Natural loss‐of‐function mutation of EDR1 conferring resistance to tomato powdery mildew in Arabidopsis thaliana accession C24

Summary To screen for potentially novel types of resistance to tomato powdery mildew Oidium neolycopersici, a disease assay was performed on 123 Arabidopsis thaliana accessions. Forty accessions were fully resistant, and one, C24, was analysed in detail. By quantitative trait locus (QTL) analysis of an F2 population derived from C24 × Sha (susceptible accession), two QTLs associated with resistance were identified in C24. Fine mapping of QTL‐1 on chromosome 1 delimited the region to an interval of 58 kb encompassing 15 candidate genes. One of these was Enhanced Disease Resistance 1 (EDR1). Evaluation of the previously obtained edr1 mutant of Arabidopsis accession Col‐0, which was identified because of its resistance to powdery mildew Golovinomyces cichoracearum, showed that it also displayed resistance to O. neolycopersici. Sequencing of EDR1 in our C24 germplasm (referred to as C24‐W) revealed two missing nucleotides in the second exon of EDR1 resulting in a premature stop codon. Remarkably, C24 obtained from other laboratories does not contain the EDR1 mutation. To verify the identity of C24‐W, a DNA region containing a single nucleotide polymorphism (SNP) unique to C24 was sequenced showing that C24‐W contains the C24‐specific nucleotide. C24‐W showed enhanced resistance to O. neolycopersici compared with C24 not containing the edr1 mutation. Furthermore, C24‐W displayed a dwarf phenotype, which was not associated with the mutation in EDR1 and was not caused by the differential accumulation of pathogenesis‐related genes. In conclusion, we identified a natural edr1 mutant in the background of C24.


INTRODUCTION
Powdery mildews are able to colonize a wide variety of plant species, including Arabidopsis, and many economically important crops, such as wheat, barley and tomato. Resistance to powdery mildews can be manifested through the action of dominantly or semi-dominantly inherited resistance genes (R genes). The most abundant dominant R genes encode proteins containing nucleotide-binding site and leucine-rich repeat (NBS-LRR) domains, such as Pm3b in wheat (Yahiaoui et al., 2004), MLA alleles in barley (Seeholzer et al., 2010) and Ol-4 in tomato (Seifi et al., 2011). As a result of specific recognition of a matching pathogen-encoded avirulence protein by the cognate R-gene product, R-gene-mediated resistance is usually race or isolatespecific (Ellis et al., 2000). In Arabidopsis, the only dominant R gene characterized to date, conferring resistance to powdery mildews, is RPW8 (Resistance to Powdery Mildew 8), which is structurally different from common R genes and imparts resistance to several isolates of powdery mildew (Xiao et al., 2001).
Another form of powdery mildew resistance is governed by recessively inherited genes conferring race-non-specific resistance. Based on the resistance mechanism, they can be generally classified into three groups. Resistance in the first group is based on the loss of function of negative regulators of immune responses. An example is the edr1 (enhanced disease resistance 1) mutation in Arabidopsis, resulting in resistance to powdery mildew Golovinomyces cichoracearum and the bacterial pathogen Pseudomonas syringae (Frye and Innes, 1998). EDR1 encodes a putative mitogen-activated protein kinase kinase kinase (MAPKKK), and is considered to be a negative regulator, because edr1 resistance is caused by the activation of multiple defence responses, including increased defence gene expression and accelerated cell death response at the site of infection (Frye and Innes, 1998;Frye et al., 2001). Xiao et al. (2005) showed that EDR1 negatively regulates RPW8. The resistance phenotype of edr1 depends on the salicylic acid (SA) signalling pathway, because double mutants combining edr1 with mutations that block SA defence responses or reduce SA production reverted to susceptibility for powdery mildew (Frye et al., 2001).
