factors, the number and frequency of GCC boxes in the promoter regions of genes from the edr1&pm-upregulated gene set were calculated. There were 52 GCC boxes, a frequency of 0.091 per gene, compared to a frequency of 0.127 per gene in the unchanged dataset and 0.11 for all Arabidopsis genes (Table 1), indicating that the majority of the genes in the edr1&pm-upregulated gene set are not regulated by AP2/ERF family transcription factors. Indeed the lower than average frequency of GCC boxes suggests that the edr1&pm-upregulated gene set is enriched in genes that lack AP2/ERF binding sites.

EDR1 localizes to the endoplasmic reticulum and the nucleus

The enrichment for WRKY transcription factors and genes containing their binding sites in the edr1&pm-upregulated gene set suggests that EDR1 negatively regulates the activity of these transcription factors. To determine whether this could be occurring directly, we analyzed the subcellular localization of full-length EDR1 protein fused to super yellow fluorescent protein (sYFP2; (Kremers et al., 2006)) using confocal microscopy. This fusion protein was shown to be functional as EDR1-sYFP expressed under the EDR1 native promoter was able to complement an edr1 Arabidopsis mutant in stable transgenic plants (Fig. S2). Unfortunately, we were unable to detect EDR1-sYFP using confocal microscopy in these plants, likely to the low level of expression from the native EDR1 promoter. To visualize EDR1-sYFP we thus transiently expressed it in Nicotiana benthamania leaves using a dexamethasone-inducible promoter. All transformed cells gave a similar pattern, displaying localization to internal membranes and to the nucleus (Fig. 3A). To confirm the nuclear localization, we co-expressed EDR1-sYFP with the nuclear protein GCN5-mCHERRY (Bhat et al., 2004). GCN5-mCHERRY fluorescence was confined to the nucleus and appeared to be excluded from the nucleolus (Fig. 3B). The nuclear portion of the EDR-sYFP fluorescence co-localized with GCN5. To determine whether the membrane localization was associated with the endoplasmic reticulum (ER), we co-expressed EDR1-sYFP with an ER marker consisting of the signal peptide of AtWAK2 (Arabidopsis thaliana Wall-Associated Kinase 2) at the N-terminus of mCherry and the ER retention signal His-Asp-Glu-Leu at its C-terminus (Nelson et al., 2007). Figures 3 D, E, and F show that EDR1-sYFP co-localized with this ER marker outside of the nucleus. To assess whether the nuclear signal from EDR1-sYFP could be due to degradation of EDR1-sYFP, we performed immunoblot analyses using an anti-GFP antibody (Fig. 3G). We observed a band of ~130 kD, the expected size for intact EDR1-sYFP, and a second band at approximately 70 kD. Untagged EDR1 is readily cleaved, releasing a C-terminal 50 kD fragment (without GFP) during extraction from either plant cells or from E. coli (data not shown), despite the use of protease inhibitors in the extraction buffer. It is thus unclear whether this 70 kD protein is present in live cells, but regardless, this EDR1-YFP fragment should be too large to diffuse into the nucleus on its own. These observations indicate that a portion of the EDR1 pool accumulates in the nucleus.

Figure 3.

Subcellular localization of EDR1. Panels a–c, EDR1-sYFP and GCN5-mCherry were transiently co-expressed in N. benthamiana leaves and imaged using confocal laser scanning microscopy. Panel a, EDR1-sYFP (a single optical section taken through the nucleus of an epidermal cell). Panel b, GCN5-mCherry expressed in the same cell. Panel c, overlay of panels a and b. Panels d–f, EDR1-sYFP and an mCherry ER marker (see Methods) were transiently co-expressed in N. benthamiana. d, EDR1-sYFP (a single optical section taken through the cell cortex of an epidermal cell); e, mCherry-HDEL, f, overlay of d and e. Panel g, immunoblot of EDR1-sYFP extracted from N. benthamiana leaves. KEG-sYFP is an unrelated YFP fusion protein included to show specificity of the antibody. Scale bar is 25 µm.

