Arabidopsis EDR1 Protein Kinase Regulates the Association of EDS1 and PAD4 to Inhibit Cell Death

Abstract

ENHANCED DISEASE SUSCEPTIBILITY1 (EDS1) and PHYTOALEXIN DEFICIENT4 (PAD4) are sequence-related lipase-like proteins that function as a complex to regulate defense responses in Arabidopsis by both salicylic acid–dependent and independent pathways. Here, we describe a gain-of-function mutation in PAD4 (S135F) that enhances resistance and cell death in response to infection by the powdery mildew pathogen Golovinomyces cichoracearum. The mutant PAD4 protein accumulates to wild-type levels in Arabidopsis cells, thus these phenotypes are unlikely to be due to PAD4 over accumulation. The phenotypes are similar to loss-of-function mutations in the protein kinase EDR1 (Enhanced Disease Resistance1), and previous work has shown that loss of PAD4 or EDS1 suppresses edr1-mediated phenotypes, placing these proteins downstream of EDR1. Here, we show that EDR1 directly associates with EDS1 and PAD4 and inhibits their interaction in yeast and plant cells. We propose a model whereby EDR1 negatively regulates defense responses by interfering with the heteromeric association of EDS1 and PAD4. Our data indicate that the S135F mutation likely alters an EDS1-independent function of PAD4, potentially shedding light on a yet-unknown PAD4 signaling function.

Loss-of-function mutations in the ENHANCED DISEASE RESISTANCE1 (EDR1) gene of Arabidopsis confer enhanced resistance to the powdery mildew pathogen Golovinomyces cichoracearum (Frye and Innes 1998). This enhanced resistance is correlated with enhanced cell death at the site of infection. The edr1-1 mutation causes a premature stop codon in the EDR1 gene, which encodes a protein kinase with homology to mitogen-activated protein kinase kinase kinases (MAPKKKs) belonging to the Raf family (Frye et al. 2001). The edr1 mutant does not display constitutive expression of defense genes in the absence of a pathogen, indicating that the enhanced resistance is not caused by constitutive activation of systemic acquired resistance (Frye and Innes 1998); however, edr1-mediated disease resistance is suppressed by mutations that block or reduce salicylic acid (SA) production or signaling (Christiansen et al. 2011; Frye and Innes 1998; Frye et al. 2001; Hiruma and Takano 2011; Hiruma et al. 2011; Tang et al. 2005), suggesting that edr1-mediated enhanced resistance against G. cichoracearum requires an intact SA signaling pathway.

In addition to enhancing resistance to powdery mildew, loss-of-function mutations in EDR1 enhance drought-induced growth inhibition, ethylene-induced senescence, and sensitivity to abscisic acid (ABA) (Tang et al. 2005; Wawrzynska et al. 2008). The enhanced drought-induced growth inhibition and enhanced ABA sensitivity phenotypes but not ethylene-induced senescence are suppressed by mutations in the ENHANCED DISEASE SUSCEPTIBILITY 1 (EDS1) and PHYTOALEXIN DEFICIENT 4 (PAD4) genes, which encode sequence-related nucleocytoplasmic lipase-like proteins (Tang et al. 2005). The inability of these mutations to suppress the ethylene-induced senescence phenotype of edr1 mutants suggests that EDR1may regulate multiple pathways.

The pad4 mutant was originally isolated in an Arabidopsis screen for enhanced disease susceptibility to Pseudomonas syringae pv. maculicola (Glazebrook et al. 1996). PAD4 physically interacts with EDS1 as a heterodimer (Feys et al. 2001; Jirage et al. 1999; Rietz et al. 2011; Wagner et al. 2013), forming a nucleo-cytoplasmic complex that promotes accumulation of the plant defense signaling molecule SA (Cui et al. 2017; Feys et al. 2001). EDS1 and PAD4 also contribute to defense responses activated by intracellular nucleotide-binding, leucine rich repeat (NLR) receptors that have an N-terminal toll-interleukin 1 receptor (TIR) domain (Aarts et al. 1998; Bhandari et al. 2019; Cui et al. 2018; Feys et al. 2001). NLR-mediated immune responses are often associated with localized host-cell death as part of the hypersensitive response (HR) (Maekawa et al. 2011). Arabidopsis pad4 mutants display a delayed HR against the oomycete pathogen Hyaloperospora arabidopsidis that is insufficient for preventing pathogen spread (Feys et al. 2001). This partially retained HR can be attributed to partial genetic redundancy between PAD4 and the nuclear SENESCENCE-ASSOCIATED GENE 101 (SAG101), another component of the EDS1 regulatory hub (Feys et al. 2005; Lipka et al. 2005). It was recently established that EDS1-SAG101 heterodimers promote HR cell death in TIR-NLR receptor immunity, whereas formation of EDS1-PAD4 heterodimers is necessary for transcriptionally mobilizing SA and other defense pathways (Bhandari et al. 2019; Feys et al. 2005; Gantner et al. 2019; Lapin et al. 2019; Rietz et al. 2011). Complementary studies have shown that EDS1 and PAD4 transduce photo-oxidative stress signals leading to cell death and the slowing of plant growth and that they are involved in plant fitness regulation (Chandra-Shekara et al. 2007; Venugopal et al. 2009; Wituszyńska et al. 2013; Xiao et al. 2001).

