Discovery of a novel powdery mildew (Blumeria graminis) resistance locus in rye (Secale cereale L.)

Powdery mildew is one of the most destructive diseases in the world, causing substantial grain yield losses and quality reduction in cereal crops. At present 23 powdery mildew resistance genes have been identified in rye, of which the majority are in wheat-rye translocation lines developed for wheat improvement. Here, we investigated the genetics underlying powdery mildew resistance in the Gülzow-type elite hybrid rye (Secale cereale L.) breeding germplasm. In total, 180 inbred breeding lines were genotyped using the state-of-the-art 600 K SNP array and phenotyped for infection type against three distinct field populations of B. graminis f. sp. secalis from Northern Germany (2013 and 2018) and Denmark (2020). We observed a moderate level of powdery mildew resistance in the non-restorer germplasm population, and by performing a genome-wide association study using 261,406 informative SNP markers, we identified a powdery mildew resistance locus, provisionally denoted PmNOS1, on the distal tip of chromosome arm 7RL. Using recent advances in rye genomic resources, we investigated whether nucleotide-binding leucine-rich repeat genes residing in the identified 17 Mbp block associated with PmNOS1 on recent reference genomes resembled known Pm genes.


K SNP genotyping of panel.
To investigate the genetics underlying powdery mildew (PM) resistance, the assayed hybrid rye breeding germplasm was genotyped on the rye 600 K SNP array. With only scaffold positional data available for the array, SNP marker sequences were anchored to the recent 'Lo7' rye reference genome and stringently filtered to ensure its accuracy. In total, 591,196 markers were successfully mapped to the reference genome, and the developed marker map was made available at https:// doi. org/ 10. 5281/ ZENODO. 50862 35. Quality filtration of markers for low minor allele frequency, missing markers, and missing individual scores across the panel led to the identification of 261,406 informative markers (Supplementary material 1). Characterization of fundamental performance-related metrics revealed a homogeneous inter-and intrachromosomal distribution of markers (Table 2). On average, each chromosome housed 32,676 markers with a mean Table 1. Location and origin of powdery mildew resistance genes in rye (Secale cereale L.) and wheat-(Triticum aestivum L.) rye translocation lines with rye being the donor parent. After 1-3. . ND: Interchromosomal position not determined. www.nature.com/scientificreports/ marker-to-marker distance of 25.54 kb. The largest marker-to-marker distance was 9.95 Mbp on chromosome 2R, with a mean of 4.05 Mbp across the chromosomes. As a quality parameter, the polymorphism information content (PIC) was calculated to estimate the ability of markers to detect polymorphisms within the assayed germplasm (Supplementary Table S2). Across the informative marker panel, a mean PIC of 0.234 was identified, with a mean interchromosomal PIC ranging from 0.204 to 0.249 ( Table 2). Visualization of these array performance metrics along the rye genome using Circos revealed a drop in marker density and PIC across the pericentromeric region on all chromosomes (Fig. 1).

Phenotyping of hybrid rye breeding germplasm.
To provide a comprehensive phenotypic dataset of PM resistance in the assayed Gülzow-type hybrid rye breeding germplasm, the lines were scored for their infection type (IT) against three distinct Blumeria graminis f. sp. secalis (Bgs) populations (Table 3, Supplementary  Table S3). The lines scoring an IT below 1 were considered 'resistant' (Supplementary Fig. S1). Across the assayed germplasm, the N13 (Nienstädt, 2013) Bgs population yielded a mean IT of 2.40 ± 1.28 standard deviations (SD), with 47 resistant lines out of which 3 were restorers. The N18 (Nienstädt, 2018) Bgs population yielded a mean IT of 2.71 ± 1.11 SD with 29 resistant lines, out of which 1 was a restorer. D20 (Dyngby, 2020) Bgs population yielded a mean IT of 2.74 ± 1.10 SD with 20 resistant lines out of which 1 was a restorer. Across the assayed germplasms, 20 out of 88 non-restorer germplasm and 1 out of 92 restorer lines were consistently resistant to all three Bgs populations. Both controls, hybrid cv. KWS Binntto ('susceptible') and KWS Serafino ('resistant') were susceptible to all three Bgs populations. KWS Binntto had a mean IT of 3.56 ± 0.54 SD and KWS Serafino 3.01 ± 1.01 SD across Bgs populations.
