A Suppressor/Avirulence Gene Combination in Hyaloperonospora arabidopsidis Determines Race Specificity in Arabidopsis thaliana

The pathosystem of Arabidopsis thaliana and diploid biotrophic oomycete Hyaloperonospora arabidopsidis (Hpa) has been a model for investigating the molecular basis of Flor's gene-for-gene hypothesis. The isolates Hpa-Noks1 and Hpa-Cala2 are virulent on Arabidopsis accession RMX-A02 whilst an F1 generated from a cross between these two isolates was avirulent. The F2 progeny segregated 3,1 (avirulent, virulent), indicating a single major effect AVR locus in this pathogen. SNP-based linkage mapping confirmed a single AVR locus within a 14 kb map interval containing two genes encoding putative effectors. The Hpa-Cala2 allele of one gene, designated H. arabidopsidis cryptic1 (HAC1), encodes a protein with a signal peptide and an RxLR/dEER motif, and triggers a defense response in RMX-A02. The second gene is heterozygous in Hpa-Cala2. One allele, designated Suppressor of HAC1Cala2 (S-HAC1Cala2) encodes a protein with a signal peptide and a dKEE motif with no RxLR motif; the other allele (s-hac1 Cala2) encodes a protein with a signal peptide, a dEEE motif and is divergent in sequence from the S-HAC1Cala2 allele. In selfed progeny from Hpa-Cala2, dominant S-HAC1Cala2 allele carrying progeny correlates with virulence in RMX-A02, whereas homozygous recessive s-hac1Cala2 carrying progeny were avirulent. Genetic investigations suggested other heterozygous suppressor loci might exist in the Hpa-Cala2 genome.

The biotrophic oomycete Hyaloperonospora arabidopsidis (Hpa, formerly known as H. parasitica) has co-evolved as a downy mildew pathogen in wild populations of Arabidopsis thaliana (Göker et al., 2004;Holub, 2006), and has been used for more than 25 years as an experimental model (Koch and Slusarenko, 1990;Tör et al., 1994) for investigating the molecular basis of the gene-for-gene theory proposed by H. H. Flor to explain genetic evidence that for each gene controlling resistance in the host, there is a corresponding gene controlling pathogenicity in the pathogen (Flor, 1946). At least seven downy mildew R-genes have been characterized in this pathosystem (Botella et al., 1998;McDowell et al., 1998;Bittner-Eddy et al., 1999;van der Biezen et al., 2002;Sinapidou et al., 2004;Eulgem et al., 2007), and all encode cytoplasmic NLR proteins that contain a conserved nucleotide-binding site (NBS) and a variable leucine-rich repeat (LRR) domain. Three of the seven predicted AVR elicitors have also been characterized (Allen et al., 2004;Rehmany et al., 2005;Bailey et al., 2011) and all encode secreted effector-like proteins.
Plant pathogens deliver effectors to establish infection by suppressing the host immune system, altering plant physiology, and colonizing the host cell (Doehlemann and Hemetsberger, 2013;Win et al., 2013). Current consensus suggests that effectors target apoplastic and cytoplasmic components (Shan et al., 2008;Dodds and Rathjen, 2010;Pedersen et al., 2012). Fungal and oomycete pathogens secrete effectors with the involvement of an N-terminal signal peptide (Kamoun, 2006;Song et al., 2009), and some studies of oomycetes suggest that RxLR and dEER motifs may be involved in the translocation of these effectors into the host cell (Whisson et al., 2007;Tyler et al., 2013). However, a recent proteomic investigation on the RxLR effector AVR3a from the potato pathogen P. infestans demonstrated that the Nterminal part of the native protein up to and including the RxLR motif but excluding the dEER motif is cleaved off before the secretion, indicating a possible role of the RxLR motif in secretion of the effector from the pathogen (Wawra et al., 2017). Effectors that elicit a defense response due to detection by an R-protein are referred to as avirulence elicitor proteins (Jones and Dangl, 2006).
Interestingly, H. H. Flor described an exception to the simplified "gene-for-gene" model in which avirulent progeny of the flax rust fungus Melampsora lini were derived from sexual mating of two virulent isolates. He used selfing, back-crossing, and inter-mating of progeny to verify that several avirulence genes each had a matching suppressor/inhibitor gene. Lawrence et al. (1981) subsequently confirmed these results in the flax rust pathosystem. A similar example was described from detailed genetic studies in the rice blast fungus Magnaporthe grisea (Ellingboe, 1992;Lau et al., 1993). In some cases, the suppressor and the avirulent determinant were tightly linked on the same chromosome and could not be separated by recombination (Lau et al., 1993). Recent studies have begun to reveal the molecular basis of such interactions in other fungi. For example, Avr1 of Fusarium oxysporum f.sp. lycopercisi is recognized by the tomato R-gene product I-1 but suppresses the recognition of Avr2 and Avr3 by I-2 and I-3, respectively (Houterman et al., 2008). Current studies suggest that the suppressor/avirulence gene combination provides the basis of specificity in fungi and oomycetes (Bourras et al., 2016).
Whilst genetic studies into Arabidopsis-Hpa interactions have shown the existence of a gene-for-gene relationship, molecular, and bioinformatics investigations have helped the cloning of several avirulent determinants in the pathogen and the corresponding R-genes in the host Holub, 2006;Baxter et al., 2010;Bailey et al., 2011). Some of the Hpa RXLR-type effector proteins have been molecularly characterized. For example, ATR1 (Rehmany et al., 2005) is recognized by RPP1 by direct association via C-terminal leucine-rich repeats (Steinbrenner et al., 2015). ATR13 (Allen et al., 2004) is a highly polymorphic and dynamic protein with two surface-exposed patches of polymorphism, only one of which is involved in specific recognition of RPP13-Nd . ATR39-1 has been identified by computational methods and is recognized by RPP39 in Arabidopsis accession Weiningen (Wei-0) (Goritschnig et al., 2012). Recently we cloned ATR5 Emoy2 , which does not have the RxLR motif but contains a dEER motif (Bailey et al., 2011). All of these Hpa effectors are avirulence determinants and, until now, no suppressor has been identified for any of them. Here, we report the genetic identification and map-based cloning of a H. arabidopsidis cryptic (HAC1 Cala2 ) avirulence determinant and a predicted matching suppressor (S-HAC1 Cala2 ) from the same interval using a sexual cross between two virulent isolates of H. arabidopsidis. HAC1 Cala2 is an RxLR-dEER type putative effector protein inherited in progeny from the parent isolate Hpa-Cala2 and can trigger an immune response in the A. thaliana accession RMX-A02. S-HAC1 Cala2 is in the same parent isolate within the genetic interval next to the HAC1 Cala2 gene and encodes a putative effector protein with a dKEE motif. S-HAC1 Cala2 is epistatic over HAC1 Cala2 .