So far, all described mutations in EDS1 and PAD4 have caused a loss of function (Feys et al. 2001; Hu et al. 2005; Jirage et al. 1999; Rietz et al. 2011; Wagner et al. 2013). Here, we describe a gain-of-function mutation in the PAD4 gene that enhances a subset of edr1 mutant phenotypes, including edr1-dependent cell death after powdery mildew infection and edr1-accelerated ethylene- and age-induced senescence. This mutation causes a serine to phenylalanine substitution at position 135 of PAD4. Furthermore, the PAD4^S135F substitution alone confers enhanced disease resistance and enhanced cell death after infection with the powdery mildew fungus G. cichoracearum. The molecular basis for these phenotypes remains unclear; however, the S135F substitution did not affect PAD4 protein accumulation, localization, or its ability to associate with EDS1. The discovery that PAD4^S135F enhances a subset of edr1 phenotypes supports previous findings that the edr1 phenotype is at least partially due to changes in SA signaling (Tang et al. 2005). Analysis of edr1 and pad4/eds1 transcriptome data revealed that a significant proportion of the PAD4/EDS1 gene network is upregulated in edr1 plants during the defense response. To follow up on these results, we investigated whether EDR1 plays a direct role in regulating PAD4. Significantly, we found that EDR1 interacts with both PAD4 and EDS1 and that EDR1 can inhibit the interaction between EDS1 and PAD4.

RESULTS

Identification of a mutation in PAD4 that enhances edr1 mutant phenotypes.

The edr1 mutant displays enhanced sensitivity to flg22, a 22-amino acid peptide derived from bacterial flagellin that is known to induce defense responses (Geissler et al. 2015). This sensitivity can be assayed in very young seedlings grown in liquid culture. We took advantage of this phenotype to screen for second site mutations that can suppress this enhanced flg22 sensitivity, restoring edr1 mutants to a wild-type phenotype. Candidate suppressor mutants obtained in this screen were assessed for the presence of mutations in genes previously shown to be required for edr1 mutant phenotypes (Tang et al. 2005; Wawrzynska et al. 2008), so that we could focus our efforts on new genes. To our surprise, all suppressor candidates analyzed (13 in total) carried an identical missense mutation in the PAD4 gene, causing a change of amino acid Ser135 to Phe135 (PAD4^S135F). Because these 13 mutants were derived from multiple different ethylmethane sulphonate–mutagenized parents, it seemed likely that the parent population (prior to mutagenesis) carried this mutation and that the mutation was not responsible for the suppressor phenotype. We therefore sequenced the PAD4 gene in the edr1-1 parental line used for suppressor mutagenesis. This analysis confirmed that the edr1-1 parental line used for the suppressor mutagenesis carried the same mutation and that this mutation had arisen at some point during the backcrossing process of the original edr1-1 mutant, which lacks this mutation (discussed below). We have designated this new pad4 mutation as pad4-13, as it represents the 13th mutant allele of pad4 to be described. To separate the pad4-13 mutation from the edr1-1 mutation, the double mutant line was backcrossed to wild-type Col-0 and F2 plants identified that were homozygous mutant at one locus and homozygous wild type at the other. Each separate mutant was then back-crossed to wild-type Col-0 three times to eliminate any other unlinked or loosely linked mutations.

The pad4-13 mutation confers enhanced disease resistance and contributes to edr1-dependent enhanced cell death.

Because we had previously shown that loss-of-function mutations in PAD4 suppressed edr1-1 mutant phenotypes (Tang et al. 2005), the discovery that a missense mutation in PAD4 was present in the edr1-1 mutant suggested that the pad4-13 mutation might be contributing to edr1 mutant phenotypes. To test this hypothesis, we infected wild-type Col-0, edr1-1 (lacking pad4-13), edr1-3 (contains a T-DNA insertion in EDR1), pad4-13, and edr1-1 pad4-13 plants with G. cichoracearum and quantified fungal growth by counting conidiospores at 8 days postinoculation (dpi). As expected, edr1-1 pad4-13 plants had a reduced spore count compared with wild-type Col-0 (Fig. 1A). This enhanced disease resistance was not influenced by the presence of the pad4-13 mutation, as the edr1-1 and edr1-3 mutants had comparable spore counts (Fig. 1A). Interestingly, the pad4-13 mutant also had a reduced spore count, similar to that of the edr1 mutants (Fig. 1A). These results indicate that the pad4-13 mutation alone confers an enhanced disease resistance similar to edr1 mutations and that the mutations are not additive in their effects.

Fig. 1.

The pad4-13 mutation confers enhanced disease resistance and contributes to edr1-associated cell death. A, Quantitative analysis of powdery mildew conidia (asexual spores) on Col-0, edr1-1, pad4-13, edr1-1pad4-13, edr1-3 lines. Plants were inoculated with powdery mildew, and conidia production was determined 8 days postinoculation (dpi). Bars indicate the mean of three samples, each with three technical replicates. Error bars indicate standard deviation. Results are representative of three independent experiments. B, Trypan blue staining of powdery mildew-infected Col-0, edr1-1, pad4-13, edr1-1pad4-13, and edr1-3 lines. Four-week old plants were infected with Golovinomyces cichoracearum and trypan blue was used to reveal fungal hyphae and patches of dead mesophyll cells (arrows) at 8 dpi. Bars = 50 μm. Pictures are representative of three independent experiments. C, Quantification of cell death in powdery mildew–infected leaves. The indicated lines were assessed for leaf mesophyll cell death at 8 dpi, using trypan blue staining. Dead cells stained with trypan blue were quantified using ImageJ. For quantification, six pictures from five independent experiments were randomly chosen (n = 30). Results are provided as means with 10th and 90th percentiles (box) and range (whiskers). Statistical outliers are shown as circles. Lower case letters indicate values that are significantly different (P < 0.01; one-way analysis of variance test, using the Bonferroni method).

Loss-of-function mutations in PAD4 have been shown to enhance disease susceptibility (Feys et al. 2001; Frye et al. 2001; Glazebrook et al. 1997; Zhou et al. 1998). Indeed, upon G. cichoracearum infection, pad4-1 plants accumulate more fungal spores than wild type (Supplementary Fig. S1). These data indicate that the pad4-13 mutation causes a gain-of-function that enhances resistance to G. cichoracearum.