To visualize the resistance spectrum of breeding lines, a circular neighbor-joining dendrogram was constructed, and concentric circles were added to integrate the scored IT (Fig. 2).
Statistical analysis of the line infection-type distribution across Bgs population by ANOVA showed that the N13 Bgs population differed significantly from the N18 and D20 Bgs populations (p val < 0.00016) (Fig. 3). Nine non-restorer germplasm lines were found to exhibit a differential resistance response to the N13 Bgs population. These lines were categorized as 'partially resistant' , with a mean infection type of 1.52 ± 0.58 SD; they exhibited a mean infection type of 2.74 ± 0.70 and 2.80 ± 0.77 against the N18 and D20 populations, respectively. Genome-wide association study. For the identification of SNP markers associated with PM resistance, GWAS using MLM and the BLINK method was used for each of the three Bgs population trials and across trials for the entire germplasm and individual parental populations (Supplementary Figs. S2, S3). Isolated single SNPs in GWAS-MLM results were removed from the analysis even if they were significantly associated, as these were interpreted as having spurious associations or incorrect mapping positions. Instead, dense peaks comprising a large number of significant associated markers within a confined region in GWAS-MLM were selected for further analysis. GWAS-MLM on the entire panel led to the identification of a haplotype block on chromosome arm 7RL that was significantly associated (− log 10 = 14.6) with powdery mildew resistance spanning from 882 to 898 Mbp (Fig. 4B, C). The haplotype block harbored 244 markers exhibiting an association above the Bonferroni adjusted significance threshold (− log 10 ≥ 6.7) (Supplementary Table S4. Successive GWAS-BLINK led to the identification of the top-most resistance-associated (− log 10 = 37.9) marker within the haplotype block on chromosome arm 7RL at 892.09 Mbp, which explained 16.8% of the phenotypic variance (Fig. 4A).
In the non-restorer germplasm population, lines were found to carry a highly conserved haplotype across the significantly associated markers on chromosome arm 7RL (Supplementary Table S6). A resistant haplotype was conserved in 42 out of 48 resistant lines and a susceptible haplotype was found in 28 out of 31 susceptible lines, and the remaining 9 lines showed a differential resistance response (Supplementary material S3).
GWAS using BLINK led to the disappearance of the marker position on the 'Unmapped' chromosome that was in linkage disequilibrium with the genomic region on chromosome arm 7RL (Fig. 4A, B). In the non-restorer germplasm, an additional nonsignificant (− log 10 = 3.62) peak was identified in GWAS-MLM spanning from     www.nature.com/scientificreports/ Phylogenetic analysis using the NB-ARC domain of NLRs residing in the PM resistance-associated haplotype block on chromosome arm 7RL led to the finding that the majority of the NLRs were represented in both reference genomes (Fig. 5). The 'Weining' reference genome exhibited one unique NLR not present in 'Lo7' , while the latter exhibited eight unique NLRs, of which five formed a distinct clade. Eight homogeneously represented NLRs from each reference genome clustered in a large clade ('clade 1').
An additional phylogenetic analysis was performed using the entire reference genome NLR repertoire including NLR genes of characterized R genes as references (Fig. 6, Supplementary Fig. S4). The NLR genes residing within the haplotype block were found to span much of the reference NLR repertoire diversity. Clade 1 remained intact in both reference genome NLR repertoire trees positioned in a section harboring four out of five reference Pm genes included, with the closest being Pm60. In both reference genome NLR repertoire trees, a single Characterization of an Rpp13-like NLR gene residing in the resistance-associated block on chromosome arm 7RL. While several of the NLR genes residing within the PM resistance-associated haplotype block on chromosome arm 7RL chromosome showed evidence of an evolutionary relationship with known Pm genes, we selected the Rpp13-like NLR gene for further investigation on the basis of its close proximity. The Rpp13-like NLR gene (Lo7_chr7R_nlr_94, Wei_chr7R_nlr_139) was the closest NLR gene, residing approximately 2 Mbp from the top-most PM resistance-associated marker identified by BLINK (Fig. 4C, Supplementary Table S8). The homologous Rpp13-like NLR genes in the 'Lo7' and 'Weining' reference genomes encoded canonical NLR proteins of 922 to 947 aa, showing 97% sequence similarity, with sequences differing by three single amino acid variants and an indel of 25 aa in the NB-ARC domain (Supplementary material 2). Protein BLAST of the NLR gene led to 88% sequence similarity with disease resistance Rpp13-like protein 4 in T. urartu, which encodes a 924 aa NLR protein sequence.

