Genes Encoding Recognition of the Cladosporium fulvum Effector Protein Ecp5 Are Encoded at Several Loci in the Tomato Genome.

The molecular interactions between tomato and Cladosporium fulvum have been an important model for molecular plant pathology. Complex genetic loci on tomato chromosomes 1 and 6 harbor genes for resistance to Cladosporium fulvum, encoding receptor like-proteins that perceive distinct Cladosporium fulvum effectors and trigger plant defenses. Here, we report classical mapping strategies for loci in tomato accessions that respond to Cladosporium fulvum effector Ecp5, which is very sequence-monomorphic. We screened 139 wild tomato accessions for an Ecp5-induced hypersensitive response, and in five accessions, the Ecp5-induced hypersensitive response segregated as a monogenic trait, mapping to distinct loci in the tomato genome. We identified at least three loci on chromosomes 1, 7 and 12 that harbor distinct Cf-Ecp5 genes in four different accessions. Our mapping showed that the Cf-Ecp5 in Solanum pimpinellifolium G1.1161 is located at the Milky Way locus. The Cf-Ecp5 in Solanum pimpinellifolium LA0722 was mapped to the bottom arm of chromosome 7, while the Cf-Ecp5 genes in Solanum lycopersicum Ontario 7522 and Solanum pimpinellifolium LA2852 were mapped to the same locus on the top arm of chromosome 12. Bi-parental crosses between accessions carrying distinct Cf-Ecp5 genes revealed putative genetically unlinked suppressors of the Ecp5-induced hypersensitive response. Our mapping also showed that Cf-11 is located on chromosome 11, close to the Cf-3 locus. The Ecp5-induced hypersensitive response is widely distributed within tomato species and is variable in strength. This novel example of convergent evolution could be used for choosing different functional Cf-Ecp5 genes according to individual plant breeding needs.

The pathosystem of tomato (Solanum lycopersicum) and the leaf mold pathogen Cladosporium fulvum is a well-studied model of genefor-gene interactions and plant disease resistance gene evolution (Rivas and Thomas 2005;De Wit 2016). The fungus is considered an asexual non-obligate biotroph of the Mycosphaerellaceae family that infects plants via conidia that settle on the abaxial leaf side, germinating and entering the plant through open stomata, leading to reduced respiration, defoliation and even host death (Thomma et al. 2005). The fungus abundantly secretes effector proteins in the leaf apoplast and several of these effectors can be recognized by corresponding R genes in specific tomato accessions. The tomato R genes (designated Cf genes, for resistance genes to Cladosporium fulvum) encode receptor-like proteins (RLPs) that localize to the plasma membrane and contain extracellular leucine-rich repeats (eLRRs), a membrane spanning domain, and a non-signaling short cytoplasmic domain. This system has acted as a model for investigating the structure and evolution of plant disease resistance gene loci (Thomas et al. 1998;Rivas and Thomas 2005;De Wit et al. 2012;Lin et al. 2014;De Wit 2016). To date, these genes have been mapped to just two chromosomal segments in the tomato genome. The Milky Way (MW) locus on the short arm of chromosome 1, and several genetically linked loci (like ORION; OR) containing functional Cf genes, encode a large number of genes with distinct recognition specificities including Cf-4, Hcr9-4E, Cf-9, Hcr9-9B, Cf-19, Cf-Ecp1, Cf-Ecp2, Cf-Ecp3, Cf-Ecp4, Cf-Ecp5 Thomas et al. 1997;Parniske et al. 1999;Takken et al. 1999;Haanstra et al. 2000;Panter et al. 2002;Kruijt et al. 2004;Soumpourou et al. 2007;Zhao et al. 2016). These genes were designated as Hcr9s (homologs of Cladosporium resistance gene 9). Another complex locus (Cf-2/Cf-5) encoding Cf genes has been characterized on the short arm of chromosome 6, that includes Cf-2.1, Cf-2.2 and Cf-5 (Dixon et al. 1996) with their genes having been designated Hcr2s (homologs of Cladosporium resistance gene 2) respectively (Rivas and Thomas 2005). However, the chromosomal locations of several other Cf genes, which may comprise new complex loci, have not yet been reported (Rivas and Thomas 2005;De Wit 2016).
Here, we deployed genetic mapping to investigate the genetics of the tomato hypersensitive response-type (HR) or cell death to C. fulvum extracellular protein 5 (Ecp5). Ecp5 is 115 aa long with 6 cysteine residues and is also one of the least polymorphic C. fulvum effectors (Haanstra et al. 2000;Stergiopoulos et al. 2007;De Wit et al. 2009). In contrast to C. fulvum avirulence proteins (Avrs) that are race-specific, Ecps are secreted by all C. fulvum strains and are known to play an essential role in infection, since their mutation or deletion results in decreased virulence of C. fulvum on tomato, to which Ecp5 is suspected to be no exception (Luderer et al. 2002;De Wit et al. 2009;De Wit et al. 2012;Mesarich et al. 2017). A closely related fungus to C. fulvum, Dothideomycetes septosporum contains only a pseudogenised Ecp5 homolog, despite having functional homologs of other C. fulvum effectors like Avr4 and Ecp2 (De Wit et al. 2012).
A single gene controlling the Ecp5-induced HR in tomato was previously designated Cf-Ecp5 in the line S. lycopersicum G1.1161 (introgressed from S. pimpinellifolium) and mapped at the AURORA locus (AU), proximal to MW (Haanstra et al. 2000). Additional S. pimpinellifolium and S. lycopersicum accessions appeared to also carry a single dominant Cf-Ecp5 gene and develop a HR following inoculation with recombinant potato virus X (PVX) strains expressing Ecp5 (Haanstra et al. 1999;Haanstra et al. 2000).