In addition to enhancing resistance to G. cichoracearum, the edr1 mutation causes an increase in mesophyll cell death following infection by this fungus (Frye and Innes 1998). To assess whether the pad4-13 mutation contributes to this cell-death phenotype, we used trypan blue staining to score cell death at 5 dpi. The edr1-1 pad4-13 mutant displayed large patches of mesophyll cell death (Fig. 1B). In comparison, the edr1-1 and edr1-3 mutants displayed fewer patches of dead cells and these patches were smaller. Significantly, the pad4-13 mutant also displayed patches of dead mesophyll cells, similar in appearance to the edr1 mutants. No mesophyll cell death was detected in wild-type Col-0 plants. To further characterize the cell death response, the patches of dead mesophyll cells positive for trypan blue staining were quantified. The edr1-dependent cell death was enhanced by the presence of the pad4-13 mutation, indicating that the two mutations are additive in their effect on powdery mildew–induced cell death (Fig. 1C). Notably, pad4-13 plants displayed a significantly higher level of cell death than edr1 plants.

EDR1 physically interacts with EDS1 and PAD4.

The conclusion that pad4-13 can enhance some but not all edr1 phenotypes prompted us to investigate whether EDR1 and PAD4 are part of a common regulatory complex. In support of this hypothesis, both proteins were previously shown to localize partially to the nucleus (Christiansen et al. 2011; Feys et al. 2005). To test whether EDR1 interacts with PAD4, we performed yeast two-hybrid analyses. Counter to expectations, we could not detect an interaction between wild-type EDR1 and PAD4 (Fig. 2A). As described above, however, PAD4 is known to interact with EDS1, and this interaction is required for both basal disease resistance and TIR-NLR-mediated resistance (Feys et al. 2005; 2001; Rietz et al. 2011; Wagner et al. 2013), suggesting that the genetic interaction between EDR1 and PAD4 could be mediated by EDS1. We thus tested whether EDR1 interacts with EDS1 and observed a positive yeast two-hybrid interaction (Fig. 2A). One possible reason we could not detect the interaction between PAD4 and EDR1 is that PAD4 could be a substrate of EDR1, and this interaction may be very transient. We therefore tested whether a substrate-trap mutant form of EDR1, EDR1^ST (Gu and Innes 2011), interacts with PAD4. EDR1^ST results from an aspartic acid to alanine substitution in the catalytic site of EDR1 (amino acid 810), which inhibits phosphotransfer and, thus, stabilizes interactions with substrates. Indeed, EDR1^ST was found to interact with both EDS1 and PAD4. However, the enhanced interaction of EDR1^ST with PAD4 is possibly explained by enhanced stability of the mutant protein compared with wild-type EDR1 (Fig. 2A).

Fig. 2.

EDR1 physically interacts with EDS1 and PAD4. A, Yeast two-hybrid analysis of EDR1 interactions with EDS1 and PAD4. AD = GAL4 activation domain fusion, BD = GAL4 DNA binding domain fusion, T = SV40 large T antigen, LAM = lamin. Protein expression was verified through immunoblotting. AD-tagged proteins also contain a hemagglutinin (HA) tag, which was used for detection. B, EDR1 coimmunoprecipitates with PAD4. C, EDR1 coimmunoprecipitates with EDS1. For both B and C, the indicated constructs were transiently expressed in Nicotiana benthamiana and were then immunoprecipitated using GFP-Trap beads. These experiments were all repeated three times with similar results.

We then sought to determine whether the interactions observed in yeast also occur in planta. Coimmunoprecipitation (co-IP) assays in Nicotiana benthamiana were performed. EDS1-3xHA (hemagglutinin) and PAD4-mCherry were independently coexpressed with either EDR1-sYFP or 5xMYC-sYFP as a negative control. 5xMYC-sYFP was used as a negative control because it displays a nucleocytoplasmic distribution similar to PAD4 and EDR1 but would not be expected to interact with PAD4. Both PAD4 and EDS1 were found to coimmunoprecipitate with EDR1 but not when coexpressed with 5xMYC-sYFP (Fig. 2B and C). These assays indicate that both PAD4 and EDS1 can form complexes with EDR1 in planta. Although PAD4 did not interact with wild-type EDR1 in yeast two-hybrid, we did observe a PAD4-EDR1 interaction in co-IP experiments. Based on these observations, we propose that EDR1 directly interacts with both EDS1 and PAD4.

EDR1 inhibits the interaction between EDS1 and PAD4.

The interaction between EDR1 and both PAD4 and EDS1 raised the question of whether EDR1 regulates PAD4-EDS1 heterodimer association. Formation of the EDS1-PAD4 heterodimer brings together α-helical coil surfaces in the partner C-terminal domains that are essential for basal and TIR-NLR immunity signaling (Bhandari et al. 2019; Lapin et al. 2019). To test whether EDR1 can affect this interaction, we performed a yeast three-hybrid analysis in which the kinase domain of EDR1 (EDR1-KD) was expressed as a third protein in the yeast cell under control of the methionine-regulated promoter Met25 (repressed in the presence of 1 mM methionine and induced in its absence). However, we still observed accumulation of EDR1-KD in the absence of methionine, perhaps due to leakiness of the promoter (Fig. 3). EDR1-KD expression inhibited the interaction between EDS1 and PAD4 (Fig. 3A). To test whether this effect of EDR1 was dependent on EDR1 kinase activity, we also performed the assay using EDR1-KD^ST, which is kinase-inactive. EDR1-KD^ST also blocked the EDS1-PAD4 interaction (Fig. 3A). Expression of EDR1-KD and EDR1-KD^ST had no noticeable effect on the interaction between the bacterial effector AvrB and the soybean R protein RIN4b, indicating that the effect on the EDS1-PAD4 interaction was specific. Immunoblotting demonstrated that EDR1-KD and EDR1-KD^ST accumulated in yeast to similar levels and that EDR1 expression did not interfere with the accumulation of EDS1 or PAD4 (Fig. 3B). That EDR1 kinase activity was dispensable for blocking the EDS1-PAD4 interaction suggests that EDR1 may be interfering with EDS1-PAD4 association by competing for a common EDS1 binding site rather than by phosphorylation of either protein.

Fig. 3.