Pathogen Isolates and Pathology Methods
All isolates of H. arabidopsidis were maintained on Ws-eds1 (Parker et al., 1996). Generation of a cross (CaNo F 1 ) between Hpa-Cala2 and Hpa-Noks1 was described previously (Bailey et al., 2011). Preparation of inoculum for experiments, and the assessment of sporulation were as described previously in Tör et al. (2002).
selected CaNo F 2 isolates. DNA was extracted separately from each individual F 2 isolate, and DNA from fifteen virulent (hac1/hac1) and fifteen avirulent (HAC1/±) F 2 isolates was pooled in equal concentrations to make up the virulent and avirulent bulks, respectively. One lane of 100 bp paired-end Illumina HiSeq2500 sequencing data was generated from each bulked pool, comprising 104 million reads for the virulent bulk and 110 million reads for the avirulent bulk. The Illumina reads were first trimmed based on their quality scores using Btrim (Kong, 2011) with a cut-off of 25 for average quality scores within a moving window of 5 bp. The minimum acceptable read length was 25 bp (that is, reads that were shorter than 25 bp after trimming were discarded). Other parameters for Btrim were set to default values. SPAdes v. 3.6.1; SSPACE v. 3.0 were then employed to de novo assemble both Hpa-Cala2 and Hpa-Noks1 genome sequences. We then Used BWA-mem to do the alignment of reads against the Hpa-Noks1 genome assembly as a reference to map sequence reads from both virulent and avirulent bulks and the frequency of major allele at 26,722 sites have been examined. The contig 23137 (GenBank, LLKM01000918) had the highest percentage of SNP hit representing the largest difference between the bulks and used for subsequent mapping.

Map-Based Cloning of HAC1 and S-HAC1
SNPs identified between the virulent and avirulent bulks using the genomic contig 23137 from Hpa-Noks1 as a reference were converted to CAPS markers using dCAPS (http://helix.wustl. edu/dcaps/dcaps.html) (Neff et al., 2002). They were then used to map the HAC1 locus. Once linkage was confirmed, flanking markers were used to identify the Hpa-Emoy2 genomic Scaffold 41, and were also used for screening further F 2 isolates to identify recombinants. A total of 190 F 2 isolates were screened with the molecular markers. Recombinant CaNo F 2 isolates from either side of the interval were used to narrow the interval down to 14 kb on the Scaffold 41, 298,000-312,000. This region was checked on Ensembl Protist site (http://protists.ensembl.org) for possible ESTs. There were two gaps in the interval and both were filled manually by long PCR and sequencing the products. All PCR amplifications for mapping were performed in a total volume of 20 µl containing 20 ng of genomic DNA, forward and reverse primers (Supplemental Table 1) each at 0.2 µM, BioMix Red (Bioline). The PCR reaction consisted of a first step at 94 • C for 3 min followed by 35 cycles of 30 s denaturation at 94 • C, 30 s annealing at 50-60 • C (based on T m of primers) and 1 min extension at 72 • C. Finally, an extension step was carried out at 72 • C for 5 min. A 3 µl sample of each reaction volume was loaded onto a 1.5% agarose gel to ascertain whether PCR amplification was successful. The remaining 10-15 µl of PCR reactions were digested with relevant restriction enzymes following manufacturer's instructions. Digested products of PCR amplicons were separated on a 2% agarose gel containing TAE buffer at 110 V for 2 h, and visualized under UV light after staining with GelRed.

Expression Analysis
Total RNA was isolated from uninfected (control) or infected with Hpa-Cala2, Hpa-Noks1, Hpa-Emoy2, at 1, 2, 3, 4, 5, 6, and 7 day post inoculation (dpi) using RNeasy plant mini kit (Qiagen Ltd., West Sussex, UK) according to manufacturer's instructions. RT-PCR was performed with gene specific or Hpa-Actin primers (Supplemental Table 1), using 0.5 µg of total RNA as template. The PCR product was separated on a 1.5% agarose gel and stained with ethidium bromide. The products were verified by sequencing.

RACE and DNA Sequencing
Total RNA was isolated from infected plant materials using RNeasy plant mini kit (Qiagen Ltd., West Sussex, UK) according to manufacturer's instructions. RACE was performed using the GeneRacer kit (Life Technologies Ltd, UK), following manufacturer's instructions using gene specific and nested primers (Supplemental Table 6). PCR regime and subsequent cloning was carried out as described by Bailey et al. (2011). PCR products or plasmid clones were sequenced using Mix2Seq Kit by Eurofins Genomics (Wolverhampton, UK).

