|Home | About | Journals | Submit | Contact Us | Français|
Obligate biotrophic fungi are major pathogens of cultivated crops. They establish a long‐term feeding relationship with a living host throughout their life cycle, suggesting highly effective and long‐lasting host defence suppression and physiological reprogramming (Kemen & Jones, 2012). Two of the most agronomically important groups of obligate biotrophic fungal pathogens are basidiomycete rusts and ascomycete powdery mildews. Powdery mildews are pathogens that infect monocot and dicot species including important crops such as cereals, grape and tomato. Cereal mildews are grouped into one species, Blumeria graminis (B. g.), that is divided in different formae speciales corresponding to pathogens adapted to a specific host species (Schulze‐Lefert & Panstruga, 2011). Of these, five grow on cultivated cereals: B. g. tritici on wheat, B. g. hordei on barley, B. g. secalis on rye, B. g. avenae on oat and the newly identified B. g. triticale form that arose from a hybridization between B. g. tritici and B. g. secalis and grows on triticale, bread wheat and durum wheat (Troch et al., 2014; Menardo et al., 2016).
Resistance against obligate biotrophic pathogens is commonly mediated by R genes encoding intracellular immune receptors mostly belonging to the conserved protein family of nucleotide‐binding, leucine‐rich repeat receptor (NLR) proteins (Dodds & Rathjen, 2010). Forty‐five genetic loci have been identified in cereal crops that confer agronomically effective resistance against the powdery mildew fungus B. graminis, but only a few of them have been cloned (Yahiaoui et al., 2006; Bhullar et al., 2009; Seeholzer et al., 2010; Hurni et al., 2013; Sánchez‐Martín et al., 2016). Mildew resistance genes have been introgressed from rye and wild grasses into wheat, indicating high conservation of effector recognition in different cereal species despite host specialization on the pathogen side. Two of the best characterized R genes in cereals are Mla in barley and Pm3 in wheat, both forming allelic series (Yahiaoui et al., 2006; Bhullar et al., 2009; Seeholzer et al., 2010). R proteins interact directly or indirectly with cognate pathogen avirulence factors (AVRs), and upon recognition many induce a hypersensitive response (HR), a form of programmed cell death that is particularly effective against obligate biotrophs (Moffett et al., 2002). Domain swap studies have shown that the LRR domain is a major determinant of AVR recognition specificity, and a few amino acid differences in this domain can result in distinctly different resistance spectra (Brunner et al., 2010; Ravensdale et al., 2011).
There are few Avr genes isolated from obligate biotrophic fungi and they typically encode for small proteins with a predicted signal peptide and no homology to any known protein function (Ravensdale et al., 2011; Bourras et al., 2015; Lu et al., 2016). Two exceptions are AVRa10 and AVRk1 from barley powdery mildew: they are encoded within long interspersed nuclear elements (LINE) retrotransposons (Ridout et al., 2006; Amselem et al., 2015). Most of the Avr genes cloned from obligate biotrophic fungi have been isolated from the flax rust fungus Melampsora lini (AvrL567, AvrM, AvrP123 and AvrP4) and control race‐specific recognition by the L, M and P allelic series of R genes in flax (Ravensdale et al., 2011). Amino acid polymorphisms between Avr variants are associated with differences in recognition specificity. For example, Dodds et al. (2006) showed that seven out of 12 variants of AvrL567 are differentially recognized by the L5, L6 or L7 allele, while the other five are not recognized. In powdery mildews, the first Avr coding for a typical effector protein has been cloned recently (Bourras et al., 2015). AvrPm3 a2/f2 is highly expressed at the stage of haustorium formation and confers dual recognition specificity towards the wheat Pm3a and Pm3f resistance gene alleles. In contrast to flax rust where Avr variants with different allelic specificities can be encoded within one cluster (e.g. AvrL567 variants A, B, F and L; Ellis et al., 2007), there is only one AvrPm3 a2/f2 gene in the mildew genome, and the closest effector homologues are not recognized by any alleles of the Pm3 resistance gene (Bourras et al., 2015). Thus, while the structural and molecular basis of race‐specific resistance to obligate biotrophic fungi is conserved on the plant side, there seem to be major differences in the way avirulence and gain of virulence are controlled on the pathogen side.
The powdery mildew genomes are some of the largest among fungi with an estimated size of 150–180 Mb, 90% of which corresponds to transposable element (TE) sequences (Spanu et al., 2010; Wicker et al., 2013). Contrasting with genome size, the gene content of powdery mildews is one of the lowest in fungi with c. 6500 genes, compared to an average of 11 000 in ascomycetes and 15 000 in basidiomycetes (Wicker et al., 2013; Mohanta & Bae, 2015). The mildew genomes encode a very large complement of candidate secreted effector proteins (CSEPs) that account for almost 10% of all predicted coding genes (Spanu et al., 2010; Pedersen et al., 2012; Wicker et al., 2013). Mildew effectors are predominantly expressed in the haustorium, and there is increasing molecular evidence of their direct implication in pathogen virulence (Zhang et al., 2012; Pliego et al., 2013; Schmidt et al., 2014; Ahmed et al., 2015). A functional screen of barley powdery mildew effectors using host‐induced gene silencing resulted in the identification of two CSEPs, BEC1054 (syn. CSEP0064) and BEC1011 (syn. CSEP0264) whose down‐regulation led to a reduction of 60 and 70%, respectively, in the ability of the mildew to form haustoria (Pliego et al., 2013). BEC1054 targets several barley proteins including a pathogen‐related‐5 protein isoform, suggesting a role in suppression of pathogen‐associated molecular pattern (PAMP)‐triggered immunity (PTI) (Pennington et al., 2016). Based on structural homologies to known fungal ribonucleases, BEC1054 and BEC1011 were classified as RNase‐like proteins, which make up the largest class of mildew CSEPs, accounting for almost 15% of all B. g. hordei effectors (Pedersen et al., 2012; Pliego et al., 2013). In B. g. tritici, SvrPm3 a1/f1 is a ribonuclease‐like effector involved in suppressing the AvrPm3 a2/f2 –Pm3a/f effector‐triggered immunity (ETI) (Bourras et al., 2015; Parlange et al., 2015). Together, these results suggest a central role of RNase‐like effectors in controlling mildew virulence and race specificity. Their exact mode of action is poorly understood, but a possible role as modulator of host immunity via interactions with host RNAs was proposed (Pedersen et al., 2012; Spanu, 2015).
