A highly differentiated region of wheat chromosome 7AL encodes a Pm1a immune receptor that recognizes its corresponding AvrPm1a effector from Blumeria graminis

Summary Pm1a, the first powdery mildew resistance gene described in wheat, is part of a complex resistance (R) gene cluster located in a distal region of chromosome 7AL that has suppressed genetic recombination. A nucleotide‐binding, leucine‐rich repeat (NLR) immune receptor gene was isolated using mutagenesis and R gene enrichment sequencing (MutRenSeq). Stable transformation confirmed Pm1a identity which induced a strong resistance phenotype in transgenic plants upon challenge with avirulent Blumeria graminis (wheat powdery mildew) pathogens. A high‐density genetic map of a B. graminis family segregating for Pm1a avirulence combined with pathogen genome resequencing and RNA sequencing (RNAseq) identified AvrPm1a effector gene candidates. In planta expression identified an effector, with an N terminal Y/FxC motif, that induced a strong hypersensitive response when co‐expressed with Pm1a in Nicotiana benthamiana. Single chromosome enrichment sequencing (ChromSeq) and assembly of chromosome 7A suggested that suppressed recombination around the Pm1a region was due to a rearrangement involving chromosomes 7A, 7B and 7D. The cloning of Pm1a and its identification in a highly rearranged region of chromosome 7A provides insight into the role of chromosomal rearrangements in the evolution of this complex resistance cluster.

The corresponding avirulence effectors for several wheat NLR resistance genes, such as Pm2 and the Pm3 allelic series, were recently identified from Bgt. Map-based cloning using haploid F 1 populations enabled the identification of AvrPm2, AvrPm3 a2/f2 , AvrPm3 b2/c2 and a suppressor locus of the Pm3 allelic series SvrPm3 a1/f1 (Bourras et al., 2015Praz et al., 2017). These Avr proteins show very little sequence similarity to each other or with known proteins. However, these Bgt Avr proteins are small (109-130 amino acids) and may, from modelling, share some structural similarities (Praz et al., 2017;Bourras et al., 2019).
Previous attempts to isolate Pm1 and Lr20 by map-based cloning were unsuccessful due to restricted genetic recombination, which may be due to a translocation from an unknown source to the distal region of chromosome 7AL (Neu et al., 2002). In this study, new techniques were employed to overcome this limitation for positional gene isolation. Specifically, the method of mutagenesis and resistance (R) gene enrichment sequencing (MutRenSeq) is a targeted approach that does not depend on recombination but instead utilizes a sequence bait library to enrich NLR encoding sequences from knockout mutants for the target R gene. Comparing random mutations between independent mutants allows the identification of a candidate gene. Herein, we report the successful cloning of Pm1a from common wheat using MutRenSeq, and the corresponding AvrPm1a effector from the wheat powdery mildew fungus Bgt using genetic mapping and RNA sequencing (RNAseq). Additionally, the method of chromosome sequencing (ChromSeq) allowed us to isolate sequences from Pm1a bearing chromosome 7A, showing evidence of rearrangement leading to suppressed recombination.

Plant materials
Six M 5 putative point mutations for Pm1a in Chinese Spring*5/ Axminster 7A (CS/Ax7A) substitution line carrying the Pm1a locus on chromosome 7A from Axminster (Sears & Briggle, 1969) were used in RenSeq analysis (mutants #396,404,428,435,446 and 650,see Supporting Information Fig. S1). F 5 RILs used in genetic analysis and confirmation of a diagnostic marker developed from the gene candidate were derived from a combination of three different crosses segregating for Pm1a, namely, CS 9 CS/Ax7A, CS 9 Thew and Thew 9 CS. Additional mutants from Thew, along with cultivars Norka and Schomburgk were also used to test Pm1 markers. Transformable wheat cultivar Fielder, which does not carry Pm1, was used for transgenic complementation. CS/Ax7A was used in flow sorting of chromosome 7A and sequencing.

Screening for powdery mildew response
The avirulent Bgt inoculum used for screening Pm1a was collected from the glasshouse area of the CSIRO Canberra campus (Australia) and maintained on seedlings of susceptible cultivar Morocco for the duration of study. Screening of the mutagenized and genetic populations took place in an isolated glasshouse at the University of Sydney Camden Campus. Additional powdery mildew tests were similarly conducted at CSIRO Canberra with CS/Ax7A and Thew included as resistant controls. Conidiospores were shaken or brushed directly onto 2-week-old seedlings grown in a disease-free glasshouse. CS/Ax7A and CS were used as resistant and susceptible controls, respectively. Phenotyping of seedlings was simply as resistant or susceptible as Pm1a confers a completely immune response (infection type (IT) 0) compared to CS (IT 3+).