Presuming that all Ecp5-responding accessions carried a similar Cf-Ecp5 gene at AU, a transposon tagging strategy was deployed to isolate it from S. lycopersicum line Ontario 7522 (Ont7522). However, unexpected genetic ratios during initial crossing lead to further investigation and reassignment of Cf-Ecp5 in G1.1161 to the MW locus. Three other Cf-Ecp5 loci from S. pimpinellifolium LA0722 and LA2852, and S. lycopersicum Ont7522 were shown to be genetically unlinked to MW. We used AFLP bulked segregant analysis and mapped Cf-Ecp5 in LA0722 to the bottom arm of chromosome 7 and Cf-Ecp5 in LA2852 and Ont7522 to the top arm of chromosome 12. Differential cell death symptoms were also observed among the S. pimpinellifolium and S. lycopersicum Cf-Ecp5-carrying accessions and allelism crosses between them revealed distinct Ecp5-dependent host regulators. We screened a total of 139 domesticated and wild tomato accessions for Ecp4-and Ecp5-induced HR and identified multiple accessions carrying Cf-Ecp4 or Cf-Ecp5 genes that could be used in future studies. We found that Cf-Ecp5 was distributed in a wider geographical region than Cf-Ecp4. Lastly, we report the chromosomal locations of Cf-11, which maps to the top arm of chromosome 11, close to the previously reported Cf-3 gene (Kanwar et al. 1980). All in all, this study revealed a unique example of multiple convergently evolved tomato loci in different accessions associated with the HR to C. fulvum Ecp5 effector that have not been previously reported in plant-pathogen interactions.
A number of other stocks were used in this study for the genetic mapping of Cf-Ecp5 genes. The following lines were supplied by The Sainsbury Laboratory, Norwich, UK; S. lycopersicum variety 'Moneymaker' (MM), which contains no Cf genes (Cf0); the FT33 line (Rommens et al. 1992), which contains a T-DNA located 3 centimorgans (cM) proximal to the MW locus on the short arm of chromosome 1 and which harbors the maize transposon Dissociation (Ds) that carries the Escherichia coli uidA gene (GUS); line M18 contains an insertion of Ds in the Cf-9 gene (Ds::Cf-9) at the MW locus . Forty-nine S. pennellii introgression lines (ILs) in the S. lycopersicum M82 background (Eshed and Zamir 1994) were used for AFLP mapping of Cf-Ecp5 genes to defined chromosomal intervals as described previously (Thomas et al. 1995).
C. fulvum race 3 (CAN 84) and race 2.4.5.9.11 were obtained from Dr MHA Joosten at the University of Wageningen, Netherlands. C. fulvum race 4 GUS (Oliver et al. 1993) and race 5 were maintained and prepared for plant infections as described previously . The S. lycopersicum stock Ont7716 and near-isogenic lines (NILs) containing single introgressed Cf genes were obtained from The Sainsbury Laboratory in Norwich, UK (Tigchelaar 1984).
DNA preparation and marker sequence analysis DNA was prepared from individual tomato lines and bulked segregant pools as described previously (Thomas et al. 1995;Soumpourou et al. 2007). AFLP analysis of tomato genomic DNA was performed as described previously (Thomas et al. 1995). Cleaved amplified polymorphic sequence (CAPS) and simple sequence repeat (SSR) analyses were all performed as described previously (Soumpourou et al. 2007).

DNA gel blot analysis
Two to five mg of tomato DNA was digested at 37°for 16 h, extracted with phenol-chloroform and ethanol precipitated. DNA was electrophoresed for 20 h at 2.5 V/cm in 0.8% w/v agarose gels in a vertical gel apparatus using 40 mM Tris-acetate, pH 7.9, 1 mM EDTA as running buffer. Nucleic acids were transferred to Hybond-N membrane (Amersham) and cross-linked by irradiation with UV light. Filters were hybridized with 32 P-labeled probes at 65°for 16-20 h and washed four times for 15 min in 2x SSC (1x SSC is 0.15 M sodium chloride, 0.015 M sodium citrate) and 1% w/v SDS at 65°, and for 30 min in 0.2x SSC and 0.1% w/v SDS at 65°.
Assaying Cf-Ecp5 function by infiltration with recombinant potato virus X Cf-Ecp5 function can be assayed in several ways based on expression of the C. fulvum Ecp5 protein in plants. In a genetic test, the F 1 progeny of crosses between Cf-Ecp5 containing lines and MM plants stably expressing Ecp5 exhibit a characteristic seedling lethal phenotype. Transgenic MM lines expressing the C. fulvum Ecp5 protein (MM-Ecp5) were described previously (Soumpourou et al. 2007). Alternatively, Cf-Ecp5 function can be determined by delivering ECP5 in the form of recombinant Potato Virus X, expressing ECP5 (Soumpourou et al. 2007). Plants expressing Cf-Ecp5 exhibit systemic necrosis following virus replication and spread. Stocks of A. tumefaciens GV3101 were streaked onto Luria-Broth (LB) agar plates containing 40 mg/ml kanamycin. Colonies were selected and cultured in 10 ml LB medium containing 40 mg/ml kanamycin on a shaking platform at 28°overnight. The cultures were centrifuged at 2000 x g and the pellets re-suspended in a solution containing 1x Murashige & Skoog salts and 2% w/v sucrose. To initiate the expression of vir genes, acetosyringone (39-59-Dimethoxy-4'-hydroxy-acetophenone) was added to a final concentration of 150 mM and the bacteria were left at room temperature for 3 hr. Bacteria were infiltrated into a single cotyledon of 7-10 day old seedlings. Plants containing the corresponding Cf gene showed signs of systemic necrosis at 7-8 days post-infection (dpi) Soumpourou et al. 2007).
Marker analysis AFLP analysis of bulked-segregant pools for identifying markers linked in trans with various Cf-Ecp5 genes was performed as described previously (Thomas et al. 1995). Generation of CAPS (Cleaved Amplified Polymorphic Sequence) markers between specific haplotypes involved PCR amplification of sequences based on the tomato EXPEN2000 genetic map (http://solgenomics.net/) and sequencing of haplotype-specific products to detect sequence polymorphisms. PCR was carried out in a total volume of 20 ml containing 10 mM Tris-HCl, pH 8.0, 1.5 mM MgCl 2 , 0.2 mM dNTPs and 0.2 units Taq DNA polymerase, 50 ng genomic DNA and 1 mM of each primer. PCR products were purified using Qiaspin (Qiagen) columns and were sequenced on a 3730XL sequencer. Details of primers and CAPS markers used in this study are shown in Table S1. In reference to previous studies, genetic distances were estimated by converting recombination fractions using the Haldane function.