Quantification of Pathogen Biomass and High-Resolution Melt Curve Analysis
Samples for genomic DNA isolation were collected from infected seedlings 7 dpi. Fifteen seedlings from each genotype made up a sample and three replicates were used for each sample. DNA was isolated using CTAB method (Doyle, 1987) and the PCR was performed in a total volume of 25 µl containing 50 ng of gDNA, 12.5 µl of SybrGreen Mastermix (ABI, Carlsbad, California), Hpa-Actin or At-Actin primers (Supplemental Table 1) and water on a Roche LightCycle 480 device. PCR conditions were as follows, 95 • C for 4 min, then 10 cycles touchdown of 95 • C for 30 s, annealing temperature of 65 • C, decreasing 1 • C every cycle to 56 • C, and extension at 72 • C for 30 s. After 10 cycles of touchdown, a further 25 cycles of 95 • C for 30 s, 60 • C for 30 s and 72 • C for 30 s and a final extension at 72 • C for 5 min were carried out. Relative abundance of Hpa-Actin to At-Actin was calculated as 2 ∧ -dCt (Livak and Schmittgen, 2001). High-resolution melt curve analysis (HRM) was carried out using gene specific primers for s-hac1 Cala2 (Supplemental Table 1). The 20 µl reaction mix consisted of 30 ng of DNA, 10 µl HRM mix (Bioline Sensifast), 250 nM of each primer and sterile distilled water. PCR conditions were as follows, 95 • C for 2 min, followed by 8 cycles of 95 • C for 5 s, 65 • C for 10 s, 72 • C 25 s; then 37 cycles of 95 • C for 5 s, 58 • C for 10 s 72 • C for 25 s. This was followed by a melt cycle as follows, 95 • C for 15 s, 40 • C for 15 s, 70 • C for 1 s, then increasing to 95 • C with a ramp rate of 0.2 • Cs −1 with 25 fluorescence data acquisitions per • C.

Light Microscopy
Seedlings of infected and non-inoculated controls were stained with a solution of phenol, lactic acid, glycerol, and water (1,1,1,1) supplemented with 1 mg/ml trypan blue, decolourised in chloral hydrate and visualized under a compound microscope according to a previously described method (Koch and Slusarenko, 1990).

Agrobacterium-Mediated Transformation
Full-length HAC1, S-HAC1, and s-hac1 candidate alleles were amplified from Hpa-Cala2 and Hpa-Noks1 genomic DNA using gene specific primers and cloned into pDONR/zeo vector (Life technologies) by Gateway cloning technology. Candidate genes were transferred to the binary vectors pEarleyGate100 (Earley et al., 2006) that has the constitutive 35S CMV promoter or pBAV150 (Vinatzer et al., 2006), which has Dexamethasone inducible promoter by LR recombination (Life technologies). The constructs were electroporated into E. coli strain DH10B and positive clones were identified by PCR and sequencing. The construct was then introduced into Agrobacterium tumefaciens strain GV3101 by electroporation and the RMX-A02 and Col-0 plants were transformed by the floral dip method (Clough and Bent, 1998). Transformants were selected by growing plants in soil soaked with 0.1% Basta (AgrEvo, Norfolk, UK) and inserts were confirmed by PCR using gene specific primers. Homozygous T 3 plants were obtained and used for the subsequent experiments.

Statistical Analysis
For statistical analysis, two-tailed students t-tests were performed on data obtained in plant infection assays and qRT-PCR.

Accession Numbers
The raw sequence reads from the genomics sequencing of bulks are available from the Sequenced Read Archive (SRA) under accession numbers SRX1646609 (avirulent) and SRX1646555 (virulent). Genomic sequences of parental isolates can be found under accession numbers SRX1646645 for Hpa-Cala2 and SRX1646646 for Hpa-Noks1. Accession numbers for the HAC1 and S-HAC1 alleles are KX523274, KX523275, KX523276, KX523277, KX523278, KX523279, and KX523280. Genome sequence assemblies for the parental isolates are available from GenBank under accession numbers LKIA00000000.1 and LLKM00000000.1 (Studholme and Tör, to be published elsewhere). Accession number of Hpa-Noks1 23137 contig is LLKM01000918. Accession number for Hpa-Emoy2 reference genome v8.3 Scaffold 41 is ABWE00000000.2.

Discovering Cryptic Avirulence Determinant from Mating of Virulent Hpa Isolates
Two Hpa isolates, Cala2 and Noks1, were used to screen the worldwide diversity collection of 96 A. thaliana accessions (Nordborg et al., 2005) for variation in downy mildew response. Twenty-one accessions from this collection were susceptible in cotyledons to both of these isolates. The susceptible accessions were then screened with an F 1 isolate (referred to hereafter as CaNo F 1 ) that was generated from a laboratory mating of Hpa-Cala2 and Hpa-Noks1 (Bailey et al., 2011). Three North American accessions (RMX-A02, RMX-A180, and Yo-0) and one Swedish accession (Bil-7) were found to be resistant in cotyledons to CaNo F 1 (Figure 1; Supplemental Table 2).
We used an available F 2 mapping population from this F 1 isolate for further investigations. Accessions that were susceptible to both Hpa-Cala2 and Hpa-Noks1 but resistant to CaNo F 1 also showed resistance to some of the F 2 isolates, indicating the presence of a possible "cryptic" avirulence locus in Hpa. Using the A. thaliana accession RMX-A02, a total of 54 randomly selected CaNo F 2 isolates were screened to determine the inheritance of the avirulence determinant. A single dominant avirulence determinant could explain the phenotypic variation that segregated in the CaNo F 2 population (avirulence,virulence ratio was 43,11, with chi-square = 0.717 and P = 0.30, Supplemental Table 3). The predicted avirulence determinant was designated as Hyaloperonospora arabidopsidis cryptic1 (HAC1). This result suggested the presence of a suppressor, designated Suppressor of HAC1 (S-HAC1) originating from either Hpa-Cala2 or Hpa-Noks1 but that was not inherited by the F 1 isolate that was used to generate the F 2 population.
Fine Mapping Defines a 14 kb Interval for the HAC1 Locus DNA from 15 virulent and 15 avirulent F 2 isolates was pooled in equal concentrations to make up the virulent and avirulent DNA bulks, respectively, and bulk segregant mapping analysis was performed. We generated 100 bp paired-end Illumina HiSeq2500 FIGURE 1 | Avirulent determinant HAC1 was identified from the segregating F 2 population. The parental isolates Hpa-Noks1 and Hpa-Cala2 displayed a compatible interaction while CaNo F 1 showed an incompatible interaction with RMX-A02 and CaNo F 2 isolates segregating in a 3,1 (avirulent, virulent) ratio.
sequencing data from the two bulked (virulent and avirulent) pools, comprising 104 million reads for the virulent bulk, and 110 million reads for the avirulent bulk. We then used the Hpa-Noks1 genome assembly as a reference to map sequence reads from both virulent and avirulent bulks. A total of 26,722 SNP sites were then identified and examined for frequency of major alleles in the bulks. Hpa-Noks1 contig 23137 had the highest SNP frequency (Supplemental Table 4). Some of these SNPs were converted to CAPS marker and were then used to map the HAC1 locus. Once linkage was confirmed, flanking markers were used to identify the Hpa-Emoy2 reference genome (v8.3.) Scaffold 41. A total of 190 F 2 isolates were then screened with further markers and the HAC1 locus was fine mapped to a 14 kb interval on the reference genome Hpa-Emoy2, Scaffold 41, 298.000-312.000 (Figure 2A).