MutRenSeq pipeline
Treatment with EMS was performed as described in Sharp & Dong (2014) on 600 and 300 seeds of CS/Ax7A in 0.5% and 0.6% EMS, respectively. Seeds of Thew were also treated with 0.5% and 0.6% EMS using 300 seeds in each. Thus, 572 and 373 M 1 plants of CS/Ax7A and Thew were harvested, respectively. Homozygous CS/Ax7A mutants retaining the closely linked Lr20 and Sr15 from six different M 2 plants were used in subsequent R gene enrichment. DNA was prepared from leaves of uninfected seedlings of the six mutants and one wild type (WT). DNA extraction, target enrichment using Triticeae RenSeq Bait Library V2 (https://github.com/steuernb/Muta ntHunter), Illumina sequencing, de novo assembly and read mapping were carried out as described in Steuernagel et al. (2016). Single nucleotide polymorphism (SNP) calling and candidate identification were performed using the MUTRIGO pipeline with default parameters (https://github.com/TC-Hewitt/MuTrigo).

Pm1a structure confirmation
The partial candidate contig identified by RenSeq (contig #8725) was aligned to the IWGSC Chinese Spring reference REFSEQ v.1.0 (CSv1) (IWGSC, 2018) using BLASTN (Zhang et al., 2000) and the top matching gene was aligned to the RenSeq WT assembly using BLASTN. High scoring pairs were filtered using a custom UNIX script and the corresponding candidate contig was chosen based on presence of expected domains and a SNP in the outstanding mutant. Both candidate contigs were confirmed to belong together based on Sanger sequencing of the PCR product from genomic DNA bridging the two sequences. RNA was extracted from leaves of CS/Ax7A using the RNeasy Plant Mini Kit (Qiagen, Chadstone Centre, VIC, Australia). Transcript structure, including flanking untranslated regions (UTRs), was obtained by 5 0 and 3 0 RACE (rapid amplification of complementary DNA ends) using a SMARTer RACE 5 0 /3 0 Kit (Clontech, Mountain View, CA, USA). The upstream promoter region, intron sequences and downstream terminator region were inferred from a large contig from the Ax7A de novo ChromSeq assembly bearing the Pm1a sequence. Coding sequence and translation were predicted using FGENESH (Solovyev et al., 2006).

Mutant confirmation
EMS-induced SNPs identified in mutants by MutRenSeq were confirmed with Sanger sequencing of PCR-amplified coding regions from mutants. Nonsynonymous amino acid substitutions were confirmed using CODONCODE ALIGNER 8.0 (https://www.c odoncode.com/aligner/).

Transformation confirmation of Pm1a
A 9386 bp genomic sequence of Pm1a, including all introns, 2 kb of 5 0 UTR and 1.5 kb of 3 0 UTR encompassing the native promoter and terminator, was synthesized and cloned (Epoch Life Sciences, Missouri City, TX, USA) into NotI/SgsI-digested binary vector VecBarIII. The Pm1a gene was introduced into cv Fielder using the Agrobacterium-transformation protocol (Ishida et al., 2015) and phosphinothricin as the selective agent. Ten independent transgenic plants (T 0 ) carrying the Pm1a gene as well as three nontransgenic sibling control lines were recovered. The T 0 and sibling controls were acclimatized to glasshouse conditions. After 2 weeks, sibling controls, WT controls and T 0 plants were inoculated with Bgt as described earlier. Then, 10 cm leaf samples were taken from the second leaf at 10 d post-inoculation (dpi) and again from the third or fourth leaves at 14 dpi. Leaves were scanned using an Epson Perfection V600 Photo scanner immediately after sampling. For quantification of fungal biomass, 10 and 14 dpi leaf samples were processed and stained for chitin as described by Ayliffe et al. (2013). Relative fluorescence was measured on a FLUOstar Omega spectrophotometer (BMG Labtech, Mornington, VIC, Australia) with excitation filter 485-12 and emission filter Em520.
Genetic fine mapping was performed manually based on nine recombinant progeny in the QTL interval (defined by flanking markers snp64149 and snp64379) that showed virulence scores of 0 or 1. Four recombinant progeny showing intermediate phenotypes were excluded. A single recombinant progeny did not fit into our proposed high-resolution map and would place AvrPm1a upstream of marker snp64348. However, this recombinant would exclude any of the candidates in the QTL interval. Given the partially quantitative nature of the AvrPm1a phenotype this recombinant was tentatively excluded from the analysis.