Gene ontology analysis of genes in the edr1-upregulated dataset

Gene Ontology (GO) annotations are assigned to nearly every gene in the Arabidopsis genome (Berardini et al., 2004). These annotations provide information about putative structure, function, and cellular localization for the predicted protein products. By analyzing the GO annotations of the edr1-upregulated genes at each time point, it is possible to determine which categories of genes are enriched, and how these change temporally after infection with powdery mildew. For this analysis, genes whose expression was at least two-fold higher in edr1 compared to wild-type Col-0 at any time point were selected (blue circle in Fig. 1). At 0 hours, prior to inoculation with pathogen, the GO categories that were most significantly enriched (p≤0.0001) were ‘endomembrane system’, ‘cellulase activity’, ‘cell wall’, ‘external encapsulating structure’, ‘extracellular region’, ‘membrane’, ‘nutrient reservoir’, ‘response to heat’, and ‘apoplast’ (Table S2), suggesting that in the absence of pathogen, the edr1 mutation primarily affects expression of genes associated with secretion and the cell wall. Notably, the endomembrane system category remained highly enriched at 18, 36 and 96 hpi, suggesting that secretory system may play an important role in edr1-mediated defenses. This would be consistent with the ER localization of EDR1.

At 18 hpi, the number of categories that were enriched increased dramatically. Of the 25 GO categories enriched with a P value ≤0.0001, the majority were associated with defense responses (e.g. ‘defense response’, ‘response to other organism’, ‘response to biotic stimulus’, ‘response to fungus’, ‘immune system process’, ‘response to chitin’, ‘innate immune response’, ‘systemic acquired resistance’, etc.) Additionally, the ‘kinase activity’ category was highly enriched (P=1.85E-04). These observations suggest that the early response to G. cichoracearum in edr1 plants is a specific defense response regulated in part by kinase cascades.

At 36 hpi, many of the same categories were still enriched, with the notable exception of the kinase activity related categories (Table S2). Additionally, at 36 hpi, the three new categories with the highest significance were ‘peroxidase activity’, ‘oxidorectase activity’ and ‘antioxidant activity’. This implies that by 36 hpi, edr1 plants are producing significantly more ROS than wild-type plants. After 96 hours, there was a reduction in the number of GO categories significantly enriched in edr1 plants because gene expression in wild-type Col-0 had caught up with the levels of expression in edr1 (Table S2). The GO analysis suggests that the response to G. cichoraceaum in edr1 mutant plants is primarily a more rapid and robust activation of defense genes, likely mediated by kinase signaling cascades that includes increased ROS production.

BiMax clustering of edr1-upregulated genes

Genes that are involved in defense pathways are induced in response to many different stimuli. To determine if the edr1&pm-upregulated gene set included genes that were also induced by other pathogens or stimuli, we used the BiMax clustering algorithm within the Genevestigator V3 web toolbox to analyze the expression of these genes across all Arabidopsis datasets involving biotic or abiotic stimulation, or comparing mutant plants to wild type (Prelic et al., 2006). These analyses revealed a subset of genes from the edr1&pm-upregulated gene set that were regulated in a similar manner in ten different treatments (Fig. 4A). These ten different experiments included infection with various virulent and avirulent pathogens, and responses to abiotic stresses such as ozone and wounding. Many of the genes that share regulation in these different categories are defense associated, including a flavin monooxygenase, a TIR-NB-LRR gene, two glutathione S-transferases, a WRKY transcription factor, and a cytochrome P450 gene that has been associated with defense responses (Fig. 4A). Additionally, the FAD binding domain containing gene At1g30700, the gene most similar to Ha-CHOX, was present in this Bimax cluster (Fig. 4A).

Figure 4.

BiMax cluster analysis of edr1&pm-upregulated genes. a, cluster analysis across the set of experiments housed within Genevestigator under the ‘stimulus’ category. b, cluster analysis across the set of experiments housed under the ‘mutant’ category. Only the largest clusters are shown, bounded by the yellow box in each.

BiMax clustering was also performed to compare gene regulation in different mutant backgrounds (Fig. 4B). Interestingly, the largest BiMax mutant cluster included the constitutive defense mutant cpr5 (Bowling et al., 1997). Over one-third of the genes from the edr1&pm-upregulated gene set are also upregulated in cpr5. This implies that enhanced resistance controlled by edr1 and cpr5 may be mediated by many of the same genes. CPR5 was identified in a screen for plants with enhanced resistance to pathogens. Unlike the edr1 mutant, however, expression of defense genes such as PR-1 in the cpr5 mutant is high in the absence of pathogen.