The second group is defined by loss of a host susceptibility factor required for pathogen growth. In a screen for Arabidopsis mutants showing resistance to powdery mildew G. cichoracearum independent of the constitutive expression of PR1 (Pathogenesis-Related protein 1) or the formation of lesions, the pmr6 (powdery mildew resistant 6) mutant was identified (Vogel et al., 2002). PMR6 encodes a putative pectate lyase and the loss-of-function mutation causes altered cell wall composition (Vogel et al., 2002). pmr6-mediated resistance is independent of known defence responses, because mutations in genes encoding components of SA or jasmonate/ethylene pathways do not alter pmr6 resistance status (Vogel et al., 2002). Furthermore, pmr6 controls resistance to two powdery mildew species, but retains full susceptibility to unrelated pathogens, such as bacterium and oomycete species, suggesting that PMR6 may be a true powdery mildew compatibility factor (Micali et al., 2008;Vogel et al., 2002). Therefore, pmr6 probably confers resistance as a result of a loss of a susceptibility factor rather than by activation of known host defence responses.
In the third group, well-defined signalling pathways are not engaged, but resistance to unrelated pathogens is displayed. An example is dmr1 (downy mildew resistant 1), which mediates resistance to both downy mildew Hyaloperonospora arabidopsidis and powdery mildew Oidium neolycopersici (On) (Huibers et al., 2013;Van Damme et al., 2009). DMR1 encodes a homoserine kinase, and its impairment results in the accumulation of homoserine, which is responsible for the resistance to downy mildew (Van Damme et al., 2009). dmr1-mediated resistance to downy mildew might trigger a novel defence pathway because the exogenous application of homoserine still induces resistance in the single mutant impaired in immune responses or double mutants combining pmr4 (defective in the production of pathogen-induced callose) with mutations that impair SA signalling pathways (Van Damme et al., 2009).
On is a powdery mildew species causing worldwide disease on tomato. Resistance genes have been identified in wild tomato species, including six monogenic genes comprising five dominant  and one recessive (ol-2) loci, and three polygenic resistance quantitative trait loci (QTLs) (Bai et al., 2003(Bai et al., , 2005. However, to date, only the identities of ol-2, Ol-4 and Ol-6 have been (partially) revealed Seifi et al., 2011). ol-2 was shown to encode a non-functional MLO protein which causes resistance as a result of enhanced cell death response and the deposition of a callose-rich barrier (papilla) at the site of invasion. Hence, MLO is considered to be a negative regulator. Ol-4 and Ol-6 are probably Mi-1 homologues which encode NBS-LRRtype R proteins. They provide effective protection against three unrelated pests, i.e. powdery mildew, nematodes and aphids. However, neither gene has been cloned to date.
By studying 23 Arabidopsis thaliana accessions, Göllner et al. (2008) showed that RPW8 (located on chromosome 3) and polygenic resistance are major sources of resistance to powdery mildew G. orontii. In the same study, seven of these accessions were challenged with On. Intriguingly, Sha, which contains RPW8, was resistant to three powdery mildew species, but susceptible to On, implying that RPW8 is not effective against On. Furthermore, heterologous expression of RPW8 genes in tomato did not result in resistance to On (Xiao et al., 2003). These data indicate that genetic factors for On resistance in Arabidopsis are different from those involved in resistance to G. orontii.
In this study, we employed the On-Arabidopsis pathosystem to: (i) determine the mode of inheritance of On resistance in natural accessions; and (ii) identify novel Arabidopsis genes conferring resistance to tomato powdery mildew. We were mainly interested in recessive genes, because these are less likely to be NBS-LRRtype R genes and may confer isolate-non-specific resistance. Ultimately, our goal is to silence or induce mutations in tomato orthologues of Arabidopsis resistance genes to achieve On resistance in tomato. Here, we describe the map-based cloning of a recessive resistance locus in Arabidopsis, which turned out to be a natural mutation in the EDR1 gene.