Candidate AvrPm1a identification
AvrPm1a candidates were identified by manual curation of gene models in the candidate interval based on assembly, annotation and candidate effector definition of Bgt_genome_v3_16 reported in M€ uller et al. (2019). To verify the annotation, RNAseq data of both parental isolates were mapped against the reference genome with STAR (v.2.6.0a) (Dobin et al., 2013) as described in Praz et al. (2018) and visualized with IGV (v.2.8.0) (Robinson et al., 2011). The RNAseq dataset was generated previously, from susceptible cv Chinese Spring and triticale cv Timbo infected with 96224 or THUN-12, respectively (Menardo et al., 2016;Praz et al., 2018). Infected leaf samples were collected at 2 dpi during Bgt haustorium formation. The final set of candidate genes in the interval are listed in Supporting Information Table S1. The erroneous gene models of Bgt-50400 and Bgt-50399, two nonsignal peptide-containing genes, were excluded for lacking RNAseq support or missing start codon, respectively. To identify SNPs between the parental isolates, genomic Illumina reads of isolate THUN-12 were mapped against the reference as described in M€ uller et al. (2019). Protein domains of candidate effectors were predicted using PFAM (v.33.1) (https://pfam.xfam.org/). Signal peptide prediction was based on SIGNALP (v.3.0) (Bendtsen et al., 2004) with default settings. Alignment of the E004 candidate effector family was done with the Clustal algorithm of MEGA X (Kumar et al., 2018) with default parameters. For expression analysis and differential gene expression, RNAseq reads were aligned against the Bgt 96224 CDS (M€ uller et al., 2019) with SALMON (v.0.12.0) (Patro et al., 2017) as described in Praz et al. (2018). Subsequent expression and differential expression analysis were done using the EDGER (v.3.11) (Robinson et al., 2010) package in RSTUDIO (RStudio-Team, 2018) as described in Praz et al. (2018).

Transient co-expression assay in Nicotiana benthamiana
Effector candidate genes were codon-optimized for expression in Nicotiana benthamiana using the online tool of Integrated DNA Technologies (https//eu.idtdna.com/CodonOpt). Signal peptides were predicted using the SIGNALP algorithm (http://www.cbs.d tu.dk/services/SignalP-3.0/) and subsequently replaced with a start codon. For hemagglutinin (HA)-epitope tagged effectors the corresponding sequence was added at the C-terminus directly upstream of the stop codon. The resulting sequences predicted to encode the mature peptide, including flanking attL gateway sites, were synthesized by BioCat GmbH (https://www.biocat.com). Sequences for all synthesized DNA fragments are described in Supporting Information Dataset S1. The synthesized effector genes were subsequently cloned into the binary vector pIPKb004 New Phytologist (2021) 229: 2812-2826 Ó 2020 The Authors New Phytologist Ó 2020 New Phytologist Foundation www.newphytologist.com
The Pm1a resistance gene was synthesized in two overlapping fragments (Dataset S1) with flanking sequences matching the binary vector pIPKb004 (Himmelbach et al., 2007) by BioCat GmbH. For the HA-epitope tagged version the HA coding sequence was introduced by two subsequent rounds of PCR with the primers listed in Table S2. The two resulting fragments were cloned by In-Fusion Cloning (Takara, Tokyo, Japan) into a modified pIPKb004 plasmid (Himmelbach et al., 2007) in which the gateway cassette was removed using the restriction enzymes HindIII and BsrGI. All pIPKb004-based binary constructs were verified by Sanger sequencing and transformed into Agrobacterium tumefaciens strain GV3101 using a freeze-thaw transformation protocol (Weigel & Glazebrook, 2006). Agrobacterium mediated expression in N. benthamiana and hypersensitive response (HR) assessment was performed 5 d after Agrobacterium infiltration according to the protocol described in Bourras et al. (2019).

Protein detection
To assess whether the effector and R protein were transiently expressed in transformed N. benthamiana leaves we followed the protocol for protein extraction, sodium dodecyl sulphate-polyacrylamide gel electrophoresis (SDS-PAGE) and Western blotting in Bourras et al. (2019). For HA detection we used a peroxidase-conjugated antibody, anti-HA-HRP (rat monoclonal 3F10, Roche, Basel, Switzerland) at a dilution factor of 1 : 3000 in the presence of Supersignal Western blot enhancer solution (Thermo Scientific, Waltham, MA, USA) according to the manufacturer. Peroxidase-based chemiluminescence was imaged using WesternBright ECL HRP substrate (Advansta, San Jose, CA, USA) and a Fusion FX Imaging System (Vilber Lourmat, Eberhardzell, Germany) with default settings.

Microscopy
Infected leaf segments (3 cm) were taken at 10 and 14 dpi and treated by submersion in 1 M potassium hydroxide (KOH) and autoclaved. Treated leaves were washed and resuspended in 50 mM Tris-HCl pH 7.5. Samples were stained with WGA-FITC and aniline blue. Fluorescence microscopy was carried out with an AxioImager Epifluorescence Widefield Microscope (Zeiss, Oberkochen, Germany) fitted with FITC, GFP and DAPI filters.

Fluorescence in situ hybridization
Fluorescence in situ hybridization (FISH) was carried out with oligo probes (Oligo-pSc119.2 and Oligo-pTa535 producing green and red signals, respectively) on CS and Lr20-carrying lines (CS/Ax7A, Kenya W744, Thew, Norka) according to J. . Chromosome 7AL in Lr20-carrying lines had a weak pSc119 signal at the distal region, whereas in CS this signal was absent. Using this karyotype, flow-sorted chromosomes were confirmed to be 7A.