Bioinformatic analysis to detect LRR-encoding genes in defined regions of the tomato genome To identify candidate RLP genes for Cf-Ecp5 on tomato chromosomes 7 and 12 target regions delimited by flanking Conserved Ortholog Set II (COSII) markers were defined. For each target region protein sequences for the two markers were extracted from the TAIR website (http://www.arabidopsis.org/) and a TBLASTN search was carried out against version 2.30 of the 'Heinz 1706' tomato genome (Tomato Genome Consortium 2012). Using these results we defined target regions on chromosomes 7 and 12 which are shown in Table S7. The nucleotide sequence for each target region was extracted and sixframe translations were generated using the EMBOSS tool TRANSEQ (Rice et al. 2000). The b-sheets of Cf proteins (and other RLP and eLRR-RK proteins) contain highly conserved leucine-rich repeat motifs (LxxLxLxxNxLxGxIP) (Rivas and Thomas 2005). Using the six-frame translations, a Perl regular expression search was carried out to detect motifs with this canonical structure. From this search we found 13 motifs within the chromosome 7 target and 35 within the chromosome 12 target. Of the 48 sequences, 34 were unique. The unique sequences were then taken and using MEME a mixture model was created (Bailey and Elkan 1994) to form a search model for LRR motifs. Before using the model to search for unknown RLPs within the two target regions, the model was tested on a region of the tomato genome known to encode RLPs. The tomato resistance genes Cf-2 and Cf-5 are located on chromosome 6 ) and the nucleotide sequence for Cf-5 (AF053993) was used in a BLAST analysis of the tomato genome. Two highly homologous genes were found at the predicted location of the 'Heinz 1706' sequence between positions 2138887 -2142639 and 2161278 -2167796. An 18 kb gap separates these two genes, so the target area included the genespanning region plus an additional 18 kb on either side was used to test our model. Using FIMO (part of the MEME Suite) (Bailey et al. 2009), we then searched for LRR-like motifs in this region with a FIMO default p-value of 1e -4 . Sixty-four LRR-like motifs were identified in the target area, fifty-six within the two Cf-2/Cf-5 homologs, and eight outside it. The lowest p-value of the presumed eight false-positives outside of the Cf-2/Cf-5 homologs was 1.06e -05 . When all 64 LRR-like motifs were filtered on this p-value, we were left with 49 LRR-like motifs within the two homologs, and none outside. We then used FIMO with a q-value threshold of 1.06e -05 to identify LRR-encoding motifs within the chromosome 7 and 12 target regions and regions with consecutive blocks of LRR-encoding motifs as candidate LRR-encoding genes.

RESULTS
The Cf-Ecp5 gene from S. pimpinellifolium G1.1161 maps at MW We previously reported the mapping of two other S. pimpinellifolium genes (Cf-Ecp1 and Cf-Ecp4) at MW, where the majority of Cf genes have been mapped (Soumpourou et al. 2007). We also identified recombinants between the FT33 locus (which contains a T-DNA insertion harboring Ds:GUS and located 3 cM proximal to MW) and four Cf-Ecp genes we targeted for cloning. The observed recombination frequencies between the FT33 locus and genes located at MW (Cf-Ecp1 and Cf-Ecp4), and the Orion (OR) locus (Cf-Ecp2 being 10 cM proximal to MW), appeared consistent with their reported map locations (Haanstra et al. 1999;Soumpourou et al. 2007). The Cf-Ecp5 gene in S. pimpinellifolium G1.1161 was originally mapped at the Aurora (AU) locus, 3 cM proximal to MW, and therefore close to the FT33 locus (Haanstra et al. 2000). However, in our study the observed recombination frequency between FT33 and Cf-Ecp5 was higher than expected (3.67%) (Soumpourou et al. 2007), and similar to the recombination frequencies observed between FT33 and Cf genes at MW. This result suggested that Cf-Ecp5 is either located at MW or proximal to FT33.
An allelism test was performed where G1.1161 was crossed with line M18 which contains a stable Ds:GUS-tagged allele of Cf-9 . The F 1 plants were then test-crossed to Cf0:Ecp5, Cf0, or S. pennellii LA0716 to generate segregating populations. The resulting progeny were screened for GUS activity and either the seedling lethal phenotype (crosses to Cf0:Ecp5) or a systemic HR following agro-infiltration with PVX:Ecp5 (crosses to Cf0 or LA0716). Out of 734 progeny, no recombinants were observed between Cf-Ecp5 and the Ds::GUS-tagged Cf-9, confirming they are either allelic or very tightly linked (Table 1). This result is consistent with Cf-Ecp5 being encoded by an Hcr9 at MW in S. pimpinellifolium G1.1161 and was therefore designated Cf-Ecp5.1.
The Cf-Ecp5 gene in S. pimpinellifolium LA2852 and S. lycopersicum Ont7522 map to chromosome 12 We assumed that Cf-Ecp5 genes previously identified in other tomato accessions (LA0722, LA1689, LA2852, Ont7522) would map to the same chromosomal location as Cf-Ecp5.1 at MW (Laugé et al. 2000). Our preliminary analysis demonstrated that Cf-Ecp5 assorted independently of the FT33 locus in these accessions. To map these genes we identified S. pennellii markers linked in trans to each gene which could then be located on S. pennellii introgression lines (ILs) (Eshed and Zamir 1994). S. pimpinellifolium LA2852 and S. lycopersicum Ont7522 were each crossed to S. pennellii LA0716 and testcrossed to Cf0, to generate segregating populations. Testcross progeny were agro-infiltrated with PVX:Ecp5 and scored for Ecp5induced HR. For each cross, DNA bulked segregant pools were created from the cotyledons of forty testcross plants exhibiting either wild-type (wt) or Necrotic (N) symptoms ( Figure 1A, 1B).