EDR1 interferes with EDS1:PAD4 association. A, The EDR1 kinase domain (KD) inhibits EDS1:PAD4 interaction in a yeast three-hybrid assay. The indicated constructs were transformed into yeast strains AH109 (activation domain constructs) and Y187 (DNA binding domain and methionine promoter constructs in pBridge vector) and were then mated. Diploids were selected on –Leu, –Trp plates, then, were replated on the indicated media. Growth on –His plates indicates physical interaction between EDS1 and PAD4. Media lacking methionine induces the MET promoter. AvrB and RIN4b are positive controls for interaction. B, Immunoblot analysis confirms protein expression in yeast strains utilized in the yeast three-hybrid assay. C, Loss of EDR1 results in the upregulation of the EDS1-PAD4 network during a defense response. The edr1 only dataset is enriched for a more diverse set of biological gene ontology terms than the EDS1-PAD4 network.

edr1 plants display enhanced EDS1/PAD4 signaling during defense response.

Recently, a network of 155 core genes was demonstrated to be upregulated during the overexpression of EDS1 with PAD4 (Cui at al. 2017). Previous work has demonstrated that loss-of-function mutations in either EDS1 or PAD4 inhibit a subset of edr1 phenotypes (Tang et al. 2005). The discovery that EDR1 can interact with EDS1 and PAD4 as well as disrupt the formation of the EDS1/PAD4 complex prompted us to investigate whether EDR1 negatively regulates the EDS1-PAD4 signaling network. We have previously demonstrated that the loss of EDR1 results in the upregulation of many defense-related genes during powdery mildew infection (Christiansen et al. 2011). We found that the majority of the 155 genes that were upregulated during EDS1-PAD4 overexpression are significantly upregulated in edr1 plants relative to wild type after powdery mildew infection (Fig. 3C). A total of103 of the 155 EDS1-PAD4 upregulated transcripts were upregulated in edr1 plants during infection. This demonstrates that EDR1 has a negative impact on the induction of many EDS1-PAD4 upregulated genes during the defense response.

Gene ontology (GO) term enrichment analysis revealed that the genes belonging to both the EDS1-PAD4 upregulated and edr1 upregulated networks are enriched for processes such as SA response, response to chitin, and protein phosphorylation (Fig. 3C). Interestingly, those genes that were found to be upregulated in edr1 plants but not belonging to the EDS1-PAD4 network were enriched for a more diverse set of processes, including response to JA, ethylene, oxidative stress, hypoxia, and wounding. This correlates with the previous discovery that edr1 phenotypes are only partially suppressed by mutations in EDS1 or PAD4 (Tang et al. 2005) as well as the observation that pad4-13 enhances a subset of edr1 phenotypes (Fig. 1). These data demonstrate that EDR1 negatively regulates a broad set of defense responses that includes but is not limited to the EDS1-PAD4 network.

The S135F substitution in PAD4 does not affect protein accumulation, localization, or interaction with EDS1.

To determine the effect of the S135F substitution on PAD4 function, we investigated possible changes that could result in PAD4 overactivity. We hypothesized that an increase in the stability of the PAD4 protein caused by the S135F substitution might result in enhanced SA signaling and cell death. However, we were unable to detect an increase in the accumulation of PAD4^S135F relative to PAD4 in Arabidopsis plants undergoing a defense response elicited by the RPS4 TIR-NLR protein (unelicited plants have nearly undetectable levels of PAD4) (Fig. 4A).

Fig. 4.

The S135F substitution in PAD4 does not affect its stability, interaction with EDS1, or subcellular localization pattern. A, PAD4 protein accumulates to similar levels in wild-type Col-0, pad4-13, edr1, and double mutant Arabidopsis. Total protein was extracted from Arabidopsis rosette leaves that were either untreated or sprayed with Pseudomonas syringae DC3000(avrRps4), which induces PAD4 accumulation. B, PAD4^S135F interacts with EDS1 in a yeast two-hybrid assay. The indicated constructs were transformed into yeast strain AH109 (activation domain constructs [AD]) and yeast strain Y187 (DNA binding domain constructs [BD]) and the strains were mated, with diploids plated on the indicated media. C, The S135F mutation does not enhance the ability of PAD4^MLF to interact with EDS1 in a yeast two-hybrid assay. The indicated constructs were transformed into yeast strain AH109 (AD constructs) and yeast strain Y187 (BD constructs) and the strains were mated, with diploids plated on the indicated media. D, The S135F mutation does not increase the interaction between PAD4 and EDS1^LLIF. Constructs were expressed in Nicotiana benthamiana and protein was immunoprecipitated using anti–red fluorescent protein beads. E, PAD4^S135F displays a nucleocytoplasmic localization pattern indistinguishable from wild-type PAD4. The indicated constructs were transiently expressed in N. benthamiana and were imaged using confocal microscopy. Scale bar = 50 μM. F, PAD4-mCherry and PAD4^S135F-mCherry accumulate at similar levels without free mCherry tag. Tissue from E was subjected to immunoblotting using an antimCherry antibody. G, PAD4^S135D and PAD4^S135F both can complement a pad4-1 loss-of-function mutation. Four week-old homozygous T3 Arabidopsis plants were infected with powdery mildew. Spore counts were taken immediately following infection and 8 days postinoculation (dpi). Bars indicate the means ± standard deviation of three biological replicates per genotype. Asterisk denotes a significant difference from wild-type Col-0 at 8 dpi, using one-way analysis of variance (P < 0.0001). No other values differed significantly from wild-type Col-0 at 8 dpi, and there were no significant differences between any of the genotypes at 0 dpi.