Axminster 7A assembly comparison to the Chinese Spring reference
After trimming with cutadapt (https://pypi.org/project/cutadapt/), Ax7A ChromSeq reads were aligned to the CSv1 reference sequence assembly using MINIMAP2 (https://github.com/lh3/ minimap2). Alignment records were sorted with NOVOSORT (http://www.novocraft.com/products/novosort/) and converted to BAM format with SAMTOOLS . Read counts in nonoverlapping 1 Mb windows along the genome were determined using SAMTOOLS and standard UNIX tools. Binned read counts were imported into R and plotted along the genome using standard R functions. SNP calling was done with BCFTOOLS (https://samtools.github.io/bcftools/). SNP positions were imported into R and binned in 5 Mb windows. SNP densities were plotted along chromosome 7A using standard R functions. The functions used in this analysis are available at https://bitbucket. org/ipk_dg_public/pm1a.
Plot of alignment scores of CSv1 chromosome 7A genes against the Ax7A assembly was created by first extracting gene sequences from CSv1 using BEDTOOLS (https://github.com/ arq5x/bedtools2) and a GFF file of CSv1 gene annotations. Genes from only chromosome 7A were aligned to the Ax7A assembly using BLASTN. Outputs were filtered for top alignment over 1000 bp per query based on bit score using BLAST_FILTERV2.PYC (https://github.com/TC-Hewitt/Misc_NGS). Bit scores were normalized to bit score/kb alignment length and alignments were ordered by physical position using a custom UNIX script. Output was imported into EXCEL and plotted. CS IWGSC REFSEQ v.2.0 (CSv2) (www.wheatgenome.org) was used for coordinate-based physical comparisons due to improved assembly accuracy over CSv1 (although annotation data were not yet available). Ax7A www.newphytologist.com contig mapping to CSv2 chromosomes 2A, 3A, 7A, 7B and 7D was performed in parallel using MASHMAP (Jain et al., 2018) with identity threshold 91%, minimum segment length 5000 bp and filter mode as one-to-one. A custom UNIX script was used to filter for best alignment per query based on the longest alignment length above identity threshold. Dotplots were generated using GENERATEDOTPLOT (https://github.com/marbl/MashMap). Putative breakpoints were inferred from coordinate data in filtered MASHMAP outputs.
Axminster 7A assembly comparison with diploid species Decontamination of the Ax7A assembly of likely chromosomes 2A and 3A derived contigs was achieved using previously described MASHMAP outputs to flag contigs having top alignments with ≥ 95% identity to either 2A or 3A. These contigs were removed from the MASHMAP output which was then used as an index for GET_CONTIGS.PY (https://github.com/TC-Hewitt/ Misc_NGS) to extract filtered contigs to a separate FASTA file. A 'terminal contig set' was created by filtering the shortlisted MASHMAP output for contigs mapping distally to the respective breakpoints using a custom UNIX script. The output was then used for contig retrieval using GET_CONTIGS.PY. All 'custom UNIX scripts' cited throughout the methods can be found at https:// github.com/TC-Hewitt/Axminster7A. Mapping of genotyping-by-sequencing (GBS) data from diploid accessions was performed by first trimming raw reads using TRIMMOMATIC (http://www.usadellab.org/cms/?page= trimmomatic). Each of the 15 species had four accessions which were concatenated into a single file per species. Mapping to the whole Ax7A assembly and 'terminal contig set' was performed using BWA ). The output was processed with SAMTOOLS to remove duplicate alignments. Uniquely mapped reads were selected based on SAM tag 'XT:A:U' and exact matching reads were counted based on SAM tag 'NM:i:0'. Counts were imported into Excel and plotted as a percentage of total uniquely mapped reads per species to normalize for variation in absolute reads between libraries.
Phylogenetic tree construction R gene protein sequences with an N-terminal coiled-coil domain (CNL class) were taken from the National Centre for Biotechnology Information (NCBI) database. Accession numbers are listed in Table S3. One hundred and twenty-two sequences were aligned using MUSCLE and a phylogenetic tree was generated using the UPGMA (unweighted pair group method with arithmetic mean) method in MEGA X (Kumar et al., 2018).

Primer design and sequence resources
The F 5 lines from CS 9 CS/Ax7A were tested with a Pm1a-specific dominant sequence-tagged site (STS) marker Pm1aSTS1, designed based on the positions of SNPs in the alignment of CSv1 sequences homologous to contig #8725 harbouring the Pm1a candidate. Primers used in this study are listed in Table S4.
Genomic and transcriptomic resources used for the identification of AvrPm1a were reported previously: genome assembly and annotation of Bgt 96224 (Bgt_genome_v3_16) in M€ uller et al.