The S. pennellii markers linked in trans were similar for LA2852 and Ont7522 phenotypic bulks, suggesting the two Cf-Ecp5 genes are n■ Table 1 The Cf-Ecp5 gene in S. pimpinellifolium G1.1161 is allelic or very tightly linked to the Cf-9 gene at MW. Test cross progeny were screened using a standard histological procedure for the presence or absence of GUS activity (GUS or no GUS) and for the presence or absence of Cf-Ecp5 (Necrotic or wild type respectively) based on the development of either a systemic necrosis phenotype following agro-infiltration with PVX:Ecp5 (crosses to Cf0 and LA0716) or a seedling lethal phenotype (crosses to Cf0:Ecp5)

Phenotype
Cf0 linked in trans with Cf-Ecp5 in Ont7522 and LA2852 in the chromosomal interval between IL12-2 and IL12-3. Each phenotypic bulk is shown under the "wt" and "N" labels for each accession ("wt" = wild type, "N" = Necrotic). S. pennellii ILs (12-1 to 12-4) covering the entire chromosome 12 are shown. (D) Genetic map of the chromosome 12 region at the Cf-Ecp5 locus in LA2852 (orange box) and Ont7522 (yellow box). The interval of Cf-Ecp5 in LA2852 was delimited by genotyping 48 wild-type Cf0 x LA2852 F 2 plants with three markers: C2_At5g19690 (11/96 recombinant gametes), C2_At4g26180 (7/96 recombinant gametes), and C2_At4g16710 (3/96 recombinant gametes). The Cf-Ecp5 in Ont7522 was delimited by genotyping of 56 wild-type Cf0 x Ont7522 F 2 plants with also three markers: At1g06550 (20/112 recombinant gametes), cLET_8_K4 (8/112 recombinant gametes), C2_At4g16710 (5/112 recombinant gametes). Distances are shown in megabases (Mb) and the location of each marker on the chromosome is also shown as Mb coordinates. The recombination fraction (%) between the AN locus (black box) and each marker is shown as well for both haplotypes. located in the same chromosomal region. Three AFLP markers were selected to be characterized for the Ont7522 population using 12 bulks of S. pennellii ILs (each representing an entire tomato chromosome), with all three markers mapping to chromosome 12 ( Figure S1). The chromosomal interval for Cf-Ecp5 in both LA2852 and Ont7522 was then localized at the overlap between IL12-2 and IL 12-3 ( Figure 1C).
To further delimit the chromosomal location of these genes, two sets of 200 F 2 plants from a Cf0 x LA2852 and a Cf0 x Ont7522 cross were agro-infiltrated with PVX:Ecp5 and scored for wild-type symptoms at 14 dpi. Both sets of data were consistent with a 3:1 segregation ratio for a dominant monogenic trait (Cf0 x LA2852 F 2 , 152:48 N:wt, p-value (x 2 ) = 0.816; Cf0 x Ont7522 144:56 N:wt, p-value (x 2 )=0.497). CAPS markers were developed to distinguish the haplotypes used in these crosses (Table S1) primarily using COSII markers on the target chromosome as detailed at the Sol Genomics Network (http:// solgenomics.net/). For genetic map construction, three informative markers were identified for LA2852 haplotype (C2_At5g19690, C2_At4g26180, C2_At4g16710) that flank the locus on both sides ( Figure 1D), and for Ont7522 haplotype, the three informative markers that were identified, flanked the locus on one side only (At1g06550, cLET_8_K4, C2_At4g16710) ( Figure 1D). Only one marker was useful in mapping both genes (C2_At4g16710), which is located proximal to both Cf-Ecp5 genes at a similar distance (3.1% and 4.5% recombination respectively). This data suggest that the Cf-Ecp5 genes in LA2852 and Ont7522 are located in a 888 kb region (2.126-3.014 Mb) and are either very closely linked or allelic, and therefore define a new Cf locus that we designated ANDROMEDA (AN), following nomenclature adopted in previous studies Haanstra et al. 2000). The genes in S. pimpinellifolium LA2852 and S. lycopersicum Ont7522 were then designated Cf-Ecp5.12p and Cf-Ecp5.12l respectively. The AN locus resides in a region that appears to contain many RLPs (Kang and Yeom 2018).
Cf-Ecp5 in S. pimpinellifolium LA0722 maps to chromosome 7 To investigate if Cf-Ecp5 in S. pimpinellifolium LA0722 is also located at MW, we used the same strategy described above for line G1.1161. Two hundred and ten progeny of a (M18 x LA0722) x Cf0 cross were assayed for GUS activity and then agro-infiltrated with PVX:Ecp5 to determine their Cf-Ecp5 genotype. The progeny segregated in a 1:1:1:1 ratio for all four possible phenotypic classes (51:47:60:52 GUS/N: GUS/wt: no GUS/wt: no GUS/N, p-value(x 2 ) = 0.843). This result indicates that Cf-Ecp5 in LA0722 assorts independently of MW. A second allelism test involved a cross between a line containing Cf-Ecp5.12p and S. pimpinellifolium LA0722 that revealed putative genetic elements controlling HR (Table 2). To provide additional molecular evidence of independent assortment, 200 F 2 plants from a Cf0 x LA0722 cross were agro-infiltrated with PVX:Ecp5 and 59 wild-type individuals were identified, suggesting dominant monogenic inheritance (141:59 N:wt, p-value(x 2 ) = 0.312). We genotyped 59 wild-type plants with markers linked to known Cf gene loci on chromosomes 1, 11, and 12 (Table S2), and the data were indicative of an independently assorting Cf-Ecp5 gene in LA0722 that defined a new locus, CENTAURI (CE).