The wheat resistance gene Pm2 has been molecularly isolated by mutant chromosome sequencing (MutChromSeq) and encodes an NLR protein (Sánchez‐Martín et al., 2016). Here we report the cloning of AvrPm2, the cognate Avr of Pm2. We demonstrate specific recognition and strong induction of a typical hypersensitive cell‐death response in transient assays in Nicotiana benthamiana. We show that AvrPm2 belongs to a ribonuclease‐like effector family that is conserved among cereal mildews and structurally different from AvrPm3 a2/f2. We also show that Pm2 recognizes the close homologue BgsE‐5845 from rye powdery mildew, indicating evolutionary conservation of this avirulence effector, and providing the first molecular explanation for functional R gene introgression in cereals.
Blumeria graminis isolates and the progeny from the cross between the Swiss isolate 96224 (mating‐type MAT1‐2) and the British isolate JIW2 (mating type MAT1‐1) were maintained in the haploid asexual phase on detached leaves of the susceptible wheat (Triticum aestivum) cultivar Kanzler on benzimidazole agar as described by Parlange et al. (2011). The near isogenic lines Ulka/8*Chancellor, Federation*4/Ulka and CI12632/8*Chancellor carrying the Pm2 resistance gene were used for phenotyping the progeny (Parlange et al., 2015; Sánchez‐Martín et al., 2016).
High molecular weight DNA of fungal isolates and RNA samples from infected leaves were extracted as previously described (Bourras et al., 2015). RNA samples were extracted from infected leaf material using the Qiagen miRNeasy Mini Kit (Qiagen) according to the manufacturer's instructions. RNA quality was checked by gel electrophoresis and based on the 260 : 280 ratios measured with a spectrophotometer (Nanodrop, Thermo Fisher Scientific, Waltham, MA, USA). The 260 : 280 ratios are available in Supporting Information Table S1. Full‐length cDNA was prepared using the SuperScript III RT kit (Life Technologies) according to the manufacturer's instructions. Molecular cloning of powdery mildew and wheat genes into Gateway‐compatible entry vector was performed using the pENTR/D‐TOPO Cloning Kit (Life Technologies) according to the manufacturer's instructions (list of primers in Table S2). Recombination to the binary vector pIPKb004 was performed as described by Stirnweis et al. (2014). RACE‐PCR cDNA was prepared with the SMARTer RACE cDNA kit (Takara Bio, Shiga, Japan) according to manufacturer's instructions.
Genetic maps were constructed with Kompetitive Allele‐Specific PCR (KASP) markers as described by Bourras et al. (2015). We produced a consensus linkage group containing the AvrPm2 locus as described by Bourras et al. (2015). For fine mapping, additional markers were designed on single nucleotide polymorphisms (SNPs) predicted between the parental sequences and genotyped by direct Sanger sequencing of PCR amplicons, or by using a cleaved amplified polymorphic sequence (CAPS) marker. To genetically anchor the genes identified by collinearity between wheat and barley powdery mildew to our genetic map, SNPs located in or near those genes were scored by direct Sanger sequencing of PCR amplicons. To validate co‐segregation of BgtE‐5845 with AvrPm2, presence/absence markers were designed on and near the gene and tested by PCR. All markers/primers used for genetic mapping are listed in Table S3.
Bacterial artificial chromosomes (BACs) spanning the AvrPm2 interval were selected from the BAC library (Parlange et al., 2011) based on their position in two different assemblies, confirmed by PCR, sequenced using Illumina MiSeq technology (San Diego, CA, USA), and assembled. Genes and TEs were then annotated manually (Table S4; Methods S1). The AvrPm2 region is deposited in the GenBank/EMBL library (accession number KX765276). To identify the AvrPm2 variant in the virulent parent JIW2, genome sequencing reads were mapped on the assembly of the AvrPm2 locus using CLC genomics software with the following parameters: mismatch cost = 2, insertion cost = 2, length fraction = 0.9, similarity fraction = 0.95. Physical markers were designed every 500 bp in a 20 kb region around AvrPm2 and amplification products were obtained from three isolates. All PCR products were checked by gel electrophoresis, purified and sequenced. The physical markers used to estimate the size of the deletion are listed in Table S5.