Cloning of Pm1a by MutRenSeq
To clone Pm1a, we identified susceptible EMS-generated mutants in CS/Ax7A background. Six independent mutants together with WT CS/Ax7A were processed using the MutRenSeq pipeline. Captured reads from these six lines and WT were aligned to a de novo reference assembly of the WT reads. One contig (#8725) of 3096 bp contained a SNP in five of the six mutants. This contig contained NB-ARC and LRR motifs but no coiled-coil (CC) motifs suggesting it might be a partial NLR sequence. To identify a potentially missing segment, the contig was aligned to the CSv1 reference assembly. The top hit (87.5%) was to the 3 0 portion of a high confidence gene (TraesCS7D01G540500) predicted on chromosome 7D and functionally annotated as a disease resistance gene. The full sequence of TraesCS7D01G540500 aligned back to the CS/Ax7A RenSeq WT assembly detected an additional 10 contigs matching the 5 0 portion at 81% to 92% identity. Only one of these contigs with 91.8% identity (#3966) had a SNP in the remaining mutant and contained a CC domain. Bridging PCR was used to confirm that both contigs (#8725 and #3966) formed a single NLR gene with SNP mutations in all six mutants (Fig. S1). All SNPs were canonical for EMS mutagenesis with either C/T (one mutant) or G/A (five mutants) polymorphisms. A full transcript sequence was obtained using 5 0 and 3 0 RACE on RNA extracted from CS/Ax7A. SNPs were confirmed in all mutants by Sanger sequencing and determined to cause either amino acid substitutions or a premature stop (Fig. 1) in the translated protein. The full gene sequence of Pm1a was 5.9 kb consisting of three exons and two introns of 1816 and 88 bp. The encoded protein was 964 amino acids and contained N-terminal CC, NB-ARC and C-terminal LRR domains (Fig. 1).

The NLR gene candidate cosegregates with Pm1a resistance in RIL populations
A dominant STS marker, Pm1aSTS1, was designed from the contig #8725 sequence and amplified only in backgrounds carrying Pm1a (Fig. S2). The 3 kb amplicon based on the full-length contig #8725 did not amplify in mutants that had lost both Pm1 and Sr15/Lr20 resistance consistent with these mutants (not used for RenSeq) containing deletions (Fig. S3). Seventy-six CS/Ax7A F 5 RILs and 157 CS/ Thew F 5 RILs were then screened with the Bgt isolate avirulent to Pm1a. Resistant (n = 123), segregating (n = 6) and susceptible (n = 104) plants displayed clear phenotypes (Fig. S4)

New Phytologist
The Pm1a gene candidate confers powdery mildew resistance in transformed plants Susceptible cv Fielder was transformed with the Pm1a candidate and 10 independent transgenic lines (T 0 ), with controls, were tested with Bgt. At both 10 and 14 dpi (not shown) all T 0 lines showed resistance to Bgt similar to that observed on CS/Ax7A (Fig. 2b). The high-level immunity was confirmed by fungal biomass measurements (Fig. 2a). The T 1 progeny from four independent T 0 lines were screened with Bgt and showed a resistant : susceptible segregation ratio of 3 : 1 or greater, indicating the presence of a stable and active Pm1a transgene in one or more loci of the T 0 parents. All resistant T 1 seedlings cosegregated with marker Pm1aSTS1. These data confirmed that this NLR gene conferred Bgt resistance in wheat and is Pm1a.
The Pm1a phenotype was examined microscopically in Bgt-infected T 0 leaf tissue at 10 dpi. Hyphal growth was advanced in susceptible controls (Fig. 3a) and secondary infection was evident by the presence of haustoria within epidermal cells (Fig. 3b). Conversely, on leaves of T 0 Pm1a transgenics, only germinated conidiospores that produced appressoria were observed without development of conspicuous haustoria (Fig. 3c,d). Leaves from WT CS/Ax7A plants showed minimal infection similar to that of T 0 plants (Fig. 3e), although more advanced haustorium formation was occasionally observed (Fig. 3f). Plant cell autofluorescence, indicative of host cell death (S anchez-Mart ın et al., 2011), was frequently associated with Bgt infection sites on T 0 plants and resistant WT controls (Fig. 3d,g,h), possibly indicating HR. Autofluorescence was not associated with haustoria and secondary hyphae produced in susceptible controls.