DISCUSSION

The EDR1 kinase appears to negatively regulate cell death in response to pathogen infection, as well as in response to abiotic stress. While it has been established that EDR1 is a functional kinase (Frye et al., 2001), we still have little information on how EDR1 regulates responses to pathogen infection. The transcriptome analyses described above revealed a set of genes that are upregulated in the edr1 mutant relative to wild-type Arabidopsis after inoculation with G. cichoracearum. Of the 545 genes identified in this dataset, many are known to be involved in disease resistance from previous work. The presence of these genes indicates that the enhanced resistance to pathogen infection in edr1 is at least partly the result of derepression of defense-associated genes. This result supports the previous finding that edr1-dependent enhanced resistance requires an intact SA signaling pathway (Frye et al., 2001), as many of these known genes are involved in SA pathways.

Additionally, the genes that are induced in edr1 after pathogen inoculation encode for many signaling proteins, including putative NB-LRR proteins and receptor-like kinases. The higher expression of these genes indicates that signaling, and possibly perception of pathogens, is elevated in the edr1 mutant, and that one function of EDR1 may be to prevent unnecessary signaling. This may also serve to limit the perception of pathogen in the absence of immediate threat.

It should also be noted that while the edr1 mutation enhances expression of numerous defense and signaling genes following powdery mildew inoculation, this set of genes represents only small percentage of the total number of genes induced at least two-fold following inoculation (Fig. 1). We identified nearly 4,000 genes that were induced in both wild-type and edr1 mutant plants during at least one time point following inoculation. The great majority of these (>95%) were not significantly affected by the edr1 mutation (i.e. they were induced similarly in wild-type and edr1 plants), indicating that the edr1 mutation does not simply enhance expression of all powdery mildew-induced genes.

The number of genes we identified as being induced by powdery mildew infection (4920 in WT and 6352 in edr1) is larger than reported previously. For example, Zimmerli et al. (2004) examined gene expression in wild-type Arabidopsis infected with the same strain of powdery mildew as used in this study and identified only 13 genes that were significantly upregulated at 24 hours post inoculation. More recently, Fabro et al. (2008) examined gene expression in wild-type, npr1-1 mutant and jar1-1 mutant Arabidosis at 18 h following infection with the same powdery mildew strain and identified 117 induced genes. It is difficult to make direct comparisons between our study and those of Zimmerli and Fabro, however, as their studies employed cDNA microarrays containing only about half of the genes present on the ATH1 Affymetrix gene chip used in our study, two color dye hybridization and different statistical tests. The reduced variability associated with Affymetrix chips compared to cDNA microarrays likely increased the sensitivity of our analyses. In addition, we sampled at later time points (36 and 96 hrs), allowing us to identify genes induced later in the infection process. Finally, many of the genes identified as upregulated in both edr1 and WT plants at 18 and 36 hrs in our study may be under circadian regulation, which is one reason why genes that were not also upregulated in edr1 relative to wild-type plants were excluded from our analyses (Figure 1).

Analysis of GO annotations for the edr1-upregulated gene set (blue circle in Fig. 1) revealed a distinct temporal pattern of gene induction during powdery mildew infection. At 18 hpi, numerous defense associated gene categories were highly enriched, including the ‘kinase’ category, while at 36 hpi, the kinase category was no longer enriched, but several ROS-related categories appeared. These results suggest that immediately following pathogen inoculation, genes involved in signaling and defense responses are expressed more highly, and that as the response continues, there is a shift from initial induction of signaling to a more sustained response, perhaps through the use of ROS as signaling molecules. By 96 hours, most of the categories were no longer enriched, because gene expression in wild-type plants had caught up.