HAC1 Cala2 Is an RXLR-dEER Type Avirulence Determinant
According to the EnsemblProtists annotation, this 14 kb interval in Hpa-Emoy2 genome contains three genes without any putative effector domains and motifs, Hpa-T805676 encoding an unknown protein, Hpa-T805677 encoding a NADdependant epimerase/dehydrase (Pfam, PF01370), and Hpa-T805678 encoding an actin-like protein (PF00022). There were also two gaps in the Hpa-Emoy2 genome sequence assembly within the interval at position 41,300, 419-302,374, and 41,304, 709-305,855 (Figure 2B). Subsequent efforts therefore focused on PCR amplification and sequencing to fill these gaps from genomic DNA of Hpa-Cala2, Hpa-Noks1, Hpa-Emoy2, and a virulent and an avirulent CaNo F 2 isolates. This revealed the existence of a putative effector gene in each of the gaps. We then concentrated on these putative genes and a rapid isolation of cDNA ends (RACE) PCR was carried out for each. Full-length cDNAs for each gene from Hpa-Cala2, Hpa-Noks1, and Hpa-Emoy2 were obtained and comparison to genomic DNA revealed that none of them had any intron.
In the first gap (41,300,419-302,374), two alleles of the same gene with 96% identity at the nucleotide level were identified in Hpa-Cala2, indicating that this gene is heterozygous. Both alleles encode a predicted protein with structural similarity to other putative effector genes; a search for domains and motifs revealed a signal peptide and a dEEE motif in one allele (designated s-hac1 Cala2 , see below for details), and a signal peptide and a dKEE motif in the other (designated S-HAC1 Cala2 , see below for details). This indicates that both proteins may be secreted and act as putative effectors. Neither of the alleles had the RXLR motif ( Figure 2C). Further analysis showed Hpa-Noks1 and Hpa-Emoy2 were both homozygous for the s-hac1 allele. Additionally, the Hpa-Noks1 allele had a premature stop codon (see below for details).
In the second gap (41,304,709-305,855), a gene predicted to encode a protein with a signal peptide, an RxLR and a dEER motif was identified in Hpa-Cala2 ( Figure 2C) suggesting that this protein may also be secreted and act as a putative effector. No heterozygosity was detected in this gene at Hpa-Cala2, Hpa-Noks1, and Hpa-Emoy2 genomes. Both Hpa-Noks1 and Hpa-Emoy2 alleles had premature stop codon.
from Hpa-Noks1 and the virulent F 2 isolates revealed that both have a premature stop codon, suggesting that both of them are non-functional. In contrast, Hpa-Cala2 and the avirulent homozygous F 2 isolate have the full-length alleles indicating that the putative avirulence determinant may come from the parent isolate Hpa-Cala2.
Recognition of the HAC1 product by the corresponding R-gene in RMX-A02 protein should initiate a defense response and therefore, be amenable to identify the avirulent determinant. To test this, we cloned the full-length copies of the putative effector genes (with dEEE or DKEE motif in the first gap, and with RXLR/dEER in the second gap) from Hpa-Cala2 and Hpa-Noks1 into vectors with a Dex-inducible promoter and then transformed them into RMX-A02 and Col-0 plants. Transgenic T 1 , T 2 and homozygous T 3 plants were selected using Basta and the insert was confirmed with PCR (Supplemental Figure 1). All of the Basta-selected lines were treated with dexamethasone. Only the RMX-A02 seedlings that carried the RXLR-type effector from Hpa-Cala2 showed a chlorotic response 3 days after treatment (Figure 3) indicating an R-gene mediated cell death. Growth of these plants was stunted and in many cases individuals could not be recovered for producing seed. No altered phenotype was observed in any of the transgenic Col-0 plants indicating the defense response was race specific. Similarly, neither the Hpa-Noks1 variant of the RXLR-type effector nor the variant of the second dEEE/dKEE type putative effectors triggered a defense response in RMX-A02 (Figure 3). This supports the conclusion that HAC1 avirulence in A. thaliana accession RMX-A02 is determined by recognition of the RxLR-type protein inherited from Hpa-Cala2 and this Avr gene is designated HAC1 Cala2 .