A set of 22 Chinese isolates were resequenced (Table S6; Methods S2). We mapped Illumina reads from the genomes of 60 isolates (Table S7) on the B. g. tritici reference (Wicker et al., 2013) using Bowtie2 (Langmead & Salzberg, 2012) with the following parameters: – score‐min L, −0.6, −0.25. We used SAMtools 0.1.19 (Li et al., 2009) to convert formats and collect genotype information at every polymorphic position. Finally, we used bcftools (Danecek et al., 2011) to generate a VCF file that was parsed with in‐house Perl scripts. We considered as high‐confidence SNPs all positions with a minimum mapping score of 20, a minimum coverage of 8×, a minimum frequency of the alternative call of 0.9 and < 5 missing genotypes. The GWAS was performed with the R package Gapit (Lipka et al., 2012). We used vcftools to calculate AvrPm2 locus coverage for every isolate using sliding windows of size 1000 bp with a 1000 bp step. All genomic sequences used for this study are available at the Sequence Read Archive (SRA) under accession number SRP062198.
Transient expression by agroinfiltration in N. benthamiana was performed as described in Bourras et al. (2015). HR was visually inspected 5 d after agroinfiltration and revealed by fluorescence scanning using the Fusion FX Imaging System (Vilbert Lourmat, Eberhardzell, Germany) with the following pre‐settings: lighting, blue epi‐illumination; excitation, spectra blue 470 nm; emission, F‐535 Y2‐filter; aperture, 0.84. Fluorescence intensity was quantified using the ImageJ2 image processing program (Schindelin et al., 2015). The measure of fluorescence is in integrated density, the product of the area infiltrated and the mean grey value. These fluorescence measurements were used to conduct an ANOVA followed by Tukey HSD post‐hoc tests using R. All gene sequences used in expression constructs for transient protein expression assays in N. benthamiana are available in Notes S1.
Quantitative real‐time PCR (qRT‐PCR) ready cDNA was synthesized from 1.5 ?g total RNA using the iScript cDNA synthesis kit (Bio‐Rad) according to the manufacturer’s instructions. Glyceraldehyde 3‐phosphate dehydrogenase (Gapdh) and Actin (Act1) genes were used as internal controls as described by Bourras et al. (2015). Gene expression was normalized to that of Gapdh. Target and reference gene fragments were amplified using the KAPA SYBR FAST qPCR kit (Kapa Biosystems, Wilmington, MA, USA) and the CFX96 real‐time PCR detection system (Bio‐Rad) according to the manufacturer's instructions. Gene expression analysis was performed using the Bio‐Rad real‐time PCR CFX Manager software v.3.5, according to the manufacturer's instructions (http://www.bio-rad.com/en-ch/product/cfx-manager-software). Additional information about qRT‐PCR experiments methodology is available in Methods S3. qRT‐PCR primers for AvrPm2 are listed in Table S2 and the primer efficiency curve is depicted in Fig. S1.
We performed a de novo annotation of the B. g. tritici genome, redefined the prediction of effector families and identified the AvrPm2 family. We also identified haplotypes of AvrPm2 in other formae speciales (Methods S4). Multiple alignments using gene sequences coding for the full protein as well as truncated sequences coding for the mature peptide were performed with Muscle 1:3.8.31‐1 (Edgar, 2004) and retrotranslated with TranslatorX (Abascal et al., 2010). Raxml v8.2.8 (Stamatakis, 2006) was used to find the maximum likelihood tree using GTR + GAMMA model and bootstrap support was computed with 100 replications. To test for positive selection, we estimated the likelihood of the maximum likelihood tree under the M7 and M8 models (Yang et al., 2000) with Paml 4.8 (Yang, 2007) and subsequently used the likelihood ratio test. The sequences of the AvrPm2 effector family used for phylogenetic analyses are available in Notes S2 (coding sequences) and Notes S3 (protein sequences).
To identify AvrPm2 we used two different genetic approaches: high‐resolution genetic mapping in an F1 population and a genome wide association study using a collection of geographically distant powdery mildew isolates. We have previously produced a cross between the Pm2 avirulent isolate 96224 and the Pm2 virulent isolate JIW2 and generated a mapping population of 117 individuals (Parlange et al., 2015). This F1 haploid progeny segregated in a 1 : 1 phenotypic ratio of avirulence to virulence on a wheat genotype with the Pm2 gene, indicating the presence of a single AvrPm2 locus in the pathogen (Parlange et al., 2015). To map the genetic interval containing AvrPm2, we used 254 SNP markers that we have previously scored with KASP technology in the 117 progeny (He et al., 2014; Bourras et al., 2015). We generated a genetic map of 1275 cM that included 248 markers spread over 18 linkage groups (LGs) containing at least four markers and ranging in size from 10.3 to 173.1 cM (Table S8). AvrPm2 mapped at the end of LG_15, at 7.0 cM south of marker M051_LE, indicating that the gene was genetically linked to contig 51 of the mildew genome assembly (Fig. 1a).
Analysis of contig 51 revealed that 1.52 (76.4%) of the predicted 2 Mb assembly consisted of sequence gaps of various length (Fig. 1b; Notes S4). To reduce the genetic interval containing AvrPm2, we developed a series of markers designed on all the SNPs scored between the parental genomic sequences (CAPS and SNP markers tested by sequencing). A second approach consisted of anchoring additional sequences to the fragmented contig 51 based on collinearity with contiguous scaffolds in the B. g. hordei genome, thus allowing the development of additional SNP markers (Notes S5). All markers were tested on the 117 progeny. Together, these approaches resulted in: essential improvements of sequence assembly and gene order in contig‐51 (Fig. S2); the development of 20 additional SNP markers; and the identification of new flanking markers (F2 and BgtE‐5842) defining a 6.8 cM genetic interval containing AvrPm2 (Fig. 1c).