To locate the CE locus, we used the previously described bulked segregant AFLP mapping strategy for Cf-Ecp5.12p and Cf-Ecp5.12l. Progeny from a (LA0722 x M18) x LA0716 cross were testcrossed to Cf0 and families segregating for Cf-Ecp5 were identified. The selected progeny were scored for Ecp5-induced HR and DNA bulks were created from 30 plants exhibiting either wild-type or necrotic symptoms. AFLP analysis was used to identify markers linked in trans with Cf-Ecp5 at CE ( Figure S2). Analysis of IL chromosome pools showed that the Cf-Ecp5 in LA0722 maps on tomato chromosome 7 in the interval defined by IL7-4 (Figure 2A, 2B). As a consequence, this genetically distinct Cf-Ecp5 gene in S. pimpinellifolium LA0722 was designated Cf-Ecp5.7. To construct a genetic map, 200 F 2 plants from a Cf0 x LA0722 cross were agro-infiltrated with PVX:Ecp5 and DNA was isolated from all individuals that exhibited wild-type symptoms at 14 dpi. These data are consistent with a 3:1 segregation ratio for dominant monogenic traits (Cf0 x LA0722 F 2 , 142:58 N:wt, p-value(x 2 ) = 0.368). We used four CAPS markers for genotyping the 58 wild-type F 2 individuals: C2_At5g14520, C2_At1g78620, C2_At3g15290, and C2_At4g26750, which delimited Cf-Ecp5.7 in a 6.1 Mb region (58.902-64.993 Mb) ( Figure 2C). This region contains many different types of nucleotide-binding domain leucine-rich repeat-containing genes (NLRs), but few RLPs, in the Solanum lycopersicum genome (Jupe et al. 2013;Andolfo et al. 2014;Wei et al. 2016;Kang and Yeom 2018).
The Ecp4-and Ecp5-induced HR is conserved in the wild tomato germplasm A previous genus-wide screen of wild tomato accessions for the presence of the well-characterized Cf-4 and Cf-9 genes demonstrated these two genes are conserved throughout the Solanum genus (Kruijt et al. 2005). Previous studies have identified Cf-Ecp5 genes in S. pimpinellifolium (G1.1161, LA0722, LA1689, LA2852) and S. lycopersicum (Ont7522) (Laugé et al. 2000). In this study, we tested different accessions from multiple species that could trigger an Ecp4-or Ecp5-induced HR (Table S3). PVX-mediated delivery of C. fulvum Ecps provides a fast and reliable screen for the presence of n■ Table 2 Unexpected suppression of Ecp5-related HR in crosses between different Cf-Ecp5-carrying accessions. Data from different populations following agro-infiltration with PVX:Ecp5. Phenotypes were recorded at 10-14 dpi. G1.1161 carries a Cf-Ecp5 gene on chromosome 1, Ont7522 and LA2852 carry a Cf-Ecp5 on chromosome 12, while Cf0 and LA0716 lack Cf-Ecp5. Population B851 lacked Cf-Ecp5.12p, while B852 and B853 carried it. In this study, LA2852 was usually crossed to Cf0 initially to form a bridged cross due to difficulties in direct crosses to other accessions. The cross, phenotypes (wild-type or Necrotic), expected ratios and p-values are shown (p-values were calculated in each cross based on the expected ratio if there was no suppression for dominant traits; '5Y' = selfing; statistical significance was calculated based on chi-square tests: ' ÃÃÃ '=P £ 0.001; ' ÃÃ '=P £ 0.01; ' Ã '=P £ 0.05) their corresponding Cf genes in wild Solanum species. Despite the low germination rate in many seed stocks, of the 139 accessions tested, Cf-Ecp5 was detected in 16 accessions of 10 different tomato species (Figure 3). Five species appeared to lack Cf-Ecp5 (S. galapagense, S. habrochaites, S. ochranthum, S. lycopersicoides and S. sitiens), but this may reflect the small sample size tested and low germination rates (Table S3). Most accessions containing Cf-Ecp5, included also non-responsive plants that exhibited wild-type PVX symptoms, indicating that they were segregating and heterozygosity might be common in these. Across all the Ecp5-responsive accessions, HR symptoms appeared variable in strength, similar to the five accessions investigated in this study (G1.1161, LA0722, LA1689, LA2852, Ont7522) ( Figure S3). Despite all of them being inherited as dominant monogenic traits, differential observations in Cf-Ecp5-triggered HR symptoms in both agro-infiltrations with PVX:Ecp5 and crosses to transgenic Cf0:Ecp5 plants in G1.1161, LA0722, LA2852 and Ont7522 were indicative of the distinctiveness of their Cf-Ecp5 genes ( Figure S3). The Ecp5-induced HR in LA2852 and Ont7522 were the strongest, followed by LA0722 and LA1689, and lastly G1.1161, which exhibited the weakest HR symptoms ( Figure S3). On the other hand, Cf-Ecp4 was less prevalent by being present in only 6 accessions from 5 species (S. lycopersicum, S. chmielewskii, S. arcanum, S. neorickii and S. pennellii) (Figure 3). These are in addition to previously identified sources of Cf-Ecp4 in S. pimpinellifolium (LA1683 and LA1689) (Laugé et al. 2000). These results showed that both Cf-Ecp5 and Cf-Ecp4 are conserved within the wild tomato (Solanum) genus.
The tomato Cf-11 and Cf-3 genes are genetically linked on the top arm of chromosome 11 S. lycopersicum line Ont7716 carries the Cf-11 gene, introgressed from S. pimpinellifolium, and its location has not been previously reported.
Ont7716 is known to carry a copy of Cf-4 as well (Enya et al. 2009), so any mapping effort required an Ont7716 progeny line carrying Cf-11 only to facilitate disease resistance phenotyping and subsequent mapping. To this end, 91 plants from a (Cf0 x Ont7716) x Cf0 population were inoculated with C. fulvum race 4 GUS, which can overcome resistance conferred by Cf-4, but not Cf-11 (Oliver et al. 1993).