To determine if the genes in the edr1-upregulated dataset were also regulated in response to other pathogens, we used the publicly available microarray data and the web-based analysis tool Genevestigator V3 (Hruz et al., 2008, Zimmermann et al., 2004). Using the BiMax algorithm within Genevestigator (Prelic et al., 2006), we identified subsets of genes that are regulated in similar manners in other microarray experiments (Fig. 4). Interestingly, many of the genes upregulated in edr1 were also induced in response to different pathogens as well as abiotic stress. That these genes are regulated by EDR1 suggests that in the absence of pathogen, EDR1 serves to keep transcription of these genes low or off, and once a pathogen has been detected, EDR1 function is repressed, allowing for higher levels of transcription of pathogen-inducible genes. Interestingly, a subset of the edr1-upregulated genes was also induced at greater levels during attack by the whitefly Bemisia tabaci. The induction of defense-associated genes in response to B. tabaci may be due to a wounding response, which is consistent with the observation that these same genes are also regulated in a similar manner in response to mechanical wounding.

BiMax clustering also reveled that many of the edr1-upregulated genes are also more highly expressed in cpr5 mutants. Mutations in CPR5 cause constitutive expression of defense genes such as the PR genes and PDF1.2, elevated ROS in leaves, and formation of lesions that display deposition of autofluorescent compounds (Bowling et al., 1997). Lesions induced on edr1 mutant leaves also show deposition of autofluorescent compounds and elevated ROS (unpublished observations). Also like the edr1 mutant, cpr5 mutant plants display enhanced senescence (Jing et al., 2007, Yoshida et al., 2002). Interestingly, the majority of the genes expressed in common between cpr5 and edr1 are not suppressed by the npr1 mutation in the cpr5 mutant (Fig. 4b). Previous work demonstrated that edr1-mediated disease resistance is dependent upon NPR1 (Frye et al., 2001). It is possible that CPR5 functions downstream of NPR1, or that this subset of genes is not central to edr1-mediated resistance. Additionally, while these genes may be NPR1-independent in a cpr5 background, they may still require NPR1 in the edr1 background. Further experiments will be required to understand the signaling pathways controlling expression of this subset in edr1.

The elevated levels of ROS and ROS-associated gene expression in edr1 plants suggests that ROS may play a role in edr1 phenotypes, including enhanced sensitivity to drought (Tang et al. 2005). Recently, mutations in an EDR1-like gene in rice designated DSM1 (drought-hypersensitive mutant1) were shown to confer a similar enhanced sensitivity to drought (Ning et al. 2010). This drought sensitivity correlated with an increased sensitivity to oxidative stress and a reduction in expression of two peroxidase genes and in peroxidase activity during drought stress. Similar to the edr1 mutant, transcriptome analysis of the dsm1 mutant revealed a large number of genes (678) whose expression was significantly upregulated during stress, suggesting that the dsm1 mutation may be causing large disruptions to cellular homeostasis under stress conditions.

The set of edr1-upregulated genes should include the genes directly responsible for the enhanced disease resistance phenotype of edr1 mutant plants. A particularly intriguing gene family identified in our data set is the FAD-binding domain family, which is related to the sunflower Ha-CHOX gene. Ha-CHOX was identified for its antimicrobial properties and was found to have carbohydrate oxidase activity, which is the ability to convert glucose into H₂O₂ (Custers et al., 2004). This class of proteins may be acting to produce ROS that can act as either signaling molecules or as agents of cell death. ROS are produced in the cell during many different processes, including photosynthesis and defense responses (Apel & Hirt, 2004). During defense responses, ROS can be produced by a variety of proteins, including NADPH oxidases (also known as respiratory burst oxidase homolog (Rboh)) in the plasma membrane and peroxidases present in the apoplast (Allan & Fluhr, 1997, Torres et al., 2005b, Vera-Estrella et al., 1992). H₂O₂ production has been linked to programmed cell death (PCD) in response to pathogen (Dangl & Jones, 2001). However, overexpression of the Rboh protein AtRbohD limits PCD in response to P. syringae DC3000 and Botrytis cinerea (Torres et al., 2005a). It appears that in this case ROS may be acting instead as a signaling molecule delineating the area of infection. At the time of its discovery, Ha-CHOX represented a new class of oxidases for producing ROS with glucose as a substrate (Custers et al., 2004). This family of proteins represents another pathway for production of ROS, and may contribute to the observed resistance of edr1 plants to powdery mildew infection.