To determine the full sequence of the AvrPm2 locus, we sequenced the minimal tiling path of the BAC contigs spanning the 8.7 cM interval between markers F2 and F12 (Figs 1c, S3; Notes S4). The resulting sequence reads were assembled using the CLC genomics software, which resulted in a 217 kb contiguous sequence spanning the AvrPm2 locus, a region that was originally estimated to be 250 kb (Figs 1e, S3; Notes S4). The sequence contained mainly TEs (> 80%) and a cluster of four homologous effector genes, BgtE‐5842, BgtE‐5843, BgtE‐5845 and BgtE‐5846. To reduce the number of AvrPm2 avirulence gene candidates we specifically searched for genes that are polymorphic between the two parental isolates, genetically co‐segregate with AvrPm2, and are expressed in the avirulent parent 96224. Two genes, BgtE‐5843 and BgtE‐5846, are not expressed as determined by RNA‐Seq data (Wicker et al., 2013) and mapped outside of the genetic interval containing AvrPm2 as defined by the flanking markers. BgtE‐5842 was expressed in 96224 but mapped at three recombinants from AvrPm2, thus leaving only one possible candidate: BgtE‐5845.
To test whether BgtE‐5845 cosegregates with AvrPm2 in the F1 population, we first analysed sequence polymorphisms between the parental genomes by mapping the sequencing reads from the virulent parent JIW2 on the model assembly of the AvrPm2 locus (Fig. 1f). Consistent with results from Wicker et al. (2013), we found that BgtE‐5845 was located in a region of several kilobases that is deleted in the virulent parent JIW2. Based on PCR amplification of physical markers spanning a 20 kb interval in the model assembly, we estimated the size of the deletion in the JIW2 virulent parent to be at least 10 kb (Figs S4, S5). Therefore, we assessed the co‐segregation between BgtE‐5845 and AvrPm2 using two InDel markers, E5845Int1 and E5845Int2, designed in the coding sequence of BgtE‐5845, and two additional ones, E5845_100 and E5845_d3, designed 100 and 4.6 kb upstream of the gene, respectively (Figs 1f, S4). The four markers were tested on the parents and the 12 recombinant progeny defining the genetic interval between the markers flanking AvrPm2. The results indicated that the gene and up to 4.6 kb of upstream sequence are only present in 96224 and the Pm2 avirulent recombinant progeny, thus indicating that BgtE‐5845 co‐segregates with AvrPm2.
In our second approach to identify AvrPm2, we performed a GWAS using 60 cereal mildew genomes including 42 B. g. tritici (12 virulent and 30 virulent) and 18 B. g. triticale (15 virulent and three avirulent) isolates originating mainly from Switzerland, Israel and China (Wicker et al., 2013; Menardo et al., 2016; Table S7). Genetic association between sequence polymorphisms and differences in virulence/avirulence patterns on Pm2 was assessed using the Genome Association and Prediction Integrated Tool (GAPIT; Lipka et al., 2012). We found the best correlated SNP at position 565 988 on contig 51, which is 35 kb from the closest predicted effector gene BgtE‐5846, and 50 kb from BgtE‐5845 (Fig. 2a,b). As mentioned before, BgtE‐5846 is not polymorphic between 96224 and JIW2, and RNA‐Seq mapping indicated it is not expressed in the avirulent parent. Manual inspection of the AvrPm2 locus sequence in the 60 mildew genomes revealed that BgtE‐5845 was present in all Pm2 avirulent isolates and absent in all the virulent ones (Table S7). To identify the size of the deletion in these geographically diverse isolates, we analysed sequencing coverage over a 40 kb region around BgtE‐5845. We found that the coverage was similar among all isolates over the 40 kb region except for a segment of 12 kb where the average coverage was c. 45× in the avirulent isolates and lower than 5× in all the virulent ones, thus suggesting a 12 kb deletion that contained BgtE‐5845 (Fig. 2c). We found that the sequence outside of the deletion consisted of high copy number long terminal repeat (LTR) retrotransposons of the superfamily Copia (family Ino), while the deleted segment contained mainly low copy number solo LTRs (Wicker et al., 2007), indicating a particular structure of the AvrPm2 locus based on TE content (Fig. 2d).
In conclusion, two different approaches identified BgtE‐5845 as the best and only candidate for AvrPm2, and that the virulent allele arose from a deletion in the virulent parent. BgtE‐5845 encodes for a typical effector protein of 120 residues that is slightly shorter than AVRPM3A2/F2 (130 residues) (Bourras et al., 2015). The N‐terminal region of the native protein consists of a predicted 21 amino acid long signal peptide followed by a Y(x)xC motif commonly found in mildew CSEPs (SignalP 4.1; Petersen et al., 2011; Pedersen et al., 2012). The predicted mature protein consists of 99 residues and contains a C‐terminal cysteine that is predicted to form a disulphide bond with the N‐terminal cysteine in the Y(x)xC motif (Disulfind; Ceroni et al., 2006).
The wheat resistance gene Pm2 was recently identified by MutChromSeq in the cultivar Federation*4/Ulka (Sánchez‐Martín et al., 2016). The structure of the gene consists of three exons (3730, 58 and 46 bp) and two introns (104 and 468 bp) coding for a protein of 1278 amino acids, slightly shorter than the well‐studied PM3 alleles (1415 amino acids) (Yahiaoui et al., 2004; Sánchez‐Martín et al., 2016). We analysed the predicted protein sequence for conserved domains and identified an N‐terminal coiled‐coil (CC) domain of 198 amino acids, a central nucleotide binding domain (NB) of 390 amino acids and 23 C‐terminal leucine rich repeats (LRRs) of various length (16–47 amino acids) (Fig. 3; Methods S5).