Results confirmed the presence of Cf-11 (43:48 Resistant:susceptible, p-value(x 2 ) = 0.711) and resistant plants were randomly selected to be selfed. Seven selfed populations were infected with C. fulvum race 5 that cannot overcome resistance to either Cf-4 or Cf-11 (Table S4), thus any population segregating phenotypically in a 3:1 ratio would carry the Cf-11 gene only. Four F 2 families were identified that carried Cf-11 only and further selfing and progeny tests (J476) resulted in the identification of a Cf-11 homozygous plant, designated 'Cf11N' that was used as parent for mapping. However, preliminary attempts to identify molecular markers linked to Cf-11 using Ont7716 x S. pennellii F 2 population were unsuccessful due to challenges in disease phenotyping. Thus, to select a parent that will be polymorphic to Ont7716 for AFLP analysis and subsequent mapping of Cf-11, Ont7716 was analyzed by DNA gel blots with probes that detect Hcr9s and Hcr2s ( Figure S4A, S4B). From the results the Cf2 haplotype was chosen as a polymorphic parental line to be crossed with Ont7716. Resistant and susceptible F 2 bulks from a (Cf2 x On7716) x Cf0 cross were analyzed by AFLP analysis with 264 primer combinations and one marker (M1) linked in cis to Cf-11 was identified. BLAST analysis on the M1 sequence showed that the marker is located at 1.232 Mb on chromosome 11 (based on the S. pennellii genome sequence) (File S1).
To construct a genetic map, 938 progeny from J476 (Cf0 x Ont7716 background, segregating for Cf-11) were infected with C. fulvum race 4 GUS and 214 susceptible plants were genotyped with markers M1 and SSR136 (704:234 resistant:susceptible, p-value(x 2 ) = 0.979) ( Table S5). The results positioned the locus between the two markers (M1 and SSR136), while a subsequent screen with markers 0124F20 and 136SP6 confirmed this and further delimited Cf-11 to a 535 kb region (1.322-1.857 Mb) (Figure 4). The tomato genetic map indicated the S. lycopersicum Cf-3 gene is also located on the top arm of chromosome 11 (Kanwar et al. 1980). Preliminary marker analysis showed that the Cf3 haplotype was similar to Cf11N at the Cf-11 locus and therefore, these genes may be linked, or possibly allelic. The progeny of a (Cf3 x Cf11N) x Cf0 cross were inoculated with C. fulvum race 4 GUS that is restricted by either gene. If Cf-3 and Cf-11 are allelic, no susceptible progeny should have been recovered from this cross. However, a number of progeny (6 out of 288) were scored as susceptible, showing similar levels of fungal sporulation to Cf0 control plants at 14 dpi. Molecular analysis of these individuals confirmed they contained markers of the Cf-3/Cf-11 haplotype, hence Cf-3 and Cf-11 appear to be closely linked (Figure 4). The chromosomal region where both Cf-3 and Cf-11 reside appears to be devoid of RLPs in the Solanum lycopersicum genome (Jupe et al. 2013;Andolfo et al. 2014;Wei et al. 2016;Kang and Yeom 2018).

DISCUSSION
Expanding the Cf genetic map in tomato Breeding for resistance to C. fulvum has a long history, but breeding for durable resistance will depend on successful introduction of novel Cf genes from wild tomato species (Bailey 1947;Kerr and Bailey 1964). The Cf gene map has been extended in this study to include new loci on three different chromosomes ( Figure 5). The genetic mapping of Cf-Ecp5.1 to MW further highlights the importance of this locus in the tomato genome to generate effector recognition specificities Thomas et al. 1997;Takken et al. 1999;Panter et al. 2002;Yuan et al. 2002;Rivas and Thomas 2005;Soumpourou et al. 2007;De Wit et al. 2009;Zhao et al. 2016). The reassignment of Cf-Ecp5.1 at MW from AU is possibly due to the different functional analyses performed to assay Cf-Ecp5 function. In this study we used only PVX:Ecp5, while in the original study a combination of C. fulvum infections and PVX:Ecp5 were used in mapping that could have resulted in assaying two distinct, but linked fungal resistance genes (Laugé et al. 2000).
All previously characterized Cf genes on tomato chromosomes 1 and 6 encode RLPs (Wulff et al. 2001;Wulff et al. 2004). To accurately identify candidate RLP genes on the newly mapped tomato loci CE, Cf-11/Cf-3 and AN on chromosomes 7, 11 and 12 respectively, a high quality genome sequence is needed from each accession, as Cf loci are highly polymorphic (Wulff et al. 2009). In an early attempt, we analyzed the S. lycopersicum 'Heinz 1706' tomato genome sequence (Tomato Genome Consortium 2012) and the S. pimpinellifolium LA1589 draft genome sequence (Tomato Genome Consortium 2012) in regions delimited by the flanking markers for each locus in this study, retrieving some informative data only for 'Heinz 1706' (Table S7). However, recent studies have identified candidate R genes in the regions covering the novel Cf loci reported in this study (Table  S7) (Jupe et al. 2013;Andolfo et al. 2014;Wei et al. 2016;Kang and Yeom 2018). The CE and Cf-11/Cf-3 loci, on chromosomes 7 and 11 respectively, reside in regions poor in RLPs (Kang and Yeom 2018), Figure 3 Geographic distribution of Ecp4-and Ecp5-induced HR on the north-western area of South America. The distribution of Ecp4-(white script) and Ecp5-responsive (yellow script) accessions following an agro-infiltration of seedlings with PVX:Ecp4 and PVX:Ecp5 respectively on a collection of 139 wild tomato accessions is shown. Accessions responding to both Ecp4 and Ecp5 are shown in orange script. The figure also includes S. pimpinellifolium accessions tested in a previous work (Laugé et al. 2000), which were used for mapping in this study (black highlight). The tomato species used in this screen are also shown on the left, clustered in groups that reflect their approximate phylogenetic relationship as established previously (Kruijt et al. 2005;Rodriguez et al. 2009). The approximate distribution of Ecp4-and Ecp5-responsive accessions is shown as white and yellow dashed colored borders respectively, along with the distribution of Cf-9 alleles (blue colored border) from a previous study (Kruijt et al. 2005). The Cf-Ecp5 genes that were mapped in this study are shown along with their accessions on the top right. The responsive accessions are ordered by geographic latitude and their location was estimated by using the TGRC and CGN database in conjunction with "Google Maps" online service (overview of map on the top right). Source material for breeding lines Ont7522 and G1.1161 were PI124161 and CGN15529 respectively, for which their collection sites are used in this map to locate the origin of their corresponding genes (Kanwar et al. 1980;Laugé 1999;Laugé et al. 2000).