Genetic analysis of On resistance in Arabidopsis accessions
To explore the natural variation for On resistance, 123 accessions (five plants per accession) were inoculated with On spores and evaluated on the basis of a disease index (DI) score ranging from '0' (resistant) to '3' (susceptible). In total, 40 accessions were fully resistant to On (DI = 0), whereas the others showed varying levels of susceptibility from low to high (Table S1, see Supporting Information). To determine the genetic mode of resistance, 19 resistant accessions were crossed with susceptible Col-0 or Sha. The F1 plants (five plants per cross) from 18 crosses displayed a susceptible phenotype (DI > 0) (Table S2, see Supporting Information). To assess whether the resistance is mediated by a single gene or more than one gene, a χ 2 test was performed on respective F2 generations (Table S2). Segregation ratios (resistant : susceptible plants or resistant : intermediate : susceptible plants) following a single gene pattern were observed in four accessions. For the remaining 15 accessions, the segregation ratios were not compatible with a single-gene hypothesis (P < 0.05), suggesting that resistance to On in Arabidopsis is mostly polygenic.

Fine mapping of QTL-1 controlling resistance to On in C24
C24 is one of the accessions exhibiting absolute resistance. It was crossed with susceptible accession Sha to generate a mapping population. The F1 plants were susceptible (Fig. 1A), and the segregation ratio of F2 plants suggested the involvement of more than one resistance gene (Table S2). Preliminary QTL analysis of 96 F2 plants with 21 indel markers (Table S3, see Supporting Information) covering all five chromosomes resulted in the identification of two QTLs with a logarithm of odds (LOD) score higher than 2.5 (Fig. 1B). QTL-1 was located on chromosome 1 and acted in a recessive manner, as only plants homozygous for the C24 allele in Class C/C is homozygous for the C24 allele, C/S is heterozygous and S/S is homozygous for the Sha allele. For each class, the DI value is the average score of F2 plants with the designated genotype for marker 159 linked to QTL-1 on chromosome 1, or for marker 515 linked to QTL-2 on chromosome 2. (D) Markers used for the fine mapping of QTL-1 (Table S4) are indicated. The distance between markers is proportional to the physical distance. White bars represent regions homozygous for the C24 allele, and shaded bars represent heterozygous regions. The space in between white and shaded bars denotes a crossover event between two flanking markers for each recombinant. The arrows point towards the interval in which QTL-1 resides. For each recombinant (REC), the phenotype of F3 and/or F4 populations is indicated. S, susceptible; SNP, single nucleotide polymorphism; R, resistant. this region were resistant (Fig. 1C). QTL-2 on chromosome 2 acted in a semi-dominant manner (Fig. 1C). To separate the effects of the two QTLs, we selected single F2 plants showing a heterozygous genotype at one QTL locus and homozygous for the Sha allele at the other locus, and selfed these to produce F3 progeny. The analysis of F3 progeny showed a tight correlation between phenotype and genotype for QTL-1 (Fig. S1A, see Supporting Information); plants (n = 37) with the C24 genotype were fully resistant or only slightly infected (with a mean DI below 0.5), whereas plants with a heterozygous (n = 25) or Sha (n = 1) genotype supported a high level of fungal sporulation (with a mean DI in the range 1-3). Surprisingly, only one plant had the Sha genotype at QTL-1, showing a skewed ratio of genotypes. For QTL-2, plants with the C24 genotype and heterozygous plants showed full resistance, except for two heterozygous plants (Fig. S1B). The DI for plants with the Sha genotype was in the range 0-2 ( Fig. S1B). Considering that QTL-1 confers full resistance and shows a tight correlation between genotype and phenotype, we focused our attention on the cloning of QTL-1.
To fine map QTL-1, two flanking markers, 159 and 162 (Table S4, see Supporting Information), were used to screen 136 F3 plants derived from a single F2 plant which is heterozygous at the QTL-1 locus and homozygous for the Sha allele at the QTL-2 locus. Two recombinants REC1 and REC2 were obtained (Fig. 1D), and examination of their responses to On narrowed down the QTL-1 region between markers 30 and 38. Recombinants were sought between these two markers by the analysis of 3552 F3 plants. Five informative recombinants (REC3-REC7) were obtained (Fig. 1D), and disease assays were performed using F3 recombinants or their F4 progenies, or both. By combining the genotypic and phenotypic data, the QTL-1 interval was reduced to a 58-kb chromosomal region between markers single nucleotide polymorphism 1 (SNP1) and SNP50 (chromosome 1 nucleotides 2 754 401-2 811 983; Fig. 1D).