Two classes of transcription factors were significantly overrepresented in the edr1-upregulated dataset, WRKYs and AP2/ERFs. Both WRKY family and AP2/ERF family transcription factors have been previously shown to be induced during defense responses and are known to induce defense-related genes (Buttner & Singh, 1997, Eulgem & Somssich, 2007b, Song et al., 2005). Consistent with the overrepresentation of WRKY transcription factors, promoter scanning of the edr1-upregulated genes showed them to be enriched for W-boxes. Surprisingly, however, the same gene set had a lower than average frequency of AP2/ERF binding sites (GCC boxes), possibly indicating that the upregulated AP2/ERFs function as transcriptional suppressors.

WRKY transcription factors have been long associated with control of defense gene induction. The WRKY box was originally identified in the promoters of PR genes from parsley (Rushton et al., 1996). Most WRKYs are transcriptional activators, while there is evidence that some can also act as repressors of transcription (Eulgem & Somssich, 2007a). Two of the WRKY genes identified in the edr1-upregulated dataset, WRKY38 and WRKY59, have also been shown to be induced by overexpression of NPR1 and by treatment with the SA analog benzothiadiazole S-methylester, supporting a role for these two transcription factors in defense responses (Wang et al., 2006). Another WRKY that was present in the dataset, WRKY75, has been shown to be a regulator of phosphate (Pi) uptake in roots and is induced under Pi deficient conditions, indicating a role for this gene in nutritional stress responses (Devaiah et al., 2007).

AP2/ERF transcription factors were originally identified as proteins that modulated transcription in response to ethylene. Besides proteins that are responsive to ethylene, ERF-domain containing transcription factors can also be activated in response to pathogen, such as Pti4 from tomato (Chakravarthy, 2003), and in response to drought, such as DREB2A (Sakuma et al., 2006). While none of the AP2/ERF genes identified in our dataset have a function yet ascribed, their similarity to other AP2/ERF transcription factors suggest that they may also be involved in defense or drought related responses.

The promoters of several of the edr1-upregulated transcription factors contain W boxes, indicating that they may be regulated by a positive feedback loop (data not shown), enabling a rapid response to even slightly elevated levels of these proteins. We hypothesize that EDR1 may function to regulate the level of these proteins by phosphorylation that would then target them for proteasome-mediated degradation. This model is supported by our localization studies that showed at least a fraction of the EDR1 protein is localized to the nucleus (Fig. 3), where it could interact with these transcription factors directly. A similar model has been proposed for the CTR1 kinase (Gagne et al., 2004, Guo & Ecker, 2003, Potuschak et al., 2003), which belongs to the same kinase subfamily as EDR1 (Frye et al., 2001). CTR1 regulates the level of the EIN3 transcription factor via direct or indirect phosphorylaton of EIN3 on a specific threonine residue (T592; Yoo et al., 2008), which promotes its degradation by the proteosome (Gao et al., 2003). In this context, it is worth noting that like EDR1, CTR1 has been localized to the endoplasmic reticulum, where it is associated with ethylene receptors (Gao et al., 2003); thus, if CTR1 is directly phosphorylating transcription factors, it may also need to move between a membrane complex and the nucleus. Alternatively, it has been proposed that CTR1 may activate a MAP kinase pathway that then leads to the phosphorylation of EIN3 on T592, but this remains to be shown (Yoo et al., 2008). A function for EDR1 and CTR1 in the nucleus is further supported by the finding that DSM1 from rice, which belongs to the same subfamily of kinases as EDR1 and CTR1, is primarily located in the nucleus (Ning et al., 2010).

Like CTR1, the majority of EDR1 protein appears to be associated with the endoplasmic reticulum (Fig. 3). The significance of this localization is not yet clear, but is consistent with the GO analyses that showed genes associated with secretion and the endomembrane system are highly enriched in the edr1-upregulated dataset (supplementary Table S2).