Another possible explanation for the overactivity of PAD4^S135F is that it might have an enhanced interaction with its partner, EDS1. The EDS1-PAD4 interaction is mediated principally by conserved residues in the partner N-terminal domains, respectively, EDS1^LLIF and PAD4^MLF, which form a hydrophobic groove (Wagner et al. 2013). In an Arabidopsis EDS1-PAD4 structural model based on the EDS1-SAG101 heterodimer crystal structure (Wagner et al. 2013), PAD4^S135 is located in a loop close to but facing away from the PAD4^MLF heterodimer contact site (Supplementary Fig. S2). We therefore assessed whether the S135F substitution in PAD4 affected its interaction with EDS1 in a yeast two-hybrid assay. We observed no obvious effect on the interaction (Fig. 4B). In addition, we introduced the S135F mutation into the PAD4^MLF triple mutant, generating PAD4^MLFS. We found that the S135F mutation did not significantly enhance the weakened interaction between PAD4^MLF and EDS1 in yeast two-hybrid assays (Fig. 4C). Similarly, we observed no change in the ability of PAD4^S135F to coimmunoprecipitate with EDS1 or with EDS1^LLIF compared with wild-type PAD4 (Fig. 4D). These data indicate that the S135F mutation does not affect the ability of PAD4 to interact with EDS1.

Finally, we investigated whether the S135F mutation alters the localization of PAD4 in plant cells. Transient expression of PAD4-mCherry and PAD4^S135F-mCherry showed that both proteins displayed a nucleocytoplasmic localization (Fig. 4E). To verify that the observed localization was not the result of protein degradation, we performed immunoblotting, which also demonstrated a similar level of accumulation of the PAD4 and PAD4^S135F proteins (Fig. 4F). We thus conclude that the S135F mutation does not alter PAD4 stability, localization, or its ability to interact with EDS1 but somehow still affects PAD4 function and signaling.

Phosphorylation of PAD4^S135 is unlikely to negatively regulate PAD4 activity.

Our data indicate that EDR1 functions as a negative regulator of EDS1/PAD4 signaling. As EDR1 has been demonstrated to have kinase activity (Tang and Innes 2002), we hypothesized that EDR1-mediated regulation of EDS1/PAD4 is by direct phosphorylation. Therefore, we carried out IP mass spectrometry (MS) experiments in N. benthamiana, using transient expression of Arabidopsis PAD4, EDS1, EDR1, and EDR1^ST proteins. However, we were consistently unable to detect any phosphorylation of PAD4 or EDS1 in either the presence or absence of active EDR1. This result was repeated in three independent experiments. Importantly, the unphosphorylated S135-containing peptide was identified in all replicates, even though S135 is surface-exposed in the structural model (Supplementary Fig. S2), making it potentially amenable for phosphorylation.

Although we could not detect EDR1-mediated phosphorylation of EDS1 or PAD4 in N. benthamiana, it remains a possibility that, under specific conditions, EDR1 or some other kinase may regulate PAD4 via phosphorylation. Thus, we investigated whether the gain of function phenotype of S135F may be caused by the loss of an important phosphorylated serine residue. To test whether S135 is an important site of phosphorylation, we generated transgenic pad4-1 PAD4^S135D-MYC phosphomimic Arabidopsis. If PAD4 is indeed negatively regulated by phosphorylation at S135, then the PAD4^S135D-MYC transgene should be unable to complement the pad4-1 allele. However, we found that pad4-1 plants were fully complemented by PAD4^S135D-MYC, PAD4-MYC, and PAD4^S135F-MYC expression in resistance to powdery mildew infection (Fig. 4G). For this experiment, homozygous T3 plants were utilized. We observed a higher level of wild-type PAD4-MYC accumulation than that of PAD4^S135D-MYC and PAD4^S135F-MYC (Supplementary Fig. S3). Despite accumulating to lower levels than wild-type PAD4, the PAD4^S135D and PAD4^S135F transgenes were equally able to complement the pad4-1 mutant phenotype (enhanced susceptibility). This result demonstrates that the gain of function phenotype of S135F is unlikely to be the result of blocking phosphorylation.

DISCUSSION

Arabidopsis EDR1 acts as a negative regulator of cell death during both biotic and abiotic stress responses. Loss-of-function mutations in the EDR1 gene confer enhanced disease resistance to powdery mildew infection and more rapid senescence than wild-type plants when exposed to ethylene (Frye and Innes 1998; Frye et al. 2001; Tang et al. 2005). In this work, we report that a missense mutation in the PAD4 gene (pad4-13) that causes an S135F substitution enhances edr1-dependent cell death after pathogen attack. Moreover, the pad4-13 mutation alone confers enhanced disease resistance to the powdery mildew G. cichoracearum and accelerated cell death.

PAD4 is required for the accumulation of the signaling molecule SA (Feys et al. 2005; Jirage et al. 1999), and, thus, loss-of-function mutations in the PAD4 gene severely compromise defense against biotrophic pathogens, including powdery mildew (Gao et al. 2014). The pad4-13 mutation, in contrast, enhances resistance to G. cichoracearum, indicating that this mutation causes a gain-of-function. Moreover, this enhanced disease resistance is accompanied by enhanced cell death (Fig. 1B), similar to that observed in the edr1 mutant (Frye and Innes 1998). While the enhanced disease resistance is not additive in the edr1-1 pad4-13 double mutant, the cell death is more extensive in the double mutant than in either of the single mutants, suggesting that PAD4 and EDR1 independently regulate the cell-death pathway.

The enhanced disease resistance phenotype in both edr1 and pad4-13 without additive effects in the double mutant can be explained by both mutations causing a similar effect on SA signaling. Alternatively, PAD4^S135F might be augmenting edr1 cell death in parallel with SA, since PAD4 with EDS1 promotes both SA-dependent and SA-independent pathways in basal and TIR-NLR-mediated resistance (Bhandari et al. 2019; Cui et al. 2018). We have shown that pad4-13 does not alter PAD4 accumulation, localization, or interaction with EDS1 (Fig. 4), yet it remains unclear what effect this mutation has on PAD4. While PAD4^S135 is located close to the chief N-terminal PAD4^MLF interface with EDS1^LLIF, it is facing away from the interaction groove (Supplementary Fig. S2), consistent with the finding that the PAD4^S135F substitution does not obviously alter PAD4-EDS1 heterodimerization. It is possible that close proximity of PAD4^S135F to an α-helix of the PAD4 EP domain (Supplementary Fig. S2) creates a loosening of N-terminal restraint on the PAD4 C-terminal signaling function. Recently, it has been demonstrated that EDS1/PAD4 functions to antagonize the activity of MYC2, a master regulator of JA signaling in TIR-NLR immunity (Cui et al. 2018). It is, therefore, a formal possibility that the S135F substitution enhances the interaction between PAD4 and MYC2 or some other unknown signaling partner.