Pathogen Genetics Supports
Heterozygosity in HAC1 Locus in Hpa-Cala2 As described above, the gene discovered in the first gap of the genome sequence of Hpa-Cala2 is heterozygous, with alleles having either a dEEE or dKEE motif. Both virulent and avirulent CaNo F 2 isolates had only the putative effector with the dEEE motif, indicating that this allele does not influence the virulence of the pathogen on RMX-A02. This implies that the dEEE allele rather than the dKEE allele from Hpa-Cala2 was inherited during the generation of the F 1 . Thus, the presence of the dKEE allele in Hpa-Cala2 may somehow suppress the HAC1 Cala2 -triggered defense response. If true, then cultures derived from selfedoospores of Hpa-Cala2 should segregate for phenotypic variation of virulence/avirulence in RMX-A02. To test this, 29 selfed Hpa-Cala2 lines were screened using high-resolution melt curve (HRM) analysis. The segregation ratio obtained was 8,15,6 ratio (dKEE/dKEE, dKEE/dEEE, dEEE/dEEE; 1,2,1 ratio) confirming the segregation of alleles (Figure 4). Seven of these selfed lines were used to test interactions on RMX-A02. Five lines were virulent and two avirulent correlating well with the HRM data. These results indicate that the dKEE allele is epistatic over HAC1 Cala2 either by inducing susceptibility before HAC1 Cala2 triggers defense or by somehow interfering directly with the function of the HAC1 Cala2 effector. Since Hpa-Cala2 is virulent and heterozygous, the dKEE-containing allele is the originally predicted dominant allele, designated S-HAC1 Cala2 (Suppressor of H. arabidopsidis cryptic 1), whereas the allele with the dEEE motif would be the recessive or s-hac1 Cala2 allele.

Overexpression of S-HAC1 Cala2 in Arabidopsis Alters Interaction Phenotype
Since pathogen genetics and molecular studies suggested that S-HAC1 Cala2 plays a role in Hpa-Cala2 virulence on RMX-A02, we tested whether overexpression of this gene within the plant alters the HAC1 Cala2 -triggered defense response. The S-HAC1 Cala2 gene was cloned under a 35S promoter, transformed into RMX-A02 plants and inserts were confirmed by PCR (Supplemental Figure 2). Homozygous T 3 lines were then challenged with a representative avirulent F 2 isolate (CaNo F 2 110) and Hpa-Cala2 and Hpa-Emoy2 as virulent control isolates. In controls, transgenic lines inoculated with Hpa-Cala2 or Hpa-Emoy2 did not show any alterations in pathogen growth and the lines gave wild-type levels of susceptibility.
The typical defense response of untransformed RMX-A02 seedlings inoculated with CaNo F 2 110 was without any symptoms such as flecks, pitting, or trailing necrosis; the cotyledons remained green and very rarely sporulation was observed. RMX-A02 lines carrying the S-HAC1 Cala2 gene exhibited wild-type development of rosettes and flowering, indicating that overexpression of this gene did not have a physiological effect on the plant. However, T 3 transgenic lines challenged with avirulent CaNo F 2 110 displayed enhanced sporulation, at L3 level interaction phenotype (Tör et al., 2002). Enhanced pathogen growth was also confirmed with trypan blue staining of infected tissues (Figure 5A). High-level sporulation expected for full susceptibility was not observed, indicating the transgene did not fully suppress the avirulence conferred by HAC1 Cala2 when expressed in planta. Ws-eds1, Ler-0, and Col-0 inoculated with CaNo F 2 110 displayed full sporulation as expected ( Figure 5A).
The biomass of CaNo F 2 110 within the transgenic and nontransgenic lines was also quantified using qPCR. A significant increase in pathogen biomass was detected within the RMX-A02 lines carrying the S-HAC1 Cala2 gene lines (Figure 5B) indicating the presence of the S-HAC1 Cala2 gene contributes to pathogen growth.
FIGURE 4 | S-HAC1 Cala2 and s-hac1 Cala2 alleles segregate in selfed Hpa-Cala2. Twenty-nine cultures derived from oospores of selfed Hpa-Cala2 were subjected to high-resolution melt curve analysis using S-HAC1 Cala2 -specific primers. Blue peaks show homozygous S-HAC1 Cala2 lines and red peaks show heterozygous lines. The experiment has been repeated with the s-hac1 Cala2 specific primers and the same results were obtained.

Sequence and Expression Analysis of HAC1 Cala2 , S-HAC1 Cala2 and s-hac1 Cala2
The open reading frame of the HAC1 Cala2 gene encodes a predicted protein of 327 amino acids (molecular mass of 37.262 kDa). Domain and motif searches of the HAC1 Cala2 protein identified a signal peptide (M1-A15), an RxLR (49 to 52 aa) and dEER (59-62 aa) motifs ( Figure 6A). Transmembrane prediction using the TMHMM server (Krogh et al., 2001) did not identify any TM domain, suggesting that HAC1 Cala2 is a soluble protein.
Analysis of HAC1 Noks1 and HAC1 Emoy2 sequences indicated that these proteins are identical at the N-terminal, but both of them lack the RXLR motif and both have a premature stop codon around the dEER (58 aa) motif ( Figure 6A). This stop codon has originated from indels at the nucleotide level that cause a frame shift (Supplemental Figure 3). BLAST searches against the Hpa-Emoy2 database (http://eumicrobedb.org) did not reveal a significant alignment of HAC1 Cala2 with any unigene. Using HAC1 Cala2 , the TBLASTN search against Ensembl EST database revealed 66% identity to HpaT801867 (ATR1 NDWSB allele) and 48% identity to HpaT813746, two avirulence proteinlike proteins. Analysis of structure of the HAC1 Cala2 protein using Phyre2 (Kelley et al., 2015) revealed that 55% of the total amino acid sequences can be modeled onto the ATR1 effector protein from Hpa, supporting the hypothesis that the HAC1 Cala2 protein may act as an effector protein (Supplemental Figure 4).
The open reading frame of the S-HAC1 Cala2 allele encodes a predicted protein of 393 aa (molecular mass of 43.225 kDa). Domain and motif searches on the S-HAC1 Cala2 protein identified a signal peptide (M1-T23), and a dKEE (55-58 aa) motif (Figure 6B), and no TM domain has been identified. s-hac1 Cala2 has similar motifs but has the dEEE instead of dKEE ( Figure 6B). S-HAC1 Cala2 and s-hac1 Cala2 share 94.9% identity. s-hac Noks1 has a premature stop codon (94 aa) and analysis of S-HAC1 Cala2 , s-hac1 Cala2 , s-hac Noks1 , and s-hac Emoy2 sequences revealed that the N-termini of these proteins are conserved ( Figure 6B). Alignment of nucleotide sequences of S-HAC1 Cala2 , s-hac1 Cala2 , s-hac Noks1 , and s-hac Emoy2 indicate the level of polymorphism (Supplemental Figure 5). BLAST and TBLASTN searches with S-HAC1 Cala2 , s-hac1 Cala2 against Ensembl did not reveal any significant alignment. In addition, structure analysis of S-HAC1 Cala2 and s-hac1 Cala2 using Phyre2 did not detect any (remote) homologs for the protein.
Expression of HAC1 Cala2 and S-HAC1 Cala2 in Hpa-Cala2 was investigated over a time course. Total RNA was isolated from uninfected (control) or infected tissues at 1-7 days post inoculation (dpi). Gene specific primers were used for reverse transcriptase mediated (RT) PCR analysis (Supplemental Figure 6). Expression of both genes was detected in infected tissues at 2 dpi. To obtain further insight into the expression level of these genes, quantitative PCR was carried out with total RNA isolated from Arabidopsis seedlings infected Hpa-Cala2 at 3, 4, and 7 dpi using gene specific primers for HAC1 Cala2 , S-HAC1 Cala2 and s-hac1 Cala2 . All three genes showed increasing levels of expression over a week, and the expression levels of HAC1 Cala2 and S-HAC1 Cala2 were very similar. The expression level of s-hac1 Cala2 was significantly lower than that of HAC1 Cala2 and S-HAC1 Cala2 (Figure 7).