That EDR1 may be regulating the levels of transcription factors via targeting them to the proteosome is supported by our previous finding that all edr1-mediated phenotypes can be suppressed by a missense mutation in the KEEP ON GOING gene (KEG), which encodes a RING-finger E3 ubiquitin ligase (Wawrzynska et al., 2008). Null mutations in KEG have been shown to cause elevated levels of the ABI5 transcription factor, a central regulator of ABA signaling during post-germinative growth, and KEG and ABI5 can physically interact (Stone et al., 2006). These data suggest that KEG may ubiquitinate ABI5, targeting it for proteosome-mediated degradation. ABI5 cannot be the only target of KEG, however, as an abi5 null mutation only partially suppresses a keg null mutation (Stone et al., 2006). We have proposed a model whereby EDR1 is responsible for phosphorylating at least a subset of transcription factors that are KEG substrates, and it is this phorphorylation that promotes association with KEG (Wawrzynska et al., 2008). The transcription factors identified in the present study as being upregulated by the edr1 mutation represent candidates for testing this model.

EXPERIMENTAL PROCEDURES

Plant growth and inoculation conditions

Arabidopsis thaliana Col-0 and edr1 seeds were sown on Metromix soil and placed at 4°C for three days. Plants were then transferred to a growth room and grown under 9 hour days at a temperature of 23°C. After four weeks, the plants were inoculated with Golovinomyces cichoracearum using a settling tower ~1 meter tall. Plants to be inoculated were placed at the bottom of the tower, which contained a Nytex mesh screen at the top. Four pad4 mutants with heavy powder growth were passed over the mesh 20 times each to transfer the spores to the plants below. Spores were allowed to settle for 30 minutes and the plants were transferred to growth chambers.

Tissue collection and RNA preparations

Tissue was collected for each time point (0, 18, 36, and 96 hours) by harvesting four full rosettes per genotype per biological replicate. Four biological replicates were collected and placed in liquid nitrogen. Tissue was ground with a mortar and pestle and used for RNA preparations. High quality RNA was prepared using the Spectrum Plant Total RNA Kit (Sigma) and concentrated to >0.75 µg/µl with the RNEasy MinElute Cleanup Kit (Qiagen). RNA was then frozen in liquid nitrogen and shipped to the Center for Medical Genomics at the Indiana University School of Medicine.

Transcriptome analyses

First strand cDNA synthesis, biotinylated cRNA synthesis, hybridization to Affymetrix ATH1 GeneChips®, and chip scanning were carried out using the facilities of the Center for Medical Genomics at Indiana University School of Medicine. Data were processed using the Affymetrix MAS5 algorithm. Data analysis was carried out using ArrayAssist software (now sold under the name GeneSpring GX from Agilent Technologies) and Genevestigator (https://www.genevestigator.ethz.ch/) (Hruz et al., 2008). Data were normalized using the GC-RMA algorithm and log₂-transformed using ArrayAssist. Genes whose expression was at least 2-fold greater in edr1 than wild-type Col-0 for any time point with a p-value ≤ 0.05 using the asymptotic computation were selected. We also generated separate lists of genes that were induced at least 2-fold after inoculation with G. cichoracearum in wild type and in the edr1 mutant. Each list of genes was then subjected to correction for multiple testing errors using the Benjamini-Hochberg method and genes with a corrected p-value ≤ 0.05 were selected, representing a false discovery rate less than or equal to 5% (Benjamini & Hochberg, 1995). As a control for our promoter analyses, we also selected set of genes unresponsive to either G. cichoracearum or the edr1 mutation (the unchanged dataset) defined as all genes whose fold change was <1.155 (up or down) in all comparisons. The 1.155 fold change value was chosen in order to create a gene set of approximately the same size as the edr1-upregulated gene set.

For Gene Ontology (GO) analyses we used the ArrayAssist program to identify GO terms that were significantly enriched (P≤0.05) in the set of genes upregulated in edr1 at each time point at least two-fold relative to wild-type (blue circle in Fig. 1). For bicluster analysis of the edr1&pm-upregulated genes, we used the BiMax algorithm within the web-based program GENEVESTIGATOR V3 (Hruz et al., 2008, Prelic et 2006). Because the BiMax algorithm is limited to analyzing 100 genes at a time, we divided our set of edr1&pm-upregulated genes (yellow circle in Fig. 1) into five groups and analyzed each group independently. Each group was subjected to BiMax cluster analysis with discretization set to 1.0 (Prelic et al., 2006). The ‘stimulus’ and ‘mutant’ microarray data sets within Genevestigator were analyzed separately.