A natural mutation in EDR1 confers resistance to On
The 58-kb interval between markers SNP1 and SNP50 encompasses 15 candidate genes. Interestingly, one of these is EDR1 (At1g08720). Previously, a mutated edr1 allele was obtained by γ-irradiation in the background of Col-0 (Frye and Innes, 1998). The induced mutation caused a premature stop codon in the fourth exon of EDR1. The edr1 mutant was shown to be resistant to powdery mildew G. cichoracearum. To investigate whether EDR1 is a good candidate for QTL-1, we challenged the edr1 mutant from Frye and Innes (1998) with tomato powdery mildew On. Col-0 showed clear symptoms of infection by On, whereas edr1 was free of symptoms ( Fig. 2A). Quantification of fungal DNA indicated an approximately 20-fold decrease in fungal biomass on the edr1 mutant compared with Col-0 (Fig. 2B).
These results showed that EDR1 is a good candidate for QTL-1. Examination of the protein sequences of EDR1 from C24 (accession no. EF470629) and Col-0 (accession no. AF305913) in the National Center for Biotechnology Information (NCBI) database revealed an amino acid difference (V395E) in the fourth exon in a non-conserved region of the protein. To investigate whether this difference is associated with resistance, we sequenced the coding regions of EDR1 in parental lines C24 and Sha. Surprisingly, a premature stop codon in the second exon of EDR1 was produced in C24 as a result of the loss of two nucleotides (GT) compared with the Sha allele ( Fig. 3A, genomic sequence chromosome 1, position 2 775 090-2 775 091). EDR1 is located between markers SNP1 and SNP23 (Fig. 1C). Recombinants REC3 and REC7 were homozygous for the C24 allele of both markers, and they were both resistant to On. Therefore, we expected them to contain the edr1 mutation. Sequencing results confirmed that REC3 and REC7 indeed carried the edr1 mutation (data not shown), indicating that edr1 is correlated with resistance to On. Using an F4 population derived from an F3 plant heterozygous for QTL-1 and homozygous Fig. 2 The edr1 (enhanced disease resistance 1) mutation confers resistance to tomato powdery mildew Oidium neolycopersici. (A) Fungal growth on Col-0 plant and edr1 mutant. (B) Fungal biomass quantification. Values were normalized relative to act2, and calibrated to levels on edr1 mutants. Error bars represent standard deviation of three biological replicates and, for each replicate, rosette leaves were collected. Asterisk indicates significant difference from the control according to independent-samples t-test: *P < 0.05. A representative of two experiments is presented.
for Sha alleles at QTL-2, the association between the edr1 mutation and resistance was further confirmed by phenotyping (a disease test) and genotyping with a CAPS marker that can distinguish the C24(-W) allele from the Sha allele of EDR1 (Fig. S2, see Supporting Information). Remarkably, as observed previously (Fig. S1A), a skewed segregation of genotypes was again obtained. From 96 F4 plants, 31 were homozygous for the C24-W edr1 allele, 63 were heterozygous and only two were homozygous for the Sha EDR1 allele.
Because the deletion of dinucleotide GT (nucleotides 1033-1034) was not present in EDR1 (accession no. EF470629) of C24, we suspected that our C24 germplasm (referred to as C24-W) might be different from other C24 sources. Therefore, part of the second exon of EDR1 was sequenced from plants of the stock (C24-stock) from which we obtained C24-W, and two other C24 sources (referred to as C24-H and C24-U, see Experimental procedures). Sequencing results showed that none of these C24 sources carried the mutation (Fig. S3, see Supporting Information). One possibility is that the accession we used was not C24. To exclude this possibility, we used 12 indel markers from all five chromosomes of Arabidopsis (Table S5, see Supporting Information) to genotype C24-W, C24-H, C24-U, C24-stock and Col-0, as well as the DNA of the parental plants C24 and Sha used for crossing and for the development of all the markers for mapping in the F 2 population. The resulting genotyping data showed that all C24 sources had the same marker pattern (Fig. S4, see Supporting Information), indicating that C24-W is not different from the other sources of C24. As these data may not be conclusive, we searched for a C24-specific DNA signature, a SNP of MIR164A unique to C24 (the C on chromosome 2 at position 19 520 846 is substituted Values were normalized relative to act2, and calibrated to levels in C24-U plants. Error bars represent standard deviation of three biological replicates and, for each replicate, rosette leaves were collected. Asterisks indicate significant difference from the control according to independent-samples t-test: *P < 0.05. A representative of two experiments is presented. by T in C24), which was found by analysing 96 Arabidopsis accessions (Todesco et al., 2012). polymerase chain reaction (PCR) products containing the MIR164A SNP were obtained from the genomic DNA of several sources of C24 and also Col-0. Sequencing results showed that all C24 sources, including C24-W, carried T instead of C at this position, whereas Col-0 contained the expected T (Fig. S5, see Supporting Information). Thus, we confirmed that C24-W is truly of C24 lineage.
We chose C24-W and C24-U for further analysis. A notable difference was that C24-W plants were smaller than those of C24-U (Fig. 3B). A reduced plant size is observed in a number of Arabidopsis mutants which constitutively accumulate high levels of SA (Lu et al., 2003). As an elevated level of SA induces the expression of PR genes, the basal expression levels of PR1 and PR2 in non-inoculated plants of C24-W and C24-U were compared. The results showed that transcript levels of PR1 and PR2 were not increased significantly in C24-W compared with C24-U (Fig. 3C), suggesting that the smaller size of C24-W is not caused by the accumulation of SA. As C24-U does not carry the edr1 mutation, it was expected to be less resistant than C24-W to On, which was confirmed by quantification of the fungal biomass (Fig. 3D).

Induction of PR gene expression and cell death are associated with the resistance conferred by the C24-W edr1 mutation
On pathogen infection, the edr1 mutant derived from Col-0 (Frye and Innes, 1998) showed accelerated cell death and elevated PR1 expression, although the mutant did not express PR1 constitutively. In an attempt to study the resistance mechanism conferred by the edr1 mutation in C24-W, a histological analysis was performed on On-inoculated leaves of F 4 plants homozygous for the C24-W edr1 allele and on heterozygous EDR1edr1 F4 plants. Samples were taken 3, 6 and 8 days post-inoculation (dpi). Macroscopically, no fungal sporulation was observed on F4 plants homozygous for the mutant allele (edr1edr1) (Fig. 4A) relative to heterozygous F4 plants. Instead of fungal colonies, necrotic spots appeared on inoculated leaves of edr1edr1 plants (Fig. 4A). Microscopically, fungal growth was greatly restricted after haustorium formation on the resistant F4 plants starting from 6 dpi (Fig. 4B). Cell death was observed in many epidermal cells intruded by fungal haustoria (Fig. 4B), suggesting that the resistance conferred by the C24-W edr1 allele is post-haustorial.
Next, we investigated PR1 and PR2 gene expression levels in homozygous edr1edr1 F 4 plants at 6 dpi. A significant increase in expression level (more than three-fold) of both PR genes was observed in the F4 plants homozygous for the C24-W edr1 allele compared with the heterozygous F4 plants (Fig. 5). This suggests the involvement of the SA pathway in the resistance to On conferred by the C24-W edr1 allele.