Discussion
In Denmark, the top-yielding hybrid rye cultivars and population varieties were evaluated in Danish official trials as an advisory service for farmers 53   www.nature.com/scientificreports/ of causing substantial grain yield and quality losses in rye. Furthermore, on the basis of the trial records, we selected a hybrid cv. KWS Serafino as a resistant control in our study, as it showed less than 1% leaf area covered by PM at the site in 2017. However, under a high disease pressure, cv. KWS Serafino was found to be susceptible in the current study, suggesting a potential lower level of resistance against PM in the evaluated top-yielding hybrid cultivars than evident from the official trials.
In this study, we investigated PM resistance in the less-prevalent Gülzow-type elite hybrid rye breeding germplasm against three distinct Blumeria graminis f. sp. secalis (Bgs) populations from Denmark and Northern Germany. We observed a moderate level of powdery mildew resistance in the non-restorer germplasm population, and by performing a genome-wide association study (GWAS) using 261,406 informative SNP markers, we identified a strong PM resistance-associated site on the distal region of chromosome arm 7RL.
Hybrid rye breeding germplasms are highly secluded with little exchange of material. The existing exchange is subsequently influenced by the deployed fertility control system, which determines the compatibility of foreign material for introgression into heterotic parental populations 59 . Additionally, the 600 K SNP array was developed using lines from a German hybrid rye breeding germplasm deploying the predominant Pampa-type cytoplasmic male sterility system (CMS) 57 . In contrast, the germplasm assayed in this study deploys the less-prevalent Gülzowtype CMS system 60 . Due to the distinct nature of the different hybrid rye breeding germplasms, we investigated the performance of the 600 K SNP array on the Gülzow-type germplasm. With no physical position data available for the 600 K SNP markers, we developed a marker map by anchoring the marker sequences to the 'Lo7' reference genome. In brief, we found that the 600 K SNP array yielded a dense panel of informative SNP markers in the Gülzow-type germplasm with markers homogeneously distributed across the rye genome. The level of marker informativeness measured by the polymorphism information content (PIC) was largely comparable with observations made in maize (Zea mays L.). Here, high-density SNP genotyping of 544 diverse CIMMYT www.nature.com/scientificreports/ inbred lines yielded 362 K informative SNP markers with a mean PIC of 0.25 61 . In this study, however, we did observe a drop in marker informativeness and density across the pericentromeric region. Similar observations were made in the study by Rabanus-Wallace, et al. 50 , who reported a considerable reduction in both genetic diversity and gene density within the pericentromeric region of the inbred rye line 'Lo7' . In barley, this region has furthermore been observed to display a 20-fold lower recombination rate 62 . Thus, we concluded that the 600 K SNP array performed satisfactorily on the Gülzow-type germplasm in relation to the fundamental characteristics investigated. The high-density array constitutes a milestone in rye genomic resources transitioning SNP-based genomic studies and replaces the previous rye 5 K SNP array by Haseneyer, et al. 57,63 . GWAS on SNP genotype data has become an effective tool in genome-based plant breeding for the study of oligogenic traits often governed by a few genes with large effects, such as the major monogenic inherited resistance (R) genes 63 . The use of high-density SNP typing has allowed whole-genome scans for the identification of often small haplotype blocks significantly associated with resistance 64 . The creation of the 600 K SNP array and the chromosomal-scale reference genomes 'Lo7' and 'Weining' in rye has significantly changed the genomic toolbox available for mining novel resistance genes 50,57 . Using these recent advances in rye genomic resources, we successfully managed to identify a site on chromosome arm 7RL that was significantly associated with PM resistance. The high level of resolution provided by the dense marker panel revealed that the PM resistanceassociated haplotype block spanned 17 Mbp on the distal tip of the chromosome arm 7RL subtelomeric region.