Figure 4
Cf-11 and Cf3 loci are linked on the top arm of chromosome 11 in tomato. Genetic map of the chromosome 11 region at the Cf-11 locus in Ont7716 (blue box). The interval was delimited by genotyping 201 susceptible plants from a population of 831 J476 progeny (Cf0 x Ont7716 BC 3 ; following infection with C. fulvum race 4 GUS) with two markers, 0124F20 and 136SP6, and combining the data with previous results with another two markers (M1 and SSR136). The Cf-11 locus showed 3.9% recombination with marker 0124F20 (16/402 recombinant gametes) and 1.5% with 136SP6 (6/402 recombinant gametes). Distances are shown in megabases (Mb) and the location of each marker on the chromosome is also shown as Mb coordinates. The recombination fraction (%) between the Cf-11 locus (black box) and each marker is shown, along with the hypothetical distal and proximal locations of the Cf-3 locus (gray box) with 2.1% on either side of the Cf-11 locus.
while the AN locus resides in a hotspot for RLPs similar to the Milky Way locus on chromosome 1, which is confirmed by both our analysis (Table S7) and another study (Kang and Yeom 2018). It will be interesting to know whether all these novel Cf genes encode RLPs or something else.

The multiplicity of Ecp5 responses in tomato species
In some plant-microbe interactions, two or more loci may control the resistance phenotype in an extension of Flor's gene-for-gene hypothesis (Flor 1947;Bourras et al. 2016). However, these genes do not convergently evolve in different chromosomal locations, with the exception of the potato Rpi-blb3 and Rpi-mcq1 genes to date (Aguilera-Galvez et al. 2018), which exhibit differential effectorspecific resistance to Avr2-carrying Phytophthora infestans isolates (Aguilera-Galvez et al. 2018). This distinctiveness of these two potato R genes recognizing Avr2 has been attributed to independent evolution of the underlying recognition mechanism, since the genes originate from Mexico (Rpi-blb3) and from Peru (Rpi-mcq1) (Aguilera-Galvez et al. 2018). Avr2 is also a highly diverse effector within Phytophthora species (Vleeshouwers et al. 2011), which could facilitate convergent evolution of two R genes through overlapping recognition specificities.
In this study, from the mapping attempts on only four accessions, four Cf-Ecp5 loci were mapped on three different chromosomes, exhibiting differences in HR strength in both PVX:Ecp5 agro-infiltration assays and crosses to Ecp5-expressing transgenic tomatoes ( Figure S3), while allelism crosses between different Cf-Ecp5-carrying accessions revealed putative suppression elements that point to distinct evolution of effector recognition or defense activation mechanisms in tomato (Table 2). Crosses between S. pennellii LA0716 and S. pimpinellifolium LA2852 or S. lycopersicum Ont7522 resulted in suppression of the Ecp5-induced HR at the F 1 generation, while the same phenotype was also observed in a S. pimpinellifolium G1.1161 · (Cf0 x LA2852) background ( Table 2). The most interesting example of Ecp5-induced HR suppression was observed in a cross between S. lycopersicum Ont7522 and F 1 progeny from Cf0 x S. pimpinellifolium LA2852, in which both accessions carry a Cf-Ecp5 gene on chromosome 12 at AN locus (Table 2). We genotyped 39 randomly selected wild-type plants to investigate if the Ecp5-induced HR suppression assorts independently of the AN locus on chromosome 12. The results indicated that the genetic elements responsible for this Cf-Ecp5-induced HR suppression were genetically unlinked to the AN locus and accession-specific for each Cf-Ecp5 gene (Table S6). These data suggest that these Cf-Ecp5 genes have evolved distinct HR-regulating elements.
Just like other characterized Cf genes, Cf-Ecp5s possibly also encode RLPs, which would lack any intracellular signaling domain and they may interact with the receptor-like kinase (RLK) SOBIR1 (Liebrand et al. 2013), or a similar RLK that will stabilize them, and require the RLK BAK1 for downstream immune signaling (Van Der Burgh et al. 2019). However, functional tests will be required to determine if Ecp5 is perceived by functionally similar/distinct CfEcp5 proteins and whether they all require SOBIR1 and BAK1 for defense activation. The suppression phenotypes of Ecp5-dependent HR in different Cf-Ecp5-carrying accessions (Table 2) may also help to elucidate this question.
Ecp5 is the most monomorphic effector reported in C. fulvum, having only one variant with a single polymorphic nucleotide in its single intron (Stergiopoulos et al. 2007;De Wit et al. 2009), which raises an interesting question as to how these different accessions evolved genetically-distinct Ecp5 recognition specificities. All Ecp5responding accessions originated from different geographic regions ( Figure 5) (Kanwar et al. 1980;Laugé 1999;Laugé et al. 2000;Rivas and Thomas 2005;De Wit et al. 2009) and whether geographic origin plays a role in the evolution of strong responses or not remains to be seen. The geographic distribution of Ecp5-responding accessions appeared much wider than either Cf-Ecp4 or Cf-9 (Figure 3), which could be explained by the convergent evolution of multiple distinct Cf-Ecp5 loci under different environmental conditions or low selection pressure (Stergiopoulos et al. 2007;Bolton et al. 2008). Another interesting question is how many distinct Cf-Ecp5 loci and types of Ecp5-responses exist throughout tomato species that were not screened in our study (Figure 3) (Table S3). If Cf-Ecp5 genes are to be incorporated into breeding programs, their allelic and variant interactions need to be studied further in different backgrounds and with multiple C. fulvum strains.