Suppression of putative homologues of EDR1 in tomato
The aim of our study was to identify genes conferring broadspectrum resistance to powdery mildews in Arabidopsis, and subsequently to investigate whether putative orthologous genes in tomato also confer resistance to powdery mildew. With the protein sequence of Arabidopsis EDR1 as a query, multiple genes showing a relatively high level of homology were found in the tomato genome database SGN (Sol Genomics Network). We chose the first two genes, Solyc01g097980 (Solyc01g) and Solyc06g068980 (Solyc06g), to investigate their involvement in resistance. Phylogenetic analysis of these two tomato genes with the protein sequences of Arabidopsis EDR1 and EDR1 sequences from other species indicated that Solyc01g is much more closely related than Solyc06g to EDR1 from Arabidopsis and other species (Fig. 6A). The protein sequences encoded by Solyc01g and Solyc06g show 56% and 45% identity with the Arabidopsis EDR1 protein, respectively, whereas they show 42% identity with each other. The protein encoded by Solyc01g (accession no. AJ005077) probably is an EDR1-like MAPKKK protein, because it is more similar to Arabidopsis EDR1 protein than to any of the five Arabidopsis EDR1 paralogues (Frye et al., 2001). Furthermore, the kinase domains of the Solyc01g-encoded protein and Arabidopsis EDR1 show 86% identity (Frye et al., 2001). Tomato cultivar Moneymaker (MM) was transformed with RNAi silencing constructs (Fig. 6B), and several primary transformants (RNAi-Solyc01g and RNAi-Solyc06g) were obtained. These were selfed to produce T2 progeny. One T2 family for Solyc01g and three for Solyc06g were obtained. Nine plants harbouring the NPTII (Neomycin Phosphotransferase II) resistance gene from each T2 family were challenged with On. All supported abundant powdery mildew sporulation, comparable with the untransformed control, as judged by visual inspection. Subsequently, three plants from each T2 family were analysed for the expression of the targeted EDR1 homologues, and the fungal biomass was quantified. Although significantly reduced expression of Solyc01g and Solyc06g was detected in both the RNAi-Solyc01g and RNAi-Solyc06g lines, respectively (Fig. 6C), the level of fungal growth in the transgenic T2 plants was only slightly decreased relative to the level in MM (Fig. 6D). We repeated the disease assay with more plants per T2 family, and included non-transgenic T2 plants for fungal biomass quantification (at 15 dpi). In this experiment, we observed that all T2 plants (transgenic and nontransgenic) showed a reduced fungal biomass relative to MM (Fig. 6E). The transgenic T2 plants T2(+) did not show a significant reduction in fungal biomass compared with the non-transgenic T2 plants T2(-), except for one T2 family in which Solyc06g was silenced. This suggests that the silencing of Solyc01g and Solyc06g separately did not result in resistance against tomato powdery mildew.