Until recently, no PM resistance gene had been identified on the rye 7R chromosome. However, during the development of translocation lines for wheat improvement using a local Chinese variety of rye 'Baili' , Ren et al. 42 discovered a PM-resistant 7BS:7RL translocation line. With the recipient wheat parent being susceptible, the PM resistance gene traced back to the rye donor chromosome arm 7RL. Resistance phenotyping of the translocation line demonstrated that it displayed a high level of PM resistance against prevalent Bg. tritici pathotypes in China, making it very promising for the development of novel PM-resistant wheat cultivars. Although a novel finding in rye, several PM resistance genes have been identified in wheat chromosomal segments syntenic to rye chromosome arm 7RL. During Triticeae speciation, a series of recurrent translocation events gave rise to major patterns of chromosomal rearrangements 65 . In rye, the distal region of chromosome arm 7RL, therefore, shows high homology with the wheat 2A/B/D chromosomes 58,66 . Currently, more than five Pm genes have been identified in wheat on the 2A/B/D chromosomes 67,68 .
With the Pm gene discovered by Ren, et al. 42 originating from a forage-type population of rye in a gene pool distinct from the germplasm investigated in this study, it is reasonable to presume that the two Pm genes could be either distinct or allelic 69 . We provisionally denote the novel Pm locus residing in the subtelomeric region of chromosome arm 7RL as PmNOS1.
Comparative analysis of the three Bgs populations used in the study revealed that the N13 population significantly differed from the two more recent populations, showing a less virulent composition of pathotypes. In addition, nine non-restorer germplasm breeding lines displaying a differential resistance profile were identified, showing 'partial resistance' only to the N13 population. With none of these lines carrying the resistance haplotype associated with the resistance locus on chromosome arm 7RL, our findings suggest that these do not carry the PmNOS1 locus. While this observation can be explained by the occurrence of recombination events between the PM resistance-associated marker and the causative gene, this result seems less likely due to the consistent divergence in the haplotype 70 . Instead, it seems more likely that the differentially resistant non-restorer germplasm lines carry a distinct Pm gene that lost its effect during the period 2013-2018 in northern Germany. As a result of host genetic uniformity and the high evolutionary capacity of Bg to acquire virulence and migrate rapidly over long distances, the effectiveness of Pm genes is often rapidly lost. An example of this is the Mla13 Pm gene introduced in a former Czechoslovakian barley cultivar, 'Koral' , in 1980 17 . After years of providing effective resistance against PM disease, virulence was observed in England in 1988 in a pathotype believed to originate from Czechoslovakia, having migrated by wind across the European continent and North Sea 71 . Ongoing monitoring of the virulence structure of Bg was conducted to survey the effectiveness of deployed Pm genes in elite cultivars 72,73 . While the Pm gene present in the differentially resistant non-restorer germplasm lines has seemingly been overcome, our findings suggest that PmNOS1 remains effective.
Enabled by recent advances in rye genomic resources, we investigated whether any of the NLR genes residing in the region harboring the PmNOS1 locus resembled known Pm genes. In several crop species, including rye, NLR genes have been observed to accumulate at recombination hotspots in subtelomeric chromosomal regions 50,74,75 . Additionally, we identified several large clusters of NLR genes residing in the PM resistanceassociated block harboring PmNOS1 in both of the reference genomes. In addition to the PM resistance gene identified in our study, Fusarium head blight and leaf rust resistance genes have been mapped to the subtelomeric region of chromosome arm 7RL 76,77 . Stem and stripe rust resistance genes reported in the 7BS:7RL translocation line developed by Ren et al. 42 are furthermore likely to reside in the subtelomeric region. To investigate the likely diversity of R genes residing in the PM resistance-associated block, we conducted a phylogenetic analysis using the NB-ARC domain sequence of NLR genes residing in the block 78,79 . In contrast to the rapidly evolving LRR domain often exhibiting intraspecific polymorphism, the NB-ARC domain is largely conserved and suited for the study of evolutionary relationships among NLR genes 80,81 . As expected, the NLR genes residing in the block represented a large proportion of the NLR repertoire diversity in rye, accentuating the evolutionary plasticity of the NLRs residing in the subtelomeric region 75 . Intriguingly, guided by a panel of isolated NLR genes as a reference, we observed a predisposition of NLRs within the block to be in close proximity to known Pm genes in wheat and barley. This evolutionary relationship could hint at a common attribute among the NLRs 79 .