Although we could not detect an enhanced interaction between PAD4^S135F and EDS1 using a yeast two-hybrid assay, we did observe that coexpression of EDR1 with EDS1 and PAD4 inhibited the EDS1-PAD4 interaction in a yeast three-hybrid assay. Furthermore, EDR1 interacts strongly with EDS1 and PAD4 in yeast and in co-IP assays from N. benthamiana. Collectively, these observations suggest that EDR1 functions, at least in part, to negatively regulate the interaction between EDS1 and PAD4. Because formation of an EDS1-PAD4 heterodimer is essential for the rapid transcriptional reprogramming of host defense pathways in pathogen resistance (Bhandari et al. 2019), EDR1 might exert important negative control on EDS1-PAD4 signaling activity in response to infection. In support of this model, mutations in either EDS1 or PAD4 block edr1-mediated enhanced resistance and cell death (Frye et al. 2001). Furthermore, genes upregulated in the absence of EDR1 overlap significantly with genes upregulated by co-overexpression of EDS1 and PAD4 (Fig. 3C). Importantly, overexpression of either EDS1 or PAD4 alone does not upregulate these genes or enhance resistance (Cui et al. 2017), which indicates that it is the concentration of the EDS1-PAD4 complex and not their individual protein levels that determines the strength of defense signaling.

MATERIALS AND METHODS

Plant material and growth conditions.

Arabidopsis thaliana accession Col-0 and Col-0 mutants edr1-1 (Frye and Innes 1998), edr1-3 (salk_127158C), pad4-13, and edr1-1 pad4-13 were used in this study. The edr1-1 parental seed used for the suppressor mutagenesis was derived from a backcross 3 population. To confirm that the pad4-13 mutation was present in this population, we sequenced PAD4 amplified from multiple individuals of that population and found that the pad4-13 mutation was segregating within the population. To assess whether the pad4-13 mutation was present in our original edr1-1 mutant, we sequenced PAD4 in an edr1-1 M6 population (eight individual plants) that had never been backcrossed. Surprisingly, none of these plants carried the pad4-13 mutation, suggesting that the mutation had arisen spontaneously at some point during the backcrossing process. Consistent with this conclusion, an edr1-1 population being used by a former lab member in China also lacks this mutation (D. Tang personal communication).

To separate the pad4-13 mutation from the edr1-1 mutation, the double mutant line was backcrossed to wild-type Col-0 and F2 plants were screened for the presence of these mutations by amplifying the mutant regions using PCR and sequencing. Individual F2 plants that were homozygous wild-type EDR1 and homozygous mutant pad4-13 were back-crossed to wild-type Col-0 three times to remove any other unlinked or loosely linked mutations. Similarly, individual F2 plants that were homozygous edr1-1 and wild-type for PAD4 were also back-crossed to wild-type Col-0 three times to isolate the edr1-1 mutation.

Seeds were surface sterilized with 50% (vol/vol) bleach and were planted on one-half-strength Murashige and Skoog plates supplemented with 0.8% agar and 1% sucrose. Plates were placed at 4°C for 72 h for stratification and were then transferred to a growth room set to 23°C and 9 h light (150 μEm⁻²s⁻¹) and 15-h dark cycle. Seven-day-old seedlings were transplanted to MetroMix 360 (Sun Gro Horticulture) and were grown for the indicated time for each experiment. For transient expression experiments, Nicotiana benthamiana was grown under the same growth room conditions as Arabidopsis thaliana but potted in Pro-Mix PGX Biofungicide plug and germination mix.

Quantifying powdery mildew sporulation.

G. cichoracearum UCSC1 was maintained on hypersusceptible Arabidopsis pad4-2 mutant plants. Inoculation was carried out as described by Serrano et al. (2014). Briefly, four-week-old plants were inoculated using a settling tower approximately 0.8 m tall and covered with a 100-micron Nitex mesh screen. Plants with a heavy powdery mildew infection (leaves covered in white powder due to production of asexual spores) were passed over the mesh to transfer the conidiospores to the plants below. Twelve pad4-2 mutant plants were used for inoculating each tray of 60 plants. Conidiospores were counted as described by Serrano et al. (2014). Briefly, after inoculation, the conidiospores were allowed to settle for 30 min, and three leaves per genotype were harvested, weighed, and transferred to 1.5-ml microcentrifuge tubes. Then, 500 μl of distilled H₂O were added and conidiospores were liberated by vortexing 30 s at maximum speed. Leaves were removed and conidiospores were concentrated by centrifugation at 4,000 × g for 5 min. For each sample, conidiospores were counted in eight 1-mm² fields of a Neubauer-improved hemocytometer (Marienfeld). Spore counts were normalized to the initial weight of the leaves and results were averaged. The same procedure was repeated 8 dpi.

Quantifying cell death.

Staining with trypan blue was performed essentially as described by Serrano et al. (2010). Arabidopsis plants were inoculated with G. cichoracearum as described above, leaves were collected at 5 dpi and were boiled in alcoholic lactophenol (ethanol/lactophenol, 1:1 vol/vol) containing 0.1 mg of trypan blue per milliliter (Sigma) for 1 min. Leaves were then destained, using a chloral hydrate solution (2.5 mg ml⁻¹) at room temperature overnight. Samples were observed under a Zeiss Axioplan microscope. To quantify cell death, six pictures of each of five experimental repetitions were randomly selected (n = 30) and total leaf area and the trypan-stained area were measured, using ImageJ, and the percentage (area of cell death to total leaf area) was calculated. Cell death measurements are provided as means with 10th and 90th percentiles and range.