DISCUSSION
The reference species A. thaliana displays abundant genetic variation among wild accessions (Alonso-Blanco and Koornneef, 2000), which was illustrated by our results obtained after challenging 123 accessions with virulent tomato powdery mildew On. With the inoculum dosage routinely used, 40 accessions showed complete resistance (Table S1). Segregation analysis of 19 crosses in F1 and F2 (Table S2) indicated that polygenic resistance to On is more common than monogenic resistance. This observation was also made in a study of resistance to powdery mildew in multiple Arabidopsis accessions . Both observations support the notion that polygenic resistance seems to be more common in interactions of powdery mildews with Arabidopsis than with barley, and also over-represented relative to other Arabidopsis plant-pathogen interactions (Schulze-Lefert and Vogel, 2000).
In this study, we identified a natural mutation of EDR1 in the C24 background, resulting from the deletion of two nucleotides from a dinucleotide repeat array (GT)3. Eukaryotic genomes contain strings of DNA in which a single base or a small number of bases are repeated (microsatellites). Rearrangement can occur within repeated sequences, resulting in repeat addition and deletion (Flavell, 1986). This is probably caused by slippage during DNA replication (Ellegren, 2004). AC/GT repeats are scarce in plants in comparison with observations in mammalian genomes (Lagercrantz et al., 1993). Examination of dinucleotide repeats in Arabidopsis confirmed that AC/GT is least abundant (Marriage et al., 2009;Morgante et al., 2002). Marriage et al. (2009) estimated the mutation rate of dinucleotide repeats in Arabidopsis, and revealed that the majority of mutations are gains or losses of a single repeat, where the AC/GT motif is the least mutable. The mutation rate is positively affected by repeat length across motifs, but the AC/GT motif does not fit this general trend. Although meiotic and mitotic errors could not be distinguished in the study, they suggested that meiotic errors are more likely to contribute to the mutation rate. Our observation that C24-W contains a dinucleotide deletion from a microsatellite sequence, whereas all other sources of C24 do not contain this mutation, suggests that the mutation is a recent event.
It is notable that C24-W carrying the edr1 mutation exhibits reduced stature, whereas the edr1 mutant in the Col-0 background does not. Reduced stature of C24-W is not caused by the differential expression of PR genes (Fig. 3C), nor is it associated with the edr1 mutation, because progenies only segregating for QTL-1 did not show dwarfing. In contrast, progenies only segregating for QTL-2 showed size differences. Almost all plants homozygous for the C24 alleles of markers linked to QTL-2 showed reduced stature (Table S6, see Supporting Information). However, the QTL-2 region has not been fine mapped to date, and this prevents us from unravelling the mechanism underlying the dwarf phenotype.
Another remarkable observation was the skewed ratio of genotypes in both F 3 and F4 populations in which only the QTL-1 allele (results of the second experiment). Values were normalized relative to Elongation Factor 1α (EF), calibrated to levels in untransformed MM plants or non-transgenic T2 plants. Error bars represent standard deviation of at least three biological replicates and, for each replicate, third and fourth leaves were pooled. Asterisks indicate significant difference from the control according to one-way analysis of variance or t-test: *P < 0.05. is segregating. There was a shortage of plants homozygous for the Sha allele (the EDR1 allele) (Figs S1 and S2). Possibly, plants homozygous for the Sha allele in the QTL-1 region germinate more slowly than plants homozygous and heterozygous for the C24 allele. In our experiments, we sowed seeds together in one pot and, after germination, transplanted the small seedlings to individual pots. This may have resulted in the selection of seedlings only from the fast germinating seeds.
Accession C24 shows broad-spectrum resistance to several unrelated pathogen species. C24 exhibits resistance to three species of powdery mildew, i.e. G. orontii, G. cichoracearum and G. cruciferarum . Mapping of the gene(s) underlying this resistance has been unsuccessful . In addition, C24 provides downy mildew isolate-specific resistance and dominant resistance against the bacterium Pseudomonas syringae pv. tomato DC3000 (Lapin et al., 2012). Furthermore, C24 confers effective resistance against Cucumber mosaic virus mediated by a coiled coil (CC)-NBS-LRR-type protein RCY1 (Takahashi et al., 2002). Therefore, C24 seems to be an example of the natural pyramiding of different resistance loci.
Here, we showed that C24 without the edr1 mutation was less resistant to tomato powdery mildew On than C24-W carrying the mutation (Fig. 3D), but, compared with Col-0, it was less susceptible, as judged by visual inspection and after fungal biomass quantification (Fig. S6, see Supporting Information). This might be explained by elevated levels of SA, hydrogen peroxide and the expression of SA-mediated defence-related genes, such as PR1, in C24 (Bechtold et al., 2010;Lisec et al., 2008). However, these inherent traits do not necessarily contribute to pathogen resistance, because Col-0 introgression lines containing resistance QTLs to downy mildew H. arabidopsidis did not show enhanced expression of PR1 compared with susceptible Col-0 (Lapin et al., 2012).
The paradigm examples of naturally occurring loss-of-function mutations conferring resistance to powdery mildew are mlo orthologues in barley (mlo11; Jørgensen, 1992), tomato (ol-2; Bai et al., 2008) and pea (er1; Pavan et al., 2011;Humphry et al., 2011). For these species, mutations in a single MLO gene are sufficient to achieve full resistance. However, in Arabidopsis, silencing of three MLO genes (AtMLO2, AtMLO6 and AtMLO12) is required to obtain full resistance against powdery mildew. This indicates that, although the role of mlo mutation promoting resistance to different powdery mildew species is conserved, the mechanism in Arabidopsis may not be representative of the situation in other plant species. We observed that, although the Arabidopsis edr1 mutant conferred full resistance to tomato powdery mildew, silencing of two putative tomato homologues of EDR1 separately did not inhibit fungal sporulation (Fig. 6). There are at least two possibilities to explain this phenomenon. First, in the RNAi transformants, the tomato EDR1 homologues still retained a low level of expression, which may be sufficient to produce enough protein for sustained functionality. This holds true especially for Solyc01g097980, the most likely EDR1 orthologue of tomato based on phylogenetic analysis (Fig. 6A). Only one T2 family was obtained with the silencing construct for this gene and, although the level of silencing was significant, the expression of the gene was still relatively high (Fig. 6C). Additional tomato transformants showing a more severe reduction in expression of Solyc01g097980 are required to elucidate the role of this gene in susceptibility towards On. Second, the tomato EDR1 homologues may show redundancy, and the silencing of more than one gene may be necessary to obtain resistant plants.
In summary, we identified a natural mutation of EDR1 in Arabidopsis accession C24-W conferring full resistance to tomato powdery mildew On. We plan to investigate whether C24-W shows resistance to additional pathogens. Furthermore, it will be of interest to reveal the overall differences between C24-W and C24 sources not having the edr1 mutation, which may improve our understanding of the complex resistance in C24 in general. As described for the edr1 mutant derived from Col-0 (Frye and Innes, 1998), our data showed that accelerated cell death and elevated PR1 expression contribute to the resistance in the C24-W edr1 mutant. We plan to further study the allelic effects of the two edr1 mutations by comparing molecular resistance mechanisms in C24-W and Col-0-edr1.