Phylogenetic analysis led to the identification of an NLR gene with an evolutionary relationship and protein sequence similarity to Rpp13-like protein 4 in T. urartu. Based on its close proximity to the marker displaying the strongest association with the PmNOS1 locus, the Rpp13-like NLR gene was selected for further characterization. In A. thaliana, Rpp13 confers resistance against Peronospora parasitica, the causative agent of downy mildew www.nature.com/scientificreports/ disease 82 . Recent studies have, however, identified Rpp13-like NLR genes associated with powdery mildew resistance in cereals. In wheat, Liu et al. 83 showed that silencing of the Rpp13 homologous gene TaRPP13-3 in resistant wheat cv. 'Brock' induced susceptibility to powdery mildew. In barley, Cheng et al. 84 found that the expression of an Rpp13-like NLR gene was highly upregulated after inoculation with powdery mildew. In rye, while the rate of decay has only been determined in a few genes related to frost response, which showed a rapid linkage decay 85 , these genes have been demonstrated to decay after 3.76 kb, on average, across the genome in a similar hybrid breeding germplasm in maize 61 . Due to the heterogenic nature of outcrossing species, their rate of decay is often rapid 86 . However, in a recent population study on assayed germplasm, non-restorer germplasm was found to exhibit relatively low genetic diversity, low effective population size and high linkage disequilibrium 59 . Consistent with their observations, we found a large conserved haplotype on chromosome arm 7RL harboring the PmNOS1 locus 59 . The linkage decay in the non-restorer germplasm is likely considerably reduced by the low genetic diversity and effective population size, resulting in a similarly low frequency of effective recombination events. The large amount of linkage, while beneficial for trait discovery in GWAS at lower marker density, impedes the identification of a narrow and precise genomic region that may harbor the gene of interest, even at high marker resolution. In the case of the PmNOS1 locus, the haplotype block on chromosome arm 7RL was found to span 17 Mbp, harboring between 17 and 25 potential candidate NLR genes in the 'Lo7' and 'Weining' reference genomes. For more accurate mapping of the PmNOS1 locus, the development of multiparent mapping populations could be conducted, allowing several generations of potential effective recombination events in the region 87,88 . Identification of the causative gene could be performed by resistance gene enrichment sequencing (RenSeq) analysis followed by transformation of a susceptible non-restorer germplasm line to validate the gene 89 . This would in turn equally show whether the gene is present in the reference genomes and whether PmNOS1 encodes an Rpp13-like NLR protein.
In conclusion, our study demonstrates the immediate value of recent advances in rye genomic resources for the mining of novel resistance genes. These resources now permit accurate identification of delimited resistanceassociated haplotype blocks and scanning for trait-associated genes residing within. With pathogens such as Bg displaying a large evolutionary plasticity, shortening the process from identification of resistance-associated sites to isolation of the underlying R gene is important for the development of novel resistant cultivars. The relevance of studies in rye is accentuated by the possibility of introgressing novel R genes into the staple cereal wheat by chromosomal translocation lines. To aid further studies in the field, we have provided both a rye600K SNP array marker map anchored using the 'Lo7' reference genome and NLR repertoire information of the 'Lo7' and 'Weining' reference genomes in open-access data repositories.