Plasmid construction and generation of transgenic Arabidopsis plants.

EDS1^LLIF and PAD4^MLF clones used in this study were derived from pENTR cDNA clones (Bhandari et al. 2019). Site-directed mutagenesis was utilized to introduce the PAD4^S135F mutation into PAD4^MLF, generating PAD4^MLFS. All primers used in this study for cloning and site-directed mutagenesis are listed in Supplementary Table S1.

For yeast two-hybrid assays, the full-length open reading frames of EDR1, EDR1 (D810A), and EDS1 were cloned into the DNA-binding domain vector pGBKT7 (Clontech Match-maker System). The full-length open reading frame of PAD4, PAD4^MLF, PAD4^MLFS, and EDS1 were cloned into the activation domain vector pGADT7. The SV40 large T antigen and lamin cloned into pGADT7 and pGBKT7, respectively, were used as negative controls.

For yeast three-hybrid assays, EDS1 and RIN4b cDNA sequences were inserted into multiple cloning site I of the pBridge vector (Clontech), using the SmaI and SalI restriction sites (separate constructs). The EDR1 kinase domain (amino acids 587 to 933) and EDR1 kinase domain substrate trap mutant form (EDR1^D810A) were cloned into multiple cloning site II of the pBridge vector using NotI and BglII restriction sites. PAD4 cDNA was inserted into the pGADT7 (Clontech) plasmid using NdeI and SmaI restriction sites. To clone AvrB into pGADT7, NdeI and BamHI restriction sites were used.

For EDR1 yeast two-hybrid experiments, EDR1 full-length wild-type cDNA and EDR1^ST (D810A) was cloned into pGBKT7 using SmaI and SalI restriction sites. EDS1 and PAD4 were cloned into pGADT7 using NdeI and SmaI restriction sites.

For transient expression in N. benthamiana, PAD4-mCherry, PAD4^S135F-mCherry, EDS1-3xHA, EDS1^LLIF-3xHA, and 5xMYC-sYFP were cloned into the cauliflower mosaic virus 35S promoter vector pEarleyGate100 (Earley et al. 2006), using a modified multisite Gateway recombination cloning system (Invitrogen) as described by Qi et al. (2012). PAD4-cYFP and EDS1-nYFP were cloned into the dexamethasone-inducible vectors pTA7001-DEST (Aoyama and Chua 1997) and pBAV154 (Vinatzer et al. 2006), respectively, using multisite Gateway cloning. EDR1-sYFP and EDR1-sYFP were also cloned into pBAV154 using multisite Gateway cloning.

Transgenic pad4-1 plants expressing PAD4-5xMYC, PAD4^S135D-5xMYC, and PAD4^S135F-5xMYC were generated using the floral dip method (Clough and Bent 1998). PAD4^S135D clones were generated using site-directed mutagenesis of PAD4 cDNA. PAD4, PAD4^S135D, and PAD4^S135F full-length cDNA tagged with 5xMYC were cloned into the pEarleyGate100 vector (Earley et al. 2006) using multisite Gateway cloning. Plasmids were transformed into Agrobacterium sp. strain GV3101 (pMP90) by electroporation with selection on Luria-Bertani plates containing 50 μg of kanamycyin sulfate per milliliter (Sigma-Aldrich) and 20 μg of gentamycin per milliliter (Gibco). Selection of transgenic plants was performed by spraying 1-week-old seedlings with 300 μM BASTA (Finale). Protein expression was verified via immunoblot using mouse anti-MYC horseradish peroxidase (HRP) antibody (Thermo Fisher).

Yeast two-hybrid and three-hybrid assays.

For yeast two-hybrid assays between EDR1 and PAD4 or EDS1, pGBKT7 and pGADT7 clones were transformed into haploid yeast strain AH109 (Clontech) by electroporation and were selected on synthetic defined (SD)–Trp–Leu medium. For yeast two-hybrid assays between EDS1 and PAD4, the full-length EDS1 open reading frame was cloned into an empty pBridge vector. Full-length PAD4, PAD4^S135F, PAD4^MLF, and PAD4^MLFS open reading frames were cloned into pGADT7. Yeast strain AH109 was transformed with pGADT7 vectors by electroporation and transformants were selected on SD–Leu. Yeast strain Y187 was transformed with pBridge plasmids by electroporation and transformants were selected on SD–Trp.

For yeast three-hybrid assays, EDR1-KD and EDR1-KD^ST were cloned into pBridge vectors, under the control of the MET25 promoter. EDS1 and RIN4b were cloned into pBridge. PAD4, PAD4^S135F, and AvrB were cloned into pGADT7. Yeast strains AH109 and Y187 were transformed with pGADT7 and pBridge, respectively.

Matings between the Y187 and AH109 strains carrying the appropriate constructs were performed in yeast peptone dextrose medium at 30°C for 16 h. Mating cultures were then diluted and were plated on SD–Trp–Leu. Before carrying out yeast two-hybrid or three-hybrid assays, yeast was grown for 16 h at 30°C. Cultures were resuspended in water to an optical density at 600 nm (OD₆₀₀) of 1.0, were serially diluted, and were plated on appropriate SD media. Plates were allowed to grow for up to 5 days at 30°C.

β-galactosidase assays.

β-galactosidase assays using ortho-nitrophenyl-β-galactoside (ONPG) were performed as described by Clontech Laboratories (2009). Diploid yeast was grown overnight in SD–Leu–Trp at 30°C. A subculture was made by adding 4 ml of fresh SD–Leu–Trp to 1 ml of the overnight culture. The subculture was grown at 30°C until OD₆₀₀ = 0.3. Cells were pelleted and were resuspended in Z buffer. A 100–μl fraction was then subjected to three cycles of freezing in liquid nitrogen and thawing in a 37°C water bath. Z buffer containing β-mercaptoethanol (700 μl) was added. Then, 170 μl of Z buffer with ONPG was added to each reaction. Samples were incubated at 30°C for up to 24 h. OD₆₀₀ and OD₄₂₀ readings were taken and β-Gal units were calculated.