The susceptible Pm2 allele in cultivar Chinese Spring encodes a truncated protein (422 amino acids) that arose from a 7 bp deletion introducing a frame shift and a premature stop codon. In addition, within the 422 amino acids encoded by the truncated version of the gene, we found 14 SNPs between susceptible and resistant Pm2 alleles leading to 7 amino acid polymorphisms and one deletion (Fig. 3). Blast searches against the Wheat Survey Sequence derived from the cultivar Chinese Spring indicated that Pm2 has three homologues on chromosome 5DS sharing 98, 82 and 81% identity at the nucleotide level, respectively. The sequence of the closest homologue of Pm2 on 5DS was almost identical to that of the susceptible Pm2 allele (one SNP), suggesting that it is the same gene as the susceptible allele of Pm2 described above. We also found two homologous sequences on chromosome 5BS (87 and 81% identity), and one on chromosome 5AS (95% identity). We also searched for Pm2 homologous sequences in the barley High Confidence Genes database and found one sequence on chromosome 6H, MLOC_11605, with only 53% identity at the amino acid level (Table S9). The MLOC_11605 gene encodes for an NBS‐LRR protein annotated as a ‘Disease resistance protein’ that has 13 orthologues and three paralogues in barley (http://plants.ensembl.org). These data suggest that Pm2 is a unique, single gene on wheat chromosome 5DS and there is no orthologue or closely related homologue in barley.
To functionally validate AvrPm2, we used the transient agroinfiltration assay in N. benthamiana described by Bourras et al. (2015). The AvrPm2 gene was cloned from the avirulent parent 96224, and constructs expressing protein versions with and without signal peptide were transiently co‐expressed with Pm2 in a 4 : 1 Avr : R ratio (optical density at 600 nm (OD600) = 1.2, Fig. 4a). We also co‐expressed the experimentally validated AvrPm3 a2/f2–Pm3a Avr‐R gene pair as a control (Fig. 4b,c). A strong hypersensitive cell death response was observed when AvrPm2 was co‐expressed with Pm2, whereas no HR was observed when these constructs were co‐expressed with the GUS control (Fig. 4b,c). Visual inspection of the HR induced by Pm2 in the presence of the native AvrPm2 construct expressing the full‐length protein revealed a weaker HR compared to the AvrPm2 construct without the signal peptide (Fig. 4e,f). These results demonstrate that AVRPM2 is the cognate avirulence protein of PM2, and suggest that the AVRPM2–PM2 interaction takes place in the cytosol or the nucleus. Similarly, visual inspection of several replicates of the AvrPm2–Pm2 triggered HR indicated it was consistently stronger than the AvrPm3 a2/f2–Pm3a positive control. Therefore, we analysed the infiltrated N. benthamiana leaves by fluorescence imaging and quantified side by side the hypersensitive cell death response induced by Pm2 and Pm3 in the presence of 4 : 1 and 2 : 1 Avr : R ratios. We found that the HR induced by Pm2 was always significantly stronger than Pm3 independently of the Avr : R ratio (4 : 1, 2.3× stronger; 2 : 1, 3.5× stronger, Fig. 4d; Tables S10, S11). In addition, there was only a slight but not significant reduction in HR intensity when Pm2 was co‐infiltrated with lower amounts of AvrPm2 construct (2 : 1 ratio of Avr : R), whereas a significant HR reduction of 45% was observed for Pm3 (Fig. 4d; Tables S10, S11). The HR was even weaker and not visible by eye for Pm3 with a ratio of 2 : 1 while it was clear and strong for Pm2 (Fig. S6).
We wanted to compare gene expression of AvrPm2 and AvrPm3 a2/f2 in the avirulent parent 96224. We isolated RNA from leaf segments of the susceptible wheat cultivar Chinese Spring inoculated with the Pm2 and Pm3a avirulent isolate 96224. Samples were collected at six time points from 1 to 8 d post‐inoculation (dpi), and relative gene expression over time was assessed by qRT‐PCR. AvrPm2 is expressed at lower amounts than AvrPm3 a2/f2 and shows an almost constitutive expression over time (Figs 5a, S7). Scale magnification revealed very low mRNA amounts of AvrPm2 at 1 dpi, and a mild but sustained increase in expression between 3 and 4 dpi that is evidence for a shift of gene expression kinetics towards later infection stages as compared to AvrPm3 a2/f2 (Fig. 5b).