Materials and methods
Plant material and DNA extraction. A panel of 180 inbred rye (Secale cereale L.) lines, 92 restorer and 88 non-restorer germplasm, belonging to the elite Gülzow-type hybrid rye breeding germplasm at Nordic Seed A/S (Dyngby, Denmark) were investigated. Population structure and information on the genetic characteristics of the germplasm were presented in a recent study by Vendelbo et al. 59 . The parental populations represent genetically secluded gene pools with restorer (paternal) lines carrying a dominant allele for the restoration of male fertility and non-restorer germplasm lines carrying a recessive allele and fertile cytoplasm, which is used to maintain cytoplasmic male sterile (maternal) lines. DNA extraction was performed using an adapted SDSbased method according to the USDA 90 after Pallotta et al. 91 on an equivalent of 75 mg of plant material collected from the coleoptiles and primary leaves of two seven-day-old seedlings per line. The DNA concentration and 260/280 nm ratio of the samples were measured using an Epoch™ microplate spectrophotometer (Biotek®), and evidence of fragmentation was obtained by size visualization on a 1.2% agarose gel. Molecular marker resource and SNP genotyping. Samples of each line containing 200 ng of high molecular weight gDNA with a ≥ 1.8 260/280 nm ratio were sent for single nucleotide polymorphism (SNP) genotyping at Eurofins Genomics Europe Genotyping (Aarhus, Denmark). Genotyping was performed using a 600 K SNP array with 600,843 SNP markers on an Affymetrix GeneTitan™ Scanner platform 57 . Collection and multiplication of Blumeria graminis f. sp. secalis populations. As there was no unified information on germplasm resistance to powdery mildew, three field populations of Bgs were sampled in Northern Germany and Denmark in the period from 2013 to 2020 to screen the lines against a broad range of Bgs pathotypes prevalent in Northern Europe. Two Bgs populations were collected at the Nordic Seed rye multiplication site in Germany in 2013 (N13) and 2018 (N18) (52.29254°N, E9.14896°E). Ten leaves exhibiting PM disease were carefully collected and rinsed in 0.5 L of water to release Bgs spores. Six pots containing 15-20 susceptible 12-day-old seedlings of the restorer line R277 were then inoculated by spraying a fine mist of the spore solution using an atomizer bottle. Pots were transferred to a separate climate chamber for each population and incubated at 18 °C with 12 h of light using 400 W high-pressure Phillips SON-T Agro lamps. Populations were continuously multiplied in an overlapping two-week cycle. Each week, the tray of 2-week-old seedlings was substituted with a tray of fresh 12-day-old seedlings. Inoculation was achieved by passive dissemination of spores from the 'older' tray to the tray with new plants through a steel-grid shelf.
An additional Bgs population was collected in autumn 2020 (D20) at Nordic Seed (Dyngby, Denmark) (55.94944°N, E10.25414°E) using a mixture of hybrid cvs. KWS Binntto, KWS Bono and KWS Florano. Pots with 15 to 20 seedlings of the susceptible mixture were placed outdoors 12 days after sowing (DAS) in August-October 2020. Plants were controlled regularly, and all leaves showing PM disease within the period were collected. Prior to inoculation, leaves were placed in Petri dishes with moist filter paper and set to sporulate at room temperature in light for four to eight hours. Then, a pot containing 15-20 seedlings of the susceptible mixture was inoculated www.nature.com/scientificreports/ at 12 DAS by horizontally stroking the collected leaves across the seedlings. Inoculated pots were sprayed with a fine mist of water, placed in a container with transparent lids to ensure 100% RH and incubated in the dark at 10-15 °C for 24 h. After incubation in the dark, the pots were transferred to an isolated greenhouse cabin and incubated under 16 h of daylight at 18-24 °C and 8 h of dark at 14-16 °C. Multiplication of trial inoculum for the three Bgs populations was performed using the same procedure as described above. For each population, two trays containing 35 pots of the susceptible mixture were inoculated by brushing 3-4 highly infected pots inoculated 20 days earlier across the tray.
Infection and scoring. All lines were phenotyped for the infection-type response to the Bgs populations.
In each greenhouse trial, eight seeds per line were sown in a 28-hole tray using a completely randomized design with two repetitions for each of the two trial replicates. For each tray, a positive ('susceptible') control consisting of hybrid cv. KWS Binntto and negative ('resistant') control cv. KWS Serafino was included. At 14 DAS, trays were inoculated as described above by brushing 3-4 highly infected pots inoculated 20 days earlier across the tray. After 14 days of incubation, the lines were phenotyped by scoring the infection response on the first and second leaves for each of the eight seedlings per repetition in accordance with a 9-step 0-4 scale by Torp et al. 88 ( Table 3).