Do effector-specific R genes emerge and evolve anywhere? Cf genes have been repeatedly bred into cultivated tomato from wild relatives such as S. pimpinellifolium, S. hirsutum and S. peruvianum (Atherton and Rudich 1986) and according to a recent study (Kang and Yeom 2018) most of these types of genes (RLPs) are located on tomato chromosomes 1 and 12. Some regions appear to be the source of major Cf loci, like MW, which are able to generate multiple Figure 5 The expanded tomato Cf gene map. Schematic representation of tomato chromosomes 1, 6, 7, 11 and 12 as genetic maps that include all Cf genes mapped to date. Loci are shown relative to the physical map of each chromosome (based on the tomato physical map at Sol Genomics Network). Heterochromatic regions are shown as black boxes on each chromosome. Loci are shown as white script on black background with all their known Cf genes underneath. Scale is shown in megabases (Mb). effector-specific RLPs (Rivas and Thomas 2005;De Wit et al. 2009). Although distinct recognition of multiple effectors by one or more R proteins is known (Mackey et al. 2002;Ma et al. 2018), the opposite has only been observed as convergent evolution in different species (Ashfield et al. 2014;Saucet et al. 2015;Carter et al. 2018 preprint) with only one recent example in closely related species (Aguilera-Galvez et al. 2018). However, considering the monomorphic nature of Ecp5 and level of genetic complexity observed among Cf-Ecp5 genes in this study, we have to address this slight paradox in terms of evolution. There are hypotheses that might be able to explain the Cf-Ecp5 evolution, since conventional conservation of a Cf-Ecp5 allele through natural selection (Mauricio et al. 2003) is unlikely to be the source of four genetically distinct genes, nor are the putative overlapping specificities of distinct R proteins for a highly variable effector like Avr2 (Gilroy et al. 2011;Aguilera-Galvez et al. 2018), as Ecp5 is not.
The first hypothesis involves small-scale genomic duplications following unequal crossover events or other chromosomal anomalies that have allowed the dispersal of an original cluster harboring the ancestral Cf-Ecp5 gene throughout the tomato genome prior to speciation (Baumgarten et al. 2003;Meyers et al. 2005;Rensing et al. 2008;Flagel and Wendel 2009). The second hypothesis involves transposons and other reverse-transcriptase-mediated duplication that can lead to 'macro-transposition' of Cf-Ecp5 genes or their clusters (Freeling et al. 2008;Huang and Dooner 2008;Xiao et al. 2008). The third hypothesis for the Cf-Ecp5 evolution could be sought in the mechanisms involved in plant R gene creation, which is the moment pseudogenes or paralogues become fully functioning resistance genes capable of activating defenses. A critical factor in addressing this is probably the long-term effects of biotic stress on DNA evolution and R gene loci structure, which have only been investigated partially and indirectly (Kovalchuk et al. 2003;Biémont and Vieira 2006;Boyko et al. 2007;Boyko et al. 2010). Nevertheless, sequencing and bioinformatic analysis of all Cf-Ecp5 genes and their flanking sequences will uncover the nature of these R genes and help to understand their function.

CONCLUSIONS
In plant pathology, it is important to consider the underlying complexity of each pathosystem and its gene-for-gene interactions. The discovery of novel Cf loci in this study and expansion of the Cf repertoire with loci of variable HR strength (Cf-Ecp5s) has potential implications for breeding plant disease resistance. Vertical resistance in tomato can be defeated quickly as strains of C. fulvum that overcame several Cf genes have been reported already (Kerr and Bailey 1964;Luderer et al. 2002;Enya et al. 2009). C. fulvum secretes approximately 70 apoplastic proteins in planta (Mesarich et al. 2017), from which a significant number act as effectors and are recognized in wild tomato accessions, making it very likely that a large number of novel Cf genes can be exploited for breeding purposes in the future. Considering that Ecp5 is a core effector in C. fulvum and has no natural variability, studying the differences between distinct Cf-Ecp5 genes in the tomato germplasm can facilitate our understanding on effector-or strain-specific recognition and defense activation, while allowing us to choose or engineer optimal Cf gene variants that have the strongest possible defense responses to different C. fulvum strains with minimal cost-to-fitness. To identify the causative genes at each locus and facilitate their introgression in elite breeding lines, functional screens will be required for any candidate gene within the genomic regions of interest that has structural similarities to RLPs (Rivas and Thomas 2005). Such screens can include co-agroinfiltration of the candidate Cf-Ecp5 gene and Ecp5 in tobacco, N. benthamiana, Cf0, or agroinfiltration of the candidate gene in Cf0:Ecp5 transgenic tomatoes. Silencing SOBIR1 and BAK1 in each CfEcp5 line will provide further insights about each Cf-Ecp5 protein's downstream function like other Cf proteins or not. As for the unlinked suppressor loci, a genotyping-by-sequencing strategy of suitable phenotypic bulks (Table 2; Necrotic vs. wild type) can map them and facilitate their subsequent isolation and characterization, identifying novel genes in downstream defense activation. The commercial deployment of Cf-Ecp5 can force C. fulvum to lose Ecp5 from its secretome, but Ecp59s lack of allelic variation could reflect selective restraints by the pathogen to maintain full virulence (Mesarich et al. 2017). Furthermore, the simultaneous stacking of different variants of Cf-Ecp5 genes into one cultivar could extend the duration of resistance to C. fulvum and constitute a model for horizontal resistance in other plant pathosystems in the future (Dangl et al. 2013).

ACKNOWLEDGMENTS
We thank James Brown, Jonathan Jones, Brande Wulff, Mark Coleman, and Farid El Kasmi for advice and suggestions. Both M.I. and E.S. were supported by the Greek State Scholarships Foundation (IKY) awards for postgraduate study. Research with recombinant PVX was authorized under DEFRA license PHL 170/5467.