We found an effector family of AvrPm2 homologues that contained 24 members, 14 from B. g. tritici and ten from B. g. hordei. In B. g. tritici, the family includes all four predicted effector genes encoded within the AvrPm2 locus (Fig. 1d; Methods S4), and an additional effector, BgtAcSP‐30091, located on contig 48 that is genetically linked to contig 51 and these are hereafter referred to as ‘Cluster_1’ (Fig. 6b). We found a second genetically unlinked and physically distant cluster of four members (Cluster_2) located on the same linkage group as AvrPm2 (Fig. 6b). Five additional members were found as singletons in genetically unlinked sequences in contigs 33, 62 and 133 or in unassembled sequences for which we have no genetic data. The Y(x)xC motif was conserved among the gene family, and an additional conserved protein motif consisting of one arginine, followed by a non‐conserved residue, a phenylalanine or a tyrosine, and a highly conserved proline that we refer to as the ‘RxFP motif’ (Fig. S8). A disulphide bridge between the conserved cysteine residues was predicted in nine out of the 14 B. g. tritici family members including AvrPm2 (Table S12). Phylogenetic analyses were performed using the sequences encoding the full protein as well as the mature peptide after removal of the secretion signal and including all family members from both wheat and barley mildews. The phylogenetic trees were identical independently from the presence/absence of the signal peptide and revealed several gene duplication and gene loss events after the divergence between those two formae speciales of mildew. We then tested for positive selection as described in Yang et al. (2000, 2005) and found that the AvrPm2 family is under diversifying selection, with the sequence after the signal peptide region accumulating positively selected amino acids (Fig. S8). These results are similar to those reported for AvrPm3 a2/f2, suggesting common patterns of evolution among Avr effector families in wheat powdery mildew.
One member of the AvrPm2 family in B. g. hordei, CSEP0372, is Avr a13 that is recognized by the barley powdery mildew resistance gene Mla13 (Lu et al., 2016). Two additional members of the AvrPm2 homologous family in B. g. hordei are the well‐characterized mildew CSEPs BEC1011 and BEC1054 (Pedersen et al., 2012; Pliego et al., 2013; Pennington et al., 2016). Previous studies have shown that BEC1011 and BEC1054 belong to the RNase‐like class of mildew effectors (Pedersen et al., 2012). Therefore, we tested all the members of the AvrPm2 family from wheat and barley powdery mildew in structural protein modelling predictions using RaptorX (Källberg et al., 2012). Consistent with the previously predicted RNase‐like structure of BEC1011 and BEC1054, we found structural homologies to fungal ribonucleases throughout the AvrPm2 family in both wheat and barley mildews (Fig. 7; Table S13). This demonstrates that two members of the RNase‐like class of effectors act as avirulence factors in powdery mildews.
We also searched the rye and triticale mildew genomes for AvrPm2 orthologous genes. In rye mildew (B. g. secalis), we found an orthologue, BgsE‐5845, that differs from BgtE‐5845 by two amino acids (Fig. 4a). We found no polymorphism in the BgsE‐5845 sequence among the five B. g. secalis isolates we analysed (Menardo et al., 2016). In triticale mildew (B. g. triticale), which is genetically very similar to wheat mildew, we identified only one variant of AvrPm2 carrying a single mutation in the isolate CAP‐39‐A1 (BgtriticaleE‐5845) (Fig. 4a). By contrast, we found no direct homologue to AvrPm2 in the barley powdery mildew genome, and the closest gene was CSEP0066 (48% identity at the amino acid level).
All AvrPm2 haplotypes and direct/closest homologues identified from triticale, rye and barley powdery mildews were tested for recognition by Pm2 using transient assays in N. benthamiana. We also tested BgtE‐5842, BgtE‐5843 and BgtE‐5846, three additional effectors encoded within the AvrPm2 locus (Fig. 1c). All transient co‐expression assays involving the additional effectors encoded within the AvrPm2 locus and the closest AvrPm2 homologue in B. g. hordei resulted in no HR, indicating these effectors are not recognized by Pm2 (Fig. S6). However, all transient co‐expression assays involving the AvrPm2 haplotype in B. g. triticale (BgtriticaleE‐5845), the B. g. secalis homologue (BgsE‐5845) as well as two AvrPm2 variants harbouring either one of the SNPs identified in BgsE‐5845 (BgtE‐5845_T62I and BgtE‐5845_R118C) resulted in strong HR, indicating conservation of a functional AvrPm2 gene in powdery mildew forms specialized on wheat, triticale and rye (Figs 4a, S6).
In this study we cloned the mildew avirulence gene AvrPm2 and showed that it is not related to the only previously known avirulence gene in B. g. tritici, AvrPm3 a2/f2. AvrPm2 is a member of the family of RNAse‐like effectors that includes Avr a13, a barley mildew effector recognized by the Mla13 resistance gene. We also showed that Pm2 recognizes a homologue of AvrPm2 from rye powdery mildew, BgsE‐5845.
For this work, we combined map‐based cloning, next generation sequencing and high‐throughput genotyping to clone AvrPm2 in a similar approach to that used for the identification of AvrPm3 a2/f2 (Bourras et al., 2015). Given that map‐based cloning is time consuming, we also used GWAS to identify AvrPm2 from a panel of 60 natural isolates showing a balanced pattern of virulence on Pm2 (33 avirulent, 27 virulent). We used whole genome sequencing data for GWAS and the best associated SNP for AvrPm2 correctly identified the genetic region, but was actually below the threshold for statistical significance after standard correction for multiple testing. It was located in a non‐genic region at 50 kb from a large 12 kb deletion that contained the Avr. Thus, the use of genomic sequences instead of transcriptomics was crucial for finding AvrPm2 because the SNP was located in a non‐genic region. In barley powdery mildew, an approach using transcriptomic data led to the cloning of two novel avirulence genes, Avr a1 and Avr a13 (Lu et al., 2016). The two case studies now available demonstrate that association studies in a geographically diverse set of isolates can accelerate Avr gene cloning in powdery mildews.