Cloning of AvrPm1a
To identify AvrPm1a we used an existing mapping population with 118 sequenced haploid F 1 progeny (M€ uller et al., 2019) originating from the cross between B. g. triticale isolate THUN-12, exhibiting an avirulence phenotype on Pm1a near-isogenic wheat line Axminster/8*Chancellor, and Bgt isolate 96224, which is virulent to Pm1a (Fig. S5a,b). Single interval QTL mapping with 118 progeny identified a single QTL on chromosome 6 (Logarithm of the odds (LOD) = 5.8) associated with the avirulence phenotype to Pm1a (Figs 4a, S5c). The genetic confidence interval (1.5 LOD) encompassed 210 266 bp in the published chromosome-level assembly of isolate 96224 (M€ uller et al., 2019) and harboured a cluster of seven predicted effector genes ( Fig. 4b-d; Table S1). Using whole-genome re-sequencing and RNAseq data we determined that all seven candidate effectors were present and expressed in the avirulent isolate THUN-12 (Table S1). The high-density genetic map and nine recombinant F 1 progeny with clear avirulence/virulence patterns on Pm1a lines allowed us to further reduce the AvrPm1a interval to 90 562 bp and the number of candidates to two effector genes, BgtE-5612 and BgtE-20015, located between the flanking markers (Fig. 4c,  d). Both genes were strongly expressed during early infection stages of B. graminis (2 dpi) and exhibited sequence polymorphisms between the parental isolates THUN-12 and 96224 ( Fig. 4d; Table S1). Compared to the reference isolate 96224, BgtE-5612_THUN12 carries five SNPs resulting in four amino acid changes, whereas BgtE-20015_THUN12 carries a single nonsynonymous SNP (Fig. 4d), making BgtE-5612 and BgtE-20015 originating from Pm1a avirulent isolate THUN-12 prime candidates for AvrPm1a.
Previously, Bgt avirulence genes were functionally validated by Agrobacterium-mediated co-expression of NLR genes and Avr candidates in the heterologous N. benthamiana system (Bourras et al., 2015Praz et al., 2017). Following a similar approach, we optimized BgtE-5612 and BgtE-20015 codon sequences from both parental isolates for expression in N. benthamiana; synthesized genes lacking the signal peptide were tested for recognition by Pm1a in Nicotiana leaves. Co-expression of BgtE-5612_THUN12 but not BgtE-5612_96224 with Pm1a in Nicotiana resulted in a strong HR (Figs 5a,b, S6a), whereas no such effect was observed for either version of BgtE-20015 (Fig. S6a). We fused all effector candidates and Pm1a Cterminally with a HA epitope tag and verified protein production of all effector variants and the R protein in Nicotiana by Western blotting (Figs. 5c, S6b,c).
These results demonstrated that BgtE-5612_THUN12 is AvrPm1a, which is consistent with the genetic mapping data predicting the AVR component to originate from the THUN-12 parental isolate. The recognized variant BGTE-5612_THUN12 differs from the virulent version BGTE-5612_96224 by four amino acids (K28E, L40F, G47R, C60S) occurring as a cluster in the N-terminal part of the protein, surrounding the Y/FxC motif (Fig. 5d).

Phylogenetic analysis of Pm1a and corresponding avirulence effector AvrPm1a
Comparison of the Pm1a protein sequence with a panel of 122 cloned coiled-coil (CNL class) NLR proteins (Fig. S7) showed that Pm1a is not closely related to other known Pm proteins such as the Mla allelic series or Pm3. In fact, Pm1a is relatively dissimilar to the other NLR proteins in this panel including other wheat NLR proteins. It shows greatest similarity to Pm21 which originates from Dasypyrum villosum (He et al., 2017).
AvrPm1a is a member of the effector family E004, one of the 235 previously described Bgt and B. g. hordei effector families (M€ uller et al., 2019) (Fig. S13, see later). Interestingly, 22 of the 51 genes that encode E004 family proteins, including AvrPm1a, reside within a 900 kb genomic region on chromosome 6. This arrangement is likely to have arisen from tandem duplication and diversifying selection imposed by NLR recognition of effector sequences, such as that seen for AvrPm1a. In this gene family AvrPm1a is amongst the most highly expressed genes (Fig. S8).
AVRPM1A shows some similarity to several other cloned Bgt avirulence effector proteins (i.e. AVRPM2, AVRPM3 A2/F2 , AVRPM3 B2/C2 and AVRPM3 D3 ) which contain, like AVRPM1A, a signal peptide, a N-terminal Y/FxC motif, a single, conserved cysteine towards the C-terminus while otherwise exhibiting highly divergent amino acid sequences. Interestingly, all known Bgt avirulence effectors exhibit a similar exon/intron structure and are predicted to consist of a N-terminal a-helix followed by several b-strands (Bourras et al., 2015