Other Heterozygous Suppressor Genes Exists in Hpa-Cala2
To investigate whether any other heterozygous alleles might exist, we used the parent isolate Hpa-Cala2 and two progeny lines from selfed Hpa-Cala2, Hpa-Cala2S2, and Hpa-Cala2S5, to screen 50 of the 96 worldwide diversity collection of FIGURE 5 | Overexpression of S-HAC1 Cala2 in RMX-A02 displays altered phenotype and biomass production. The S-HAC1 Cala2 gene was cloned under a 35S promoter and introduced to RMX-A02 plants. Homozygous T 3 lines were challenged with avirulent F 2 isolate, CaNo F 2 110. (A), wild-type and transgenic seedlings inoculated with CaNo F 2 110 and stained with trypan blue 7 days post inoculation (dpi); (B), pathogen biomass production in transgenic and non-transgenic RMX-A02. Relative abundance of Hpa-Actin (Hpa-807716) to Arabidopsis At5g46630 gene was calculated from infected plant materials 2, 3, 4, and 7 dpi. Three samples were included for each analysis and the experiments were repeated three times. The increase in pathogen biomass in RMX-AO2+S-HAC1 Cala2 compared with that in the RMX-A02 control was statistically significant (*P < 0.05, **P < 0.01; Student's t-test).
A. thaliana accessions (Nordborg et al., 2005). Interestingly, the isolates displayed different interaction phenotypes with these accessions (Supplemental Table 5), indicating the parent isolate was heterozygous at other loci. Forty accessions gave the same reaction to Hpa-Cala2, Hpa-Cala2S2, and Hpa-Cala2S5. However, 10 of the accessions differed in their responses, FIGURE 6 | Alignment of amino acid sequences of HAC1 and S-HAC1. Amino acid sequences of HAC1 and S-HAC1 were predicted from cDNAs isolated from Hpa-Cala2, Hpa-Noks1 and Hpa-Emoy2 and were aligned using Geneious (v8.0). (A) Alignment of HAC1 Cala2 , HAC1 Noks1 , and HAC1 Emoy2 amino acid sequences. Black or dark gray boxes with white letters indicate identity or similarity, respectively, to HAC1 Cala2 and dashes indicate gaps in the alignment. The reading frames are adjusted to fit translation of regions downstream of stop codon. The predicted N-terminal signal peptide, the RXLR and dEER motifs are shown above the sequences, (B) Alignment of amino acid sequences of S-HAC1 Cala2 , s-hac1 Cala2 , s-hac1 Noks1 and s-hac1 Emoy2 . Black or dark gray boxes with white letters indicate identity or similarity, respectively, to S-HAC1 Cala2 and dashes indicate gaps in the alignment. The predicted N-terminal signal peptide and the dEER motifs are shown above the sequences. *indicates stop codon.