We found by transient expression in N. benthamiana that Pm2 recognizes AvrPm2 and that the HR observed in the AvrPm2–Pm2 interaction is stronger than that in the AvrPm3 a2f2 –Pm3a interaction. In addition, Pm2 can recognize smaller ratios of Avr constructs. Quantification was based on fluorescence imaging, which specifically allows the visualization of the accumulation of phenolic compounds associated with the HR in N. benthamiana (Chaerle & Van Der Straeten, 2000; Chaerle et al., 2007). This method also allows rapid inspection of the area undergoing HR and the detection of weak responses that are not visible or difficult to reveal by classical chemical or histochemical methods (e.g. Trypan blue staining; Ma et al., 2012). Fluorescence can be precisely quantified using standard image analysis software such as ImageJ (Schindelin et al., 2015) and the data can then be normalized and differences can be statistically tested. In addition, no histochemical pre‐treatment or staining is required, which makes it very suitable for screening large numbers of interactions.
On the pathogen side, avirulence to Pm2 is controlled by an effector encoding gene whose virulent allele is the result of a 12 kb deletion that includes the Avr gene. In the case of the AvrPm2 deletion, the genomic context might have played a role: the low‐copy region containing AvrPm2 is flanked by five Copia retrotransposons that all belong to the Ino family, and which all share at least 80% sequence identity. Most importantly, the two elements flanking the AvrPm2 region are in the same transcriptional orientation. Thus, it is possible that this sequence organization provided the template for an unequal recombination event that led to the deletion of the low‐copy region containing AvrPm2 (Fig. S9; Georgiev et al., 1989). Additional examples of gain of virulence by partial or complete deletion of the Avr gene have been described in barley powdery mildew (e.g. Avr a13; Lu et al., 2016) and the hemibiotroph pathogen Magnaporthe oryzae (e.g. AvrPib; Zhang et al., 2015), indicating that gene loss is a common mechanism for Avr turnover in biotrophic fungi. This is very different from the AvrPm3 a2/f2 –Pm3a/f interaction, where gain of virulence is mediated either by point mutations of the AVR protein or the action of a suppressor of Avr recognition (Bourras et al., 2015). In flax rust, another obligate biotrophic fungus, gain of virulence is exclusively based on amino acid polymorphism for all AVRs cloned so far. There, all Avr–R pairs can be described as gene‐for‐gene interactions, although an additional gene whose identity and mode of action are still unknown can act as a suppressor of Avr recognition (Ellis et al., 2007; Ravensdale et al., 2011). Thus, there are additional layers of complexity and a larger variety of genetic mechanisms controlling Avr–R interactions in powdery mildews that have not yet been identified in other fungal systems. These can range from a simple gene‐for‐gene model (AvrPm2–Pm2; Avr a13–Mla13) to epistatic interactions and multi‐loci determinism that is reminiscent of a quantitative trait (AvrPm3–Pm3).
AvrPm2 belongs to an effector family of 26 members that is sequence unrelated and structurally distinct from the AVRPM3 A2/F2 family. In addition to a conserved Y(x)xC‐like motif found at the N‐terminal end of the predicted mature peptide, the AVRPM2 family also shares an RxFP motif separating the two N‐ and C‐terminal highly variable regions, which is reminiscent of the RXLR–dEER motif in oomycetes (Dou et al., 2008). There, it was shown that these motifs are sufficient for the translocation of pathogen effectors into the host cytoplasm (Dou et al., 2008). Therefore, the strong conservation of the Y(x)xC–RxFP motifs in otherwise highly divergent effector sequences suggests an important role of these residues in effector protein function.
The AvrPm2 family encodes effectors with structural homology to ribonucleases (RNase‐like) and contains the barley Avr a13 gene as well as BEC1011 and BEC1054. The latter two effectors have been shown to be important for mildew virulence (Pliego et al., 2013; Pennington et al., 2016). AvrPm2 and Avr a13 are homologues and encode structurally conserved effectors. However, their cognate resistance genes, Pm2 and Mla13, encode sequence‐unrelated and evolutionary distinct NLRs from wheat and barley. Assuming direct AVR–R binding in these interactions, this could be explained by convergent evolution of PM2 and MLA13, resulting in recognition of structurally conserved effectors. Another possible hypothesis is that AVRPM2 and AVRa13 both interact with a conserved protein in wheat and barley that is guarded by PM2 and MLA13, respectively. The AVRPM2–PM2 and the AVRa13–MLA13 interactions can be functionally studied by site‐directed mutagenesis and domain swaps of the two AVR proteins. Such studies, in addition to biochemical work, will result in a deeper understanding of recognition specificity and reveal whether there is convergent evolution of effector recognition or conservation of Avr targets.
We found that all B. g. tritici avirulent isolates contained the same haplotype of AvrPm2. Two additional haplotypes were found in the B. g. triticale isolate CAP‐39‐A1 and in the orthologous gene of AvrPm2 in B. g. secalis, indicating conservation of AvrPm2 in wheat, triticale and rye powdery mildews. B. g. triticale is a hybrid of B. g. tritici and B. g. secalis, two mildew formae speciales that diverged c. 200 000 yr ago which is quite recent compared to their hosts, wheat and rye that diverged 4 million yr ago (Middleton et al., 2014). In all Pm2 virulent isolates, we found a very similar 12 kb deletion, independent from the geographical origin (Europe, Israel, China), indicating that the virulent avrPm2 allele is also conserved, and suggesting there is no naturally occurring virulent protein variant for AVRPM2. A contrasting example can be found in Magnaporthe oryzae where the virulent alleles of AvrPib arose from a variety of gain‐of‐virulence mechanisms including TE insertion, segmental deletion, complete gene deletion and point mutations (Zhang et al., 2015). There are different possible explanations for the presence of very similar or identical deletions in isolates from different geographical origins: either the AvrPm2 deletion is ancient and occurred before the global cultivation of wheat and rye, or the deletion is associated with a better fitness of the pathogen and spread rapidly around the world after its occurrence.