Promoter analyses

One kilobase regions upstream of the ATG start codon were collected for all the genes in the edr1-upregulated dataset and the unchanged dataset using the The Arabidopsis Information Resource (TAIR) bulk sequence retrieval tool (http://arabidopsis.org/tools/bulk/sequences/index.jsp). The promoter regions were scanned for six letter words using the TAIR motif analysis tool (http://arabidopsis.org/tools/bulk/motiffinder/index.jsp).

Construction of EDR1-sYFP fusion proteins and subcellular marker proteins

To make translational fusions of EDR1 to sYFP2 (Kremers et al., 2006), a full-length EDR1 cDNA without the stop codon and an sYFP2 cDNA with a stop codon were cloned into pDONR P1-P4 and pDONR P4r-P2 Gateway compatible vectors (Invitrogen), respectively. The sequences of the EDR1 and sYFP in the respective vectors were verified and the two pDONR vectors with EDR1 and sYFP were recombined into the pTA7002-GW destination vector (Aoyama & Chua, 1997, McNellis et al., 1998), using multisite Gateway cloning technology from Invitrogen, to generate a dexamethasone inducible EDR1-sYFP fusion protein construct. A similar cloning strategy was used to generate the dexamethasone inducible nuclear marker protein GCN5-mCherry (Bhat et al., 2004). An ER marker was created by combining the signal peptide of AtWAK2 (Arabidopsis thaliana wall-associated kinase 2) at the N-terminus of mCherry and the ER retention signal His-Asp-Glu-Leu at its C-terminus (Nelson et al., 2007). To generate an EDR1-sYFP construct expressed under the native EDR1 promoter, approximately 1.5 kb of EDR1 5’ sequence was inserted into the binary vector pMDC32-HPB in place of the 35S promoter in this Gateway compatible vector (Qi & Katagiri, 2009). EDR1-sYFP was then recombined into the resulting vector as described above. A stop codon was included after the sYFP sequence, thus the HPB tag was not added.

Subcellular localization of EDR1

Fusion proteins were transiently expressed in leaves of N. benthamiana using agroinfiltration as previously described (Ade et al., 2007). For dexamethasone-inducible constructs, leaves were imaged 24 hrs after application of 50 µM dexamethasone. Intracellular fluorescence was observed by confocal laser scanning microscopy using a Leica SP5 AOBS inverted confocal microscope (Leica Microsystems) equipped with argon ion (458-, 476-, 488-, 496-, and 514-nm laser lines) and the He-Ne (561- nm laser line) lasers and a Leica 63X NA1.2, HCX PL APO, water objective (Part# 506279). sYFP (excited by the 514-nm Argon laser) fluorescence was detected using the Leica AOBS system and a custom 522- to 545-nm band-pass emission filter, whereas mCherry (excited using 561 nm He-Ne laser) fluorescence was detected using the Leica AOBS system and a custom 595- to 620-nm band-pass emission filter.

The integrity of the EDR1-sYFP protein within N. benthamiana leaves was assessed by immunoblot analysis using rabbit polyclonal anti-GFP antisera (Thermo Scientific).

Complementation of the edr1 mutation with EDR1-sYFP

The Arabidopsis edr1 mutant was transformed with the EDR1 native promoter EDR1-sYFP construct described above using the floral dip transformation procedure (Clough & Bent, 1998). Plants containing the transgene were selected on agar plates using 30 µg/ml hygromycin. T2 generation plants were tested for complementation of drought-induced senescence and lesion phenotypes of the edr1 mutant. Plants were grown in Metromix 360 in 4 inch plastic pots under 9 hr days for 3 weeks with watering as needed to keep soil moist. At three weeks watering was stopped. Ten days after cessation of watering edr1 plants began to show yellow and brown lesions on leaves and severe chlorosis on older leaves, while all leaves on wild-type Col-0 plants and edr1 plants transformed with EDR1-sYFP remained green.

Data deposition

The raw and normalized gene expression data generated in this study have been deposited in the NCBI GEO expression database (http://www.ncbi.nlm.nih.gov/geo/) under accession number GSE26679.