DISCUSSION
Cryptic avirulence is a phenomenon that was first described by H.H. Flor during his seminal research on interactions between flax and the basidiomycete fungus M. lini (Flor, 1946) and this phenomenon has subsequently been confirmed in fungal genetic experiments by other researchers (Lawrence et al., 1981;Ellingboe, 1992;Lau et al., 1993). Here we FIGURE 7 | Expression analysis of s-hac1 Cala2 , S-HAC1 Cala2 and HAC1 Cala2 . Total RNA was isolated from infected Hpa-Cala2 at 3, 4 and 7dpi. Quantitative Real-time PCR was carried out using gene specific primers for HAC1 Cala2 , S-HAC1 Cala2 and s-hac1 Cala2 . Fifteen seedlings from each genotype made up a sample and three replicates were used for each sample, the experiments were repeated three times with similar results. Expression level of s-hac1 Cala2 was significantly lower than that of S-HAC1 Cala2 and HAC1 Cala2 in Hpa-Cala2 background. Asterisks denote statistical significance in two-tailed Student's t-test (p < 0.05). present pathology, genetic and molecular evidence that this phenomenon also occurs in a biotrophic oomycete. Using the Arabidopsis-Hyaloperonospora experimental model, we showed that the Arabidopsis accession RMX-A02 is susceptible to two standard isolates (Hpa-Cala2 and Hpa-Noks1) but resistant to an F 1 derived from these isolates (CaNo F 1 ) and that a single corresponding avirulence locus is indicated by segregation in F 2 progeny derived from this F 1 . This was confirmed using fine-mapping and genomic sequencing to clone an avirulence determinant (designated HAC1 Cala2 ) and a heterozygous avirulence suppressor gene in the same map interval (designated S-HAC1 Cala2 /s-hac1 Cala2 ). Both genes encode putative effector proteins.
In most cases, a map-based gene cloning strategy uses parental lines that differ in contrasting phenotypes (e.g., resistant vs. susceptible plants; or avirulent vs. virulent pathogen isolates) to generate a mapping population. However, the parent isolates used in the present study (Hpa-Cala2 and Hpa-Noks1) were both virulent on the same Arabidopsis accession, RMX-A02. Thus, we could not know from which parent the hidden or cryptic avirulence allele originated. Cryptic alleles are phenotypically silent DNA sequences and have been reported in prokaryotic microorganisms as well as in fungi (Hall et al., 1983;Le Gac et al., 2007). Several molecular explanations have been postulated including, (1) spontaneous occurrence by mutation or rearrangements, (2) action of a second suppressor locus, and (3) epigenetic control. They may however be activated in a few individuals of a large population by mutation, recombination, insertion elements or by other genetic mechanisms (Hall et al., 1983;Li, 1984).
In the present investigation, we used next generation sequencing to reveal the loci involved. Virulent and avirulent bulks were sequenced, as well as the genomes of the parental isolates, and we used the polymorphic nature of the isolates to screen 26722 SNP sites, enabling us to identify a contig from Hpa-Noks1 genome linked to the locus. The HAC1 locus was mapped to a 14 kb interval using an experimental population of 190 F 2 isolates, and haplotype sequence information from Hpa-Cala2, Hpa-Noks1, and Hpa-Emoy2 to characterize the high degree of polymorphism in the interval.
All of the R/AVR gene pairs that have previously been cloned from the Hpa-Arabidopsis system, including RPP1/ATR1 (Botella et al., 1998;Rehmany et al., 2005) and RPP13/ATR13 (Bittner-Eddy et al., 1999;Allen et al., 2004), have fallen into the NLR type of R-genes and RxLR-dEER class of effectors. HAC1 Cala2 has the canonical signal and translocation peptides, RxLR-dEER, and triggers a defense response in RMX-A02. Previously, we used the CaNo F2 110 to investigate the corresponding gene in RMX-A02 and mapped the locus in the Arabidopsis genome to chromosome 4 within an interval of 35 kb containing three R-genes in the RPP4/RPP5 locus (unpublished data). Thus, HAC1 Cala2 -triggered defense responses in the RMX-A02 background are very highly likely to be mediated by an NLR type R-gene.
A large number of RxLR or RxLR-like effectors have been identified in various downy mildew pathogens and Phytophthora species (Bailey et al., 2011;Anderson et al., 2015) and molecular studies on several of these effectors have shown that they suppress plant immunity (Sohn et al., 2007). It was interesting that the structure of HAC1 Cala2 could be modeled with high confidence onto the ATR1 effector protein. To investigate the RPP1/ATR1 interaction, Chou et al. (2011) determined the crystal structure of ATR1 and identified the critical recognition sites of the effector protein. They concluded that oomycete effectors that share the same lineage as ATR1 rapidly evolve to escape host detection. In this case, if HAC1 Cala2 is indeed from the same family as ATR1, it can be proposed that HAC1 Cala2 alleles, hac1 Noks1 , and hac1 Emoy2 , have already evolved to avoid recognition as they have premature stop codons, enabling Hpa-Noks1 and Hpa-Emoy2 to be virulent on RMX-A02.
A diploid organism can be heterozygous at a gene locus when its cells contain two different alleles of a gene and use this for its own advantage. The importance of heterozygosity for virulence in fungi and oomycetes has been recognized for a while and in some cases documented. For example, a high level of heterozygosity for virulence in two populations of Puccinia recondita (wheat leaf rust fungus) has been found (Kolmer, 1992). Similarly, genomic sequences of P. striiformis f. sp. tritici (Pst-wheat stripe rust fungus) revealed the presence of heterozygosity (Zheng et al., 2013). Population studies on M. grisea (rice blast fungus) using PCR markers identified six of the loci investigated to be heterozygous (Babu et al., 2013). Genomic investigations on Phytophthora capsici revealed the loss of heterozygosity, which has been proposed to be a rapid mechanism for fixing alleles and may be an important component of adaptability (Lamour et al., 2012). Similarly, screening of 652 P. infestans isolates from commercial potato fields in the Netherlands during a tenyear period using 12 informative microsatellite markers and mitochondrial haplotypes detected a low level of heterozygosity (Li et al., 2012). It has been reasoned that a pathogen could gain a selective advantage from a heterozygous genotype by increasing the pathogen's ability to adapt to a changing host environment such as overcoming the host's resistance mechanism (Zheng et al., 2013).