Many disease resistance genes in bread wheat have been introgressed from wild and cultivated related species. These include genes such as Pm21 introgressed from the wild species Haynaldia villosa (Cao et al., 2011) or the race‐specific Pm8 and Pm17 mildew resistance genes from rye (Yahiaoui et al., 2006; Hurni et al., 2013). It has long been hypothesized that the successful transfer of functional R genes from one species to another is based on the recognition of effectors conserved in pathogens specialized on different host species. AvrPm2 is highly conserved in wheat, triticale and rye mildews, and all variants in these formae speciales can act as the cognate Avr of Pm2. Therefore, AvrPm2 provides molecular evidence that functional R gene introgressions could indeed be based on recognition of effectors conserved in powdery mildew forms specialized on specific hosts. The identification of such conserved Avrs will help to detect redundant specificities of resistance genes in the gene pool used for wheat breeding. Such knowledge can be used in pre‐breeding to develop pathogen‐informed approaches for the identification of mildew resistance genes with non‐redundant effector recognition in wild and domesticated relatives that can then be introgressed into cereal crops.
C.R.P., S.B., D.Y. and B.K. designed the study. C.R.P., S.B., F.Z., J.S‐M., R.B., G.H., K.E.M., R.B‐D., F.P. and S.F. performed the experiments. C.R.P., S.B., F.Z., D.Y. and B.K. wrote and edited the manuscript. C.R.P., F.M., S.R., S.O., L.K.S. and T.W. performed analysis. F.Z., M.X., L.Y. and R.B.D. contributed data.
Please note: Wiley Blackwell are not responsible for the content or functionality of any Supporting Information supplied by the authors. Any queries (other than missing material) should be directed to the New Phytologist Central Office.
Fig. S1 Primer efficiency curve for qRT‐PCR primers for AvrPm2.
Fig. S2 Genetic and physical mapping based on the B. g. triticale/B. g. hordei genome collinearity.
Fig. S3 Assembly of the BACs spanning the AvrPm2 region.
Fig. S4 Identification of the AvrPm2 deletion in the Pm2 virulent isolate JIW2.
Fig. S5 Gel electrophoresis analysis of AvrPm2 deletion markers.
Fig. S6 Functional analysis of the interaction between Pm2 and different family members and homologues in B. g. triticale, B. g. secalis and B. g. hordei in N. benthamiana.
Fig. S7 Gene expression of AvrPm2 and AvrPm3 a2/f2 in the avirulent isolate 96224.
Fig. S8 Annotated protein alignment of the AVRPM2 effector family in B. g. tritici and B. g. hordei.
Fig. S9 Proposed molecular mechanism that could lead to the loss of the genomic region containing the AvrPm2 gene.
Table S1 Summary of the RNA samples used for qRT‐PCR
Table S2 List of markers used for gene annotation, cloning and expression analysis
Table S3 List of markers used for genetic mapping
Table S4 BAC end markers
Table S5 Markers used for estimating the size of the deletion
Table S6 Whole genome resequencing strategies for the 22 Chinese isolates
Table S9 Blast hits of PM2 against wheat and barley databases
Table S10 Results of the ANOVA for HR intensity in HR quantification assays
Table S11 Results of Tukey HSD post‐hoc test
Table S12 Signal peptide and disulphide bond predictions in the AVRPM2 family
Table S13 Structure prediction of the AVRPM2 in B. g. tritici and B. g. hordei
Methods S1 BAC selection and sequencing.
Methods S2 Illumina whole‐genome resequencing.
Methods S3 RNA extraction and qRT‐PCR assays.
Methods S4 Genome‐wide effector annotation and AvrPm2 family identification in different formae speciales.
Methods S5 In silico analysis of proteins encoded by Pm2 and AvrPm2.
Table S7 List of the B. g. tritici and B. g. triticale isolates used for genome‐wide association studies for the identification of AvrPm2
Table S8 Genetic map of the 96224 × JIW2 mapping population based on KASP markers
Notes S1 Sequences of the Blumeria graminis genes used for transient expression in Nicotiana benthamiana.
Notes S2 Coding sequences of the AvrPm2 family members in B. g. tritici and B. g. hordei.
Notes S3 Protein sequences encoded by AvrPm2 family members in B. g. tritici and B. g. hordei.
Notes S4 BAC selection, sequencing and assembly.
Notes S5 Genetic and physical mapping of AvrPm2 based on the B. g. tritici/B. g. hordei genome collinearity.
This work was supported by the Swiss National Science Foundation (grant 310030‐163260), the National Basic Research Program of China (2013CB127700) and the China Agriculture Research System (CARS0304B). We would like to acknowledge Prof. Dr Paul Schulze‐Lefert (Max Planck Institute for Plant Breeding Research, Köln, Germany) for sharing unpublished information about the cloning of Avr a13 in barley powdery mildew.
See also the Commentary on this article by Spanu, 213: 969–971.
Dazhao Yu, Email: moc.anihc@uyoahzad.
Beat Keller, Email: hc.hzu.tsnitob@rellekb.