IPs and immunoblots.

For total protein extraction, four leaves of infiltrated N. benthamiana were collected, were frozen with liquid nitrogen, and were ground in lysis buffer (50 mM Tris-HCl, pH 7.5, 150 mM NaCl, 1% Nonidet P-40, 1% plant proteinase inhibitor cocktail [Sigma], and 50 mM 2,2′-dithiodipyridine [Sigma]) or, for co-IP, IP buffer (50 mM Tris, pH 7.5, 150 mM NaCl, 50 mM NaF, 50 mM Na-β-glycerophosphate, pH 7.4, 2 mM EGTA, 1 mM EDTA, 0.1% Nonidet P-40, 0.1% Triton X-100, 10% glycerol, 1% plant proteinase inhibitor cocktail [Sigma], and 50 mM 2,2′-dithiodipyridine [Sigma]). Samples were centrifuged at 10,000 × g at 4°C for 5 min, and supernatants were transferred to new tubes.

IPs were performed as described previously (Shao et al. 2003), using GFP-Trap_A (green fluorescent protein) and RFP-Trap (red fluorescent protein) beads (Chromotek). Total proteins were mixed with 1 volume of 2× Laemmli sample buffer (Bio-Rad), were supplemented with 5% β-mercaptoethanol, 1% protease inhibitor cocktail (Sigma), and 50 mM 2,2′-dithiodipyridine (Sigma). Samples were then boiled for 5 min before loading. Either total proteins, immunocomplexes, or both were separated by electrophoresis on a 4 to 20% Mini-PROTEAN TGX stain-free protein gel (Bio-Rad). Proteins were transferred to a nitrocellulose membrane and were probed with anti-HA-HRP (Sigma,), anti-mCherry-HRP (Santa Cruz), mouse anti-GFP (Invitrogen), and goat antimouse-HRP antibodies (Invitrogen).

For protein extraction from yeast, yeast grown on solid –Leu, –Trp plates were resuspended in lysis buffer (100 mM NaCl, 50 mM Tris-Cl, pH 7.5, 50 mM NaF, 50 mM Na-β-glycerophosphate, pH 7.4, 2 mM EGTA, 2 mM EDTA, 0.1% Triton X-100, 1 mM Na₃VO₄). Glass beads were then added to the suspension and the solution was vortexed three times for 1 min. Samples were then boiled for 10 min. Immunoblots were performed using anti-HA-HRP (Sigma), mouse anti-GAL4DBD (RK5C1) (Santa Cruz Biotechnology), rabbit anti-EDS1 (Agisera), goat antimouse-HRP (abcam), and goat antirabbit-HRP (abcam) antibodies. Visualization of immunoblots from yeast strains used in three-hybrid assay were performed using the KwikQuant Imager (Kindle Biosciences).

Transcriptome analysis.

The edr1 dataset was based upon previously generated microarray data of edr1 plants 18 h postinoculation with powdery mildew (Christiansen et al. 2011) (Gene Expression Omnibus accession GSE26679). Upregulated genes were identified as having higher expression in edr1 plants compared with wild-type plants (P value < 0.05) using the National Center for Biotechnology Information GEO2R tool (Edgar et al. 2002). Gene identification numbers were converted to The Arabidopsis Information Resource, using the DAVID gene ID conversion tool (Huang et al. 2008). The EDS1-PAD4 dataset was based on 155 genes previously identified as being significantly upregulated due to EDS1 and PAD4 coexpression (Cui et al. 2017). Comparison of the edr1 and PAD4-EDS1 datasets was performed using the Venny 2.1 tool (Oliveros 2007). GO enrichment analysis was performed using PANTHER gene list analysis (Mi et al. 2019).

Coexpression of EDR1, PAD4, and EDS1 for MS.

To detect phosphorylation of PAD4 or EDS1 via EDR1, PAD4-mCherry and EDS1-3xHA were transiently coexpressed with either EDR1 or EDR1-ST(D810A)-sYFP in N. benthamiana. At 24 h after agrobacterium infiltration, plants were sprayed with dexamethasone to induce EDR1 and EDR1-ST expression. IP and gel electrophoresis were carried out, as noted above, using RFP-trap (Chromotek) beads. Following gel electrophoresis, PAD4-mCherry and EDS1-HA bands were visualized, using UV light, and were excised. EDS1-HA and PAD4-mCherry bands were then sent for MS analysis.

Gel bands were diced into 1-mm cubes and were incubated for 45 min at 57°C with 2.1 mM dithiothreitol to reduce cysteine residue side chains. These side chains were then alkylated with 4.2 mM iodoacetamide for 1 h in the dark at 21 °C. Proteins were digested with either trypsin, chymotrypsin, or pepsin. For the trypsin digestion, a solution containing 1 μg of trypsin; 25 mM ammonium bicarbonate was added, and the samples were digested for 12 h at 37°C. For the chymotrypsin digestion, a solution containing 1 μg chymotrypsin, in 25 mM ammonium bicarbonate was added and the samples were digested for 12 h at 25°C. For the pepsin digestion, a solution containing 0.5 μg of pepsin in 5% formic acid was added and the samples were digested for 12 h at 21 °C. The resulting peptides were desalted using a ZipTip (Millipore). The samples were dried down and were injected into an EasyNano high-pressure liquid chromatography coupled to an Orbitrap Fusion Lumos mass spectrometer (Thermo Fisher Scientific) operating in data dependent MS/MS selection mode. The peptides were separated using a 75-micron, 25-cm column packed with C18 resin (Thermo Fisher Scientific) at a flow rate of 300 nl/min. A 1-h gradient was run from buffer A (0.1% formic acid) to 60% buffer B (100% acetonitrile, 0.1% formic acid).