Plant growth conditions and pathogen inoculation
All the A. thaliana accessions were obtained from the Max Planck Institute in Köln, Germany. C24-H was obtained from Hanzi He (Plant Physiology, Wageningen University, Wageningen, the Netherlands) and C24-U from Dr Guido Van den Ackerveken (Plant-Microbe Interactions, Utrecht University, Utrecht, the Netherlands). The plants were grown in soil substrate in a growth chamber with a day/night cycle of 16 h/8 h at 21°C/18°C day/ night temperature. The relative humidity was kept at 70% and the light intensity was 100 W/m 2 . The Netherlands isolate of On was maintained on susceptible tomato cultivar MM plants. Fungal spores were washed off from infected MM leaves with water and diluted to a concentration of 2.5 × 10 5 spores/mL for inoculation of Arabidopsis, or 2.5 × 10 4 spores/mL for inoculation of tomato. Approximately 30-day-old plants were inoculated by spraying spores on the leaves. The DI was recorded 8-14 days after inoculation of On: 0, no sporulation; 1, slight sporulation; 2, moderate sporulation; 3, abundant sporulation.

QTL mapping and recombinant screening
To locate resistance loci, Joinmap 4 (Van Ooijen, 2006) and MapQTL 6 (Van Ooijen, 2009) were used with default settings. For recombinant screening, DNA was extracted using the protocol described by Kasajima et al. (2004). For the development of indel markers, primers were designed based on the flanking sites of known insertion and deletion polymorphisms between Col-0 and Ler, as obtained from the Cereon database administered by Monsanto (Jander et al., 2002). For the development of SNP markers, the known SNPs between C24 and Col-0 available from the 1001 genome database (http://1001genomes.org) were examined with Lightscanner™ (Idaho Technology Inc., Bioké, Leiden, Netherlands) to determine whether the SNP was applicable to distinguish C24 from Sha.

Generation of stable silenced lines
To suppress tomato genes Solyc01g097980 and Solyc06g068980 individually, fragments with lengths of 223 and 158 bp, respectively, were amplified from MM cDNA using primers Fw-caccTCAGGTGCAGCGTTGGCTGAG and Rv-TGCCCTTTGCCACATCAAGGG for Solyc01g097980, and primers Fw-caccAGTGGATGGCCCCAGAAGTGCTG and Rv-ACGGTGCTGAAACCCCACAGCG for Solyc06g068980. The fragments were recombined into the pENTR/D-TOPO vector (Invitrogen Fisher Scientific, Landsmeer, Netherlands) and sequenced. Subsequently, the fragments were introduced into the pHellsgate8 vector (Helliwell et al., 2002) and finally transformed into Agrobacterium strain AGL1+virG. For the transformation of tomato cultivar MM, the same protocol as described by Huibers et al. (2013) was used. Primary transformants (T1) were selfed to generate T2 progeny. For each segregating T2 family, a PCR using NPTII primers (Fw-NPTII-TTCCCCTCGGTATCCAATTA and Rv-NPTII-GATTGTCTGTTGTGCCCAGT) was performed to select transgenic progeny.

CAPS marker analysis of F4 progeny of C24-W × Sha
To determine the allele composition for the Edr1 gene in F4 progeny of C24-W × Sha, primers edr1-S6F (TATCCACAGACTCCGCAAAG) and edr1-S6R (TGATTCTGCGAAAACAGCAC) were designed to amplify a fragment of exon 2 and intron 2. The 526-bp (C24-W) or 528-bp (Sha) PCR product was digested with AluI, and the fragments were separated on an agarose gel (Fig. S2)

Quantitative reverse transcription-polymerase chain reaction (RT-PCR) and data analysis
In each experiment, three biological replicates per genotype were used. Samples were prepared from Arabidopsis rosette leaves or pooled third and fourth leaves per tomato plant. For the quantification of fungal biomass, DNA or RNA was used. For the quantification of transcript levels, RNA was used. DNA was isolated with the DNeasy plant mini kit (Qiagen, Venlo, Netherlands). Total RNA was extracted using the RNeasy kit (Qiagen). After removal of DNA with DNase I (Invitrogen), 1 μg of total RNA was employed for cDNA synthesis using a Superscript II reverse transcriptase kit (Invitrogen). Quantitative real-time PCR was conducted using the iQ SYBR Green Supermix (Bio-Rad, Veenendaal, Netherlands) and the CFX96 Real-Time system (Bio-Rad). The PCR amplification consisted of an initial denaturation step of 3 min at 95°C, followed by denaturation for 15 s at 95°C, annealing and extension for 1 min at 60°C for 39 cycles, and then a final melt step from 65°C to 95°C ramp with 0.5°C increments per cycle to monitor specificity. The primers used for fungal quantification were Fw-On-CGCCAAAGACCTAACCAAAA and Rv-On-AGCCAAGAGATCCGTTGTTG. The primers used for the detection of relative transcript levels were as follows: Fw-TGAAGGAGCCAGAAAATCCA and Rv-TCTTCCCATGGAATCTCACA for Solyc01g097980; Fw-TTCATGGGAGCTG TTACTCG and Rv-ACTGATTGTTGGGTCGATGG for Solyc06g068980; Fw-EF-GGAACTTGAGAAGGAGCCTAAG and Rv-EF-CAACACCAACAGCAACAGTCT for tomato reference gene Elongation Factor 1α (Løvdal and Lillo, 2009); Fw-GAACACGTGCAATGGAGTTT and Rv-GGTTCCACCATTGTTACACCT for Arabidopsis PR1 gene At2G14610; Fw-CCCGTAGCATACTCCGATTT and Rv-AAGGAGCTTAGCCTCACCAC for Arabidopsis PR2 gene At3G57260; Fw-AATCACAGCACTTGCACCA and Rv-GAGGGAAGCAAGAATGGAAC for Arabidopsis reference gene actin (act2) At3G18780.
For the analysis of the relative expression level and fungal biomass, the 2 -ΔΔCt method, as described by Livak and Schmittgen (2001), was used. Data were statistically examined using independent-samples t-test and one-way analysis of variance (ANOVA) based on post hoc comparisons using Tukey's honestly significant difference (HSD) test (P < 0.05). All analyses were performed using SPSS Statistics 20 following the instructions of the SPSS Survival Manual, 4th edn. (Pallant, 2010).