In the current investigation, it is interesting that heterozygosity of a putative effector gene was revealed in one isolate Hpa-Cala2 but not in two other standard isolates (Hpa-Noks1 and Hpa-Emoy2) (Figure 8). Heterozygote genotypes are known to have a higher relative fitness than either the homozygote dominant or homozygote recessive genotypes, known as heterozygous advantage (Hedrick, 2012). There are many examples of heterozygous advantage in living systems. For example, the Major Histocompatibility Complex (MHC) loci are highly polymorphic heterozygous loci that control immunological recognition of pathogens in animals and confer a selective advantage by enhancing resistance to infectious diseases FIGURE 8 | Genotypes of three Hpa isolates at the HAC1 locus. Sequencing and pathology results were used to determine the genotypes of isolates. In Hpa-Cala2, virulence on RMX-A02 results from the combination of an active suppressor (S-HAC1 Cala2 ) and an active avirulence (HAC1 Cala2 ) allele. In the presence of an inactive suppressor (s-hac1 Cala2 ) and an active avirulence (HAC1 Cala2 ) allele, the interaction results in avirulence. In Hpa-Noks1 and Hpa-Emoy2, the putative avirulence gene is inactive due to mutations, thus both are virulent on RMX-A02. (Penn et al., 2002). A well-established case of heterozygous advantage in humans is that of the genes involved in sickle cell anemia (Luzzatto, 2012) and Cystic Fibrosis (Modiano et al., 2007). In plant breeding programmes, hybrid vigor generated from crosses between inbred lines from different genetic backgrounds, is an example of heterozygous advantage used routinely for higher crop yield (Whitford et al., 2013). However, the majority of studies on heterozygosity in fungi and oomycete pathogens have been within the field of population genetics and there has been no investigation of heterozygosity at the level of individual locus or effector genes. Segregation of a suppressor/effector after sexual reproduction (as observed in this study) means that a proportion of the progeny will have an active Avr effector, whereas others will not. This may be advantageous to the pathogen by increasing the pathogen's ability to adapt to a changing host environment.
A secreted effector can both trigger and suppress R genebased immunity. For example, F. oxysporum f.sp. lycopersici (Fol) effector Avr1 triggers a defense response when the host plant, tomato, carries a matching R-gene (I or I-1). However, Avr1 suppresses the protective effect of two other host Rgenes, I-2 and I-3 (Houterman et al., 2008). Similarly, IPI-04, an effector protein of P. infestans, has been shown not only to elude detection by the host potato protein RB, but also to block recognition of the effector protein IPI-01, leading to suppression of RB-mediated host cell death (Halterman et al., 2010;Chen et al., 2012). A similar scenario may exist for the Hpa-Cala2 RMX-A02 interactions, where S-HAC1 Cala2 may suppress the HAC1 Cala2 -triggered host immunity or interfere with its function. Suppression of HAC1 Cala2 -triggered immunity may occur within the host cell. For this to happen, S-HAC1 Cala2 has to be delivered inside the host cell and may target the same host component as HAC1 Cala2 . Alternatively, S-HAC1 Cala2 may physically bind to HAC1 Cala2 within the pathogen or inside the host cell, promoting conformational changes within HAC1 Cala2 and preventing either the secretion of HAC1 Cala2 from the pathogen, its entry into the host cell or the recognition of HAC1 Cala2 by the host intracellular receptor. We hope to investigate this further with yeast two-hybrid and pull-down experiments.
Overexpression of S-HAC1 Cala2 in RMX-A02 plants did not confer any resistance phenotype. However, when these transgenic lines were challenged with the avirulent CaNo F 2 110 isolate (carrying HAC1 Cala2 and s-hac1 Cala2 ), phenotypic alterations were observed, indicating that S-HAC1 Cala2 may have interacted with a pathogen-delivered protein, most likely to be HAC1 Cala2 , in the apoplast or cytoplasm of the host cell. Surprisingly, full suppression of HAC1 Cala2 -triggered resistance was not observed in these transgenic lines carrying S-HAC1 Cala2 . This may indicate that S-HAC1 Cala2 needs to interact with HAC1 Cala2 within the pathogen or apoplast. Alternatively, it may require posttranslational modifications (PTM) by the pathogen, such as phosphorylation, SUMOylation, methylation, ubiquitination, or acetylation (Hewezi, 2015). Nevertheless, the results of this study suggest that the combination of HAC1 Cala2 and S-HAC1 Cala2 may dictate the race-specificity of Hpa-Cala2 on the Arabidopsis accession RMX-A02.
Hpa is a diploid, homothallic pathogen, which means that a haploid antheridium and oogonium are produced within the same thallus to generate oospores. Frequent sexual reproduction in a homothallic pathogen would be expected to give rise to a largely homozygous genome and so maintenance of heterozygosity at the HAC1 or other loci in Hpa-Cala2 may be surprising. However, this and many other experimental isolates have been originally single spored and maintained in the laboratory largely through asexual propagation. In a diploid clonal population, mutation events at each locus lead to an accumulation of heterozygosity over time (Balloux et al., 2003); thus, it would be expected that there may be considerable heterozygosity in experimental isolates where there is an absence of frequent sexual reproduction. In addition, sexual crosses between expressed and silenced Avr alleles can result in varying outcomes and unusual inheritance patterns for Avr gene expression in progeny (Na and Gijzen, 2016).
The use of selfed lines Hpa-Cala2S2 and Hpa-Cala2S5 indicated that other suppressor loci within Hpa-Cala2 might be heterozygous and so the question arises as to whether heterozygosity is a widespread phenomenon in Hpa. To investigate this, we took advantage of the fact that Hpa isolates have been used in different laboratories to screen the same Arabidopsis diversity collections and some published data are available. For example, in investigations by both Krasileva et al. (2011) andNemri et al. (2010), interaction phenotypes of isolates Hpa-Emwa1 and Hpa-Emco5 (among other isolates) on the 83 Arabidopsis accessions were assessed. However, different results were obtained by these authors for at least 12 Arabidopsis accessions (Supplemental Table 6). This discrepancy may have arisen if each group obtained their isolates from different oospores of a segregating population, indicating the existence of heterozygosity at different loci. In addition, it should be remembered that natural populations of Hpa are polycyclic, reproducing many times asexually (leading to an increase in heterozygosity) as well as sexually, before overwintering as oospores; also, despite being homothallic, in mixed populations, outcrossing with other races is possible. Thus, it is very probable that natural populations of Hpa have considerable heterozygosity. Evidence of outcrossing in other homothallic oomycetes, such as P. ultimum (Francis and St. Clair, 1993) and P. aphanidermatum (Lee et al., 2010) has also been used to explain high levels of heterozygosity in these populations.

AUTHOR CONTRIBUTIONS
MT and AW-T: Planned and designed the research; AW-T, MT, VC, and OT: Conducted the laboratory work; AW-T, MT, and DS: Analyzed the data; MT, AW-T, EH, and DS: Interpreted the data and wrote the manuscript.

FUNDING
This work was supported by grant F-09 963/A (MT) from the Leverhulme Trust. Financial support from the Turkish Ministry of Education to OT is gratefully acknowledged.

ACKNOWLEDGMENTS
We are grateful to Yiguo Hong for critically reading the manuscript.