Research
New Phytologist 2017). In the case of AVRPM1A this pattern is extended by the presence of a second a-helix at the amino terminus, partially explaining the larger size of the effector (Figs. 5d, S14a,b, see later). Strikingly, all identified Bgt avirulence proteins, including AVRPM1A, exhibit an RNase-like fold when subjected to structural modelling (Figs. 5d, S14c, see later) (Bourras et al., 2015Praz et al., 2017).
Terminal chromosome 7AL is highly diverged due to rearrangement FISH was conducted on lines CS/Ax7A, Kenya W744, Thew, and Norka carrying Pm1-Sr15-Lr20, and CS. Using Oligo-pSc119.2 (green) and Oligo-pTa535 (red) as probes a weak pSc119 signal was present in the distal region of chromosome 7AL in all four resistant lines but not in CS suggesting the presence of an introgression or translocation (Fig. S9).
ChromSeq was conducted on purified DNA of flow sorted chromosomes 7A of line CS/Ax7A. The chromosome sample purity was estimated to be 72% for the final DNA library with contamination primarily from chromosomes 2A and 3A. The sequencing output was c. 69 Gbp. A de novo Ax7A assembly created from the ChromSeq reads comprised of 383 563 contigs with a combined length of c. 1.82 Gbp and N50 of 7.85 kb.
When trimmed ChromSeq reads were aligned to the entire CSv1 reference assembly and SNP counts were plotted along chromosome 7A, few SNPs were observed in peri-centromeric regions but a distinct increase in SNP density was observed at the distal end of 7A (Fig. S10a). This was mirrored by a discrete drop in alignment scores of CSv1 7A annotated genes aligned to the Ax7A assembly (Fig. S10b). These results indicate a clear differentiation in this distal region between chromosome 7A in the CS assembly and the sequenced Axminster 7A.
Counts of uniquely mapped reads were plotted along the length of each CSv1 chromosome (Fig. 6). Majority of the reads (57%) originated from chromosome 7A, with contamination from chromosomes 2A (21%), 3A (18%) and 4A (2-%). The sharp decline in read counts at the distal end of 7AL was accompanied by a concomitant rise in read counts at the distal ends of chromosomes 7BL and 7DL. When Ax7A assembly contigs were mapped to the CSv2 reference, the dot plot of 7A showed a downward inflection near the terminus indicating a decrease in homology. This placed a putative breakpoint at c. 728 Mb relative to CSv2 7A (Fig. 7a). Conversely, upward inflections indicating increased homology occurred at the termini of 7B and 7D, which placed putative breakpoints at c. 660 and 720 Mb and 630 Mb relative to CSv2 7B and 7D, respectively (Fig. 7b,c). These results suggest the terminal part of Axminster chromosome 7AL contains segments that are more related to sequences from terminal 7BL and 7DL. Based on the positions of putative breakpoints relative to CSv2 chromosomes 7A, 7B and 7D, the physical

Possible ancestral origins of terminal 7AL
Probable contaminant contigs were removed from the Ax7A assembly based on significant alignments to either chromosome 2A or 3A in CSv2 (Fig. S11). To investigate whether the Pm1-Sr15-Lr20 linkage block has potential origins in diploid wheat relatives, GBS data was sourced from an unrelated study on a collection of diploid accessions (Bernhardt et al., 2019). GBS reads of 15 species were aligned to the decontaminated Ax7A assembly. Unsurprisingly, T. urartu, the ancestral donor of the wheat A genome (Salamini et al., 2002) had the highest proportion of exact matching reads. Alignments of the other species were not noteworthy (Fig. S12a). A set of 480 contigs, provisionally representing the terminal portion of Axminster 7A, were grouped based on primary mapping after the putative breakpoints in either CSv2 7A, 7B, or 7D. To determine if any of the diploid species became more represented in the terminal portion of Axminster 7A, GBS reads similarly aligned to this terminal contig set indicated exact matching reads of D. villosum were notably higher than in the entire Ax7A alignment and were comparable to T. urartu (Fig. S12b). Given that Pm1a was also closest to Pm21 in the phylogeny (Fig. S8), this raised the possibility that portions of terminal Axminster 7A were derived from Dasypyrum or a related lineage.

Discussion
In this study, the R/Avr gene pair of Pm1a and AvrPm1a was successfully cloned using MutRenSeq and genetic mapping with  Amino acid polymorphisms in the coding sequence of the effector genes are depicted by red lines. Details about SNPs and amino acid polymorphisms for the candidate genes can be found in Supporting Information Table S1. Gene lengths are not drawn to scale.

New Phytologist
RNAseq, respectively. We showed that resistance conferred by Pm1a can be transferred by transgenic complementation to susceptible varieties, inducing an immune reaction (Fig. 2) that limited the formation of mature haustoria in transgenic leaves. Whole cell autofluorescence in response to fungal penetration suggested a rapid onset of HR (Fig. 3). However, more rigorous staining methods are required to confirm this, and future observations of initial infection dynamics will require sampling at earlier timepoints given the rapid development of haustoria (Bushnell & Bergquist, 1974;Kunoh et al., 1982).
Interestingly, all Blumeria Avr proteins belong to the Blumeria protein superfamily of candidate secreted effector proteins (CSEPs) and exhibit common features such as a generally small size (102-130 amino acids), a signal peptide as well as a N-terminal Y/FxC motif and a C-terminal cysteine at a conserved location. Furthermore, numerous CSEP families, including Avr effectors, have a conserved exon/intron structure, indicating a common phylogenetic origin and were predicted to exhibit a RNase-like structure (Pedersen et al., 2012;Bourras et al., 2016Bourras et al., , 2019Lu et al., 2016;Praz et al., 2017;Spanu, 2017Saur et al., 2019. For Bgh CSEP BEC1054, a member of the effector family E014, which harbours both Avra13 and AvrPm2, the RNase-like structure was recently confirmed experimentally (Pennington et al., 2019). The newly identified AvrPm1a shares many of the earlier-mentioned features of AVR proteins such as conserved sequence motifs (signal peptide, Y/FxC), a similar exon/intron structure and a predicted RNase-fold (Figs S13, S14a,c). However, with 155 amino acids, it is significantly longer than all previously identified Blumeria AVRs. This difference can be attributed to the presence of an additional a-helix at the N-terminus and a slightly extended b-sheet region towards the C-terminus of the protein (Fig. S14a,b). As to how the larger effector size