Histological analyses
Three or four leaves per plant of the F4 progeny (C24-W × Sha) were collected at 3, 6 and 8 dpi with On. At each time point, five plants homozygous for the C24-W allele and three heterozygous plants were sampled. For 6 dpi, two homozygous plants for the Sha allele were included. The sampled leaf segments were fixed in acetic acid-ethanol (1:3, v/v) and stained with 0.03% trypan blue in lactophenol-ethanol, as described by Hering and Nicholson (1964).

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

Fig. S1
Relation between genotype of F3 progeny for quantitative trait locus-1 (QTL-1) and QTL-2 and resistance to Oidium neolycopersici. Disease index (DI) was scored for plants showing C24 (C), heterozygous (H) or Sha (S) genotypes. (A) QTL-1 region (both markers 159 and 162). (B) QTL-2 region (both markers 515 and 187). Each data point represents the average value from two time points of scoring per F3 plant. The total number of plants with the designated genotype is shown in parentheses. Fig. S2 Analysis of C24-W × Sha F4 plants segregating for quantitative trait locus-1 (QTL-1). (A) The polymorphism revealed by the CAPS marker analysis of the EDR1 (Enhanced Disease Resistance 1) alleles. In total, 96 F4 plants were genotyped, two of which were homozygous for the Sha allele (S), 31 were homozygous for the C24-W allele (C) and 63 were heterozygous (H). The plants were inoculated with Oidium neolycopersici (On) and scored for resistance or susceptibility. All resistant plants were homozygous for the C24-W edr1 allele. All heterozygous plants and the two plants homozygous for the Sha EDR1 allele showed clear powdery mildew symptoms. (B) Images of symptoms after On inoculation on F 4 plants homozygous (left) or heterozygous (right) for the C24-W edr1 allele. Fig. S3 Dinucleotide deletion in exon 2 of EDR1 (Enhanced Disease Resistance 1) is specific for C24-W. Partial sequence trace files of exon 2 of EDR1 from C24-W and C24-U. The dinucleotide deletion in C24-W is indicated. Trace files for C24-stock and C24-H are identical to that from C24-U. Fig. S4 Application of indel markers to verify the identity of C24-W. For each marker (see Table S5), seven plants were tested. Lane 1, Col-0; lane 2, C24-W; lane 3, C24-stock; lane 4, C24-U; lane 5, C24-H; lane 6, C24; lane 7, Sha; M, marker. DNA from lanes 6 and 7 was employed for the development of all the markers used for mapping in this population. Genotyping was repeated twice, and data from one replicate are presented here.

Fig. S5
Sequence of polymerase chain reaction (PCR) products containing MIR164A single nucleotide polymorphism (SNP) in different C24 sources and Col-0. Fig. S6 Fungal biomass quantification in C24-U and C24-W compared with Col-0 at 9 days post-inoculation. Values were normalized relative to act2, and calibrated to levels in Col-0 plants. Error bars represent standard deviation of eight biological replicates and, for each replicate, rosette leaves were collected. Asterisks indicate significant difference based on a t-test: *P < 0.05; **P < 0.01. Table S1 Disease index (DI) scores of Arabidopsis accessions inoculated with Oidium neolycopersici. Table S2 Segregation of resistance to Oidium neolycopersici in Arabidopsis accessions. Chi-squared tests were performed in all respective F2 generations. The P value is only shown when higher than 0.05, which means that the segregation ratio fits the indicated pattern. Table S3 Primers of indel markers for preliminary quantitative trait locus (QTL) analysis.

Table S4
Primers of chromosome 1 markers for the genotyping of recombinants to fine map quantitative trait locus-1 (QTL-1). Table S5 Primers of indel markers for the genotyping of different sources of C24. Table S6 Association of plant size of F 3 plants with markers defining the quantitative trait locus-2 (QTL-2) region. The F3 populations segregate for QTL-2, but not for QTL-1. They are homozygous for the Sha alleles in the QTL-1 region. C, homozygous C-24 allele; H, heterozygous; S, homozygous Sha allele.