Data analysis. Bioinformatic analysis of SNP marker data was performed with the R studio (v. 1.3.959) interface in R statistical software (v. 4.0.1) by applying various predesigned packages 92,93 . Mapping of 600 K SNP array markers to the 'Lo7' reference genome. Positional data of the 600 K SNP markers were obtained by mapping each of the 600,843 SNP marker sequences to the rye reference genome 'Lo7' using the NCBI blastn (v. 2.9.0 +) function 50,94 . The mapping positions of SNPs were hereafter stringently filtered for I) complete SNP sequence alignment and II) a maximum of 1 mismatch to ensure accurate positioning.
Molecular markers and characterization of 600 K SNP array performance. Prior to analysis, markers were filtered for a marker allele frequency ≥ 0.005, missing individual score ≤ 0.2 and missing marker score ≤ 0.1. Fundamental characteristics of SNP marker informativeness, including polymorphism information content (PIC), were calculated using the SnpReady (v. 0.9.6) R package 95 . The interchromosomal distribution of the informative marker PIC, marker-to-marker distance and marker density in 10 mb bins on the 'Lo7' rye genome were visualized using Circos (v. 0.69.8) in the Galaxy online interface 96,97 . A Circos plot was constructed using the pipeline developed by Hiltemann, et al. 98 .
Analysis of phenotypic data. The distribution of infection types against the three respective PM populations was visualized by density plots using ggplot2 (v. 3.3.3) R package 99 . To determine whether the population infection-type distribution differed significantly, ANOVA was conducted using R.
To correct the resistance phenotype for the effects of replication and population, we fitted the data to a linear mixed model using the lme4 (v. 1.1.26) package in R: where µ is the general mean, P is the population, R represents the replications, l is the line id, and ɛ is the residuals. P and R were set as fixed effects, and l was set as a random effect. The random effect and residuals were assumed to be independent normally distributed variables described as follows: l ~ N (0, I ơ 2 l ), and ɛ ~ N (0, I ơ 2 ɛ). The BLUP solutions for the line effect were used in GWAS. Data were also fitted to a linear mixed model for each of the populations to correct for the effect of replication. In this model 'P' was removed.
Genome-wide association study. Discovery of PM resistance-associated SNP markers was performed by a genome-wide association study (GWAS) using the genomic association and prediction integration tool (GAPIT) (v.3) package in R 100 . The Manhattan plot was colorized using the RColorBrewer (v.1.1-2) R package color palette 99 . GWAS using a mixed linear model (MLM) was performed to identify discrete haplotype blocks associated with powdery mildew resistance. Additionally, the Bayesian-information and linkage-disequilibrium iteratively nested keyway (BLINK) method was performed to identify the top-most powdery mildew resistanceassociated marker within each haplotype block 101 . BLINK uses a multiple loci test for MLM by combining a fixed effects model, Bayesian information content and linkage disequilibrium information to collectively improve the statistical power while simultaneously reducing the computational run time. Markers that are in linkage disequilibrium with the top-most significant marker at a site are excluded in BLINK. A standard Bonferroni-corrected threshold of α = 0.05 was used as the significance threshold.
Characterization of nucleotide-binding leucine-rich repeat proteins. The gene structure, coding sequence and NLR protein sequence of candidate genes were extracted from RNA-seq and de novo protein data provided in 'Weining' and 'Lo7' reference genome data repositories 50,58 . To investigate whether NLR genes residing in y = µ + P + R + l + ε www.nature.com/scientificreports/ PM resistance-associated sites resembled known genes, the NCBI blastx function was used for protein-protein searches in the online database 94 . For functional analysis and the prediction of protein domains, InterPro Scan was used 103 . Identification of sequence divergence between reference genome homologs was performed by multiple sequence alignment using the multiple sequence comparison by log-expectation (MUSCLE) method for coding sequences and the 'Clustal Omega' method for NLR protein sequences in Geneious Prime (v. 2020.2.3).
Phylogenetic analysis. Neighbor-joining clustering analysis of breeding lines was performed with Euclidean genetic distance measurement using the ape (v. 5.3) R package 104 . The tree was constructed after 10,000 bootstrapping iterations with weak nodes (≤ 80% recurrence) collapsed into multifurcations. A circular neighborjoining tree was generated using the iTOL (v. 5) online tool (http:// itol. embl. de/), enabling a colorful visualization of each line infection type spectrum against the three Bgs populations 105 .