Research
New Phytologist might influence the mode-of-recognition by its corresponding NLR resistance gene remains to be determined. All amino-acid polymorphisms between the avirulent AvrPm1a_THUN12 and the virulent avrPm1a_96224 variants were located in a short stretch surrounding the Y/FxC motif. In AvrPm3 a2/f2 effector family E008, multiple amino acid residues close to the Y/FxC motif were shown to be under diversifying selection, which was attributed to a potential selection pressure imposed by the Pm3a and Pm3f alleles (McNally et al., 2018). Similarly, naturally occurring sequence polymorphisms in gain of virulence alleles of AvrPm3b2/c2 were located around the Y/FxC motif and the underlying protein region was shown to be crucial for recognition of AvrPm3 b2/c2 by its corresponding NLRs Pm3b and Pm3c . The extent to which similar principles apply for recognition of AvrPm1a by Pm1a will be subject to future experiments. Cloning of additional Pm1 alleles could lead to the identification of additional AvrPm1 alleles, which may help reveal polymorphisms conferring gain or loss of virulence.
The sequencing of Axminster chromosome 7A is a step closer to understanding the true physical size and makeup of the terminal region of this chromosome in lines with Pm1a and Sr15/Lr20. Lines carrying the Pm1-Sr15-Lr20 linkage block characteristically prohibit recombination in the terminal 7AL region. Neu et al. (2002) suggested this region was an alien segment or rearranged fragment. Marker assays also showed no natural recombinants in the region of interest and Australian wheat breeders were unable to combine resistance to root lesion nematode (Rlnn1) with a preferred white flour colour allele due to linkage with the Psy-A1 locus also linked with Pm1-Sr15-Lr20 in certain cultivars (Jayatilake et al., 2013). Earlier evidence that this linkage block was due to a chromosomal rearrangement within the 7AL region was inferred from the presence of sequences that appeared to originate from the B genome (Crawford et al., 2011;Jayatilake, 2014). In this study, we clearly show this region to be abundant in not only sequences from 7B but also 7D (Figs 6, 7). However, precise physical ordering and composition of 7A, 7B and 7D sequences was not attainable due to significant divergence from current reference assemblies and notable disruption of synteny encompassing this region (Liang et al., 2016). A higher contiguity is required to achieve accurate physical resolution of Axminster 7AL, either from SMRT or Dovetail sequencing (Moll et al., 2017).
Comparison of terminal sequences of Axminster 7A with sequence data from diploid accessions pointed to a relatively high proportion of exact matching D. villosum reads (Fig. S12). Thus, introgression from Dasypyrum or a related lineage was considered. However, due to the limited number of reads available (ranging 10-60 K uniquely mapping to the terminal set), this could not be confirmed without deeper sequencing analyses or direct comparison with a genome assembly of D. villosum. The high divergence of this terminal region, which appears to be common to genotypes carrying Pm1a and Lr20-Sr15, traces back to unrelated land varieties introduced to Australia and North America in the late 19 th century with no clear links between them. Ouyang et al. (2014) did not observe characteristic recombination suppression around MlIW172 in T. turgidum subsp. dicoccoides (tetraploid). Intriguingly, Liang et al. (2016) did observe strong recombination suppression in this region of two T. monococcum accessions (diploid). It could be possible that an ancient introgression(s) predisposed this region to its present structure, which in turn promoted a diversification of NLRs that culminated in the linked resistance cluster carrying Pm1a and Sr15-Lr20.

Data availability
All high-throughput RenSeq and ChromSeq sequencing data described in this article have been deposited under BioProject PRJEB39498. Pm1a sequence is available under GenBank accession MW531535. [Correction added after online publication 15 December 2020: an accession number was added in the preceding sentence.] AvrPm1a variant 96224 and variant THUN-12 sequences are available under GenBank accession nos. MT773601 and MT773602, respectively. GBS sequencing data for diploid accessions came from the study by Bernhardt et al. (2019) and can be accessed at http://dx.doi.org/10.5447/IPK/2019/18.

Supporting Information
Additional Supporting Information may be found online in the Supporting Information section at the end of the article.
Dataset S1 Sequences produced by gene-synthesis used in this study.

Fig. S3
Amplification of a large product based on the candidate contig from Triticum aestivum cv Chinese Spring/Axminster*7A (CS/Ax7A) was absent in double mutants of pm1 and sr15/lr20.